Canonical covariate columns

This file is the authoritative register of covariate column names used in nlmixr2lib models. Every covariate referenced inside a model’s model block must have an entry here. The register is seeded from a full audit of inst/modeldb/ and extended whenever a new paper introduces a covariate that isn’t yet registered.

How to use this register

1. Before adding a covariate to a new model, search this file (by canonical name and by source alias) for the concept you need.
2. If the canonical name exists, use it exactly. Document any source-paper rename in the model’s covariateData[[name]]$source_name field.
3. If the source paper uses an alias listed under an existing canonical name, use the canonical name and note in covariateData[[name]]$notes whether the values must be transformed (e.g., SEXM -> SEXF inverts values; the effect coefficient sign / reference category must be inverted as well).
4. If the covariate is not in this register at all, propose a new entry with a canonical name, description, units, type, scope, and source aliases. Verify with the maintainers before committing. The addition is part of the model’s PR.
5. Do not modify existing model files when you discover an alias; simply document the mapping in the register. Retrofitting existing models is a separate effort.

Scope: general vs specific

Each entry has a Scope: field declaring whether it is general (any model may use it without warning) or specific (only the models listed under Example models may use it; other usage is flagged by checkModelConventions). This prevents accidentally reusing a covariate name whose meaning is tied to a particular paper.

• Scope: general – the covariate has a stable, paper-independent meaning and any future model may use it. Examples: WT, AGE, SEXF, CREAT, ADA_POS, CRP, CRCL.
• Scope: specific – the covariate’s semantics depend on a particular study’s design (a specific study indicator, a drug-product variant, a composite race grouping, a tumor-type decomposition, etc.). If a new paper needs the same concept with different semantics, register a new canonical name; if the concept matches, extend the Example models list (and consider promoting to general).

When adding or updating an entry, choose the most conservative scope: if in doubt, start with specific and promote when a second model legitimately ratifies the name.

Case convention

Covariate column names are ALL CAPS by default. checkNamingRegisters() enforces this mechanically: any ### canonical in this file that is not all caps and is not listed below is reported as a case-convention issue, and any name listed below that no longer has an entry is reported as stale-case-exemption. The list is therefore the complete and current set of exceptions – it cannot drift out of date the way the hand-maintained sentence it replaced did (that sentence named two exceptions when there were ten).

An exception is only for a name whose case is load-bearing: upper-casing it would change or destroy the meaning. “The source paper spelled it that way” is not a reason – source spellings belong in Source aliases: and in the model’s source_name field, not in the canonical. The six legacy lower-case canonicals that were carried for that reason (dilution, nonECZTRA, ooc1-ooc4) were renamed on 2026-09-09, along with CONMED_RTV_AUC_12h, whose h was an ordinary unit initialism rather than a case-significant one.

To add an exception, add the name as a nested bullet here and say why in the entry’s own Notes: field.

• Case-significant unit suffix – the trailing unit is only unambiguous in mixed case (picomolar pM is not PM, which is a different quantity):
  • L_ANTAGONIST_pM
  • L_OPIOID_pM
• Case-significant proper noun – a construct, reagent, or gene-product name whose published spelling carries the meaning:
  • STUDY_d2eGFP

Entry schema

- name: <CANONICAL_NAME>
  description: <one-sentence definition>
  units: <unit string, or "(binary)" / "(categorical)">
  type: continuous | binary | categorical | count
  scope: general | specific
  reference_category: <the 0 group for binary/categorical, or NULL>
  source_aliases:
    - <ALIAS_NAME> (<transformation if any>) -- used in <model.R>
example_models:
    - <model.R>
notes: <free text>

---

Demographics

WT (canonical for body weight (baseline or time-varying))

• Description: Body weight (baseline or time-varying).
• Units: kg
• Type: continuous
• Scope: general
• Reference category: n/a – used with allometric scaling (WT / ref_wt)^exponent. Reference weights observed: 70 kg (adults), 75 kg, 84.8 kg, 56 kg (Kloprogge 2014 quinine cohort typical), 5 kg (infants), 25 kg (Wang 2012 Chinese pediatric epilepsy cohort median), 37.1 kg (Damnjanovic 2023 Serbian paediatric epilepsy cohort mean).
• Source aliases:
  • WEIG – weight column abbreviation used by Wang 2012 (Acta Pharmacol Sin 33:845-851); same biological quantity in kg, no value transformation. Used in Wang_2012_levetiracetam.R.
• Example models: Clegg_2024_nirsevimab.R, Hu_2026_clesrovimab.R, Zhu_2017_lebrikizumab.R, Kovalenko_2020_dupilumab.R, CarlssonPetri_2021_liraglutide.R, Cirincione_2017_exenatide.R, Grimm_2023_gantenerumab.R, Grimm_2023_trontinemab.R, Kyhl_2016_nalmefene.R, Soehoel_2022_tralokinumab.R, Xie_2019_agomelatine.R, PK_2cmt_mAb_Davda_2014.R, phenylalanine_charbonneau_2021.R, Chua_2025_mirikizumab.R, Jackson_2022_ixekizumab.R, Kotani_2022_astegolimab.R, Ma_2020_sarilumab_anc.R, Ma_2020_sarilumab_das28crp.R, Moein_2022_etrolizumab.R, Tiraboschi_2025_amlitelimab.R, Robbie_2012_palivizumab.R, Bajaj_2017_nivolumab.R, Quartino_2019_trastuzumab.R, Wang_2020_ontamalimab.R, Fau_2020_isatuximab.R, Okada_2025_rocatinlimab.R, Kunisawa_2014_olprinone.R, Xu_2020_daratumumab.R (reference 78.6 kg; power exponents 0.451 on linear CL and 0.375 on V1), Dirks_2008_cetuximab.R (reference 60 kg; additive linear-deviation effect 0.0083 per kg on V1 of a two-compartment Michaelis-Menten model for cetuximab in SCCHN; source column WGT), Struemper_2017_belimumab.R (reference 67 kg; fixed allometric exponents 0.75 on CL and Q, 1.00 on Vc, 0.8 on Vp; baseline-only, source column BWT), MedellinGaribay_2015_gentamicin.R (linear (not allometric) weight scaling on both CL and Vc: CL = theta1 * BW + theta5 * (CRCL/75), Vc = theta2 * BW; no reference weight used because the scaling is linear, not divisive; source column BW; cohort mean 6.4 +/- 2.2 kg, infants 1-24 months), Wahlby_2004_pefloxacin.R (time-varying; reference 65 kg; linear-exp form on V with coefficient 0.014), Decker_2024_baricitinib.R (baseline-only, source column WTE = weight at entry; reference 74 kg carried from the upstream adult rheumatoid-arthritis model rather than taken from the pediatric cohort; exponents fixed at 0.75 on all clearance arms and 1 on both volumes), Mouksassi_2015_thrombomodulinAlfa.R (baseline; reference 68 kg, the Table 1 combined median of the pooled 392-subject cohort; power exponents ESTIMATED rather than fixed to allometric values – 0.56 (14.15% SE) on CL and 0.50 (21.14% SE) on Vc – and body weight was screened both continuously and as a categorical < / >= 100 kg split, the continuous power form being retained), FernandezTeruel_2024_capivasertib.R (baseline, source column BBW, reference 67 kg; the SAME column enters in TWO different functional forms in one model, which is why the reference weight cannot be read off a single equation: a LINEAR-deviation effect on initial apparent clearance, CL0/F * (1 + (WT - 67) * 0.00585), and a POWER effect on the logit-scale first-order absorbed fraction, logitffo * (WT / 67)^(-1.11); Fernandez-Teruel 2024 equations (a) and (d)), Ekobena_2025_bictegravir.R (baseline, source column BW; reference 70 kg is a ROUNDED value named in the text, NOT the cohort median, which Table 1 gives as 74 kg – a reminder that the reference weight must be read off the covariate equation’s own prose rather than inferred from the demographics table; the exponent 0.35 is ESTIMATED, not fixed to an allometric 0.75, and the paper writes the term as exp(beta * log(BW / 70)) rather than as the algebraically identical power form), Isbister_2015_snake_antivenom.R (baseline, source column wt; power effect on the CENTRAL volume only, with an ESTIMATED and very small exponent of 0.132 whose 84% RSE means it is barely distinguishable from zero. The centring value is the one thing the paper does not print: the Methods say the covariate was “centred to the average weight” but no average appears anywhere, so the model uses the Table 1 MEDIAN of 57 kg as the stand-in. That substitution is nearly free here and the reason is worth recording, because it is the general test for whether an unprinted centring constant matters: the exponent is small and the cohort weight range narrow (40-70 kg), so any plausible mean in that range rescales V by (57/wt_av)^0.132, at most about 3%. Contrast the Ekobena entry above, where the reference weight is named in the text and differs from the cohort median – when the exponent is large, that distinction is load-bearing; here it is not), Francis_2019_rifapentine.R (reference 56 kg, the Table 1 pooled-cohort median; theory-fixed exponent 1 on the apparent central volume of a one-compartment rifapentine model. Worth noting for the head-to-head it records: total body weight beat fat-free mass as the size descriptor for VOLUME by 20 OFV points in the same analysis where fat-free mass beat total body weight for CLEARANCE by 23, so the model scales the two parameters on two different columns. Observed weight range 38-94 kg), Bi_2018_testosteroneCypionate.R (time-varying, entering ONLY through the within-subject deviation (WT - WT_BASE) under the Wahlby 2004 decomposition, in an exponential clearance term centred at +2.90 kg; the size effect proper is carried by the paired WT_BASE column, so a consumer supplying a constant weight column contributes nothing through this term by design).
• Notes: Universal. Verify time-varying vs. baseline-only against the source paper.

AGE (canonical for subject age)

• Description: Subject age in years.
• Units: years
• Type: continuous
• Scope: general
• Reference category: n/a
• Source aliases: none.
• Example models: Archary_2019_lamivudine.R, Budha_2023_tislelizumab.R, Chakraborty_2012_canakinumab.R, Chen_2020_luspatercept.R, Conrado_2014_alzheimer.R, Diepstraten_2013_propofol.R, Gandhi_2021_abatacept.R, Goel_2016_Sonidegib.R, Hennig_2013_tobra.R, Hong_2025_datopotamab.R, Ide_2020_elotuzumab.R, Koopman_2023_factorix.R, Kuchimanchi_2024_dostarlimab.R, Kunarajah_2017_doxorubicin.R, Kyhl_2016_nalmefene.R, Lahu_2010_roflumilast.R, Li_2006_meropenem.R, Li_2017_cediranib.R, Li_2019_abatacept.R, Lin_2024_casirivimab.R, Martinez_2019_alirocumab.R, Melhem_2022_dostarlimab.R, Mulyukov_2018_ranibizumab.R, NA_NA_tte_gompertz.R, NA_NA_tte_lognormal.R, Retlich_2015_linagliptin.R, Rosario_2015_vedolizumab.R, Svensson_2016_bedaquiline.R, Thakre_2022_risankizumab.R, Wu_2024_inotuzumab.R, Yassen_2025_asundexian.R, Yu_2022_ofatumumab.R, Zhong_2026_abatacept.R, Zhou_2021_belimumab.R, Zhu_2017_lebrikizumab.R, Ekobena_2025_bictegravir.R (median-centred FRACTIONAL-deviation form on apparent clearance), GarciaHervalejo_2026_teicoplanin.R (median-centred RAW-deviation linear multiplier (1 + 0.012 * (AGE - 62)) on the clearance of a one-compartment teicoplanin model in adults with haematological malignancies; centring value 62 years, observed range 18-85. Contrast with the Ekobena form noted above: this one is linear in the raw deviation in years and is NOT divided by the median, so the coefficient carries units of per year – which is exactly the distinction that entry warns about, seen from the other side. Two further points. The sign is POSITIVE, i.e. clearance rises with age, and the paper argues in its Discussion that this is not renal ageing but a surrogate for the hypoalbuminaemia and altered body composition of older cachectic patients: teicoplanin is 90-95% protein bound, albumin itself failed screening in this cohort, and the paper attributes that failure partly to collinearity with age. And because the multiplier is linear rather than exponential it turns negative below 62 - 1/0.012 = 8.7 years, so the term is adults-only by construction).
• Notes: Zhu 2017 normalizes as AGE/40. Ekobena 2025 uses a form worth recognising because it is easy to mis-transcribe: the paper calls it a “linear model”, but it is linear in the fractional deviation from the cohort median and sits inside an exponential – exp(beta_Age * (AGE - AgeM) / AgeM) with AgeM = 51 years. The division by AgeM is load-bearing. Dropping it and applying the coefficient to a raw deviation in years turns the paper’s stated “11% decreased CL” at age 80 into a 99.7% decrease. When a paper describes a continuous covariate effect as “linear” and centres it on a median, check whether the printed equation also normalises by that median before encoding it.

AGE_GE70 (canonical for age >= 70 years indicator)

• Description: 1 = subject is aged 70 years or older at baseline, 0 = subject is younger than 70 years. Time-fixed at study entry per subject. Used when the source paper reports an age effect as a threshold indicator (>= 70 y vs < 70 y) rather than as a continuous covariate on AGE.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (< 70 years).
• Source aliases:
  • AGE >= 70 – derived from the continuous AGE column via AGE_GE70 = as.integer(AGE >= 70); used in Overgaard_2016_liraglutide.R (Overgaard 2016 Fig. 2 / Table S1 covariate label “Age >= 70 years”).
• Example models: Overgaard_2016_liraglutide.R (multiplicative log-scale effect on CL/F, coefficient -0.10 per Table S1).
• Notes: Sibling to the ECOG_GE1 / ECOG_GE2 binary threshold family; captures a paper-defined age band as a decomposed binary indicator rather than a continuous AGE term. When the source paper describes the age band as “>= 70 y” (as in Overgaard 2016), use AGE_GE70; for other thresholds (e.g., >= 65 y, >= 75 y) register a sibling canonical (AGE_GE65, AGE_GE75). Use the AGE_GT<n> spelling instead when the paper’s threshold is strictly greater than the cut point (see AGE_GT65). The underlying continuous AGE may still be recorded alongside for downstream re-derivation.

AGE_GT65 (canonical for age > 65 years indicator)

• Description: 1 = subject is older than 65 years at baseline, 0 = subject is 65 years old or younger. Time-fixed at study entry per subject. Used when the source paper reports an age effect as a threshold indicator (> 65 y vs <= 65 y) rather than as a continuous covariate on AGE. This is the conventional geriatric cut point in anti-infective and oncology popPK.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (<= 65 years).
• Source aliases:
  • AGE_GROUP (coded 1 = age <= 65 y, 2 = age > 65 y) – derived from the continuous AGE column via AGE_GT65 = as.integer(AGE > 65), equivalently AGE_GT65 = AGE_GROUP - 1; used in Zakria_2026_voriconazole.R (Zakria 2026 Table 2 rows “CL-AGE_GROUP 1” / “CL-AGE_GROUP 2”).
• Example models: Zakria_2026_voriconazole.R (additive-fractional effect on CL, coefficient -0.519 per Table 2 and Eq 2), Lalovic_2020_lemborexant.R (multiplicative factor 0.739 on CL/F; the Methods and Table 3 define elderly as > 65 years although Table 1 prints ‘>= 65 years’).
• Notes: Sibling to AGE_GE70 and to the ECOG_GE1 / ECOG_GE2 binary threshold family. The GT (strictly greater) rather than GE (greater or equal) spelling is load-bearing: a subject aged exactly 65 years belongs to the reference group under AGE_GT65 but to the indicator group under a hypothetical AGE_GE65. Match the spelling to the inequality the source paper prints – Zakria 2026 dichotomises as “<= 65 years” versus “> 65 years”, so AGE_GT65 is correct and AGE_GE65 would misclassify the 65-year-olds. Register AGE_GE65 separately if a paper genuinely uses the “>= 65 y” band. The underlying continuous AGE may still be recorded alongside for downstream re-derivation.

AGE_GT60 (canonical for age > 60 years indicator)

• Description: 1 = subject is older than 60 years at baseline, 0 = subject is 60 years old or younger. Time-fixed at study entry per subject. Used when the source paper reports an age effect as a threshold indicator (> 60 y vs <= 60 y) rather than as a continuous covariate on AGE.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (<= 60 years).
• Source aliases:
  • Age>60 – derived from the continuous AGE column via AGE_GT60 = as.integer(AGE > 60); used in Rini_2013_axitinib.R (Rini 2013 Table 2 row “Age >60-yr effect on CL” and the typical-value clearance equation printed in Results “PK model”, in which Age>60 “is 1 if applicable and 0 otherwise”).
• Example models: Rini_2013_axitinib.R (linear-proportional effect on CL: (1 - 0.213 * AGE_GT60), i.e. 21.3% lower clearance above the cut point, per Rini 2013 Table 2 and the Results “PK model” equation).
• Notes: Sibling to AGE_GT65 and AGE_GE70, and to the ECOG_GE1 / ECOG_GE2 binary threshold family; registered under the instruction in the AGE_GE70 entry to register a sibling canonical for each new threshold rather than reusing a neighbouring cut point. The GT (strictly greater) rather than GE spelling matches the inequality Rini 2013 prints (“age > 60 years”, “>60-yr”), so a subject aged exactly 60 belongs to the reference group; register AGE_GE60 separately if a paper genuinely uses the “>= 60 y” band. The threshold is paper-chosen rather than physiological: Rini 2013 first modelled age continuously on CL, then dichotomised it to avoid overestimating clearance at younger ages given a non-uniform age distribution across its 17 pooled studies, and selected 60 years as the most significant of several tested cut points (Supplementary Figures S2-A and S2-B). Do not treat the effect size as transportable to a different cut point. The underlying continuous AGE may still be recorded alongside for downstream re-derivation.

AGE_GE60 (canonical for age >= 60 years indicator)

• Description: 1 = subject is aged 60 years or older at baseline, 0 = subject is younger than 60 years. Time-fixed at study entry per subject. Used when the source paper reports an age effect as a threshold indicator (>= 60 y vs < 60 y) rather than as a continuous covariate on AGE.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (< 60 years).
• Source aliases:
  • OLD – derived from the continuous AGE column via AGE_GE60 = as.integer(AGE >= 60); used in Jayanti_2026_pyrazinamide.R (Jayanti 2026 Methods “geriatric (>= 60 years) with DM” and the deposited NONMEM dataset S2 File, whose OLD column spans ages 17-59 where it is 0 and 60-80 where it is 1).
• Example models: Jayanti_2026_pyrazinamide.R (enters only as the product AGE_GE60 * DIS_DIAB, giving Korean patients who are both diabetic and >= 60 years old a 26% higher CL/F; never a main effect).
• Notes: Sibling to AGE_GT60, AGE_GT65 and AGE_GE70, registered under the standing instruction in the AGE_GT60 and AGE_GE70 entries to register a separate canonical for each threshold and each inequality rather than reusing a neighbouring one. The GE (greater or equal) rather than GT spelling is load-bearing and distinguishes this entry from AGE_GT60: a subject aged exactly 60 belongs to the indicator group here but to the reference group under AGE_GT60. Match the spelling to the inequality the source actually uses, and be aware that a paper may print both forms inconsistently – Jayanti 2026 prints “Age > 60 years old” in its Table 1 row header and its S3 / S4 Table footnotes but “>= 60 years” in Methods, Results and the Fig 3 caption, and only the deposited dataset resolves it. When a paper is ambiguous and no dataset is available, prefer the form used in the Methods / Results text that describes the covariate model over the form used in a demographics-table row header. As with the sibling entries, the threshold is paper-chosen rather than physiological, so an effect size estimated at this cut point is not transportable to another one. The underlying continuous AGE may still be recorded alongside for downstream re-derivation.

LBM (canonical for lean body mass)

• Description: Lean body mass.
• Units: kg
• Type: continuous
• Scope: general
• Reference category: n/a
• Source aliases:
  • LBW (lean body weight) – synonym; same biological quantity (total body weight minus body fat). Hemophilia popPK literature typically uses LBW (Hume or James formula) where mAb / general literature uses LBM. Used in Garmann_2017_BAY81_8973.R (reference 51.1 kg) and Li_2023_TQB3203.R (reference 49.53 kg).
• Example models: Kyhl_2016_nalmefene.R (reference 56.28 kg, exponent 0.626 on CL), Garmann_2017_BAY81_8973.R (alias LBW; reference 51.1 kg, exponents 0.610 on CL and 0.950 on Vc), Schoemaker_2017_brivaracetam.R (alias LBW; paediatric cohort, reference 50 kg adult typical value, fixed theoretical allometric exponents 0.750 on CL/F and 1.00 on V/F), Lee_2025_levofloxacin.R (James formula from sex, total body weight, and height; reference 47.91 kg cohort median, range 37.7-60.3; power exponent 1.75 on the PERIPHERAL volume V2 rather than on CL or Vc, which is unusual for this column – the source screened total body weight, FFM, and NFM as alternatives and retained none), Li_2023_TQB3203.R (alias LBW; reference 49.53 kg cohort median, range 32.09-59.40; screened because of the drug’s high lipophilicity and retained on THREE parameters with exponents of opposing sign – +1.18 on the central volume V1, -1.41 on the shallow peripheral volume V2, and -2.55 on the inter-compartmental clearance CL2 – but NOT on CL, which the authors attribute to insufficient sample size; the underlying body-composition formula is not stated in the paper), Yu_2020_inotersen.R (reference 51.6 kg cohort median, range 31.3-80.3; power exponents fixed at 1 on all four disposition parameters CL/F, Q/F, Vc/F and Vp/F; formula not stated in the paper, which tabulates lean body mass separately from lean body weight).
• Notes: The body-composition formula varies by source paper (James, Boer, Hume) and materially changes the value at a given weight and height; record the formula the source used in covariateData[[LBM]]$notes. The James formulae are LBM (female) = 1.07 * TBW - 148 * (TBW / height)^2 and LBM (male) = 1.1 * TBW - 128 * (TBW / height)^2, with TBW in kg and height in cm. When a paper reports LBM/LBW without naming the formula (as in Li_2023_TQB3203.R), say so explicitly in the per-model notes and record the cohort median and range so a downstream user can check that whatever formula they apply reproduces the source distribution – the formulae differ by several kg at a given height and weight, which directly rescales any median-normalized power effect.

FFM (canonical for fat-free mass)

• Description: Fat-free mass derived from body weight, height, and sex via the Janmahasatian et al. formula (Clin Pharmacokinet 2005;44:1051-1065).
• Units: kg
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (FFM / ref)^exponent. Reference values observed: 33 kg (Rolsma 2025 cefepime, paediatric + adult cystic fibrosis), 40.69 kg (Zhou 2021 belimumab pooled adult+pediatric SLE), 45 kg (Aguiar 2021, Crohn’s disease cohort median), 50 kg (Sinha 2026 oxcarbazepine, the “reference adult” fat-free mass), 58.9 kg (Abdul-Aziz 2025 caspofungin, critically ill adult ECMO cohort median).
• Source aliases:
  • LBW (“lean body weight”) – used in Rolsma_2025_cefepime.R and AbdulAziz_2025_caspofungin.R. Same biological quantity in kg with no value transformation. Registered here rather than under LBM because the register discriminates fat-free mass from lean body mass by the estimating formula, not by the source paper’s label, and both papers cite Janmahasatian et al. – the fat-free-mass equation. Abdul-Aziz 2025 makes the identification explicit in its Discussion: “Although the Janmahasatian et al. equation was developed to estimate fat-free mass, the estimate is considered to be a representation of lean body weight and is commonly used interchangeably.” Expect this alias frequently in the critical-care and antimicrobial-dosing literature, where “LBW” is the conventional label for the Janmahasatian quantity.
  • PKWT (“pharmacokinetic weight”) – used in Sinha_2026_oxcarbazepine.R. Same biological quantity in kg with no value transformation, but PKWT is a composite fat-free-mass metric: it equals FFM for subjects aged >= 3 years and total body weight (WT) for children younger than 3 years. The WT fallback exists because the Al-Sallami et al. paediatric FFM equation was developed in children >= 3 years, so Sinha 2026 Methods 2.3.2 assumes a WT-equivalent FFM below that age. A downstream user computing the column for a cohort that includes children under 3 years must apply the fallback; above 3 years PKWT and FFM are the same quantity.
• Example models: Zhou_2021_belimumab.R (reference 40.69 kg; exponents 0.673 on CL and 0.891 on V1), Aguiar_2021_ustekinumab.R (reference 45 kg; power exponents 0.598 on CL, 0.590 on Vc, 0.586 on Vp), Sinha_2026_oxcarbazepine.R (alias PKWT; reference 50 kg; exponent 1 fixed on oxcarbazepine clearance, 0.671 on MHD-metabolite clearance, and a shared 0.752 on both central volumes; paediatric cohort spanning 44 days to 20.9 years including 52% with obesity, where the FFM-based descriptor outperformed total body weight by ~7 AIC points), Rolsma_2025_cefepime.R (alias LBW; reference 33 kg; linear normalisation (FFM / 33) on the central volume with no exponent reported, in children and adults with cystic fibrosis), AbdulAziz_2025_caspofungin.R (alias LBW; reference 58.9 kg, the cohort median; estimated power exponent 1.24 on the volume of a one-compartment model, and the ONLY covariate retained – no covariate was kept on clearance – in 8 critically ill adults on ECMO. Note the exponent’s bootstrap 95% CI is 0.08-1.91, so it is barely distinguishable from the theory-based value of 1 for a volume; a useful precedent that a supra-linear FFM exponent on volume can be reported without being well identified), Tsirizani_2025_ritonavir.R (reference 21.0 kg, the cohort median, corresponding to a child of 26 kg – the anchor Table 3 footnote b names – in 170 African children aged 3.16-15.6 years on boosting-dose ritonavir; theory-fixed exponents 0.75 on CL/F and Q/F and 1 on Vc/F and Vp/F. FFM beat total body weight by 5 OFV points and also beat fat mass. Notable as a case where the reference value is recoverable ONLY from the supplemental NONMEM control stream (TVFFM = 21.0 ;MEDIAN): the main text gives the 26 kg weight anchor but never the corresponding FFM, and the paper never reports which equation produced the FFM data column, citing only Holford & Anderson 2017 for the allometric-size theory. A downstream user must therefore supply FFM directly or pick an equation; the paper’s own 21.0 kg / 26 kg pair is the only in-paper calibration available), Nyangwa_2026_pretomanid.R (reference 45.5 kg, the cohort median, in 94 adults with rifampicin-resistant TB; theory-fixed exponents 0.75 on CL/F and 1 on V/F, and the ONLY covariate retained – FFM beat both total body weight and BMI on base-model fit. A second case, alongside Tsirizani 2025, where the reference value is not printed anywhere: the authors published their nlmixr2 code as supplementary Appendix 1, but it scales on a pre-computed logFFM data column whose centring constant is never stated. The reference is recoverable by arithmetic rather than assumption – exp(lcl) and exp(lvc) equal the Table 2 typical values, which forces logFFM to be zero at the reference, and only a cohort-median reference reproduces the Table 3 median AUC(0-24) of 64,000 ug*h/L; 70 kg would sit 39% high and an uncentred log(FFM) 94% low. Like Tsirizani 2025 the paper never reports which equation produced the FFM column, leaving the cohort’s median weight of 56.8 kg against a median FFM of 45.5 kg as the only in-paper calibration point), Francis_2019_rifapentine.R (reference 46 kg; theory-fixed exponent 0.75 on CL/F in 326 southern African adults with pulmonary tuberculosis. The first registered instance where FFM and total body weight are used as size descriptors for DIFFERENT parameters in the SAME model rather than competing for both: FFM won for clearance by 23 OFV points over total weight, and total weight won for volume, so the model carries (FFM/46)^0.75 on CL and (WT/56)^1 on V. Note also that the reference value is reported inconsistently – Table 2 footnote b and the Results body both say 46 kg, while the Abstract says 45 kg, which is also the Table 1 cohort median. 46 kg is used because it appears in the parameter table’s own footnote; the choice moves typical clearance by only (46/45)^0.75 = 1.7 percent, which is the same “does the unprinted or contested reference actually matter” test the Isbister 2015 entry above works through for weight. Like Tsirizani 2025 and Nyangwa 2026, the paper does not report which equation produced the FFM column), Goyal_2022_vancomycin.R (reference 45 kg, the cohort median, in 34 hospitalized pregnant women; exponent 0.75 on CL and Q, 1 on Vc and Vp. Janmahasatian et al. for patients >= 18 years and Al-Sallami et al. for patients < 18 years, so one column spans two equations; FFM beat total body weight by 12 OFV points. Note that the Janmahasatian formula uses total body weight, which in pregnancy includes the fetus, placenta and amniotic fluid – the source does not adjust for this).
• Notes: Distinct from LBM (lean body mass) which is sometimes computed by the Boer or Hume formulae. When the source paper reports the body-composition formula it used (e.g., Janmahasatian for FFM), record it in covariateData[[FFM]]$notes. FFM is preferred over total body weight when scaling monoclonal-antibody PK because mAb distribution is largely confined to extracellular fluid; muscle / lean tissue tracks extracellular volume better than total weight in heavier patients. Paediatric extractions should record which FFM equation applies over which age band: Janmahasatian et al. (Clin Pharmacokinet 2005;44:1051-1065) is an adult formula, and Al-Sallami et al. (Clin Pharmacokinet 2015;54:1169-1178) supplies the paediatric age-correction multiplier a + (1 - a)/(1 + (PNA/b)^-c) applied to the adult form, with a = 0.88, b = 13.4, c = 12.7 for males and a = 1.11, b = 7.1, c = 1.1 for females. Note that the female numerator (1 - 1.11) is NEGATIVE; Sinha 2026 ESM1 Eq. S4 prints it as +0.11, which is a transcription error – the multiplier must converge to 1 at adolescence so the paediatric form reduces to the adult one.

IBW (canonical for ideal body weight)

• Description: Ideal body weight in kg, typically derived from height and sex using the Devine formula or its variants. Time-fixed at baseline unless the source paper states otherwise. Used in size-normalisation of clearance / dose-rate in adult popPK models where the source paper reports IBW as the preferred size descriptor over total body weight (e.g., when overweight subjects pull clearance scaling away from the typical pattern).
• Units: kg
• Type: continuous
• Scope: specific
• Reference category: n/a – used with a linear ratio (IBW / ref) for clearance / dose-rate normalisation. Reference values observed: 60 kg (Holford 1992 tacrine adult Alzheimer’s population mean IBW).
• Source aliases:
  • IBW – standard abbreviation used directly in Holford 1992.
• Example models: Holford_1992_tacrine.R (reference 60 kg; tacrine “clearance” relative to IBW = 60 enters as the dose-rate normalisation factor (60 / IBW) driving the tacrine effect-compartment input; the Holford-Peace Devine variant is documented in covariateData[[IBW]]$notes), VanWart_2004_garenoxacin.R (reference 64.2 kg; additive linear-deviation effect 0.764 mL/min per (IBW - 64.2) kg on CL/F, plus an in-model obesity flag derived as WT > 1.3 * IBW per Van Wart 2004’s >130% IBW obesity definition), Shoji_2011_pregabalin.R (reference 62 kg; power-of-ratio scaling on CL/F (exponent 0.354) and V/F (exponent 0.819) per Shoji 2011 Table 3 final model; the IBW formula is the Devine-family variant cited as reference [12] of the paper), Dirks_2008_cetuximab.R (reference 64 kg; additive linear-deviation effect 0.0108 per kg on Vmax of a two-compartment Michaelis-Menten model for cetuximab in SCCHN; the source dataset supplied IBW pre-computed, so the Devine-family variant used is not identified in the paper), Stoschus_2025_phenobarbital.R (reference 68.8 kg, the cohort median; classical allometric scaling of a one-compartment model, exponent 1 on V and 0.75 on CL, in critically ill adults with refractory / superrefractory status epilepticus. IBW was the only covariate retained of the demographic, clinical, laboratory, and comedication set screened. The source computed IBW with the formula of Brower et al. (Stoschus 2025 reference 25) but does not reproduce the formula, so the variant is not identifiable from the paper alone), GarciaHervalejo_2026_teicoplanin.R (reference 61 kg, the rounded development-cohort median of 61.4 kg; power-of-ratio scaling (IBW / 61)^3.2 on the clearance of a one-compartment teicoplanin model in adults with haematological malignancies. The steepest IBW effect registered against this canonical by a wide margin – the exponent of 3.2 is over four times the classical allometric 0.75, and it is well identified (RSE 21%, bootstrap 95% CI 2.13-4.78, so the interval excludes any allometric reading). It makes clearance span 0.33-3.05 L/h across IBW 40-80 kg and dominates the model: both of the paper’s headline dosing findings – that IBW <= 40 kg patients are predicted to exceed the 60 mg/L safety threshold at steady state, and that IBW >= 80 kg patients cannot hold 15 mg/L on once-daily dosing – fall straight out of this term. Two consequences for reuse. First, the steepness makes descriptor substitution unusually costly: total body weight exceeds IBW by only 11% at the cohort medians (68 vs 61.4 kg) yet supplying it inflates clearance by 1.11^3.2 = 39%. Second, the source prints the Devine formula with the final operator mis-set as a subtraction, 50.0 + 2.3 * ((height_cm - 152.4) - 2.54), which gives a 170 cm man an IBW of 84.6 kg; the intended division (height_cm - 152.4) / 2.54 gives 65.9 kg and is the only reading compatible with the cohort’s own median IBW of 61.4 kg at a median total body weight of 68 kg. The paper additionally chose IBW over total body weight, adjusted body weight, BMI and BSA in a head-to-head anthropometric screen, attributing the result to the cachexia, fluid overload and malnutrition that make total body weight overstate metabolically active mass in this population), Garreau_2021_vancomycin.R (reference 64.1 kg, Devine formula; enters the central volume of a continuous-infusion ICU vancomycin model as the exponential of a centred power term, V1 = V1pop * exp((IBW / 64.1)^1.88), exactly as printed in the paper’s Table 2 final-model equations, and IBW also enters clearance indirectly through the IBW-based Cockcroft-Gault CRCL. The paper retained IBW over total body weight, arguing that daily fluid shifts make TBW more variable within a patient than vancomycin V1 itself), Periclou_2021_cariprazine_initial.R (reference 64.5 kg, the cohort median; additive-linear (IBW - 64.5) effects on cariprazine CL/F and Vc/F, DCAR CL/F and DDCAR Vc/F, and power-of-ratio (IBW / 64.5)^e effects on DCAR Vc/F (exponent 3.16) and DDCAR CL/F (1.12). The paper’s updated model replaced IBW with total body weight ‘for practical use’ because both performed similarly, so the updated Periclou 2021 model carries WT instead), Duong_2022_gentamicin_hodiamont2017ijaa.R (reference 70 kg; allometric (IBW / 70)^0.75 on CL and linear (IBW / 70) on V1 of a two-compartment gentamicin model in critically ill adults – the Hodiamont 2017 IJAA structure refit by Duong 2022 to a Quebec ICU cohort. The IBW formula is not stated, and the refit dataset recorded no height, so the refit probably held IBW at the 70 kg typical value; the re-estimated IBW coefficients are therefore not informed by Quebec data).
• Notes: Specific scope until a second adult-popPK model ratifies the name; at that point promote to general. Per-model covariateData[[IBW]]$notes should record the formula the source paper used. The Holford-Peace 1992 variant is: men IBW (kg) = 52 + 0.75 * (height_cm - 152); women IBW (kg) = 49 + 0.67 * (height_cm - 152). The classic Devine 1974 formula is: men 50 + 2.3 * (height_in - 60); women 45.5 + 2.3 * (height_in - 60). Other variants (Robinson 1983, Miller 1983, Hamwi 1964) exist; the per-paper choice should be recorded so a user simulating against IBW can match the source’s derivation. When the source dataset supplies IBW pre-computed, the column name is typically IBW directly. When only HT + SEXF are provided, the user must compute IBW externally using the source-paper formula before passing it to the model.

ABW (canonical for adjusted body weight)

• Description: Adjusted (dosing) body weight in kg – the obesity-dosing size descriptor that interpolates between ideal and total body weight, conventionally ABW = IBW + 0.4 * (TBW - IBW). Time-fixed at baseline unless the source paper states otherwise. Distinct from WT (total/actual body weight), from IBW (ideal body weight alone), and from FFM (fat-free mass, a physiologically derived composition estimate rather than a dosing convention): ABW is the weight a prescriber substitutes for total body weight when dosing a hydrophilic drug in an overweight patient, and popPK papers screen it head-to-head against TBW and IBW as competing size descriptors.
• Units: kg
• Type: continuous
• Scope: specific
• Reference category: n/a – used with a linear-deviation form (intercept + slope * (ABW - ref)) or with a ratio/power form (ABW / ref)^exponent. Reference values observed: 60 kg (Jaruratanasirikul 2021, critically ill Thai adults with and without ECMO).
• Source aliases:
  • ABW – the standard abbreviation, glossed as “adjusted body weight” in the Zhang 2025 imipenem review’s Table 3 footnote.
• Example models: Jaruratanasirikul_2021_imipenem.R (reference 60 kg; linear-deviation effect on the central volume, V1 = 13.8 - 0.348 * (ABW - 60) L, retained over actual body weight, ideal body weight and BMI, which were screened alongside it. The effect is linear with a negative slope and therefore has a bounded domain: V1 reaches zero at ABW = 99.7 kg and is negative above it, so a virtual cohort driving this model must be restricted to the study’s body-weight range – see covariateData[[ABW]]$notes in that file).
• Notes: The interpolation factor 0.4 is the usual convention but is not universal (0.3 and 0.5 appear in the literature), and the IBW variant feeding it (Devine, Robinson, Hamwi, …) also varies; record the formula the source paper used in covariateData[[ABW]]$notes, or state explicitly that it is unreported. A source that reports ABW alongside WT and IBW in its screened-covariate list is using ABW in this dosing-convention sense; do not alias it to either. Specific scope until a second model ratifies the name, at which point promote to general. Note that a linear-in-ABW effect on a volume or clearance is only interpretable inside the fitted body-weight range – check the sign of the slope and the implied zero-crossing before simulating outside it.

HT (canonical for body height at baseline)

• Description: Subject body height at baseline. Time-fixed unless the source paper states otherwise.
• Units: cm
• Type: continuous
• Scope: general
• Reference category: n/a – used with a linear-deviation form (HT - ref) or with a power-style scaling. Reference values observed: 167 cm (Naik 2016, vortioxetine adult MDD/GAD population median); 165 cm (Zhang 2018 flurbiprofen and Angeli 2016 healthy non-menopausal women).
• Source aliases:
  • HGT – height (cm) abbreviation appearing in some NONMEM control streams.
  • HEIGHT – spelled-out form (used by Angeli 2016).
• Example models: Naik_2016_vortioxetine.R (reference 167 cm; linear-additive effect 0.40 L/hr per (HT - 167) cm on CL/F, retained over weight and BMI in stepwise selection because it produced the larger reduction in CL IIV), Zhang_2018_flurbiprofen.R (reference 165 cm; linear-multiplicative effect 1 + theta_height * (HT - 165) on the effect-compartment equilibration rate Ke alongside a paired linear-multiplicative WT effect), Angeli_2016_iron_hepcidin.R (reference 165 cm; power-law multiplier (HT / 165)^32.70 on the hepcidin post-menses rebound parameter krel_hep; the very large exponent reflects the narrow height range across the cohort, 158-173 cm at the 10th-90th percentile), Wu_2014_FEV1_asthma.R (per-observation HT in cm enters directly into the exponential FEV1 disease-progression model: FEV1 = exp(theta_age*AGE + theta_ht*HT - theta_int) + ... with theta_ht = 0.0169 1/cm; paediatric population, ages 5-13 y).
• Notes: Height is sometimes retained as a size covariate when allometric scaling on weight performs less well; it is also an input to BSA, BMI, FFM, and LBM derivations, so a model that retains HT alongside one of those derived covariates should document the dependency in covariateData[[HT]]$notes. Promoted from specific to general on 2026-06-03 with the Angeli 2016 iron / hepcidin extraction (the third adult-popPK model to register HT, satisfying the original promotion condition documented when the canonical was introduced).

BSA (canonical for body surface area)

• Description: Body surface area (typically computed by DuBois, Mosteller, or Haycock from height and weight).
• Units: m^2
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (BSA / ref)^exponent or with linear centering (1 + slope * (BSA - ref)). Reference values observed: 1.70 m^2 (Yamada 2025, adult oncology), 1.11 m^2 (Park 2023, paediatric HSCT cohort median).
• Source aliases: none.
• Example models: Yamada_2025_zolbetuximab.R (reference 1.70 m^2; power scaling, exponents 1.06 on clearances and 0.968 on volumes), Roepcke_2023_rezafungin.R (reference 1.90 m^2 = the pooled-cohort median; separately estimated exponents 0.882 on CL, 1.56 on the central volume V1, and 1.17 on the shared peripheral volume V23; BSA replaced body weight and BMI, which the paper notes were highly correlated with it and would have fit comparably; the BSA formula used is unspecified in the source), Park_2023_mycophenolic_acid.R (reference 1.11 m^2; centered-linear scaling Vd/F = 89.83 * (1 + 0.854 * (BSA - 1.11)) on the apparent volume of distribution in paediatric HSCT recipients; founding centered-linear example), Zhang_2015_methotrexate.R (reference 1.62 m^2; centered-linear scaling on BOTH the intercompartmental clearance, (1 + 0.880 * (BSA - 1.62)), and the peripheral volume, (1 + 0.874 * (BSA - 1.62)), in Chinese osteosarcoma patients spanning 0.62-2.21 m^2. Three points worth carrying forward. First, the centring constant 1.62 is the printed-equation value and is deliberately NOT the Table 1 median of 1.63 – the paper’s Methods template centres on the MEAN – another instance of the equation-over-demographics-table rule recorded throughout the CRCL entry. Second, the signs above are CORRECTED: the paper prints a minus in both brackets, which would make the peripheral volume shrink four-fold across the observed BSA range, while the paper’s own Methods covariate template carries a plus and its Discussion states the correlation is positive. Third, with a linear rather than power form and a cohort reaching down to 0.62 m^2, the multiplier falls to 0.12 at the smallest BSA – still positive, but a reminder that a centred-linear size term has a zero crossing that a power term does not, so check it before extrapolating below the fitted range), Yao_2025_flurbiprofen_r.R (reference 2.6 m^2 = the cohort median; power exponent 1.37 on the apparent central volume Vc of R(-)-flurbiprofen; BSA formula unstated and the tabulated values are NOT reproducible from the source’s own height and weight – see Notes), Xu_2026_caspofungin.R (reference 0.79 m^2 in a paediatric intensive-care cohort whose median BSA is 0.660 m^2 – the standardisation constant is NOT the cohort median, and is confirmed only by cross-checking the paper’s own BSA-normalised CL, V1 and ke against the Table 2 thetas; Mosteller formula, stated and reproducible from the source’s own height and weight; FIXED exponents 0.66 shared by CL and Q and 1 shared by V1 and V2, selected over weight, lean body weight and fat-free mass on OFV and AIC).
• Notes: Oncology mAbs dosed by BSA (mg/m^2) often use BSA in place of body weight for allometric-style scaling. Document the BSA computation formula (DuBois / Mosteller / Haycock) the source paper used; if unstated, record “unspecified.” Sanity-check the tabulated BSA against the source’s own height and weight before using it as a normalisation constant. Yao 2025 reports a median BSA of 2.6 m^2 (range 2.0-3.2) for a cohort whose median height and weight are 1.61 m and 70 kg, which give about 1.76 m^2 by DuBois or 1.79 m^2 by Mosteller – the tabulated values run roughly 1.45x high and no published BSA formula reproduces them. Because a covariate exponent is estimated against whatever BSA scale the source used, a model carrying such an effect must be driven with BSA on that same scale (and its virtual cohort must sample BSA from the reported distribution rather than recomputing it from height and weight); silently substituting a correctly computed BSA would shift every predicted volume.

BMI (canonical for body mass index)

• Description: Body mass index at baseline.
• Units: kg/m^2
• Type: continuous
• Scope: general
• Reference category: n/a – used with a linear-deviation form (1 + e * (BMI - ref)) or a power form ((BMI / ref)^e). Document the reference value in covariateData[[BMI]]$notes.
• Source aliases: none known.
• Example models: Chua_2025_mirikizumab.R (reference 24.75 kg/m^2; linear-deviation effect on logit of bioavailability), NA_NA_lidocaine.R (DDMODEL00000281; binary stratification at threshold 27.93 kg/m^2 adding +0.939 to the GX rate constant K30 in the BMI > 27.93 cohort), Struemper_2017_belimumab.R (kg/m^2, reference 24.7; power exponent -0.610 on Vc; baseline-only, source column BBMI), Lalovic_2020_lemborexant.R (power exponent -0.428 on CL/F; reference 25 kg/m^2, unprinted and back-solved from the paper’s quoted BMI effect sizes).
• Notes: Universal clinical-trial demographic. Derived as WT / (height_m)^2; assume time-fixed at baseline unless the source paper states otherwise.

BLOOD_VOLUME (canonical for estimated total blood volume)

• Description: Subject’s estimated total circulating blood volume (plasma plus red-cell volume), in litres, derived from body size and sex rather than measured. A body-size descriptor in the same family as BSA and LBM; baseline unless a model’s notes say otherwise.
• Units: L
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (BLOOD_VOLUME / ref)^exponent. Reference values observed: 3.89 L (Savic 2022, close to the 3.9 L cohort median of adults and adolescents with sickle cell disease).
• Source aliases:
  • BLV – Savic 2022 Table 2 row label ‘Blood volume on Vc/F, (BLV/3.89)^TH’ (Savic_2022_voxelotor.R).
• Example models: Savic_2022_voxelotor.R (baseline blood volume, reference 3.89 L, power exponent 0.74 on voxelotor apparent central volume; Vc/F 269 L at the 10th-percentile 2.9 L and 413 L at the 90th-percentile 5.2 L).
• Notes: Several formulas are in use (Nadler from height, weight and sex; weight-and-sex-only mL/kg rules; Lemmens for obesity) and they differ by several hundred mL at a given body size, so record the formula in covariateData[[BLOOD_VOLUME]]$notes. When the source does not print it – Savic 2022 says only ‘calculated based on body weight and sex’ – say so, and record the cohort median and range so a user can check that the formula they apply reproduces the source distribution. Distinct from organ or tissue blood-volume fractions inside a PBPK model (frb_<organ> in parameter-names.md), which are model parameters, not subject covariates.

BMIZ (canonical for body-mass-index z-score (age- and sex-standardised))

• Description: Age- and sex-standardised body-mass-index z-score (number of standard deviations above or below the reference-population mean BMI for the subject’s age and sex). Distinct from raw BMI (kg/m^2): BMIZ is unitless and centred at 0 in the reference population, so the reference value used in linear-deviation effects is 0 (not a population BMI in kg/m^2). Time-varying when the source paper carries a per-visit z-score; document baseline-vs-time-varying status in covariateData[[BMIZ]]$notes.
• Units: unitless (z-score; standard-deviation units)
• Type: continuous
• Scope: general
• Reference category: n/a – used with a linear-deviation form (1 + e * (BMIZ - 0)) so the reference is 0 (population mean for the subject’s age/sex). Effect coefficients are interpreted as fractional change per 1 z-score-unit deviation from the reference.
• Source aliases:
  • BMI – when a paper uses the column name BMI for what is actually a z-score (e.g., the Harun 2019 NMTRAN control stream column BMI is documented in the dataset header as “body-mass index z-score”). The canonical column is BMIZ; the source-paper column name is recorded in covariateData[[BMIZ]]$source_name.
• Example models: Harun_2019_cysticFibrosis.R (time-varying per-visit BMI z-score; linear-deviation effect on baseline FEV1% predicted with reference 0 and coefficient +0.0382 per z-score unit), Muhamad_2025_cholecalciferol_pbpk.R (baseline WHO 2007 BMI-for-age z-score driving the Monasor-Ortola 2021 cubic that splits body weight into the fat-mass and lean-mass 25(OH)D3 distribution compartments; source column ZBMI).
• Notes: Distinct from BMI (raw kg/m^2 used in adult populations). Paediatric and adolescent studies routinely report BMI as a z-score relative to a growth reference (WHO 2007 Growth Reference for school-aged children, CDC 2000, etc.); document the reference standard the source paper used in covariateData[[BMIZ]]$notes. Promoted to general scope when Muhamad_2025_cholecalciferol_pbpk.R became the second paediatric model to use it (WHO 2007 reference; Harun 2019 is the other).

HAZ (canonical for height-for-age z-score (age- and sex-standardised))

• Description: Height-for-age (or length-for-age) z-score; the standard anthropometric marker of chronic malnutrition / stunting. Unitless and centred at 0 in the reference population. Same conventions as WAZ.
• Units: unitless (z-score; standard-deviation units)
• Type: continuous
• Scope: general
• Reference category: n/a – linear-deviation form, reference 0 unless the source centres on a cohort median.
• Source aliases:
  • LAZ – length-for-age z-score, the same quantity for children measured recumbent rather than standing.
• Example models: Wallender_2021_piperaquine.R (screened on relative oral bioavailability, 4.7% per z-score, delta-OFV -5.93; NOT retained in the final model – carried in covariatesDataExcluded), Chotsiri_2019_lumefantrine.R (screened in the stepwise covariate search; not significant and not retained).
• Notes: Stunting axis. See WAZ Notes for the WT / MAL_NOURISH distinctions, which apply identically.

WHZ (canonical for weight-for-height z-score (sex-standardised))

• Description: Weight-for-height (or weight-for-length) z-score; the standard anthropometric marker of acute malnutrition / wasting, and a WHO severe-acute-malnutrition criterion at WHZ < -3. Unitless and centred at 0 in the reference population.
• Units: unitless (z-score; standard-deviation units)
• Type: continuous
• Scope: general
• Reference category: n/a – linear-deviation form, reference 0 unless the source centres on a cohort median.
• Source aliases:
  • WLZ – weight-for-length z-score, the same quantity for children measured recumbent rather than standing.
• Example models: Wallender_2021_piperaquine.R (screened on relative oral bioavailability, 4.4% per z-score, delta-OFV -6.47; NOT retained – carried in covariatesDataExcluded), Chotsiri_2019_lumefantrine.R (screened alongside MUAC and WAZ; MUAC gave the larger drop in objective function and was retained instead).
• Notes: Wasting axis; distinct from MUAC, the other continuous acute-malnutrition measure (raw mm), and from the binary MAL_NOURISH. See WAZ Notes.

SEXF (canonical for sex)

• Description: Biological sex indicator, 1 = female, 0 = male.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (male).
• Source aliases:
  • SEXM (values inverted: SEXF = 1 - SEXM; effect coefficient sign and reference category both invert) – used in CarlssonPetri_2021_liraglutide.R.
  • SEX with "M"/"F" strings – derive SEXF = as.integer(SEX == "F").
  • SEX with 1=male / 2=female numeric coding – derive SEXF = as.integer(SEX == 2). Used in Netterberg_2017_docetaxel.R and NA_NA_miridesap.R (DDMODEL00000262 source bundle; Sahota 2015 NONMEM convention).
  • FEM (1 = female, 0 = male; same orientation as canonical, no transformation) – used in Guiastrennec_2016_gastric_emptying.R.
• Example models: Zhu_2017_lebrikizumab.R (canonical), CarlssonPetri_2021_liraglutide.R (alias SEXM), Bajaj_2017_nivolumab.R (male-indicator source; effect applied as exp(coef * (1 - SEXF)) to preserve the paper’s female-reference CL_REF / VC_REF), Fau_2020_isatuximab.R (exponential effect on Vc; reference category 0 = male), Netterberg_2017_docetaxel.R (multiplicative effect on baseline ANC: BACOV *= (1 + theta * SEXF); source column SEX with 1 = male, 2 = female encoding, decomposed via SEXF = as.integer(SEX == 2)), NA_NA_miridesap.R (DDMODEL00000262 / Sahota 2015; multiplicative effect on baseline SAP via SAP_BASE_ref * (1 + e_sexf_sap0 * SEXF) with e_sexf_sap0 = -0.30; female baseline is ~30% lower than male), Xu_2020_daratumumab.R (additive shift on V1 (1 + e_sexf_vc * SEXF) with e_sexf_vc = -0.205: female V1 is 20.5% lower than male, reference category 0 = male), Guiastrennec_2016_gastric_emptying.R (multiplicative +40.7% strengthening of the caloric-feedback slope SLPCAL on gastric emptying in females; SLPCAL_eff = SLPCAL * (1 + 0.407 * SEXF)), Wada_2023_sparsentan.R (male-indicator source with a FEMALE reference subject, so the published +0.139 log-scale male effect on CL/F is applied as exp(e_sexf_cl * (1 - SEXF)) to preserve the verbatim coefficient while keeping the canonical 1 = female orientation – the same construction as Bajaj_2017_nivolumab.R; men had ~15% higher CL/F), Wang_2019_oral1cmt_covariate.R (male-indicator source with a FEMALE reference subject, in the FRACTIONAL rather than exponential form: the authors’ TVCL=THETA(1)*CLWT*(1 + CLSEX*SEX) with CLSEX = -0.3 is applied as (1 + e_sexf_cl * (1 - SEXF)), so lcl = log(15) is the clearance of a 70 kg woman and men clear 30% more slowly; verified numerically in the vignette, which asserts CL = 15 L/h for a 70 kg woman and 10.5 L/h for a 70 kg man).
• Notes: When translating a model that used SEXM, document the sign/reference-category inversion in the model’s covariateData notes.

PREG (canonical for pregnancy status indicator)

• Description: 1 = pregnant, 0 = non-pregnant. Usually time-fixed per subject, in trial cohorts that enrol pregnant and non-pregnant women in parallel. It is genuinely time-varying in designs that re-sample the same women after delivery: Sawe_2025_levofloxacin.R pools matched antepartum and postpartum profiles, so a subject carries PREG = 1 on the third-trimester visit and PREG = 0 on the postpartum visit.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (non-pregnant).
• Source aliases: none known; source NONMEM control streams typically use PREG directly.
• Example models: Sawe_2025_levofloxacin.R (time-varying across matched antepartum / postpartum visits; fractional multiplicative effect 1 + e_preg_cl * PREG raising levofloxacin clearance 38.1% in the third trimester, on top of the separately-modelled FFM and serum-creatinine effects. Pregnancy was screened as a three-level categorical covariate – pregnant / postpartum / never pregnant – but only the pregnant level was retained, so postpartum records take the reference category and their intermediate exposures are reproduced by the size and renal-function covariates alone), Birgersson_2019_artesunate.R (multiplicative effect on dihydroartemisinin clearance; the published structural CLM = 190 L/h is reported with the source-paper reference category PREG = 1, so the model file applies the effect via (1 + e_preg_cl_dha * (1 - PREG)) with e_preg_cl_dha = -0.214 to preserve verbatim source values; non-pregnant women have ~21% lower CLM relative to pregnant women).
• Notes: Use this canonical for adult clinical-trial models that test a pregnancy-vs-non-pregnancy contrast (typical settings: malaria-in-pregnancy PK, antiviral-in-pregnancy PK). Trimester or gestational-age stratification within the pregnant cohort should use a separate canonical (e.g., gestational-age weeks via GA or a trimester indicator, ratified separately when needed). The canonical convention is reference category 0 (non-pregnant) following the broader pharmacology default; source papers that use the pregnant cohort as the reference (Birgersson 2019) preserve their published structural values via a (1 - PREG) form on the effect coefficient.

NEUTERED (canonical for surgical neutering / gonadectomy status)

• Description: 1 = the animal has been surgically neutered (a spayed female, i.e. ovariectomy or ovariohysterectomy, or a castrated male, i.e. orchiectomy); 0 = sexually intact. The routine companion-animal demographic descriptor, reported in veterinary papers alongside sex as “spayed female” / “castrated male” / “intact”. Time-fixed per animal in practice, since the procedure precedes enrolment in every study that has used it so far.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (sexually intact).
• Source aliases:
  • sterilization status / sterilized – Beguin 2024 Table 1 and Results wording for the same 0/1 indicator. Same orientation, no value transformation. Prefer the canonical NEUTERED: “sterilized” collides with the microbiological sense of the word, whereas “neutered” is unambiguous in the companion-animal literature.
  • neutered status – Beguin 2024 Methods wording; the source column name recorded in that model’s source_name.
• Example models: Beguin_2024_carboplatin_dog.R, Beguin_2024_carboplatin_thrombocytopenia_dog.R (carboplatin clearance in client-owned dogs; multiplicative log-scale effect exp(e_neutered_cl * NEUTERED) with e_neutered_cl = -0.22, so neutered dogs clear carboplatin about 20% more slowly, reference category 0 = intact).
• Notes: Orthogonal to SEXF, and both may be retained in the same model: a cohort routinely holds intact females, spayed females, intact males and castrated males, and Beguin 2024 screened the two separately, retaining NEUTERED and discarding SEXF. Distinct from the PRIOR_<procedure> surgical-history family (PRIOR_GAST, PRIOR_SPLEN): those record a disease-driven resection that forms part of the treatment history, whereas neutering is an elective demographic attribute recorded at intake. Ratified as a bare-word demographic canonical alongside SEXF and PREG (PMC11664934, 2026-09-11). Interpret a retained effect cautiously: Beguin 2024 found neutering significant on clearance but argue it is most likely confounded with age and plasma creatinine rather than causal, since neutering is not expected to alter GFR and a 2021 iohexol study in dogs found no neutering effect.

CHILD (canonical for child age-cohort indicator)

• Description: 1 = subject is a child, 0 = not a child. In non-human models this is the general juvenile-versus-adult age-cohort indicator (1 = juvenile animal, 0 = adult animal); the concept is the same binary immature-versus-mature contrast, only the species-specific age cutoff changes.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (not child, i.e., adult baseline).
• Source aliases:
  • PED – used in the Schoemaker 2018 LEV / BRV pediatric extrapolation (DDMODEL00000239) as the pediatric-vs-adult indicator that gates the Markov-amplitude term, the overdispersion IIV, and the four pediatric offsets on log baseline rate / mixture / placebo / Emax / EC50.
  • adult_calve – used in Winter_2024_oxytetracycline_cattle.R (and in the raw data set distributed with Winter 2024 as code_adult_calve) as the calf-vs-adult-cattle indicator on all three volumes and all three clearances. Same coding as the canonical: 0 = adult, 1 = calf.
  • AGEgr – used in CarlssonPetri_2015_liraglutide.R as the paper’s pediatric-vs-adult age-category contrast (ESM Table 1 row Cov CL-AGEgr). Same coding as the canonical: 0 = adult, 1 = pediatric.
  • adult – used in Papathanasiou_2021_somatropin.R (Methods section 2.6, ‘adult is a discrete value taking 1 or 0’), switching the typical Emax and its weight relationship and the GH residual error between adults and children. INVERTED coding relative to the canonical: CHILD = 1 - adult.
• Example models: CarlssonPetri_2015_liraglutide.R, CarlssonPetri_2021_liraglutide.R, Papathanasiou_2021_somatropin.R, Schoemaker_2018_levetiracetam.R (DDMODEL00000239), Winter_2024_oxytetracycline_cattle.R.
• Notes: Paper’s age cutoffs must be captured in covariateData[[CHILD]]$notes. That requirement is load-bearing for veterinary and preclinical models, where the cutoff is species-specific and bears no relation to the human paediatric ranges (Winter 2024 defines a calf as under 6 months of age, or any animal the original study authors declared to be a calf). CHILD is often paired with ADOLESCENT when a paper splits the immature population into two bands (CarlssonPetri_2021_liraglutide.R: children 7-11 y, adolescents 12-17 y), but it is equally the right column when a paper models a single pediatric-vs-adult contrast and ADOLESCENT would have no referent – CarlssonPetri_2015_liraglutide.R uses CHILD alone for its 10-17 y pediatric cohort against a 33-73 y adult cohort, and Schoemaker_2018_levetiracetam.R likewise. Record which band the indicator spans in the per-model notes so the two usages are never confused.

ADOLESCENT (canonical for adolescent age-cohort indicator)

• Description: 1 = subject is an adolescent, 0 = not.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0.
• Example models: CarlssonPetri_2021_liraglutide.R.
• Notes: Paired with CHILD. Document age cutoffs.

FM (canonical for fat mass)

• Description: Fat mass in kg derived as the complement of fat-free mass, FM = TBW - FFM, where FFM is computed by the Janmahasatian et al. (2005) formula (or, equivalently, by direct measurement via bioelectrical-impedance analysis / DXA when the source paper specifies it). Time-fixed at baseline.
• Units: kg
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (FM / ref)^exponent. Reference values observed: 19 kg (Robarge 2017 efavirenz; cohort median calculated FM).
• Source aliases: FAT, FATMASS.
• Example models: Robarge_2017_efavirenz.R (reference 19 kg; exponent 1 fixed on V_p/F).
• Notes: Natural complement of FFM. Use both in the same model when the paper scales different disposition parameters by FFM vs FM (e.g., Robarge 2017 efavirenz: CL/F by FFM, V_p/F by FM, motivated by lipophilic partitioning into adipose tissue). When the source paper reports the body-composition formula it used (Janmahasatian for FFM is the most common; direct DXA / BIA also acceptable), record it in covariateData[[FM]]$notes.

VISCERAL_ABDOMINAL_FAT (canonical for visceral abdominal fat area from imaging)

• Description: Visceral abdominal fat area measured from a single computed-tomographic (CT) slice at the L2 / L3 vertebral level (standard IVGTT / metabolic-imaging protocol; Jensen et al. 1995). Captures the metabolically active visceral adipose-tissue depot that drives insulin resistance independently of subcutaneous fat. Time-fixed at baseline.
• Units: cm^2 (single CT slice). Document per-model via covariateData[[VISCERAL_ABDOMINAL_FAT]]$units when a paper uses a different anatomical level (L3 / L4, L4 / L5) or a different summary form (volumetric L1-L5 cm^3, MRI-derived).
• Type: continuous
• Scope: specific
• Reference category: n/a – used with mean-centred linear-in-log form log(SI) = log(SI_ref) + theta * (VISCERAL_ABDOMINAL_FAT - ref). Reference values observed: 141.8 cm^2 / CT-slice (Denti 2010 Table 1 pooled-population mean across 204 healthy adults aged 18-87).
• Source aliases:
  • VAF – universal abbreviation in the IVGTT / glucose-metabolism literature (Denti 2010, Basu 2003, Basu 2006). The canonical column is VISCERAL_ABDOMINAL_FAT; the source paper’s column name is recorded in covariateData[[VISCERAL_ABDOMINAL_FAT]]$source_name.
• Example models: Denti_2010_glucoseMinimal.R (introduces canonical; reference 141.8 cm^2 / CT-slice; linear-in-log effect coefficient -0.00208 per cm^2 deviation on insulin sensitivity lsi).
• Notes: Specific scope because the value reported is tied to the imaging modality (single-slice CT at L2 / L3 in the founding example). Volumetric CT, multi-slice averages, and MRI-derived volumetric adiposity from future papers can ratify under the same canonical with the assay method and units recorded per-model in covariateData[[VISCERAL_ABDOMINAL_FAT]]$notes. Distinct from BODYFAT_PCT (whole-body adiposity fraction by DEXA) and from FFM / LBM (fat-free mass / lean body mass). The H3-name uses underscores to preserve readability; future authors may abbreviate the source column as VAF in their own datasets and declare source_name = "VAF".

BODYFAT_PCT (canonical for percent total body fat (whole-body adiposity fraction))

• Description: Percent total body fat, measured as the fraction of total body mass classified as fat tissue by an imaging assay (typically dual-energy X-ray absorptiometry, DEXA). Expressed as a percentage (0-100). Time-fixed at baseline unless the source paper states otherwise. Captures whole-body adiposity, distinct from regional adiposity (visceral, abdominal, subcutaneous) and from fat-free mass.
• Units: % (percent; 0 = no body fat, 100 = all body fat). Document per-model via covariateData[[BODYFAT_PCT]]$units when a paper uses a fractional 0-1 scale.
• Type: continuous
• Scope: general
• Reference category: n/a – used with mean-centred linear-in-log form log(VD) = log(VD_ref) + theta * (BODYFAT_PCT - ref), or as the direct determinant of an adipose-compartment volume in whole-body PBPK. Reference values observed: 32.39 % (Denti 2010 Table 1 pooled-population mean across 204 healthy adults); 25 % (Levitt PKQuest standard human, i.e. 17.5 kg adipose of a 70 kg total, used by Pei_2023_tacrolimus_pbpk.R).
• Source aliases:
  • %TBF – paper symbol in the IVGTT / glucose-metabolism literature (Denti 2010, Basu 2003). The leading % character is dataset-incompatible (NMTRAN / R column-name rules); the canonical column is BODYFAT_PCT and the paper’s symbol is recorded in covariateData[[BODYFAT_PCT]]$source_name.
  • PTBF – numeric-safe rendering of the paper symbol used in some derived datasets.
• Example models: Denti_2010_glucoseMinimal.R (introduces canonical; reference 32.39 %; linear-in-log effect coefficient -0.0101 per percentage-point deviation on glucose distribution volume lvd), Pei_2023_tacrolimus_pbpk.R (whole-body PBPK organ-volume scaling: the adipose compartment weight is set directly as WT * BODYFAT_PCT / 100 and the remaining lean organ weights and blood flows scale by WT * (1 - BODYFAT_PCT / 100) / 52.5 kg; Pei 2023 Methods 2.5 names “the proportion of adipose tissue” as one of the two determinants of organ volume and Table 5 reports its local sensitivity).
• Notes: Promoted from Scope: specific to Scope: general on 2026-08-05 alongside the Pei 2023 tacrolimus PBPK extraction. The original specific scope reflected the founding example’s DEXA assay, but percent total body fat is an assay-independent body-composition quantity and the second registered use (a whole-body PBPK adipose-compartment volume) is unrelated to how the value was measured; the assay method belongs in the per-model covariateData[[BODYFAT_PCT]]$notes, not in the scope. Bioimpedance- or skinfold-derived %BF from future papers ratifies under the same canonical on the same terms. Distinct from FFM (fat-free mass in kg) and from VISCERAL_ABDOMINAL_FAT (regional visceral adiposity in cm^2 / CT slice). The negative effect coefficient on glucose distribution volume captures the physiological observation that glucose distributes preferentially into lean rather than adipose tissue, so per-kg distribution volume decreases as fat fraction rises.

WT_BASE (canonical for per-subject baseline body weight (time-fixed, paired with time-varying WT))

• Description: Per-subject baseline body weight, time-fixed. Used when a source paper applies Wahlby 2004’s extended covariate-model split (Br J Clin Pharmacol 2004;58(4):367-377) and the data column carries the subject’s baseline weight as a constant alongside a time-varying WT column. Operationally, WT_BASE is equal to WT at the subject’s first observation and remains constant for all subsequent records; the within-subject delta is computed in model as (WT - WT_BASE).
• Units: kg
• Type: continuous
• Scope: general
• Reference category: n/a – typically used as (WT_BASE - ref) in an exponential or linear-deviation effect on a structural parameter, or as the ratio (WT_BASE / ref) in a power effect, where ref is the cohort median baseline weight.
• Source aliases:
  • BWT (baseline weight) – Wahlby 2004 source-column convention; also the Verrest 2023 $INPUT column name, annotated there as “BWT: baseline body weight (kg)”; used in Wahlby_2004_pefloxacin.R and Verrest_2023_paromomycin.R.
• Example models: Wahlby_2004_pefloxacin.R (reference 65 kg; saturating “up to median WT” qualifier from Wahlby 2004 Methods plateaus the effect above 65 kg, encoded as min(WT_BASE, 65) - 65), Verrest_2023_paromomycin.R (reference 27.5 kg, the cohort median; standard allometric power scaling, exponent 0.75 on CL/F and Q/F and 1.00 on Vc/F and Vp/F, in Eastern African children and adults with visceral leishmaniasis), Yang_2025_iruplinalkib.R (reference 63 kg, the pooled cohort median and the paper’s own typical-subject value; estimated power exponents 0.441 on CL/F and 1.36 on V1/F in Chinese healthy volunteers and patients with ALK/ROS1-positive solid tumors), Osawa_2018_daclatasvir.R (reference 56 kg, the cohort median; estimated power exponent 0.605 on V/F only, in Japanese adults with chronic genotype-1 HCV infection; the Verrest/Yang shape – the paper names the covariate BBWT and uses time-varying laboratory covariates elsewhere in the same analysis, but does not use Wahlby’s within-subject delta form), Bi_2018_testosteroneCypionate.R (reference 85 kg on CL and V of the testosterone popPK and 84.70 kg on the LH30 inhibitory potency – the two differ because the PK and PD layers were separate NONMEM runs; estimated power exponents 0.785 on CL/F, 1.71 on V/F and -1.14 on the LH30 IC50, paired with an explicit (WT - WT_BASE) delta effect on CL), Bi_2018_testosteroneCypionate_sperm.R (reference 85 kg; power exponent -1.27 on the inhibitory potency of the spermatogenesis indirect-response model).
• Notes: Specific scope because the BCOV/DCOV decomposition pattern is paper-defined: WT_BASE exists conceptually as a per-subject snapshot of baseline weight that the original modeller chose to enter separately from the time-varying weight column. Distinct from time-fixed adult-cohort weight (where the per-subject weight is just constant by data design); use WT_BASE only when the source paper deliberately separates the BCOV from the time-varying COV. Verrest 2023 is a second, weaker ratification: its dataset carries both a time-varying WT and a baseline BWT column and its control stream deliberately writes the allometry against BWT, but it does not use Wahlby’s within-subject delta form (WT - WT_BASE) – it simply prefers baseline weight as the size descriptor. (The paper’s own Table 3 equation prints the subscript as WT_i,t, which would imply the time-varying column; the control stream is the executed model and disagrees.) Yang 2025 is a third ratification of the same Verrest-shaped kind, and the clearest one on the question of deliberateness: that paper screened every laboratory covariate “at baseline and at the time of measurements” and retained three of them (ALB, CRCL, LDH) in their explicitly time-varying form, while naming the weight covariate BBWT – baseline body weight – in the same breath and in the same printed equations. The baseline-versus-time-varying distinction is therefore demonstrably a modelling choice the authors were making consciously, column by column, which is exactly the condition this canonical exists to record. It still does not use the delta form. Scope is left at specific; promote to general once a paper ratifies the column with Wahlby’s within-subject delta form (WT - WT_BASE), which is the semantic none of the three current examples exercises.

BSA_BASE (canonical for per-subject baseline body surface area (time-fixed))

• Description: Per-subject baseline body surface area, time-fixed. Used when a source paper enters the subject’s baseline BSA as a constant alongside a time-varying BSA column under Wahlby 2004’s extended covariate-model decomposition (Br J Clin Pharmacol 2004;58(4):367-377). Time-fixed at the subject’s first observation.
• Units: m^2
• Type: continuous
• Scope: specific
• Reference category: n/a – typically used as a linear scaling factor on a volume parameter.
• Source aliases:
  • BBSA (baseline BSA) – Wahlby 2004 source-column convention; used in Wahlby_2004_gentamicin.R and Garrett_2019_inotuzumab.R.
• Example models: Wahlby_2004_gentamicin.R (replaces time-varying BSA in the final-model V1 equation V1 = 8.63 * BSA_BASE * (ALB/34)^-0.41 because BBSA was the better predictor than time-varying BSA, RSE 2.5% vs 110% on delta-BSA), Garrett_2019_inotuzumab.R (power exponents 1.54 on cl_exp_inf and 1.64 on cl_exp_component, both centered on the 1.84 m^2 population median, plus a LINEAR effect on vc in the centered deviation, 1 + 0.774 * (BSA_BASE - 1.84); baseline BSA displaced body weight in that paper’s covariate screen and, because InO is dosed per m^2, the retained effect supports BSA-based dosing).
• Notes: Specific scope because the BCOV/DCOV split is a paper-defined modelling choice (Wahlby 2004 demonstrated that BBSA was the relevant predictor for gentamicin V1 because delta-BSA was uninformative). Garrett 2019 ratifies the same baseline-BSA column with consistent semantics in a second, unrelated therapeutic area (oncology ADC popPK), and is a reminder that the functional form attached to this covariate is not always a power model – read the source equation rather than assuming (BSA_BASE / ref)^theta. Promote to general if a third paper ratifies it.

BSA_AFFECTED_PCT (canonical for baseline percentage of body surface area affected by a skin disease)

• Description: Per-subject baseline percentage of total body surface area affected by a skin disease (psoriatic plaque, atopic-dermatitis lesion, vitiligo, burn). Time-fixed baseline value from the screening / baseline skin assessment. An extent-of-disease measure, not a body-size measure.
• Units: %
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (BSA_AFFECTED_PCT / <ref>)^exponent. Reference value observed: 31% (Li_2022_gumokimab.R, Li 2022 Eq 7).
• Source aliases:
  • BSA / BSA (%) / “body surface area (BSA) involvement” – Li 2022 gumokimab (AK111) population PK/PD; used in Li_2022_gumokimab.R.
  • TBSA (total body surface area burned) – the burn-literature form of the same quantity.
• Example models: Li_2022_gumokimab.R (power exponent -0.572 on kout, the first-order rate of psoriatic plaque loss in the PASI indirect-response model, normalized to 31%; cohort mean 33.3%, SD 14.7%, in Chinese adults with moderate-to-severe plaque psoriasis).
• Notes: NOT body surface area in m^2: papers in dermatology routinely abbreviate the affected percentage as “BSA”, which collides with the canonical BSA (m^2) and BSA_BASE. Read the units in the source’s demographics table before mapping a “BSA” column. Named without a _BL suffix by analogy with BMBLAST_PCT (a per-subject baseline percentage); a model needing a time-varying affected percentage should propose a separate column. Scope general because the same quantity recurs across skin diseases and burns.

BLOOD_GROUP_O (canonical for ABO blood group O indicator)

• Description: 1 = subject has ABO blood group O (either O-positive or O-negative Rh subtype); 0 = subject has ABO blood group A, B, or AB (non-O). Time-fixed per subject. Enters factor VIII (FVIII) and von Willebrand factor (VWF) population PK models as a surrogate for baseline VWF plasma concentration: blood group O individuals have ~25% lower circulating VWF than non-O individuals due to accelerated VWF clearance by the ASGR / CLEC4M receptor system (VWF’s ABH glycan antigens are recognized differently by clearance receptors depending on the ABO glycosyltransferase phenotype). Because VWF binds and stabilizes circulating FVIII, lower VWF in blood group O individuals translates to faster FVIII clearance.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (non-O ABO blood group: A, B, or AB).
• Source aliases:
  • blood group (paper-prose 0/1 indicator, 1 = O) – used in Hazendonk_2016_factor_viii.R (Hazendonk 2016 Table 5: 1.26 ^ blood_group on CL, so blood group O subjects have 26% higher FVIII clearance than non-O).
• Example models: Hazendonk_2016_factor_viii.R (multiplicative effect on CL: (1 + 0.26 * BLOOD_GROUP_O) – equivalent to 1.26 ^ BLOOD_GROUP_O when binary; reference category 0 = non-O ABO blood group).
• Notes: General scope because ABO blood group is a universally recorded clinical variable that recurs in FVIII / VWF / VWF-related-protein PK models (Hazendonk 2016 Discussion cites the well-established 25% lower baseline VWF in blood group O as the mechanism; also documented in Gill 1987, Miller 1985). The canonical is orthogonal to VWF (the actual plasma VWF concentration): when both are available, papers typically retain VWF and drop BLOOD_GROUP_O because VWF captures the effect more directly (e.g., Nestorov 2014 uses VWF directly; Hazendonk 2016 uses BLOOD_GROUP_O because VWF measurements were only available for ~50% of the cohort). Founding example: Hazendonk 2016.

WAIST (canonical for waist circumference)

• Description: Waist circumference measured by tape at baseline; an abdominal-adiposity anthropometric that is distinct from whole-body size. Enters population PK models as a normalised power term (WAIST / ref)^exponent.
• Units: cm
• Type: continuous
• Scope: general
• Reference category: n/a – normalised power. Reference value observed: 101 cm (Pan 2026, the cohort median from Table 1).
• Source aliases:
  • WAIST – Pan 2026 NONMEM $INPUT (Data S1 Code S1).
  • Waist (cm) – Pan 2026 Table 1 row label.
• Example models: Pan_2026_adalimumab.R (founding example; (WAIST/101)^0.888 on apparent clearance of subcutaneous adalimumab in psoriasis, the strongest single covariate on CL after body weight).
• Notes: Distinct from [[MUAC]] (mid-upper arm circumference), a nutritional-status proxy measured on the arm and stored in millimetres, used chiefly in paediatric and undernutrition cohorts. Distinct from [[VISCERAL_ABDOMINAL_FAT]], an imaging-derived cross-sectional area rather than a tape measurement. Distinct from [[BMI]], which is a whole-body ratio: the two are correlated, but they are not interchangeable, and Pan 2026 screened both and retained waist while rejecting BMI – so a model that retains a waist effect must use this column rather than substituting BMI. Where a source paper codes missing waist measurements with a sentinel (Pan 2026 uses -99 and sets the whole covariate term to 1), supply the model’s reference value instead of the sentinel: the reference reproduces the “no effect” branch exactly, whereas a negative sentinel raised to a fractional power is undefined. Ratified 2026-09-02 alongside the Pan 2026 adalimumab extraction.

RESIDENCE_RURAL (canonical for rural-versus-urban settlement type of the subject’s residence)

• Description: 1 = the subject resides in a rural area; 0 = urban. A subject-level baseline attribute describing the character of the settlement a subject lives in, not its location. Deliberately disease- and country-neutral: a rural/urban split recurs across indications and geographies, and burning the founding study’s disease or country into the name would wrongly imply the concept is specific to it. What each model must document in its own covariateData[[RESIDENCE_RURAL]]$notes is how the source study drew the line, since the rural/urban boundary is defined by national census convention and is not comparable across countries without comment.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (urban). Chosen to match the printed reference row of the founding source, so that the published hazard ratio for rural residence transcribes directly with no polarity flip. Note that urban is the minority stratum in the founding study (16 of 222 episodes, 7.2%), so the reference group there is the smaller one and the estimate is correspondingly imprecise.
• Source aliases:
  • Residence – Tchaparian 2016 Table 3 row label, with Rural / Urban as the two printed levels and Urban carrying the 1.00 reference.
  • Episodes involving urban-resident children, no. (%) – Tchaparian 2016 Table 1 row label, which reports the complement of this indicator.
• Example models: Tchaparian_2016_lumefantrine_recurrence.R (founding example; adjustment covariate of the adjusted Cox proportional-hazards model for 28-day recurrent Plasmodium falciparum parasitaemia after artemether-lumefantrine in Ugandan children, e_residence_rural_haz = log(1.21) on the log-hazard scale, Table 3).
• Notes: Orthogonal to the REGION_<PLACE> family (REGION_TANZANIA, REGION_MOZAMBIQUE, REGION_JAPAN, …), whose members name a geographic PLACE – a study site, enrollment country or continent. Settlement type is a cross-cutting attribute within any such place: a single-country study has one REGION_ value and can still vary on this column, which is exactly the case in the founding study. Do not fold a rural/urban split into a REGION_ indicator or vice versa. Also distinct from SITE-style study-centre identifiers, which index an administrative unit rather than describe it. General scope on first use because the concept carries no study-specific scale or units; only the national convention behind the split is study-specific, and that belongs in the per-model notes. Ratified 2026-09-21 alongside the Tchaparian 2016 lumefantrine extraction.

Pediatric / maturation

PAGE (canonical for postmenstrual age)

• Description: Postmenstrual age in months (GA_weeks / 4.35 + postnatal_months). Time-varying.
• Units: months
• Type: continuous
• Scope: general
• Example models: Clegg_2024_nirsevimab.R, Robbie_2012_palivizumab.R, Germovsek_2018_meropenem.R (weeks), Riccobene_2017_ceftaroline.R (weeks). Franzese_2022_ceftazidime.R / Franzese_2022_avibactam.R (weeks; Rhodin maturation TM50 47.7 weeks, Hill 3.4, active only for AGE <= 2 years; PMA taken as postnatal age + 40 weeks where unknown).
• Notes: Months is the register default, but the neonatal maturation literature (Rhodin 2009 and everything built on it) states postmenstrual age in WEEKS, and its reference constants – TM50 = 47.7 weeks, a 33-week pivot, a 5.51-week maturational half-life – are only meaningful on that scale. A model whose source equations are written in weeks therefore declares units = "weeks" in its own covariateData entry and uses weeks throughout; the canonical name is unchanged. Always read the model file’s covariateData$PAGE$units before supplying this column.

PNA (canonical for postnatal age)

• Description: Postnatal age (chronological since birth). Time-varying.
• Units: months
• Type: continuous
• Scope: general
• Source aliases:
  • PNA – used in Zhao 2018 (paper Methods ‘Population pharmacokinetic-pharmacogenetic modelling’ and Table 2 report PNA in DAYS; the canonical PNA carries months, so Zhao 2018’s F_PNA = (PNA_days / 38)^0.472 is reparameterised inside model() as F_PNA = (PNA_months / 1.249)^0.472 using the conversion PNA_months = PNA_days / 30.4375 and reference 1.249 months = 38 days / 30.4375).
  • Age (h) – used in Bardhi 2026 (Bardhi_2026_ampicillin_foal.R), whose neonatal-foal cohort spans 9-194.5 h of life and whose covariate is centred on the weighted-mean age of 61.8 h. Same reparameterisation as Zhao 2018 but from HOURS: model() recovers hours with PNA * 730.5 (1 month = 30.4375 days = 730.5 h) before forming the age ratio, so the centring value 61.8 h corresponds to 0.0846 months. Bardhi 2026 Table 2’s legend prints “61.8 years”, a typographical slip for hours.
• Example models: Hu_2026_clesrovimab.R, Zhao_2018_omeprazole.R (power effect on the formation clearance of 5-hydroxy-omeprazole: (PNA / 1.249)^0.472; PNA reference 1.249 months / 38 days from Zhao 2018 Table 2 cohort median), Bardhi_2026_ampicillin_foal.R (veterinary; power effects on both clearance, exponent 0.40, and peripheral volume, exponent 1.38, of (age_h / 61.8)).
• Notes: Sub-daily postnatal ages are legitimate members of this canonical – Bardhi 2026’s hospitalized neonatal foals are enrolled from 9 h of life – and are carried in months like every other PNA value, with the source unit recovered inside model.

GA (canonical for gestational age at birth)

• Description: Gestational age at birth. Time-fixed per subject.
• Units: weeks
• Type: continuous
• Scope: general
• Example models: Hu_2026_clesrovimab.R, Clegg_2024_nirsevimab.R (folded into PAGE).

WT_BIRTH (canonical for birth weight)

• Description: Body weight at birth. Time-fixed per subject. Distinct from current body weight (WT), which is time-varying.
• Units: kg
• Type: continuous
• Scope: general
• Reference category: n/a – used with linear-deviation (1 + e * (WT_BIRTH - ref)) or power scaling. Reference value observed: 2.59 kg (Voller 2017 newborn cohort).
• Source aliases:
  • BWEIGHT – used in Voller_2017_phenobarbital.R (Voller 2017 source data column for birth weight in kg).
  • Bwb – used in Wu_2025_paracetamol.R (Wu 2025 source-paper symbol for birth weight in kg; same biological quantity, no value transformation).
• Example models: Voller_2017_phenobarbital.R (linear-deviation effect on CL: clbw = 1 + 0.369 * (WT_BIRTH - 2.59)), Wu_2025_paracetamol.R (power-law scaling of PTNA CL_birth and GFR_birth terms with reference 1.75 kg).
• Notes: Time-fixed at birth; characterises pre-/term-newborn cohorts. Pairs with GA (gestational age at birth) when both are reported. The conventional clinical-PK abbreviation BWT is intentionally NOT used as the canonical name because it is already used across the codebase (Gandhi 2021, Li 2019, Chen 2022, Wojciechowski 2022, Lu 2019) as a source-name alias for body weight (WT). The WT_BIRTH form keeps the WT root consistent with the existing body-weight canonical and avoids the BWT ambiguity.

HT_BIRTH (canonical for birth length)

• Description: Body length measured at birth. Time-fixed per subject. Distinct from HT (baseline or current body height), on exactly the principle that distinguishes WT_BIRTH from WT: a neonatal dataset can carry both a daily body length and the value recorded at birth, and a model that fitted one must not be driven with the other.
• Units: cm
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (HT_BIRTH / ref)^exponent or with linear centering. Reference value observed: 33 cm (Padavia 2024, extreme-preterm cohort median 32.75 cm rounded to the value the published equation uses).
• Source aliases:
  • BL – Padavia 2024 source-paper symbol for birth length in cm (beta_V2_BL); same quantity, no value transformation.
• Example models: Padavia_2024_paracetamol.R (reference 33 cm; power exponent 6.46 on the paracetamol peripheral volume of distribution V2, the only size covariate the paper retained on a volume).
• Notes: Registered because the source dataset separately carries a time-varying daily body length – Padavia 2024 Methods 2.2 states that “Bodyweight and body length were also collected every day during treatment”, while Results 3.2 and the Conclusion both name the retained covariate specifically as birth length. Reusing HT here would make one canonical name mean two different measurements in the same dataset, so the fixed-at-birth value gets its own column just as birth weight does. Pairs with WT_BIRTH and GA in pre-/term-newborn cohorts. A model retaining HT_BIRTH alongside the time-varying WT (as Padavia 2024 does) should say so in covariateData, because the two size descriptors then differ in whether they are fixed at birth and a user joining the wrong column into either one gets a silently wrong model.

AGE_DPF (canonical for zebrafish-larval age in days post-fertilization)

• Description: Age of a zebrafish (Danio rerio) larva in days post-fertilization (dpf). Time-fixed per subject in destructive-sampling designs (each larva is harvested at exactly one observation, so its dpf is fixed at the value assigned at exposure-start).
• Units: days post-fertilization (dpf)
• Type: continuous
• Scope: specific
• Reference category: n/a – used in van Wijk 2019 with a step-form effect on the absorption rate (IF (AGE_DPF > 3) k12 = k12_3 * (1 + e_age_dpf_k12)) and a per-day power-form effect on the elimination rate (k25 = k25_3 * (1 + e_age_dpf_k25)^(AGE_DPF - 3)). Reference age is 3 dpf (the youngest cohort in the study).
• Source aliases:
  • AGE – van Wijk 2019 NONMEM column. Renamed to canonical AGE_DPF because the human-PK canonical AGE denotes subject age in years; the zebrafish dpf semantic is incompatible and would silently corrupt any future model that mixed them. Same orientation, no value transformation.
• Example models: vanWijk_2019_paracetamol.R.
• Notes: Specific scope because the meaning is bounded to zebrafish-larval-development PK studies. Distinct from canonical AGE (human age in years), PNA (postnatal age in months), PAGE (postmenstrual age in months), and GA (gestational age in weeks) – none of those are appropriate for a non-mammalian organism whose developmental clock is anchored at fertilization rather than birth. Integer values 3, 4, 5 in the van Wijk 2019 dataset, but treated as a continuous covariate in the elimination-rate power form. Future zebrafish-PK or other non-mammalian-developmental-age models should reuse this canonical only when the covariate is indeed dpf-anchored; other developmental-time conventions (e.g., somite-stage, hpf, dph) would warrant separate canonicals.

ASPHYXIA (canonical for perinatal asphyxia indicator (low 5-minute Apgar score))

• Description: Binary indicator of perinatal asphyxia at birth defined by a 5-minute Apgar score < 5; 1 = 5-minute Apgar score was less than 5 (i.e., the neonate met the paper’s perinatal-asphyxia criterion), 0 = 5-minute Apgar score was 5 or higher. Time-fixed per subject (an event at birth, not a time-varying physiologic state). Used by neonatal popPK models that report a structural-PK shift in subjects who experienced perinatal asphyxia relative to those who did not.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no perinatal asphyxia; 5-minute Apgar >= 5). Effect form in Grasela 1985 is a linear-deviation increment on the volume of distribution: V = V_typical * (1 + e_asphyxia_vc * ASPHYXIA) with e_asphyxia_vc = 0.13 (+13% V in subjects with 5-minute Apgar < 5; p < 0.05).
• Source aliases:
  • None registered (Grasela 1985 abstract names the covariate asphyxia in prose; the abstract does not report the source NONMEM column label).
• Example models: Grasela_1985_phenobarbital.R (Grasela & Donn 1985; preterm-neonate phenobarbital popPK; +13% on V when 5-minute Apgar < 5; no detected effect on CL).
• Notes: Distinct from the continuous Apgar score itself (which has not been registered as a canonical because no model in nlmixr2lib uses the continuous form as of this entry). The < 5 cutoff at 5 minutes is the perinatal-asphyxia convention Grasela 1985 used; future papers that adopt a different Apgar cutoff (e.g., < 7 at 1 minute, or a different time point) should register a separate canonical (e.g., APGAR1_LT7) rather than reusing ASPHYXIA with relaxed semantics.

Mother-infant dyad partner

In a dyad model the modelled SUBJECT is the mother, so the general WT, AGE, and PAGE canonicals already denote her. Describing the breastfed infant with those same names would put two different people’s demographics on one record under one canonical name, and nothing in the column name would say which person it referred to. The _INFANT suffix marks a column as belonging to the dyad PARTNER rather than to the modelled subject. It parallels the infant_<canonical> compartment namespace in compartment-names.md.

Use these columns only in a genuine dyad model, i.e. one that carries maternal and infant quantities simultaneously. A standalone paediatric popPK model whose subject is the infant uses the plain WT / AGE / PAGE canonicals, because there is no second person to disambiguate against.

WT_INFANT (canonical for breastfed-infant body weight in a mother-infant dyad model)

• Description: Body weight of the breastfed infant paired with the modelled lactating subject. Distinct from WT, which in a dyad model is the mother’s body weight.
• Units: kg
• Type: continuous
• Scope: general
• Reference category: n/a – used both to size the maternal breast-milk compartment (milk ingested per feed = daily milk intake per kg x WT_INFANT / feeds per day) and, in a full dyad model, to scale the infant’s clearances and volumes allometrically from the mother’s, (WT_INFANT / WT)^0.75 on clearance and (WT_INFANT / WT) on volume.
• Source aliases:
  • INFWT – NONMEM $INPUT column name in Wattanakul_2024_primaquine.R / Wattanakul_2024_primaquine_motherinfant.R (INFWT; INFANT BODY WEIGHT (KG)); same quantity in kg, no value transformation.
• Example models: Wattanakul_2024_primaquine.R (enters only through the breast-milk compartment volume V_M = (0.15 L/kg/day * WT_INFANT) / feeds-per-day; cohort median 6.8 kg, range 4.13-10.8, giving milk-compartment volumes of 0.062-0.162 L), Wattanakul_2024_primaquine_motherinfant.R (additionally scales the infant’s apparent clearances and volumes from the mother’s estimates).
• Notes: Baseline in the founding models, but time-varying in principle – infant weight changes materially over a multi-month lactation period, and dyad simulations that sweep infant age (Wattanakul 2024 uses WHO weight-for-age curves with z-scores -3 to +3 from birth to 24 months) pair each AGE_INFANT with the corresponding WT_INFANT. Verify baseline vs. time-varying against the source paper.

AGE_INFANT (canonical for breastfed-infant postnatal age in a mother-infant dyad model)

• Description: Postnatal age of the breastfed infant paired with the modelled lactating subject. Distinct from AGE, which in a dyad model is the mother’s age.
• Units: months
• Type: continuous
• Scope: general
• Reference category: n/a – typically enters an enzyme-maturation function on the infant’s clearance. When the maturation function is written in postmenstrual age, the model derives PMA internally by adding an assumed term gestation to AGE_INFANT rather than requiring a separate column, because individual gestational ages are rarely reported in lactation studies; the assumed gestation must be stated in the model file.
• Source aliases:
  • INFAGE – NONMEM $INPUT column name in Wattanakul_2024_primaquine_motherinfant.R (INFAGE; INFANT AGE (MONTHS)); same quantity in months, no value transformation.
• Example models: Wattanakul_2024_primaquine_motherinfant.R (drives monoamine-oxidase-A maturation on the infant’s primaquine clearance, MF = PMA / (TM50 + PMA) with PMA = AGE_INFANT + 9.2 months for assumed full-term gestation of 40 weeks and TM50 = 7.6 months; cohort median 5.0 months, range 1.6-21.7).
• Notes: Months, not years – lactation cohorts are dominated by infants under two years old, where a year scale loses resolution, and paediatric maturation functions are conventionally written in months or weeks of postmenstrual age. This matches the units of the general PAGE canonical. Distinct from PAGE: AGE_INFANT is postnatal, so a model that needs postmenstrual age must add the gestational term explicitly.

NAT2_SLOW_INFANT (canonical for breastfed-infant NAT2 slow-acetylator phenotype in a mother-infant dyad model)

• Description: 1 = the breastfed infant paired with the modelled lactating subject is an NAT2 (arylamine N-acetyltransferase 2) slow acetylator; 0 = the infant is an intermediate or rapid (fast) acetylator. Same phenotype definition, same slow-versus-nonslow pooling and same _SLOW = 1 orientation as the general NAT2_SLOW canonical, which in a dyad model denotes the MOTHER. Time-fixed per infant (germline genotype-derived phenotype).
• Units: (binary)
• Type: categorical
• Scope: general
• Reference category: 0 (nonslow infant: intermediate or rapid acetylator pooled).
• Source aliases:
  • metaboliser status (child) / child: fast metaboliser / child: slow metaboliser – prose form used by Garessus_2019_isoniazid_pbpk.R when it enumerates the four mother-child metaboliser pairs; NAT2_SLOW_INFANT = as.integer(child status == "slow").
• Example models: Garessus_2019_isoniazid_pbpk.R (selects the infant’s hepatic clearance between the two values observed by Rey 2001 in children under 6 months, 2.55 L/h fast and 0.76 L/h slow; crossed with NAT2_SLOW to produce the four dyad phenotype pairs of Figures 4 and 5).
• Notes: Member of the _INFANT dyad-partner family established by WT_INFANT and AGE_INFANT, and registered for the same reason: the mother and the infant each have their own acetylator phenotype, they are genetically independent, and a dyad model must carry both simultaneously, so one column name cannot serve both people. A lactation-transfer model in which only the mother’s phenotype matters needs NAT2_SLOW alone; this canonical is required only when the infant side of the dyad has its own metabolism. Do not use it for a standalone paediatric popPK model whose subject is the infant – that model has no second person to disambiguate against and uses the plain NAT2_SLOW. The companion NAT2_RAPID canonical has a dyad-partner form NAT2_RAPID_INFANT under the same rule, reserved here but not yet exercised; a dyad model that resolves all three infant acetylator levels should register it at that point.

BREASTFEED_EXCLUSIVE (canonical for exclusive-breastfeeding indicator)

• Description: Whether the modelled lactating subject was breastfeeding her infant exclusively (1) or mixing breast milk with formula or complementary food (0). Screened in lactation popPK analyses as a proxy for milk maturity and for the volume and composition of milk produced, both of which can influence drug concentration in milk. Unlike WT_INFANT and AGE_INFANT, this describes the feeding RELATIONSHIP rather than the dyad partner’s own demographics, so it carries no _INFANT suffix.
• Units: (binary)
• Type: categorical
• Scope: general
• Reference category: 0 (not exclusively breastfeeding, i.e. mixed feeding).
• Source aliases:
  • breastfeeding exclusivity – prose form used in Baklouti_2026_amoxicillin.R (Baklouti 2026 Methods section 2.3, “breastfeeding exclusivity (yes/no, categorical variable)”) and screened under the same wording by Melander_2025_cetirizine.R.
• Example models: Baklouti_2026_amoxicillin.R (screened in the stepwise covariate search and not retained; documented in covariatesDataExcluded).
• Notes: Registered by the maintainers 2026-09-02 (PMC13206287) from a documentation-only use, deliberately ahead of the first model that RETAINS it. Lactation popPK is now a recurring family in this library (Wattanakul_2024_primaquine.R, Wattanakul_2024_primaquine_motherinfant.R, Ojara_2024_lamivudine.R, Melander_2025_cetirizine.R, Deferm_2025_lactation_physiology.R, Baklouti_2026_amoxicillin.R) and exclusivity is a standard screened covariate across it, so registering the name now pre-empts a divergent coinage later. checkModelConventions() does not validate names appearing only in covariatesDataExcluded, so this entry is documentation rather than a lint target until a model references it in model().

Nutritional status

MAL_NOURISH (canonical for malnutrition status indicator)

• Description: 1 = subject is malnourished at study entry per a paper-defined anthropometric criterion (e.g., WHO height-for-age and weight-for-age Z-scores both < -2.0 in Tikiso 2021); 0 = not malnourished. Subject-level baseline indicator; the time-decaying recovery during nutritional supplementation is carried by the paired T_NUT_SUPP column rather than as a time-varying value of MAL_NOURISH.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (not malnourished).
• Source aliases:
  • MAL – used in Tikiso_2021_abacavir.R (the dataset’s paper-defined indicator, 1 = malnourished, 0 = not malnourished).
  • NUT – used in CatalanLatorre_2018_taurine_rat.R (the dataset’s paper-defined indicator, 1 = undernourished UN, 0 = well-nourished WN).
  • Group – used in Kir_2025_atenolol_rat_pbpk.R and Kir_2025_metoprolol_rat_pbpk.R (the Monolix regressor named Group that selects the control vs malnourished parameter set).
• Example models: Tikiso_2021_abacavir.R (gates the time-decaying malnutrition effect: mal_decay = MAL_NOURISH * exp(-T_NUT_SUPP * log(2) / 12.2), which then drives multiplicative shifts of +115% on F and -64% on CL at the start of nutritional supplementation, decaying with a 12.2-day half-life).
  • Tikiso_2021_abacavir.R (gates the time-decaying malnutrition effect: mal_decay = MAL_NOURISH * exp(-T_NUT_SUPP * log(2) / 12.2), which then drives multiplicative shifts of +115% on F and -64% on CL at the start of nutritional supplementation, decaying with a 12.2-day half-life).
  • CatalanLatorre_2018_taurine_rat.R (static baseline indicator – no T_NUT_SUPP pairing because there was no nutritional rehabilitation phase in the preclinical Wistar-rat study; MAL_NOURISH = 1 reduces the saturable tubular secretion Vmax of taurine by 9.4% relative to well-nourished animals).
  • Kir_2025_atenolol_rat_pbpk.R (static baseline indicator in a preclinical Sprague-Dawley mPBPK model; no T_NUT_SUPP pairing because there was no nutritional rehabilitation phase. Unusually, MAL_NOURISH here gates ABSORPTION rather than disposition: it switches all three sequential zero-order absorption rates, the presence and end time of the third absorption window, the blood-to-plasma ratio, and the proportional residual-error magnitude. The structural disposition parameters fd1, Kp1 and CL were deliberately shared across nutrition groups).
  • Kir_2025_metoprolol_rat_pbpk.R (companion to the atenolol model above, from the same paper; MAL_NOURISH switches both zero-order absorption rates, the end time of the second absorption window, the blood-to-plasma ratio, the additive residual-error magnitude, and the fraction of hepatic intrinsic clearance operating in the oral arm).
• Notes: Specific scope because the malnutrition definition (WHO Z-score thresholds in Tikiso 2021; end-of-adaptation body weight below 80% of the well-nourished mean AND serum albumin below 23 g/L in Catalan-Latorre 2018; a 5% protein isocaloric diet fed for 17-20 days, confirmed by significant falls in body weight, serum albumin and total cholesterol, in Kir 2025; mid-upper arm circumference, weight-for-height vs height-for-age, etc.) is paper-defined; per-model covariateData[[MAL_NOURISH]]$notes must document the criterion used. Pairs with T_NUT_SUPP (days on nutritional supplementation) when the model uses a time-decaying recovery function; otherwise MAL_NOURISH alone serves as a static baseline indicator. Distinct from generic body-weight Z-scores (which are continuous anthropometric metrics rather than a binarised malnutrition indicator). For a model that distinguishes malnutrition SEVERITY levels, use the graded MAL_NOURISH_MILD / MAL_NOURISH_MOD / MAL_NOURISH_SEV indicators instead of this pooled one.

MAL_NOURISH_MILD (canonical for mild malnutrition indicator)

• Description: 1 = mild malnutrition per a paper-defined grading, 0 = not malnourished or a different malnutrition level. One of three mutually exclusive severity indicators (MAL_NOURISH_MILD / MAL_NOURISH_MOD / MAL_NOURISH_SEV); all three 0 is the non-malnourished reference.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (not malnourished, when MAL_NOURISH_MOD and MAL_NOURISH_SEV are also 0).
• Source aliases:
  • M (“mild nutritional impairment”) – used in Sjogren_2021_malnutrition_physiology.R (the Barac-Nieto et al. grading, body weight / height 89.5 % of standard).
• Example models: Sjogren_2021_malnutrition_physiology.R (selects the ‘Mild’ column of the physiological scaling parameters that transform a non-malnourished PBPK physiology into a malnourished one).
• Notes: Specific scope because the grading scheme is paper-defined (Barac-Nieto body weight / height bands in Sjogren 2021; Gomez weight-for-age or WHO z-score bands elsewhere); per-model covariateData[[MAL_NOURISH_MILD]]$notes must document it. Follows the HEPIMP_MILD / HEPIMP_MOD / HEPIMP_SEV severity-suffix family. Distinct from the pooled binary MAL_NOURISH, which does not separate levels; a model that estimates or tabulates per-level effects must use these graded indicators.

MAL_NOURISH_MOD (canonical for moderate / intermediate malnutrition indicator)

• Description: 1 = moderate (or ‘intermediate’) malnutrition per a paper-defined grading, 0 = not malnourished or a different malnutrition level. Mutually exclusive with MAL_NOURISH_MILD and MAL_NOURISH_SEV.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (not malnourished, when MAL_NOURISH_MILD and MAL_NOURISH_SEV are also 0).
• Source aliases:
  • I (“intermediate nutritional impairment”) – used in Sjogren_2021_malnutrition_physiology.R (the Barac-Nieto et al. grading, body weight / height 82.7 % of standard).
• Example models: Sjogren_2021_malnutrition_physiology.R (selects the ‘Intermediate’ column of the physiological scaling parameters).
• Notes: The middle level of a three-level grading, whatever the paper calls it (‘moderate’, ‘intermediate’); the _MOD suffix follows the HEPIMP_MOD / RENALIMP_MOD family. See MAL_NOURISH_MILD for scope and the distinction from MAL_NOURISH.

MAL_NOURISH_SEV (canonical for severe malnutrition indicator)

• Description: 1 = severe malnutrition per a paper-defined grading, 0 = not malnourished or a different malnutrition level. Mutually exclusive with MAL_NOURISH_MILD and MAL_NOURISH_MOD.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (not malnourished, when MAL_NOURISH_MILD and MAL_NOURISH_MOD are also 0).
• Source aliases:
  • S (“severe nutritional impairment”) – used in Sjogren_2021_malnutrition_physiology.R (the Barac-Nieto et al. grading, body weight / height 73.9 % of standard).
• Example models: Sjogren_2021_malnutrition_physiology.R (selects the ‘Severe’ column of the physiological scaling parameters; the level the paper uses for every pediatric PK evaluation, giving a virtual weight-for-height z-score of about -3).
• Notes: See MAL_NOURISH_MILD for scope and the distinction from MAL_NOURISH. WHO ‘severe acute malnutrition’ (WHZ < -3 and/or MUAC < 115 mm) is one clinical definition of this level; where a model uses the continuous measurements themselves, carry WHZ / MUAC instead.

MUAC (canonical for mid-upper arm circumference)

• Description: Mid-upper arm circumference, the continuous anthropometric measurement of the circumference of the upper arm at its midpoint. A standard WHO nutritional-status measure in children aged 6-59 months; the WHO severe-acute-malnutrition threshold is MUAC < 115 mm. Time-fixed at admission in the models that use it, though it is measurable repeatedly during nutritional rehabilitation.
• Units: mm
• Type: continuous
• Scope: general
• Reference category: n/a
• Source aliases:
  • MUAC – used in Chotsiri_2019_lumefantrine.R (Chotsiri 2019 Table 1 and Table 2 report MUAC in mm and the covariate coefficient per cm; the model therefore converts MUAC / 10 inside model).
• Example models: Chotsiri_2019_lumefantrine.R (continuous exponential effect on the relative bioavailability of oral lumefantrine: fmuac = exp(e_muac_f * (MUAC / 10 - muac_ref)) with e_muac_f = -log(1 - 0.254) per cm and muac_ref = 13.0 cm, i.e. bioavailability falls 25.4% per 1 cm reduction in MUAC).
• Notes: Store the raw measurement in millimetres – the unit WHO uses for the SAM threshold and the unit source tables almost always tabulate – and convert inside model when a paper publishes its coefficient per centimetre. Distinct from the binary MAL_NOURISH: MAL_NOURISH is a paper-defined malnourished/not-malnourished indicator, whereas MUAC is the underlying continuous measurement and is one of the criteria papers use to set MAL_NOURISH. A model that retains a continuous MUAC effect must use MUAC; binarising it onto MAL_NOURISH would discard the dose-response. Where a paper reports both, carry both columns. Also distinct from the weight-for-height / weight-for-age / height-for-age z-scores (WHZ / WAZ / HAZ families), which are population-standardised scores rather than a raw circumference.

WAZ (canonical for weight-for-age z-score (age- and sex-standardised))

• Description: Age- and sex-standardised weight-for-age z-score: the number of standard deviations a child’s body weight lies above or below the median weight of a reference population of the same age and sex. Unitless and centred at 0 in the reference population. The WHO classifies WAZ < -2 as underweight and WAZ < -3 as severely underweight, and papers routinely make the -3 cut-off a breakpoint rather than using the score linearly across its whole range. Distinct from the raw body weight WT (kg), which is carried separately and is what allometric scaling uses – a model that adjusts for both size (WT) and nutritional status (WAZ) needs both columns. Baseline in the founding model, but time-varying in principle during nutritional rehabilitation; document baseline-vs-time-varying status in covariateData[[WAZ]]$notes.
• Units: unitless (z-score; standard-deviation units)
• Type: continuous
• Scope: general
• Reference category: n/a – the reference is 0 (the reference-population median for the child’s age and sex). Where a paper applies the effect only below a threshold (commonly the WHO severe-underweight cut-off -3), the effect multiplier must evaluate to exactly 1 at and above that threshold; encode the below-threshold depth as max(0, threshold - WAZ) so the covariate is inert for adequately-nourished children.
• Source aliases:
  • ZWFA – used in SemereGebreyesus_2024_rifabutin.R (the source paper’s column name; “weight-for-age z score”, computed from WHO growth charts at ages <= 10 years and US CDC growth charts above 10 years).
• Example models: SemereGebreyesus_2024_rifabutin.R (threshold effect on rifabutin oral bioavailability, active only below the WHO severe-underweight cut-off: waz_below3 <- max(0, -3 - WAZ) then fdepot *= (1 + e_waz_fdepot)^waz_below3 with e_waz_fdepot = -0.26, i.e. bioavailability falls 26.0% per z-score unit below -3 and is unchanged at WAZ >= -3; Semere Gebreyesus 2024 Table 2 and Results “Covariate model”), Wallender_2021_piperaquine.R (time-varying; linear-deviation effect on relative oral bioavailability, (1 + 0.113 * (WAZ - (-0.5))), centred on the cohort median -0.5, so each 1 z-score decrease lowers F by 11.3%), Chotsiri_2019_lumefantrine.R (screened alongside MUAC and WHZ and found significant, but not retained in the final model; carried as documentation).
• Notes: Belongs to the same anthropometric z-score family as BMIZ (body-mass-index z-score) and, when future papers register them, HAZ (height/length-for-age, reported as ZHFA by some papers) and WHZ (weight-for-height). Record which growth reference the source paper used – WHO Child Growth Standards, the WHO 2007 Growth Reference for school-aged children, or CDC 2000 – in covariateData[[WAZ]]$notes, because the same measured weight maps to materially different z-scores across references, and paediatric papers frequently switch reference at age 5 or 10 years. Distinct from the binary MAL_NOURISH (a paper-defined malnourished/not-malnourished indicator, which a WAZ threshold is often used to set) and from MUAC (a raw circumference in mm): a model that retains a continuous or threshold-graded z-score effect must use WAZ, since binarising it onto MAL_NOURISH would discard the dose-response. Where a source paper states a per-unit percentage change without printing the equation, prefer the compounding form (1 + e)^depth over the linear form (1 + e * depth): it matches the literal reading of “X% for each unit”, it keeps a bioavailability or clearance multiplier strictly positive at the extreme z-scores reachable in severely malnourished cohorts, and it is the form the register already uses for the closely analogous per-unit nutritional-anthropometry effect on bioavailability in Chotsiri_2019_lumefantrine.R (MUAC).

T_NUT_SUPP (canonical for time on nutritional supplementation)

• Description: Time elapsed since the start of nutritional / refeeding supplementation, in days. 0 at the start of supplementation; increases with time. Used by population PK models that describe the time-varying recovery of PK parameters during nutritional rehabilitation in malnourished cohorts.
• Units: days
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as the argument of an exponential decay function exp(-T_NUT_SUPP * log(2) / T_half) whose half-life T_half is an estimated parameter (12.2 days in Tikiso 2021).
• Source aliases:
  • TNUTRI – used in Tikiso_2021_abacavir.R (the dataset’s paper-defined column for days since start of nutritional supplementation; same orientation as the canonical, 0 = start of supplementation, increasing with time on supplementation).
• Example models: Tikiso_2021_abacavir.R (paired with MAL_NOURISH; drives the recovery decay exp(-T_NUT_SUPP * log(2) / 12.2) of the malnutrition effect on F and CL).
• Notes: Specific scope because the underlying recovery dynamics are tied to the nutritional-rehabilitation protocol of the source study. For non-malnourished subjects (MAL_NOURISH == 0) the value is irrelevant because the malnutrition effect is gated by MAL_NOURISH; supply 0 as a default. For fully-recovered malnourished subjects, supply a large value (e.g., >= 100 days, well beyond the 12.2-day Tikiso 2021 half-life) so the decay function reaches near zero and the effect vanishes.

DRINK_OGTT (canonical for oral-glucose-tolerance-test-only drink indicator)

• Description: Binary indicator that the postprandial test drink is glucose-only (oral glucose tolerance test, OGTT). 1 = the drink is glucose-only with no fat content (e.g., 25 / 75 / 125 g OGTT); 0 = otherwise (water, or any drink containing fat). Per-occasion (per-test-drink-administration) covariate – a single subject in a crossover challenge protocol receives different drink types across occasions, so the indicator varies per dose-record rather than per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (water or fat-containing drink). Mutually exclusive with DRINK_FAT – both indicators cannot be 1 simultaneously.
• Source aliases: paper narrative “glucose solution” / “OGTT” cohort labels driving the gastric-emptying-onset T50OGTT selection in Guiastrennec 2016.
• Example models: Guiastrennec_2016_gastric_emptying.R (selects the OGTT-specific half-onset time T50OGTT = 15.7 min for the gastric-emptying delay Hill function; water is recovered when DRINK_OGTT = DRINK_FAT = 0 with the onset factor pinned to 1).
• Notes: Specific scope because the OGTT / fat-containing partition is tied to the Guiastrennec 2016 postprandial-challenge design (Studies B and C OGTT arms). Set to 1 for Study B’s 25 / 75 / 125 g OGTT drinks and the Study C 75 g OGTT arm; set to 0 for Study A water and for all fat-containing drinks. Pairs with DRINK_FAT: the two indicators jointly select the appropriate gastric-emptying-delay T50 parameter (T50OGTT vs T50Fat) for the Hill onset function.

DRINK_FAT (canonical for fat-containing test-drink indicator)

• Description: Binary indicator that the postprandial test drink contains fat (any nonzero fat content). 1 = the drink contains fat (e.g., the Study C low / medium / high-fat isocaloric drinks, or the Study D medium-high-fat drink); 0 = otherwise (water or glucose-only OGTT drinks). Per-occasion covariate – a single subject in a crossover challenge protocol receives different drink types across occasions.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (water or glucose-only drink). Mutually exclusive with DRINK_OGTT – both indicators cannot be 1 simultaneously.
• Source aliases: paper narrative “low-fat” / “medium-fat” / “high-fat” / “medium-high-fat” cohort labels driving the gastric-emptying-onset T50Fat selection in Guiastrennec 2016.
• Example models: Guiastrennec_2016_gastric_emptying.R (selects the fat-specific half-onset time T50Fat = 23.1 min for the gastric-emptying delay Hill function).
• Notes: Specific scope because the fat-containing partition is tied to the Guiastrennec 2016 postprandial-challenge design. Pairs with DRINK_OGTT: the two indicators jointly select the appropriate gastric-emptying-delay T50 parameter (T50OGTT vs T50Fat) for the Hill onset function.

GE_T10_BL (canonical for baseline (drug-free) gastric emptying t10)

• Description: Per-subject time, in minutes, for 10% of a standardized test meal to leave the stomach, measured before treatment (drug-free). In Kim 2021 it comes from the 13C-Spirulina gastric emptying breath test (GEBT): the baseline kPCD values are converted to gastric emptying fractions at fixed timepoints with the Szarka 2008 regressions (which also use sex and BMI), and t10 is interpolated from those fractions. A short t10 means fast early gastric emptying.
• Units: min
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as a power term normalised to a reference, (GE_T10_BL / 30)^theta; Kim 2021 uses 30 min. Record the per-model reference in covariateData[[GE_T10_BL]]$notes.
• Source aliases:
  • HT10 – used in Kim_2021_YH12852.R (supplementary NONMEM control stream, SLP = THETA(9) * EXP(ETA(8)) * (HT10/30)**THETA(11)); same orientation and units, no transformation.
• Example models: Kim_2021_YH12852.R (power effect on the linear slope of plasma YH12852 on the stomach emptying rate, exponent 3.57; Kim 2021 Table 1).
• Notes: Member of the <X>_BL baseline family (QTC_BL, MAP_BL, OFFTIME_BL, HAERATE_BL). Not <X>_BASE, which is reserved for a time-fixed baseline paired with a time-varying column of the same quantity; no time-varying gastric emptying column exists. A gastric emptying half-time (t50) covariate, screened but not retained in Kim 2021, would be a sibling GE_T50_BL. Time-fixed per subject. Distinct from the test-drink indicators DRINK_OGTT / DRINK_FAT, which describe the meal rather than the subject. Specific scope until a second gastric emptying model ratifies the name.

D25OH_BL (canonical for baseline (pre-supplementation) serum 25-hydroxyvitamin D concentration)

• Description: Per-subject serum concentration of 25-hydroxyvitamin D (25(OH)D, calcifediol) measured before the first vitamin D dose. 25(OH)D is the circulating storage metabolite of vitamin D and the analyte on which vitamin D status is clinically defined (deficiency below 50 nmol/L, severe deficiency below 25 nmol/L per the Institute of Medicine); every subject carries a substantial endogenous pool sustained by cutaneous synthesis and dietary intake, so a supplemental dose is superimposed on a large pre-existing baseline rather than starting from zero. In vitamin D PK models the baseline therefore plays a dual role: it is the initial condition of the 25(OH)D compartment(s), and it identifies the otherwise-unobservable per-subject endogenous vitamin D input rate, which is back-solved so that formation and elimination balance exactly at the measured baseline. Because the elimination of 25(OH)D is concentration-dependent, the baseline also sets where on the nonlinear clearance curve each subject starts and hence how much a given dose raises them.
• Units: nmol/L
• Type: continuous
• Scope: specific
• Reference category: n/a – not a centred covariate effect. Used directly as an initial condition and as the argument of the endogenous-input back-calculation, so no reference or normalisation value applies. Papers reporting ng/mL must be converted (1 ng/mL = 2.496 nmol/L for 25(OH)D3, molar mass 400.64 g/mol); record the conversion in covariateData[[D25OH_BL]]$notes.
• Source aliases:
  • D25BASE – used in Muhamad_2025_cholecalciferol_pbpk.R (Muhamad 2025 Figures S2/S3 equation symbol) and Huang_2021_cholecalciferol_pbpk.R (Huang 2021 Methods and supplement Final Model section); same orientation and units, no transformation.
• Example models: Muhamad_2025_cholecalciferol_pbpk.R (initial condition of all five 25(OH)D3 compartments, each scaled by that compartment’s volume and partition coefficient, and the argument of the per-child endogenous input ENDOG = CLmax * D25BASE^gamma / (C50^gamma + D25BASE^gamma) * D25BASE / Fm), Huang_2021_cholecalciferol_pbpk.R (the healthy-adult model Muhamad 2025 descends from: the same two roles, initial condition of all four 25(OH)D3 compartments and argument of the arm-level ENDOG, which also sets the pre-dose vitamin D3 steady state).
• Notes: Member of the <ANALYTE>_BL baseline-concentration family (GLN_BL, HGB_BL, FERRITIN_BL, INS_BL, PKK_BL, TRACP5B_BL, TRAST_BL, AT_BL_UDL, PLAQUE_BL). The closest structural precedents are GLN_BL and AT_BL_UDL, which likewise carry the endogenous baseline of the very substance being dosed and likewise anchor the modelled scale. Distinct from the dosed amount, which belongs in the DOSE_<drug>_<units> family. The canonical names the measured metabolite 25(OH)D, not parent vitamin D (cholecalciferol / ergocalciferol) and not the active hormone 1,25-dihydroxyvitamin D; a model needing a baseline for either of those should ratify a separate name rather than overload this one. Time-fixed per subject in Muhamad 2025, which sampled once at baseline, and per study arm in Huang 2021, which fitted arm means; note that seasonal variation in 25(OH)D is real (Muhamad 2025 Section 4.3 infers an 11.5 nmol/L sinusoidal amplitude in the Cape Town control arm), so a study that resamples the baseline across seasons should carry it as time-varying and say so. Specific scope, although two vitamin D models now share the name.

GLN_BL (canonical for baseline (pre-dose) endogenous plasma L-glutamine concentration)

• Description: Per-subject pre-dose plasma concentration of endogenous L-glutamine, measured before the first study dose. L-glutamine is the most abundant free amino acid in plasma and is present at several hundred umol/L in every subject, so an exogenous oral dose is superimposed on a large endogenous pool. In popPK models of administered L-glutamine the baseline plays a dual role: it is subtracted from every measured concentration to isolate the exogenous increment being modelled (so it defines the zero of the modelled concentration scale), and it enters the clearance covariate equation, because a higher endogenous pool competes with exogenous glutamine for metabolism and other elimination pathways.
• Units: umol/L
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as a power (Eq. 5) term normalised to a standard baseline, (GLN_BL / GLN_BL_standard)^theta. Sadaf 2024 uses GLN_BL_standard = 683 umol/L. Record the per-model normalisation constant in covariateData[[GLN_BL]]$notes, since it is a modelling choice rather than a physiological constant.
• Source aliases:
  • Glu_BSL – used in Sadaf_2024_glutamine.R (Sadaf 2024 Table 2; the paired standard value is written Glu_BSLstardard in the table footnote, a typographical variant of “standard”). Note the Glu stem here abbreviates glutamine, not glutamate: Sadaf 2024 measured both analytes and screened baseline L-glutamate separately (it was not retained), so a future extraction of that paper’s sibling analyte must not reuse this alias.
• Example models: Sadaf_2024_glutamine.R (power effect on apparent oral clearance: cl = 78.5 * (WT/70)^0.75 * (GLN_BL/683)^-0.96, Sadaf 2024 Table 2 theta2 = -0.96; the same per-subject value is the quantity subtracted from each measurement before fitting).
• Notes: Specific scope until a second L-glutamine or amino-acid supplementation model ratifies the canonical. Member of the <ANALYTE>_BL baseline-concentration family (HGB_BL, FERRITIN_BL, INS_BL, PKK_BL, TRACP5B_BL, TRAST_BL); the closest structural precedent is AT_BL_UDL, which likewise carries the endogenous baseline of the very substance being dosed and likewise serves both as a power covariate normalised to a reference and as the endogenous anchor of the observation scale. Distinct from AA_* general amino-acid panel entries (none registered at the time of writing) and from the dosed amount, which is DOSE_GLN_GKG. Time-fixed per subject in Sadaf 2024, which verified that the pre-dose baseline did not drift across the four study visits; a study in which the endogenous pool does shift over time should carry it as time-varying and say so in the model’s covariateData notes.

Pregnancy / hormonal status

EGA (canonical for maternal estimated gestational age during pregnancy)

• Description: The mother’s estimated gestational age at the time of an observation or dose, in weeks. Time-varying across a pregnancy (though usually effectively constant within a single treatment cycle). EGA = 0 is the non-pregnant anchor: semi-physiological pregnancy models are constructed so that every gestational relation collapses to its non-pregnant value at EGA = 0, which makes the covariate a continuous pregnant-vs-non-pregnant contrast rather than a within-pregnancy-only stratifier.
• Units: weeks
• Type: continuous
• Scope: general
• Reference category: n/a – EGA = 0 is the non-pregnant reference value, not a category. Enters as the argument of the gestational physiology polynomials (serum albumin, alpha-1-acid glycoprotein, GFR, haematocrit, CYP3A4 activity, plasma volume, extracellular water, total body water), each of which is written so that it returns the non-pregnant value at EGA = 0.
• Source aliases:
  • EGA – Janssen 2023 symbol for estimated gestational age; same orientation and units, no transformation.
  • GAWK, GESTAGE, GAWEEK – common NONMEM column spellings for maternal gestational age in weeks; same orientation, no transformation.
• Example models: Janssen_2023_docetaxel.R, Janssen_2023_paclitaxel.R, Janssen_2023_doxorubicin.R, Janssen_2023_epirubicin.R (semi-physiological enriched cytotoxic models in which EGA drives all ten gestational physiology relations reprinted as Janssen 2023 Eqs 1-17).
• Notes: Distinct from GA (gestational age at birth), which is a time-fixed maturity descriptor of the neonate used in paediatric models – EGA describes the mother and varies over her pregnancy. Also distinct from PREG (binary pregnancy status) and from TPP (time postpartum, which is 0 during pregnancy and increases after delivery). PREG gives a step contrast; EGA gives the continuous within-pregnancy trajectory that PREG explicitly defers to (see the PREG entry’s note that “trimester or gestational-age stratification within the pregnant cohort should use a separate canonical… ratified separately when needed”). When a source paper reports trimesters rather than weeks, record the week value used per trimester in covariateData[[EGA]]$notes rather than introducing a trimester-indicator canonical.

TPP (canonical for time postpartum (time after delivery))

• Description: Time elapsed since delivery, in weeks. 0 during pregnancy and at delivery; increases after delivery. Used by popPK models that describe the time-varying postpartum recovery of pregnancy-induced PK changes (e.g., renal blood flow, hepatic enzyme activity) that return gradually to the prepregnant baseline over weeks to months rather than instantaneously at delivery.
• Units: weeks
• Type: continuous
• Scope: general
• Reference category: n/a – enters as the argument of a sigmoidal recovery function (TPP^gamma) / (TPP^gamma + T50^gamma) whose T50 (weeks-to-50%-recovery) and gamma (shape parameter) are estimated. For pregnant visits (TPP = 0) the sigmoid evaluates to 0 and the pregnancy-state PK parameter is unchanged; for far-postpartum visits (TPP >> T50) the sigmoid approaches 1 and the parameter reaches its asymptotic non-pregnant value.
• Source aliases:
  • TPP – de Kock 2017 NONMEM column for time after delivery in weeks; same orientation, no transformation; assigned 0 for samples collected during pregnancy.
• Example models: deKock_2017_sulfadoxinePyrimethamine.R (sigmoidal effect on sulfadoxine CL with asymptotic fractional change -0.757, T50 = 6.35 weeks, gamma = 4.90; the sigmoid approaches its asymptote ~13 weeks postpartum, consistent with the literature for return of GFR and renal blood flow to prepregnant values within 6-12 weeks postpartum).
• Notes: Paired with PREG (pregnancy status indicator) when the source paper models pregnancy as a step contrast on one PK parameter and as a sigmoidal time-decay on another. During pregnancy TPP = 0; after delivery TPP > 0. Document the postpartum sampling window in covariateData[[TPP]]$notes per model. Distinct from T_NUT_SUPP (time on nutritional supplementation, days) and GA (gestational age at birth, weeks) – those are different timescale covariates anchored to different events.

TERM_BIRTH (canonical for term-vs-preterm birth indicator)

• Description: Binary indicator of term-vs-preterm birth status; 1 = term birth, 0 = preterm birth. Time-fixed per subject. The gestational-age cutoff separating the two levels is the source paper’s own: 37 weeks is the conventional obstetric definition (Allegaert 2015), but papers do use other cutoffs (Knebel 2011 defines preterm as gestational age < 38 weeks) and the cutoff actually used must be recorded per-model in covariateData[[TERM_BIRTH]]$notes. In Allegaert 2015 (paracetamol PK in young women) the indicator selects between two typical-value clearances for the sulphate-formation pathway; in Knebel 2011 it selects between two values of the maturation half-time AG50.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: varies by effect form; document per-model. Allegaert 2015 uses the symmetric CL = TERM_BIRTH * theta_term + (1 - TERM_BIRTH) * theta_preterm selection – so neither category is “the multiplicative reference”, and both per-stratum clearances are estimated parameters. Knebel 2011 instead makes term birth the reference and multiplies the maturation half-time by an estimated factor raised to (1 - TERM_BIRTH), so 0 (preterm) is the effect-carrying level.
• Source aliases:
  • TERM (Allegaert 2015 NONMEM column; same orientation, no transformation) – used in Allegaert_2015_paracetamol.R.
  • a preterm indicator (Knebel 2011: 1 = gestational age < 38 weeks) – orientation is INVERTED relative to the canonical, so TERM_BIRTH = 1 - PRETERM and the published coefficient applies to (1 - TERM_BIRTH). Used in Knebel_2011_pantoprazole.R.
• Example models: Allegaert_2015_paracetamol.R, Knebel_2011_pantoprazole.R (multiplicative 1.38-fold inflation of the sigmoid-Emax maturation half-time AG50 for preterm infants: mat_age50 * 1.38^(1 - TERM_BIRTH), taking AG50 from 0.153 to 0.211 years; 77 of 202 patients preterm).
• Notes: Distinct from GA (continuous gestational age in weeks): TERM_BIRTH is the binarized version. Use GA when the source paper carries gestational age as a continuous covariate; use TERM_BIRTH only when the paper itself dichotomizes. Do not derive TERM_BIRTH from GA programmatically inside model – the cutoff convention belongs in data assembly, not the model file, and papers disagree on the cutoff (37 vs 38 weeks). Scope promoted from specific to general on 2026-08-26 alongside the Knebel 2011 pantoprazole extraction, a second and independent paper dichotomizing gestational age at birth into term vs preterm.

CONMED_BIRTHCONTROL (canonical for oral-contraceptive use indicator)

• Description: Binary indicator of oral hormonal contraceptive use; 1 = currently taking an oral contraceptive (estrogen-progestin or progestin-only pill), 0 = not on hormonal contraception. Time-varying as women cycle on/off contraception across study occasions.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no oral contraceptive). Effect form in Allegaert 2015 is multiplicative: CL_glucuronide *= theta_CONMED_BIRTHCONTROL when CONMED_BIRTHCONTROL == 1, with theta_CONMED_BIRTHCONTROL = 1.46 (estrogen-driven UGT2B7 induction).
• Source aliases:
  • BC (Allegaert 2015 NONMEM column; same orientation, no transformation) – used in Allegaert_2015_paracetamol.R.
  • OC (Csajka 2005 NONMEM column; same orientation, no transformation) – used in Csajka_2005_ephedrine_caffeine.R.
• Example models: Allegaert_2015_paracetamol.R, Csajka_2005_ephedrine_caffeine.R, Angeli_2016_iron_hepcidin.R (multiplicative effect on iron elimination kout_iron; women on oral contraception have ~18% lower iron elimination, consistent with the paper’s narrative that contraception limits menstrual blood loss; e_conmed_birthcontrol_kout_iron = -0.20).
• Notes: Specific scope because the canonical encoding pools all oral contraceptive types (combined / progestin-only) into a single binary; future models that need to distinguish formulations should register a finer-grained canonical (e.g., CONMED_BIRTHCONTROL_COMBINED, CONMED_BIRTHCONTROL_PROGESTIN). The full-word canonical name was chosen over a shorter BC_USE form for clarity in source traces.

CONMED_TICLOPIDINE (canonical for ticlopidine co-medication / study-phase indicator)

• Description: Binary indicator that the record belongs to a ticlopidine pre-dosing phase; 1 = ticlopidine phase, 0 = placebo phase. TIME-VARYING in principle (a crossover subject contributes records in both phases), time-fixed within one phase of one study period.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (placebo phase). In Ashraf_2018_s_ketamine_ticlopidine.R the column selects between the two phase-specific proportional residual SDs for S-ketamine and for norketamine, sd = propSd * (1 - CONMED_TICLOPIDINE) + propSd_ticlo * CONMED_TICLOPIDINE; it carries no structural effect.
• Source aliases:
  • TICLO – the $INPUT data item of the Ashraf 2018 deposited NONMEM control stream (Supplementary Information S3), where it both keys $ERROR on to EPS(1)/EPS(2) versus EPS(3)/EPS(4)/EPS(5) and gates the $DES interaction block with IF (TICLO.EQ.1).
• Example models: Ashraf_2018_s_ketamine_ticlopidine.R (founding example; doi:10.1002/psp4.12346).
• Notes: Member of the CONMED_<INN> family, auto-approved as a well-formed member during the Ashraf 2018 extraction (PMC6202471). Deliberately distinct from DOSE_TICLOPIDINE_MG, which carries the administered ticlopidine dose magnitude that drives the interaction term: the indicator marks the study phase (and stays 1 through the washout that follows the last ticlopidine dose), whereas the dose column falls to 0 as soon as dosing stops. A consumer who uses the indicator to gate the interaction will never see the enzyme-activity recovery the model predicts. The structural gate in the source control stream is redundant – with no ticlopidine in the system the mechanistic-static interaction term evaluates to exactly 1 – so the founding model uses this column only for the residual-error switch.

CONMED_CYPROTERONE (canonical for cyproterone acetate co-medication indicator)

• Description: Binary indicator of concomitant cyproterone acetate (CPA), the steroidal antiandrogen given as flare prophylaxis around the start of GnRH-agonist therapy in prostate cancer; 1 = the CPA effect is in force at this record’s time, 0 = not. TIME-VARYING: the column marks the window over which the model assumes CPA is suppressing endogenous GnRH activity, not merely whether the subject ever received CPA, so it opens before the first GnRH-agonist dose and closes when the assumed effect duration elapses.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no cyproterone acetate effect in force). Effect form in Snelder 2019 is a divisor on the endogenous-agonist activity ratio: agonist_kd_ratio / (1 + cpa * CONMED_CYPROTERONE) with cpa = 4.13, i.e. an 80.5% reduction in endogenous GnRH activity and a consequent 67.3% reduction in typical testosterone.
• Source aliases:
  • ICPA / CPA – Snelder 2019 Supplement 4 NONMEM control stream, where ICPA is the per-subject CPA flag and the $DES block gates CPA = CPA0 to a 336 h window opening 168 h before the leuprorelin dose.
• Example models: Snelder_2019_leuprorelin.R (founding example; gates the constant CPA suppression of endogenous GnRH activity in the modified Romero testosterone model).
• Notes: Member of the CONMED_<INN> family. General scope because CPA (and the equivalent nonsteroidal antiandrogens used the same way) is standard flare prophylaxis across the GnRH-agonist literature that this model family draws on, so the same column will recur in triptorelin, goserelin and leuprorelin models. The window is supplied as data rather than derived inside model() because its start and duration are protocol facts, not fitted parameters: EC403 gave CPA 7 to 3 days before the first leuprorelin dose (Methods 2.1) and the analysis assumed a two-week effect (Methods 2.5). Record the window actually used in covariateData[[CONMED_CYPROTERONE]]$notes. Distinct from a DOSE_CPA_MG style column, which would carry the administered magnitude; register that separately if a future model estimates a dose-dependent rather than constant CPA effect. Auto-approved as a well-formed CONMED_<INN> member during the Snelder 2019 leuprorelin extraction (PMC6533438).

DIS_EOPE (canonical for early-onset pre-eclampsia indicator)

• Description: Binary indicator of early-onset pre-eclampsia (eoPE); 1 = eoPE diagnosed before 34 weeks gestation, 0 = not eoPE. Time-fixed per subject within the gestational PK study window. Used by population PK models that compare drug disposition in pregnant women with vs without early-onset pre-eclampsia.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no eoPE). Effect form in Schoenmakers 2025 is multiplicative on CL: CL_eoPE = CL * ThetaPE^DIS_EOPE with ThetaPE = 0.617 (38% reduction in betamethasone CL when eoPE is present); encoded in nlmixr2 as the log-additive shift cl <- exp(lcl + etalcl + e_eope_cl * DIS_EOPE) with e_eope_cl = log(0.617).
• Source aliases:
  • PE – common abbreviation in obstetric pharmacology papers when the cohort restriction is to early-onset PE only; used in Schoenmakers_2025_betamethasone.R (paper notation: eoPE / ThetaPE).
• Example models: Schoenmakers_2025_betamethasone.R (multiplicative effect on CL, encoded via the log-additive form on lcl; reduces apparent betamethasone clearance from 15.6 L/h to 9.6 L/h).
• Notes: Distinct from a broader PREECL indicator that would pool early-onset, late-onset and postpartum pre-eclampsia. The “early-onset” specifier corresponds to diagnosis before 34 weeks gestation, the conventional clinical cutoff (Phipps 2019 Nat Rev Nephrol). Future papers that enrol mixed early-/late-onset cohorts or that report PE status without the 34-week stratification should register a separate canonical (e.g., PREECL for any-onset PE, or LOPE for late-onset PE) rather than reusing DIS_EOPE with relaxed semantics. Distinct from PREG (pregnancy status indicator): DIS_EOPE is a complication-of-pregnancy stratifier within a pregnant cohort, whereas PREG discriminates pregnant-vs-non-pregnant subjects.

SCORE_HIGHAM (canonical for Higham pictorial blood-loss assessment chart score)

• Description: Higham’s score (Pictorial Blood-loss Assessment Chart, PBAC), a semi-quantitative pictorial scoring of menstrual blood loss summed across all sanitary protection items used during a menstrual period. Higher scores correspond to heavier menstrual bleeding; a per-menstrual-period score >= 100 is the conventional clinical threshold for menorrhagia (Higham 1990 BJOG). Time-fixed per subject when computed as the mean of the last few cycles preceding the study (the Angeli 2016 use), time-varying when scored per cycle.
• Units: (score; unitless)
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as a power-law multiplier (SCORE_HIGHAM / tHIGHAM)^exponent on hepcidin synthesis and elimination rate constants in Angeli 2016, with the reference value tHIGHAM set near the population median (96.6 in the Angeli 2016 HEPMEN cohort).
• Source aliases:
  • HIGHAM – prior canonical name (pre-2026-06-19 standardization audit).
  • HIGHAM – prior canonical name (pre-2026-06-19 SCORE_ family standardization).
  • HiS – Angeli 2016 paper notation; same orientation, no transformation.
• Example models: Angeli_2016_iron_hepcidin.R (per-subject mean of the three most recent cycles’ Higham scores; enters ksynH and koutH via power-law effects with exponents 0.66 and 0.83 respectively).
• Notes: Specific scope because the Higham / PBAC score is a reproductive-health-specific instrument. Promote to general if a second menstrual-cycle or iron-status paper registers the same instrument. Distinct from generic blood-loss quantities reported in mL (which would warrant a separate MENSTRUAL_BLOOD_LOSS_ML canonical). The threshold for menorrhagia is 100 (Higham 1990 BJOG 97:734); the Angeli 2016 cohort had a mean of 96.6 with SD 60.5 (Table I), so most subjects sat near the menorrhagia threshold.

DLOSS (canonical for menstrual-period duration)

• Description: Per-subject length of the menstrual period (days of menstrual bleeding) used as a time bound on the increased-elimination phase in joint iron / hepcidin turnover models. Time-fixed per subject within a single menstrual cycle (the Angeli 2016 use); set to the individual’s observed menses length over the study cycle rather than estimated from the data.
• Units: days
• Type: continuous
• Scope: specific
• Reference category: n/a – enters the model as the upper bound of the loss-phase indicator window t < DLOSS. Reference values observed: Angeli 2016 inclusion criterion required menses length between 3 and 5 days (Methods p. 491); the per-subject value was fixed to the observed individual menses length.
• Source aliases:
  • dloss – Angeli 2016 paper notation; same orientation, no transformation.
• Example models: Angeli_2016_iron_hepcidin.R (per-subject menses length in days; bounds the loss phase during which kloss adds to both kout_iron and kout_hep).
• Notes: Specific scope because the menses-length covariate is meaningful only inside joint iron / hepcidin turnover models that include a discrete loss-phase indicator. Promote to general if a second reproductive-cycle PK / PD paper uses an equivalent per-subject menses-length time-bound. Distinct from TPP (time postpartum, weeks; unbounded postpartum time after delivery) and from a hypothetical CYCLE_LENGTH covariate (full menstrual cycle length in days, which would warrant a separate canonical).

DAY_DELIVERY (canonical for day-of-delivery indicator)

• Description: Binary indicator that the PK sample was collected on the day of delivery (the labour-and-childbirth occasion). 1 = sample taken on the day of delivery; 0 = otherwise (pregnant non-delivery, or non-pregnant). In Hirt 2007 the indicator is mutually exclusive with PREG = 1: per the paper, on the day of delivery the coding was PREG = 0 and DAY_DELIVERY = 1, so the two indicators jointly partition women into non-pregnant (PREG = 0, DAY_DELIVERY = 0), pregnant non-delivery (PREG = 1, DAY_DELIVERY = 0), and delivery (PREG = 0, DAY_DELIVERY = 1) cohorts. Time-fixed per sampling occasion.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (not day of delivery).
• Source aliases:
  • DEL – Hirt 2007 NONMEM symbol for the day-of-delivery indicator; same orientation, no value transformation.
• Example models: Hirt_2007_nelfinavir.R (multiplicative effect 1 + 1.92 * DAY_DELIVERY applied to both maternal nelfinavir clearance CL_Nm_No and distribution volume V; gates a body-weight power effect (WT/73)^2.81 on CL_Nm_No within the delivery cohort only).
• Notes: Distinct from PREG (pregnant non-delivery, 1 = pregnant; see entry above), TPP (time postpartum, a continuous time-since-delivery covariate for postpartum recovery), and TERM_BIRTH (term-vs-preterm birth at any postnatal time). Use DAY_DELIVERY when the source paper carries a contrasted day-of-delivery cohort against pregnant non-delivery women in the same PK analysis (placental-transfer / labour-PK studies). Specific scope until a second model ratifies the name; promote to general when a future placental-transfer or labour-PK paper uses the same indicator.

SEX_HORMONE_BINDING_GLOBULIN (canonical for serum sex hormone-binding globulin)

• Description: Serum sex hormone-binding globulin (SHBG) concentration, the high-affinity binding protein for testosterone and estradiol (baseline or time-varying). Higher SHBG restricts the unbound sex-steroid fraction available for hepatic extraction, so it typically enters a sex-steroid clearance with a negative power exponent.
• Units: nmol/L. Document per-model via covariateData[[SEX_HORMONE_BINDING_GLOBULIN]]$units; papers reporting ug/dL or mg/L must be converted on data ingestion.
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (SEX_HORMONE_BINDING_GLOBULIN / ref)^exponent. Reference values observed: 20 nmol/L (Pastuszak 2021, hypogonadal men; Table 1 mean 20.8 nmol/L).
• Source aliases:
  • SHBG – Pastuszak 2021 Table 1 and Table 2 (‘SHBG, CL/F (power)’); baseline value, no value transformation.
• Example models: Pastuszak_2021_testosteroneUndecanoate.R (baseline-only; power exponent -0.219 on the apparent clearance of total testosterone, referenced to 20 nmol/L; effect parameter e_sex_hormone_binding_globulin_cl).
• Notes: Spelled out rather than abbreviated, per the maintainers’ policy for new canonicals. A plain analyte name like ALB, used for both baseline and time-varying values; each model’s covariateData notes say which one it uses. Distinct from the shbg turnover-STATE compartment in compartment-names.md (Jensen 2023, Reinecke 2018), which is a simulated amount, not a data column.

Vital signs

HR (canonical for heart rate)

• Description: Subject heart rate, in beats per minute. Captured in popPK studies where hemodynamic state modifies hepatic blood flow and hence clearance of high-extraction-ratio drugs (e.g., propofol). May be time-varying when serial intra-operative or intensive-monitoring values are recorded; many studies summarise as the per-subject median across the observation window and treat the covariate as time-fixed. Document baseline-vs-time-varying status in covariateData[[HR]]$notes per model.
• Units: beats/min
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (HR / ref)^exponent or linear-deviation forms (1 + e * (HR - ref)). Reference values observed: 158 beats/min (Ngamprasertwong 2016; population reference encoded in the Table 2 equation CL = theta1 * (HR/158)^theta2).
• Source aliases:
  • HR – same orientation as the canonical, no value transformation; used in Ngamprasertwong_2016_propofol_sheep.R (per-subject median HR over the propofol-infusion observation window, treated as time-fixed in line with the cohort-typical sheep hemodynamic state).
  • Baseline HR – Mukker 2026 Table S1; the per-subject pre-dose (time-zero) heart rate, treated as a time-fixed covariate entering the C-DeltaHR linear-mixed-effects intercept as a centered deviation. Same orientation, no value transformation; the baseline-versus-observation-time reading is documented per-model per the instruction in this entry’s Description.
• Example models: Ngamprasertwong_2016_propofol_sheep.R (power effect on maternal propofol clearance: CL_indiv = theta1 * (HR/158)^theta2 with theta2 = 0.764; clearance increases with heart rate, plausibly reflecting heart-rate-driven increases in hepatic blood flow that govern propofol’s high hepatic-extraction-ratio elimination), Mukker_2026_tuvusertib_HR.R (per-subject baseline HR entering a centered linear-deviation term e_hr_bl_e0 * (HR - hr_bl_ref) on the concentration-DeltaHR intercept; note that the model’s observable is named d_hr, the change from baseline, so that it does not shadow this covariate column).
• Notes: General scope because heart rate is a universally applicable vital sign suitable for any model where hemodynamic state modulates clearance. Future models can use a different reference HR (typical adult human is ~70 beats/min vs the sheep cohort 158 beats/min); document the reference in covariateData[[HR]]$notes. Distinct from HR_BAND or HRV (not yet registered) which would be a heart-rate-band stratifier or heart-rate variability metric, respectively.

SBP (canonical for systolic blood pressure)

• Description: Subject systolic blood pressure, in mmHg. Captured at baseline or serially during a study. Universal vital sign used in popPK / disease-progression models where hemodynamic state serves as either (a) a prognostic covariate on non-PK endpoints (overall survival, tumor dynamics), or (b) a covariate on clearance for drugs whose disposition is sensitive to blood pressure. Document baseline-vs-time-varying status per model in covariateData[[SBP]]$notes.
• Units: mmHg
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (SBP / ref)^exponent, linear-deviation forms (1 + e * (SBP - ref)), or additive-log forms beta * log(SBP). Reference values observed: 120 mmHg (Terranova 2022 avelumab JAVELIN Gastric 100 cohort median; also the population-adult clinical normal midpoint).
• Source aliases:
  • Log systolic blood pressure – Terranova 2022 Table S2 OS TTE coefficient label; enters as log(SBP) * beta (an additive-log form on log-median OS).
  • SBP – source-paper column name; same orientation, no value transformation.
• Example models: Terranova_2022_TGD_OS_gastric.R (additive-log effect on log-median OS: log_median_OS += 0.532 * log(SBP), screened by ML and retained in the parametric TTE model despite the credible interval including zero, in keeping with the paper’s hypothesis-generating scope).
• Notes: General scope because systolic BP is a universally applicable vital sign. Sibling of HR (heart rate, beats/min) and DBP_REL (relative change in diastolic BP from baseline, unitless). The two BP siblings are physiologically related but independent covariate columns: some source papers report SBP only, some report DBP or DBP-change only, some report both. SBP here is the raw absolute value; if a future model needs a relative-change (SBP_REL) or a diastolic sibling (DBP) it should register a new sibling canonical rather than overloading this entry.

BODYTEMP (canonical for body temperature)

• Description: Subject body temperature (typically axillary or oral) at the relevant clinical observation. Captured at study admission in acute-infection PK studies (fever as a marker of acute illness severity); may be time-varying when serial temperature measurements are recorded across visits. Document baseline-vs-time-varying status in covariateData[[BODYTEMP]]$notes per model.
• Units: degC
• Type: continuous
• Scope: general
• Reference category: n/a – used with linear-deviation forms (1 + e * (BODYTEMP - ref)) or exponential forms exp(e * (BODYTEMP - ref)). Reference values observed: 36.9 degC (Kloprogge 2013 lumefantrine; pooled-cohort median in Ugandan pregnant + non-pregnant women with uncomplicated P. falciparum malaria), 37.2 degC (Kloprogge 2014 quinine; pregnant Ugandan women with uncomplicated P. falciparum malaria, cohort median at admission).
• Source aliases:
  • TEMP – common short form in malaria / infectious-disease NONMEM control streams; used in Kloprogge_2013_lumefantrine.R and Kloprogge_2014_quinine.R (same orientation as the canonical, no value transformation).
  • Temp – rearing-water temperature in Yang_2020_enrofloxacin_rainbowTrout_pbpk.R (acslXtreme code constant Temp). For a poikilothermic animal the body temperature is the ambient water temperature, so the water-temperature column is carried as BODYTEMP with no value transformation.
• Example models: Kloprogge_2013_lumefantrine.R (linear-deviation effect on mean absorption transit time MTT: MTT_indiv = TVMTT * (1 + e_bodytemp_mtt * (BODYTEMP - 36.9)) with e_bodytemp_mtt = 0.165 per degC; mean transit time increases ~16.5% per degC over 36.0-39.8 degC, plausibly reflecting reduced gut motility / prolonged absorption in feverish malaria patients), Kloprogge_2014_quinine.R (exponential effect on elimination clearance: CL_indiv = TVCL * exp(e_bodytemp_cl * (BODYTEMP - 37.2)) with e_bodytemp_cl = -0.243 per degC, centered at the cohort median; clearance decreases ~21.6% per degC increase in admission body temperature over 36.0-38.9 degC, reflecting reduced metabolic CYP3A4 activity during acute febrile malaria), Hahn_2019_sufentanil.R (exponential effect on systemic clearance: CL = 37.8 * exp(e_bodytemp_cl * (BODYTEMP - 36.9)) with e_bodytemp_cl = +0.207 per degC, centered at the cohort median 36.9 degC; clearance INCREASES ~23.0% per degC over the 33-38.7 degC range observed in critically ill VA-ECMO patients – the OPPOSITE sign to Kloprogge 2014, and the two are not in conflict: Kloprogge’s range is febrile malaria at or above normothermia, whereas Hahn’s is dominated by therapeutic and incidental hypothermia in ECMO patients, where reduced total hepatic blood flow and slowed CYP3A4 activity both cut the clearance of a high-extraction drug), Yang_2020_enrofloxacin_rainbowTrout_pbpk.R (rainbow trout; not centred – drives cardiac output through the linear relationship CO (mL/min/kg) = 3.95 * BODYTEMP - 12.9, so every tissue blood flow rises with water temperature over the 5-17 degC calibration range; CO reaches zero at 3.27 degC, so the model cannot be used below about 3.3 degC).
• Notes: General scope because body temperature is a universally applicable vital sign; the Kloprogge 2013 reference value 36.9 degC is cohort-specific (Ugandan malaria cohort median) and future models should document their own reference in covariateData[[BODYTEMP]]$notes. Units are degrees Celsius; convert from Fahrenheit (degF) at data-assembly time, not inside model(). Distinct from BODYTEMP_FEBRILE (not yet registered) which would be a binary fever indicator if a future paper dichotomises at the conventional 37.5 / 38.0 degC threshold. For fish and other poikilotherms, whose body temperature equals the water (or ambient) temperature, record that temperature as BODYTEMP; the physiological role is the same (temperature-dependent blood flow and enzyme activity), only the measurement differs.

PACO2 (canonical for arterial blood carbon dioxide tension (PaCO2))

• Description: Arterial blood partial pressure of carbon dioxide, a routinely-measured blood-gas variable. Elevated in hypercapnia and respiratory acidosis (chronic obstructive pulmonary disease, restrictive lung disease, hypoventilation, severe asthma exacerbation). Used in population-PK analyses of theophylline and other drugs whose hepatic metabolism is sensitive to acid-base status / pulmonary disease severity. Typically captured at study admission via an automatic blood-gas analyzer on an arterial puncture sample; document baseline-vs-time-varying status in covariateData[[PACO2]]$notes per model.
• Units: mmHg
• Type: continuous
• Scope: general
• Reference category: n/a – used as a continuous covariate in exponential / log-linear effect forms exp(e_paco2_param * PACO2) or centered exp(e_paco2_param * (PACO2 - ref)). Reference values observed: 42.5 mmHg (Yano 1993 stable chronic airway obstruction cohort mean; also the laboratory normal-range midpoint of 35-45 mmHg in adults).
• Source aliases:
  • PaCO2 – the standard physiological abbreviation; used directly in Yano 1993 Table I and Eqs 7-8.
• Example models: Yano_1993_theophylline.R (log-linear effect on both CL and Vd of theophylline: CL = exp(lcl + e_hepimp_cl * HEPIMP + e_paco2_cl * PACO2 + etalcl) with e_paco2_cl = 0.0233 per mmHg, and Vd = exp(lvc + e_paco2_vc * PACO2 + etalvc) with e_paco2_vc = 0.00934 per mmHg; both estimated by forward-selection likelihood-ratio testing).
• Notes: Companion to arterial PaO2, arterial blood pH, serum bicarbonate, and base excess in respiratory-physiology covariate panels (not yet registered; future models retaining these as final covariates should propose them as new canonicals at extraction time). Yano 1993 found PaCO2 to be the most informative blood-gas covariate when entered in a forward-selection competition against PaO2, pH, hematocrit, and albumin. General scope because PaCO2 is a routinely-measured laboratory variable suitable for any model where pulmonary disease severity or acid-base status modulates drug disposition.

PEEP (canonical for positive end expiratory pressure)

• Description: Positive end expiratory pressure (PEEP) applied at the end of the expiratory phase during positive-pressure mechanical ventilation. Airway pressure setting maintained above atmospheric to prevent alveolar collapse; range 0-20 mmHg (0-25 cmH2O) in adult ICU cohorts. Time-varying within an ICU stay as ventilator settings are adjusted; source papers may use it as an admission value or a time-varying covariate – document per-model in covariateData[[PEEP]]$notes. Higher PEEP raises intrathoracic pressure, decreases venous return and cardiac output, and can compromise hepatic blood flow (total hepatic blood flow decreases 4%, 12%, 32% at PEEP 10, 15, 20 mmHg respectively).
• Units: mmHg (the unit reported in Swart 2004 Table 3). Some ICU literature uses cmH2O (1 cmH2O = 0.736 mmHg); when the source paper uses cmH2O the convention should be documented in covariateData[[PEEP]]$notes and values converted to mmHg for consistency across models.
• Type: continuous
• Scope: specific
• Reference category: n/a – typically used with centred linear scaling. Reference values observed: 5 mmHg (Swart 2004 lorazepam CL formula centring value; representative of routine “protective” PEEP in mechanically-ventilated ICU adults).
• Source aliases:
  • PEEP – printed column label in Swart 2004 Table 3 and formula label in Table 4; used in Swart_2004_lorazepam.R.
• Example models: Swart_2004_lorazepam.R (centred linear effect on CL for non-alcohol-abuse subjects with reference 5: cl = 4.13 - (PEEP - 5) * 0.417 L/h; each 1 mmHg above 5 lowers CL by 0.417 L/h, capturing decreased hepatic blood flow at higher PEEP; alcohol-abuse subjects instead take a flat CL = 0.74 L/h with no PEEP effect).
• Notes: Specific scope because PEEP is meaningful only for mechanically-ventilated cohorts. Distinct from MECH_VENT (binary invasive-mechanical-ventilation status) which is a treatment-status flag, not a quantitative pressure – MECH_VENT cannot substitute for PEEP because every subject in a PEEP-covariate model already has MECH_VENT = 1. The Swart 2004 rationale (Discussion) links PEEP to hepatic blood flow, so PEEP as a covariate is expected to matter most for low-hepatic-extraction-ratio drugs (lorazepam ER ~ 0.06). Future ICU popPK papers reusing PEEP with the same quantitative encoding can extend Example models; scope may be promoted to general after multiple papers. Per-model covariateData[[PEEP]]$notes must document whether the value is admission PEEP, time-varying PEEP, or a cohort summary.

STO2 (canonical for tissue oxygen saturation)

• Description: Regional tissue (as opposed to arterial or venous) oxygen saturation, the fraction of haemoglobin in the microvasculature of a named tissue bed that is oxygen-saturated, expressed as a percent. Measured non-invasively by near-infrared spectroscopy (NIRS), which is the instrument; STO2 names the quantity, on the same principle that a drug concentration is not named after the assay that measures it. Used in trauma / critical-care popPK analyses as a marker of peripheral perfusion and shock severity: a falling tissue oxygen saturation signals inadequate microvascular oxygen delivery even while arterial saturation and blood pressure remain near normal. Document the tissue bed (skeletal muscle, cerebral, somatic / renal) and whether the value is an admission reading, a nadir, or time-varying in each model’s covariateData[[STO2]]$notes – these are materially different quantities and are not interchangeable.
• Units: % (percent of haemoglobin saturated; 0-100)
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (STO2 / ref)^exponent or linear-deviation forms. Reference values observed: 88 % (Stitt 2026; median admission skeletal-muscle NIRS StO2 in the TAMPITI adult severe-trauma cohort).
• Source aliases:
  • NIRS – the instrument abbreviation, used as the column label in Stitt 2026 Table 1 / Table 2 and Equation 1. Same orientation, no value transformation.
  • SmO2 – skeletal-muscle oxygen saturation, the sports-physiology and trauma-monitoring spelling.
  • NIRS StO2, StO2, rSO2 – punctuation / capitalisation variants seen in publication tables and device output.
• Example models: Stitt_2026_tranexamicAcid.R (median admission skeletal-muscle NIRS StO2, reference 88 %; exponent -0.29 on TXA clearance, so lower tissue oxygen saturation predicts faster clearance).
• Notes: The canonical names the measured quantity rather than the measurement modality, consistent with the analyte-naming style of IL6, CRP and HGB; NIRS is retained as a source alias so paper tables and NONMEM control streams using the instrument abbreviation resolve without value transformation. This opens a clean family alongside future SPO2 (peripheral pulse-oximetry saturation) and SAO2 (arterial blood-gas saturation) entries, which are different quantities measured at different sites and must be registered separately rather than aliased onto this one. General scope because tissue oxygen saturation is a routine critical-care measurement expected to recur across trauma, sepsis, cardiac-surgery and ECMO popPK models. Tissue-bed and summary-statistic mismatches are a real hazard, not a theoretical one: Stitt 2026 pairs an adult reference of 88 % (median value at admission) with a paediatric virtual-cohort range of 51-80 % taken from the literature as the median lowest recorded value, and the paper itself flags (Discussion, S262) that the two are not the same quantity and that the mismatch inflates the extrapolated paediatric clearance. Any model reusing this column must record which of the three – admission, nadir, or time-varying – its values are.

DBP_REL (canonical for relative change in diastolic blood pressure from baseline)

• Description: Relative (fractional) change in diastolic blood pressure from the per-subject baseline, unitless (e.g., 0.10 = +10% above baseline). Typically a time-varying driver simulated from an upstream diastolic-blood-pressure indirect-response model as DBP_REL(t) = (dbp(t) - dbp0) / dbp0, then consumed as a covariate by a downstream model. Distinct from a raw diastolic-BP value (mmHg); this is the normalized deviation.
• Units: fraction (unitless; positive when dBP is elevated above baseline)
• Type: continuous
• Scope: specific
• Reference category: n/a – enters a log-linear hazard term exp(e_dbprel_haz * DBP_REL).
• Source aliases:
  • DBPREL – Hansson 2013 e85 NONMEM column for the relative dBP change; same orientation, no value transformation.
• Example models: Hansson_2013_sunitinib_os.R (time-varying relative dBP change, simulated from Hansson_2013_sunitinib_dbp and entered as an overall-survival hazard modulator: a greater relative dBP increase lowers the hazard).
• Notes: Specific scope because the value semantics are tied to the upstream Hansson 2013 indirect-response dBP model’s relative-change definition; a future model needing the same relative-dBP-change driver should add itself to the Example models list (and consider promotion) rather than reuse the name with different semantics. Structurally analogous to RCFB1MAX (relative change from baseline in SUVmax driving the Schindler 2016 OS hazard). The related raw-state PD output is the dbp compartment in compartment-names.md; the placebo-arm indicator paired with that model family is PLACEBO.

QTC_BL (canonical for subject’s heart-rate-corrected QT interval at pre-dose baseline)

• Description: Subject’s pre-dose baseline QT interval corrected for heart rate, treated as a per-subject time-fixed covariate. Used in concentration-DeltaQTc PD models to capture each subject’s intrinsic QTc level entering the linear-mixed-effects intercept as a centered term e_qtc_bl_e0 * (QTC_BL - QTC_BL_ref). The QT correction method (Bazett, Fridericia, individual, study-population-specific) is model-specific and documented per-model in covariateData[[QTC_BL]]$notes; the canonical name does NOT carry the correction method, on the same precedent as HGB_BL (units / convention documented per-model) and IGG (units documented per-model). Distinct from HR (heart rate at observation, beats/min) and BODYTEMP (vital-sign covariate): QTC_BL is a derived per-subject ECG summary statistic, not a per-observation vital-sign reading.
• Units: ms (the standard ECG / QT interval unit; document the correction method per-model in covariateData[[QTC_BL]]$notes).
• Type: continuous
• Scope: general
• Reference category: n/a – enters as a centered linear deviation e_qtc_bl_e0 * (QTC_BL - QTC_BL_ref). Reference values observed: 390 ms (Darpo 2014; rounded standard reflecting the male-dominated 28 men / 11 women cohort with day-0 QTcI ranges 384-394 ms (men) / 398-412 ms (women) and similar QTcF ranges; the specific cohort median was not quoted in the paper). Future popPK-DeltaQTc models centering on a different reference (e.g., 400 ms in older / sicker cohorts where QTc is elevated) should document the per-model reference in covariateData[[QTC_BL]]$notes.
• Source aliases:
  • baseline QTcI – used in Darpo_2014_racSotalol_QTcI.R (day-0 individually-corrected QT interval, ms; values transformed via centering against the cohort-median reference inside model).
  • baseline QTcF – used in Darpo_2014_racSotalol_QTcF.R (day-0 Fridericia-corrected QT interval, ms; same covariate canonical, correction method documented per-model).
  • QTcF_i0 / Baseline QTcF – Mukker 2026 Equation 1 and Table 2; the per-subject time-zero QTcF, entering as theta3 * (QTcF_i0 - QTcF0) where QTcF0 is the cohort mean baseline. No value transformation.
• Example models: Darpo_2014_racSotalol_QTcI.R, Darpo_2014_racSotalol_QTcF.R (PD-only linear mixed-effects E-R model for DeltaQTcI / DeltaQTcF after single 160 mg oral rac-sotalol; QTC_BL is each subject’s pre-dose day-0 mean QTc by the corresponding correction method, time-fixed for the analysis), Mukker_2026_tuvusertib_QTcF.R (Garnett LME C-DeltaQTcF model in patients with advanced solid tumors; centered on the published cohort mean baseline QTcF of 422 ms rather than a rounded standard – Mukker 2026 defines QTcF0 in the Equation 1 legend as “the overall mean QTcF_i0, that is, the mean of all the baseline (time zero) QTcF values” and Table 1 reports that mean (SD) as 422 (21.0) ms). Sasaki_2022_delamanid_QTc_dm6705.R and Sasaki_2022_delamanid_QTc_parent.R (pediatric DeltaQTcB models; Bazett correction; the unpublished cohort-mean baseline centering constant is taken as 421 ms from the Figure S5 baseline QTcB-RR regression at the reported ~100 bpm mean heart rate).
• Notes: General scope because pre-dose QTc baseline is a universally applicable ECG-derived per-subject covariate in any concentration-DeltaQTc model. The QT correction method (Bazett, Fridericia, individual, study-population-specific) is model-specific and is the per-model documentation responsibility of covariateData[[QTC_BL]]$notes; do NOT register parallel canonicals like QTCI_BL or QTCF_BL for different correction methods (the unit, role, and centering pattern are identical – only the precise QT-to-QTc conversion formula differs upstream of the model). Future concentration-DeltaQTc papers using a non-zero centering reference different from the Darpo 2014 rounded standard 390 ms should document the per-model reference in covariateData[[QTC_BL]]$notes.

The _PRESURG / _BL haemodynamic anchor pair

Perioperative haemodynamic models frequently need a subject’s vital signs at two distinct pre-treatment anchors, and the distinction is load-bearing rather than cosmetic: the two anchors bracket the physiological insult the model is about to describe, and a model can carry both as covariates on the same structural parameter with opposite signs.

• _PRESURG – the last value recorded before the anaesthetic / surgical intervention, i.e. the subject’s own undisturbed physiological set-point.
• _BL – the value at the moment treatment starts (here, at hypotension diagnosis, immediately before the first antihypotensive bolus). This keeps _BL’s established at-first-dose meaning used throughout this register (QTC_BL, HGB_BL, FERRITIN_BL).

Do not collapse the two onto one column, and do not use a single _PRE suffix for both – _PRE reads as either anchor and has blurred them in draft extractions. When a paper reports only one pre-treatment value, use _BL unless it is explicitly pre-intervention.

MAP_PRESURG (canonical for pre-surgery mean arterial pressure)

• Description: Mean arterial pressure recorded before the anaesthetic / surgical intervention – the subject’s undisturbed baseline set-point. See the _PRESURG / _BL anchor-pair note above.
• Units: mmHg
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (MAP_PRESURG / ref)^exponent. Reference values observed: 100 mmHg (Dings 2026 population median; cohort mean 101 mmHg in both treatment arms).
• Source aliases:
  • MAP (pre-surgery row of Dings 2026 Table 2) – same orientation, no value transformation. The source distinguishes the two anchors by table row rather than by column name, so the row label must be carried into the column name at data-assembly time.
• Example models: Dings_2026_cafedrine_theodrenaline_ephedrine.R (power effect on the maximum attainable MAP: rmax_map = 119 * (MAP_PRESURG/100)^0.499 * (MAP_BL/64)^-0.53; a higher undisturbed set-point raises the attainable ceiling).
• Notes: General scope – a pre-intervention vital-sign anchor applies to any perioperative or procedural model. Sibling of SBP_PRESURG and HR_PRESURG; paired with MAP_BL. The corresponding model output state is map in compartment-names.md – this column is an input, not a prediction.

MAP_BL (canonical for mean arterial pressure at treatment baseline)

• Description: Mean arterial pressure at the moment treatment starts – in the founding model, at hypotension diagnosis, immediately before the first antihypotensive bolus. See the _PRESURG / _BL anchor-pair note above.
• Units: mmHg
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling and as the additive baseline anchor of an Emax equation. Reference values observed: 64 mmHg (Dings 2026 population median; cohort means 64.2 and 64.5 mmHg).
• Source aliases:
  • MAP (baseline row of Dings 2026 Table 2) – same orientation, no value transformation.
• Example models: Dings_2026_cafedrine_theodrenaline_ephedrine.R (serves two distinct roles: the additive baseline of the MAP Emax equation map = MAP_BL + (rmax_map - MAP_BL) * ..., and a power covariate on the ceiling with exponent -0.53, opposite in sign to MAP_PRESURG’s +0.499). Sang_2021_doxorubicin_rat_qsp.R, Sang_2021_doxorubicin_human_qsp.R (baseline MAP sets baseline TPR = MAP_BL / (HR * SV_BL) and scales the MAP feedback constant as (MAP_BL / 106.596)^-1.98).
• Notes: General scope. The opposite-signed pairing with MAP_PRESURG on one parameter is the clearest demonstration of why the two anchors must stay separate columns – collapsing them would cancel most of the covariate model. Sibling of SBP_BL, HR_BL, DBP_BL.

SBP_PRESURG (canonical for pre-surgery systolic blood pressure)

• Description: Systolic blood pressure recorded before the anaesthetic / surgical intervention. See the _PRESURG / _BL anchor-pair note above.
• Units: mmHg
• Type: continuous
• Scope: general
• Reference category: n/a – power scaling. Reference values observed: 139 mmHg (Dings 2026 population median; cohort means 139 and 138 mmHg).
• Source aliases:
  • SBP (pre-surgery row of Dings 2026 Table 2) – same orientation, no value transformation.
• Example models: Dings_2026_cafedrine_theodrenaline_ephedrine.R (rmax_sbp = 169 * (SBP_PRESURG/139)^0.577 * (SBP_BL/92)^-0.494).
• Notes: General scope. The absolute-value pre-intervention sibling of the already-registered SBP (which is the unqualified / time-varying systolic pressure).

SBP_BL (canonical for systolic blood pressure at treatment baseline)

• Description: Systolic blood pressure at the moment treatment starts. See the _PRESURG / _BL anchor-pair note above.
• Units: mmHg
• Type: continuous
• Scope: general
• Reference category: n/a – power scaling and additive Emax baseline anchor. Reference values observed: 92 mmHg (Dings 2026 population median; cohort means 92.3 and 93 mmHg).
• Source aliases:
  • SBP (baseline row of Dings 2026 Table 2) – same orientation, no value transformation.
• Example models: Dings_2026_cafedrine_theodrenaline_ephedrine.R (additive baseline of the SBP Emax equation and a power covariate on the ceiling, exponent -0.494).
• Notes: General scope.

HR_PRESURG (canonical for pre-surgery heart rate)

• Description: Heart rate recorded before the anaesthetic / surgical intervention. See the _PRESURG / _BL anchor-pair note above.
• Units: beats/min
• Type: continuous
• Scope: general
• Reference category: n/a – power scaling. Reference values observed: 92 beats/min (Dings 2026 population median; cohort means 92.6 and 91.7 beats/min).
• Source aliases:
  • HR (pre-surgery row of Dings 2026 Table 2) – same orientation, no value transformation.
• Example models: Dings_2026_cafedrine_theodrenaline_ephedrine.R (rmax_hr = 77.8 * (HR_PRESURG/92)^0.399 * (1 + 0.15 * TRT_EPHEDRINE)).
• Notes: General scope. The pre-intervention anchor sibling of the already-registered HR (unqualified / time-varying heart rate used as a covariate on clearance).

HR_BL (canonical for heart rate at treatment baseline)

• Description: Heart rate at the moment treatment starts. See the _PRESURG / _BL anchor-pair note above.
• Units: beats/min
• Type: continuous
• Scope: general
• Reference category: n/a – additive Emax baseline anchor. Reference values observed: 84 beats/min (Dings 2026 population median; cohort means 85.4 and 87.5 beats/min).
• Source aliases:
  • HR (baseline row of Dings 2026 Table 2) – same orientation, no value transformation.
• Example models: Dings_2026_cafedrine_theodrenaline_ephedrine.R (additive baseline of the HR Emax equation, hr = HR_BL + (rmax_hr - HR_BL) * conc/(conc + ec50_hr) + ...; no covariate effect is estimated on it). Sang_2021_doxorubicin_human_qsp.R (initial condition of the HR turnover state; 70 beats/min typical, Sang 2021 Table 1).
• Notes: General scope. Completes the _BL anchor set with MAP_BL, SBP_BL and DBP_BL. HR_BL was not in the ratified name list of, but Dings 2026 Eq. A12 requires the at-diagnosis heart-rate anchor as the additive baseline of the HR Emax term, so it is registered here as the mechanical fourth member of the same ratified family rather than as a new concept.

DBP_BL (canonical for diastolic blood pressure at treatment baseline)

• Description: Diastolic blood pressure at the moment treatment starts. See the _PRESURG / _BL anchor-pair note above.
• Units: mmHg
• Type: continuous
• Scope: general
• Reference category: n/a – power scaling. Reference values observed: 50 mmHg (Dings 2026 typical-parturient value; see notes).
• Source aliases:
  • DBP (baseline row) / DBPBL – same orientation, no value transformation.
• Example models: Dings_2026_cafedrine_theodrenaline_ephedrine.R (power effect on the K/PD apparent volume: vc = 1 * (DBP_BL/50)^-1.49; a lower baseline diastolic pressure – a more profound sympathetic block – inflates the apparent distribution volume and so blunts the effective driving concentration).
• Notes: General scope. The absolute-value counterpart of the already-registered DBP_REL (a relative change from baseline, unitless), and the diastolic member of the _BL anchor set. Dings 2026 does not tabulate baseline DBP in Table 2, so the 50 mmHg centring value is taken from the typical-parturient definition in the paper’s Section 3.4, which Section 2.3 defines as the population median covariate values; it is internally consistent with the tabulated baseline MAP and SBP via MAP = DBP + (SBP - DBP)/3 (50 + (92 - 50)/3 = 64 mmHg, matching the reported baseline MAP of 64 mmHg).

LVEDV_BL (canonical for left ventricular end-diastolic volume at treatment baseline)

• Description: Pre-dose left ventricular end-diastolic volume (echocardiographic), used as the initial condition and steady-state anchor of an LVEDV (preload) turnover state in cardiovascular-systems models.
• Units: mL
• Type: continuous
• Scope: general
• Reference category: n/a – enters as the initial condition of the edv state and in its steady-state production rate. Reference values observed: 0.385 mL (typical rat, Sang 2021 Table S5 from Snelder 2014); 113 mL (cardiovascular-healthy human) and 141 mL (LVEDV-enlarged human) in Sang 2021 Table 1.
• Source aliases:
  • LVEDVbase – regressor name in the Sang 2021 deposited Mlxtran code.
  • LVEDV0 – Sang 2021 Table 1 / Table S5 row name.
• Example models: Sang_2021_doxorubicin_rat_qsp.R, Sang_2021_doxorubicin_human_qsp.R (initial condition of the LVEDV turnover state; baseline SV = LVEDV_BL - LVESV_BL).
• Notes: General scope. Member of the _BL haemodynamic anchor set (MAP_BL, HR_BL, SBP_BL, DBP_BL). Units are mL even for rodents (a rat LVEDV is about 0.3 mL); convert microlitre echocardiography values (Sang 2021 Table S2 reports uL) by dividing by 1000. Distinct from the edv model state, which is the predicted time course.

LVESV_BL (canonical for left ventricular end-systolic volume at treatment baseline)

• Description: Pre-dose left ventricular end-systolic volume (echocardiographic). Together with LVEDV_BL it fixes the baseline stroke volume (LVEDV_BL - LVESV_BL) and baseline ejection fraction of a cardiovascular-systems model.
• Units: mL
• Type: continuous
• Scope: general
• Reference category: n/a – used to derive baseline stroke volume and to anchor an LVESV feedback term. Reference values observed: 0.085 mL (typical rat, Sang 2021 Mlxtran LVESV0); 48 mL (cardiovascular-healthy human, Sang 2021 Table 1 LVEDV0 - SV0 = 113 - 65).
• Source aliases:
  • LVESVbase – regressor name in the Sang 2021 deposited Mlxtran code.
• Example models: Sang_2021_doxorubicin_rat_qsp.R, Sang_2021_doxorubicin_human_qsp.R (baseline SV and the steady-state production rate of LVEDV through the (1 - FB_LVESV * LVESV) dissipation feedback).
• Notes: General scope; sibling of LVEDV_BL. When a source reports baseline LVEF instead, derive LVESV_BL = LVEDV_BL * (1 - LVEF/100).

DUR_MAP_BELOW_PRESURG (canonical for cumulative duration of mean arterial pressure below the pre-surgery anchor)

• Description: Cumulative minutes for which a subject’s mean arterial pressure stays below their own MAP_PRESURG value, over a defined window (in the founding model, from antihypotensive treatment until umbilical-cord clamping). A per-subject exposure-to-hypotension metric rather than an instantaneous measurement: it integrates both the depth-independent time spent hypotensive and, implicitly, the adequacy of treatment. Typically derived from a simulated or observed MAP-time profile rather than recorded directly.
• Units: min
• Type: continuous
• Scope: general
• Reference category: n/a – enters linear regressions directly with a per-minute coefficient. Reference values observed: cohort means 10.1 min (cafedrine/theodrenaline) and 10.2 min (ephedrine) in Dings 2026 Table 2.
• Source aliases:
  • TIMEMAP – Dings 2026 Section 2.3 symbol; same orientation and units, no transformation.
  • Duration MAP<pre-surgery MAP – Dings 2026 Tables A2 / A3 row label.
• Example models: Dings_2026_neonatal_acidosis.R (the single strongest predictor of both neonatal umbilical-arterial pH, -0.00177 per minute, and base excess, -0.105 mmol/L per minute).
• Notes: General scope – a duration-below-own-baseline exposure metric generalises to any endpoint driven by cumulative hypotension or hypoperfusion. Named against the _PRESURG anchor because that is the comparator that defines it; a variant measured against a different anchor should be registered as a separate canonical (e.g. DUR_MAP_BELOW_BL) rather than silently reusing this one, since the two are numerically very different. The founding paper’s convention when MAP never recovers within the window is documented in covariateData[[DUR_MAP_BELOW_PRESURG]]$notes of the founding model.

ORGVOL_SALGLAND (canonical for measured total salivary gland volume)

• Description: Measured total volume of the salivary glands. The founding source delineates the left plus right parotid glands only; a model that includes the submandibular or sublingual glands should say so in covariateData[[ORGVOL_SALGLAND]]$notes rather than registering a separate canonical.
• Units: mL
• Type: continuous
• Scope: general
• Reference category: n/a – used directly as a physiologic input that sets a compartment’s sub-volumes, its perfusion and its permeability-surface-area product; not a covariate-effect coefficient.
• Source aliases:
  • `Salivary Glands (measured volume) – Golzaryan 2025 Table S1 column header; reported in mL, no value transformation.`
• Example models: Golzaryan_2025_lu177psmaIT_pbpk.R.
• Notes: Founding member of the ORGVOL_<ORGAN> family. The salivary glands are a dose-limiting organ at risk in PSMA-targeted radioligand therapy, so their volume is measured per patient rather than scaled from body size. Cohort range 17-54 mL in five mCRPC patients.

ORGVOL_TUMOR1 (canonical for measured individually delineated tumour lesion 1 volume)

• Description: Measured volume of the first of the individually delineated tumour lesions that a model carries as its own compartment. Lesion numbering follows the source paper.
• Units: mL
• Type: continuous
• Scope: general
• Reference category: n/a – used directly as a physiologic input that sets a compartment’s sub-volumes, its perfusion and its permeability-surface-area product; not a covariate-effect coefficient.
• Source aliases:
  • `Tumor 1 (measured volume) – Golzaryan 2025 Table S1 column header; reported in mL, no value transformation.`
• Example models: Golzaryan_2025_lu177psmaIT_pbpk.R.
• Notes: Founding member of the ORGVOL_<ORGAN> family; see the family note above for the distinction from TUM_VOL. Lesion indices are positional labels from the source, not anatomical sites, so a model that names its lesions anatomically should register named members (ORGVOL_TUMOR_BONE,…) instead of numbered ones. Cohort range 0.5-4 mL in five mCRPC patients.

ORGVOL_TUMOR2 (canonical for measured individually delineated tumour lesion 2 volume)

• Description: Measured volume of the second of the individually delineated tumour lesions that a model carries as its own compartment.
• Units: mL
• Type: continuous
• Scope: general
• Reference category: n/a – used directly as a physiologic input that sets a compartment’s sub-volumes, its perfusion and its permeability-surface-area product; not a covariate-effect coefficient.
• Source aliases:
  • `Tumor 2 (measured volume) – Golzaryan 2025 Table S1 column header; reported in mL, no value transformation.`
• Example models: Golzaryan_2025_lu177psmaIT_pbpk.R.
• Notes: Founding member of the ORGVOL_<ORGAN> family. Extend the family with ORGVOL_TUMOR3 and beyond as models resolve more lesions. Cohort range 1-34 mL in five mCRPC patients.

ORGVOL_TUMORREST (canonical for measured lumped remaining-tumour-lesion volume)

• Description: Volume assigned to the single compartment that represents every tumour lesion a model does not delineate individually. Unlike the other members of this family this is an assumption rather than a measurement, because the lesions it stands for were not imaged individually; record how the source arrived at it in covariateData[[ORGVOL_TUMORREST]]$notes.
• Units: mL
• Type: continuous
• Scope: general
• Reference category: n/a – used directly as a physiologic input that sets a compartment’s sub-volumes, its perfusion and its permeability-surface-area product; not a covariate-effect coefficient.
• Source aliases:
  • `R_TU,Rest,0 / [R_TU,Rest,0] – Golzaryan 2025 Table S2 states the lumped compartment through its total binding sites at an assumed 266 nmol/L density, which is the same quantity divided by that density.`
• Example models: Golzaryan_2025_lu177psmaIT_pbpk.R.
• Notes: Founding member of the ORGVOL_<ORGAN> family. The founding source assumes either 10 mL or 50 mL per patient.

Renal / hepatic function

URINE_FLOW (canonical for instantaneous urine flow rate)

• Description: Instantaneous urine flow rate (mL/h) measured over the urine collection interval that includes the current observation. Time-varying.
• Units: mL/h
• Type: continuous
• Scope: general
• Reference category: n/a. Two registered roles, both keyed on the same URINE_FLOW == 0 sentinel for “no urine collected during the interval”:
  • Covariate-effect role – a centered-linear effect CL_renal = base + theta_URINE_FLOW * (URINE_FLOW - URINE_FLOW_ref) with URINE_FLOW_ref = 100 mL/h in Allegaert 2015. The linear-effect term is gated by URINE_FLOW > 0 and not extrapolated below the centering reference.
  • Flow-denominator role – the column is not a covariate effect at all but the physical flow that converts a renal excretion RATE into the urine CONCENTRATION observable, C_Ur = CL_renal * C_p / URINE_FLOW, and the same zero sentinel gates the renal arm off entirely in a subject with no diuresis.
• Source aliases:
  • UF (Allegaert 2015 NONMEM column; same orientation, no transformation) – used in Allegaert_2015_paracetamol.R.
  • UVol – Butragueno-Laiseca 2022 supplementary material, “the measured volume of urine excreted in each urine recovery interval”. Reported as a per-interval VOLUME (mL); divide by the interval duration on ingestion to obtain the mL/h this column carries.
• Example models: Allegaert_2015_paracetamol.R (covariate-effect role), ButraguenoLaiseca_2022_piperacillin.R (flow-denominator role; Curine <- cl_renal * Cc / (URINE_FLOW / 1000) reproduces supplementary eq. 3’s A_Ur/UVol in rate form, and (URINE_FLOW > 0) supplies the paper’s “CL_Renal … absent in patients without diuresis” gate).
• Notes: Promoted from specific to general alongside the Butragueno-Laiseca 2022 piperacillin extraction. The earlier version of this entry warned that a second model using a different EFFECT FORM should register its own canonical rather than reuse URINE_FLOW; the maintainers ruled that this second model may reuse it. That warning still stands for a genuinely different COVARIATE-EFFECT parameterization (e.g. direct URINE_FLOW / URINE_FLOW_ref proportional scaling on a clearance), which should register a sibling such as URINE_FLOW_PROP. Distinct from URINE_VOL_24H, which is a 24-hour cumulative volume used as a residual-renal-function descriptor rather than a per-interval rate; a paper that fits urine concentrations needs the per-interval rate this column carries, not the daily summary. The full-word canonical name was chosen over the bare UF source-data abbreviation for clarity in source traces.

URINE_VOL_24H (canonical for 24-hour residual diuresis (total urine volume per day))

• Description: Total urine volume collected over a 24-hour interval, in mL per 24 h. Used as an indicator of preserved native renal function in CRRT-dependent / oliguric / anuric critically ill patients, where it characterises the residual non-CRRT renal-excretion pathway. Time-fixed per study day in the founding example (a single 24-hour summary collected by nursing staff during the day of PK sampling). Distinct from URINE_FLOW (instantaneous urine flow rate in mL/h measured per collection interval).
• Units: mL/24h
• Type: continuous
• Scope: general
• Reference category: n/a – used with an additive linear effect CL = CL_base + e_URINE_VOL_24H_cl * (URINE_VOL_24H / 100) centered at 100 mL/24h (the anuria cutoff), or as a binary preserved-diuresis gate (URINE_VOL_24H > 100) multiplying a renal-clearance arm. Clinical cutoffs used in the founding example: anuria < 100 mL/24h, oliguria 100-500 mL/24h, preserved diuresis > 500 mL/24h.
• Source aliases:
  • residual diuresis (Ulldemolins 2015 prose); typical NONMEM column abbreviation DIUR. Same value orientation, no transformation.
• Example models: Ulldemolins_2015_meropenem.R (critically ill adults with septic shock and continuous renal replacement therapy; reference 100 mL/24h; additive linear effect 0.22 L/h per (URINE_VOL_24H / 100) on meropenem total CL on top of the CRRT-mediated baseline 3.68 L/h), Huppe_2023_fosfomycin.R (critically ill adults with renal insufficiency on continuous venovenous hemodialysis; used as the binary preserved-diuresis gate (URINE_VOL_24H > 100) that switches the renal clearance arm off entirely in anuric patients, per the source’s own definition of preserved diuresis as residual diuresis exceeding 100 mL/24h), Yu_2023_vancomycin.R (critically ill adults on CVVH or CVVHDF; the sole retained covariate, entering vancomycin CL as the multiplicative power term 1.90^(log10(URINE_VOL_24H + 10) / 2.3) where 2.3 is the cohort median-normalising constant log10(160 + 10) and the +10 mL offset keeps the term finite for the 46.5% of the cohort at or below 100 mL/24h; the anuric typical CL of 1.39 L/h is the CRRT-mediated clearance and a subject producing 3 L/day reaches 2.77 L/h).
• Notes: Promoted from specific to general alongside the Huppe 2023 fosfomycin extraction, the second model to retain a 24-hour urine-volume covariate with consistent semantics. The two registered effect forms are both multiplicative-or-additive in the same 100 mL/24h anuria cutoff and so share this entry: Ulldemolins 2015 uses the additive linear base + slope * (URINE_VOL_24H / 100), and Huppe 2023 uses the binary gate (URINE_VOL_24H > 100) on a renal-clearance arm. Distinct from URINE_FLOW because (a) URINE_VOL_24H is a 24-hour cumulative volume (mL/24h), not an instantaneous rate (mL/h), and (b) the additive-linear effect form base + slope * (URINE_VOL_24H / 100) is structurally different from the centered-linear-effect-with-sentinel-zero form used by URINE_FLOW. A future model that uses a proportional or multiplicative effect form on a 24-hour volume should evaluate whether the semantics still match this entry (proportional / multiplicative forms are compatible); a centered-linear-with-sentinel form (analogous to the URINE_FLOW convention) would diverge and should register a sibling canonical. Yu 2023 vancomycin exercises that multiplicative branch: it is a power model in log10(URINE_VOL_24H + offset) normalised at the cohort median rather than at the 100 mL/24h anuria cutoff, which stays within this entry’s semantics (a monotone multiplicative effect of 24-hour urine volume on a total clearance whose zero-diuresis limit is the extracorporeal clearance). Note that the offset used to keep the log finite (+10 in Yu 2023) and the normalising volume are both paper-specific and belong in covariateData[[URINE_VOL_24H]]$notes of the model file, not in this entry. The full-word URINE_VOL_24H form was chosen over the shorter URINE_24H per the maintainers’ instruction.

ANURIA (canonical for binary anuria indicator)

• Description: 1 = the subject is anuric (produces no clinically meaningful urine and has no residual renal elimination); 0 = the subject has residual urine production. Bare per-subject binary, for source papers that ascertain and tabulate anuria as a yes/no patient characteristic rather than reporting a urine volume. Its operational role is to switch a renal-clearance arm off entirely, so that an anuric subject’s total clearance reduces to the non-renal and extracorporeal / peritoneal routes.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (residual urine production preserved). Note the orientation: the covariate is 1 for the subgroup in which the renal arm is ABSENT, so a model gates its renal clearance by (1 - ANURIA) rather than by ANURIA.
• Source aliases:
  • Anuric – Patel 2015 Table 1 column header, carried per patient as Yes / No. Maps directly onto the canonical with no value transformation (Yes -> 1).
• Example models: Patel_2015_oseltamivir.R (gates the renal arm of oseltamivir carboxylate elimination in end-stage-renal-disease patients on automated peritoneal dialysis: cl_renal_oselcarb <- (1 - ANURIA) * exp(lcl_renal_oselcarb + etalcl_renal_oselcarb). Five of the nine analysed patients were anuric and four had residual urine production; the paper reports the renal arm as the dominant elimination route in the urine producers, at 0.736 L/h/70 kg with 117% between-subject variability, and its dosing simulations partition the cohort into anuric / low-urine / high-urine strata precisely because the gate changes the achievable exposure by more than fourfold).
• Notes: Deliberately registered as a bare binary rather than derived from an existing urine-volume column, because the source papers this entry exists for do not report a volume. Distinct from URINE_VOL_24H, which is a CONTINUOUS 24-hour residual diuresis in mL/24h and whose own register entry already sanctions a (URINE_VOL_24H > 100) preserved-diuresis gate – when a paper reports the volume, use URINE_VOL_24H and its threshold gate, because the continuous form carries strictly more information and its anuria cutoff is explicit; use ANURIA only when the paper ascertained the binary and nothing finer. Deriving ANURIA from URINE_VOL_24H at ingestion time would be lossy in the other direction too: a paper’s clinical adjudication of “anuric” is not always the same as its measured volume falling below 100 mL/24h. Distinct from URINE_FLOW (instantaneous mL/h) and URINE_VOL_INTERVAL (per-collection-interval volume feeding an observation equation). Distinct from RENALIMP_ESRD and DIS_ARF, which are disease-state classifications: an end-stage-renal-disease cohort can contain both anuric and urine-producing patients, which is exactly the split this column encodes, so the two are complementary rather than substitutable. Scope: specific; promote to general if a second paper carries a bare anuria binary with the same renal-arm-gating semantics. Ratified by the maintainers 2026-09-02 (PMC4386947).

URINE_VOL_INTERVAL (canonical for urine volume collected in a single urine recovery interval)

• Description: Volume of urine actually collected during the urine recovery interval that a urinary observation belongs to, in mL. Time-varying: one value per collection interval, changing whenever the interval changes. Its role is to convert a modelled cumulative urinary AMOUNT into the urinary CONCENTRATION the assay reports – Curine = A_urine / URINE_VOL_INTERVAL – so it appears in an OBSERVATION equation, not as a covariate effect on a structural parameter. A value of 0 means no urine was produced in the interval (anuria) and there is consequently no urinary concentration to observe; models must guard the division.
• Units: mL
• Type: continuous
• Scope: general
• Reference category: n/a – a per-record divisor, not a centered or normalized covariate effect. No reference value applies.
• Source aliases:
  • UVol – Butragueno-Laiseca 2024 Methods, Base population model: “CUr was obtained as AUr/UVol, where UVol is the measured volume of urine excreted in each urine recovery interval”. Same quantity in mL, no value transformation.
• Example models: ButraguenoLaiseca_2024_meropenem.R (founding example; the urinary observable of a five-matrix meropenem model in critically ill children, Curine <- urine / (URINE_VOL_INTERVAL / 1000), converting mL to L so the modelled mg amount yields mg/L. Urine was collected before and 2, 4, 6 and 8 h after the start of the infusion, so the recovery intervals are roughly 2 h; the urine state is reset to zero at each interval boundary with an evid = 5, amt = 0 replacement record placed a numerical epsilon AFTER the boundary, because rxode2 applies dose-type records before same-time observations and a reset sitting exactly on the boundary would zero the state before the end-of-interval sample is taken), Trang_2021_meropenem.R, Trang_2021_vaborbactam.R (the urinary observables of the two co-fitted plasma-and-urine models for meropenem and vaborbactam in noninfected adults, Curine <- urine / (URINE_VOL_INTERVAL / 1000); urine was collected over 0-4, 4-8, 8-12, 12-24, 24-48 and 48-72 h intervals in the phase 1 studies, so the same interval-boundary reset applies), Nayak_2021_rivipansel.R (the urinary observable of a two-compartment rivipansel model fit jointly to plasma and urine, Curine <- urine / (URINE_VOL_INTERVAL / 1000); the source NONMEM stream sets S3 = UVOL on the ADVAN3 output compartment and observes it with CMT = -3 records, which read and then reset the compartment – the same per-interval reset as the evid = 5 idiom above).
• Notes: Deliberately NOT folded into either existing urine canonical, both of which were checked before ratification. URINE_VOL_24H is fixed to a 24-hour window and is used as a renal-function DESCRIPTOR (an anuria / oliguria / preserved-diuresis marker feeding a clearance covariate), whereas this column is the collection volume of one short PK recovery interval and feeds an observation equation. URINE_FLOW is an instantaneous RATE in mL/h whose own entry explicitly warns against reuse under different effect semantics. The two are related – URINE_VOL_INTERVAL divided by the interval duration is a urine flow – but the interval duration is not itself a column, so deriving one from the other at ingestion time would be lossy. Scope promoted to general once a second paper (Trang 2021) fitted urinary concentrations from interval collections.

FLUID_IN_24H (canonical for total daily fluid intake volume)

• Description: Total volume of fluid administered to (or ingested by) a subject over a 24-hour window – enteral plus parenteral, summed across all routes. A whole-body INTAKE volume: not weight-normalized, not a rate, and not a net balance (nothing is debited for urine, insensible or drain losses). Used as a renal-haemodynamic covariate on the clearance of renally eliminated, hydrophilic drugs, where a larger fluid load raises urine output and glomerular filtration and therefore raises clearance. Most natural in neonatal / paediatric intensive care, where daily fluid intake is prescribed and charted per patient per day and is a first-class clinical variable rather than a derived one.
• Units: mL/24h
• Type: continuous
• Scope: specific
• Reference category: n/a – used algebraically. Zhao 2025 enters it as a power term normalized to the cohort MEAN, CL = 0.14 * ... * (FLUID_IN_24H / 367.18)^0.14. Future papers may center on a median or use a linear form; record the per-paper parameterisation in covariateData[[FLUID_IN_24H]]$notes.
• Source aliases:
  • DFI (daily fluid input) – used in Zhao_2025_vancomycin.R (Zhao 2025 Microbiol Spectr Table 1 “Daily fluid input (mL)”; cohort median 364 mL, IQR 252.91-473.10, mean 367.18). Same quantity in mL per 24 h, no value transformation.
• Example models: Zhao_2025_vancomycin.R (founding example; vancomycin popPK in 126 non-extremely preterm NICU neonates, where daily fluid input is one of the two covariates the paper contributes as novel. Enters clearance as (FLUID_IN_24H / 367.18)^0.14 alongside body weight, serum creatinine and a concomitant-diuretic indicator; univariate dOFV -20.894 on CL, entered at forward-inclusion step 7 and survived backward elimination).
• Notes: Specific scope until a second paper ratifies the name; at that point promote to general. Distinct from URINE_VOL_24H, which is the OUTPUT counterpart over the same 24-hour window: the two are not interconvertible (they differ by insensible losses, stool water and net fluid retention), they can move in opposite directions in a fluid-overloaded or oliguric patient, and the founding paper screened BOTH and retained only the intake – Zhao 2025 Discussion records that daily urine volume “was not a significant covariate affecting clearance”, so collapsing the two would destroy the paper’s result. Distinct from PFA (perioperative intra-operative fluid administration volume), which is a gross cumulative volume over one surgical procedure rather than a per-day intake, and from PFA_NET_RATE, which is a net hourly balance that explicitly debits urine output. Distinct from the extracorporeal circuit flows (RRT_CRRT_EFFLUENT_FLOW, Q_CVVH, DFR), which are machine settings rather than a patient’s charted intake. Deliberately named FLUID_IN_24H rather than a bare FLUID_24H so the direction of flow is unambiguous against the registered urine-output family; a future paper reporting a charted net fluid BALANCE per day (intake minus output) should register a sibling FLUID_NET_24H rather than overloading this entry, and one reporting weight-normalized intake (mL/kg/24h, the more common neonatal prescribing unit) should record the normalisation in per-model notes rather than silently rescaling this column.

CSF_DRAIN_VOL_24H (canonical for the volume of cerebrospinal fluid removed through a CSF drain over 24 hours)

• Description: Total volume of cerebrospinal fluid drained out of a patient through an external ventricular drain, a lumbar cistern drain or an equivalent CSF diversion device over a 24-hour window. A charted clinical observation, not a derived quantity: neurosurgical units record the drain output per day and titrate it against the patient’s intracranial pressure and post-operative symptoms. Its modelling role is mechanistic rather than covariate-effect: for a drug with a CSF compartment, the drained fluid carries drug out of that compartment, so the column enters model() directly as the CSF compartment’s drainage clearance (CL_csf = CSF_DRAIN_VOL_24H / 24, converted from mL/h into the model’s volume units) rather than as a coefficient on a structural parameter. This is the same “measured prescription quantity enters model() directly” role that RRT_PERIT_DIAL_FILL_VOLUME already sanctions. A patient with no drain in situ takes the value 0, which removes the drainage term entirely; the column is therefore safe to default to 0 for a non-drained cohort.
• Units: mL/24h
• Type: continuous
• Scope: general
• Reference category: n/a – used algebraically, not as a centered effect. Lu_2016_meropenem.R divides it by 24 to obtain the CSF drainage clearance in mL/h. Clinically relevant values from the founding example: 0 (no drainage), 50, 150 and 250 mL/24h are the Monte Carlo scenarios, and the paper’s dosing recommendation is conditioned on keeping drainage below 150 mL/24h.
• Source aliases:
  • CSF daily drainage vol – Lu 2016 Table 1 row label (“CSF daily drainage vol, ml”). Same quantity in mL per 24 h, no value transformation.
  • daily CSF drainage volume – Lu 2016 Results prose form.
  • QEVD – Chauzy 2022 (Antibiotics 11:1293) Eq. 6-8 and Table S3, the EVD flow in L/h measured per 0.5-1 h collection interval and supplied to the model as a time-varying covariate. Multiply by 24000 to obtain mL/24h (QEVD = CSF_DRAIN_VOL_24H / 1000 / 24).
• Example models: Lu_2016_meropenem.R (founding example; meropenem plasma + CSF popPK in 82 adults with post-neurosurgical bacterial meningitis, cohort mean 126 +/- 81 mL/24h and range 0-350. The CSF compartment’s drainage clearance is CSF_DRAIN_VOL_24H / 1000 / 24 L/h, which at the cohort mean removes roughly a third as much drug from the CSF compartment as back-transfer to plasma does, and which is the mechanism behind the paper’s headline recommendation to limit drainage); Chauzy_2022_metronidazole_pbpk.R (minimal CNS PBPK of unbound metronidazole in 8 brain-injured adults, 4 with an EVD. The drain flow removes drug from the cranial CSF compartment and displaces the physiological CSF sink flow, Qsink = max(0, 0.024 - QEVD) L/h; per-interval flows of 0-0.040 L/h, i.e. 0-960 mL/24h, are tabulated in Table S3. Time-varying within a dosing interval, unlike the founding example’s per-day value).
• Notes: Named by composing three shapes the register already uses: the CSF_<quantity> prefix from CSF_TPRO (a quantity measured in, or pertaining to, the CSF matrix), the <FLUID>_VOL_24H form from URINE_VOL_24H, and the direction-explicit naming that FLUID_IN_24H‘s entry insists on. The maintainers’ ruling (PMC5075067, 2026-09-21) chose the full-word CSF_DRAIN_VOL_24H over the shorter CSF_DRAIN_24H and over a rate-valued CSF_DRAIN_RATE in mL/h, for consistency with URINE_VOL_24H (whose own Notes record the same full-word preference) and because the source paper charts, tabulates and simulates the quantity strictly per day. Distinct from URINE_VOL_24H and FLUID_IN_24H, which are whole-body renal output and whole-body intake respectively and are renal-function / haemodynamic descriptors feeding covariate effects; this column is a matrix-specific efflux from one modelled compartment and is not interconvertible with either. Distinct from CSF_TPRO, which is a concentration measured in CSF rather than a volume removed from it. A future paper reporting an instantaneous CSF drain flow rate in mL/h rather than a per-day total should record the conversion in covariateData[[CSF_DRAIN_VOL_24H]]$notes if the semantics are otherwise identical, or register a sibling rate canonical if it is used with different effect semantics – the same boundary that URINE_FLOW and URINE_VOL_24H already draw. general scope: used by two independent papers, one with a per-day value and one with a time-varying per-interval flow converted to the 24-hour basis.

TER_MAG3 (canonical for tubular extraction rate measured by 99mTc-MAG3 renography)

• Description: Tubular extraction rate of 99mTc-mercaptoacetyltriglycine (MAG3), the whole-kidney tubular secretion clearance measured by MAG3 renography. Not BSA-normalized and not a filtration measurement: MAG3 is handled predominantly by active proximal-tubular secretion, whereas creatinine clearance and 51Cr-EDTA / iohexol clearance measure glomerular filtration. Papers that use TER as their renal-function input typically convert it to a filtration rate with a published regression before it enters the model.
• Units: mL/min
• Type: continuous
• Scope: general
• Reference category: n/a – used as a direct model input, not as a covariate-effect coefficient.
• Source aliases:
  • TER – the bare symbol used by Golzaryan 2025 Table S1 and by the upstream Kletting 2016 PLoS One Table 1 (“TER = tubular extraction rate as determined with the 99mTc Mag3 method”); same quantity in mL/min, no value transformation.
• Example models: Golzaryan_2025_lu177psmaIT_pbpk.R (Golzaryan 2025 whole-body 177Lu-PSMA I&T PBPK; per-patient TER of 136-252 mL/min is converted to the 51Cr-EDTA glomerular filtration rate as GFR = TER / 3 * 20 / 15 per Table S2, then scaled to peptide molecular size by the sieving ratio phi = 0.66 to give the filtration flow used by the kidney sub-model – founding example).
• Notes: Deliberately NOT folded into CRCL. CRCL pools creatinine-based and tracer-measured estimates of glomerular filtration, normalized to 1.73 m^2; TER_MAG3 is non-normalized tubular extraction, a physiologically distinct process measured by a different tracer, and the two are related only through a cohort-specific regression. Reusing CRCL would require an undocumented conversion at data-ingestion time and would silently mix two renal processes under one column. The tracer is named in the canonical because the number is uninterpretable without it – a tubular extraction rate measured with 131I-OIH is not numerically interchangeable with a MAG3 one.

CRCL (canonical for BSA-normalized renal function (creatinine-based estimate OR tracer-measured GFR))

• Description: BSA-normalized renal function expressed in mL/min/1.73 m^2. Accepts either (a) a creatinine-based estimate – MDRD- or CKD-EPI-estimated glomerular filtration rate, or a measured creatinine clearance that has been BSA-normalized as 1.73 x CrCl / BSA – or (b) a tracer-measured glomerular filtration rate using an exogenous filtration marker (iohexol clearance, inulin clearance, 99mTc-DTPA clearance, 51Cr-EDTA clearance), which is the clinical gold standard for measured GFR. All variants enter popPK models with the same operational role (a covariate on clearance) and the same units; the per-model covariateData[[CRCL]]$description and notes must state which assay the source paper used.
• Units: mL/min/1.73 m^2
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (CRCL / ref)^exponent or with linear centering (1 + slope * (CRCL - ref)). Reference values observed: 76 mL/min/1.73 m^2 (Hamren 2008, iohexol-clearance-measured GFR), 80 mL/min/1.73 m^2 (Cirincione 2017, MDRD eGFR), 90 mL/min/1.73 m^2 (Li 2019, calculated GFR; Bajaj 2017, CKD-EPI eGFR), 100 mL/min/1.73 m^2 (Xu 2019, measured-CrCl BSA-normalized), 141 mL/min/1.73 m^2 (Jung 2024, bedside Schwartz eGFR in critically ill children, whose renal function skews supranormal), 96 mL/min/1.73 m^2 (Shiau 2026, CKiD U25 eGFR in critically ill children with multiple organ dysfunction syndrome).
• Source aliases:
  • eGFR – estimated glomerular filtration rate; the estimating equation varies by cohort and must be documented per model. MDRD variant used in Cirincione_2017_exenatide.R and Kotani_2022_astegolimab.R; creatinine-based CKD-EPI variant in Bajaj_2017_nivolumab.R; the cystatin-C-based CKD-EPI variant in Patel_2025_eteplirsen.R, where creatinine is explicitly rejected as a renal marker because it is a by-product of the muscle breakdown that defines Duchenne muscular dystrophy (cystatin C is produced by all nucleated cells and is therefore independent of muscle mass – see CYSC); the paediatric bedside Schwartz variant 0.413 x height (cm) / Scr (mg/dL) in Jung_2024_vancomycin.R; the combined creatinine-cystatin C CKD-EPI 2021 variant in Taylor_2026_methotrexate.R, where the paper’s whole purpose was to show that the combined equation beats both single-marker CKD-EPI equations and either raw laboratory value as a predictor of clearance (dOFV 117.1 against the base model, versus 81.5 for cystatin C alone and 42.9 for creatinine alone). The same combined-equation variant, under the source column name eGFRCre-Cys, is used in Tseng_2026_piperacillin.R, which ran the same three-way comparison in the opposite population – adults with LOW body weight, where creatinine-based equations OVERestimate filtration because serum creatinine is a product of skeletal muscle metabolism – and reached the same ranking (dOFV 12.13 for the combined equation, versus 10.50 for cystatin C alone and 6.48 for Cockcroft-Gault creatinine clearance).
  • EGFR – all-caps variant.
  • CECYS – the CKD-EPI cystatin C eGFR in mL/min/1.73 m^2 (BSA-normalized; same units and orientation, no value transformation). Used in Kim_2022_piperacillin.R and Kim_2022_tazobactam.R (critically ill adults with and without ECMO) as the exponential effect exp(theta * (CRCL - 52.77)) on CL; selected over Cockcroft-Gault, MDRD, creatinine CKD-EPI, combined creatinine-cystatin C CKD-EPI and the BSA-de-normalized (mL/min) variants of each.
  • U25, GFR (CKiD U25) – the CKiD “Chronic Kidney Disease in Children Under 25” serum-creatinine eGFR equation, BSA-normalized to mL/min/1.73 m^2. An age- and sex-specific paediatric/young-adult successor to the bedside Schwartz equation (whose variant is recorded above under eGFR for Jung_2024_vancomycin.R), developed to give continuous estimates across the paediatric-to-adult transition rather than breaking at age 18. Used in Marianski_2026_ceftriaxone.R under the source column name GFR, where it enters clearance as the through-origin linear ratio (CRCL / 120). Note that the 120 reference there is an external normal-renal-function anchor, NOT the cohort median of 100 – so unlike most entries in this register the printed typical CL is not the typical patient’s CL. Same units and orientation as every other member of this canonical, so no value transformation is needed.
  • CLCRN / NCrCL (Franzese 2022) – BSA-normalized creatinine clearance, mL/min/1.73 m^2, computed with the updated bedside Schwartz equation in children and as Cockcroft-Gault creatinine clearance x 1.73 / BSA in adults, so one column spans a pediatric-to-adult cohort under two estimating equations on a single normalized scale. Used in Franzese_2022_ceftazidime.R (two-segment linear spline hinged at 100, capped at 150) and Franzese_2022_avibactam.R (power below / linear above an 80 hinge, capped at 150); in both it is read only for AGE > 2 years, a postmenstrual-age maturation function replacing it at younger ages. Note the parent adult models Li_2019_ceftazidime.R / Li_2019_avibactam.R used raw Cockcroft-Gault mL/min, so the two generations are not interchangeable on the same input.
  • CRCL_BSA – BSA-normalized creatinine clearance (measured CrCl / BSA x 1.73); used in Xu_2019_sarilumab.R.
  • GFR (Chinese-modified MDRD-type eGFR, raw mL/min) – serum-creatinine eGFR with a 0.742 female factor and a 1.233 Chinese correction, reported in mL/min and NOT BSA-normalized. Used in Ma_2020_vancomycin.R (adult kidney transplant recipients; power term (CRCL / 36.67)^0.698 on CL, where 36.67 is not the Table 1 per-patient median of 39.91 mL/min). The source’s typeset estimating equation has lost its exponent formatting, so the equation cannot be reconstructed from that paper; users supply the eGFR value directly.
  • eGFR (creatinine CKD-EPI as printed with a 144 multiplier for both sexes, raw mL/min) – Wei 2022 Eq. (4) writes 144 x (Scr/a)^b x 0.993^age (Scr in mg/dL; a = 0.7 female / 0.9 male) with no sex-specific 141 multiplier and no 1.73 m^2 term, and labels the result mL/min. Used in Wei_2022_vancomycin.R (postoperative neurosurgical adults; power term (CRCL / 115.2)^0.80 on CL, 115.2 mL/min being the cohort median). Supply the value exactly as that equation computes it.
  • 1.73*CrCl/BSA – the formula form appearing in Xu 2019 Eq. for Vm.
  • cGFR – calculated/estimated GFR, BSA-normalized; used in Li_2019_abatacept.R.
  • eGFR-LMR18rel / LMR18rel – the Lund-Malmo Revised 2018 creatinine-based estimating equation, BSA-normalized (mL/min/1.73 m^2). A Scandinavian-derived alternative to MDRD and CKD-EPI; used in Beijer_2026_cloxacillin.R. The companion LMR18abs / eGFR-LMR18abs is the SAME equation de-normalized back to absolute mL/min via a Du Bois BSA, so the two are not interchangeable – see the Notes on size normalisation below.
  • CRCL3 (CKD-EPI / bedside Schwartz eGFR de-indexed by BSA) – raw mL/min, NOT BSA-normalized. Nayak 2021 computes CKD-EPI eGFR for adults and bedside Schwartz eGFR for children younger than 18 years, then multiplies by the Du Bois BSA over 1.73 (the paper’s ‘normalized by the BSA’); Table 2 fixes the direction, since children aged 6-11 years with a Schwartz eGFR near 140 mL/min/1.73 m^2 are tabulated at 91.6 mL/min. Used in Nayak_2021_rivipansel.R (reference 119 mL/min, the control stream’s median, on the renal clearance arm; the exponent rises by 0.413 below 60 mL/min).
  • CLCR – source-paper column name; underlying assay form varies. Used in Delattre_2010_amikacin.R (raw Cockcroft-Gault, NOT BSA-normalized; median 55.5 mL/min in critically ill septic adults), in MedellinGaribay_2015_gentamicin.R (Schwartz formula CLCR = K * length / SCr with K in {0.33, 0.45, 0.55}, BSA-normalized to mL/min/1.73 m^2), and in Zhou_2019_vancomycin.R (raw Cockcroft-Gault, NOT BSA-normalized; reference 56.28 mL/min, the cohort mean, in Chinese geriatric patients with pulmonary infections – note that the reference here is a MEAN rather than the median most entries in this register use, because the source describes its normally distributed continuous covariates as mean +/- SD), and in Chauzy_2022_ceftaroline.R (creatinine clearance measured from a 24 h urine collection, raw mL/min, NOT BSA-normalized; time-varying, measured on each of two PK occasions; reference 180 mL/min, the cohort median, in ventilated ICU adults with augmented renal clearance, 83-309 mL/min), and in Sima_2022_ciprofloxacin.R (MEASURED 24-h urine creatinine clearance UCR x V / SCR, NOT BSA-normalized, reported in mL/s; the model divides the mL/min column by 60 and applies an uncentered exponential effect on the parent-to-metabolite transfer rate constant). Document the assay form per model in covariateData[[CRCL]]$description.
  • CRCL (Cockcroft-Gault on total body weight, serum creatinine in umol/L with a 0.818 divisor) – raw mL/min, NOT BSA-normalized. Jin 2022 Table 1 footnote: [(140 - Age) x weight (kg)] / [0.818 x Scr (umol/L)] x k, k = 1 for men and 0.85 for women. Used in Jin_2022_lithium.R as the median-centred power (CRCL / 116)^0.186 on lithium CL/F (cohort median 116 mL/min, range 61.7-226, in Chinese patients with bipolar disorder).
  • CrCL (Cockcroft-Gault >= 19 years, modified Schwartz < 19 years, both raw mL/min, UNCAPPED in pregnancy) – not BSA-normalized. Used in Goyal_2022_vancomycin.R (hospitalized pregnant women; reference 175 mL/min with exponent 1 on CL; cohort median 176, range 43-389 mL/min). The source deliberately does not cap the estimate at the conventional 120-150 mL/min because glomerular filtration rises in pregnancy: capping at 120 or 150 mL/min raised the OFV and biased clearance upward. A user supplying a capped value to this model will under-predict clearance in the high-CRCL tail.
  • CrCL (Cockcroft-Gault on a BMI-SELECTED body weight) – raw mL/min, NOT BSA-normalized. Used in Jin_2026_colistinSulfate.R (reference 116.3 mL/min, the cohort median, in critically ill adults with carbapenem-resistant Gram-negative pneumonia). The distinguishing feature is that the weight entering the Cockcroft-Gault numerator is chosen by body-mass-index band rather than being total body weight throughout: actual body weight if BMI < 18.5 kg/m^2, ideal body weight (IBW = Constant + 0.91 x (height_cm - 152.4), Constant 50 for men and 45.5 for women) if BMI is 18.5-24.9 kg/m^2, and adjusted body weight (ABW = IBW + 0.4 x (total body weight - IBW)) if BMI >= 25 kg/m^2. The source’s serum creatinine is in umol/L, so its equation carries the standard 0.814 divisor. Note that the typeset equation names “ideal body weight” in its numerator while the surrounding prose specifies the three-way BMI rule; the prose is operative because it is the description that defines all three cases. A model carrying this alias is NOT interchangeable with one built on a total-body-weight Cockcroft-Gault column in an obese cohort, where the ABW branch can run well below the total-weight value.
  • COGBWA (Cockcroft-Gault on body weight at ICU admission) – raw mL/min, NOT BSA-normalized. Used in Swartling_2022_cefotaxime.R (critically ill adults; centred at the cohort median 94 mL/min and truncated at 120 mL/min, above which the linear CL effect is flat). The weight in the Cockcroft-Gault numerator is the ICU-ADMISSION weight, which includes weight gained from resuscitation fluids, not the pre-admission weight.
  • CG / eCrClCG – Cockcroft-Gault estimated creatinine clearance, raw mL/min, NOT BSA-normalized; the body-weight convention inside the Cockcroft-Gault equation is not stated by the source. Used in Gijsen_2022_meropenem.R (reference 111.7 mL/min, printed in the covariate equation without saying which cohort statistic it is; power exponent 0.637 on clearance in critically ill septic adults with preserved or increased renal function).
  • CrCL (Cockcroft-Gault on total body weight, raw mL/min) – (140 - age) x weight / (72 x SCr in mg/dL), times 0.85 for women, NOT BSA-normalized. Used in Zhou_2022_simurosertib.R (adults with advanced solid tumors; power exponent 0.325 on CL/F centred on 90.45 mL/min, the analysis-data-set median printed in the covariate equation, whereas the Table 2 per-patient median is 89.9 mL/min).
  • CLcr (Bjornson Rcr / SCr, INCOMPLETE) – the source computes Bjornson’s creatinine production rate Rcr (males 27 - 0.173 x age, females 25 - 0.175 x age, both mg/kg/24 h) and divides it by serum creatinine in mg/dL, then labels the result mL/min. The published Bjornson method finishes with CrCl = Rcr x weight / (SCr x 14.4); the weight multiplier and the 14.4 divisor are both omitted because the source dataset recorded no body weight, so the column runs about weight / 14.4 = 4.9-fold BELOW a Cockcroft-Gault clearance for the same patient and is dimensionally (mg/kg/24 h)/(mg/dL), not mL/min. Used in Ahmed_2024_vancomycin.R. This alias is a warning, not just a naming note: because the covariate enters as a power term, supplying a conventional mL/min creatinine clearance to such a model inflates typical CL by 4.9^exponent. Verify the arithmetic against the source’s own cohort median before reusing any model that carries this alias – for Ahmed 2024 the Table 1 median patient (male, 65 years, SCr 1.2 mg/dL) gives (27 - 0.173 x 65) / 1.2 = 13.1 against a reported median of 12.7, which identifies the form unambiguously. A cohort that looks uniformly and implausibly renally impaired is the tell.
  • CLCR,LBW – creatinine clearance computed on LEAN body weight rather than total body weight, raw mL/min and NOT BSA-normalized. Rolsma 2025 builds it with the Cockcroft-Gault equation substituting lean body weight for subjects over 12 years of age and the Schwartz bedside equation for subjects up to 12 years, so a single column spans a paediatric-to-adult cohort under two different estimating equations. Used in Rolsma_2025_cefepime.R (reference 77 mL/min) and Rolsma_2025_meropenem.R (reference 90 mL/min).
  • CrCL (Cockcroft-Gault >= 13 years in mL/min, Schwartz < 13 years in mL/min/1.73 m^2, RRT from ultrafiltration rate and blood flow) – one column spanning a paediatric-to-adult hospitalized-influenza cohort under two estimating equations on two scales, like CLCR,LBW above. Used in Zuo_2020_zanamivir.R as a piecewise-linear hinge on CL: factor 1 at or above an estimated knot of 97 mL/min and 1 + 0.00929 x (CRCL - 97) below it.
  • CrCL (Cockcroft-Gault on total body weight, UNCENTRED exponential) – raw mL/min, NOT BSA-normalized. Used in AlQurain_2022_tramadol.R (older inpatients, cohort mean 65.1 mL/min, range 16-351) as exp(beta * CRCL) on both tramadol CL/F (beta 0.00498) and V2/F (beta 0.0119) with no reference value, the Monolix default; the printed typical CL/F and V2/F are therefore the values at CrCL = 0, not at the typical patient.
  • Clcr (measured 8-h urinary creatinine clearance) – raw mL/min, NOT BSA-normalized. Used in BeraldiMagalhaes_2021_ethambutol.R as the through-origin ratio (CRCL / 101) on clearance (reference 101 mL/min; cohort medians 92.3 mL/min in ICU patients and 113.88 mL/min in outpatients, Table 1). The paper’s simulation methods label the same quantity mL/min/1.73 m^2, which conflicts with its Table 1; the Table 1 unit is taken as authoritative.
  • CLCRurinary (measured urinary creatinine clearance) – raw mL/min, NOT BSA-normalized. Used in NgougniPokem_2022_temocillin.R as the through-origin ratio (CRCL / 39.9) on the clearance from the central compartment (reference 39.9 mL/min, the cohort median, range 20.55-149.3, in critically ill adults with septic shock and ascitic fluid effusion; Supplementary Table S1 Ke = CLi*(CLCRurinary/39.9)/V).
  • CLiohexol – plasma iohexol clearance, the clinical gold-standard tracer-measured glomerular filtration rate; used in Hamren_2008_tesaglitazar.R (tesaglitazar / acyl glucuronide interconversion popPK in subjects with varying renal function). Iohexol is an exogenous contrast agent cleared exclusively by glomerular filtration (no tubular secretion or reabsorption); its clearance is a direct measurement of GFR.
  • CLinulin – plasma inulin clearance, the historical gold standard for measured GFR; analogous to iohexol clearance but requires constant intravenous infusion of inulin.
  • CL_DTPA – plasma 99mTc-DTPA clearance, a radiolabelled tracer of glomerular filtration commonly used in clinical practice.
  • CL_EDTA – plasma 51Cr-EDTA clearance, another radiolabelled glomerular-filtration tracer.
  • CrCl1 – pre-dose creatinine clearance on a scale whose printed unit label is WRONG. Zhang 2015 Table 1 heads the row ml/min with median 1.88 and range 0.94-4.64, magnitudes that are anuric as mL/min and are instead consistent with the SI unit mL/s (1.88 mL/s = 113 mL/min). Used in Zhang_2015_methotrexate.R. This alias is a warning, not just a naming note, and belongs with the CLcr (Bjornson) alias above as the second registered case of a renal-function column that is NOT interchangeable with a conventional mL/min value: the coefficient is calibrated per unit on the source’s own numeric scale and centred at 1.89 on that scale, so supplying mL/min inflates the covariate deviation about 60-fold. Two independent tells identify it. First, the same Table 1 mislabels serum creatinine as mg/dl with median 49 and range 19-81, values that are only sane as umol/L – when one row of a demographics table is demonstrably in SI units, check every other row before trusting its unit column. Second, a Cockcroft-Gault estimate for the paper’s own median patient gives about 178 mL/min against a tabulated median equivalent to 113 mL/min, the expected direction and magnitude of Cockcroft-Gault bias when creatinine production is low, which is consistent with a MEASURED 24 h clearance (the protocol collected 24 h urine). A weight-normalized reading (mL/min/kg, as in Zhang_2026_tebipenem.R) gives essentially the same absolute clearance for the median patient and cannot be excluded from the paper alone; what matters for reuse is that the model be driven on the source’s numeric scale either way.
  • CREACL – ‘creatinine clearance (calculated by CKD-EPI equation)’ reported in mL/min with no statement of whether the CKD-EPI value was de-normalised from mL/min/1.73 m^2; used in Michalickova_2020_riociguat.R (reference 70 mL/min, the cohort median; enters the riociguat non-M1 clearance through the origin as CLe,r/F = 0.66 * (CRCL / 70)).
• Example models: Cirincione_2017_exenatide.R (MDRD eGFR), Xu_2019_sarilumab.R (measured CrCl BSA-normalized), Kotani_2022_astegolimab.R (MDRD eGFR), Li_2019_abatacept.R (cGFR), Bajaj_2017_nivolumab.R (CKD-EPI eGFR, reference 90 mL/min/1.73 m^2), NA_NA_lidocaine.R (DDMODEL00000281; binary stratification at threshold 52.7 mL/min adding -0.319 to the GX rate constant K30 in the CRCL <= 52.7 cohort; the source .ctl does not state the BSA-normalisation method), Delattre_2010_amikacin.R (raw Cockcroft-Gault mL/min, NOT BSA-normalized; reference 55.5 mL/min population median; additive linear effect 1.42 L/h per (CRCL/55.5) on CL), MedellinGaribay_2015_gentamicin.R (Schwartz BSA-normalized CLCR; reference 75 mL/min/1.73 m^2 (population mean 76.7); additive linear effect 0.06 L/h per (CRCL/75) on CL in infants 1-24 months), Georges_2009_ceftazidime.R (raw MDRD-eGFR mL/min, NOT BSA-normalized; mean 121 mL/min; additive linear effect 0.024 L/h per mL/min on CL, no centering), Hamren_2008_tesaglitazar.R (iohexol-clearance-measured GFR; reference 76 mL/min/1.73 m^2; linear centered slope 0.99 %/(mL/min/1.73 m^2) on renal clearance of parent tesaglitazar AND simple linear normalised scaling (CRCL / 76) on the saturable Michaelis-Menten Vmax of acyl-glucuronide renal elimination; founding tracer-GFR example), Chen_2023_nemonoxacin.R (raw Cockcroft-Gault mL/min, NOT BSA-normalized; cohort 114.4 +/- 29.2 mL/min, median 116.2, range 50.7-200.7; additive linear effect 0.026 L/h per mL/min on a 10.3 L/h intercept with no centering, per Chen 2023 Eq. 8 CL = (10.3 + 0.026 x CLcr) x (BW/70)^0.75; source column CLcr), Wada_2023_sparsentan.R (raw Cockcroft-Gault mL/min, NOT BSA-normalized; reference 112 mL/min overall cohort median, range 26-361; power effect (CRCL / 112)^0.222 on CL/F, so lower renal function raises sparsentan exposure even though urinary excretion accounts for only 2.18% of the dose – the paper notes the association may be partly confounded by sex, which is a Cockcroft-Gault input and also a retained covariate on CL/F), Shu_2024_posaconazole.R (raw Cockcroft-Gault mL/min, NOT BSA-normalized; reference 103.81 mL/min cohort median, range 14.44-240.64 in Chinese HSCT recipients spanning ages 3-65; power effect (CRCL / 103.81)^0.68 on CL/F for posaconazole oral suspension, of which only about 13% of a dose is renally excreted – as with Wada 2023 the effect is more plausibly a marker of general physiological reserve than a mechanistic renal-elimination pathway, and the authors flag it as a departure from earlier posaconazole popPK models), Suzuki_2024_mycophenolic_acid.R (Cockcroft-Gault standardised to a fixed 70 kg body weight – i.e. mL/min/70 kg, NOT BSA-normalised and NOT a raw value; cohort 2.5th / 50th / 97.5th percentiles 6.6 / 52.8 / 148.8 in renal-transplant recipients at a single Japanese centre; enters as the ratio CRCL / 100 multiplying the renal arm of an additive renal-plus-non-renal apparent clearance, with body size handled separately by an allometric (WT/70)^0.75 term), Jung_2024_vancomycin.R (bedside Schwartz paediatric eGFR, BSA-normalized; reference 141 mL/min/1.73 m^2; power effect (CRCL / 141)^0.5259 on CL in critically ill children – a less-than-proportional renal effect, and the sole retained covariate on clearance), Nakai_2025_tranexamicAcid.R (raw Cockcroft-Gault mL/min computed with actual body weight, NOT BSA-normalized; reference 61.0 mL/min cohort median, range 21.8-147.5 in adults on cardiopulmonary bypass; power effect (CRCL / 61.0)^0.752 on CL; source column CLcr, and note that the paper’s summary equation misannotates the covariate as “CLcr (L/h)” while dividing by the mL/min median – the Discussion’s worked example resolves the ratio to mL/min), Rolsma_2025_cefepime.R, Rolsma_2025_meropenem.R (alias CLCR,LBW; raw Cockcroft-Gault-on-lean-body-weight / Schwartz-bedside mL/min, NOT BSA-normalized; references 77 and 90 mL/min respectively, both being the cohort typical value; through-origin linear normalisation (CRCL / ref) on the RENAL arm of an additive renal-plus-non-renal clearance, with the non-renal arm carrying the body-size covariate instead), Patel_2025_eteplirsen.R (cystatin-C-based CKD-EPI eGFR, BSA-normalized; reference 145 mL/min/1.73 m^2 cohort median, range 85.5-180 in boys with Duchenne muscular dystrophy; power effect (CRCL / 145)^1.60 on CL. Two features are unusual and worth carrying forward: the cohort is uniformly supranormal – no subject fell below 85.5 – so the model carries no information about renal impairment and the paper says so explicitly; and the estimating equation was chosen because creatinine is invalid in this disease, not merely as a convenience), Lee_2025_levofloxacin.R (raw Cockcroft-Gault mL/min, NOT BSA-normalized; reference 105.71 mL/min cohort median, observed range 74.8-113 in 12 healthy Korean adults; power effect (CRCL / 105.71)^0.901 on CL, the paper having tested and rejected an additional non-CrCl clearance component as negligible – note that the paper’s own Monte Carlo dosing simulations extrapolate this term from the fitted 74.8-113 range out to 10-170 mL/min), Taylor_2026_methotrexate.R (CKD-EPI creatinine-cystatin C 2021 eGFR, BSA-normalized; reference 76 mL/min/1.73 m^2; power effect (CRCL / 76)^0.80 on CL in adults with lymphoma receiving high-dose methotrexate. Two features are worth carrying forward: the covariate is genuinely LONGITUDINAL – creatinine and cystatin C were drawn alongside every methotrexate sample and re-converted to eGFR at each timepoint, so the 76 is the median of the longitudinal record and sits BELOW the baseline medians of 82.5 and 80.2 mL/min/1.73 m^2 because renal function declines across the post-infusion monitoring window; and the source paper’s Table 1 heads the renal block “Estimated kidney function (mL/min)”, pooling the BSA-normalized CKD-EPI equations with a raw Cockcroft-Gault estimate, so that header alone is not a safe unit source – Methods Sect. 2.2.2 is the authority that the CKD-EPI values are mL/min/1.73 m^2), Zhang_2026_tebipenem.R (WEIGHT-normalized mL/min/kg, Sato 2008 paediatric predictive equations from serum creatinine; cohort mean 3.72 +/- 0.97, median 3.73, range 1.46-6.97; additive linear effect 0.104 (L/h/kg) per (mL/min/kg) on a 0.363 L/h/kg intercept with no centering, the whole bracket then multiplied by body weight to give CL/F in L/h – the founding weight-normalized example), Mouksassi_2015_thrombomodulinAlfa.R (raw Cockcroft-Gault mL/min, NOT BSA-normalized, and additionally TRUNCATED at 150 mL/min – the source truncated the upper tail for the covariate analysis “in order to better represent upper physiological range of CrCL”, which bound on 1.6% of Phase I and 9.5% of Phase IIb subjects, so a user supplying real data must apply the same truncation or the term extrapolates past what was fitted; reference 72.6 mL/min; power effect (CRCL / 72.6)^0.27 on the CL of renally-eliminated thrombomodulin alfa, pooled healthy adults plus adults with sepsis and suspected DIC spanning 5-150 mL/min; BASELINE value, the paper having screened baseline and time-varying CrCL separately and retained baseline; subjects on renal replacement therapy for ESRD or acute kidney injury were excluded, so the model carries no information about dialysis), Chandorkar_2015_ceftolozane.R, Chandorkar_2015_tazobactam.R (raw Cockcroft-Gault mL/min on ACTUAL body weight, NOT BSA-normalized; BASELINE, the paper having verified that serum creatinine was stable over the short treatment course; power effects (CRCL / 109)^0.715 and (CRCL / 115)^0.67 on the clearance of the two analytes of one fixed-ratio combination. The two references differ because the two analytes were fitted to different subject sets – 376 subjects for ceftolozane against 243 for tazobactam, the cUTI study having dosed ceftolozane alone – so a single paper can legitimately carry two centring constants for what is nominally the same column, and they are not interchangeable. Both models are also a rare instance of a paper publishing enough of its own answer key to verify the transcription: the Figure 1 tornado panels print relative CL at six CrCL values per analyte, and (CrCL/109)^0.715 reproduces all six ceftolozane labels exactly), Tseng_2026_piperacillin.R (2021 CKD-EPI creatinine-cystatin C eGFR, BSA-normalized; power effect (CRCL / 104.368)^1.12 on CL in hospitalized adults with a body mass index at or below 18.5 kg/m^2. Three points are worth carrying forward. First, the reference constant 104.368 in the paper’s printed Table 2 formula does NOT equal the Table 1 cohort median of 109.5 mL/min/1.73 m^2 even though the prose calls it “the median eGFR value”; the printed constant is the one this model uses, and replicating the paper’s own Figure 3 target-attainment table confirms it (RMSE 2.06 percentage points against 4.40 for 109.5). Second, this is the mirror image of the Patel_2025_eteplirsen.R rationale: both papers reject creatinine as a renal marker because their cohort’s muscle mass is atypical, but in Duchenne muscular dystrophy creatinine is elevated by muscle breakdown while in low body weight it is depressed by low muscle mass, so the creatinine-based equation errs in opposite directions. Third, the cohort skews preserved-to-augmented (median 109.5, IQR 36) and patients who developed acute kidney injury or needed renal replacement therapy were excluded, so the model carries no information about unstable renal function; the paper’s own dosing simulations nevertheless extrapolate the term across 20-130 mL/min/1.73 m^2), Ma_2026_colistinSulfate.R (raw Cockcroft-Gault mL/min, NOT BSA-normalized; reference 94.76 mL/min, cohort mean 100.42 +/- 68.87, observed range 12.14-246.91 in Chinese ICU adults with carbapenem-resistant infections; power effect (CRCL / 94.76)^0.232 on the CL of intravenous colistin sulfate, the sole retained covariate. Two features are worth carrying forward: the centring constant 94.76 appears ONLY inside the final-model equation printed in the Results text and is neither tabulated nor equal to the reported cohort mean, so a table-only read of the paper recovers the exponent but not the reference; and patients on CRRT or ECMO were excluded by design, so the term carries no information about renal replacement therapy even though the cohort is otherwise a renal-impairment-rich ICU population), Xie_2022_colistinSulfate.R (raw Cockcroft-Gault mL/min, NOT BSA-normalized; reference 57.5 mL/min, printed only inside Eq. 3 and not the Table 1 cohort median of 48.8 (range 6.5-193.8) in Chinese ICU adults; power effect (CRCL / 57.5)^0.353 on the CL of intravenous colistin sulfate. Unlike Ma 2026, two of the 20 subjects were on CRRT with no separate dialysis term, and the paper’s own dosing tables are not reproduced in level by the printed parameters – see the model vignette), Ahmed_2024_vancomycin.R (Bjornson Rcr / SCr with the weight and 14.4 factors OMITTED – see the CLcr (Bjornson) alias above; reference 12.7, the cohort median, the source’s own Methods defining the reference as mean(CLcr) but never printing a mean; power effect (CRCL / 12.7)^0.49 on CL in Sudanese adult inpatients. Two features are worth carrying forward. First, this is the founding example of a renal-function column that is NOT interchangeable with any other member of this canonical: it is deflated about five-fold by two omitted factors, so it is the one case where reusing the model with a standard CrCl silently biases clearance by more than two-fold. Second, the paper prints the covariate model twice in mutually inconsistent forms – the Methods give the centered (CLcr / mean(CLcr))^beta that was fitted, while Equation 5 and the Table 3 footnote drop the denominator; the uncentered reading overshoots the paper’s own Table 4 simulated clearances roughly three-fold, which settles it), Peng_2025_meropenem.R (raw Cockcroft-Gault mL/min, NOT BSA-normalized; the lowest-renal-function cohort registered against this canonical – median 13.6 mL/min, IQR 6.9-22.2, in Chinese critically ill adults all of whom were on CVVH, so the column is RESIDUAL native function sitting alongside the extracorporeal QEFF arm rather than the whole of renal elimination. Enters as a median-CENTERED EXPONENTIAL exp(0.0183 * (CRCL - 13.6)) on the endogenous clearance arm – note this is the log-linear form, not the power-of-a-ratio form that most entries above use, so the coefficient carries units of per mL/min and is not a dimensionless exponent. The paper’s dosing simulations extrapolate the term to 10-50 mL/min, roughly twice the upper end of the fitted range), Bai_2024_imipenem.R (raw Cockcroft-Gault mL/min, NOT BSA-normalized, computed in SI creatinine units – the Table 1 footnote prints the exact form used, CrCl = [(140 - age) x weight (kg)] / [0.818 x Cr (umol/L)], multiplied by 0.85 for women, so the 0.818 divisor plays the role of the familiar 72 when serum creatinine is in mg/dL; reference 99.896 mL/min; power effect (CRCL / 99.896)^0.473 on CL in critically ill Chinese ICU adults with sepsis, the sole covariate retained after forward inclusion and backward elimination, dropping the objective function from 776.148 to 734.330 and the CL between-subject CV from 38.562% to 35.748%. Two points are worth carrying forward. First, this is a third instance of the centring-constant-versus-tabulated-median mismatch already recorded for Tseng_2026_piperacillin.R and Ma_2026_colistinSulfate.R: the text below Eq. 8 calls 99.896 “the median value of CrCl” while Table 1 gives the cohort median as 99.34, and the equation constant is the one the fitted typical CL of 11.357 L/h is conditioned on. The discrepancy is immaterial here – at the exponent 0.473 it moves CL by 0.26% – but the pattern is now frequent enough that the printed equation, not the demographics table, should be treated as the authority for the centring value. Second, the paper states an explicit applicability window, “the predication model can only be used within the range (17.80-256.22 mL/min)”, and unlike several of the models above its own dosing simulations stay inside that window), Beijer_2026_cloxacillin.R (Lund-Malmo Revised 2018 creatinine-based eGFR, BSA-normalized – the founding LMR18 example; reference 67 mL/min/1.73 m^2, cohort median 72, IQR 61-84, range 18-142 in Swedish adults undergoing hip or knee arthroplasty; power effect (CRCL / 67)^0.51 on the UNBOUND clearance of cloxacillin, alongside a (WT / 84)^0.67 body-weight term. Three points are worth carrying forward. First, this is the one entry so far that registers the estimating-equation choice as a result: the paper’s Table S1 correlates seven eGFR variants against the empirical-Bayes eta_CL and LMR18 wins in both relative (r = 0.32) and absolute (r = 0.43) form, beating CKD-EPI 2021 (0.15 / 0.29) and Cockcroft-Gault (0.21 weight-adjusted 0.18), so a user substituting a CKD-EPI value is supplying a measurably worse predictor and not merely a different flavour of the same column. Second, the BETTER-correlating absolute LMR18 estimate was deliberately REJECTED in favour of the relative one, because absolute eGFR left every anthropometric covariate (adjusted / ideal / total body weight, BMI, BSA) independently correlated with clearance at p < 0.001 – the relative estimate plus a separate weight term is the decollinearized parameterisation, and the two arms of that pair must be supplied together or body size is double-counted. Third, this is a fourth instance of the centring-constant-versus-tabulated-median mismatch recorded for Tseng_2026_piperacillin.R, Ma_2026_colistinSulfate.R and Bai_2024_imipenem.R: Table S2’s printed CL equation divides by 67 while Table 1 gives the cohort median as 72, and here the Figure 1 caption independently names “eGFR 67 mL/min/1.73 m2” as the typical patient, corroborating the equation over the demographics table), GarciaHervalejo_2026_teicoplanin.R (CKD-EPI eGFR, BSA-normalized; reference 92.15 mL/min/1.73 m^2, the development-cohort median, observed range 26-141 in adults with haematological malignancies treated for febrile neutropenia; power effect (CRCL / 92.15)^0.35 on the CL of a one-compartment teicoplanin model. Two points worth carrying forward. First, this is another instance of the Table-1-header-versus-Methods unit hazard already recorded for Taylor_2026_methotrexate.R, but running the opposite way: here Table 1 heads the row mL/min/1.73m2 while the Table 2 abbreviation footnote writes in mL/min., and Table 1 governs because CKD-EPI is BSA-normalized by construction and the Methods never mention de-normalising it. Second, the renal term is the WEAKEST and least precisely estimated fixed effect in that model – RSE 47%, bootstrap 95% CI 0.06-0.67, a lower bound near zero – and it is comprehensively dominated by the co-retained IBW term at exponent 3.2; across the full observed eGFR range the renal multiplier spans only 0.64-1.16. Patients with end-stage renal disease or on renal replacement therapy were excluded by design, so the model carries no dialysis information), Hadigol_2026_talazoparib_enzalutamide.R (raw un-normalised mL/min, source column BCCL, reference 86.85 mL/min; power effect (CRCL / 86.85)^0.455 on talazoparib apparent BASE clearance, i.e. on the interaction-free intercept of a drug-drug-interaction model rather than on total clearance. Two cautions specific to this entry: the estimating equation is NOT named anywhere in the paper – the raw-mL/min renal-disease cutoffs of its Table 4 are the only evidence for the normalisation – and the reference 86.85 mL/min is presented only as the value for the typical patient, never as the cohort median, with the demographics table not open-access, so it must not be cited as a median), Komatsu_2016_penicillinG.R (raw Cockcroft-Gault mL/min, NOT BSA-normalized; cohort mean 82.52 +/- 33.29, range 11-144 in Japanese adults with infective endocarditis. The most extreme functional form registered against this canonical: a bare THROUGH-ORIGIN product with no intercept and no centring, CL (L/h) = 0.21 x CRCL (mL/min), so the covariate is not a multiplier on a typical clearance but the whole of it and the parameter carries the compound unit L/h per mL/min. Two consequences a reuser must know. First, there is no non-renal clearance arm at all, so the model predicts zero elimination at zero renal function – which is false for penicillin G – and it must not be used at or near anuria even though the paper’s own nomogram recommends a regimen for CLcr below 20 mL/min and its Monte Carlo simulations extrapolate to 5 mL/min, below the observed minimum of 11. Second, the absence of a reference value means there is no centring constant to get wrong, which is the one hazard recorded for Tseng_2026_piperacillin.R, Ma_2026_colistinSulfate.R, Bai_2024_imipenem.R and Beijer_2026_cloxacillin.R that this entry is structurally immune to. The form is verifiable from the paper’s own Discussion, which states a half-life of 0.79 h “calculated by keeping the CLcr fixed at 120 mL/min”: 0.693 x 28.9 / (0.21 x 120) = 0.795 h, which pins the slope, the volume and the mL/min unit simultaneously), Kuchimanchi_2022_diroximelFumarate.R (MDRD eGFR denormalized to absolute mL/min with each participant’s body surface area, per ESM Table S2 footnote b, so the per-model units are mL/min rather than the register default; power (CRCL / 111.9)^0.547 on 2-hydroxyethyl succinimide clearance only, 111.9 mL/min being the cohort median), Wang_2021_mycophenolic_acid.R (MDRD eGFR modified for Chinese patients, Ma 2006; linear centring 1 + 0.00791 x (CRCL - 57) on CL/F), He_2020_phosphocreatine.R (bedside Schwartz paediatric eGFR, with the original Schwartz 0.45 constant for infants under 1 year, BSA-normalized, source column GFR; reference 127.78 mL/min/1.73 m^2, the cohort median, range 66.33-224.01 in Chinese children with myocarditis; power effect (CRCL / 127.78)^0.311 on the clearance of the METABOLITE creatine formed from infused phosphocreatine, not on the parent drug), Royer_2021_palbociclib.R (raw Cockcroft-Gault mL/min, NOT BSA-normalized; reference 78.9 mL/min, the cohort MEAN (the paper’s Methods normalize every continuous covariate by the population mean; median 72.1, range 23.4-282.3) in French women with metastatic breast cancer sampled during routine therapeutic drug monitoring; power effect (CRCL / 78.9)^0.419 on palbociclib CL/F, retained over body weight and age in backward deletion although renal excretion is under 8% of palbociclib clearance – the authors read CRCL as a composite of the weight, age and creatinine it is computed from), WerumeusBuning_2021_ceftazidime.R (CKD-EPI 2009 eGFR, BSA-normalized; reference 73 mL/min/1.73 m^2, the cohort median, range 6-153 in critically ill adults; power effect (CRCL / 73)^0.772 on the ceftazidime clearance of patients OFF continuous veno-venous hemofiltration only – the term is switched off together with the eGFR-driven branch when RRT_CRRT_STATUS = 1, so a CVVH patient’s CRCL value is ignored but must still be finite. Table 1 prints the unit as mL/min/m 2, a dropped 1.73), BilbaoMeseguer_2021_levetiracetam.R (creatinine clearance MEASURED from a timed urine collection, raw mL/min, NOT BSA-normalized; cohort median 117 mL/min, range 54-239, in Spanish ICU adults, 37% with augmented renal clearance; enters as a POWER term with no multiplying coefficient added to a fixed non-renal arm, CL = 3.5 + (CRCL / 120)^2.5 L/h, so the renal arm is 1 L/h at the 120 mL/min divisor. Two points worth carrying forward. First, the exponent is a typeset superscript in both the Table 3 header and the Results equation, and a plain-text extraction flattens it into what reads as a linear slope 3.5 + 2.5 x CrCl/120; the paper’s Discussion settles the form, quoting mean CL of 4.5 L/h at 120 and 9.2 L/h at 240 mL/min, which only the power reading reproduces. Second, with an exponent of 2.5 the term is steeply supra-proportional, so a Cockcroft-Gault or BSA-normalized value substituted for the measured urinary clearance moves CL far more than it would under the near-proportional exponents most entries above carry; CrCl below 50 mL/min was an exclusion criterion), NgougniPokem_2022_temocillin.R (creatinine clearance MEASURED from a urine collection, raw mL/min, NOT BSA-normalized; reference 39.9 mL/min, the cohort median, range 20.55-149.3, in Belgian ICU adults in septic shock with intra-abdominal infection and ascitic fluid effusion; through-origin linear ratio (CRCL / 39.9) on the unbound temocillin clearance from the central compartment, with no non-renal arm on that clearance – the only other elimination route is the abdominal drain from the ascitic fluid compartment, so the model must not be used near anuria), Duong_2022_gentamicin_rea.R (MDRD eGFR 186.3 x (Scr/88.4)^-1.154 x Age^-0.203 x 1.212 [black] x 0.742 [female], Scr in umol/L, labelled mL/min by the source with no BSA de-normalisation described; enters gentamicin clearance as the sigmoid clmax * CRCL^1.2 / (crcl50^1.2 + CRCL^1.2) with crcl50 = 129 mL/min, the Rea 2008 structure refit to a Quebec ICU cohort), Duong_2022_gentamicin_bos.R (raw Cockcroft-Gault (140 - Age) x WT x 1.23 x 0.85 [female] / Scr, Scr in umol/L, mL/min, NOT BSA-normalised; linear-centred 1 + 0.00925 * (CRCL - 92) on gentamicin CL, centred at the 92.2 mL/min cohort mean – the Bos 2019 structure refit to the same Quebec cohort), Yu_2022_colistinSulfate.R (raw Cockcroft-Gault mL/min, NOT BSA-normalized; cohort 79.54 +/- 53.99 mL/min in Chinese ICU adults with carbapenem-resistant infections; additive linear effect CL = 0.994 + 0.525 x CRCL / 66.47 L/h on intravenous colistin sulfate, the same divisive-normalization form as Delattre_2010_amikacin.R. The divisor 66.47 appears only in the printed equation and is not the tabulated cohort mean; concentrations drawn during renal replacement therapy or ECMO were excluded).
• Notes: All estimation methods (creatinine-based MDRD / CKD-EPI / measured CrCl, and tracer-based iohexol / inulin / DTPA / EDTA clearance) produce values in the same units and are operationally interchangeable as a covariate on clearance. Three distinct size normalisations appear across the example models and must be documented per model, because they are not interchangeable: BSA-normalised (mL/min/1.73 m^2, the canonical default), raw un-normalised (mL/min, e.g. Delattre_2010_amikacin.R, Chen_2023_nemonoxacin.R), and per-70-kg-body-weight (mL/min/70 kg, e.g. Suzuki_2024_mycophenolic_acid.R). Supplying a value on the wrong normalisation silently rescales the renal-function term, and in a model that also carries an allometric weight term it double-counts body size. Document the method explicitly in each model’s covariateData[[CRCL]]$description so future reviewers can trace the source assay; the assay choice affects accuracy of the absolute GFR value but does not change the covariate’s role as a linear / power scalar on renal clearance. Tracer-measured GFR is the clinical gold standard (unaffected by muscle mass, malnutrition, and tubular-secretion confounders that perturb creatinine-based estimates) but is rarely available outside research and renal-impairment-cohort studies; creatinine-based estimates dominate routine clinical popPK datasets. Tracer-measured-GFR scope codified on 2026-06-17 alongside the Hamren 2008 tesaglitazar extraction (by the maintainers’ decision).

RENALFUNC_REL (canonical for relative renal function, 1 = normal)

• Description: Dimensionless per-subject renal function expressed as a FRACTION OF NORMAL, so that 1 = normal renal function, 0.5 = half of normal, and values above 1 describe supranormal filtration. This is the organ-function scalar of the Groningen / Medimatics MwPharm and Edsim++ modelling tradition, where it lives in a dedicated organ-function object (the ORG / P object) and multiplies the renal arm of a decomposed clearance: CLi = CL * (fe * RENALFUNC_REL + (1 - fe) * HEPFUNC_REL), with the paper-named parameter fe supplying the renal share of total clearance. Compute it as measured GFR / reference normal GFR (or as the ratio of any consistent renal-function measure to its own normal reference) – the ratio, not the absolute measurement, is what this column carries.
• Units: (dimensionless)
• Type: continuous
• Scope: general
• Reference category: n/a – the column IS the multiplier and enters the model directly; 1 denotes normal renal function. No centering or power transform is applied.
• Source aliases:
  • RF – Visscher 2025 Methods section 2.3 and Edsim++ ORG object symbol for relative renal function. Same orientation and scale, no value transformation. NOTE: the bare two-letter name RF must NOT be used as the canonical, because it is already recorded in this register as a rejected alias of the rheumatoid-factor column.
  • RENAL_FUNCTION_FRACTION, GFR_REL, GFR/GFR_normal – descriptive forms of the same ratio.
• Example models: Visscher_2025_parathyroidHormone.R (founding example; single-patient rhPTH(1-84) PK-PD in postsurgical hypoparathyroidism. Multiplies the renal fraction fe = 0.3 of the 39.93 L/h population clearance per Visscher 2025 Eq. 3. Held at 1 in the validation vignette because the source states that the equation corrects for the variations in GFR but reports NO numeric RF values and NO reference-GFR denominator from which the ratio could be recovered – and Table S3 omits the 24-h urine volume that the paper’s own GFR equation needs, so the per-interval ratio cannot be reconstructed from the supplement either.).
• Notes: Genuinely distinct from every pre-existing renal-function entry, which is why it needed its own canonical rather than an alias: CRCL and CRCL_BASE are ABSOLUTE renal function in their own units (mL/min/1.73 m^2) and converting one to a relative multiplier requires a reference denominator that a source paper often does not state; KBF is an absolute organ blood flow; and RENALIMP_MILD / RENALIMP_MOD / RENALIMP_SEV are CATEGORICAL impairment bands that cannot express a continuous multiplier. A model that reports an absolute GFR should use CRCL; use RENALFUNC_REL only when the source’s own parameterisation is the dimensionless fraction-of-normal. Sister canonical to HEPFUNC_REL; the two are normally used together, since the point of the parameterisation is to let the renal and non-renal arms of clearance vary independently. Expect reuse: the relative-organ-function scalar is the standard parameterisation of the MwPharm / Edsim++ family (Edsim++ author N.C. Punt is a co-author of the founding paper, whose reference list cites a sibling model from the same group and framework).

HEPFUNC_REL (canonical for relative liver function, 1 = normal)

• Description: Dimensionless per-subject hepatic function expressed as a FRACTION OF NORMAL, so that 1 = normal liver function and 0.5 = half of normal metabolic capacity. Companion to RENALFUNC_REL in the MwPharm / Edsim++ organ-function (ORG / P) object, multiplying the non-renal (hepatic-metabolic) remainder of a decomposed clearance: CLi = CL * (fe * RENALFUNC_REL + (1 - fe) * HEPFUNC_REL).
• Units: (dimensionless)
• Type: continuous
• Scope: general
• Reference category: n/a – the column IS the multiplier and enters the model directly; 1 denotes normal liver function.
• Source aliases:
  • LF – Visscher 2025 Methods section 2.3 and Edsim++ ORG object symbol for relative liver function. Same orientation and scale, no value transformation.
  • HEP_FUNCTION_FRACTION, LIVERFUNC_REL – descriptive forms of the same ratio.
• Example models: Visscher_2025_parathyroidHormone.R (founding example; multiplies the hepatic remainder 1 - fe = 0.7 of total clearance. Held at 1 because Visscher 2025 Methods states explicitly that “LF is the relative liver function (equal to 1 in this model)” – the patient had normal hepatic function and no hepatic-impairment data inform the parameter. Retained in the model rather than folded into CL so the paper’s own renal / hepatic decomposition of clearance stays visible and user-perturbable.).
• Notes: Distinct from the HEPIMP / HEPIMP_MILD / HEPIMP_MOD / HEPIMP_SEV / HEPIMP_MODSEV family, which are binary Child-Pugh-style impairment INDICATORS, and from the underlying laboratory measures AST / ALT / BILI / ALB: this column is a continuous fraction-of-normal metabolic-capacity multiplier, not a category or a lab value. A source that reports a Child-Pugh class should use the HEPIMP_* family; use HEPFUNC_REL only when the source’s own parameterisation is the dimensionless fraction-of-normal. Sister canonical to RENALFUNC_REL. A value of exactly 1 is a legitimate and common entry – it records that the source’s structural decomposition of clearance exists and was exercised at normal hepatic function, which is information a downstream user needs in order to perturb it.

KBF (canonical for native kidney (renal) blood flow)

• Description: Whole-organ blood flow perfusing the kidneys, used as a physiological system input in mechanistic kidney / renal PBPK models, where it sets the peritubular vascular flow that carries drug past the tubular epithelium. May be measured (para-aminohippurate or isovalerylglycine renal clearance, phase-contrast MRI, Doppler ultrasound) or assumed from a population physiological default. Distinct from the extracorporeal-circuit flows BFR / DFR / Q_CVVH / QBL / QEFF, which describe dialysis or ECMO hardware rather than native renal perfusion, and from CARDIAC_OUTPUT (whole-body hemodynamic output, of which renal blood flow is roughly one fifth).
• Units: mL/min (raw whole-organ flow, NOT BSA-normalized; the Huang & Isoherranen mechanistic kidney framework consumes it internally as L/h via the 1 mL/min = 0.06 L/h conversion)
• Type: continuous
• Scope: general
• Reference category: n/a – enters the model directly as the peritubular vascular flow, not as a ratio or centered deviation. Healthy-adult physiological default in the founding framework: 60 L/h (= 1000 mL/min), scaled proportionally with GFR in the population version of the model (Huang & Isoherranen 2020 distributed MATLAB code, Q_kidney0 = 60 * Proportional_Reduction). Granda 2024 replaces that population default with a per-subject measurement, which is the paper’s central methodological contribution.
• Source aliases:
  • Isovalerylglycine CL_r – Granda 2024 Table 3 column heading; the renal clearance of the endogenous solute isovalerylglycine used as the KBF surrogate (high extraction ratio, low protein binding, clearance more than fourfold GFR). Same orientation and units, no value transformation.
  • Q_kidney / Qk / Q_kidney0 – Huang & Isoherranen 2018 / 2020 and Chang 2023 symbol for the same quantity, expressed in L/h (divide by 0.06 to obtain the canonical mL/min).
  • RBF / RPF – renal blood flow / renal plasma flow shorthand common in renal-physiology and PAH-clearance literature. RPF is PLASMA flow and must be converted to blood flow (KBF = RPF / (1 - haematocrit)) before use.
• Example models: Granda_2024_kynurenicacid_pbpk.R, Granda_2024_tenofovir_pbpk.R, Granda_2024_oseltamivircarboxylate_pbpk.R (all three: per-subject KBF estimated from measured isovalerylglycine renal clearance, mL/min; cohort mean 575 +/- 387, range 76-1692 across 27 adults spanning CKD stages 1-5; enters as the peritubular blood flow qk <- KBF * 0.06 in the 35-state mechanistic kidney PBPK).
• Notes: General scope because native kidney blood flow is a paper-independent physiological quantity that any renal-PBPK or renal-physiology model may use. The canonical name states the quantity rather than a paper symbol, per the standing canonical-naming rule. Note the physiological constraint the founding models enforce: secretory clearance cannot exceed kidney blood flow, so a subject whose measured secretory-marker clearance exceeds their measured KBF cannot be fitted (Granda 2024 participants 38, 41 and 45, whose fitted secretory clearance therefore railed at the 1000 L/h search bound); a model consuming KBF should surface that violation rather than silently saturating. The companion per-subject quantity in Granda 2024 – unbound intrinsic secretory clearance – is deliberately NOT registered as a covariate: it is a fitted model parameter (lclintsec), not a measurement, per the same ruling.

CREAT (canonical for serum creatinine)

• Description: Serum creatinine concentration (baseline or time-varying).
• Units: umol/L or mg/dL – document the unit used in each model via covariateData[[CREAT]]$units.
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (CREAT / ref)^exponent.
• Source aliases:
  • CRE (umol/L, reference 70.73) – used in Thakre_2022_risankizumab.R.
  • SCR – common clinical-PK abbreviation; also Llanos-Paez 2020 source column for the patient’s individual serum creatinine.
  • Serum_creatinine – the NONMEM $INPUT column in Ding_2026_vancomycin.R (umol/L), abbreviated Scr in that stream’s $PK block.
• Example models: Thakre_2022_risankizumab.R, Hennig_2013_tobra.R (umol/L; paired with CREAT_REF for the SCR_mean / SCR ratio used in the Hennig 2013 renal-function factor), Llanos_2017_gentamicin.R (umol/L; standardized per-patient against CREAT_REF rather than a fixed cohort reference), LlanosPaez_2020_gentamicin.R (umol/L; used as the patient’s SCR_i in the renal-function ratio (CREAT_REF / CREAT)^0.58 on CL).
• Notes: CREAT chosen over the shorter CRE/SCR as the NONMEM/clinical-PK convention that is unambiguous. Per-model reference values must be documented in covariateData[[CREAT]]$notes.

CREAT_REF (canonical for sex/age/size-expected normal-mean serum creatinine)

• Description: Externally-computed reference serum creatinine for the individual (the expected normal SCR for a healthy person of the same sex, age and body size). Used as the numerator of a ratio against the patient’s measured CREAT to define a renal-function factor on clearance.
• Units: umol/L or mg/dL – must match the unit of the paired CREAT column. Document via covariateData[[CREAT_REF]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – used as (CREAT_REF / CREAT)^exponent so that a patient with measured SCR equal to the population-expected normal SCR has factor 1.
• Source aliases:
  • SCR_mean – used in Hennig 2013 (Eq. 5: f_SCR = (SCR_mean / SCR)^theta_SCR); also Llanos-Paez 2020 paper notation for the Ceriotti 2008 age/sex-matched physiological mean SCR.
  • Scrmean – Llanos-Paez 2017 paper notation; computed from Ceriotti et al. 2008 age- and sex-stratified medians (Clin Chem 54:559-566, doi:10.1373/clinchem.2007.099648).
  • SCR_standardised – Germovsek 2018 paper notation; PMA-adjusted standardisation of raw SCR per the paper’s reference 28 (a previously-developed Standing-style PMA stratification).
  • CCR,adj – Ruhs 2012 paper notation; age- and gender-adjusted reference creatinine derived from the paper’s reference [23] (a Schwartz-style paediatric maturation adjustment); the main text does not give the explicit formula.
• Example models: Hennig_2013_tobra.R, Llanos_2017_gentamicin.R (umol/L; computed externally per Ceriotti et al. 2008), LlanosPaez_2020_gentamicin.R (umol/L; ratio (CREAT_REF / CREAT)^0.58 multiplies the maturation-scaled CL), Germovsek_2018_meropenem.R (umol/L; ratio (CREAT_REF / CREAT)^0.40 multiplies the maturation-scaled CL for renal meropenem clearance in neonates and young infants; PMA-stratification reference per Germovsek 2018 Methods reference 28), Ruhs_2012_methotrexate.R (mg/dL; age- and gender-adjusted CCR,adj per the paper’s reference [23]; ratio (CREAT_REF / CREAT)^0.314 multiplies the BSA-scaled MTX CL in paediatric ALL patients).
• Notes: Specific scope because the formula used to derive the reference value is paper-defined (Hennig 2013 cites a combination of Ceriotti 2008, Junge 2004 and Johansson 2011 reference-interval relationships; Llanos-Paez 2017 and 2020 both use Ceriotti 2008); a future paper that uses a different reference-SCR derivation (e.g., a CKD-EPI-style adult-only reference, or a Schwartz-derived paediatric-only reference) should pin its formula in covariateData[[CREAT_REF]]$notes so that a user assembling a virtual cohort can reproduce it. When no covariate data are available to compute CREAT_REF, set CREAT_REF = CREAT so the renal-function factor evaluates to 1 (matching the Hennig 2013 ‘covariate set to 1 for missing data’ rule).

BUN (canonical for blood urea nitrogen)

• Description: Blood urea nitrogen concentration (baseline or time-varying). Reflects the nitrogenous waste burden being cleared by the kidneys; rises with reduced glomerular filtration, dehydration, increased protein catabolism, or GI bleeding. Distinct from CREAT (the other commonly-reported renal-function marker) because urea reabsorption is flow-dependent in the renal tubule, so BUN is sensitive to volume status as well as GFR.
• Units: mg/dL or mmol/L – document the unit used in each model via covariateData[[BUN]]$units (1 mmol/L urea ~= 2.80 mg/dL BUN).
• Type: continuous
• Scope: general
• Reference category: n/a – used either as a linear / hinge effect on a PK parameter, or with power scaling (BUN / ref)^exponent. Reference values observed: 7 mg/dL (Hall 2017 MARS hinge knot for ka; not a population median); 4.2 mmol/L (Chen 2017 cohort median for the tacrolimus power-of-ratio CL effect); 5 mmol/L (Wang 2026 cohort median for the ciprofol power-of-ratio V1 effect – not printed in that paper and recovered algebraically, see below).
• Source aliases: none known.
• Example models: Hall_2017_dapsone.R (mg/dL; population median 13 mg/dL [range 7-28]; enters the MARS-based covariate model on the absorption rate constant via the basis function BF1 = max(0, BUN - 7), which interacts with a weight hinge to drive Ka), Chen_2017_tacrolimus.R (mmol/L; cohort median 4.2 mmol/L [range 1.7-10.4]; enters CL/F via standard power-of-ratio scaling (BUN/4.2)^1.42 in low-dose oral tacrolimus for Chinese myasthenia-gravis patients), Wang_2026_ciprofol.R (mmol/L; cohort median 5 mmol/L [range 2.9-7.6]; enters the central volume via power-of-ratio scaling (BUN/5)^-0.821 in Chinese children aged 1-9 years given a single ciprofol bolus).
• Notes: Promoted to general scope on 2026-06-03 alongside the Chen 2017 tacrolimus extraction, the second model registering BUN. Hall 2017 enters BUN only through a piecewise-linear MARS hinge (max(0, BUN - 7) * max(0, 63.7 - WT)), not as a power scaling – the per-model covariateData[[BUN]]$notes documents this is part of a machine-learning-driven hinge model, not a standard popPK covariate transform. Chen 2017 uses the conventional centred power-of-ratio form (BUN/ref)^exponent referenced to the cohort median; the positive exponent there is interpreted via a urea-driven protein-carbamylation mechanism that reduces albumin binding of the ~99%-protein-bound tacrolimus. Wang 2026 is the first model to enter BUN on a distribution volume rather than a clearance, and the first with a negative exponent: higher urea is interpreted there as a marker of reduced effective plasma volume, shrinking the initial distribution space of a highly protein-bound anaesthetic. It is also a worked example of recovering an unprinted normalising median – that paper prints only the cohort mean (5.2 mmol/L), but quotes V1 at two BUN values, and solving 506 * (BUN/ref)^-0.821 against both quoted volumes returns ref = 5.00 twice over.

UACR (canonical for urine albumin-to-creatinine ratio)

• Description: Urine albumin-to-creatinine ratio (UACR), a renal-damage biomarker quantifying albuminuria. UACR is the spot-urine albumin concentration divided by the spot-urine creatinine concentration; it approximates 24-hour urinary albumin excretion (in mg/day) because creatinine excretion is roughly constant at ~1 g/day in an adult. UACR is the principal albuminuria measure in KDIGO chronic-kidney-disease (CKD) staging, with thresholds A1 (< 30 mg/g, normal-to-mildly increased), A2 (30-300 mg/g, moderately increased), and A3 (> 300 mg/g, severely increased). Distinct from the function markers CRCL / CREAT / BUN: UACR captures upstream glomerular albumin leak (kidney damage), whereas CRCL / CREAT / BUN capture downstream filtration rate (kidney function). Both axes are routinely reported together in diabetic-kidney-disease (DKD) and CKD popPK / PD studies. Time-fixed at baseline by default; document baseline-vs-time-varying status per model via covariateData[[UACR]]$notes.
• Units: mg/g (urine albumin in mg per gram of urine creatinine). Equivalent to mg/mmol scaled by 8.84 (1 mg/g ~= 0.113 mg/mmol). Document the unit used in each model via covariateData[[UACR]]$units.
• Type: continuous
• Scope: general
• Reference category: n/a – used either as a centred linear effect on (UACR - ref) or as a power scaling (UACR / ref)^exponent. Reference values observed: 800 mg/g (Goulooze 2022 serum-potassium analysis centring value) and 850 mg/g (Goulooze 2022 UACR / eGFR analysis centring value); both are close to the FIDELIO-DKD cohort median of 852 mg/g.
• Source aliases:
  • UACR0 – Goulooze 2022 paper convention for baseline UACR (the 0 suffix denotes the baseline / pre-randomisation value).
• Example models: Goulooze_2022_finerenone.R (centred linear effects on Emax and on the disease-progression slope TSLOPE, both centred at 800 mg/g; coefficients 9.31e-5 g/mg on Emax and 1.14e-3 g/mg on TSLOPE per Goulooze 2022 Table 1), Goulooze_2022_finerenone_uacr.R (power scaling (UACR / 850)^0.877 on the baseline of the UACR disease-progression state), Goulooze_2022_finerenone_egfr.R (same scaling on the embedded UACR sub-model baseline, plus (UACR / 850)^0.306 on the magnitude of the inter-individual variability of the eGFR decline rate – a covariate on a variance term rather than on a typical value).
• Notes: Promoted from specific to general scope alongside the Goulooze 2022 UACR / eGFR extraction, which was the second paper to register the column (as the original entry anticipated). Distinct from the uacr COMPARTMENT, which is the modelled UACR time course; this column is the time-fixed observed baseline.

UPRO_BL (canonical for baseline 24-hour urinary protein excretion)

• Description: Baseline (pre-treatment) total urinary protein excretion over 24 hours, the “daily urinary protein excretion” or “proteinuria” reported by glomerular-disease trials. Measured from a timed 24-hour urine collection (or, where a paper says so, estimated from a spot urine protein-to-creatinine ratio) and expressed as mass per day. Time-fixed at baseline; in an MBMA it is the study arm’s median or mean baseline rather than a single patient’s value.
• Units: g/day. Document per-model via covariateData[[UPRO_BL]]$units; a source in mg/day converts by dividing by 1000.
• Type: continuous
• Scope: specific
• Reference category: n/a – continuous baseline. Reference values observed: 1.82 g/day (Yu 2022 centring value, close to the median arm baseline of its 40 IgA-nephropathy trials).
• Source aliases:
  • Baseline / daily urinary protein excretion at baseline – Yu 2022 Equation 7 and Results.
• Example models: Yu_2022_igaNephropathy_mbma.R (study-arm level; centred linear effect -0.63 * (UPRO_BL - 1.82) on the drug-arm Emax of the change from baseline in urinary protein excretion, none on the placebo Emax).
• Notes: Member of the <MARKER>_BL baseline family (FERRITIN_BL, BMD_BL, HGB_BL, QTC_BL). Distinct from UACR (urine ALBUMIN-to-creatinine ratio, mg/g) and from a protein-to-creatinine ratio (UPCR, g/g): total protein includes non-albumin proteins, and a ratio is dimensionally different from a daily excretion rate, so none of the three may be substituted for another without the paper’s own conversion. A future model using baseline UPCR should register a sibling UPCR_BL rather than overload this name. The paired change-from-baseline endpoint is the uprocfb output in compartment-names.md.

UNGALCR (canonical for urine NGAL-to-creatinine ratio)

• Description: Urinary neutrophil gelatinase-associated lipocalin (NGAL) concentration divided by the urine creatinine concentration of the same spot-urine sample. NGAL is a 25 kDa lipocalin released by injured renal tubular epithelium; it rises in urine hours to days before serum creatinine moves, so it is a marker of tubular kidney injury rather than of glomerular filtration. Normalising to urine creatinine removes the dependence on urine volume / dilution, which is what makes a spot sample interpretable. Belongs to the same “damage marker, not function marker” axis as [[UACR]], and is distinct from the function markers [[CRCL]] / [[CREAT]] / [[CYSC]]; a model can carry one of each, and Downes_2023_vancomycin_full.R does exactly that (eGFR-Hoek for filtration, UNGALCR for injury), reporting a correlation of only -0.27 between the two.
• Units: ng/mg creatinine. Document the unit used in each model via covariateData[[UNGALCR]]$units; urinary biomarker ratios are also reported as ng/mmol creatinine and as ug/g creatinine in the wider literature, and the three differ by more than a decimal shift.
• Type: continuous
• Scope: specific
• Reference category: n/a. Two forms are anticipated: an UNNORMALISED exponential on the log ratio, base^log(UNGALCR), in which case there is no reference value and the intercept parameter is the value at UNGALCR = 1 ng/mg (the Downes 2023 form); or a conventional median-normalised power, (UNGALCR / ref)^exponent. Reference values observed: none (Downes 2023 uses the unnormalised form); the Downes 2023 training-group median for reference is 60.3 ng/mg creatinine.
• Source aliases:
  • uNGAL/uCr – Downes 2023 Table 2 and Table S4 footnotes (“the natural logarithm of the urinary NGAL concentration normalized to urinary creatinine”).
  • UNGAL – Downes 2023 Table S2 step 3, where the same quantity is screened in an untransformed power form (50.1/UNGAL)^TH2 before the log form was preferred. Note the column is the RATIO in both rows despite the bare UNGAL spelling.
  • LUNGAL – Downes 2023 Table S2 step 3 name for log(UNGALCR), i.e. the transformed value rather than the column itself. Store the ratio on its natural scale in this column and take the logarithm inside model().
• Example models: Downes_2023_vancomycin_full.R (founding example; ng/mg creatinine, no reference normalisation; exponential factor 0.94^log(UNGALCR) on vancomycin clearance in critically ill children, so clearance falls about 22 percent at the cohort median of 60.3 ng/mg relative to the UNGALCR = 1 intercept. Two features are worth carrying forward. First, the covariate spans more than three orders of magnitude within one small cohort – 1.4 to 2809 ng/mg in the 30-subject training group and 7.8 to 10034 in the 20-subject testing group – so a simulated cohort must draw it on the log scale, and the authors themselves flag the spread as a plausible reason the biomarker did not improve individual-level AUC estimation. Second, the unnormalised form means the intercept parameter is anchored at a value far below anything observed, so CL0 here is not interpretable as a typical clearance and must not be compared directly against the CL0 of the paper’s two reduced sibling models.).
• Notes: Scope is specific until a second model ratifies the column; promote to general at that point. The sibling urinary biomarkers Downes 2023 screened on the same footing – urinary KIM-1, urinary cystatin C and urinary osteopontin, each also creatinine-normalised and log-transformed (Table S2 step 3) – are NOT registered here because none was retained in a final model. A future paper that retains one should add a parallel UKIM1CR / UCYSCCR / UOPNCR entry rather than widening this one, since the analytes are biologically distinct injury markers. Do not confuse this column with plasma NGAL, which Downes 2023 screened separately as (95/PNGAL)^TH1 and also did not retain; a future model needing plasma NGAL should register NGAL (a plasma concentration in ng/mL) as its own canonical, because the plasma and creatinine-normalised-urine forms have different units, different dynamic ranges and different biological meaning.

SOD (canonical for serum sodium concentration)

• Description: Serum (or plasma) sodium concentration. Most commonly captured at study entry as a baseline biochemistry value, but can be time-varying when serial electrolytes are recorded. Reflects fluid / electrolyte balance; hyponatraemia is common in severe acute illness, severe malnutrition, GI losses, SIADH and adrenal insufficiency, while hypernatraemia accompanies dehydration and excessive sodium intake.
• Units: mmol/L (equivalent to mEq/L for sodium).
• Type: continuous
• Scope: general
• Reference category: n/a – used with a centered-linear-deviation form (1 + e_sod_<param> * (SOD - ref)). Reference values observed: 136 mmol/L (Thuo 2011 ciprofloxacin; cohort median in Kenyan children with severe malnutrition, normal-range lower bound).
• Source aliases:
  • Na+ / NA / SODIUM – common source-paper printed forms; renamed to canonical SOD when assembling input data. Used in Thuo_2011_ciprofloxacin.R (the paper writes “Na+ (mmol/L)” in Table 1 and structural-model equations).
• Example models: Thuo_2011_ciprofloxacin.R (linear centered-deviation effects on apparent CL and apparent Vc: 1 + 0.0368*(SOD - 136) and 1 + 0.0291*(SOD - 136); reference 136 mmol/L is the cohort median).
• Notes: General scope because serum sodium is a universally applicable serum-electrolyte covariate. Reference value is paper-specific (cohort median); future models should document their own reference in covariateData[[SOD]]$notes. Distinct from any “sodium content of dosed formulation” concept (e.g., sodium-rich oral rehydration solution) – that would warrant a separate canonical (DOSE_NA_MGML, etc.) if a future model retains it.

POT (canonical for serum potassium concentration)

• Description: Serum (or plasma) potassium concentration. Most commonly captured at study entry as a baseline biochemistry value, but can be time-varying when serial electrolytes are recorded. Central to mineralocorticoid-receptor-antagonist, potassium-sparing-diuretic and RAAS-inhibitor programmes, where hyperkalaemia is the dose-limiting toxicity and drives titration and discontinuation rules; KDIGO and ESC hyperkalaemia thresholds are 5.5 and 6.0 mmol/L. Direct sibling of SOD (serum sodium). Distinct from the serum_k COMPARTMENT, which is the modelled potassium time course when potassium is the PD output rather than an explanatory covariate.
• Units: mmol/L (equivalent to mEq/L for potassium).
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (POT / ref)^exponent or linear centring (1 + slope * (POT - ref)). Reference values observed: 4.4 mmol/L (Goulooze 2022 FIDELIO-DKD, the cohort median).
• Source aliases:
  • K0 – Goulooze 2022 convention for observed baseline serum potassium (the 0 suffix denotes the baseline / pre-randomisation value).
  • K, K+, POTAS, SERUM_K – common source-paper printed forms.
• Example models: Goulooze_2022_finerenone_egfr.R (power scaling (POT / 4.4)^-1.38 on DSLOPE, the slope of the acute finerenone effect on eGFR; ESM Table S2 theta_K0,DSLOPE = -1.38, reproducing the paper’s Table 2 range of +31.9% to -18.4% over the 3.6-5.1 mmol/L 5th-95th percentile band).
• Notes: General scope because serum potassium is a universally applicable serum-electrolyte covariate, matching the scope granted to SOD. Reference value is paper-specific (cohort median); future models should document their own reference in covariateData[[POT]]$notes.

Renal-replacement-therapy (RRT) family – section-header policy

All RRT-related canonicals follow the RRT_<MODALITY>_<KIND> shape, where MODALITY is the specific RRT type (HEMODIAL for intermittent hemodialysis, CRRT for continuous / extended modalities, PERIT_DIAL for peritoneal dialysis) and KIND is STATUS for subject-level treatment-status indicators (time-fixed within the analysis window) or ACTIVE for per-time-point session gates (time-varying within subject, 1 only during an active session). KIND may also name a measured circuit quantity rather than an indicator, in which case it spells the quantity out (EFFLUENT_FLOW); the shape and the modality slot are unchanged, and such a member is continuous rather than binary. The 2026-06-19 canonical-register standardization audit renamed the prior HEMODIAL, HEMODIALYSIS, and CRRT_STATUS canonicals into this family to make the modality + kind contrast explicit at the column name; pre-2026-06-19 names are preserved as source_aliases so existing data CSVs continue to work for one release cycle.

RRT_HEMODIAL_STATUS (canonical for intermittent-hemodialysis treatment-status indicator)

• Description: 1 = the subject was undergoing intermittent hemodialysis during the modeled period; 0 = no intermittent hemodialysis. Treatment-status flag rather than a measured renal-function value; used as a multiplicative covariate on PK parameters that change with chronic dialysis (typically CL and Vc – intermittent hemodialysis decreases vancomycin-class CL and reduces interstitial volume overload, lowering Vc). Per-subject indicator in the source data; in Goti 2018 it is treated as time-fixed at the subject level (the cohort either was or was not receiving intermittent hemodialysis during the admission), and Goti 2018 explicitly notes that actual hemodialysis-session timing was not used because of documentation limitations.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no intermittent hemodialysis).
• Source aliases:
  • HEMODIAL – prior canonical name (pre-2026-06-19 RRT__ standardization).
  • DIAL – used in Goti_2018_vancomycin.R (binary indicator on CL and Vc in a 2-compartment vancomycin popPK model). Goti 2018 Methods notes the indicator was created for the routine-TDM cohort (n = 336 hemodialysis subjects of 1812 total) and that all hemodialysis procedures were intermittent and used high-flux membranes.
  • DIT1 – used in Knebel_2008_epoetinDelta.R (the first of two dialysis-modality switches in a three-level covariate; DIT1 = 1 selects the hemodialysis stratum, with predialysis rather than “no hemodialysis” as the reference – see the Notes below).
• Example models: Goti_2018_vancomycin.R (multiplicative factors on CL and Vc: 0.7^RRT_HEMODIAL_STATUS on CL and 0.5^RRT_HEMODIAL_STATUS on Vc, so dialysis subjects have 30% lower CL and 50% lower central volume than non-dialysis subjects), VanWart_2025_telavancin.R (used purely as a gate, not as an effect in its own right: it restricts the additive T_POST_HEMODIAL > 48 h central-volume increase to the 8 chronic-kidney-disease-stage-5 subjects, while the dialysis CLEARANCE arm of the same model is gated by the time-varying RRT_HEMODIAL_ACTIVE instead – the clearest available illustration of the STATUS-versus-ACTIVE split), Knebel_2008_epoetinDelta.R (exponentiated-switch factor 4.53^RRT_HEMODIAL_STATUS on central volume only, paired with PERIT_DIAL against a predialysis reference in a pediatric CKD erythropoietin popPK model), Li_2019_avibactam.R (selects an ABSOLUTE on-dialysis clearance of 20.8 L/h in place of the reference-clearance construction, Li 2019 Table 2 theta6; mutually exclusive in effect with that model’s RENALIMP_ESRD term, which supplies the much lower off-dialysis clearance).
• Notes: Specific to intermittent hemodialysis (IHD). Distinct from peritoneal dialysis (PD) and from continuous renal replacement therapy (CRRT), each of which has different drug-extraction kinetics and would warrant its own canonical (RRT_PERIT_DIAL_STATUS, RRT_CRRT_STATUS) if a future paper retains them as covariates. Three-level dialysis-modality cohorts: when a paper’s cohort spans predialysis, hemodialysis, and peritoneal dialysis and takes predialysis as the reference, encode the two non-reference strata as RRT_HEMODIAL_STATUS and PERIT_DIAL together, with both indicators 0 for the predialysis stratum (Knebel_2008_epoetinDelta.R). The 1-levels still mean exactly what the two canonical definitions say; what shifts is the reference the estimated effects are read against, which is a per-paper fact and must be documented in each model’s covariateData[[...]]$reference_category and $notes rather than inferred from the register. This differs from the two-level chronic-dialysis case (Takama_2007_darbepoetin.R) where the two indicators are perfectly anti-correlated and only one is needed. Goti 2018 treats RRT_HEMODIAL_STATUS as time-fixed per subject because session-level dialysis timing was not reliably documented in the source EHR data; a future paper that resolves drug clearance during versus between dialysis sessions would use a time-varying form (RRT_HEMODIAL_ACTIVE) or a separate per-session covariate. When pairing RRT_HEMODIAL_STATUS with CRCL, note that the Cockcroft-Gault CRCL of an anuric hemodialysis patient is by convention very low or set per institution to a small floor value (Goti 2018 truncated CRCL > 150 mL/min to 150 mL/min and corrected SCr < 1 mg/dL in elderly subjects); residual renal function in hemodialysis subjects is highly variable and the dialysis indicator captures the bulk PK shift on top of the CRCL covariate.

RRT_CRRT_STATUS (canonical for continuous / extended renal-replacement-therapy treatment-status indicator)

• Description: 1 = the subject was undergoing a continuous or extended (long-session) extracorporeal renal-replacement-therapy modality during the modeled period (continuous venovenous hemofiltration CVVH / CVVHF, continuous venovenous hemodiafiltration CVVHDF, sustained low-efficiency dialysis SLED, extended daily diafiltration EDD-f, or similar slow-clearance / long-duration extracorporeal therapies); 0 = no such therapy. Treatment-status flag rather than a measured renal-function value; used as a multiplicative or piecewise covariate on PK parameters that change when slow extracorporeal solute removal is active (typically CL). Per-subject indicator in the source data; in Shekar 2014 it is treated as time-fixed at the subject level (the cohort either was or was not receiving RRT during the entire PK sampling period; all RRT patients were on CVVH or EDD-f continuously / daily during sampling) – the indicator captures the subject-level RRT status, not session-level on/off timing.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no continuous / extended RRT).
• Source aliases:
  • CRRT_STATUS – prior canonical name (pre-2026-06-19 RRT__ standardization).
  • RRT – used in Shekar_2014_meropenem.R (binary indicator selecting between the RRT-fixed-CL term and the CrCL-driven non-RRT CL term in a 2-compartment meropenem popPK model). Shekar 2014 Methods describes the RRT cohort as mixed CVVH (control RRT subjects, true CRRT) and EDD-f (ECMO RRT subjects, extended daily diafiltration; pharmacokinetically CRRT-like for slow-clearance solutes such as meropenem) and the model treats the modalities as a single binary covariate without distinguishing them.
• Example models: Shekar_2014_meropenem.R (piecewise CL: TVCL = exp(lcl) * RRT_CRRT_STATUS + e_crcl_cl * CRCL_in_Lh * (1 - RRT_CRRT_STATUS), with CRCL in raw Cockcroft-Gault mL/min converted to L/h inside model(); 5/11 ECMO patients and 5/10 controls were on RRT), Wi_2017_teicoplanin.R (power-model multiplier on the peripheral volume, vp <- exp(lvp) * 1.50^RRT_CRRT_STATUS; 5/10 subjects on CVVHDF via Prismaflex), Sun_2025_teicoplanin.R (exponential effect on the CENTRAL volume, exp(-0.71 * RRT_CRRT_STATUS), i.e. CRRT reduces V by 50.8%; 20/86 sepsis patients, modality mix 14 CVVH / 5 CVVHD / 1 both), WerumeusBuning_2021_ceftazidime.R (the same piecewise switch as Shekar 2014 on a one-compartment ceftazidime model: cl = cl_off * (1 - RRT_CRRT_STATUS) + exp(lcl_cvvh) * RRT_CRRT_STATUS, where the on-CVVH clearance is a single fixed 2.9 L/h carrying NO between-subject variability and the off-CVVH branch carries the eGFR and comorbidity effects and the eta; 20/96 critically ill adults on CVVH), Lin_2021_vancomycin.R (additive-linear term in an additive-linear CL equation, +0.41 L/h * CRRT-S with CRRT-S = 2 - RRT_CRRT_STATUS, i.e. CRRT LOWERS vancomycin CL by 0.41 L/h in adult ICU patients; the paper notes the direction contradicts earlier reports).
• Direction is not fixed by the canonical. Most published CRRT effects act on clearance and increase it, but the registry now holds examples going both ways and on either volume: Wi_2017_teicoplanin.R has CRRT raising the peripheral volume 1.5-fold, while Sun_2025_teicoplanin.R has it lowering the central volume by half, with no retained effect on clearance at all. Sun 2025 attributes the reduction to fluid overload being the indication for starting CRRT, so that CRRT-treated subjects had an expanded interstitial volume corrected by the therapy. Do not “correct” a source paper’s sign against an expected direction; transcribe it and record the authors’ stated mechanism in covariateData[[RRT_CRRT_STATUS]]$notes.
• Notes: Distinct from RRT_HEMODIAL_STATUS (intermittent hemodialysis IHD only) and from RRT_HEMODIAL_ACTIVE (per-time-point session gate in within-subject time-varying dialysis-clearance models such as Liesenfeld 2013 dabigatran). Anticipated as a future canonical RRT_PERIT_DIAL_STATUS for peritoneal dialysis. Shekar 2014 ratification uses a mixed CVVH + EDD-f cohort because the source paper treats them identically as a single binary RRT covariate; a future paper that retains modality as a separate covariate (e.g. CVVH vs SLED vs CVVHDF) would either reuse RRT_CRRT_STATUS with finer per-modality columns layered on top, or warrant its own modality-specific canonical (RRT_CVVH_STATUS, RRT_SLED_STATUS, etc.). When pairing RRT_CRRT_STATUS with CRCL, note that Cockcroft-Gault CrCL is conventionally not defined / not reported for RRT-dependent subjects; Shekar 2014 records CrCL only for non-RRT subjects and the model formula switches off the CrCL term when RRT_CRRT_STATUS = 1.

RRT_CVVHDF_STATUS (canonical for the continuous-venovenous-hemodiafiltration modality indicator within a continuous-RRT cohort)

• Description: 1 = the subject’s continuous renal-replacement modality is continuous venovenous hemodiafiltration (CVVHDF), which combines a diffusive dialysate stream with a convective ultrafiltrate / substitution stream; 0 = the subject is on the reference continuous modality, continuous venovenous hemodialysis (CVVHD), which is purely diffusive. This is a modality discriminator layered on top of RRT_CRRT_STATUS / RRT_CRRT_ACTIVE, not a treatment-status flag in its own right: it says which continuous modality, never whether a continuous modality is running. It is therefore meaningful only while the circuit is active, and the pair (RRT_CRRT_ACTIVE, RRT_CVVHDF_STATUS) is what distinguishes off-circuit from CVVHD – the modality indicator alone cannot. Use it when a paper estimates a separate extracorporeal clearance per modality over a cohort that shares one body-PK model; a paper that pools modalities under a single binary needs RRT_CRRT_STATUS instead, and a paper that derives the extracorporeal clearance from circuit flows needs RRT_CRRT_EFFLUENT_FLOW (which CVVHDF and CVVHD both use, the CVVHDF effluent simply being the larger by the ultrafiltration rate).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (CVVHD, the purely diffusive continuous modality).
• Source aliases:
  • Table footnote markers distinguishing modality per subject (Broeker 2018 Table 1 footnotes a = CVVHD, b = CVVHDF); papers frequently encode the modality only this way rather than as a named data column.
  • CVVHDF_STATUS – the pre-RRT_<MODALITY>_<KIND> form of this name, as first proposed when the concept was ratified.
  • MODE, CRRT_MODE, RRT_MODE – multi-level modality columns appearing in the critical-care CRRT popPK literature. Decompose a multi-level modality column into one binary per non-reference modality rather than carrying an integer code.
• Example models: Broeker_2018_tigecycline.R (founding example; 11 critically ill adults with acute kidney injury, 8 on CVVHD and 3 on CVVHDF, sharing one two-compartment body-PK model while Table 2 estimates two separate dialysis clearances – 1.69 L/h for CVVHD and 2.71 L/h for CVVHDF. The column selects between them, cl_hemodialysis <- RRT_CRRT_ACTIVE * ((1 - RRT_CVVHDF_STATUS) * exp(lcl_hemodialysis_cvvhd + eta) + RRT_CVVHDF_STATUS * exp(lcl_hemodialysis_cvvhdf)). Only the reference CVVHD arm carries IIV; the CVVHDF arm’s IIV “tended to zero during estimation” at n = 3).
• Notes: Ratified as the first member of the anticipated modality family that the RRT_CRRT_STATUS entry names (“a future paper that retains modality as a separate covariate (e.g. CVVH vs SLED vs CVVHDF) would … warrant its own modality-specific canonical (RRT_CVVH_STATUS, RRT_SLED_STATUS, etc.)”); the maintainers ruled 2026-05-30 for a single model file carrying one new binary modality canonical, CVVHD as reference. Siblings RRT_CVVH_STATUS and RRT_SLED_STATUS remain anticipated-but-unratified; add each on the same pattern when a paper retains it. Mutual exclusivity across a future multi-modality set is a convention, not a data constraint – if a cohort spans three or more continuous modalities, carry one binary per non-reference modality with all of them 0 denoting the reference, exactly as FILT_SA_MED / FILT_SA_LARGE handle three filter sizes in ButraguenoLaiseca_2022_piperacillin.R, and record the per-paper reference modality in each model’s covariateData[[...]]$reference_category. Which modality is the reference is a per-paper fact: Broeker 2018 takes CVVHD because it carries 8 of its 11 subjects, but a cohort dominated by CVVHDF would reasonably invert that, and the canonical’s 1-level always means CVVHDF regardless.

RRT_CRRT_EFFLUENT_FLOW (canonical for total effluent flow rate of a continuous-RRT circuit)

• Description: Total effluent flow rate leaving a continuous / extended renal-replacement circuit – the sum of spent dialysate, substitution (replacement) fluid, and net ultrafiltration. This is the standard clinical “CRRT dose” metric, normally prescribed weight-normalized in mL/kg/h and recorded absolutely in mL/h. It is the circuit OUTPUT, and therefore distinct from the circuit INPUTS BFR (blood flow into the filter) and DFR (dialysate flow into the filter): in a hemodiafiltration modality the effluent exceeds the dialysate inflow by the ultrafiltration and substitution rates. Because solute removal by a continuous modality is effluent-flow-limited for small, minimally protein-bound drugs, the effluent flow rate is the natural covariate on an extracorporeal clearance arm; the sieving / saturation coefficient that converts it into a clearance is absorbed into the estimated clearance parameter (lcl_crrt) rather than being applied to this column.
• Units: mL/h
• Type: continuous
• Scope: general
• Reference category: n/a – enters as a through-origin linear ratio (RRT_CRRT_EFFLUENT_FLOW / ref) or a power-centered ratio on the CRRT clearance arm, and is meaningful only while RRT_CRRT_ACTIVE = 1. Reference values observed: 2000 mL/h (2 L/h; Zurawska 2026, the value the CRRT clearance estimate is standardized to).
• Source aliases:
  • FLOW – Zurawska 2026 Eq. 6 notation.
  • phi_Effl, phi_Effluent – Butragueno-Laiseca 2022 supplementary material eq. 5 and the companion design-optimisation paper (Garcia-Orueta 2026, doi:10.1007/s40262-025-01594-1); same orientation, no transformation.
  • Effluent flow, Q_eff, Q_effluent – variant abbreviations used across the critical-care CRRT popPK literature.
  • EFFLUENT_FLOW, Q_EFFLUENT, EFR, QEFFL – printed forms appearing in adjacent critical-care CRRT popPK literature. None was adopted as the canonical: the RRT section-header policy requires the RRT_<MODALITY>_<KIND> shape (maintainers’ ruling 2026-08-28).
  • Weight-normalized prescriptions (mL/kg/h) are NOT this column – multiply by body weight on ingestion and record the conversion in covariateData[[RRT_CRRT_EFFLUENT_FLOW]]$notes.
• Example models: Zurawska_2026_piperacillin.R (through-origin linear ratio on the estimated CRRT clearance arm, cl_crrt <- exp(lcl_crrt + etalcl_crrt) * (RRT_CRRT_EFFLUENT_FLOW / 2000) * RRT_CRRT_ACTIVE, so the 3.4 L/h population CRRT clearance is the value at a 2 L/h effluent flow; the cohort mean effluent was 2.75 +/- 1.1 L/h, i.e. 32 +/- 7.8 mL/kg/h), ButraguenoLaiseca_2024_meropenem.R (sibling model from the same group; supplies the denominator of the effluent-concentration observable of a five-matrix meropenem model in critically ill children on CKRT, Ceffluent <- cl_hemodialysis * Cc / (RRT_CRRT_EFFLUENT_FLOW / 1000), the division by 1000 converting mL/h to the L/h in which the clearance is expressed. Cohort total ultrafiltration rate median 1,253 mL/h, IQR 335-1,870 (Table 2); per-filter mean effluent flows 44 / 60 / 46 mL/kg/h for the low / medium / high surface-area filters (Table 3)), ButraguenoLaiseca_2022_piperacillin.R (founding example; critically ill children on continuous venovenous haemodiafiltration. Normalises the effluent observable Ceffluent <- cl_hemodialysis * Cc / (RRT_CRRT_EFFLUENT_FLOW / 1000) per supplementary eq. 5. Supplementary Table 3 reports the flow per filter surface area as 78 (SD 24) mL/kg/h for the 0.2 m^2 filter, 54 (5) for 0.6 m^2 and 60 (15) for 1.2 m^2; the absolute values used in the validation vignette are these multiplied by the weight-band body weight).
• Notes: Distinct from QBL / QEFF, which are extracorporeal CLEARANCES in L/h – Zhang_2025_fluconazole.R supplies QEFF as effluent flow ALREADY multiplied by a sieving coefficient, whereas this column is the raw prescribed flow with the sieving absorbed into the estimated parameter. Distinct from Q_CVVH, which is scoped to a CVVH filter placed parallel to an ECMO circuit (Ahsman 2010 cefotaxime). Distinct from URINE_FLOW (native urine output). Pair with RRT_CRRT_ACTIVE as the on/off gate; a model that instead derives the extracorporeal clearance from circuit flows via the Michaels equation should use BFR / DFR and an estimated lkoa rather than this column. Note that BFR and DFR are canonically mL/min because intermittent-hemodialysis circuits are prescribed in mL/min, whereas continuous modalities are prescribed in mL/h or mL/kg/h – each canonical follows its own modality’s clinical convention, so converting between them requires care.

RRT_PERIT_DIAL_FILL_VOLUME (canonical for the instilled fill volume of a peritoneal-dialysis exchange)

• Description: Volume of peritoneal dialysate instilled into the peritoneal cavity at the start of a dwell – the “fill volume” or “exchange volume” of a CAPD / APD exchange. This is the circuit INPUT (VOL IN in a peritoneal-dialysis model schematic) and is distinct from the drained volume at the end of the dwell (VOL OUT), which exceeds it by net ultrafiltration. Its usual role is not a covariate effect on a structural parameter at all: for a drug administered intraperitoneally the peritoneal cavity is a compartment whose volume is a measured prescription quantity rather than something the data can identify, so the fill volume IS the peritoneal compartment volume and enters model() directly as v1 <- RRT_PERIT_DIAL_FILL_VOLUME. Time-varying in principle (a patient’s prescription can change between exchanges) but time-fixed per subject in the founding example.
• Units: L
• Type: continuous
• Scope: general
• Reference category: n/a – used as an absolute volume, not a centered or normalized covariate effect. No reference value applies.
• Source aliases:
  • V1 – Hartinger 2023 Table 2, whose entry for the peritoneal compartment volume reads FIXED * with the footnote “Fixed to the actual volume of peritoneal solution used (range 1-2 L)”.
  • VOL IN – Hartinger 2023 Figure 1 schematic label for the instilled volume.
  • exchange volume, fill volume, dwell volume – prose forms across the peritoneal-dialysis literature. Same orientation, no transformation; convert mL to L on ingestion.
• Example models: Hartinger_2023_vancomycin.R (founding example; intraperitoneal vancomycin in adults on CAPD treated for PD-associated peritonitis. The column supplies the volume of the peritoneum compartment directly – v1 <- RRT_PERIT_DIAL_FILL_VOLUME, Cip <- peritoneum / v1 – and no IIV was estimated on it, the source having excluded the peritoneal volume from IIV testing precisely because it is a measured quantity. Observed prescriptions: 2000 mL in 56% of patients, 1500 mL in 39%, 1200 mL and 1000 mL in one patient each; the paper’s simulations use 2 L at 75 and 100 kg and 1.5 L at 50 kg. The value is verifiable against the paper’s own figure axes – a 1500 mg intraperitoneal loading dose reads exactly 750 mg/L on the Figure 4 peritoneal-concentration axis.).
• Notes: Family-conforming member of the RRT_<MODALITY>_<KIND> shape defined by the section-header policy above, with PERIT_DIAL as the modality and a spelled-out measured circuit quantity as the kind – the same construction as RRT_CRRT_EFFLUENT_FLOW, and continuous rather than binary for the same reason. Distinct from the binary PERIT_DIAL (a treatment-modality indicator answering whether a subject is on peritoneal dialysis) and from the anticipated RRT_PERIT_DIAL_STATUS; this column answers how much fluid an exchange instils. A model that needs the drained volume as well – e.g. one that resolves ultrafiltration, or that computes drug removal as Cip * VOL_OUT rather than emptying the compartment – should register a sibling RRT_PERIT_DIAL_DRAIN_VOLUME rather than overload this entry, since the two differ by the ultrafiltrate and are not interchangeable. Note that the end-of-dwell drain and the next instillation belong in the EVENT table (a replacement event zeroing the peritoneal compartment followed by the next dose), not in the ODEs; this column sizes the compartment, it does not empty it.

ALB (canonical for serum albumin)

• Description: Serum albumin concentration.
• Units: g/L (SI; canonical as of 2026-06-19). US-convention papers report in g/dL; convert via ALB_gL = ALB_gdL * 10 on data ingestion. Models calibrated to g/dL values must apply an inline conversion in model() (e.g., alb_gdL <- ALB * 0.1 # SI g/L -> US-convention g/dL) so the structural coefficients stay aligned with their original calibration.
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (ALB / ref)^exponent.
• Source aliases:
  • BALB (baseline albumin) – used in Zhou_2021_belimumab.R. Maps directly to ALB; baseline-vs-time-varying status documented in per-model notes.
  • HSA (human serum albumin) – used in Fauchet_2015_lopinavir_unbound.R where the column name follows the paper’s protein-binding-equation notation distinguishing serum albumin from alpha-1 acid glycoprotein. Maps directly to canonical ALB; no value transformation. Reported in g/L; converted to umol/L inside model() via molecular weight 66500 g/mol for the K_HSA linear-binding term.
• Example models: Fasanmade_2009_infliximab.R (g/dL, reference 4.1), Thakre_2022_risankizumab.R (g/L, reference 45), Chua_2025_mirikizumab.R, Moein_2022_etrolizumab.R, Tiraboschi_2025_amlitelimab.R, Yamada_2025_zolbetuximab.R, Li_2019_abatacept.R (g/dL, reference 4.0; the Li 2019 Methods states ‘mg/dL’ which is a publication typo – see the model’s covariateData[[ALB]]$notes), Quartino_2019_trastuzumab.R (g/dL, reference 4; source column ALBU; negative exponent -0.998 on linear CL), Wang_2020_ontamalimab.R (g/L, reference 39), Zhou_2021_belimumab.R (g/L, reference 40; baseline-only, source column BALB), Okada_2025_rocatinlimab.R (g/L, reference 44; source column ALBU; power exponent -1.30 on linear CL), Xu_2020_daratumumab.R (g/L, reference 37.0; power exponent -1.149 on linear CL), Struemper_2017_belimumab.R (g/L, reference 41; baseline-only, source column BALB; power exponent -0.736 on linear CL), Fauchet_2015_lopinavir_unbound.R (g/L; source column HSA; enters the saturable-binding submodel via a linear K_HSA * [ALB] * Cunbound term with K_HSA = 0.036 L/umol, not as a power scaling on CL), Roepcke_2023_rezafungin.R (g/dL, reference 3.2; power exponent -0.708 on the shared peripheral volume V23 rather than on CL; model() applies the inline alb_gdL <- ALB * 0.1 SI-to-US conversion required by this entry’s Units note. Roepcke 2023 Table 2 prints the exponent magnitude without a sign and the footnote-a equation supplies the minus; the negative orientation is corroborated by Table 3, where the low-albumin patient cohort has the larger steady-state volume of distribution), Zhang_2025_abemaciclib_qsp.R (g/L, reference 31 = the source paper’s 3.1 g/dL mean in breast-cancer patients; ALB does NOT scale a clearance or a volume here – it scales the fraction unbound of four analytes through the paper’s Eq 2, fup = 1/(1 + B*ALB), so the effect is on the free concentration driving target engagement rather than on a disposition parameter), Bi_2018_testosteroneCypionate.R (g/dL calibration with the mandated inline ALB * 0.1 conversion; enters ONLY as the within-subject delta (ALB - ALB_BASE) under the Wahlby 2004 decomposition, in a LINEAR multiplier (1 + -0.27 * (dALB_gdL + 0.2)) on the apparent central volume, with the baseline half carried by the paired ALB_BASE column. This is the register’s clearest case of ALB appearing in a model in its delta role only – a consumer supplying a constant albumin column contributes nothing through this term, which is the intended behaviour, not a bug), Jiang_2021_ivosidenib.R (g/L; enters ONLY through the within-subject albumin ratio ALB / ALB_BASE, with power exponents 0.99 on CL/F and 1.1 on Vc/F; the baseline half is carried by ALB_BASE), KhanAsa_2020_voriconazole.R (g/dL calibration with the inline ALB * 0.1 conversion; LINEAR term 1 + 0.249 * (ALB_gdL - 3.2) on voriconazole CL/F, centred on the cohort median).
• Notes: Canonical units standardized to g/L (SI) on 2026-06-19. The per-model covariateData[[ALB]]$units field is load-bearing for model-file source-trace; effect-coefficient magnitude is meaningless without the unit. Per-model conversion to US-convention g/dL (where the structural coefficients were calibrated) is required in model() via an inline alb_gdL <- ALB * 0.1 line - see the per-model notes for the conversion factor used. Conversion factor: 1 g/dL = 10 g/L.

ALB_LT35 (canonical for serum albumin below 35 g/L (3.5 g/dL) indicator)

• Description: Binary indicator dichotomising serum albumin at 35 g/L (equivalently 3.5 g/dL), 1 = albumin below 35 g/L (hypoalbuminaemia; the poor-prognosis category), 0 = albumin at or above 35 g/L. Time-fixed per subject when derived from a baseline albumin measurement. Distinct from the continuous canonical ALB (g/L), which captures albumin magnitude and is used with power or linear-centred scaling; ALB_LT35 is the dichotomised form used by source papers whose published coefficient is defined only on the binary contrast, so the continuous value cannot reconstruct the reported effect.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (serum albumin >= 35 g/L, i.e. >= 3.5 g/dL – normal albumin, the better-prognosis category).
• Source aliases:
  • Albumin < 3.5 (g/dL) – Claret 2014 (J Clin Pharmacol 54(3):253-257) Table 2 covariate row label; the source study reports albumin in US-convention g/dL and dichotomises at 3.5 g/dL, which is 35 g/L in the canonical SI unit. Coded 1/0.
  • ALBLOW / HYPOALB – common NONMEM column-name variants for a baseline hypoalbuminaemia flag dichotomised at the same clinical threshold.
• Example models: Claret_2014_DCR_OS_nsclc.R (additive effect on the log-normal AFT location parameter of overall survival: mu_os += e_alb_lt35_mu_os * ALB_LT35 with e_alb_lt35_mu_os = -0.270, i.e. median OS is exp(-0.270) = 0.763 times shorter – 23.7% shorter – in patients with baseline albumin below 3.5 g/dL).
• Notes: Derive from a continuous albumin column with ALB_LT35 = as.integer(ALB < 35) when ALB is in canonical SI g/L, or as.integer(ALB_gdL < 3.5) when the source column is in US-convention g/dL. Register both ALB and ALB_LT35 in a dataset when downstream users may want either form; a model file should declare only the one its published coefficient is defined on. 3.5 g/dL is the conventional clinical-chemistry lower limit of normal for serum albumin and is the threshold used across the oncology-prognosis literature, so this canonical is expected to be reused rather than re-thresholded; a paper dichotomising at a different cut-point should register a sibling (ALB_LT30, ALB_LT40) following the AGE_GE70 / AGE_GT65 and MET_GE3 / MET_GE4 precedent for co-existing thresholds on one underlying continuous variable.

ALB_BASE (canonical for per-subject baseline serum albumin (time-fixed))

• Description: Per-subject baseline serum albumin, time-fixed. Used when a source paper enters the subject’s baseline albumin as a constant column alongside the time-varying ALB column under Wahlby 2004’s extended covariate-model decomposition (Br J Clin Pharmacol 2004;58(4):367-377). The within-subject delta is computed in model as (ALB - ALB_BASE).
• Units: g/L (SI), matching the paired ALB column. US-convention papers report in g/dL; convert on ingestion via ALB_BASE_gL = ALB_BASE_gdL * 10 and apply the inline alb_base_gdL <- ALB_BASE * 0.1 conversion in model() when the published coefficients were calibrated in g/dL. Document per-model via covariateData[[ALB_BASE]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – used either as the ratio (ALB_BASE / ref) in a power effect or as (ALB_BASE - ref) in a linear or linear-exp effect. Reference value observed: 4.60 g/dL = 46.0 g/L (Bi 2018 control-stream constant).
• Source aliases:
  • ALBUMIN (baseline albumin, paired with a separate CALBUMIN change-from-baseline column) – Bi 2018 $INPUT convention; used in Bi_2018_testosteroneCypionate.R.
• Example models: Bi_2018_testosteroneCypionate.R (g/dL calibration, reference 4.60 g/dL; power exponent -1.55 on the apparent central volume of a depot-testosterone popPK model, paired with the delta effect -0.27 dL/g on (ALB - ALB_BASE) centred at -0.2 g/dL. The centring constants appear only in the Supplementary S6 control stream, not in the main article), Jiang_2021_ivosidenib.R (g/L, reference 37 g/L; power exponents 0.82 on CL/F and 0.73 on Vc/F, paired with the albumin ratio ALB / ALB_BASE (exponents 0.99 and 1.1) – the ratio form of the same baseline-plus-within-subject decomposition, and a second paper using ALB_BASE with consistent semantics), Nguyen_2022_cemiplimab.R (g/L, reference 38.6 g/L; power exponent -0.217 on the central volume V1, alongside a separate time-varying ALB column carrying exponent -1.11 on CL; the paper keeps the baseline and time-varying albumin columns separate without a delta decomposition).
• Notes: Specific scope because the BCOV/DCOV decomposition is a paper-defined modelling choice. Promote to general if a second paper ratifies the column with consistent semantics. Distinct from time-fixed adult-cohort ALB (where the per-subject albumin is constant merely by data design); use ALB_BASE only when the source paper deliberately separates the baseline snapshot from a time-varying albumin column. Also distinct from ALB_LT35, which is a dichotomised prognostic indicator rather than a methodologic baseline-snapshot column.

TPRO (canonical for total serum protein)

• Description: Total serum protein concentration (sum of albumin + globulins; baseline or time-varying).
• Units: g/L (SI; canonical as of 2026-06-19). US-convention papers report in g/dL; convert via TPRO_gL = TPRO_gdL * 10 on data ingestion. Models calibrated to g/dL must apply an inline conversion in model (e.g., tpro_gdL <- TPRO * 0.1). 1 g/dL = 10 g/L.
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (TPRO / ref)^exponent. Reference value observed: 74 g/L (Frey 2010 pooled-cohort median).
• Source aliases:
  • PROT – Frey 2010 abbreviation in the final-model equation.
  • TP – common clinical-chemistry abbreviation.
• Example models: Frey_2010_tocilizumab.R (g/L, reference 74; exponent -1.1 on V1), Hahn_2019_sufentanil.R (g/L, reference 45 = the source paper’s 4.5 g/dL cohort median; exponent +2.46 on the PERIPHERAL volume V2 rather than on V1 or a clearance. The positive orientation is unusual – higher total protein predicts a larger peripheral volume, the opposite sign to Frey 2010 – and Hahn 2019’s Discussion acknowledges the discrepancy explicitly, proposing that sufentanil binding is governed mainly by alpha-1 acid glycoprotein rather than by total protein, and that low total protein may instead mark impaired hepatic function. The same Discussion cautions that V2 shrinkage was 33%, so the effect should be treated as observational), Jeong_2020_tiropramide.R (paper in g/dL; the model converts inline via TPRO / 10 and applies a LINEAR centred form 1 + (TP - 7.6) * slope on both CL/F (-0.804 per g/dL) and V/F (-1.049 per g/dL), the first linear rather than power use of this column).
• Notes: Distinct from ALB (serum albumin, the largest single component of total protein). Frey 2010 retains both TPRO and ALB on V1 as separate covariates with opposite signs (TPRO negative, ALB positive) and notes there is no clear mechanistic explanation; the joint effect may reflect serum-volume modifications. TPRO ratified canonically on 2026-04-28 alongside the Frey 2010 extraction.

FU (canonical for fraction unbound of parent drug in plasma supplied as a data column)

• Description: Fraction of the parent drug unbound in plasma, supplied to the model as a data column rather than as an internal parameter. Two established usages: (a) an individual subject’s measured fraction unbound, typically determined by ultrafiltration of a per-subject ex-vivo plasma sample (occasionally by equilibrium dialysis) – the founding Hamren 2008 usage; and (b) a scenario-level fraction unbound that a mechanistic / PBPK model varies across simulated patient types, where the paper assigns one fu value per scenario rather than per subject – the Abdullah-Koolmees 2024 usage. In both cases the value is an input the user supplies, not a quantity the model derives. Distinct from binding-protein concentrations (ALB / AAG / TPRO), which would derive fu indirectly through a binding equation – FU is the fraction itself.
• Units: % (percent; e.g., 0.1 = 0.1% fraction unbound = 0.001 expressed as a fraction) OR unitless fraction, depending on the source paper’s convention. The per-model covariateData[[FU]]$units field documents which, and MUST be set; effect-coefficient magnitudes and Kp scalings are meaningless without the unit.
• Type: continuous
• Scope: general
• Reference category: n/a – used with linear centering (1 + slope * (FU - ref)), with power scaling (FU / ref)^exponent, or directly as a multiplicative physiological factor (PBPK Kp = Kpu * FU and unbound-drug clearance terms). Reference values observed: 0.1% (Hamren 2008 tesaglitazar cohort median imputed for three subjects with missing fu measurements; range 0.06-0.2%); 0.42 unitless (Abdullah-Koolmees 2024 voriconazole, healthy / DDI scenarios at plasma albumin 40 g/L, with 0.49 in the ICU scenarios at albumin 30 g/L).
• Source aliases:
  • f_u – Hamren 2008 paper notation (italicised fu).
  • FRAC_UNBOUND – common NONMEM column-name form when the value is reported as a unitless fraction.
  • fup – Abdullah-Koolmees 2024 supplement R source (fup, fup_icu, fup_ddi); unitless fraction, same orientation as the canonical.
• Example models: Hamren_2008_tesaglitazar.R (%, reference 0.1%; linear centered effect 555 %/unit fu on the parent metabolic clearance CLmt – per-subject fu measurements at day 42 by ultrafiltration drive the realised CLmt range from approximately 1.5 to 3.0 L/h across the cohort fu range 0.06-0.2%; founding example), AbdullahKoolmees_2024_voriconazole_pbpk.R (unitless fraction, scenario-level; enters twice in a whole-body PBPK – once converting every tissue Kpu to a tissue:plasma Kp via Kp = Kpu * FU, and once selecting the unbound drug presented to saturable hepatic CYP metabolism and to renal clearance), Panday_2025_sert_tremor_cavg_mbma.R, Panday_2025_sert_tremor_cmax_mbma.R (unitless fraction, per-DRUG rather than per-subject: the human plasma fraction unbound of each of 20 SERT inhibitors, 0.02 for sertraline to 0.8 for tramadol and tapentadol, converting the reported total plasma concentration to an unbound one in Eq. S1).
• Notes: Rare in popPK literature – only papers that do per-subject ultrafiltration measurements report the per-subject form. Most popPK papers either (a) assume a single fixed population-typical fu, in which case use a parameter-side fu paper-named parameter rather than this covariate, or (b) parameterise the effect through measured binding-protein concentrations (ALB, AAG, TPRO). Use FU when the fu value is a per-record model input the user is expected to supply – either because the paper measured it per subject, or because the paper varies it across named simulation scenarios. Promoted from specific to general alongside the Abdullah-Koolmees 2024 voriconazole whole-body PBPK extraction, the second model to use FU with consistent “fraction unbound of parent in plasma, supplied as data” semantics across two distinct mechanistic uses.

CSF_TPRO (canonical for cerebrospinal-fluid total protein)

• Description: Total protein concentration measured in cerebrospinal fluid (baseline or time-varying). CNS-compartment analogue of TPRO; used in popPK models of CSF-penetrating drugs as a surrogate for blood-brain-barrier integrity (elevated CSF protein indicates inflammation or barrier breakdown and typically correlates with increased CNS penetration of small-molecule drugs).
• Units: g/L (SI; canonical as of 2026-06-19). US-convention reporting in mg/dL converts via CSF_TPRO_gL = CSF_TPRO_mgdL * 0.01. 1 g/L = 100 mg/dL.
• Type: continuous
• Scope: general
• Reference category: n/a – enters either as an additive term on the logit-scale CSF uptake / barrier parameter, or as a power scaling (CSF_TPRO / ref)^exponent on a penetration fraction. Reference value observed: 1.2 g/L (Germovsek 2018 typical-infant value; sick-neonate cohort median).
• Source aliases:
  • CSF_protein – Germovsek 2018 paper notation.
  • CSFPROT / CSF_PROT – compact column-name forms common in NONMEM control streams.
• Example models: Germovsek_2018_meropenem.R (g/L, reference 1.2; additive on the logit CSF barrier parameter with coefficient theta_CSFproteins = -0.17 per g/L deviation from 1.2; ratified canonically on 2026-05-21 alongside the Germovsek 2018 meropenem extraction), Svensson_2020_rifampicin.R (mg/dL in the source, converted inside model() as CSF_TPRO * 100; rifampicin CSF partition coefficient multiplied by 1 + (log10(prot_mgdl) - log10(165)) / log10(165) * 0.631, centred on the cohort median 165 mg/dL).
• Notes: Distinct from TPRO (serum total protein) – the two are biologically independent because the blood-brain barrier prevents free equilibration of serum protein into CSF. Normal CSF protein is approximately 0.15-0.45 g/L in healthy adults; sick neonates and meningitis patients can reach several g/L. The covariate is typically time-varying because CSF protein evolves over the course of CNS inflammation; missing values are commonly imputed to the cohort median when the source paper does not report a per-sample CSF protein measurement.

IGG (canonical for serum immunoglobulin G)

• Description: Serum total immunoglobulin G concentration (baseline or time-varying). Used in mAb PK analyses as a competition-for-FcRn-recycling covariate on therapeutic-mAb clearance – high endogenous IgG is hypothesized to displace the therapeutic mAb from FcRn salvage and increase its catabolic clearance.
• Units: g/L (SI; canonical as of 2026-06-19). US-convention papers report in mg/dL; convert via IGG_gL = IGG_mgdL * 0.01 on data ingestion. 1 g/L = 100 mg/dL. Models calibrated to mg/dL must apply an inline conversion in model (e.g., igg_mgdL <- IGG * 100).
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (IGG / ref)^exponent. Reference values observed: 14.8 g/L (Zhou 2021), 9.65 g/L (Yang 2021).
• Source aliases:
  • BIGG (baseline IgG) – used in Zhou_2021_belimumab.R.
  • IGGBL (baseline IgG) – used in Yang_2021_cemiplimab.R.
• Example models: Zhou_2021_belimumab.R (g/L, reference 14.8; baseline-only; exponent 0.293 on CL), Yang_2021_cemiplimab.R (g/L, reference 9.65; small positive exponent 0.184 on shared CL/Q), Struemper_2017_belimumab.R (g/L, reference 13.7; baseline-only; exponent 0.347 on CL).
• Notes: Mechanistically meaningful for monoclonal-antibody PK because endogenous IgG competes with the therapeutic mAb for FcRn-mediated recycling. The per-model covariateData[[IGG]]$units field is load-bearing (1 g/L ~= 100 mg/dL). Baseline-vs-time-varying status documented in covariateData[[IGG]]$notes. Distinct from lIgG0 / IgG-as-a-state in mechanistic FcRn-competition TMDD models (e.g., Valenzuela_2025_nipocalimab.R), where IgG is a dynamic state, not a baseline covariate; use IGG only when the source paper treats IgG as a static (baseline) covariate column.

IGM (canonical for serum immunoglobulin M)

• Description: Serum total immunoglobulin M (IgM) concentration (baseline). Used in IgRT population-PK analyses as a proxy for B-cell antibody-producing capacity / humoral function – IgM is the first antibody produced after B-cell activation, so circulating IgM reflects ongoing B-cell activity prior to class-switching to IgG.
• Units: g/L (SI; canonical as of 2026-06-19). US-convention reporting in mg/dL converts via IGM_gL = IGM_mgdL * 0.01. 1 g/L = 100 mg/dL.
• Type: continuous
• Scope: specific
• Reference category: n/a – used with power scaling (IGM / ref)^exponent. Reference values observed: 0.21 g/L (Cheng 2026, pooled PID + SAD pediatric cohort median).
• Source aliases: none known.
• Example models: Cheng_2026_immunoglobulin.R (g/L, reference 0.21; baseline-only; power exponent 0.11 on baseline IgG (CBAS) – IgM enters as a humoral-capacity proxy that informs the endogenous-IgG baseline rather than directly modifying clearance).
• Notes: IgM is the immune-globulin class produced by activated B cells before class-switching, so it remains detectable in patients with hypogammaglobulinaemia who still have residual B-cell function. Scope: specific because the relevance of IgM as a covariate depends on the paper’s mechanistic interpretation (in Cheng 2026 it acts on the endogenous-IgG baseline; future use cases may differ). Promote to general if a second paper retains IgM with consistent semantics.

TBILI (canonical for total bilirubin)

• Description: Total serum bilirubin concentration.
• Units: umol/L (SI; canonical as of 2026-06-19). US-convention papers report in mg/dL; convert via TBILI_umolL = TBILI_mgdL * 17.1 on data ingestion. 1 mg/dL = 17.1 umol/L. Models calibrated to mg/dL must apply an inline conversion in model() (e.g., tbili_mgdL <- TBILI / 17.1 # SI umol/L -> US-convention mg/dL).
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (TBILI / ref)^exponent.
• Source aliases:
  • BIL (legacy NONMEM short label for total bilirubin) – used in NA_NA_lidocaine.R (DDMODEL00000281; binarised at threshold 0.53 mg/dL with BIL_HIGH = as.integer(BIL > 0.53)).
  • BILT (Urien 2005 capecitabine paper’s NONMEM short label for “total bilirubin”) – used in Urien_2005_capecitabine.R (umol/L, reference 8.8; power scaling on the capecitabine non-transformation CL10 and on the 5’-DFUR -> 5-FU rate constant K34).
• Example models: Yamada_2025_zolbetuximab.R (mg/dL, reference 0.38; small positive exponent 0.0347 on V1), NA_NA_lidocaine.R (mg/dL, source column BIL; binary effect at threshold 0.53 mg/dL on the GX elimination rate constant K30), Urien_2005_capecitabine.R (umol/L, reference 8.8; source column BILT; positive exponent +0.32 on capecitabine non-transformation CL10 and negative exponent -0.36 on the 5’-DFUR -> 5-FU rate constant K34), Cheng_2020_sirolimus.R (umol/L, reference 11.29 cohort median; source column TBIL; negative exponent -0.32 on sirolimus CL/F in children), Michalickova_2020_riociguat.R (mg/dL, reference 0.69 = cohort median, source column BILTOT; exponent -0.462 on the riociguat formation clearance to its M1 metabolite, CLf,M1/F).
• Notes: Hepatic-function marker. Unit varies by paper (US convention mg/dL, SI convention umol/L; 1 mg/dL ~= 17.1 umol/L). The per-model covariateData[[TBILI]]$units field is load-bearing.

DBIL (canonical for direct (conjugated) bilirubin)

• Description: Direct (conjugated) serum bilirubin concentration. Distinct from TBILI: direct bilirubin is the water-soluble glucuronide-conjugated fraction processed by hepatocytes and excreted in bile, so a rise in DBIL specifically flags impaired biliary excretion / cholestasis or intrahepatic shunting, whereas total bilirubin also captures unconjugated (indirect) hyperbilirubinaemia from haemolysis or Gilbert-type conjugation defects.
• Units: umol/L (SI; canonical as of 2026-06-19). US-convention papers report in mg/dL; convert via DBIL_umolL = DBIL_mgdL * 17.1 on data ingestion. 1 mg/dL = 17.1 umol/L. Models calibrated to mg/dL must apply an inline conversion in model (e.g., dbil_mgdL <- DBIL / 17.1).
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (DBIL / ref)^exponent. Reference values observed: 2.6 umol/L (Chen 2015 voriconazole Chinese ICU cohort population median), 2.82 umol/L (Li 2023 TQ-B3203 Chinese advanced-solid-tumor cohort population median).
• Source aliases: none known; source papers commonly write the abbreviation DBIL directly.
• Example models: Chen_2015_voriconazole.R (umol/L, reference 2.6; negative exponent -0.40 on CL: CL = TVCL * (DBIL / 2.6)^-0.40), Li_2023_TQB3203.R (umol/L, reference 2.82; negative exponent -0.24 on CL and -1.77 on the inter-compartmental clearance CL2: CL = TVCL * (BMI / 23.44)^0.78 * (DBIL / 2.82)^-0.24).
• Notes: Hepatic-function / cholestasis-specific marker. Unit varies by paper (US convention mg/dL, SI convention umol/L; 1 mg/dL ~= 17.1 umol/L). Distinct entry from TBILI because direct vs total are not interchangeable: total = direct + indirect, and the two fractions track different pathophysiologic processes. Promoted from Scope: specific to Scope: general alongside the Li 2023 TQ-B3203 extraction, which is the corroborating second model the prior note required: it uses DBIL with the same semantics as the founding example (SI umol/L, median-normalized negative power exponent on clearance, interpreted as impaired biliary excretion lowering clearance) in an unrelated drug and cohort. Both registered uses are median-normalized power effects on a clearance term, so the reference value is always a cohort median rather than a fixed clinical threshold – record the per-model median in covariateData[[DBIL]]$notes. Li 2023 additionally illustrates that a paper may retain DBIL on an inter-compartmental clearance as well as on elimination clearance; an inter-compartmental effect changes the concentration-time shape (and therefore Cmax) without changing AUC0-inf, which depends only on dose and elimination clearance.

AST (canonical for aspartate aminotransferase)

• Description: Serum aspartate aminotransferase activity (baseline or time-varying).
• Units: U/L (SI; canonical as of 2026-06-19; IU/L is used interchangeably). Document per-model via covariateData[[AST]]$units.
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (AST / ref)^exponent.
• Source aliases:
  • SGOT (serum glutamic-oxaloacetic transaminase; the legacy clinical-chemistry name for AST) – used in Quartino_2019_trastuzumab.R.
• Example models: Lu_2014_trastuzumabemtansine.R (U/L, reference 27; small positive exponent 0.071 on CL), Quartino_2019_trastuzumab.R (IU/L, reference 24; source column SGOT; positive exponent 0.205 on linear CL), Xu_2026_caspofungin_optimalDesign.R (U/L, reference 52.5 = the fitted subset median; exponent 0.898 on the intercompartmental clearance Q rather than on CL – the only example on Q; the effect was retained in the paper’s supplemental design model but did NOT survive into its final model, so treat it as subset-specific rather than a class effect), Osawa_2018_asunaprevir.R (U/L, reference 52 = the cohort median; genuinely time-varying, exponent -0.382 on CL/F, and the only example paired with a separate AST_BASE term – exponent -0.598 – normalised to the SAME reference constant), Krishnatry_2021_molibresib.R (IU/L, reference 28; time-varying, exponent -0.194 on the clearance of the active metabolite composite GSK3529246 rather than of the parent).
• Notes: Hepatic-function marker. Commonly reported alongside ALT and TBILI; register a separate ALT canonical if a future paper requires it. SGOT is the older lab-reporting name; values and units are identical to AST. When a paper carries a time-fixed baseline AST column alongside this one and gives the two distinct effects, the baseline column is AST_BASE; do not collapse the pair onto this entry.

ALT (canonical for alanine aminotransferase)

• Description: Serum alanine aminotransferase activity (baseline or time-varying).
• Units: U/L (SI; canonical as of 2026-06-19; IU/L is used interchangeably). Document per-model via covariateData[[ALT]]$units.
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (ALT / ref)^exponent.
• Source aliases:
  • SGPT (serum glutamic-pyruvic transaminase; the legacy clinical-chemistry name for ALT, paralleling SGOT -> AST) – used in NA_NA_lidocaine.R (DDMODEL00000281; binarised at threshold 11 with SGPT_HIGH = as.integer(SGPT > 11)).
• Example models: Nikanjam_2019_siltuximab.R (U/L, reference 19; small negative exponent -0.096 on CL), Melhem_2022_dostarlimab.R (U/L, reference 18; small negative exponent -0.0585 on CL, time-varying), NA_NA_lidocaine.R (source column SGPT; binary effects at threshold 11 on the GX rate constant K30 and on the 2,6-xylidide rate constant K40), Nguyen_2022_cemiplimab.R (IU/L, reference 19; baseline ALT, exponent -0.0729 on CL), Xie_2022_colistinSulfate.R (U/L, reference 37, cohort median 37.5, range 7-495 in critically ill adults; power exponent 0.635 on the PERIPHERAL volume of colistin sulfate rather than on a clearance, so it alters the time to steady state but not the steady-state average concentration).
• Notes: Hepatic-function marker. Commonly reported alongside AST and TBILI. SGPT is the older lab-reporting name; values and units are identical to ALT.

ALP (canonical for alkaline phosphatase)

• Description: Serum alkaline phosphatase activity (baseline or time-varying). Liver-function / cholestasis marker; often used in popPK covariate models either as a continuous concentration with power scaling or as a binary above/below upper-limit-of-normal (ULN) indicator. When binarized inline, document the ULN threshold used (typically ~120 U/L for adults; varies by lab, age, and sex).
• Units: U/L (SI; canonical as of 2026-06-19; IU/L is used interchangeably). Document per-model via covariateData[[ALP]]$units.
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (ALP / ref)^exponent, with a linear-deviation form, or binarized inline as alp_high <- (ALP > uln) for a binary >ULN indicator.
• Source aliases: none known.
• Example models: Gupta_2016_lenvatinib.R (binarized inline as alp_high <- (ALP > 120); the source paper enters ALP as a 0/1 NONMEM indicator with ALP = 1 when the ratio ALP/ULN > 1; multiplicative effect on CL/F: 0.883^alp_high), Wada_2023_sparsentan.R (continuous power scaling on CL/F normalised to the 68 U/L overall cohort median: (ALP / 68)^-0.208; elevated alkaline phosphatase lowers apparent clearance, and ALKP was the only one of four hepatic markers – ALKP, ALT, AST, total bilirubin – retained after backward elimination in a study where hepatic impairment itself was not tested as a covariate), Lalovic_2020_lemborexant.R (power exponent -0.118 on CL/F referenced to the 71 U/L cohort median).
• Notes: Liver-function / cholestasis marker; routine clinical-chemistry covariate. Commonly tested alongside ALT / AST / GGT / TBILI.

PHOS (canonical for serum phosphate)

• Description: Serum inorganic phosphate concentration, reported clinically as elemental phosphorus (baseline or time-varying). A routine serum-chemistry analyte and the pharmacodynamic biomarker of fibroblast growth factor receptor (FGFR) inhibition: FGFR1 blockade suppresses FGF23 signalling in the proximal tubule, reducing urinary phosphate excretion and raising serum phosphate, so on-treatment hyperphosphatemia is an on-target class effect of the FGFR-inhibitor drug class. Used both as a baseline covariate (predicting the on-treatment rise) and, in its change-from-baseline form, as an exposure surrogate.
• Units: mmol/L (SI; canonical). US-convention papers report in mg/dL; convert via PHOS_mmolL = PHOS_mgdL * 0.3229 on data ingestion (elemental phosphorus, atomic mass 30.974 g/mol, so 1 mg/dL = 0.32285 mmol/L and 1 mmol/L = 3.0975 mg/dL). Models calibrated against mg/dL values must either apply an inline conversion in model (e.g., phos_mgdL <- PHOS * 3.0975) or carry the coefficient rescaled to per-mmol/L units, with the conversion shown in the model file – the same treatment ALB receives (see the ALB entry). Adult reference range is roughly 2.5-4.5 mg/dL (0.81-1.45 mmol/L).
• Type: continuous
• Scope: general
• Reference category: n/a – entered so far in an uncentered linear form theta * PHOS. Reference values observed: baseline serum phosphate was not tabulated by Gong 2023, but the observed quartile boundaries of the change from baseline at 13.5 mg once daily were 0.5, 2.1, 2.7, 3.5, and 6.3 mg/dL (0.16, 0.68, 0.87, 1.13, and 2.03 mmol/L; Gong 2023 Figure 3a).
• Source aliases:
  • PHOS – the NONMEM $INPUT column name in the Gong 2023 Appendix S1 control stream, in mg/dL. Maps directly to the canonical name; unit conversion to SI mmol/L is required.
  • PO4 / PHOSPHORUS / IPHOS (inorganic phosphorus) – common clinical-laboratory reporting names for the same analyte.
• Example models: Gong_2023_pemigatinib_phosphate.R (mmol/L; uncentered linear coefficient -0.5730 mg/dL per mmol/L on the change-from-baseline endpoint, rescaled from the published -0.185 per mg/dL; the model’s endpoint is retained in the paper’s mg/dL so Table S2 remains usable verbatim).
• Notes: Distinct from ALP (alkaline phosphatase, an enzyme activity) and from CPK (creatine phosphokinase) despite the shared word stem – PHOS is the electrolyte. Housed in the renal / hepatic function section because phosphate homeostasis is renally regulated and because source papers report it inside the same serum-chemistry covariate panel as ALB / ALP / ALT / AST / TBILI (Gong 2023 tests exactly that panel); the register has no separate electrolyte section. When a source paper models the change from baseline in serum phosphate as an endpoint rather than as a covariate, that is a model output, not a covariate column – register only the baseline (or time-varying) input value here.

GGT (canonical for gamma-glutamyltransferase)

• Description: Serum gamma-glutamyltransferase activity (baseline or time-varying); hepatic / cholestatic biliary-enzyme marker.
• Units: U/L (SI; canonical as of 2026-06-19; IU/L is used interchangeably). Document per-model via covariateData[[GGT]]$units.
• Type: continuous
• Scope: general
• Reference category: n/a – used with linear-deviation form 1 + theta * (GGT - ref) or power scaling (GGT / ref)^exponent. Reference values observed: 33 U/L (Retlich 2015 popPK linagliptin median), 32.3 U/L (Retlich 2015 popPK/PD linagliptin median).
• Source aliases: none known.
• Example models: Retlich_2015_linagliptin.R (U/L, reference 33; linear-deviation effect on linagliptin CL with coefficient -0.0339 % per U/L deviation. The PK/PD layer uses GGT (reference 32.3 U/L) as a piecewise covariate on baseline DPP-4 activity BSL with a linear-deviation effect below GGT = 175 U/L and a constant +21.3% effect above the threshold).
• Notes: Liver-function / cholestasis marker; routine clinical-chemistry covariate. Commonly tested alongside ALT / AST / ALP / TBILI. The piecewise above/below-threshold form in Retlich 2015 reflects empirical saturation of the GGT-vs-DPP-4-activity relationship at extreme values.

TBA (canonical for total serum bile acids)

• Description: Total serum (or plasma) bile-acid concentration, the summed pool of primary and secondary conjugated and unconjugated bile acids. A sensitive marker of hepatocellular uptake / excretory function and of cholestasis: because bile acids undergo near-complete first-pass hepatic extraction from portal blood, the systemic concentration rises early when sinusoidal uptake or canalicular secretion is impaired, often before the transaminases or bilirubin move. Both baseline and time-varying usages are covered; document per-model in covariateData[[TBA]]$notes which applies.
• Units: umol/L (the conventional clinical reporting unit; document per-model via covariateData[[TBA]]$units).
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (TBA / ref)^exponent or exponential effects exp(coef * (TBA - ref)). Reference values observed: 3.95 umol/L (Liu 2025, cohort median in an elderly Chinese voriconazole TDM cohort).
• Source aliases:
  • TBA – used in Liu_2025_voriconazole.R (Table 1 / Table 2 and the published final-model equation).
  • SBA, TSBA, BA – total-(serum)-bile-acid variants seen in clinical-chemistry panels and NONMEM $INPUT columns; the same analyte, no value transformation.
• Example models: Liu_2025_voriconazole.R (median-normalized power effect on apparent voriconazole clearance, (TBA / 3.95)^-0.158, so a cholestatic rise in bile acids predicts slower clearance and higher voriconazole exposure).
• Notes: Registered on the same terms as the established general-scope liver-panel analytes TBILI, DBIL, AST, ALT, ALP and GGT – a routine clinical-chemistry measurement that recurs across hepatically-cleared-drug population-PK analyses – and is not an alias of any of them: total bile acids report hepatic uptake and canalicular transport capacity, whereas the transaminases report hepatocyte injury and bilirubin reports conjugation plus excretion, so the four can move independently (bile acids commonly rise with normal transaminases in early or mild cholestasis). Closest sibling is GGT, the other cholestasis-weighted marker, but GGT is an induced membrane enzyme and TBA a transported substrate pool; they are separate assays. Values against this column must be a total bile-acid pool, not an individual species (cholic, chenodeoxycholic, ursodeoxycholic acid), and not a post-prandial-stimulated measurement unless the source paper states so – record the fasting-vs-postprandial status in per-model notes, since the postprandial pool can be several-fold the fasting one. Right-skewed in unselected inpatient cohorts (Liu 2025: median 3.95, IQR 2.50-7.45 umol/L), so median normalization rather than mean centring is the usual encoding.

LDH (canonical for serum lactate dehydrogenase)

• Description: Serum lactate dehydrogenase activity (baseline or time-varying). General-purpose marker of tissue / cellular turnover; in oncology PK analyses it is interpreted as a disease-burden / cell-turnover proxy.
• Units: U/L (SI; canonical as of 2026-06-19; IU/L is used interchangeably). Document per-model via covariateData[[LDH]]$units.
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (LDH / ref)^exponent or with an additive linear-on-log form exp(coef * (log(LDH) - log(ref))) (algebraically equivalent to (log(LDH) / log(ref))^coef). Reference values observed: 217 U/L (Sanghavi 2020).
• Source aliases:
  • BLDH (baseline LDH) – used in Sanghavi_2020_ipilimumab.R.
• Example models: Sanghavi_2020_ipilimumab.R (linear-on-log form on CL with reference 217 U/L; coefficient 0.703), NA_NA_lidocaine.R (DDMODEL00000281; binary stratification at threshold 195 U/L switching the typical-value baseline of the 2,6-xylidide rate constant K40).
• Notes: Universal lab marker. Sanghavi 2020 log-transforms LDH because the distribution is heavily right-skewed (range 74-6,245 U/L over a median of 217); other papers may use a simple (LDH/ref)^exponent form. Document the functional form in covariateData[[LDH]]$notes. When a paper supplies LDH already normalized to each laboratory’s upper limit of normal (a unitless ratio rather than a U/L activity), use the sibling canonical LDHR instead – the two are NOT interchangeable, because LDH carries U/L units and LDHR is dimensionless.

LDHR (canonical for serum lactate dehydrogenase normalized to the laboratory’s upper limit of normal)

• Description: Baseline (or time-varying) serum lactate dehydrogenase activity divided by the reporting laboratory’s upper limit of normal (LDH_observed / ULN_LDH). Dimensionless. Used in multi-site oncology analyses so LDH from laboratories with different reference ranges can be pooled on a common scale, and because elevated-LDH-versus-normal is the clinically meaningful contrast in advanced melanoma rather than the absolute activity.
• Units: (unitless ratio)
• Type: continuous
• Scope: general
• Reference category: n/a – LDHR = 1 is the upper limit of normal, so a log-linear term (1 + log(LDHR) * coef) is centered at exactly the ULN and contributes nothing for a patient at the ULN. Also usable as a power term (LDHR / ref)^exponent. Feng 2019 reports a cohort median LDHR of 1.0 (range 0.4-40.5).
• Source aliases:
  • BLDHU – Feng 2019 Table 2 footnote f (“BLDHU is the normalized baseline LDH with upper limit of normal”).
  • LDH ratio – Feng 2019 Table 1 (“LDH ratio indicates patient’s actual value divided by the upper limit of normal”).
• Example models: Feng_2019_ipilimumab.R (log-linear form on both baseline tumor burden and the linear tumor-growth rate: TB0_TV = TB0_REF * (1 + log(LDHR) * 0.868) and TG_TV = TG_REF * (1 + log(LDHR) * 0.771) * (1 + CAV * -0.00342); note the linear-in-coefficient form (1 + log(LDHR) * coef), not the exponential exp(coef * log(LDHR)) form more common elsewhere in the register).
• Notes: Structurally parallel to ALBR (serum albumin normalized to ULN) – the same “analyte / ULN” construction applied to a different analyte. Distinct from LDH, which is the absolute activity in U/L; do not substitute one for the other, since a model calibrated on a ratio near 1 will be badly wrong if fed a U/L value near 200. Papers log-transform this ratio (as both LDH and LDHR users do) because the distribution is heavily right-skewed – Feng 2019 Table 1 footnote a: “Log-transformed LDH ratio was used in tumor growth dynamics and overall survival model development due to skewed distribution”, with an observed range spanning two orders of magnitude (0.4-40.5). General scope because ULN-normalized LDH is a standard multi-site oncology covariate and is not tied to any one analysis plan.

T4 (canonical for serum total thyroxine)

• Description: Serum total (protein-bound plus free) thyroxine concentration, the principal thyroid hormone. Baseline or time-varying. A member of the plain clinical-chemistry analyte family alongside ALB, TPRO, TBILI, AST, ALT, ALP, GGT and LDH.
• Units: ug/dL (the conventional clinical-chemistry unit for total T4; adult reference interval roughly 4.5-12 ug/dL). Papers reporting SI nmol/L convert via T4_ugdL = T4_nmolL / 12.87. Document per-model via covariateData[[T4]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – used with power scaling (T4 / ref)^exponent. Reference value observed: 7.40 ug/dL (Bi 2018 control-stream constant, near the middle of the adult reference interval).
• Source aliases:
  • THYROXIN – Bi 2018 $INPUT column name and Figure 1 forest-plot label; used in Bi_2018_testosteroneCypionate.R.
• Example models: Bi_2018_testosteroneCypionate.R (reference 7.40 ug/dL; power exponent 1.19 on the zero-order synthesis rate of LHRH-stimulated luteinizing hormone in a depot-testosterone HPG-axis PK/PD model. Bi 2018 retained the covariate in the final model on statistical grounds but reports that “No difference in simulated LH suppression was identified among volunteers with different thyroxin values”).
• Notes: Do not confuse with thyroxine-BINDING GLOBULIN (TBG), the transport protein, which is a different analyte on a different scale (adult reference roughly 12-30 ug/mL). Bi 2018 is the cautionary case: its Table 1 tabulates thyroxine-binding globulin (group medians 19.5 / 18.5 / 21 ug/mL) while its control stream’s covariate is THYROXIN referenced to 7.40, and the supplementary LH dataset carries thyroxin = 7.4 for its example subject – a value consistent with total T4 in ug/dL and inconsistent with TBG. A future extraction needing the binding protein should register a separate TBG canonical rather than reuse this one. Free thyroxine (FT4, ng/dL) is likewise a distinct analyte and needs its own sibling entry.

AMYL (canonical for blood amylase activity)

• Description: Blood (serum or plasma) total amylase activity (baseline or time-varying). Routine clinical-chemistry enzyme, conventionally read as a pancreatic / salivary marker, but amylase is also cleared renally and therefore accumulates as glomerular filtration falls. In population PK analyses of renally-impaired cohorts it is used in the latter role – as a surrogate for uremic burden rather than for pancreatic disease – so the direction of any effect should be interpreted against the paper’s stated mechanism, not assumed to be pancreatic.
• Units: U/L (SI; canonical, matching the AST / ALT / ALP / GGT / LDH enzyme family; IU/L is used interchangeably). Document per-model via covariateData[[AMYL]]$units.
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (AMYL / ref)^exponent. Reference values observed: 59.5 IU/L (Kim 2025 healthy-subject median).
• Source aliases:
  • Amyl – Kim 2025 Table 1 / Table 2 and NONMEM $INPUT column abbreviation.
• Example models: Kim_2025_evogliptin.R (IU/L, reference 59.5; positive exponent 0.363 on relative bioavailability F1, paired with TRIG; both markers rise with worsening renal impairment and jointly express the paper’s finding that uremia inhibits CYP3A4-mediated first-pass metabolism and so raises oral bioavailability).
• Notes: Routine clinical-chemistry covariate; follows the bare-abbreviation naming of the established enzyme canonicals AST, ALT, ALP, GGT, and LDH. Distinct from lipase, which is the companion pancreatic enzyme and was recorded but not tested in Kim 2025 – register a separate LIPASE canonical if a future paper retains it. Because amylase can move for either pancreatic or renal reasons, covariateData[[AMYL]]$notes should state which interpretation the source paper intends.

HEPIMP_MILD (canonical for mild hepatic impairment indicator)

• Description: 1 = mild hepatic impairment per the National Cancer Institute Organ Dysfunction Working Group (NCI ODWG) criteria, 0 = normal hepatic function or non-mild category. NCI ODWG mild = total bilirubin <= ULN with AST > ULN, OR total bilirubin > 1.0xULN to <= 1.5xULN with any AST.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (normal hepatic function; the moderate / severe categories are typically pooled into the reference for population PK analyses where mild impairment is the only category with non-trivial sample size).
• Source aliases:
  • HEPIMP (with values 1 = mild / 0 = others) – used in Lin_2024_casirivimab.R.
  • Child-Pugh A – used in Desai_2016_isavuconazole.R and in the Luo_2024_*_pbpk.R CES1 semi-PBPK family, where the three HEPIMP_MILD / HEPIMP_MOD / HEPIMP_SEV indicators are mutually exclusive and select the Child-Pugh A / B / C physiology columns respectively, with all three 0 selecting the normal-hepatic-function column.
• Example models: Lin_2024_casirivimab.R (NCI ODWG mild; multiplicative fractional change on CL), Lu_2022_patritumab.R (NCI ODWG mild; paired with HEPIMP_MOD_OR_MISSING; multiplicative fractional effect 0.706 on CLDXd for mild impairment vs the normal-hepatic-function reference), Desai_2016_isavuconazole.R (Child-Pugh A; log-additive shift e_hepimp_mild_cl = log(1.55 / 2.54) = -0.494 on CL and e_hepimp_mild_q = log(38.8 / 33.678) = +0.142 on Q vs the healthy reference, paired with HEPIMP_MOD for the parallel Child-Pugh B stratum), Nguyen_2019_cabozantinib_liver_dysfunction.R (NCI ODWG mild; direct multiplicative factor 1.12^HEPIMP_MILD on CL/F and 1.04^HEPIMP_MILD on Vc/F, paired with HEPIMP_MODSEV so that both indicators zero is the normal-hepatic-function reference; 558 / 2023 = 28% of the cohort were mild. This is the rare case where the covariate’s scientific value is a NEGATIVE result: the 12% clearance increase was judged clinically negligible and the authors retained the model WITHOUT these covariates as final, which is why the companion Nguyen_2019_cabozantinib.R omits them).
• Notes: Use this column when a model dichotomizes hepatic-impairment status as “mild vs. others” (i.e., normal + the rare moderate/severe cases pooled into the reference). For models that test moderate or severe as separate categories, register additional canonicals HEPIMP_MOD / HEPIMP_SEV rather than overloading this entry.

HEPIMP_MOD_OR_MISSING (canonical for composite moderate-or-data-missing hepatic impairment indicator)

• Description: 1 = moderate hepatic impairment per the NCI ODWG criteria OR baseline hepatic-function data missing/unknown; 0 = normal hepatic function or any other (non-moderate, non-missing) category. Composite indicator used by source papers that pool the moderate-impairment subgroup with patients whose hepatic-function data are missing because both subgroups are individually too small to estimate as separate effects.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (normal hepatic function; mild impairment is typically captured by a separate HEPIMP_MILD indicator paired with this column, so all-zero corresponds to NCI ODWG group 1 = normal).
• Source aliases:
  • HEPIMP_MOD_MISSING – prior canonical name (pre-2026-06-19 OR-disambiguation standardization).
  • HEPATIC / HEPATIC_MOD_MISSING – informal NONMEM names for the composite group.
• Example models: Lu_2022_patritumab.R (paired with HEPIMP_MILD; multiplicative fractional effect 0.532 on CLDXd; the composite group pools n = 6 moderate-impairment patients with n = 6 missing/unknown patients per Lu 2022 Table S5).
• Notes: Specific scope because the composition of the “moderate or missing” group is paper-defined and the missing/unknown subgroup may have a different distribution of true hepatic-function status across studies. Use only when the source paper explicitly pools the moderate-impairment cases with missing-data cases under a single coefficient; for models that estimate moderate impairment separately (without pooling missing data), use the HEPIMP_MOD canonical instead.

B2M (canonical for serum beta-2-microglobulin)

• Description: Serum beta-2-microglobulin concentration. Low-molecular-weight (~12 kDa) protein freely filtered at the glomerulus and reabsorbed in the proximal tubule; serum levels rise with renal impairment, with increased plasma-cell turnover in multiple myeloma, and with broader lymphoid-cell turnover. Used in oncology PK analyses both as a renal-function proxy and as a tumor-burden / disease-severity covariate.
• Units: mg/L
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (B2M / ref)^exponent. Reference values observed: 3.90 mg/L (Fau 2020 multiple-myeloma cohort median).
• Source aliases: none; B2M is the universal abbreviation.
• Example models: Fau_2020_isatuximab.R (mg/L, reference 3.90; exponent 0.343 on the steady-state linear clearance CLinf), Ide_2022_elotuzumab.R (thresholded B2M >= 3.5 mg/L indicator, factor 1.11 on VC; source column B2MICG_Z in mg/dL thresholded at 0.35).
• Notes: In multiple myeloma B2M is part of the International Staging System (ISS); in routine PK analyses it is interpreted simultaneously as a renal-function and disease-burden marker. Document the interpretation per-model via covariateData[[B2M]]$notes.

CYSC (canonical for serum cystatin C)

• Description: Serum cystatin C concentration. Low-molecular-weight (~13 kDa) protease-inhibitor protein produced at a near-constant rate by all nucleated cells; freely filtered at the glomerulus and almost entirely reabsorbed and catabolized in the proximal tubule, so serum cystatin C reflects glomerular filtration rate (GFR) more directly than serum creatinine. Used as a renal-function covariate on the clearance of renally-eliminated drugs, especially in populations where creatinine-based estimates of GFR are unreliable (low muscle mass, normal-creatinine concentrations masking impaired GFR, ICU patients).
• Units: mg/L
• Type: continuous
• Scope: general
• Reference category: n/a – used with either power scaling (CYSC / ref)^exponent (Chung 2013) or centred-linear scaling on the reciprocal 1 + e * (1/CYSC - ref_inv) (Viberg 2006). Reference values observed: 0.91 mg/L (Chung 2013 vancomycin Korean adults with SCr <= 1.2 mg/dL; cohort median); 1.32 mg/L equivalent to 1/CYSC = 0.758 (mg/L)^-1 (Viberg 2006 cefuroxime adult patients with broad renal-function range; population-typical).
• Source aliases:
  • Cystatin C / cystatin – Chung 2013 paper narrative and Table 2 footnote.
  • CysC – Viberg 2006 paper narrative and Table 4 footnote.
  • CYSC – used in Wang_2019a_tacrolimus.R (Wang 2019 Methods ‘Patients and data collection’ and Table I; 0.85 +/- 0.25 mg/L, median 0.8, range 0.4-2.3 mg/L in Chinese children with refractory nephrotic syndrome). This is the first registered use of cystatin C as a disease-progression marker rather than as a renal-elimination covariate: tacrolimus is not appreciably renally cleared, and Wang 2019 interprets the negative coefficient on apparent oral clearance as nephrotic-syndrome severity rather than as reduced glomerular filtration. Serum creatinine was screened in the same model and was NOT retained.
• Example models: Chung_2013_vancomycin.R (mg/L, reference 0.91; power exponent -0.780 on CL: CL_pop * (CYSC / 0.91)^-0.780), Viberg_2006_cefuroxime.R (mg/L; centred-linear effect on 1/CYSC with coefficient 1.43 per (mg/L)^-1 and reference 0.758 (mg/L)^-1 on CL: CL_pop * (1 + 1.43 * (1/CYSC - 0.758))), Ding_2026_vancomycin.R (mg/L, reference 1.5; the only registered model in which cystatin C enters both disposition parameters, as two separately-estimated power exponents: -0.650 on CL and -0.294 on V. Ding 2026 attributes the volume effect to reduced excretion at higher CYSC producing a relative fall in apparent V, and notes it is the first report of the association), Huang_2025_colistinSulfate.R (mg/L, reference 2.31 – the cohort MEDIAN, explicitly distinguished in the Results text from the Table 1 mean of 2.87; power exponent -0.626 on the CL of colistin sulfate, observed range 1.05-5.11 in critically ill adults. This is the founding continuous-renal-replacement-therapy example, and the reason it matters is mechanistic rather than incidental: the cohort was entirely on CVVHDF, and creatinine clearance, serum creatinine, blood urea nitrogen and uric acid were ALL screened on CL and none was retained, while cystatin C was. The authors attribute this to molecular size – CVVHDF combines convection and diffusion and therefore clears creatinine (113 Da) efficiently, so serum creatinine and any Cockcroft-Gault estimate derived from it stop tracking residual renal function once the filter is running, whereas cystatin C (13.25 kDa) is not appreciably removed (cited clearance 17 mL/min, under 30% of its production rate) and so remains interpretable on dialysis. A model needing a renal-function covariate in a CRRT population should prefer this column over CRCL for that reason).
• Notes: Cystatin C is freely filtered at the glomerulus and is not secreted by the renal tubule (unlike creatinine), so it is less sensitive to muscle mass, body composition, and tubular-secretion blockers. Reference ranges 0.57-0.97 mg/L for adult females and 0.65-1.10 mg/L for adult males (Chung 2013 Methods; Roche Cobas 6000 particle-enhanced immunoturbidimetric assay). Distinct from CREAT (serum creatinine) – the two are commonly reported alongside each other and can enter the same model as separate covariates (as in Chung 2013, where CYSC explains 62% of CL variability vs SCr 13%). The functional form (power on CYSC vs centred-linear on 1/CYSC) is paper-specific and lives in the model file; the canonical column is the underlying biomarker concentration in mg/L.

NTPROBNP (canonical for serum N-terminal pro-B-type natriuretic peptide)

• Description: Serum (or plasma) concentration of N-terminal pro-B-type natriuretic peptide, the biologically inactive cleavage fragment co-secreted with BNP by cardiac ventricular myocytes in response to increased ventricular wall tension. The standard laboratory index of cardiac stress / ventricular dysfunction, and the covariate through which impaired cardiac function – and the reduced cardiac output and organ perfusion that accompany it – enters a pharmacokinetic model. Baseline or time-varying.
• Units: ng/mL. Read the source paper’s unit carefully: the near-universal clinical reporting unit is pg/mL (equivalently ng/L), and 1 ng/mL = 1000 pg/mL, so a paper reporting a cohort median of “1.095” in ng/mL and one reporting “1095” in pg/mL describe the same cohort. Document the unit actually used via covariateData[[NTPROBNP]]$units and convert the centring constant with it.
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (NTPROBNP / ref)^exponent. Reference values observed: 10 ng/mL (Ding 2026 cardiac-surgery cohort; a round value well above the cohort median of 1.095 ng/mL, chosen in the region of the elevated-cardiac-stress scenarios the paper simulates rather than at the cohort centre).
• Source aliases:
  • NT_proBNP – the NONMEM $INPUT column name in Ding_2026_vancomycin.R; abbreviated BNP inside that stream’s $PK block (BNP=NT_proBNP).
  • NT-proBNP / NTproBNP – the common narrative and table spellings.
• Example models: Ding_2026_vancomycin.R (founding example; ng/mL, reference 10; power exponent -0.0823 on vancomycin CL, so clearance falls as cardiac stress rises. This is the paper’s headline finding: in cardiac-surgery patients, cardiac function measured by NT-proBNP is a covariate on vancomycin clearance independently of the two renal-function covariates CREAT and CYSC already in the model, and eliminating it costs 41.35 OFV points on 1 df).
• Notes: Distinct from LVEF (left ventricular ejection fraction), the imaging-based measure of systolic function: the two are related but not interchangeable, and Ding_2026_vancomycin.R is a worked case where NT-proBNP was retained while LVEF was rejected in the same covariate search, both as a continuous covariate (dOFV 1.074) and dichotomised at 0.4 and 0.3. An assay upper limit of detection is common and matters for the power form: Ding 2026’s assay saturated at 35 ng/mL and results reported as “> 35” were set to 35 before modelling, which truncates the covariate’s upper tail. A cut-off of 0.3 ng/mL is conventionally used to identify dyspnoea caused by heart failure. Because the power model is undefined at zero and NT-proBNP spans orders of magnitude within a single cohort (0.117-19.12 ng/mL at the 2.5-97.5 percentiles in Ding 2026), a simulated cohort must draw this covariate on the log scale.

HEPIMP (canonical for hepatic-impairment indicator (NCI ODWG classification))

• Description: Baseline hepatic-impairment indicator per the National Cancer Institute Organ Dysfunction Working Group (NCI ODWG) classification: 1 = mild or worse hepatic impairment (group >= 2 = mild, moderate, or severe), 0 = normal hepatic function (group 1).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (normal hepatic function, NCI ODWG group 1).
• Source aliases:
  • BHPTGRPN (categorical: 1 = normal, 2 = mild, 3 = moderate, 4 = severe; 9999 = missing) – used in Lu_2019_polatuzumab.R. Decompose: HEPIMP = as.integer(BHPTGRPN > 1.5 & BHPTGRPN != 9999).
  • HEP_IMP – retired canonical name; replaced by HEPIMP for consistency with the HEPIMP_MILD family.
• Example models: Lu_2019_polatuzumab.R (multiplicative effect on FRAC_NS = 1.19, applied as 1.19^HEPIMP).
• Notes: NCI ODWG classification (Ramalingam SS et al., J Clin Oncol 2010;28:4507) groups subjects by total bilirubin and AST: group 1 = normal, group 2 = mild (TBILI <= ULN and AST > ULN, or TBILI > 1-1.5 x ULN), group 3 = moderate (TBILI > 1.5-3 x ULN), group 4 = severe (TBILI > 3 x ULN). Source papers typically pool groups 2-4 versus group 1 for a binary indicator because the impaired-liver subgroups are individually small. If a future model needs finer resolution (separate effects for mild vs moderate-or-worse), add a parallel HEPIMP_MOD canonical rather than overloading this one.

NASF (canonical for nonalcoholic steatohepatitis severity score combining NAS and fibrosis staging)

• Description: Integer composite severity score for nonalcoholic steatohepatitis (NASH), summing the NAFLD activity score (NAS: steatosis 0-3 + hepatocyte ballooning 0-2 + lobular inflammation 0-3, total 0-8) and the fibrosis staging score (0 = absent, 1 = perisinusoidal / pericellular, 2 = periportal, 3 = bridging, 4 = cirrhosis). Total range 0-12; healthy subjects without biopsy-confirmed NASH are assigned NASF = 0 by convention. Scores below 5 reflect a benign form of NAFLD; scores >= 5 reflect biopsy-confirmed NASH.
• Units: (count, 0-12)
• Type: count
• Scope: specific
• Reference category: n/a – used with a linear effect on log(NASF / 4) for NASF >= 4 and zero contribution for NASF < 4 (so NASF = 4 and any NASF < 4 reduce to the typical-value reference). The cutoff of 4 distinguishes patients with a benign form of NAFLD from those with biopsy-confirmed NASH (Pierre 2017 Methods ‘Covariate analysis’ and references 31 and 34).
• Source aliases:
  • NASF – used in Pierre_2017_morphine.R (Pierre 2017 Methods ‘Covariate analysis’).
• Example models: Pierre_2017_morphine.R (linear effect on log(NASF / 4) for NASF >= 4 with coefficient -0.628 on M3G clearance: CL_M3G_i = CL_M3G_pop * (1 + e_nasf_cl_m3g * log(NASF / 4)) for NASF >= 4 and CL_M3G_i = CL_M3G_pop for NASF < 4; higher NASF reduces M3G clearance via reduced biliary excretion and increased basolateral efflux of M3G into systemic circulation).
• Notes: The NAFLD activity score (NAS) component is the histology score described by Bondini 2007 / Kleiner 2005 and references therein; the fibrosis staging is the Brunt / NASH-CRN system. The combined NASF score is the noninvasive staging proposed by Santiago-Rolon 2015 (Proc R Health Sci J 34:189-194) and used by Angulo 2007 as the NAFLD Fibrosis Score cutoff. Scope: specific because the precise cutoff (NASF >= 4) and the linear-on-log functional form are Pierre 2017’s modeling choice; future papers may model NASF or its components differently. Distinct from HEPIMP* (NCI ODWG oncology-trial hepatic-impairment categories) and from continuous liver enzymes (ALT, AST, ALP) – NASF is a biopsy-derived disease-severity ordinal specific to NAFLD / NASH.

HEPIMP_SEV (canonical for severe hepatic impairment indicator)

• Description: 1 = severe hepatic impairment, 0 = normal hepatic function or less-than-severe category. The classification scheme that defines “severe” is paper-specific and must be documented in per-model covariateData[[HEPIMP_SEV]]$notes. Two schemes are commonly encountered:
  • NCI ODWG group 4: total bilirubin > 3 x ULN with any AST (Ramalingam SS et al., J Clin Oncol 2010;28:4507).
  • Child-Pugh Class C: composite score 10-15 across bilirubin, albumin, INR, ascites, and encephalopathy.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (any non-severe category: normal, mild, or moderate; the model typically uses HEPIMP_SEV alongside other severity-specific indicators that partition the non-severe pool further).
• Source aliases:
  • Child-Pugh Class C – used in vanderWalt_2013_dapagliflozin.R (covariate effects on CLP_M15 and V2M; the paper dichotomizes severe hepatic impairment per the Child-Pugh classification).
• Example models: vanderWalt_2013_dapagliflozin.R (Child-Pugh Class C; multiplicative fractional effects -0.422 on the dapagliflozin -> D3OG metabolic clearance and +1.33 on the D3OG central volume of distribution; paper text “With severe HI (Child-Pugh Class C), CLP M15 decreased by 41% and V2M increased by 134%”), Comisar_2025_rimegepant.R (multiplicative fractional effect on CL/F = -0.423 per Comisar 2025 Table 3, a 42.3% clearance decrease giving a 1.7-fold rise in simulated steady-state AUCtau, which supports the label recommendation to avoid rimegepant in severe hepatic impairment; paired with HEPIMP_MOD (-0.229) as two mutually exclusive indicators sharing a normal-OR-mild reference group; classification scheme not stated in the paper).
• Notes: Use this column when a model dichotomizes severe hepatic impairment as a separate indicator from milder categories. The classification scheme (NCI ODWG vs Child-Pugh vs other) is paper-specific and must be documented per-model. For composite “moderate-or-severe” pooled indicators, use the parallel HEPIMP_MODSEV canonical rather than overloading this entry. Companion to HEPIMP_MILD (mild only) and HEPIMP_MODSEV (moderate + severe pooled); the register anticipates each severity level as its own canonical when the source paper tests them as separate covariates.

HEPIMP_MODSEV (canonical for composite moderate-or-severe hepatic impairment indicator)

• Description: 1 = moderate or severe hepatic impairment, 0 = normal hepatic function or mild impairment. Composite indicator used by source papers that pool the moderate and severe subgroups because the severe subgroup alone is too small to support a separate covariate-effect estimate. Distinct from HEPIMP_MOD_OR_MISSING (which pools moderate cases with missing-data cases, not with severe cases). The classification scheme that defines the cut points is paper-specific and must be documented in per-model covariateData[[HEPIMP_MODSEV]]$notes. Two schemes are commonly encountered:
  • NCI ODWG groups 3-4 pooled: total bilirubin > 1.5 x ULN with any AST.
  • Child-Pugh Class B or C pooled: composite score >= 7.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (normal hepatic function or mild impairment; the indicator is mutually exclusive with HEPIMP_MILD, so all-zero on both indicators corresponds to the normal-function reference and HEPIMP_MILD=1 with HEPIMP_MODSEV=0 corresponds to mild-only).
• Source aliases:
  • Child-Pugh Class B,C – used in vanderWalt_2013_dapagliflozin.R (covariate effects on V3P and CLM; the paper dichotomizes moderate-or-severe hepatic impairment per the Child-Pugh classification).
• Example models: vanderWalt_2013_dapagliflozin.R (Child-Pugh Class B or C; multiplicative fractional effects -0.600 on the dapagliflozin peripheral volume of distribution V3P and -0.293 on the D3OG renal clearance CLM; paper text “Moderate or severe HI (Child-Pugh Class B or C) decreased CLM and the peripheral volume of distribution of dapagliflozin (V3P) by 29 and 60%, respectively”), Nguyen_2019_cabozantinib_liver_dysfunction.R (NCI ODWG groups 3-4 pooled; direct multiplicative factor 0.978^HEPIMP_MODSEV on CL/F and 1.06^HEPIMP_MODSEV on Vc/F, paired with HEPIMP_MILD. An extreme example of why the pooling exists and of its limits: only 15 moderate and 1 severe subject out of 2023, so both effects are indistinguishable from unity and the Vc/F 90% CI spans 0.658 to 1.71. The paper’s Conclusions state that the limited data “preclude providing any dosing recommendations for these subpopulations”).
• Notes: Use this column when a model pools moderate-and-severe hepatic impairment under a single coefficient (typically because the severe subgroup alone is too small to estimate as its own effect). The classification scheme (NCI ODWG vs Child-Pugh vs other) is paper-specific and must be documented per-model. Companion to HEPIMP_MILD (mild only) and HEPIMP_SEV (severe only). Distinct from HEPIMP_MOD_OR_MISSING (which pools moderate cases with subjects whose hepatic-function data are missing/unknown, not with severe cases). The composite mod-or-sev pooling is a different load-bearing convention than the mod-or-missing pooling, so the two canonicals must remain separate.

LIVER_RESECT_MAJOR (canonical for major hepatic-resection indicator)

• Description: 1 = major hepatic resection (three or more liver segments resected), 0 = minor hepatic resection (two liver segments resected). Time-fixed per subject, describing the surgical exposure rather than a chronic hepatic-impairment state. Distinct from the HEPIMP_* family (which classifies chronic liver disease by NCI ODWG / Child-Pugh criteria) because major hepatic resection produces a transient reduction in metabolic capacity that recovers within 2-3 weeks in patients with normal underlying liver function (Ollier 2015 Discussion citing Fortner et al., Ann Surg 1987).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (minor resection; two liver segments)
• Source aliases:
  • NSH – used in Ollier_2015_ropivacaine.R (Ollier 2015 paper notation, dichotomised: NSH = 0 for 2 segments, NSH = 1 for 3, 4 or 5 segments per Methods ‘Covariate model’).
• Example models: Ollier_2015_ropivacaine.R (multiplicative exponential effect on free-ropivacaine clearance: Cl = exp(lcl + etalcl) * exp(beta_NSH * LIVER_RESECT_MAJOR) with beta_NSH = log(620/1310) = -0.7480; Cl drops from 1310 L/h at LIVER_RESECT_MAJOR = 0 to 620 L/h at LIVER_RESECT_MAJOR = 1, a 53% reduction).
• Notes: Specific scope pending a second liver-resection-popPK model to ratify the name. The dichotomisation at >= 3 segments reflects the Ollier 2015 analytical choice; a future model that parameterises resection extent as a continuous count of segments should introduce a separate canonical rather than repurpose LIVER_RESECT_MAJOR. Do not confuse with the HEPIMP_* family: HEPIMP_MILD / HEPIMP_MOD / HEPIMP_SEV etc. describe chronic hepatic impairment classified from labs (bilirubin, AST) or Child-Pugh scores, whereas LIVER_RESECT_MAJOR is a surgical-exposure covariate that acts on drug clearance for the transient postoperative window.

LSM (canonical for liver stiffness measured by transient elastography)

• Description: Baseline liver stiffness measurement (LSM) obtained by vibration-controlled transient elastography (FibroScan-type), the standard non-invasive surrogate for hepatic fibrosis stage in chronic liver disease. Time-fixed per subject.
• Units: kPa
• Type: continuous
• Scope: general
• Reference category: n/a – typically enters as a power term normalised to a cohort reference, (LSM / ref)^exponent. Reference value observed: 8.1 kPa (Ooi 2026 Supplementary Datafile S6 and the Figure S4 forest plots).
• Source aliases:
  • LIVSTBL (baseline liver stiffness) – the NONMEM $INPUT column in Ooi 2026 Supplementary Datafiles S5 and S6; same quantity in kPa, no value transformation.
• Example models: Ooi_2026_elafibranor_alptb.R (power exponent 0.333 on the typical baseline total bilirubin of the joint ALP / total-bilirubin indirect-response model; Ooi 2026 Table 2).
• Notes: A bare clinical-measurement abbreviation in the same group as ALB, ALT, ALP, TBILI and CRCL, and registered at general scope because liver stiffness is a standard hepatology measurement rather than a paper-defined construct. Distinct from the categorical fibrosis-stage and hepatic-impairment indicators (HEPIMP, HEPIMP_MILD, …), which classify rather than measure. When a source imputes a missing LSM by setting the covariate factor to 1 (as Ooi 2026 does), supply the reference value instead so the power term evaluates to 1.

RENALIMP (canonical for renal-impairment indicator (any degree))

• Description: 1 = renal impairment of any degree (mild, moderate, severe, or kidney failure / ESRD), 0 = normal renal function. Pooled parent of the RENALIMP_* severity family, for source papers that carry renal impairment as a single binary covariate rather than resolving the individual strata. The threshold separating “normal” from “impaired” is paper-specific and must be documented in per-model covariateData[[RENALIMP]]$notes; a common scheme (used by Zhang 2025 and mirrored in FDA / EMA renal-impairment labeling guidance) is eGFR < 90 mL/min/1.73 m^2 as impaired.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (normal renal function; matched healthy controls in the founding example).
• Source aliases:
  • Renal Impairment (binary, “0 representing controls and 1 indicating renal impairment”) – used in Zhang_2025_remdesivir.R (Zhang 2025 Methods 2.3 ‘Mixed effect model and covariant analysis’).
• Example models: Zhang_2025_remdesivir.R (multiplicative shift on the remdesivir-to-GS-704277 and GS-704277-to-GS-441524 metabolic conversion rate constants, from the healthy-control column to the severe-renal-impairment column of Zhang 2025 Table 1).
• Notes: Exact structural analogue of HEPIMP (the pooled “mild or worse hepatic impairment” indicator that sits alongside HEPIMP_MILD / HEPIMP_MOD / HEPIMP_SEV / HEPIMP_MODSEV), and completes the same symmetry for the renal family, whose members already state that they “follow the HEPIMP_* family precedent”. Use this column only when the source paper genuinely pools every impairment stratum under one coefficient – Zhang 2025 says so explicitly: “To maintain simplicity, varying degrees of renal impairment were not further distinguished as separate covariates.” If a future model needs finer resolution, use the existing RENALIMP_MILD / RENALIMP_MOD / RENALIMP_SEV / RENALIMP_ESRD indicators rather than overloading this one. Companion to continuous CRCL. Distinct from RRT_HEMODIAL_STATUS / RRT_HEMODIAL_ACTIVE, which record dialysis treatment rather than the underlying renal-function stratum.

RENALIMP_MOD (canonical for moderate renal impairment indicator)

• Description: 1 = moderate renal impairment, 0 = normal renal function, mild renal impairment, or any non-moderate category. Binary categorical indicator used by source papers that test renal-impairment status as a discrete covariate rather than via continuous creatinine clearance (CRCL). The classification scheme that defines “moderate” is paper-specific and must be documented in per-model covariateData[[RENALIMP_MOD]]$notes. Two schemes are commonly encountered:
  • Cockcroft-Gault CrCl 30-50 mL/min (Morris 2011 ZP-006 study eligibility criteria).
  • FDA / EMA labeling: CrCl 30-59 mL/min (US FDA Guidance for Industry: PK in Patients with Impaired Renal Function, 2020).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (normal renal function or mild renal impairment; severe and end-stage categories are typically pooled with the reference or excluded from the source cohort).
• Source aliases:
  • RENAL (binary 1 = moderate / 0 = healthy or mild) – used in Morris_2011_telapristone.R (Morris 2011 Methods “Covariate Analysis”, paragraph after Equation for the proportional covariate model).
• Example models: Morris_2011_telapristone.R (proportional fractional effect on the telapristone absorption rate constant Ka; coefficient -0.744 yielding Ka = Ka_typ * (1 + e_renalimp_mod_ka * RENALIMP_MOD), i.e., a 74% decrease in Ka in moderate-renal-impaired subjects vs the reference healthy/mild-renal cohort), CarlssonPetri_2018_semaglutide.R.
• Notes: Use this column when a model dichotomizes moderate renal impairment as a separate indicator (typically because severe / ESRD subjects are not in the source cohort and the moderate group is the only renal-impairment stratum with non-trivial sample size). The classification scheme (Cockcroft-Gault vs MDRD vs CKD-EPI; mL/min vs mL/min/1.73 m^2) is paper-specific and must be documented per-model. Companion to CRCL (continuous Cockcroft-Gault clearance), RENALIMP_MILD and RENALIMP_SEV (parallel canonicals reserved for future extractions following the HEPIMP_* family precedent). The Morris 2011 Discussion notes that the effect is “not directly attributed to renal impairment (in terms of decrease glomerular filtration rate/changes in creatinine clearance)” but “attributed to some other disease precluding/resulting in renal impairment or other pathophysiologic states induced by renal impairment (e.g., delayed gastric emptying as a result of diabetes)”, so the binary indicator captures the cohort allocation rather than a mechanistic GFR effect.

RENALIMP_MILD (canonical for mild renal impairment indicator)

• Description: 1 = mild renal impairment, 0 = normal renal function or any non-mild category. Binary categorical indicator used by source papers that test renal-impairment status as a discrete covariate rather than via continuous creatinine clearance (CRCL). The classification scheme that defines “mild” is paper-specific and must be documented in per-model covariateData[[RENALIMP_MILD]]$notes. A common scheme (used by Carlsson Petri 2018 and mirrored in FDA / EMA labeling guidance for renal-impairment PK) is eGFR 60-89 mL/min/1.73 m^2 as mild versus eGFR >= 90 mL/min/1.73 m^2 as normal.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (normal renal function; the moderate / severe categories are typically encoded as separate RENALIMP_MOD / RENALIMP_SEV indicators alongside this one, with all three set to 0 for the normal-renal-function reference).
• Source aliases:
  • RENAL (multi-level categorical with mild as one level; decompose to binary indicators) – used in CarlssonPetri_2018_semaglutide.R (Carlsson Petri 2018 Methods ‘Renal impairment groups were categorized [by estimated glomerular filtration rate (eGFR)] as normal function (eGFR >= 90 ml/min/1.73 m^2), mild (eGFR = 60-89 ml/min/1.73 m^2), moderate (eGFR = 30-59 ml/min/1.73 m^2), or severe (eGFR < 30 ml/min/1.73 m^2) renal impairment’).
• Example models: CarlssonPetri_2018_semaglutide.R (multiplicative CL/F ratio 0.948 for mild renal impairment vs normal reference, encoded as e_renal_mild_cl^RENALIMP_MILD per Table S3).
• Notes: Sister canonical to RENALIMP_MOD and RENALIMP_SEV; the trio decomposes a four-level renal-function classification (normal / mild / moderate / severe) into three binary indicators with normal as the implicit reference. Follows the HEPIMP_* family precedent (mild / moderate / severe as parallel binary canonicals). Companion to continuous CRCL. Founded alongside the CarlssonPetri_2018_semaglutide extraction where all three renal-impairment categories were tested simultaneously.

RENALIMP_SEV (canonical for severe renal impairment indicator)

• Description: 1 = severe renal impairment, 0 = normal renal function or any non-severe category. Binary categorical indicator used by source papers that test renal-impairment status as a discrete covariate rather than via continuous creatinine clearance (CRCL). The classification scheme that defines “severe” is paper-specific and must be documented in per-model covariateData[[RENALIMP_SEV]]$notes. A common scheme (used by Carlsson Petri 2018 and mirrored in FDA / EMA labeling guidance for renal-impairment PK) is eGFR < 30 mL/min/1.73 m^2 as severe.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (normal renal function; typically paired with RENALIMP_MILD and RENALIMP_MOD so the trio of indicators is mutually exclusive and normal renal function is the shared reference).
• Source aliases:
  • RENAL (multi-level categorical with severe as one level; decompose to binary indicators) – used in CarlssonPetri_2018_semaglutide.R.
• Example models: CarlssonPetri_2018_semaglutide.R (multiplicative CL/F ratio 0.920 for severe renal impairment vs normal reference, encoded as e_renal_sev_cl^RENALIMP_SEV per Table S3; the Discussion notes the severe stratum came only from SUSTAIN 6 and might be confounded by trial effect, and that a dedicated renal-impairment clinical pharmacology trial found no clinically relevant effect).
• Notes: Sister canonical to RENALIMP_MILD and RENALIMP_MOD; the trio decomposes a four-level renal-function classification (normal / mild / moderate / severe) into three binary indicators with normal as the implicit reference. Follows the HEPIMP_* family precedent. Companion to continuous CRCL. Founded alongside the CarlssonPetri_2018_semaglutide extraction.

RENALIMP_ESRD (canonical for end-stage renal disease indicator)

• Description: 1 = end-stage renal disease (ESRD), 0 = any renal function above ESRD. Fourth and most severe member of the RENALIMP_* binary renal-function family, for source papers that carry ESRD as its own discrete covariate rather than letting continuous CRCL extrapolate into the anuric range. The threshold that defines ESRD is paper-specific and must be documented in per-model covariateData[[RENALIMP_ESRD]]$notes; a common scheme (used by Xie 2025 and mirrored in FDA / EMA renal-impairment labeling guidance) is CrCL < 15 mL/min.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (any renal function above ESRD; typically paired with RENALIMP_MILD / RENALIMP_MOD / RENALIMP_SEV so the set of indicators is mutually exclusive and normal renal function is the shared reference).
• Source aliases:
  • ESRD – used in Xie_2025_aztreonam_avibactam.R (Table S3 row ESRD on CL_AVI).
• Example models: Xie_2025_aztreonam_avibactam.R (proportional shift of -0.923 on avibactam clearance, leaving 7.7% of the reference value; aztreonam carries no ESRD term).
• Notes: Sister canonical to RENALIMP_MILD, RENALIMP_MOD and RENALIMP_SEV, completing the four-level family; general scope for the same reason those carry it, namely that the renal-impairment strata have a standard regulatory definition any future model can reuse. Two traps worth documenting per model. First, an ESRD indicator commonly replaces rather than multiplies the continuous CRCL relationship – Das 2024 prints its avibactam power arm as applying when “nCrCL < 80 mL/min and not ESRD”, and multiplying the two in Xie 2025 would give a physiologically impossible ~0.7% of reference clearance instead of 7.7%. Second, papers routinely define the ESRD stratum on raw CrCL (mL/min) while the model’s continuous renal covariate is the BSA-normalized CRCL (mL/min/1.73 m^2), so the column is generally not derivable from CRCL by a threshold; supply it from the raw value. Distinct from RRT_HEMODIAL_STATUS / RRT_HEMODIAL_ACTIVE, which record dialysis treatment rather than the underlying renal-function stratum – a model may carry both, and Xie 2025 makes them mutually exclusive because dialysis patients take a separate absolute clearance.

RENAL_ARC (canonical for augmented-renal-clearance indicator)

• Description: 1 = the subject meets the study’s definition of augmented renal clearance (ARC), a supraphysiologic glomerular filtration rate seen in younger, less comorbid critically ill and infected patients; 0 = not augmented. Counterpart at the opposite pole of the RENALIMP_* family, for source papers that carry ARC as its own discrete covariate rather than letting the continuous CRCL relationship extrapolate into the supranormal range. ARC is conventionally established from a measured clearance (a timed urine collection), not from a Cockcroft-Gault or MDRD estimate, so the column is generally not derivable from CRCL by a threshold and must be supplied from the measured value; the threshold and the assay must both be documented in per-model covariateData[[RENAL_ARC]]$notes.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (renal function not augmented). Where a source paper records the flag in only one contributing study, every subject outside that study takes the reference level, which must be stated per model.
• Source aliases:
  • ARC – data-set column name in the Li 2019 avibactam NONMEM control stream (supplementary Data S1, CLCLCR block); used in Li_2019_avibactam.R.
• Example models: Li_2019_avibactam.R (founding example; ARC defined as a measured creatinine clearance >= 140 mL/min by 8-hour urine collection and recorded only in study CXL-PK-04. The flag does two structurally distinct things: it scales the supra-80 CRCL slope on clearance by 0.992, and it EXCLUDES the subject from the sub-80 power arm, so an ARC subject with CRCL < 80 takes a renal factor of exactly 1).
• Notes: Scope specific until a second model ratifies it. Deliberately NOT folded into CRCL, and deliberately not expressed as a RENALIMP_* level: the RENALIMP_* family partitions reduced renal function into severity strata that share one reference, whereas ARC is a separate binary flag that coexists with the continuous renal covariate and modifies how that covariate acts rather than replacing it. Distinct from a high CRCL value. The two are not interchangeable even when a paper’s ARC threshold is expressed in mL/min, because ARC is established by a measured clearance while CRCL normally holds an estimate; a subject can have an estimated CRCL above the ARC threshold without being flagged, and a flagged subject’s estimated CRCL can fall below it. Where a source paper’s ARC flag interacts with the CRCL relationship (as in the founding model, where it both rescales one arm and gates the other), encode both behaviours – a model that reproduces only the rescaling will silently mis-predict ARC subjects at low estimated clearance.

CPK (canonical for serum creatine phosphokinase / creatine kinase)

• Description: Serum creatine phosphokinase (also called creatine kinase, CK) activity (baseline or time-varying). Skeletal-muscle / cardiac-muscle injury and turnover marker; in macrophage-targeted PK/PD analyses (axatilimab, anti-CSF-1R) it is interpreted as a Kupffer-cell / tissue-macrophage clearance surrogate because Kupffer cells participate in the elimination of circulating muscle-derived enzymes.
• Units: U/L (SI; canonical as of 2026-06-19; IU/L is used interchangeably). Document per-model via covariateData[[CPK]]$units.
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (CPK / ref)^exponent. Reference values observed: 63 U/L (Yang 2024 axatilimab; pooled-cohort median).
• Source aliases:
  • BLCPK (baseline CPK) – informal usage in Yang 2024.
• Example models: Yang_2024_axatilimab.R (baseline-only covariate on baseline NCMC concentration BL_NCMC with power exponent 0.376; reference 63 U/L).
• Notes: Muscle-origin enzyme distinct from AST / ALT (hepatic) and LDH (general tissue turnover). Yang 2024 uses CPK alongside AST and LDH as tracked safety biomarkers. Per-model covariateData[[CPK]]$notes should document baseline-vs-time-varying status and the clinical interpretation in the source population (skeletal-muscle injury, macrophage-clearance surrogate, or both). Distinct from any model state variable representing CPK time-course dynamics – covariate column is the pre-dose laboratory observation.

MYO (canonical for plasma myoglobin)

• Description: Plasma or serum myoglobin concentration, an oxygen-binding haem protein of skeletal and cardiac muscle used clinically as a muscle-damage / muscle-mass marker. Baseline or time-varying; document which via per-model covariateData[[MYO]]$notes.
• Units: ng/mL (equivalently ug/L; SI)
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (MYO / ref)^exponent. Reference values observed: 25 ng/mL (Zhao 2025 paracetamol; pooled-cohort mean).
• Source aliases:
  • myoglobin – spelled-out form used in Zhao_2025_paracetamol.R (Table 1 baseline characteristics, Table 2 covariate rows).
• Example models: Zhao_2025_paracetamol.R (power exponent -1.10 on the paracetamol leftover clearance CL_p/F, normalised to the pooled mean of 25 ng/mL; baseline medians 34 ng/mL in healthy controls vs 17 ng/mL in spinal-muscular-atrophy patients, with the paper quoting a normal range of 15-75 ng/mL, and the effect carrying dOFV = 18.5 as the strongest covariate in that analysis).
• Notes: Sibling of the existing muscle / tissue-turnover markers CPK and LDH, and distinct from both – myoglobin is a protein mass concentration in ng/mL while CPK and LDH are enzyme activities in U/L, and Zhao 2025 screens all three independently (only myoglobin is retained). Scope: general because the concept is assay-standard and paper-independent, matching the other routine plasma clinical-chemistry analytes ALB, CRP, CREAT, HGB, ALT, AST, TBILI, ALP, LDH, and CPK. In muscle-wasting populations myoglobin is low because muscle mass is low, so a negative exponent on a clearance term should be read as a muscle-mass surrogate rather than as muscle injury; the direction of the clinical interpretation is paper-specific and belongs in per-model notes. Distinct from any model state variable representing a myoglobin time course – the covariate column is the laboratory observation.

BFR (canonical for blood flow rate through the extracorporeal circuit during dialysis)

• Description: Instantaneous blood flow rate through the extracorporeal circuit during an active dialysis session. Time-varying within subject; meaningful only when RRT_HEMODIAL_ACTIVE = 1 – in the interdialytic period the value is sentinel and the Michaels-equation term is gated off by RRT_HEMODIAL_ACTIVE.
• Units: mL/min
• Type: continuous
• Scope: general
• Reference category: n/a – enters the Michaels equation together with DFR and a hemodialyzer mass-transfer-area coefficient. Values investigated in the ratification source were 200, 300, and 400 mL/min (Liesenfeld 2013 Methods, Study Design; Table 1). A second registered role is as the plasma-flow denominator of a post-filter concentration observable, C_Post = C_Pre * (1 - CL_CRRT / (BFR * BPR)), where the blood-to-plasma ratio BPR converts the charted whole-blood pump flow to the plasma flow actually presented to the membrane.
• Source aliases:
  • phi_Blood – Butragueno-Laiseca 2022 supplementary material eq. 4 and the sibling Butragueno-Laiseca 2025 teicoplanin analysis; same orientation, no transformation.
• Example models: Liesenfeld_2013_dabigatran.R (Michaels-equation role), ButraguenoLaiseca_2022_piperacillin.R (post-filter plasma-flow denominator; paediatric CVVHDF, supplementary Table 3 blood flow 70 (SD 31) mL/min. Note that this paper’s main-text Results misprint the same quantity weight-normalised as “5 (SD 2) ml/kg/h” – the per-minute reading is the one that reproduces the paper’s own published extraction ratio, see the model file and vignette Errata).
• Notes: Pairs with RRT_HEMODIAL_ACTIVE (binary on/off gate) and DFR (dialysate flow rate).

DFR (canonical for dialysate flow rate through the extracorporeal circuit during dialysis)

• Description: Instantaneous dialysate flow rate through the extracorporeal circuit during an active dialysis session. Time-varying within subject; meaningful only when RRT_HEMODIAL_ACTIVE = 1.
• Units: mL/min
• Type: continuous
• Scope: general
• Reference category: n/a – enters the Michaels equation together with BFR. The ratification source fixed DFR at 700 mL/min throughout (Liesenfeld 2013 Methods, Study Design) and additionally simulated 500 mL/min (Methods, Simulations).
• Source aliases: none known.
• Example models: Liesenfeld_2013_dabigatran.R.
• Notes: Pairs with RRT_HEMODIAL_ACTIVE and BFR.

VASCACC_AVF1N (canonical for single-needle arteriovenous-fistula vascular-access indicator)

• Description: 1 = the haemodialysis session was performed through an arteriovenous fistula cannulated in a SINGLE-needle setup (one needle alternately draws and returns blood, so a fraction of already-dialysed blood is recirculated); 0 = any other access type. Per-subject (or per-session) indicator describing how the patient is cannulated for extracorporeal therapy, as distinct from the circuit-setting covariates BFR / DFR / Q_CVVH (flow rates) and the therapy-presence flags RRT_HEMODIAL_STATUS / RRT_HEMODIAL_ACTIVE. Access type matters pharmacokinetically because it changes the effective solute presentation to the dialyser: a single-needle fistula recirculates blood, so the nominal pump blood flow rate overstates the flow of undialysed blood reaching the membrane.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (tunnelled dialysis catheter, TDC), shared with the sister canonical VASCACC_AVF2N; the pair decomposes the three-level access classification (TDC / AVF single-needle / AVF double-needle) into two binary indicators with TDC as the implicit reference (both indicators 0).
• Source aliases:
  • CVC0SN1DN2 (three-level integer selector, 0 = TDC, 1 = single-needle AVF, 2 = double-needle AVF; decompose to binary indicators) – used in Kong_2025_piperacillin_tazobactam.R (ESM NONMEM $PK: LGT_TAZ = THETA(18); IF(CVC0SN1DN2.EQ.1) LGT_TAZ = THETA(19); IF(CVC0SN1DN2.EQ.2) LGT_TAZ = THETA(20)).
  • AVF 1N – Kong 2025 Table 1 / Table 3 row label (“AVF 1N single-needle arteriovenous fistula”).
• Example models: Kong_2025_piperacillin_tazobactam.R (selects the tazobactam dialyser extraction ratio: 73.9% for single-needle AVF versus the 80.1% TDC reference and 73.5% for double-needle AVF, Table 3; retained in backward elimination at p < 0.001, dOFV -22.88).
• Notes: Sister canonical to VASCACC_AVF2N. Decomposed-binary-indicator encoding follows the register’s settled convention for multi-level categoricals (RENALIMP_MILD / RENALIMP_MOD / RENALIMP_SEV, the parallel HEPIMP_* family, and RACE_<GROUP>); an integer selector column was explicitly rejected because it would imply an ordering between access types that does not exist. Kong 2025 additionally applied a time-averaged blood flow rate for single-needle sessions when computing BFR (Methods 2.4), so a model that carries both VASCACC_AVF1N and BFR should document in covariateData notes whether the recirculation correction has already been folded into the supplied BFR values. General scope because vascular access type is a universally recorded haemodialysis patient characteristic and is a plausible determinant of effective dialyser clearance in any ESKD antibiotic popPK.

VASCACC_AVF2N (canonical for double-needle arteriovenous-fistula vascular-access indicator)

• Description: 1 = the haemodialysis session was performed through an arteriovenous fistula cannulated in a DOUBLE-needle setup (separate arterial-draw and venous-return needles, so there is no single-needle recirculation); 0 = any other access type. Per-subject (or per-session) indicator describing how the patient is cannulated for extracorporeal therapy; see the sister entry VASCACC_AVF1N for the family rationale and the distinction from the circuit-flow and therapy-presence covariates.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (tunnelled dialysis catheter, TDC), shared with VASCACC_AVF1N; both indicators 0 selects the TDC reference level.
• Source aliases:
  • CVC0SN1DN2 (three-level integer selector, level 2; decompose to binary indicators) – used in Kong_2025_piperacillin_tazobactam.R.
  • AVF 2N – Kong 2025 Table 1 / Table 3 row label (“AVF 2N double-needle arteriovenous fistula”).
• Example models: Kong_2025_piperacillin_tazobactam.R (selects the tazobactam dialyser extraction ratio: 73.5% for double-needle AVF versus the 80.1% TDC reference, Table 3).
• Notes: Sister canonical to VASCACC_AVF1N; the two are mutually exclusive (a subject with both set to 1 is invalid). Kong 2025 found the access-type effect statistically significant but clinically limited (Discussion: “the difference in extraction ratio between the vascular access types were limited and therefore not expected to be clinically relevant”), and the analogous access-type effect on the PIPERACILLIN extraction ratio did NOT survive backward elimination (dOFV +11.26), so piperacillin keeps a single 64.0% extraction ratio.

ECMO_PUMP_SPEED (canonical for extracorporeal-membrane-oxygenation centrifugal-pump rotational speed)

• Description: Rotational speed of the extracorporeal-membrane-oxygenation (ECMO) centrifugal blood pump during VA-ECMO or VV-ECMO support. Continuous covariate; treated as time-fixed per subject in Yang 2017 (the per-subject pump speed reported in the source data was the prevailing speed during the PK sampling window; pump-speed adjustments during sampling were not modelled as time-varying). For future models that resolve session-level changes in pump speed, the covariate is naturally time-varying and the per-model covariateData[[ECMO_PUMP_SPEED]]$notes should document the time resolution.
• Units: RPM (revolutions per minute)
• Type: continuous
• Scope: general
• Reference category: n/a – enters as a power-centered effect (ECMO_PUMP_SPEED / ref)^exponent. The reference value is paper-specific (median pump speed in the source cohort): Yang 2017 uses 2350 RPM (cohort median; Table 1 / Results: “median ECMO pump speeds of 2350 RPM”).
• Source aliases: none known.
• Example models: Yang_2017_remifentanil.R (power effect on remifentanil CL: (ECMO_PUMP_SPEED / 2350)^2.04; higher pump speed associated with higher CL, hypothesised mechanism is increased spontaneous drug degradation at high centrifugal-pump shear).
• Notes: Distinct from blood flow rate (BFR, mL/min) and from dialysate flow rate (DFR, mL/min) which characterise renal-replacement-therapy circuits. ECMO circuits use a centrifugal pump whose rotational speed sets the cardiac-output augmentation; the resulting blood flow rate (LPM) is a separate measured quantity that depends on circuit resistance and patient hemodynamics. Yang 2017 tested both ECMO pump speed (RPM) and ECMO flow rate (LPM) and only pump speed was retained as a significant covariate (ECMO flow rate was not significantly associated with PK parameters).

Q_CVVH (canonical for continuous venovenous hemofiltration circuit flow rate during ECMO)

• Description: Flow rate through the continuous venovenous hemofiltration (CVVH) filter when it is placed parallel to the ECMO circuit. The filter contributes solute removal that adds to the drug’s apparent CL whenever CVVH is active; the column is zero in subjects whose ECMO circuit has no CVVH filter or whose CVVH is currently off. Continuous covariate; naturally time-varying although the source paper treats the per-subject median as time-fixed (the median Q_CVVH per individual did not vary much during sampling).
• Units: mL/min
• Type: continuous
• Scope: general
• Reference category: n/a – enters as a power-centered effect (Q_CVVH / ref)^exponent with the source paper additionally gating the effect to 1 when Q_CVVH = 0 (no CVVH active). The reference value is paper-specific (cohort median during CVVH operation): Ahsman 2010 uses 193 mL/min (cohort median; Table 1).
• Source aliases:
  • QCVVH – used in Ahsman_2010_cefotaxime.R (paper notation, subscript dropped).
• Example models: Ahsman_2010_cefotaxime.R (power effect on DACT metabolite CL: (Q_CVVH / 193)^0.72 when CVVH active, 1 otherwise).
• Notes: Distinct from ECMO_PUMP_SPEED (the centrifugal-pump rotational speed setting cardiac-output augmentation) and from BFR / DFR (blood-flow / dialysate-flow rates characterising standalone hemodialysis circuits): Q_CVVH is the slow-clearance flow through a CVVH filter integrated INTO an ECMO circuit. Distinct from RRT_CRRT_STATUS (binary on / off indicator for continuous RRT) which captures presence-of-therapy rather than its rate. Future ECMO + CVVH popPK extractions that quantify CVVH flow should reuse this canonical and extend the example list.

Q_ECMO (canonical for extracorporeal-membrane-oxygenation circuit blood flow rate)

• Description: Blood flow rate delivered through the extracorporeal-membrane-oxygenation (ECMO) circuit – the volume of blood per unit time actually passing through the pump head and oxygenator, as read from the circuit flow probe. This is the delivered flow, a measured quantity distinct from the pump-speed setting that produces it: for a given centrifugal-pump speed the resulting flow depends on circuit resistance, cannula size, preload, and patient hemodynamics. Continuous covariate and naturally time-varying, because flow is deliberately titrated during ECMO support and stepped down during weaning; a source paper that reports only a per-subject representative flow during the PK sampling window is treating it as time-fixed, and each model must say which applies in covariateData[[Q_ECMO]]$notes.
• Units: L/min
• Type: continuous
• Scope: general
• Reference category: n/a – enters as a median-centered effect. Kim 2026 uses the median-normalized proportional form 1 + theta * (Q_ECMO - ref) / ref on 1-OH-midazolam clearance with ref = 2.7 L/min. The reference value is paper-specific (cohort median flow); note that Kim 2026 does not report its own median, so the 2.7 L/min there is back-solved from the paper’s Figure 1A rather than read from a table (see that model’s covariateData notes and vignette Errata).
• Source aliases:
  • ECMO flow rate – Kim 2026 paper notation (main text, Table 1 theta ECMO flow rate on CL1-OH MDZ, and Supplementary Methods “Covariate model development”).
  • ECMO flow rate (LPM) – Yang 2017 notation for the same quantity (tested, not retained).
• Example models: Kim_2026_midazolam_ecmo.R (median-normalized proportional effect on 1-OH-midazolam clearance: 1 + 0.336 * (Q_ECMO - 2.7) / 2.7; higher ECMO flow associated with higher metabolite clearance, so a higher midazolam infusion rate is needed to hold the 24-h concentration), Kang_2026_ticagrelor.R (power effect on the ticagrelor apparent central volume, 0.518^(Q_ECMO / 3), nested inside the ECMO_STATUS indicator so it applies only while cannulated; the 3 L/min reference is printed in the paper’s own final-model equation rather than being a cohort median, and because 0.518 < 1 a HIGHER flow gives a SMALLER volume – the opposite sign to Kim 2026’s clearance effect. The first register example where the flow term is gated by the ECMO indicator; the paper’s main text prints it ungated, which is algebraically identical only when off-ECMO flow is carried as 0).
• Notes: Named in the Q_ circuit-flow family (alongside Q_CVVH, Q_TOTAL_LPM, QBL / QEFF) rather than the device-prefixed ECMO_ family (ECMO_PUMP_SPEED, T_ECMO, T_POST_ECMO): Q_CVVH is already registered as the blood flow of a CVVH extracorporeal circuit, and ECMO circuit blood flow is that same physical quantity for a different circuit, so Q_ECMO is its exact sibling. The ECMO_ prefix groups by device, which is a different axis; a flow rate belongs with flow rates. Explicitly distinct from ECMO_PUMP_SPEED (RPM – the pump setting, whose own Notes already flag delivered flow as “a separate measured quantity”), from Q_CVVH (mL/min through a CVVH filter placed parallel to an ECMO circuit), and from BFR / DFR (mL/min through a standalone dialysis circuit). Both Q_ECMO and ECMO_PUMP_SPEED can appear in the same dataset and the same model; Yang 2017 recorded both and retained only pump speed, whereas Kim 2026 recorded both and retained only flow rate. Note the unit difference from the dialysis-circuit flows: ECMO flows are conventionally reported in L/min, dialysis flows in mL/min.

T_ABT (canonical for time elapsed on hormonal add-back therapy)

• Description: Time since the start of hormonal add-back therapy (ABT) – an oestrogen plus progestogen, or a progestogen alone, co-administered with a GnRH agonist or antagonist to partially replenish circulating oestrogen and so prevent bone-mineral-density loss and vasomotor symptoms during hypoestrogenaemia. 0 for subjects never receiving add-back therapy, which zeroes any term built on it.
• Units: weeks. Document per-model via covariateData[[T_ABT]]$units; days is the other common reporting unit.
• Type: continuous
• Scope: specific
• Reference category: 0 (no add-back therapy). In Pohl 2022 this is the entire EDELWEISS phase 2b population and therefore the whole dose-selection analysis; add-back therapy (1 mg oestradiol / 0.5 mg norethisterone acetate) was given only in the phase 1 studies 16-OBE2109-011 and 17-OBE2109-008, which used concurrent and delayed add-back arms respectively.
• Source aliases:
  • add-back therapy / ABT – Pohl 2022 Table 1 and Table 5 (“E2 increase rate on add-back therapy”).
• Example models: Pohl_2022_linzagolix_e2.R (additive oestradiol increment of 1.58 pg/mL per week on add-back therapy).
• Notes: Member of the T_<event> time-since-event family (T_NUT_SUPP, T_ECMO, T_POST_ECMO, T_POST_HEMODIAL). A time-since-event column rather than a binary flag because the published Pohl 2022 effect is a RATE (pg/mL per week), so the covariate must carry elapsed time; a bare on/off indicator cannot express it. This also handles the delayed-add-back arm of 17-OBE2109-008 correctly, where dosing of linzagolix and of add-back therapy start at different times. Specific scope because the column is meaningful only for GnRH-modulator models in endometriosis and uterine fibroids. CAUTION for anyone reusing this column from Pohl 2022: the functional form of the add-back term is not printed anywhere in that source – the Supporting Information equation omits the term entirely and only the tabulated pg/mL/week unit implies an additive linear-in-time increment.

T_ECMO (canonical for time since ECMO cannulation start (beginning of extracorporeal circulation))

• Description: Wall-clock time elapsed since the patient was cannulated onto ECMO (the beginning of extracorporeal circulation). Zero before cannulation; positive and monotonically increasing throughout the ECMO support period. Time-varying. Whether the value continues to increase after decannulation, resets at decannulation, or is censored at decannulation is paper-specific and must be documented in covariateData[[T_ECMO]]$notes; when the source model uses a saturating Hill / Emax form (Kleiber 2017) it does not matter whether T_ECMO continues past decannulation because the effect is already saturated.
• Units: hour
• Type: continuous
• Scope: general
• Reference category: n/a – enters as a sigmoidal Emax multiplier 1 + Emax * T_ECMO^gamma / (T50_EC^gamma + T_ECMO^gamma) (Kleiber 2017 Eq 9) or a saturating half-life form (Ahsman 2010 Eq 10-style). At T_ECMO = 0 (pre-ECMO baseline) the Hill ratio evaluates to 0 and the multiplier collapses to 1.
• Source aliases:
  • tec / t_EC – used in Kleiber_2017_clonidine.R (paper notation, Eq 9 and Table 3).
• Example models: Kleiber_2017_clonidine.R (sigmoidal Emax effect on V: V x (1 + 0.55 * T_ECMO^18.5 / (51.7^18.5 + T_ECMO^18.5)); V approaches 1.55 * V_pop * (WT/70) on full ECMO and recovers to V_pop * (WT/70) before ECMO).
• Notes: Conceptually the on-ECMO mirror of T_POST_ECMO – T_ECMO measures time since cannulation, T_POST_ECMO measures time since decannulation. A model that uses BOTH typically gates each by an additional ECMO-on indicator or by the natural floor (T_ECMO = 0 before cannulation, T_POST_ECMO = 0 before-and-during ECMO). Kleiber 2017 uses only T_ECMO; Ahsman 2010 uses only T_POST_ECMO. Distinct from on-ECMO duration covariates (per-run summary of total ECMO hours) and from ECMO_PUMP_SPEED (instantaneous centrifugal-pump rotational speed). Anticipated by the existing T_POST_ECMO entry’s Notes (which referenced t_EC as “tested but not retained” in Ahsman 2010); ratified canonically alongside the Kleiber 2017 clonidine extraction where the covariate is retained.

T_POST_ECMO (canonical for time after ECMO decannulation (end of extracorporeal circulation))

• Description: Wall-clock time elapsed since the patient was decannulated from ECMO (the end of extracorporeal circulation). Zero before and during ECMO support; becomes positive after decannulation and increases linearly with time. Naturally time-varying. Used as a power-centred recovery-time covariate on PK parameters whose values change as the patient transitions off ECMO back to native circulation.
• Units: hour
• Type: continuous
• Scope: general
• Reference category: n/a – enters as a power-centered effect (T_POST_ECMO / ref)^exponent with the source paper additionally gating the effect to 1 when T_POST_ECMO = 0 (before or during ECMO). The reference value is paper-specific: Ahsman 2010 uses 100 h (Appendix).
• Source aliases:
  • tEND – used in Ahsman_2010_cefotaxime.R (paper notation, subscript dropped).
• Example models: Ahsman_2010_cefotaxime.R (power effect on parent CTX CL: (T_POST_ECMO / 100)^0.16 when post-ECMO, 1 otherwise; power effect on DACT metabolite CL: (T_POST_ECMO / 100)^0.53 when post-ECMO, 1 otherwise).
• Notes: Conceptually a time-since-event covariate similar to TPP (time postpartum) and T_ENTRY (per-subject study-entry time) but anchored to ECMO decannulation. Distinct from t_EC (time since ECMO START / beginning of extracorporeal circulation, which the source paper tested but did not retain) and from on-ECMO duration covariates. The gating-at-zero convention is structural: during ECMO the recovery-time covariate is undefined / not yet started, so the (0 / ref)^exponent expression would collapse pathologically and the implementer must short-circuit the multiplier to 1. Future ECMO popPK extractions that include a post-decannulation recovery phase should reuse this canonical and extend the example list.

T_HEMODIAL_INIT (canonical for time since the initiation of maintenance hemodialysis therapy (“dialysis vintage”))

• Description: Elapsed time since the subject was first established on maintenance intermittent hemodialysis for end-stage kidney disease – the quantity nephrology calls dialysis vintage. A per-subject baseline duration measured in months or years, NOT a within-record clock: it is fixed for the duration of a typical popPK study. Its pharmacokinetic role is as a surrogate for residual renal function, which decays progressively after dialysis is started and which cannot be measured directly in maintenance-dialysis patients because serum creatinine there is dominated by time since the last session. A drug with a residual-renal elimination route therefore shows FALLING interdialytic clearance as this covariate rises.
• Units: month. Document per-model via covariateData[[T_HEMODIAL_INIT]]$units when a paper reports years or days.
• Type: continuous
• Scope: general
• Reference category: n/a – supplied as a duration and normally entered as a centred inverse-power term against the cohort median. Duke_2024_cefazolin.R uses (59 / T_HEMODIAL_INIT)^0.28, where 59 months is that cohort’s median. Must be strictly positive whenever it enters a denominator.
• Source aliases:
  • TOH – “time on haemodialysis”, the Duke 2024 Methods and Table 1 label and the symbol used in its Table S3 Pmetrics model file. Read the definition, not the abbreviation: “time ON haemodialysis” here means months since therapy was STARTED, not minutes elapsed within the current session.
• Example models: Duke_2024_cefazolin.R (inverse-power effect on the interdialytic unbound-cefazolin clearance arm, CL = CLnHD * (59 / TOH)^0.28, exponent hard-coded in the Table S3 model file; reported inverse-power fit r^2 = 0.433; the paper’s PTA simulations sweep TOH = 6, 12, 24, 36 and 60 months and PTA rises monotonically with it).
• Notes: Third and slowest member of the intermittent-hemodialysis time family, and the one that must not be confused with the other two. RRT_HEMODIAL_ACTIVE is the seconds-to-hours gate saying whether a session is running right now; T_POST_HEMODIAL is the hours-scale interdialytic clock measuring time since the last session ENDED and resetting at every session; T_HEMODIAL_INIT is the months-to-years clock measuring time since the THERAPY began and never resets. The three are dimensionally and physiologically distinct – session gating, interdialytic fluid accumulation, and progressive loss of residual renal function respectively – so a value from one can never substitute for another. Named in the T_<event> elapsed-time family alongside T_NUT_SUPP (time since start of nutritional supplementation) and T_ABT (time since start of add-back therapy), both of which likewise measure time from the initiation of an ongoing therapy; the explicit _INIT token is what keeps it unambiguous against a “duration of the current session” reading, which a bare T_HEMODIAL would invite. Duke 2024 states this covariate had not previously been included in a pharmacokinetic model for patients requiring intermittent hemodialysis, so this entry has one founding example; a future paper reporting it should reuse this column and record its functional form in covariateData[[T_HEMODIAL_INIT]]$notes. A peritoneal-dialysis or continuous-RRT paper reporting the same quantity should register a modality-matched sibling rather than reusing this column, following the RRT_<modality>_<kind> precedent.

T_POST_HEMODIAL (canonical for time since the end of the last intermittent-hemodialysis session)

• Description: Wall-clock time elapsed since the END of the subject’s most recent intermittent-hemodialysis (IHD) session. Resets to zero at the end of each session and increases through the interdialytic interval, so it is naturally time-varying within subject. Undefined for subjects never dialysed; set to 0 for them and pair the effect with RRT_HEMODIAL_STATUS so the value is inert outside the dialysis-dependent stratum. Captures interdialytic physiology that the on/off session gate cannot – principally fluid accumulation between sessions, which expands the central volume of distribution.
• Units: hour
• Type: continuous
• Scope: general
• Reference category: n/a – enters either as a threshold indicator (T_POST_HEMODIAL > <cut>) or, if a future paper resolves it, as a continuous centred / power effect. VanWart_2025_telavancin.R uses the threshold form at 48 h.
• Source aliases: none known; Van Wart 2025 describes the covariate in prose (“PK samples were collected more than 48 hours after the last active IHD session”) rather than naming a data-set column.
• Example models: VanWart_2025_telavancin.R (48-hour threshold gating a fixed additive central-volume increase of +1.55 L in dialysis-dependent subjects: vc <- ... + e_t_post_hemodial_vc * RRT_HEMODIAL_STATUS * (T_POST_HEMODIAL > 48); Van Wart 2025 attributes the rise to “fluid depletion during IHD and accumulation between sessions”, telavancin plasma concentrations being higher immediately off IHD and falling thereafter).
• Notes: The INTERDIALYTIC time axis of the intermittent-hemodialysis covariate set: RRT_HEMODIAL_STATUS says whether the subject is dialysis-dependent at all, RRT_HEMODIAL_ACTIVE says whether a session is running right now, T_POST_HEMODIAL says how long it has been since the last one ended, and T_HEMODIAL_INIT says how many months the subject has been on the therapy altogether (dialysis vintage, a residual-renal-function surrogate). Do not substitute one for another: this entry resets at every session and is measured in hours, while T_HEMODIAL_INIT never resets and is measured in months. The RRT_HEMODIAL_STATUS entry anticipated exactly this column (“a future paper that resolves drug clearance during versus between dialysis sessions would use a time-varying form (RRT_HEMODIAL_ACTIVE) or a separate per-session covariate”). Named in the T_<event> time-since-event family alongside T_POST_ECMO (time after ECMO decannulation), T_ECMO, T_CPB and T_NUT_SUPP, which is the same shape: time measured from the end of an extracorporeal therapy. Distinct from FILT_AGE_HI, which also uses a 48-hour threshold but measures the age of the DIALYSER MEMBRANE while a therapy is running (declining dialysis efficiency from membrane fouling), not the time since a session ended (interdialytic fluid accumulation) – the two thresholds coincide numerically and mean opposite things, so do not substitute one for the other. General scope because time since the last dialysis session is a paper-independent quantity any IHD popPK model may use. A future paper that retains the effect as a continuous function rather than a threshold should reuse this column and record the functional form in per-model covariateData[[T_POST_HEMODIAL]]$notes.

FILT_AGE_HI (canonical for dialysis-filter-age above-48-hour indicator)

• Description: Binary indicator that the in-use extracorporeal renal-replacement-therapy (CVVHDF / CVVH / SLED / EDD-f / IHD) hemofilter or dialyser membrane has been in continuous use for more than 48 hours at the time of dosing / sampling. 1 = filter is older than 48 h; 0 = filter is within the first 48 h of use (the fresh-filter reference). Subject-level binary in the source data (Patel 2011 had 1 of 10 patients on a > 48 h filter at the start of CVVHDF treatment; the rest had fresh filters), but the covariate is naturally time-varying within subject as the filter ages and is replaced – record the per-model time resolution in covariateData[[FILT_AGE_HI]]$notes.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (fresh filter, less than 48 h in use). Models that apply a multiplicative reduction to the dialysis clearance arm should encode the effect as cl_dialysis * e_filt_age_hi_<arm>^FILT_AGE_HI so the FILT_AGE_HI = 0 case recovers the fresh-filter baseline.
• Source aliases: none known; Patel 2011 retains the effect via the parameter ffCL_CVVHDF (a multiplicative factor on the CVVHDF clearance arm) rather than naming a data-set column. When ratified the column was assigned the canonical name FILT_AGE_HI in the model file.
• Example models: Patel_2011_fluconazole.R (multiplicative factor e_filt_age_hi_cl_renal^FILT_AGE_HI on the CVVHDF clearance arm: e_filt_age_hi_cl_renal = 0.368, so > 48 h filters give 36.8% of fresh-filter CVVHDF efficiency; bootstrap 95% CI 0.326-0.426 from Patel 2011 Table 2; Delta-OBJ = -11.46 vs base model).
• Notes: Distinct from RRT_CRRT_STATUS (subject-level binary indicator that ANY continuous / extended renal-replacement therapy is occurring) and from RRT_HEMODIAL_ACTIVE (per-time-point on/off gate for the dialysis-clearance term in within-subject time-varying dialysis-clearance models). FILT_AGE_HI gates WITHIN the CRRT-on subset by membrane age – meaningful only when RRT_CRRT_STATUS = 1 or RRT_HEMODIAL_ACTIVE = 1. The > 48 h threshold is paper-specific (Patel 2011 Methods note that membrane fouling and clot formation in the hemofilter pores progressively reduce dialysis efficiency, with the inflexion typically observed near 48 h). Future extractions that retain a finer filter-age structure (continuous filter_age_hours or multiple thresholds) should reuse FILT_AGE_HI for the binary > 48 h case and add a separate canonical for the continuous form. Scope kept specific pending a second model that ratifies the same > 48 h threshold; promote to general once corroborated.

FILT_SA_MED (canonical for medium-haemofilter membrane-surface-area indicator)

• Description: Binary indicator that the in-use extracorporeal renal-replacement-therapy haemofilter is the manufacturer’s MEDIUM membrane-surface-area size, within a paper’s discrete filter-size range. 1 = medium filter; 0 = the small (reference) filter, the large filter, or no extracorporeal therapy. Membrane surface area is the physical determinant of the diffusive/convective solute-transfer capacity of the filter, so papers that stock only a few filter sizes retain size as a categorical covariate on the extracorporeal clearance arm rather than as a continuous area. Paired with FILT_SA_LARGE; the two indicators are mutually exclusive and both are 0 for the small reference filter.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 for the COLUMN, meaning the smallest filter the source paper stocked (0.2 m^2 in both registered sources). The MODEL’s reference level, by contrast, is paper-specific and must not be assumed: Butragueno-Laiseca 2025 fixes the small filter to a multiplier of 1 and estimates the medium and large multipliers above it, whereas Butragueno-Laiseca 2022 – the same group, the same three filter sizes, the same PICU – fixes the LARGE filter to 1 and estimates the medium (0.74) and small (0.28) multipliers below it. Keep the column encoding fixed (FILT_SA_MED and FILT_SA_LARGE both 0 => small filter) and absorb the paper’s choice of reference level into the effect expression. For a small-filter reference that is cl_hemodialysis * e_filt_sa_med_<arm>^FILT_SA_MED * e_filt_sa_large_<arm>^FILT_SA_LARGE; for a large-filter reference, derive the small indicator inside model and write cl_hemodialysis * e_filt_sa_med_<arm>^FILT_SA_MED * e_filt_sa_small_<arm>^(1 - FILT_SA_MED - FILT_SA_LARGE). Both forms let every multiplier the paper prints appear verbatim in ini.
• Source aliases:
  • FILT, theta_FILT – Butragueno-Laiseca 2025 Table 2 notation for the three-level categorical covariate that these indicators decompose.
  • Filter surface area (m2) – Butragueno-Laiseca 2025 Table 1 row label.
• Example models: ButraguenoLaiseca_2025_teicoplanin.R (multiplicative factor e_filt_sa_med_cl_hemodialysis = 3.58 on the CKRT clearance arm; Butragueno-Laiseca 2025 Table 2, RSE 13%, SIR 95% CI 2.78-4.40; small-filter reference), ButraguenoLaiseca_2024_meropenem.R (same three physical filter sizes – 0.2 / 0.6 / 1.2 m^2 – but the paper references the MEDIUM filter rather than the small one, so the model uses the column only through the derived low-filter indicator filt_sa_small <- 1 - FILT_SA_MED - FILT_SA_LARGE and the medium level contributes the reference multiplier of 1; Butragueno-Laiseca 2024 Table 4, theta_FILT_Low = 0.27, theta_FILT_Med = 1 (reference), theta_FILT_High = 1.5), ButraguenoLaiseca_2022_piperacillin.R (large-filter reference: e_filt_sa_med_cl_hemodialysis = 0.74 (bootstrap 95% CI 0.6-0.9) and e_filt_sa_small_cl_hemodialysis = 0.28 (0.2-0.4) against theta_CLCKRT = 1.34 L/h for the 1.2 m^2 filter, reproducing the Results text’s per-size clearances of 1.34, 1.01 and 0.38 L/h; Butragueno-Laiseca 2022 Table 2).
• Notes: The absolute surface areas that the size labels denote are paper-specific and MUST be recorded in per-model covariateData[[FILT_SA_MED]]$notes – in the founding source, small = 0.2 m^2, medium = 0.6 m^2, large = 1.2 m^2, assigned by patient weight band (small 3-10 kg, medium 10-30 kg, large 30-60 kg). Because size is assigned by weight in paediatric practice, filter size and body weight are strongly collinear and a paper may describe one as a surrogate for the other; record which the model actually uses. Distinct from BSA (body surface area, a patient property) and from FILT_AGE_HI (membrane age, not size). Meaningful only when RRT_CRRT_ACTIVE = 1 (or the equivalent hemodialysis gate); the whole extracorporeal arm is gated off otherwise. A source paper that reports a genuinely continuous membrane surface area, or more than three sizes, should use the continuous canonical FILT_SA rather than extending the indicator set indefinitely (that canonical was founded on this instruction by Berthaud_2025_cefazolin.R, whose six paediatric patients span six distinct areas in a power model). Butragueno-Laiseca 2025 also tested effluent flow and blood flow on the extracorporeal clearance arm and could not retain them alongside filter size because of strong mutual correlation, so filter size is doing the work those flows would otherwise do. Scope kept specific pending a second model that ratifies the same three-level decomposition.

FILT_SA_LARGE (canonical for large-haemofilter membrane-surface-area indicator)

• Description: Binary indicator that the in-use extracorporeal renal-replacement-therapy haemofilter is the manufacturer’s LARGE membrane-surface-area size. 1 = large filter; 0 = the small (reference) filter, the medium filter, or no extracorporeal therapy. The large-size partner of FILT_SA_MED; see that entry for the full rationale.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 for the COLUMN, meaning the smallest filter the source paper stocked (0.2 m^2 in both registered sources). The MODEL’s reference level is paper-specific – Butragueno-Laiseca 2022 makes the LARGE filter its reference (multiplier fixed to 1) even though FILT_SA_LARGE = 1 still denotes that filter. See the FILT_SA_MED entry for both encodings.
• Source aliases:
  • FILT, theta_FILT – Butragueno-Laiseca 2025 Table 2 notation.
  • Filter surface area (m2) – Butragueno-Laiseca 2025 Table 1 row label.
• Example models: ButraguenoLaiseca_2025_teicoplanin.R (multiplicative factor e_filt_sa_large_cl_hemodialysis = 5.04 on the CKRT clearance arm; Butragueno-Laiseca 2025 Table 2, RSE 14%, SIR 95% CI 3.80-6.26; only 2 of the 12 CKRT patients used a large filter), ButraguenoLaiseca_2024_meropenem.R (multiplicative factor e_filt_sa_large_cl_hemodialysis = 1.5 on the CKRT clearance arm – a much smaller effect than the sibling’s 5.04 because it is measured against the MEDIUM reference rather than the small one, not because the filters differ; Butragueno-Laiseca 2024 Table 4, theta_FILT_High = 1.5, RSE 23%; 2 of the 7 CKRT patients used the high filter), ButraguenoLaiseca_2022_piperacillin.R (the 1.2 m^2 filter is the REFERENCE level there – theta_FILT_High = 1 Fixed – so no e_filt_sa_large_* parameter appears in ini(); the indicator is still required, because it is what distinguishes the reference large filter from the small filter that both indicators being 0 would otherwise denote; 3 of the 13 CKRT patients used a large filter).
• Notes: Mutually exclusive with FILT_SA_MED. A source paper may reference any of the three sizes: the sibling teicoplanin model references the small filter (so FILT_SA_MED and FILT_SA_LARGE carry both non-reference multipliers), whereas ButraguenoLaiseca_2024_meropenem.R references the medium filter (so the LOW level carries a multiplier and the model derives its indicator as 1 - FILT_SA_MED - FILT_SA_LARGE). The data columns are identical either way – each indicator always means “this filter is the medium / large one” – and only the model parameterisation’s reference level differs; encode whichever reference the paper printed so its published thetas stay verbatim. See the FILT_SA_MED entry for the reference-category encoding, the requirement to record the paper-specific absolute surface areas in per-model notes, the distinction from BSA and FILT_AGE_HI, and the collinearity with body weight and circuit flows. Scope kept specific pending a second ratifying model.

FILT_SA (canonical for continuous haemofilter / dialyser membrane surface area)

• Description: Membrane surface area of the extracorporeal renal-replacement-therapy filter (dialyser / haemofilter) in use, as a continuous quantity in m^2. Membrane surface area is the physical determinant of the diffusive and convective solute-transfer capacity of the filter, so it is the natural driver of an extracorporeal clearance arm. This is the continuous counterpart of the discrete FILT_SA_MED / FILT_SA_LARGE indicator pair: use those when a paper stocks only two or three filter sizes and retains size as a categorical covariate, and use FILT_SA when a paper spans enough distinct areas to fit a continuous (typically power) relationship. The FILT_SA_MED entry anticipated this split explicitly (“A source paper that reports a genuinely continuous membrane surface area, or more than three sizes, should add a continuous canonical rather than extending the indicator set indefinitely”); Berthaud 2025 is that paper.
• Units: m^2
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as a power-centred ratio (FILT_SA / ref)^e_filt_sa_<arm> on the extracorporeal clearance arm, and is meaningful only while RRT_HEMODIAL_ACTIVE = 1 (or the equivalent CRRT gate). Reference values observed: 1 m^2 (Berthaud 2025, the area the published dialysis-clearance estimate is standardized to).
• Source aliases:
  • DMSA, Dialysis membrane surface area – Berthaud 2025 abstract, Table 1 row label, and Table 2 covariate-effect row.
  • Filter surface area, dialyser surface area, membrane area – variant forms used across the intermittent-hemodialysis and CRRT popPK literature.
• Example models: Berthaud_2025_cefazolin.R (founding example; power effect on the dialysis-clearance arm, cl_hemodialysis <- exp(lcl_hemodialysis) * (FILT_SA / 1)^e_filt_sa_cl_hemodialysis with e_filt_sa_cl_hemodialysis = 1.26 (RSE 1.06%) against CLdialpop = 1.98 L/h at 1 m^2; Berthaud 2025 Table 2. Six paediatric maintenance-hemodialysis patients spanning DMSA 0.2, 0.3, 1, 1, 1.5 and 1.4/1.7 m^2. Adding DMSA to the dialysis arm made allometrically scaled body weight on that arm non-significant, decreased the BIC by 6 units and drove the BSV on the arm towards 0, so the final model carries no IIV there).
• Notes: Distinct from BSA (body surface area – a patient property, not a device property). The two are related in practice but must not be conflated: paediatric dialysis prescriptions commonly SIZE the membrane from the patient’s BSA, and Berthaud 2025 exploits this in its dosing-regimen simulations by approximating FILT_SA = 0.85 * BSA with BSA = (4 * BW + 7) / (BW + 90) when a measured area is unavailable. That approximation is a simulation convenience and does NOT reproduce the fitted data (the paper’s own Table 1 areas differ from it by up to a factor of two), so record the measured area in this column and keep any BSA-based imputation in the model’s validation vignette, flagged as an assumption. The Berthaud 2025 Discussion notes 0.85 sits mid-range of the recommended 75-100% of BSA and that dialysis clearance changes proportionally under a different coefficient. Because membrane size tracks patient size in paediatric practice, FILT_SA and WT are strongly collinear – record which the model actually uses, and expect an allometric weight term on the extracorporeal arm to lose significance once FILT_SA is added (exactly what happened in the founding source). Naturally time-varying within subject when a patient’s membrane is changed mid-course (Berthaud 2025 patient 6 switched between 1.4 and 1.7 m^2). Scope kept specific pending a second model that ratifies the continuous form; promote to general once corroborated.

FILT_FLUX_LOW (canonical for low-flux dialyser-membrane indicator)

• Description: Binary indicator that the in-use dialyser / haemofilter membrane is a LOW-FLUX membrane – small pore size, low water permeability (ultrafiltration coefficient below roughly 10-12 mL/h/mmHg), and therefore poor clearance of middle-molecular-weight solutes. 1 = low-flux membrane; 0 = high-flux membrane or membrane type not recorded. Paired with FILT_FLUX_HIGH; the two are mutually exclusive and both are 0 for the “not recorded” reference, which in the founding source is by far the largest stratum. Member of the FILT_<attribute> family alongside FILT_SA / FILT_SA_MED / FILT_SA_LARGE (membrane size) and FILT_AGE_HI (membrane age); flux is the membrane’s PERMEABILITY class, orthogonal to its surface area.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 for the COLUMN. The MODEL’s reference level in a decomposed pair is the all-zero state, which for the founding source means “dialysis membrane not available” rather than a physical membrane – 424 of 459 subjects (92%). Encode as cl_dialysis * e_filt_flux_low_<arm>^FILT_FLUX_LOW * e_filt_flux_high_<arm>^FILT_FLUX_HIGH so both printed multipliers appear verbatim in ini().
• Source aliases:
  • DIAM [Low Flux] – Xu 2017 supplementary Table S2 row label for theta 13; DIAM is the source’s dialysis-membrane covariate. Same orientation, no value transformation.
• Example models: Xu_2017_daptomycin.R (multiplicative factor e_filt_flux_low_cl = 0.96 on the dialysis clearance equation only; Xu 2017 Table S2, 14% RSE; 7 of 459 subjects).
• Notes: The flux term acts only inside the founding source’s DIALYSIS clearance equation (Eq 1), not in its not-on-dialysis equation (Eq 6), because a membrane exists only when an extracorporeal circuit is running – meaningful only when one of the dialysis-modality indicators is 1. Record the manufacturer’s membrane and its ultrafiltration coefficient per model in covariateData[[FILT_FLUX_LOW]]$notes when the paper gives them; the low/high-flux boundary is a clinical convention rather than a single numeric threshold and has drifted over the decades the pooled datasets span. Distinct from FILT_SA (surface area) and FILT_AGE_HI (membrane age). A paper reporting a genuinely continuous ultrafiltration coefficient should register a continuous canonical rather than extending this pair. Ratified by the maintainers’ ruling 2026-09-21 (PMC5306496): the decomposed pair with an all-zero ‘flux not recorded’ reference was preferred over a single high-vs-low binary, which cannot represent the 92%-of-cohort reference level the thetas were estimated against.

FILT_FLUX_HIGH (canonical for high-flux dialyser-membrane indicator)

• Description: Binary indicator that the in-use dialyser / haemofilter membrane is a HIGH-FLUX membrane – large pore size, high water permeability, and substantially better clearance of middle molecules. 1 = high-flux membrane; 0 = low-flux membrane or membrane type not recorded. The high-permeability partner of FILT_FLUX_LOW; see that entry for the full rationale.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (low-flux membrane or not recorded).
• Source aliases:
  • DIAM [High Flux] – Xu 2017 supplementary Table S2 row label for theta 14.
• Example models: Xu_2017_daptomycin.R (multiplicative factor e_filt_flux_high_cl = 1.36 on the dialysis clearance equation; Xu 2017 Table S2, 8% RSE; 28 of 459 subjects, including all 8 CVVHDF subjects and 2 of 9 CVVHD subjects).
• Notes: This factor is load-bearing when reproducing a source’s published exposures: in Xu 2017 the whole CVVHDF cohort carried a high-flux membrane, so the exposures printed in that paper’s Table 3 are generated at 0.528 x 1.36 = 0.72 L/h and cannot be recovered from the bare CVVHDF clearance estimate alone. Note also that the high- and low-flux multipliers are NOT symmetric about 1 in the founding source (1.36 vs 0.96): the reference is the unrecorded-membrane stratum, not the low-flux stratum, so the pair does not partition a two-level covariate.

RRT_HEMODIAL_ACTIVE (canonical for hemodialysis-active indicator (time-varying per-session gate))

• Description: Within-subject time-varying indicator for whether an extracorporeal renal-replacement-therapy session (intermittent hemodialysis, hemofiltration, or hemodiafiltration) is currently running. 1 during the session; 0 in the interdialytic interval and in non-dialysed subjects.

• Units: (binary)

• Type: binary

• Scope: general

• Reference category: 0 (no dialysis running). Models that compose an additional cl_hemodialysis term should gate it by RRT_HEMODIAL_ACTIVE = 1 so that the interdialytic clearance reduces to the intrinsic body clearance. Write the gated total into cl itself – see the encoding rule in Notes below; a separate cl_total variable silently disables the dialysis arm.

• Source aliases:

  • HEMODIALYSIS – prior canonical name (2026-06-09 – 2026-06-19); the 2026-06-19 RRT family standardization renamed it to RRT_HEMODIAL_ACTIVE to align with the sibling RRT_HEMODIAL_STATUS / RRT_CRRT_STATUS shape.
  • DIAL – legacy form (used in Liesenfeld 2013, Jacobs 2016, Veinstein 2013 prior to the 2026-06-09 rename to the more explicit HEMODIALYSIS).
  • IHD_ON, HD_ACTIVE, RRT_ACTIVE – variant abbreviations used in adjacent ESRD / CRRT popPK literature.

• Example models:

  • Veinstein_2013_gentamicin.R (additive arm with an estimated primary ini parameter; cl <- exp(lcl + etalcl) * WT + RRT_HEMODIAL_ACTIVE * cl_hemodialysis).
  • Liesenfeld_2013_dabigatran.R (Michaels-equation gate; the dialysis arm is derived from blood flow rate, dialysate flow rate, and a hemodialyzer mass-transfer-area coefficient rather than estimated directly. Retains a separate gated cl_total; its explicit ODE is honoured – verified by solving both gate states, not assumed – but do not copy the shape, for the reason in Notes).
  • Jacobs_2016_colistin.R (additive on/off gate of fixed device-level hemodialysis-clearance constants for CMS and colistin, in a parent-metabolite chain; likewise retains a separate gated cl_tot whose explicit ODE is honoured, and likewise is not a shape to copy).
  • Berthaud_2025_cefazolin.R (additive arm in a paediatric maintenance-hemodialysis cohort, with the arm driven by a continuous membrane surface area instead of a bare estimate: cl <- exp(lcl + etalcl) * (WT / 70)^0.75 + RRT_HEMODIAL_ACTIVE * cl_hemodialysis where cl_hemodialysis <- exp(lcl_hemodialysis) * (FILT_SA / 1)^1.26. Note that Berthaud 2025 does NOT print the gate in its final-equation block – the block states only CLtot_i = CL_i + CLdial_i – and the on/off structure is established by the Introduction (“during hemodialysis sessions, a dialysis clearance … is added to the patient residual elimination clearance”), by the Discussion’s contrast between an “interdialytic” and a “dialysis” half-life, and by Figure 4’s three-sessions-per-week simulation design).
  • VanWart_2025_telavancin.R (same additive-arm shape as Veinstein 2013, with IIV on the dialysis arm: cl_hemodialysis <- exp(lcl_hemodialysis + etalcl_hemodialysis) * RRT_HEMODIAL_ACTIVE, CL_DL = 1.77 L/h; Van Wart 2025 states CL_DL “was estimated only during those periods where intermittent hemodialysis (IHD) was active and was fixed to a value of zero when IHD was not operative”). Paired in the same model with the interdialytic T_POST_HEMODIAL on central volume.

• Notes: Distinct from a renal-impairment indicator (subject-level baseline class) and from RRT_HEMODIAL_STATUS (the subject-level treatment-status indicator) – RRT_HEMODIAL_ACTIVE is the per-time-point gate that turns the dialysis-clearance term on and off WITHIN a single hemodialysis subject’s record. Pair with BFR and DFR when the dialysis clearance depends on flow rates; pair with a filter-specific mass-transfer coefficient (estimated lkoa in the model, not a covariate) when the Michaels parameterisation is used. Pair with lcl_hemodialysis (canonical) when the dialysis arm is a primary estimated structural parameter rather than a derived expression.

  Encoding rule – write the gated total into cl, never into a separate cl_total. rxode2 does not always integrate the d/dt() a model file declares: for some model shapes it solves the linear-compartment system with its analytic kernel driven by variables named cl (or CL) and vc (or v), and the explicit right-hand side is discarded. When that happens, the idiom

  cl       <- <off-therapy arm>
  cl_total <- cl + RRT_HEMODIAL_ACTIVE * cl_hemodialysis
  kel      <- cl_total / vc
  d/dt(central) <- -kel * central

  produces a silently inert dialysis arm: the reported cl_total and kel output columns switch correctly with the gate while the simulated amounts decay at the off-therapy rate in BOTH states. Write it as a single assignment instead:

  cl <- <off-therapy arm> + RRT_HEMODIAL_ACTIVE * cl_hemodialysis
  kel <- cl / vc

  Under the REPLACEMENT rule (see the lcl_hemodialysis entry in parameter-names.md) the same principle applies: cl <- (1 - RRT_HEMODIAL_ACTIVE) * <off-therapy arm> + RRT_HEMODIAL_ACTIVE * cl_hemodialysis.

  You cannot tell by inspection whether a given model is affected – defining both cl and vc is necessary but not sufficient, and Liesenfeld_2013_dabigatran.R and Jacobs_2016_colistin.R both define the pair yet have their explicit ODEs honoured, while Eyler_2014_ertapenem.R (a plain two-compartment model) did not. So treat the single-assignment form as mandatory for new models rather than as a case-by-case judgement. Nothing else in the package catches a violation: checkModelConventions() does not, modeldb$linCmt reads FALSE for an affected model, and a validation vignette that simulates only one gate state cannot see it. The mechanical gate is tests/testthat/test-modeldb-active-gate.R, which solves every model carrying an RRT_*_ACTIVE covariate at gate 0 and gate 1 and fails if no ODE state moves. Three shipped models were found inert by that sweep and repaired: Veinstein_2013_gentamicin.R, Dohmann_2025_piperacillin.R and Eyler_2014_ertapenem.R. The same hazard applies verbatim to the sibling gates RRT_CRRT_ACTIVE, RRT_CAPD_ACTIVE and RRT_CCPD_ACTIVE.

RRT_CRRT_ACTIVE (canonical for continuous-renal-replacement-therapy-active indicator (time-varying gate))

• Description: Within-subject time-varying indicator for whether a continuous or extended extracorporeal renal-replacement-therapy modality is currently running – continuous venovenous hemodialysis (CVVHD), hemofiltration (CVVH), hemodiafiltration (CVVHDF), sustained low-efficiency dialysis (SLED), or extended daily diafiltration (EDD-f). 1 while the therapy is running; 0 while it is interrupted and in subjects never treated. Completes the RRT_<modality>_<kind> family as the continuous-modality counterpart of RRT_HEMODIAL_ACTIVE.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no continuous RRT running). Models that compose an additional dialysis-clearance term should gate it by RRT_CRRT_ACTIVE = 1 so that clearance during an interruption reduces to the intrinsic body clearance.
• Source aliases:
  • CVVHD – Huppe 2023 prose and figure legends.
  • CRRT_ON, CVVH_ACTIVE – variant abbreviations used in adjacent critical-care popPK literature.
• Example models: Huppe_2023_fosfomycin.R (Michaels-equation gate on the dialysis clearance arm: cl_cvvhd <- RRT_CRRT_ACTIVE * 60 * Michaels(BFR, DFR, K0A); each of the 15 patients contributed one PK profile during CVVHD and one after CVVHD was interrupted for tubing-system replacement, so the covariate genuinely varies within subject), Zhang_2025_fluconazole.R (gates the covariate-supplied extracorporeal clearance: clcrrt <- QEFF * RRT_CRRT_ACTIVE, reproducing the supplement’s $PK block IF (CRRTYN.EQ.0) CL_crrt = 0 / IF (CRRTYN.EQ.1) CL_crrt = CLCRRT. Every subject in the model-building dataset was on CRRT, so the gate does not vary within that dataset; it carries the paper’s off-CRRT simulation arm – the 0 mL/kg/h row of Tables 3 and 4 – and the intermittent-session scenarios of the companion Shiny application).
• Notes: Distinct from RRT_CRRT_STATUS, which is the SUBJECT-LEVEL treatment-status flag for the same modality family (Shekar 2014 records whether a patient was on RRT at all during sampling, not when a session was running). Distinct from RRT_HEMODIAL_ACTIVE, which is the per-session gate for INTERMITTENT hemodialysis (Liesenfeld 2013, Jacobs 2016, Veinstein 2013). The intermittent-versus-continuous split is retained at the ACTIVE level for the same reason the register already retains it at the STATUS level: the two modality classes have materially different drug-extraction kinetics – intermittent hemodialysis runs high blood and dialysate flows for a few hours, whereas continuous modalities run flows an order of magnitude lower around the clock, so a coefficient or mass-transfer coefficient calibrated on one does not transfer to the other. Pair with BFR and DFR when the dialysis clearance is derived from circuit flow rates via the Michaels equation, and with lkoa as the estimated filter mass-transfer-area coefficient in model (not a covariate). Note that BFR and DFR are canonically both mL/min, while continuous-modality papers commonly report dialysate flow in L/h – convert on ingestion and document any in-model back-conversion, as Huppe_2023_fosfomycin.R must do.

HEPIMP_MOD (canonical for moderate hepatic impairment indicator)

• Description: 1 = moderate hepatic impairment, 0 = normal hepatic function or any other (non-moderate) category. The classification scheme that defines “moderate” is paper-specific and must be documented in per-model covariateData[[HEPIMP_MOD]]$notes. Two schemes are commonly encountered:
  • NCI ODWG group 3: total bilirubin > 1.5-3 x ULN with any AST (Ramalingam SS et al., J Clin Oncol 2010;28:4507).
  • Child-Pugh Class B: composite score 7-9 across bilirubin, albumin, INR, ascites, and encephalopathy.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (any non-moderate category: normal, mild, or severe; the model typically uses HEPIMP_MOD alongside HEPIMP_MILD and/or HEPIMP_SEV to partition the non-moderate pool further).
• Source aliases:
  • HEP2 – used in Desai_2016_isavuconazole.R (paper’s liver-function index HEP2 for moderate Child-Pugh B; same orientation as the canonical).
• Example models: Desai_2016_isavuconazole.R (Child-Pugh Class B; log-additive shift e_hepimp_mod_cl = log(1.326 / 2.54) = -0.650 on CL and e_hepimp_mod_q = log(63.554 / 33.678) = +0.635 on Q vs the healthy reference, paired with HEPIMP_MILD for the parallel Child-Pugh A stratum; encodes the paper’s three-typical-value form CL = theta_g * exp(eta_jCL) as a healthy-reference baseline plus group-specific log shifts), Comisar_2025_rimegepant.R (multiplicative fractional effect on CL/F = -0.229 per Comisar 2025 Table 3, a 22.9% clearance decrease giving a 1.3-fold rise in simulated steady-state AUCtau; paired with HEPIMP_SEV (-0.423) as two mutually exclusive indicators whose shared reference group is normal-OR-mild hepatic function, because mild impairment was screened and found not statistically significant; the paper does not state whether its mild / moderate / severe classification follows Child-Pugh or NCI ODWG).
• Notes: Use this column when a model dichotomizes moderate hepatic impairment as a separate indicator from milder or more-severe categories. The classification scheme (NCI ODWG vs Child-Pugh vs other) is paper-specific and must be documented per-model. Companion to HEPIMP_MILD (mild only) and HEPIMP_SEV (severe only). Distinct from HEPIMP_MOD_OR_MISSING (pools moderate cases with missing-data cases) and HEPIMP_MODSEV (pools moderate with severe). Anticipated by the HEPIMP_MILD entry’s Notes: “For models that test moderate or severe as separate categories, register additional canonicals HEPIMP_MOD / HEPIMP_SEV rather than overloading this entry.” HEPIMP_SEV was registered separately; HEPIMP_MOD completes the parallel.

CRCL_BASE (canonical for per-subject baseline creatinine clearance (time-fixed))

• Description: Per-subject baseline renal function (creatinine clearance or BSA-normalized eGFR), time-fixed. Used when a source paper enters the subject’s baseline renal-function value as a separate constant column alongside a time-varying CRCL column under Wahlby 2004’s extended covariate-model decomposition (Br J Clin Pharmacol 2004;58(4):367-377). The within-subject delta is computed in model as (CRCL - CRCL_BASE).
• Units: mL/min or mL/min/1.73 m^2 – match the units of the paired CRCL column and document per-model via covariateData[[CRCL_BASE]]$units.
• Type: continuous
• Scope: general
• Reference category: n/a – used either as (CRCL_BASE - ref) in a linear or linear-exp effect, or as the ratio (CRCL_BASE / ref), where ref is a cohort or upstream-analysis median baseline renal function. Reference values observed: 71.7 mL/min (Wahlby 2004 BCLC median), 93 mL/min/1.73 m^2 (Decker 2024, median eGFR of the upstream adult rheumatoid-arthritis analysis).
• Source aliases:
  • BCLC (baseline creatinine clearance) – Wahlby 2004 source-column convention; used in Wahlby_2004_gentamicin.R (raw mL/min) and Wahlby_2004_pefloxacin.R (concept tested but not retained in final model).
  • baseline eGFR – Decker 2024 source-table wording for the time-fixed bedside-Schwartz pediatric eGFR entering the renal-clearance arm as (baseline eGFR / 93); used in Decker_2024_baricitinib.R.
• Example models: Wahlby_2004_gentamicin.R (raw mL/min; reference 71.7 mL/min = BCLC median; coefficient 0.0098 per mL/min on CL alongside paired DCLC effect 0.0068 per mL/min), Decker_2024_baricitinib.R (bedside-Schwartz pediatric eGFR in mL/min/1.73 m^2; enters the apparent renal clearance arm multiplicatively as the ratio (CRCL_BASE / 93) rather than as a centred linear deviation, paired with the delta effect 0.00586 per mL/min/1.73 m^2 on (CRCL - CRCL_BASE); the reference 93 is the median eGFR of the upstream adult analysis, not a statistic of the pediatric cohort, whose mean baseline eGFR is 119), Tsuchiwata_2026_tofacitinib.R (raw Cockcroft-Gault CrCl in mL/min, source column BCCL; enters CL/F as the power function (CRCL_BASE / 126)^0.233), Osawa_2018_daclatasvir.R (raw CrCl in mL/min, source column BCRCL; enters CL/F as the power function (CRCL_BASE / 86.48)^0.235; reference is the cohort median, printed as 86.48 in the equation text and 86.5 in Table 1 – the Figure 3 caption’s ‘88.48’ is a typo).
• Notes: Promoted to general scope on 2026-08-04 alongside the Decker 2024 baricitinib extraction, the second model registering the column with consistent Wahlby-style BCOV/DCOV semantics. The two ratifying papers differ in the functional form applied to the baseline arm – Wahlby 2004 uses a centred linear deviation (CRCL_BASE - 71.7), Decker 2024 a multiplicative ratio (CRCL_BASE / 93) – and in the underlying assay (Cockcroft-Gault-style raw CrCl vs bedside-Schwartz pediatric eGFR); both are within scope because the column’s role is identical (the time-fixed baseline half of a time-varying renal-function covariate). Document the assay, the units, and the functional form per model in covariateData[[CRCL_BASE]]$notes. Distinct from time-fixed adult-cohort CRCL (where per-subject CRCL is just constant by data design); use CRCL_BASE only when the source paper deliberately separates the BCOV from the time-varying COV.

TBILI_BASE (canonical for per-subject baseline total bilirubin (time-fixed))

• Description: Per-subject baseline total serum bilirubin, time-fixed. Used when a source paper enters the subject’s baseline TBILI as a constant column alongside the time-varying TBILI column under Wahlby 2004’s extended covariate-model decomposition (Br J Clin Pharmacol 2004;58(4):367-377). The within-subject delta is computed in model as (TBILI - TBILI_BASE).
• Units: mg/dL or umol/L – match the units of the paired TBILI column and document per-model via covariateData[[TBILI_BASE]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – typically used as (TBILI_BASE - ref) in a linear or linear-exp effect.
• Source aliases:
  • BBIL (baseline total bilirubin) – Wahlby 2004 source-column convention; used in Wahlby_2004_pefloxacin.R and Wahlby_2004_paclitaxel_myelosuppression.R.
• Example models: Wahlby_2004_pefloxacin.R (umol/L; centered at 25 umol/L; coefficient -0.0068 per umol/L on CL via exp[e * (TBILI_BASE - 25)]), Wahlby_2004_paclitaxel_myelosuppression.R (umol/L; used together with TBILI as the within-subject delta TBILI - TBILI_BASE entering the Slope-DBIL effect).
• Notes: Specific scope because the BCOV/DCOV decomposition is paper-defined. Promote to general if a second paper ratifies. Distinct from DBIL (direct/conjugated bilirubin, a clinical-chemistry concept) – TBILI_BASE is a methodologic baseline-snapshot column, not a separate biomarker.

ALP_BASE (canonical for per-subject baseline alkaline phosphatase (time-fixed))

• Description: Per-subject baseline serum alkaline phosphatase, time-fixed. Used when a source paper enters the subject’s baseline ALP as a constant column alongside the time-varying ALP column under Wahlby 2004’s extended covariate-model decomposition (Br J Clin Pharmacol 2004;58(4):367-377). The within-subject delta (or log-ratio) is computed in model from ALP and ALP_BASE.
• Units: U/L (IU/L; interchangeable). Match the units of the paired ALP column and document per-model via covariateData[[ALP_BASE]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – typically used inside a log-ratio log(ALP / ALP_BASE) or a linear delta (ALP - ALP_BASE).
• Source aliases:
  • BALKP (baseline alkaline phosphatase) – Wahlby 2004 source-column convention; used in Wahlby_2004_voriconazole.R.
• Example models: Wahlby_2004_voriconazole.R (IU/L; population median 136 IU/L; paired with time-varying ALP to form log(ALP/ALP_BASE), the paper’s log(DALKP) term, with coefficient 0.59 on CL inside Eq 7 Final-Model column).
• Notes: Specific scope because the BCOV/DCOV decomposition is paper-defined. Promote to general if a second paper ratifies.

AST_BASE (canonical for per-subject baseline aspartate aminotransferase (time-fixed))

• Description: Per-subject baseline serum aspartate aminotransferase activity, time-fixed. Used when a source paper enters the subject’s baseline AST as a separate constant column alongside the time-varying AST column, so that the two can carry distinct effects on the same structural parameter. Operationally equal to AST at the subject’s first observation and constant thereafter.
• Units: U/L (SI; IU/L is used interchangeably). Match the units of the paired AST column and document per-model via covariateData[[AST_BASE]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – used as the ratio (AST_BASE / ref) inside a power function, or as a centred linear deviation (AST_BASE - ref), where ref is the cohort median baseline AST. Reference value observed: 52 U/L (Osawa 2018, the asunaprevir cohort median).
• Source aliases:
  • BAST (baseline aspartate aminotransferase) – Osawa 2018 source-column convention, defined in the Table 2 footnote as “BAST, baseline aspartate aminotransferase” against “AST, aspartate aminotransferase at each time”.
• Example models: Osawa_2018_asunaprevir.R (U/L, reference 52 = the cohort median; power exponent -0.598 on CL/F, paired with the time-varying AST exponent -0.382 on the same parameter; both terms normalised to the SAME constant 52 U/L).
• Notes: The hepatic-transaminase member of the baseline-snapshot family alongside WT_BASE, BSA_BASE, CRCL_BASE, TBILI_BASE and ALP_BASE, and registered for the same reason: the split between a baseline value and a time-varying value is a modelling choice the source authors made deliberately, and collapsing the two loses it. Osawa 2018 is explicit that the split earned its place – adding the time-varying AST term to a model already carrying baseline AST dropped the objective function by a further 139.192 units. One feature to watch when transcribing this family: Osawa 2018 normalises the time-varying term by BASTref as well, so a single reference constant serves both terms and a subject whose AST never moves carries the SUM of the two exponents. Do not assume a separate reference for the time-varying column without checking the printed equation. Specific scope because the baseline/time-varying decomposition is paper-defined; promote to general if a second paper ratifies the column with consistent semantics. Distinct from AST, which is the time-varying (or single-value) transaminase column.

RRT_CCPD_ACTIVE (canonical for continuous-cycler-assisted-peritoneal-dialysis-active indicator (time-varying per-exchange gate))

• Description: Within-subject time-varying indicator for whether a continuous cycler-assisted peritoneal dialysis (CCPD) exchange is currently running. 1 during a cycler-driven exchange; 0 during ambulatory exchanges, during dry periods, and in subjects not on peritoneal dialysis. Completes the RRT_<modality>_<kind> family as the cycler-assisted peritoneal member, alongside its ambulatory counterpart RRT_CAPD_ACTIVE.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no cycler-assisted exchange running). Models that compose a CCPD clearance arm should gate it by RRT_CCPD_ACTIVE = 1 so that clearance outside a cycler exchange reduces to the other active routes.
• Source aliases:
  • CCPD – Patel 2015 prose, Figure 1 (CL_OC,CCPD/CAPD) and Table 2 (CL_OCCCPD). Same orientation, no value transformation.
• Example models: Patel_2015_oseltamivir.R (gates the cycler arm of a two-arm peritoneal dialysate clearance in end-stage-renal-disease patients on automated peritoneal dialysis: cldial <- RRT_CCPD_ACTIVE * cl_ccpd_oselcarb + RRT_CAPD_ACTIVE * cl_capd_oselcarb. The study regimen ran three CCPD exchanges over an 8 h daytime block followed by two CAPD exchanges over 16 h overnight, repeating daily, so the gate genuinely alternates within every subject; the paper’s Monte Carlo simulations vary the two gates’ schedules across three different automated-peritoneal-dialysis prescriptions, which is what makes the split load-bearing rather than cosmetic).
• Notes: Distinct from PERIT_DIAL, which is the SUBJECT-LEVEL chronic-modality flag (is this patient a peritoneal-dialysis patient at all) and cannot express which exchange type is running at a given time. Distinct from RRT_HEMODIAL_ACTIVE and RRT_CRRT_ACTIVE, which gate EXTRACORPOREAL circuits: peritoneal dialysis moves drug across the peritoneal membrane into instilled dialysate with no blood circuit, so neither a blood flow rate nor a hemodialyzer mass-transfer coefficient applies and a coefficient calibrated on an extracorporeal modality does not transfer. Registered as a pair with RRT_CAPD_ACTIVE rather than as one peritoneal gate plus a modality selector because the two arms carry separate estimated clearances (Patel 2015: 0.319 vs 0.170 L/h/70 kg, a 1.9-fold difference driven by exchange frequency, not membrane properties) and because a pair of binaries composes additively in model() without a nested conditional. In a schedule with dry periods both gates are 0 and only the residual routes clear drug. Pair with the clearance-component tokens _ccpd / _capd (R/conventions.R::clComponents). Ratified by the maintainers 2026-09-02 (PMC4386947).

RRT_CAPD_ACTIVE (canonical for continuous-ambulatory-peritoneal-dialysis-active indicator (time-varying per-exchange gate))

• Description: Within-subject time-varying indicator for whether a continuous ambulatory peritoneal dialysis (CAPD) exchange is currently dwelling. 1 during an ambulatory exchange; 0 during cycler-assisted exchanges, during dry periods, and in subjects not on peritoneal dialysis. Ambulatory counterpart of RRT_CCPD_ACTIVE.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no ambulatory exchange dwelling). Models that compose a CAPD clearance arm should gate it by RRT_CAPD_ACTIVE = 1.
• Source aliases:
  • CAPD – Patel 2015 prose, Figure 1 (CL_OC,CCPD/CAPD) and Table 2 (CL_OCCAPD). Same orientation, no value transformation.
• Example models: Patel_2015_oseltamivir.R (gates the ambulatory arm of the same two-arm peritoneal dialysate clearance; see the RRT_CCPD_ACTIVE entry for the shared model form and schedule).
• Notes: See the RRT_CCPD_ACTIVE Notes for the full family rationale, the extracorporeal-versus-peritoneal distinction, and why the two exchange types are registered as a pair of binaries. One caution specific to CAPD: the osmotic agent differs across exchanges in practice (dextrose versus icodextrin), and this column does NOT resolve it. Patel 2015 measured both and found them comparable for oseltamivir carboxylate (geometric mean CAPD clearance 0.195 L/h with icodextrin versus 0.177 L/h with dextrose), so a single CAPD gate was sufficient there; a drug whose peritoneal transport does depend on the osmotic agent would need a separate column rather than an overload of this one. Ratified by the maintainers 2026-09-02 (PMC4386947).

PERIT_DIAL (canonical for peritoneal-dialysis treatment-status indicator)

• Description: 1 = the subject was undergoing chronic peritoneal dialysis (PD) during the modeled period; 0 = not on peritoneal dialysis. Treatment-status flag rather than a measured renal-function value; used as a multiplicative covariate on PK parameters that change between dialysis modalities (typically central volume Vc – chronic peritoneal dialysis is associated with higher extracellular fluid volume than intermittent hemodialysis, so plasma volume markers are larger in PD subjects). Per-subject indicator in the source data; typically time-fixed at the subject level (chronic dialysis modality is a months-to-years-stable patient characteristic, not a per-session attribute).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (not on peritoneal dialysis). In a chronic-dialysis cohort comparing PD vs HD modalities, the reference category corresponds to the HD subjects; PERIT_DIAL = 1 selects the PD subgroup.
• Source aliases:
  • DIA – used in Takama_2007_darbepoetin.R (dialysis-modality indicator: DIA = 0 for hemodialysis, DIA = 1 for peritoneal dialysis; maps directly onto PERIT_DIAL with no value transformation). Takama 2007 Table 4 and Methods describe the covariate equation V1 = theta_V1 * [1 + theta_V1_WT * (WT - 54) + theta_V1_DIA * DIA] where theta_V1_DIA = 0.170 increments V1 by +17% in PD subjects relative to HD subjects.
  • DIT2 – used in Knebel_2008_epoetinDelta.R (the second of two dialysis-modality switches in a three-level covariate; DIT2 = 1 selects the peritoneal-dialysis stratum, with predialysis rather than hemodialysis as the reference – see the Notes below).
• Example models: Takama_2007_darbepoetin.R (+17% multiplicative-deviation effect on central volume V1; chronic-dialysis cohort comparing 63 HD vs 68 PD Japanese adult patients), Knebel_2008_epoetinDelta.R (exponentiated-switch factor 2.48^PERIT_DIAL on central volume only, paired with RRT_HEMODIAL_STATUS against a predialysis reference in a pediatric CKD erythropoietin popPK model).
• Notes: Distinct from RRT_HEMODIAL_STATUS (intermittent-hemodialysis IHD indicator; RRT_HEMODIAL_STATUS = 1 selects HD subjects) and from RRT_CRRT_STATUS (continuous / extended RRT indicator). When a cohort spans three modalities (predialysis / HD / PD) and the paper takes predialysis as the reference, PERIT_DIAL and RRT_HEMODIAL_STATUS are both used, are no longer anti-correlated, and are both 0 for the predialysis stratum; see the three-level paragraph in the RRT_HEMODIAL_STATUS Notes. In a chronic-dialysis cohort where every subject is on either HD or PD, RRT_HEMODIAL_STATUS and PERIT_DIAL are perfectly anti-correlated (PERIT_DIAL = 1 - RRT_HEMODIAL_STATUS); the choice of which to use as the covariate is dictated by which modality the source paper treats as the reference category. Takama 2007 took HD as the reference and reports a +17% effect on V1 for PD, so PERIT_DIAL is the natural canonical for that paper (V1 increases when PERIT_DIAL = 1). Anticipated in the RRT_HEMODIAL_STATUS register-entry notes (“would warrant its own canonical (RRT_PERIT_DIAL_STATUS, RRT_CRRT_STATUS) if a future paper retains them as covariates”) and ratified canonically on 2026-06-12 alongside the Takama 2007 darbepoetin alfa extraction. Future audit may rename this to RRT_PERIT_DIAL_STATUS for consistency with the RRT__ family; left unchanged for now to avoid disturbing the Takama 2007 ingestion before broader peritoneal-dialysis coverage in the registry.

HEMOADSORB_ACTIVE (canonical for hemoadsorption-active indicator (time-varying cartridge-in-circuit gate))

• Description: Within-subject time-varying indicator for whether an extracorporeal hemoadsorption (hemoperfusion) cartridge is currently in the blood circuit. 1 while a sorbent cartridge is perfused; 0 before and after the session, and throughout for subjects whose extracorporeal circuit carried no cartridge (sham circuit) or who were never treated. Hemoadsorption is a third extracorporeal modality alongside renal replacement therapy and ECMO: solute is removed by binding to a porous polymer sorbent rather than by diffusion or convection across a membrane, so the removal is saturable – the arm it gates is normally a decreasing function of the amount already adsorbed rather than a constant clearance.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no hemoadsorption cartridge in circuit). Models that compose an additional cl_hemoadsorption term should gate it by HEMOADSORB_ACTIVE = 1 so that clearance outside the session reduces to the intrinsic body clearance.
• Source aliases:
  • CytoSorb, CS, HP_ON, HA_ACTIVE – variant abbreviations used in the hemoperfusion / blood-purification PK literature. Leber 2023 has no data column for the indicator at all: it encodes the paper’s intervention-versus-control group assignment (five animals per combination with a CytoSorb cartridge in the circuit versus three with an identical sham circuit and no cartridge), so the ratifying models set source_name = NULL.
• Example models: Leber_2023_tacrolimus_sheep.R, Leber_2023_ciclosporin_sheep.R, Leber_2023_mycophenolateMofetil_sheep.R, Leber_2023_everolimus_sheep.R, Leber_2023_methylprednisolone_sheep.R (founding examples; all five gate the saturable arm cl_hemoadsorption <- HEMOADSORB_ACTIVE * clmax_hemoadsorption * (1 - adsorbed / amax_hemoadsorption) over a single 6-hour session, and setting the covariate to 0 throughout reproduces the paper’s sham-circuit control arm).
• Notes: Named ACTIVE rather than STATUS on the same split the RRT family uses: hemoadsorption is administered as a discrete, usually short session (6 h in Leber 2023, and commonly 8-24 h in critical care), so the per-time-point gate is the useful form and a subject-level HEMOADSORB_STATUS would be registered separately if a paper ever retained one. Deliberately outside the RRT_<MODALITY>_<KIND> namespace: hemoadsorption is not a renal-replacement modality – it provides no solute clearance by dialysis or filtration, no fluid removal, and is frequently run in series WITH an RRT circuit or an ECMO circuit rather than instead of one, so a model may legitimately carry HEMOADSORB_ACTIVE alongside RRT_CRRT_ACTIVE or ECMO_STATUS. Distinct from ECMO_STATUS for the same reason. Pair with BFR when the extracorporeal blood flow rate is a covariate on the adsorption arm (Leber 2023 held it constant at 120 mL/min and it therefore does not appear in the ratifying models), and with the adsorbed compartment plus lclmax_hemoadsorption / lamax_hemoadsorption parameters that make up the saturable sub-model. Because the gated arm depends on a cumulative state, switching the covariate back to 0 does not reset the cartridge load: a model that simulates two successive sessions with a fresh cartridge must reset adsorbed explicitly.

APHERESIS_LIPO_ACTIVE (canonical for lipoprotein-apheresis-active indicator (time-varying per-session gate))

• Description: Within-subject time-varying indicator for whether a lipoprotein apheresis (LA) session is currently running. 1 while blood is being processed through the apheresis column; 0 between sessions and throughout for subjects never treated. Lipoprotein apheresis is a selective extracorporeal procedure that removes apolipoprotein-B-containing particles (LDL and Lp(a)) by adsorption, precipitation or filtration, and is used in homozygous familial hypercholesterolemia when drug therapy alone does not reach LDL-C goals. The arm it gates is normally an additional first-order loss from an LDL-C turnover pool, so the covariate belongs on the PD side of a lipid model rather than on drug clearance.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no apheresis session running). Models that compose an additional LDL-C elimination term should gate it by APHERESIS_LIPO_ACTIVE = 1 so that the between-session turnover reduces to the intrinsic kout.
• Source aliases:
  • APHON – used in Bihorel_2025_evinacumab.R (Bihorel 2025 NM-TRAN $INPUT column, apheresis on/off).
  • LA_ON, APH_ACTIVE, LDLAPH – variant abbreviations used in the adjacent lipid-apheresis literature.
• Example models: Bihorel_2025_evinacumab.R (founding example; gates the second elimination arm of the LDL-C indirect-response pool, - kout_apheresis * APHERESIS_LIPO_ACTIVE * ldl, reproducing the supplement’s $DES term -KAPH*APHON*A(4). The estimated rate of 8.36 /day corresponds to roughly a 58% LDL-C reduction over the paper’s typical 2.5-hour session, matching its reported “approximately 60%”).
• Notes: Named ACTIVE rather than STATUS on the same split the RRT family uses – apheresis is delivered as discrete sessions of a few hours at weekly to monthly frequency, so the per-time-point gate is the useful form. A subject-level “receives lipoprotein apheresis at baseline” flag and a session-frequency column are distinct concepts and would be registered separately if a paper ever retained one as a covariate (Bihorel 2025 carries both as dataset columns, APH and APHFRQN, but neither enters the final model). Deliberately outside the RRT_<MODALITY>_<KIND> namespace, for the same reason as HEMOADSORB_ACTIVE: apheresis is not a renal-replacement modality and may legitimately be carried alongside one. Session duration is not part of this column – it is expressed by how long the indicator stays at 1 in the event table, so a model reusing this covariate must state the assumed session length (2.5 hours is the average Bihorel 2025 reports, and the value its supplement uses to impute missing start/stop times). Distinct from HEMOADSORB_ACTIVE: hemoadsorption gates a saturable sorbent arm driven by a cumulative adsorbed state, whereas lipoprotein apheresis is modelled here as a plain first-order loss that leaves no cumulative state to reset between sessions.

ECMO_STATUS (canonical for extracorporeal-membrane-oxygenation treatment-status indicator)

• Description: 1 = the subject was supported by extracorporeal membrane oxygenation (veno-venous VV-ECMO or veno-arterial VA-ECMO) during the modeled period; 0 = no ECMO support. Treatment-status flag rather than a measured circuit quantity; used as a multiplicative covariate on the PK parameters that shift when a patient is cannulated – most often the central volume, which expands because the extracorporeal circuit adds several litres of circulating fluid for a hydrophilic drug to distribute into, and sometimes clearance. Completes the extracorporeal-support covariate family as the presence-of-therapy counterpart of the circuit-quantity columns ECMO_PUMP_SPEED and Q_CVVH and of the time-since-event columns T_ECMO / T_POST_ECMO.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (not receiving ECMO).
• Source aliases:
  • ECMO – the bare source-column form used by both ratifying papers; same orientation (1 = on ECMO), no value transformation. Used in Kang_2020_cefpirome.R, Valadez_2025_cefepime.R, Xu_2026_caspofungin.R and Pressiat_2022_amikacin.R.
  • ECMO – the bare source-column form used by both ratifying papers; same orientation (1 = on ECMO), no value transformation. Used in Kang_2020_cefpirome.R, Valadez_2025_cefepime.R, Xu_2026_caspofungin.R, Kim_2022_piperacillin.R and Kim_2022_tazobactam.R (the last two select a separate typical central volume, each with its own IIV, per ECMO status rather than applying a multiplicative effect).
• Example models: Kang_2020_cefpirome.R (power-form multipliers on both parameters, CL = 5.71 * 0.487^(CREAT/1.6) * 1.41^ECMO_STATUS and V1 = 2.74 * 4.22^ECMO_STATUS, in VA-ECMO patients sampled both during support and after weaning), Valadez_2025_cefepime.R (exponential multiplier on the central volume only, Vd = V1 * (WT/70) * exp(1.043 * ECMO_STATUS), a 2.8-fold expansion; an ECMO effect on CL was tested and rejected for imprecision, and the cohort excluded renal-replacement patients so the indicator is not confounded by CRRT), Xu_2026_caspofungin.R (power-form multiplier on the central volume only, V1 = 2.22 * (BSA/0.79) * 18.2^ECMO_STATUS, the largest ECMO volume effect in the register at 18.2-fold; the only paediatric example, and the only one where the effect’s magnitude is barely identified – 4 of 29 patients on ECMO, 62% RSE, bootstrap 95% CI 4.00-298.30), Kang_2026_ticagrelor.R (power-form multipliers on both parameters in the SAME direction the mechanism predicts but OPPOSITE to Kang 2020 on clearance: CL/F = 12.2 * 0.35^ECMO_STATUS – a 65% clearance REDUCTION on ECMO, the only decrease in the register – and Vd/F = 157 * 2.74^ECMO_STATUS * (0.518^(Q_ECMO/3))^ECMO_STATUS, the only example that pairs the indicator with a circuit-flow term nested inside it; same first author, institution and VA-ECMO cohort design as Kang_2020_cefpirome.R, which makes the opposite clearance sign a drug-property contrast – renally cleared cefpirome speeds up, hepatically metabolised ticagrelor slows down – rather than a cohort difference), Pressiat_2022_amikacin.R (exponential multiplier on the PERIPHERAL volume only, V2 = 15.9 * exp(0.76 * ECMO_STATUS), a 2.14-fold expansion, in a V-A ECMO versus non-ECMO ICU control cohort; subject-level indicator for a single studied dose).
• Notes: Time resolution is paper-specific and must be documented in each model’s covariateData[[ECMO_STATUS]]$notes. Kang 2020 treats it as time-varying within subject – PK samples were drawn both during ECMO and after successful weaning, so the per-record indicator switches at decannulation. Valadez 2025 treats it as a subject-level indicator because the source does not resolve cannulation / decannulation timing. Distinct from ECMO_PUMP_SPEED (continuous centrifugal-pump rotational speed in RPM, which quantifies the degree of support rather than its presence) and from T_ECMO / T_POST_ECMO (time since cannulation / decannulation, which model a gradual onset or recovery rather than a step change); a paper that reports both presence and a circuit quantity should register both columns. Distinct from RRT_CRRT_STATUS – ECMO and continuous renal replacement therapy are separate extracorporeal circuits that frequently co-occur, and conflating them is the confounder Valadez 2025 was designed to remove by excluding RRT patients entirely. When pairing ECMO_STATUS with CRCL, note that the two ratifying papers disagree on whether ECMO alters clearance (Kang 2020 retains a 1.41-fold CL increase; Valadez 2025 rejects an ECMO-on-CL term as a null effect with CV > 200%), so the covariate’s effect on CL should never be assumed from class precedent.

RRT_CVVHD_STATUS (canonical for continuous veno-venous haemodialysis treatment-status indicator)

• Description: 1 = the subject was undergoing continuous veno-venous haemodialysis (CVVHD) – purely DIFFUSIVE continuous renal replacement, dialysate counter-current across the filter with no replacement fluid – during the modeled period; 0 = not on CVVHD. Treatment-status flag rather than a measured renal-function value. The modality-resolved sibling of RRT_CRRT_STATUS, for the case the RRT_CRRT_STATUS Notes anticipate: a source paper that retains individual continuous modalities as SEPARATE covariates because it estimates different disposition for each. Per-subject, time-fixed in the founding example.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (not on CVVHD). In a paper that strata-codes every dialysis modality, the all-zero state across RRT_HEMODIAL_STATUS, PERIT_DIAL, RRT_CVVHD_STATUS and RRT_CVVHDF_STATUS is the not-on-dialysis reference.
• Source aliases:
  • CVVHD – Xu 2017 Table 2 and supplementary Table S2 row labels (CL CVVHD, V1 CVVHD, Q CVVHD, V2 CVVHD); same orientation, no value transformation.
• Example models: Xu_2017_daptomycin.R (REPLACEMENT rule on all four disposition parameters – the indicator selects lcl_cvvhd 0.936 L/h, lvc_cvvhd 5.736 L, lq_cvvhd 7.111 L/h and lvp_cvvhd 4.891 L in place of the not-on-dialysis / other-stratum values; 9 of 459 pooled subjects).
• Notes: Use RRT_CRRT_STATUS when a paper lumps every continuous / extended modality into one indicator, and these modality-resolved columns only when the paper itself estimates modality-specific parameters. The distinction is not cosmetic: in the founding source CVVHD clearance is 1.25-fold the not-on-dialysis typical value while CVVHDF clearance is 29% BELOW it, so collapsing the two onto a single CRRT indicator would average away a 1.8-fold contrast. A model carrying these columns should leave RRT_CRRT_STATUS out rather than setting all three, to avoid double-counting. Paired with RRT_CVVHDF_STATUS; mutually exclusive with it and with RRT_HEMODIAL_STATUS / PERIT_DIAL. The clinical procedure the estimate is conditioned on must be recorded per model in covariateData[[RRT_CVVHD_STATUS]]$notes – filter surface area, blood flow and dialysate flow determine the extracorporeal clearance, and Xu 2017 states explicitly that its dose recommendation applies only to comparable procedures (high-flux 1.4-1.8 m^2 filter, blood flow 100-200 mL/min, dialysis flow 30-40 mL/kg/h). Where the paper records membrane flux separately, pair with FILT_FLUX_LOW / FILT_FLUX_HIGH. Ratified by the maintainers’ ruling 2026-09-21 (PMC5306496): adopt the two modality siblings ONLY and do NOT additionally set RRT_CRRT_STATUS on a model carrying them.

AST_BL (canonical for per-subject baseline (pre-treatment) aspartate aminotransferase (time-fixed, paired with time-varying AST))

• Description: Serum aspartate aminotransferase activity at treatment baseline, time-fixed per subject. Register this column alongside the time-varying AST whenever a source paper enters BOTH the pre-treatment anchor and the on-treatment value as separate covariates – typically because the on-treatment value enters only as the ratio AST / AST_BL, which decomposes a hepatic-function covariate into a time-fixed severity component and a time-varying recovery component. Operationally AST_BL equals AST at the subject’s first record and stays constant thereafter. Carries _BL’s established at-first-dose meaning used throughout this register (QTC_BL, HGB_BL, FERRITIN_BL, INS_BL).
• Units: U/L (SI; IU/L is used interchangeably). Document per-model via covariateData[[AST_BL]]$units.
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (AST_BL / ref)^exponent. Reference values observed: 60 U/L (Zhu 2018; the reference subject’s value, close to the 62 U/L cohort mean and above the 51 U/L median).
• Source aliases:
  • Baseline AST – the printed covariate-equation term in Zhu_2018_asunaprevir.R; same orientation, no value transformation.
• Example models: Zhu_2018_asunaprevir.R (U/L, reference 60; exponent -0.458 on CL/F, paired with the time-varying ratio term (AST / AST_BL)^-0.291 on the same parameter – higher baseline AST means lower apparent clearance and therefore higher asunaprevir exposure, and clearance then rises again as on-treatment AST falls toward baseline-normal during antiviral therapy).
• Notes: Hepatic-function marker. General scope because the baseline-versus-time-varying AST decomposition is a generic popPK construct rather than a paper-specific one, and because _BL is already an established register-wide suffix with exactly this meaning. Use plain AST when a paper enters only one AST column, whether that column holds a baseline or a time-varying value – AST_BL exists specifically for the two-column case, where using AST for both would be ambiguous about which one a covariate effect refers to. Semantically the same construction as WT_BASE (per-subject baseline paired with a time-varying WT), but named with the register’s _BL lab-value suffix rather than _BASE, matching QTC_BL / HGB_BL / FERRITIN_BL. A dataset with no on-treatment AST measurements should set AST = AST_BL on every record, which reduces the ratio term to 1.

Hematology

HGB (canonical for hemoglobin)

• Description: Blood hemoglobin concentration.
• Units: g/L or g/dL – document the unit used in each model via covariateData[[HGB]]$units.
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (HGB / ref)^exponent.
• Source aliases: none; HGB is the common NONMEM / clinical-PK abbreviation.
• Example models: Yamada_2025_zolbetuximab.R (g/L, reference 118; exponent -0.374 on V1), Zou_2020_leuprorelin.R (g/dL, reference 13.6; exponent 2.30 on the resistant-fraction parameter rp of a PSA disease-progression model).
• Notes: Unit varies by paper (SI g/L, US g/dL; 1 g/dL = 10 g/L). The per-model covariateData[[HGB]]$units field is load-bearing.

HGB_BL (canonical for per-subject baseline hemoglobin concentration (initial-condition use))

• Description: Pre-treatment (or per-subject anchor-time) hemoglobin concentration used as the initial condition for a hemoglobin state variable in indirect-response / turnover anaemia models. Distinct from HGB (which is the time-varying or baseline-only hemoglobin value when entering a covariate effect on a structural PK parameter): HGB_BL is a static per-subject covariate that supplies the steady-state baseline for an Hb-state ODE, typically together with a kin = kout * HGB_BL synthesis-rate-from-baseline coupling so the Hb state starts at steady state. Parallel to FERRITIN_BL (baseline ferritin) and WT_BASE (baseline weight).
• Units: g/L, g/dL, or mmol/L – document the unit used in each model via covariateData[[HGB_BL]]$units.
• Type: continuous
• Scope: general
• Reference category: n/a – subject-level baseline supplied as a covariate column. Reference values observed: 8.3 mmol/L (Mulder 2025 chronic-HEV SOT cohort median).
• Source aliases:
  • HBBASE – used in Mulder_2025_ribavirin.R.
  • Hbsw – used in Chanu_2010_methoxyPolyethyleneGlycolEpoetinBeta.R (“the Hb concentration at the switch from the previous ESA treatment to C.E.R.A.”).
• Example models: Mulder_2025_ribavirin.R (mmol/L; initial condition for the hb_state compartment; per-subject kin = kout * HGB_BL so the Hb state is at steady state pre-RBV; the source paper substitutes the cohort typical (median) baseline for the one subject without an individual baseline value), Chanu_2010_methoxyPolyethyleneGlycolEpoetinBeta.R (g/dL; initial condition of the hb state and the driver of the SESA = (Hb0 - Hbsw) / LS prior-ESA-withdrawal term in a red-blood-cell life-span model; read only for previously ESA-treated subjects, with ESA-naive subjects starting at their own drawn pre-ESA baseline parameter instead).
• Notes: Promoted from specific to general on the Chanu 2010 extraction, which is the second paper to use a per-subject baseline hemoglobin covariate as the initial condition of a hemoglobin state – the condition this entry originally named for promotion. The two founding models instantiate the same idiom through different structures: Mulder 2025 couples it to an explicit kin = kout * HGB_BL turnover balance, Chanu 2010 uses it directly as hb(0) in a life-span (delay differential) model where the anchor differs from the model’s own estimated pre-ESA baseline parameter. Distinct from HGB (time-varying or baseline hemoglobin entering a covariate effect on a structural PK parameter) and from a PD baseline PARAMETER such as lrbase (a fitted typical value with IIV, not a data column). The unit varies widely across papers (SI g/L, US g/dL, Dutch / European convention mmol/L; 1 g/dL = 10 g/L; 1 mmol/L of tetrameric haemoglobin ~ 16.114 g/L); the per-model covariateData[[HGB_BL]]$units field is load-bearing.

TRANSF_BURDEN_BL (canonical for baseline red-blood-cell transfusion burden)

• Description: Pre-treatment red-blood-cell (RBC) transfusion burden, the number of RBC units transfused over a fixed look-back window before the first dose. A disease-severity covariate for transfusion-dependent anemias (beta-thalassemia, lower-risk myelodysplastic syndromes). Time-fixed per subject.
• Units: RBC units/24 weeks – the look-back window varies between trials (12, 16 or 24 weeks); document the window in covariateData[[TRANSF_BURDEN_BL]]$units and rescale before reusing a model with data on a different window.
• Type: continuous
• Scope: specific
• Reference category: n/a – continuous. Reference values observed: centred at 14 RBC units/24 weeks (Chen 2021, the rounded dataset median 14.1; range 0-34.0).
• Source aliases:
  • RBCT / RBCT burden (units/24 weeks) – Chen 2021 Table 1, Table 2 and the final covariate equations.
• Example models: Chen_2021_luspatercept.R (exponential effects exp(-0.0118 * (TRANSF_BURDEN_BL - 14)) on CL/F and exp(-0.0141 * (TRANSF_BURDEN_BL - 14)) on V1/F; higher transfusion burden, lower CL/F and V1/F).
• Notes: Distinct from any time-varying transfusion EVENT (an RBC transfusion given during the study, which is a dosing-like input to a hemoglobin model, not a baseline covariate) and from a binary transfusion-dependence indicator; a model that uses only “transfusion-dependent yes/no” should register a separate binary canonical. Promote to general if a second drug’s model uses the same count covariate.

WBC (canonical for white blood cell count)

• Description: Total white blood cell count (baseline or time-varying). In chronic lymphocytic leukaemia (CLL) populations the value is elevated because circulating leukaemic B-cells make up the majority of the count, so WBC can serve as a biomarker of target-cell burden rather than general hematology.
• Units: 10^9 cells/L (equivalent to 10^3 cells/uL). Document per-model via covariateData[[WBC]]$units.
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (WBC / ref)^exponent. Reference values observed: 10 x 10^9/L (Mould 2007, typical CLL Vmax normalization).
• Source aliases: none known (WBC is the universal clinical-PK abbreviation).
• Example models: Mould_2007_alemtuzumab.R (reference 10 x 10^9/L; exponent 0.194 on Vmax), Dirks_2008_cetuximab.R (reference 6.8 x 10^9/L; additive linear-deviation effect 0.0216 per 10^9/L on Vmax; baseline only – cetuximab does not deplete circulating leukocytes, so WBC is treated as a static covariate), Gibiansky_2021_rituximab.R (CLL; reference 100 x 10^9/L; baseline only; power exponent 0.223 on the time-dependent target-mediated clearance component cl_exp_component, consistent with leukaemic B cells carrying the CD20 target).
• Notes: Time-varying in treatment studies where the drug depletes the leukaemic clone (e.g., alemtuzumab in CLL): WBC must be supplied at every observation time in the event dataset. In diseases where WBC is not therapeutically targeted the column can be treated as a baseline-only covariate; record the per-model convention in covariateData[[WBC]]$notes.

NLR (canonical for neutrophil-to-lymphocyte ratio)

• Description: Ratio of absolute neutrophil count to absolute lymphocyte count from a complete blood count with differential. Used as a peripheral inflammation marker. May be reported as baseline only or as a time-varying covariate.
• Units: ratio (unitless)
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (NLR / ref)^exponent or exponential effects. Reference values observed: 2.11 (Lin 2024, median in pooled COVID-19 + non-infected cohort).
• Source aliases: none; NLR is the universal abbreviation in clinical-PK and inflammation-biomarker literature.
• Example models: Lin_2024_casirivimab.R (time-varying; reference 2.11; small positive exponent +0.029 on CL).
• Notes: Document baseline-vs-time-varying status in covariateData[[NLR]]$notes. Although it derives from WBC differential counts, register it as its own canonical because the ratio (not the absolute counts) is what the model uses.

HCT (canonical for hematocrit)

• Description: Hematocrit – packed red blood cell volume fraction (baseline or time-varying).
• Units: % (volume fraction times 100). Document per-model via covariateData[[HCT]]$units.
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (HCT / ref)^exponent. Reference values observed: 45 % (Nestorov 2014, study-population median for severe hemophilia A adults).
• Source aliases: none; HCT is the universal NONMEM / clinical-PK abbreviation. Note that source papers report hematocrit on either the percent scale (45) or the volume-fraction scale (0.45); the canonical column is percent, so a paper reporting fractions must have both its per-subject values and its reference value multiplied by 100 before they are recorded against this column (Zhou_2025_tacrolimus.R does exactly this: paper median 0.265 -> canonical reference 26.5 %).
• Example models: Nestorov_2014_factorviii.R (reference 45 %, exponent -0.419 on V1), Zhou_2025_tacrolimus.R (reference 26.5 % = the Zhou 2025 cohort median of 0.265 rescaled to percent, exponent -0.71 on tacrolimus CL/F; a negative exponent on clearance rather than on volume – see Notes).
• Notes: Higher HCT (more red-cell volume) leaves a smaller plasma fraction within total body volume; for plasma-restricted distribution (e.g., factor VIII activity, which circulates in plasma) the central volume of distribution decreases as HCT rises, so the exponent is negative. Document baseline-vs-time-varying status in covariateData[[HCT]]$notes. Distinct from HGB (mass concentration of hemoglobin); the two correlate but enter different mechanistic relationships. Two distinct mechanisms put HCT on different parameters with the same negative sign. (a) On volume, via the plasma-fraction argument above (Nestorov 2014 factor VIII). (b) On apparent clearance of an erythrocyte-partitioned drug: tacrolimus is extensively bound to red blood cells, so whole-blood clearance falls as HCT rises because a larger share of the measured whole-blood drug is sequestered in erythrocytes and unavailable to hepatic metabolism. Because the assay measures whole blood, the negative exponent on CL/F is an artefact-free consequence of the partitioning, not a change in intrinsic clearance (Zhou 2025 Discussion: ‘Recipients with low levels of HCT have more free tacrolimus in the plasma, resulting in a corresponding increase in clearance’). The exponent magnitude is reproducible across independent lung-transplant cohorts: -0.71 (Zhou 2025) versus -0.868 (Cai et al., cited in Zhou 2025 Discussion). Because the effect enters as a ratio (HCT / ref)^exponent, a model that mistakenly mixes the percent and fraction scales between the column and the reference is off by a factor of 100^exponent – a 24-fold error at exponent -0.71 – so the scale of both must be checked together, not separately.

THB_MASS (canonical for total hemoglobin mass)

• Description: Subject-level total hemoglobin mass in grams. Plasma-volume-independent quantity measured by the optimised CO-rebreathing method (Schmidt 2005, Pottgiesser 2008); distinct from HGB (mass concentration in plasma) and HCT (volume fraction). Used in erythropoiesis / RBC-regeneration models as the steady-state set point Base that drives the negative-feedback term and seeds the steady-state initial conditions of the precursor compartments.
• Units: g
• Type: continuous
• Scope: specific
• Reference category: n/a – subject-level baseline supplied as a covariate column. Reference values observed: 885.42 g (Tetschke 2018 Table 1 example subject); Pottgiesser 2008 cohort mean ~870 g across 28 estimable adult-male volunteers.
• Source aliases: Base (Tetschke 2018 paper symbol).
• Example models: Tetschke_2018_erythropoiesis.R (reference 885.42 g; Pottgiesser 2008 dataset of 29 healthy adult male volunteers).
• Notes: Specific scope because total hemoglobin mass requires the optimised CO-rebreathing method to obtain (Schmidt 2005), which is a specialised technique not present in routine clinical labs; promote to general if a second model registers this quantity. Distinct from HGB (g/L or g/dL plasma concentration) and HCT (RBC volume fraction): THB_MASS is the absolute body-pool mass and is not perturbed by short-term plasma-volume fluctuations (Pottgiesser 2008 Section 3.2 explicitly motivates the choice of mass over concentration). Sex-dimorphic: typical value in adult males is meaningfully higher than in adult females; document the sex composition of the population in covariateData[[THB_MASS]]$notes.

LYMPH_ABS (canonical for absolute peripheral-blood lymphocyte count (total))

• Description: Total absolute peripheral-blood lymphocyte count (all lymphocyte subsets pooled: T, B, and NK cells). Baseline or time-varying; document the time resolution per model via covariateData[[LYMPH_ABS]]$notes. Used as a continuous covariate on clearance and other PK parameters when the source paper carries the count as an exogenous exposure covariate rather than as a modelled PD state. Common in allo-HSCT / GvHD-prophylaxis popPK studies where circulating-lymphocyte burden is a mechanistic driver for lymphocyte-targeting biologics (integrin blockers, T-cell / B-cell-depleting mAbs).
• Units: cells/uL (equivalent to 10^9 cells/L = K/uL x 1000; document the per-paper reporting unit in covariateData[[LYMPH_ABS]]$units). Reporting conventions vary: some papers report as K/uL (i.e., 10^3 cells/uL, numerically identical to 10^9 cells/L) and some as cells/uL – both encode the same numerical value once the K/uL suffix is expanded. Prefer cells/uL as the canonical unit for consistency with CD4_ABS, CD19_ABS, and NEUT.
• Type: continuous
• Scope: general
• Reference category: n/a – typically enters as a power term (LYMPH_ABS / ref)^exponent against a paper-specific reference. Reference values observed: 100 cells/uL (Waterhouse 2024, 0.1 K/uL, exponent -0.0180 on CL; the paper further replaces any zero lymphocyte counts with 0.01 K/uL = 10 cells/uL to avoid a log-of-zero in the power form).
• Source aliases:
  • LYMPH – Waterhouse 2024 NM-TRAN column, time-varying absolute lymphocyte count in K/uL.
  • ALC – clinical laboratory abbreviation for “absolute lymphocyte count”; common in hematology / transplant papers. Same semantics; no value transformation.
• Example models: Waterhouse_2024_vedolizumab.R (time-varying absolute lymphocyte count on CL: cl *= (LYMPH_ABS / 100)^-0.0180; reference 100 cells/uL = 0.1 K/uL per paper Table 2; zeros replaced with 10 cells/uL to avoid log(0)).
• Notes: Distinct from CD4_ABS (a subset – CD4+ T-lymphocytes only), CD19_ABS (a subset – CD19+ B-lymphocytes only), and NLR (a ratio, neutrophil-to-lymphocyte). LYMPH_ABS is the total pooled lymphocyte count and is the correct canonical for papers that measure “absolute lymphocyte count” without further flow-cytometry subsetting. Distinct from WBC (total white blood cell count, of which lymphocytes are one differential fraction). Zero-value handling: allo-HSCT / lymphocyte-depleting-therapy cohorts routinely have observed lymphocyte counts of exactly 0 K/uL at some time points; source papers commonly substitute a small floor (e.g., Waterhouse 2024 uses 0.01 K/uL) before evaluating a power-form covariate effect, per Waterhouse 2024 Supplement Equation S2. Document any per-paper zero-floor convention in covariateData[[LYMPH_ABS]]$notes.

NEUT (canonical for absolute neutrophil count)

• Description: Absolute neutrophil count, typically as a baseline covariate (entered via centred-deviation (NEUT - ref) or power scaling (NEUT / ref)^exponent) or, in semi-mechanistic myelosuppression models, as a per-subject initial-condition value for the proliferation, transit, and circulating compartments.
• Units: cells/mm^3 (equivalent to cells/uL; i.e., the same value reported in 10^9/L x 1000). Document per-model via covariateData[[NEUT]]$units if the source paper uses a different unit (e.g., 10^9 cells/L for Ozawa_2007_docetaxel.R per the paper’s Table-3 reporting unit).
• Type: continuous
• Scope: general
• Reference category: n/a – used in centred-deviation form exp(coef * (NEUT - ref)), in power scaling (NEUT / ref)^exponent, or as a direct per-subject initial-condition assignment in semi-mechanistic Friberg-family models. Reference values observed: 4133 cells/mm^3 (BAST PTTE 2017 simulated cohort median; NA_NA_tte_gompertz.R Event 1 base hazard model); 5 x 10^9/L (Ozawa 2007 typical Japanese cancer cohort, used as the initial condition for the proliferation, transit, and circulation compartments).
• Source aliases:
  • BASE – per-subject baseline ANC supplied as a NONMEM data column (used in Ozawa_2007_docetaxel.R; Appendix I $INPUT).
  • ANC – the absolute-neutrophil-count source-column spelling (10^9/L); time-varying hazard modulator in Hansson_2013_sunitinib_os.R (the overall-survival Weibull hazard term exp(e_anc_haz * ANC +...)). No value transformation beyond the per-paper unit (document 10^9/L in covariateDataANC$units).
  • Neutrophil – the NONMEM $INPUT column in Ding_2026_vancomycin.R, annotated ; 10^9/L in that stream’s $PK block.
• Example models: NA_NA_tte_gompertz.R (BAST PTTE 2017 / DDMODEL00000243 Event 1 hazard model; centred at NEUT = 4133/mm^3; coefficient -1.56e-4 on the NONMEM rescaled scale, equivalent to exp(-1.56e-4 * (NEUT - 4133)) on the hazard), Ozawa_2007_docetaxel.R (Friberg-extension myelosuppression PD; per-subject baseline ANC supplied via the NEUT column, used as the initial condition for the proliferating, three transit, and circulating compartments per the Methods text ‘Circ (t = 0) was fixed at its observed value’), Hansson_2013_sunitinib_os.R (time-varying ANC as ANC, simulated from the upstream Hansson 2013 myelosuppression model and consumed as an OS-hazard predictor).
• Notes: General scope because absolute neutrophil count is a routine clinical-laboratory measurement that recurs across cytotoxic-chemotherapy myelosuppression models (centred-deviation hazard models, Friberg-family per-subject baseline initial conditions, and time-varying ANC outputs). Promoted to general scope on 2026-05-10 to support Ozawa_2007_docetaxel.R. The NEUT canonical units are cells/mm^3, but the reporting unit 10^9 cells/L (numerically NEUT_per_mm3 / 1000) is also common in oncology papers; per-model covariateData[[NEUT]]$units documents the per-paper unit. Distinct from WBC (total white blood cell count, of which neutrophils are the largest fraction in healthy adults) – NEUT is a specific differential-count subfraction. Also distinct from NLR (neutrophil-to-lymphocyte ratio), which is a derived ratio.

FERRITIN_BL (canonical for baseline serum ferritin concentration)

• Description: Pre-treatment (or per-subject anchor-time) serum ferritin concentration. Two complementary uses across iron-related models: (1) as the per-subject initial condition for a ferritin state variable in iron-overload disease-progression models (ferritin(0) <- FERRITIN_BL; Bellanti 2015), and (2) as a static per-subject covariate entering a power-law multiplier on hepcidin turnover rate constants in iron-status / menstrual-cycle turnover models (Angeli 2016). Distinct from a state-output ferritin trajectory: FERRITIN_BL is a static per-subject covariate (one value, supplied at simulation start); a state-output ferritin would evolve over time per a disease ODE.
• Units: ug/L (clinical reporting convention in iron-related papers; numerically equivalent to ng/mL). Document per-model via covariateData[[FERRITIN_BL]]$units.
• Type: continuous
• Scope: general
• Reference category: n/a – subject-level baseline supplied as a covariate column. Reference values observed: 2260 ug/L (Bellanti 2015 thalassaemia cohort median; range 393-8500 ug/L across 27 transfusion-dependent beta-thalassaemia major patients); 53 ug/L (Angeli 2016 healthy non-menopausal women; mean 53.14, range at 10th-90th percentile 18.6-97.4 ug/L).
• Source aliases: none known.
• Example models: Bellanti_2015_deferoxamine.R (ug/L; initial condition for the ferritin compartment; n=27 transfusion-dependent beta-thalassaemia major paediatric / adolescent cohort, median 2260, range 393-8500), Angeli_2016_iron_hepcidin.R (ug/L; per-subject end-of-cycle baseline used as a power-law multiplier (FERRITIN_BL / 53)^exponent on hepcidin elimination kout_hep (exponent -0.60) and on the hepcidin post-menses rebound krel_hep (exponent -1.95); higher baseline ferritin -> slower hepcidin elimination and a smaller rebound, consistent with iron-regulatory feedback).
• Notes: Iron storage protein; elevated in transfusional iron overload (beta-thalassaemia major, sickle-cell disease on chronic transfusion, MDS on transfusion support) and in some inflammatory states (acute-phase reactant); near normal in healthy adults (typically ~30-300 ug/L). Promoted from specific to general on 2026-06-03 with the Angeli 2016 iron / hepcidin extraction (the second model to register FERRITIN_BL with consistent baseline-ferritin semantics across two distinct mechanistic uses).

IRON_BL (canonical for baseline (pre-dose) serum total iron concentration)

• Description: Per-subject baseline serum total iron concentration (transferrin-bound plus any non-transferrin-bound iron), measured before drug administration. Static per-subject (or per-administration) covariate. The baseline definition varies by paper – a single pre-dose value, or an average over a drug-free baseline sampling window when serum iron shows a marked diurnal fluctuation – and MUST be recorded per model in covariateData[[IRON_BL]]$notes.
• Units: ng/mL (Zhang 2022 reporting convention). Clinical laboratories commonly report serum iron in ug/dL (1 ug/dL = 10 ng/mL) or umol/L (1 umol/L = 55.845 ug/L = 55.845 ng/mL); convert before use. Document per-model via covariateData[[IRON_BL]]$units.
• Type: continuous
• Scope: general
• Reference category: n/a – used with linear-deviation forms (1 + slope * (IRON_BL - ref)). Reference values observed: 1110 ng/mL (Zhang 2022 M1 healthy adults, median 6-h pre-baseline average); 642.00 ng/mL (Zhang 2022 M2 CKD-5HD, median pre-dose value); 660.80 ng/mL (Zhang 2022 M3 CKD-5HD, median pre-dose value).
• Source aliases:
  • Fe.av – Zhang 2022 M1: the average serum total iron over the 6 h of baseline-period sampling preceding the dosing day (healthy subjects; used because serum iron fluctuates over 24 h).
  • Fe baseline, Febaseline – Zhang 2022 M2 / M3: serum total iron at 0 h before administration (CKD-5HD patients). Same orientation, no value transformation.
• Example models: Zhang_2022_ferricPyrophosphateCitrate_iv.R (linear effect on Vd: vc *= 1 + 0.000771 * (IRON_BL - 1110); Table 2 footnote a), Zhang_2022_ferricPyrophosphateCitrate_dialysate.R (linear effect on CL/F: cl *= 1 - 0.000728 * (IRON_BL - 642); footnote b), Zhang_2022_ferricPyrophosphateCitrate_predialyzer.R (linear effect on CL/F: cl *= 1 - 0.000702 * (IRON_BL - 660.8); footnote c).
• Notes: Distinct from FERRITIN_BL (the iron-storage protein, ug/L) and from transferrin saturation; serum iron is the circulating transport pool. For iron-replacement products whose model output is the drug-derived increment above the endogenous baseline, IRON_BL is the baseline that must be added back to predict total serum iron. The covariate-effect parameter form is e_iron_bl_<param>.

RBC (canonical for red blood cell (erythrocyte) count)

• Description: Red blood cell (erythrocyte) count (baseline or time-varying).
• Units: 10^12 cells/L (clinical reporting convention; equivalent to 10^6 cells/uL). Document per-model via covariateData[[RBC]]$units.
• Type: continuous
• Scope: general
• Reference category: n/a – used with linear-deviation (RBC - ref) or power scaling (RBC / ref)^exponent. Reference values observed: 4.40 x 10^12/L (Zhang 2012 healthy Chinese adult cohort median across the four dosing groups in Table 1); 3.7 x 10^12/L (Jiang 2023 postoperative Chinese GIST cohort central value, Table 1).
• Source aliases:
  • RBC* – used in Zhang_2012_bivalirudin.R (the asterisk in Zhang 2012 Table 1 denotes the per-subject baseline demographic value).
• Example models: Zhang_2012_bivalirudin.R (linear-deviation effect on EC50: EC50_i = theta_EC50 * exp(eta_EC50) * (1 + 1.70 * (RBC - 4.40))), Jiang_2023_imatinib.R (power scaling on apparent oral clearance: CL/F = 9.72 * (RBC / 3.7)^0.49 * ...; the paper attributes the positive association to imatinib binding to and distributing into red blood cells, and reports that halving RBC lowers CL/F by 29%).
• Notes: Universal CBC component, distinct from HGB (hemoglobin mass concentration), HCT (hematocrit volume fraction), and WBC (white blood cell count). Document time-varying-vs-baseline-only status in covariateData[[RBC]]$notes. RBC and HCT correlate via mean corpuscular volume (HCT ~ RBC * MCV / 10); models that retain both should record the dependency in notes.

PLT (canonical for platelet count)

• Description: Circulating platelet (thrombocyte) count from a complete blood count. Baseline or time-varying; document the time resolution per model in covariateData[[PLT]]$notes. Used both as a hematologic covariate on drug disposition and as a marker of coagulation-system status in trauma, cardiac-surgery, oncology and hepatology cohorts. In severe traumatic injury an elevated admission platelet count marks an intact (or reactively-activated) hemostatic system, and it predicts faster clearance of antifibrinolytics.
• Units: 10^9 cells/L (numerically identical to K/uL and to 10^3 cells/uL; the three labels encode the same value). Document the per-paper reporting label via covariateData[[PLT]]$units.
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (PLT / ref)^exponent or linear-deviation forms. Reference values observed: 196 x 10^9/L (Stitt 2026; the centring constant printed inside Equation 1. Note that Stitt 2026 Table 2 gives the TAMPITI cohort median admission platelet count as 197 K/uL – a 0.5% discrepancy against the equation’s own divisor, resolved in favour of the printed equation); 200 x 10^9/L (Wang 2019; the centring constant printed inside Equation (e), against a Table I cohort median of 213 – here the divisor is an explicitly rounded median, the paper’s Methods stating that “Covconstant was fixed at a value similar to the population median of the covariate”, so the rounding is intentional and the printed equation again governs).
• Source aliases:
  • PLT – the universal clinical and NONMEM abbreviation; used as-is in Stitt 2026 Table 1 / Table 2 and Equation 1.
  • PLTC, PLAT, THROMBO – longer spellings seen in some control streams and lab-extract column headers. Same orientation, no value transformation.
  • K/uL – a unit label, not a column name, but it appears often enough in publication tables that it is recorded here: K/uL values are already on the canonical 10^9 cells/L scale and must NOT be rescaled.
• Example models: Stitt_2026_tranexamicAcid.R (median admission platelet count, reference 196 x 10^9/L; exponent +0.468 on TXA clearance – the largest of the three clearance covariates in that model, and the one the paper credits with outweighing allometric scaling in the paediatric extrapolation because the paediatric virtual cohort’s 205-433 K/uL range sits well above the adult median), Fukae_2024_valemetostat_plt_decrease.R (founding example; centred-and-scaled (PLT - 200) / 100 entering a Bayesian logistic exposure-response model for grade >= 3 platelet-count decrease, both on the logit intercept (odds ratio 0.230 per 100 x 10^9/L increase – the largest single covariate effect in the paper) and on the unbound-AUCss exposure slope (1.14)), Wang_2019_linezolid.R (power effect on linezolid clearance, (PLT / 200)^0.261, in 37 Chinese critically ill adults with and without shock; the only one of seventeen screened covariates to survive forward inclusion and backward elimination, and the shallowest of the three power effects registered here – a 3.15-fold clearance range across the whole 11-895 x 10^9/L observed span. Note the confounding direction: linezolid causes thrombocytopenia and accumulation from low clearance drives that toxicity, so in a cross-sectional TDM dataset a low count may be a consequence of low clearance rather than a cause, which the positive exponent cannot distinguish).
• Notes: The bare-name form follows the register’s established convention for routine clinical analytes (ALB, CRP, CREAT, HGB, WBC, RBC, HCT are all bare names at general scope) and the drop-the-BL-prefix convention documented under IL6 – baseline-vs-time-varying status belongs in per-model notes, not in the column name. Jiang_2023_imatinib.R carries a placeholder PLT entry in its own covariateData that predates this registration and describes itself as having “no canonical register entry”; that note is now stale and the entry can be reconciled against this canonical when that file is next touched. Distinct from WBC, RBC, HGB and HCT (the other CBC components) and from NEUT / LYMPH_ABS (differential subfractions of WBC) – platelets are a separate cell line with separate kinetics. Distinct from FIB and the other entries under Coagulation / hemostasis biomarkers, which are plasma protein activities rather than a cell count, and from MPV (mean platelet volume, not registered) which is a per-platelet size measure rather than a count. Platelet count is time-varying and therapeutically targeted in thrombopoietic and myelosuppression models; the observable platelet-count state predicted by such a model belongs in compartment-names.md, not here, so a model that both takes admission platelet count as a covariate and predicts a platelet-count endpoint should read this column for the covariate and use a state name for the endpoint.

PLT_BASE (canonical for per-subject observed baseline platelet count (time-fixed))

• Description: Per-subject observed baseline platelet count, time-fixed. Used when a model’s time-varying output (or a time-varying covariate) is the platelet count itself, so the observed pre-treatment value needs a column of its own – typically as a covariate on the individual baseline of a thrombocytopenia (Friberg-type myelosuppression) model.
• Units: 10^9 cells/L (same scale as PLT).
• Type: continuous
• Scope: general
• Reference category: n/a – continuous. Reference values observed: 197 x 10^9/L (Srimani 2022, safety-dataset median; Equation 19).
• Source aliases:
  • BL_i,obs – Srimani 2022 Equation 19 symbol for the observed individual baseline.
• Example models: Srimani_2022_ixazomib_platelets.R (linear centred effect on the individual baseline of a Friberg chain, BL_i = BL + 0.837 * (PLT_BASE - 197) + eta).
• Notes: Follows the <X>_BASE pattern of WT_BASE and CRCL_BASE for a time-fixed baseline paired with a time-varying quantity of the same analyte. Distinct from PLT, the time-varying (or baseline-only, when nothing else carries the analyte) platelet-count covariate, and from the modelled platelet-count state, which belongs in compartment-names.md.

CD8_PP65_MID, CD8_PP65_HI, CD8_PP65_NR (canonical for the baseline CD8+CD69+ pp65-stimulated T-cell category set)

• Description: Mutually exclusive binary indicators decomposing the baseline CMV-specific cell-mediated immunity assay: the percentage of CD8+ T cells co-expressing the early activation marker CD69 after ex-vivo stimulation with CMV pp65 antigen. CD8_PP65_MID = 1 for the >= 0.5% to < 2% stratum, CD8_PP65_HI = 1 for the >= 2% stratum, CD8_PP65_NR = 1 when the assay result was not reported. All three are 0 for the < 0.5% reference stratum. The quantity is a measure of pre-existing CMV-specific cytotoxic T-cell immunity, not a general leukocyte count: a higher percentage means the recipient retains more functional CMV-directed T-cell response, which predicts spontaneous viral control independently of antiviral drug exposure.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 for all three indicators = the < 0.5% stratum (the lowest CMV-specific T-cell immunity group).
• Source aliases:
  • Baseline CD8+ CD69+ pp65 stimulation group – printed row label in Sun 2025 Table S2 and in the Figure S3 / Table S3 parameter tables.
• Example models: Sun_2025_maribavir_cmv_clearance_wk8.R, Sun_2025_maribavir_cmv_clearance_wk16.R, Sun_2025_maribavir_neutropenia.R, Sun_2025_maribavir_sae.R (all four are logistic exposure-response regressions on the 238-patient AURORA maribavir arm; the indicator set enters the linear predictor additively, and the sign flips between endpoints – higher CMV-specific immunity raises the probability of confirmed CMV clearance but lowers the probability of neutropenia and of serious adverse events).
• Notes: Not reported is a FITTED level with its own estimated coefficient, not a missing-data code to be dropped: in Sun 2025 the assay was unreported for 46 of 238 patients (19.3%), and the fitted Not reported coefficient is materially non-zero for three of the four endpoints. A downstream user must therefore supply CD8_PP65_NR = 1 for such subjects rather than imputing them into the reference stratum. Distinct from CD8_CD38DR_PCT, which is a different marker panel (CD38/HLA-DR co-expression, a general immune-activation phenotype) reported as a continuous percentage; the two must not be aliased. The parallel CD4+CD69+ pp65 assay is reported in Sun 2025 Table S2 but was not retained in any final model, so no CD4_PP65_* canonical is registered here. Scope: specific until a second CMV / transplant-immunology paper ratifies the same three-cut decomposition – the 0.5% and 2% cut-points are assay- and laboratory-specific.

WBC_CAT_MID, WBC_CAT_HI, WBC_CAT_NR (canonical for the categorised baseline total white-blood-cell immune-function set)

• Description: Mutually exclusive binary indicators decomposing baseline total white blood cell count into the immune-function strata used by Sun 2025: WBC_CAT_MID = 1 for >= 2.7 to < 7 x 10^9/L, WBC_CAT_HI = 1 for >= 7 x 10^9/L, WBC_CAT_NR = 1 when the baseline count was not reported. All three are 0 for the < 2.7 x 10^9/L (leukopenic) reference stratum.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 for all three indicators = < 2.7 x 10^9/L, i.e. the leukopenic stratum, which carries the HIGHEST modelled neutropenia risk.
• Source aliases:
  • Immune function status based on white blood cells at baseline – printed row label in Sun 2025 Table S2; Total baseline WBC in Table S4.
• Example models: Sun_2025_maribavir_neutropenia.R (all three indicators carry negative coefficients on the neutropenia logit, i.e. every stratum above the leukopenic reference has lower treatment-emergent neutropenia risk, monotonically decreasing with baseline count).
• Notes: Registered as a decomposed indicator set rather than derived inside model() from the continuous general canonical WBC, for one reason that is not a matter of taste: the Not reported stratum is a FITTED level with its own coefficient (14 of 238 patients in Sun 2025) and cannot be expressed as a threshold on a numeric WBC column. Once one level must be supplied as data the whole set is supplied as data, so that the reference stratum is unambiguous. A downstream user holding a continuous baseline WBC derives WBC_CAT_MID = as.integer(WBC >= 2.7 & WBC < 7) and WBC_CAT_HI = as.integer(WBC >= 7) with WBC in 10^9 cells/L, and sets WBC_CAT_NR = 1 only where the count is genuinely absent. Distinct from WBC (general scope, continuous, the underlying measurement) – a model needing a smooth exposure-response in leukocyte count should use WBC, not this set. Scope: specific; the 2.7 and 7 x 10^9/L cut-points are this analysis’s immune-function definition, not a general clinical convention.

Coagulation / hemostasis biomarkers

AT_BL_UDL (canonical for per-subject baseline plasma antithrombin activity)

• Description: Per-subject baseline (predose) plasma antithrombin (AT) activity level, expressed as units of AT activity per deciliter. Because the Stachrom chromogenic AT assay is calibrated to pooled normal plasma, units/dL is numerically equivalent to % of normal (100 units/dL = 100% of normal). Time-fixed per subject: the single predose value measured before the first AT-concentrate dose. Used both as a covariate on the volume of distribution ((AT_BL_UDL / 60)^exponent) and as an additive endogenous baseline in the observation equation to represent ongoing endogenous AT production alongside dosed exogenous AT (Dansirikul B3 method).
• Units: units/dL (equivalent to % of normal on chromogenic anti-FIIa or anti-FXa AT-activity assays).
• Type: continuous
• Scope: specific
• Reference category: n/a – used with power scaling (AT_BL_UDL / ref)^exponent on VD in the source paper’s final model. Reference values observed: 60 units/dL (Moffett 2017 cohort baseline mean 59 +/- 17, cohort median approximately 60). Normal healthy paediatric range 85-130 units/dL; the cohort baseline mean 59 is depressed relative to healthy range because the patients under study had consumptive AT depletion secondary to critical illness / concurrent UFH.
• Source aliases:
  • BASE – used in Moffett_2017_antithrombin.R (Moffett 2017 Table 3 baseline antithrombin activity covariate). The paper reports the covariate value as units/dL; the same numerical value is equivalent to % of normal on the STA-Stachrom AT chromogenic assay.
• Example models: Moffett_2017_antithrombin.R (power effect on VD: VD_dL = 67.9 * (WT/70)^1 * (AT_BL_UDL / 60)^(-0.389); reference 60 units/dL from Moffett 2017 Table 3. Also enters the observation equation as an additive endogenous baseline: Cc_measured = central / vc + AT_BL_UDL, per Dansirikul 2008 B3 method described in Moffett 2017 Methods “Data handling”).
• Notes: Specific scope until a second paediatric or adult AT / anticoagulation-supplementation model ratifies the canonical. Distinct from INR_BASE (baseline international normalized ratio, an integrated coagulation-cascade readout) and from other coagulation biomarkers (VWF, FVIIIRECENT) which measure different clotting-cascade components. The dual role – both a VD covariate AND an additive constant in the observation equation – reflects the Dansirikul B3 method’s approach to modelling exogenous drug administration on top of an endogenous baseline. When simulating, the same subject-level AT_BL_UDL value must be supplied for both effects; the additive baseline term captures the ongoing endogenous production of AT during the drug-derived-concentration decay window.

ACT_BASE (canonical for baseline activated clotting time)

• Description: Pre-heparin baseline activated clotting time (ACT) in seconds – the clotting time of a whole-blood sample activated with a particulate activator (celite, kaolin or glass beads), read on a point-of-care coagulometer at the bedside or in the procedure room. Time-fixed per subject: measured once, before the first intraprocedural heparin dose. ACT is the standard intraprocedural monitor of high-dose unfractionated-heparin anticoagulation (cardiopulmonary bypass, percutaneous coronary intervention, catheter ablation) precisely because aPTT saturates at the heparin concentrations those procedures require, and because a whole-blood point-of-care assay returns a result in minutes. In heparin PK-PD models the baseline enters the ACT prediction as an additive constant, so the model returns to the subject’s own pre-heparin value when heparin is eliminated.
• Units: s
• Type: continuous
• Scope: general
• Reference category: n/a – subject-specific measured baseline. Document the per-model simulation default in covariateData[[ACT_BASE]]$notes; Edrich_2015_heparin.R uses the Edrich 2015 Table 1 group medians 144 s (no chronic anticoagulant), 155 s (dabigatran), 169 s (warfarin INR < 2) and 182 s (warfarin INR >= 2), which differ significantly by group and are one of that paper’s findings.
• Source aliases:
  • BASEACT – used in Edrich_2015_heparin.R (the per-subject data item added to every prediction in the Edrich 2015 supplement $ERROR block, Predicted_ACT = F*KACT + BASEACT).
  • ACT0, ACT_0, BL_ACT, baseline ACT – other spellings seen in cardiac-anaesthesia and electrophysiology data extracts and prose. Same orientation and scale, no value transformation.
• Example models: Edrich_2015_heparin.R (founding example; additive baseline in the ACT observation equation ACT = ACT_BASE + k_ACT * Cc, where k_ACT carries a four-level multiplicative factor for the patient’s chronic oral anticoagulant. Edrich 2015 fits no population baseline parameter at all – the baseline is read from the data – and imputed the group median for the 21-33% of patients whose pre-heparin ACT was missing).
• Notes: The bare token ACT must NOT be used as the canonical column name. A heparin PK-PD model’s own observation variable is the paper-named ACT (following the aPTT / ECT precedent for non-PK paper-named outputs in compartment-names.md), so a covariate column called ACT would collide with the observable inside model(). The _BASE suffix follows the established <ANALYTE>_BASE family (WT_BASE, BSA_BASE, CRCL_BASE, TBILI_BASE, INR_BASE, NPT_BASE, ALP_BASE) and is what disambiguates the two. The plain <ANALYTE>_BASE shape is preferred over a unit-suffixed ACT_SEC_BASE: PT_SEC needed its unit token because the register already recorded bare PT with an unrelated meaning under DIS_GERD, whereas nothing else in the register claims ACT, and the _BASE suffix alone resolves the observable collision. Not interconvertible with any other clotting-time canonical in this register. ACT, PT_SEC (prothrombin time in seconds, citrated plasma plus tissue factor, extrinsic arm) and aPTT are three different assays on different specimens with different activators and different dynamic ranges; INR_BASE and PTR are unitless ratios derived from prothrombin time, and NPT_BASE is a protein mass concentration in mg/L. An ACT is a number in the hundreds of seconds, so a baseline value near 1 is itself sufficient evidence that a paper’s column is an INR rather than an ACT. Because ACT is activator- and instrument-dependent, models using this column should record the device and its measurement range where the source states them; Edrich 2015 used a Hemochron Signature Elite whole-blood microcoagulation system with a Hemochron Jr cartridge (International Technidyne, Edison, NJ) with a 0-400 s range, and that 400 s ceiling right-censors much of its post-heparin data. Naming ratified by the maintainers (PMC4327399).

INR_BASE (canonical for baseline international normalized ratio)

• Description: Pre-medication baseline INR (international normalized ratio of prothrombin time). Time-fixed per subject (measured once, before the first warfarin dose). Used directly in the warfarin K-PD INR equation as an additive constant (INR = INR_BASE + inrmax * (1 - (coag_s3 + coag_l3)/2) per Xia 2024 supplement Section 1.1) so the simulated INR returns to the subject-specific baseline when the drug is removed.
• Units: (unitless ratio; INR has no units)
• Type: continuous
• Scope: general
• Reference category: n/a – subject-specific baseline. Default simulation value documented per-model in covariateData[[INR_BASE]]$notes; the Xia 2024 simulation uses the total-cohort mean of 1.13 (Table 1).
• Source aliases:
  • INR_BASE, BL_INR, INRBASE – pre-medication INR column in NONMEM data sets; document the source-column name per-model in covariateData[[INR_BASE]]$source_name.
• Example models: Xia_2024_warfarin.R (additive baseline in the INR observation equation; cohort mean 1.13, SD 0.59 per Xia 2024 Table 1), Ohara_2014_warfarin_s.R (additive baseline in Ohara 2014 Eq 3; cohort mean 1.05, SD 0.10 per Ohara 2014 Table 1).
• Notes: Distinct from a time-varying INR observation (the model’s observed INR variable). Healthy subjects with no anticoagulation typically have INR around 1.0; the Hamberg / Xia 2024 model treats deviations from 1.0 as a subject-specific covariate rather than an estimated parameter so the model returns to the observed baseline when warfarin is withdrawn.

NPT_BASE (canonical for baseline normal (fully carboxylated) prothrombin concentration)

• Description: Pre-medication baseline plasma concentration of normal prothrombin (NPT) – the fully gamma-carboxylated, coagulation-competent fraction of coagulation factor II, measured before the first vitamin-K-antagonist dose. Time-fixed per subject. This is a mass concentration of a specific protein species, not a clotting-time-derived index: it is distinct from INR_BASE and PTR, which are both unitless ratios derived from prothrombin time. Warfarin lowers NPT by inhibiting the vitamin-K-dependent carboxylation step, so NPT falls while total (carboxylated plus des-carboxy) factor II is comparatively unchanged; an assay that measures only the carboxylated fraction, such as the carinactivase-1 method of Ohara 2014, is therefore required.
• Units: mg/L (equivalently ug/mL, the unit used in the source literature)
• Type: continuous
• Scope: general
• Reference category: n/a – subject-specific measured baseline. Where a model centers a covariate effect on NPT_BASE, document the centering value per-model; Ohara_2014_warfarin_s.R centers on the 119 mg/L cohort median given in Shi 2024 Table 3 footnote d.
• Source aliases:
  • NPT0, NPT_0, NPT0_individual – Ohara 2014 / Shi 2024; the baseline value of the NPT turnover pool.
  • normal prothrombin / normal prothrombin concentration – prose form in the Japanese and Taiwanese warfarin pharmacometrics literature.
• Example models: Ohara_2014_warfarin_s.R (load-bearing in three places at once: the initial condition of the npt state, the zero-order synthesis rate kin = kout * NPT_BASE, and the covariate on the INR exponent lambda = 3.48 * exp(0.00588 * (NPT_BASE - 119)); cohort mean 118.2 mg/L, SD 22.1, n = 99 per Ohara 2014 Table 1).
• Notes: In indirect-response vitamin-K-antagonist models NPT0 is measured data rather than an estimated parameter – Ohara 2014 fitted no inter-individual variance term for it and excluded from the PD analysis the three patients whose NPT0 was missing – so it belongs in the data as a covariate column, exactly like INR_BASE. The _BASE suffix follows the established <ANALYTE>_BASE family (WT_BASE, CRCL_BASE, TBILI_BASE, INR_BASE,…). The abbreviation NPT is retained over a descriptive alternative (PROTHROMBIN_BASE) or coagulation-factor nomenclature (FII_BASE) because it is the standard term in this literature and because “factor II” would overstate what the column holds: NPT is specifically the carboxylated subfraction, not total factor II.

PT_SEC (canonical for raw prothrombin time in seconds)

• Description: Prothrombin time (PT) as reported by the coagulation laboratory in SECONDS – the clotting time of citrated plasma after recalcification in the presence of tissue factor, which reads out the extrinsic and common arms of the coagulation cascade. Typically time-varying, measured alongside each blood draw; document the time resolution per model in covariateData[[PT_SEC]]$notes. In critical-care population PK it functions as a marker of hepatic synthetic function and of sepsis-associated coagulopathy: the vitamin-K-dependent clotting factors are hepatically synthesised, so PT lengthens as synthetic capacity falls, and drugs cleared by hepatic metabolism show reduced clearance in the same patients. This column carries the RAW laboratory value, not a ratio and not an internationally normalised index.
• Units: seconds
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (PT_SEC / ref)^exponent. Reference values observed: 15 s (Wang 2024, the divisor printed inside the final-model equation; note that Wang 2024 does not tabulate the PT distribution anywhere, so the divisor is recoverable only from the equation itself and a downstream user has no published cohort spread to sample from).
• Source aliases:
  • PT – the universal clinical and NONMEM abbreviation, used as-is in Wang 2024 prose, its covariate-selection narrative, and its final-model equation. The bare token PT must NOT be used as the canonical column name, because this register already records PT as a source alias with an unrelated meaning: under DIS_GERD, “disease status / PT – Yang 2024 narrative and Table 2 footnote a (PT = 1 for patient, 0 for healthy volunteer)”. A dataset column named PT is therefore ambiguous between a clotting time and a patient-versus-healthy-volunteer indicator, and the _SEC unit suffix is what disambiguates it.
  • PTSEC, PT_S, PROTHROMBIN_TIME – longer or punctuation-free spellings seen in lab extracts and control streams. Same orientation and scale, no value transformation.
• Example models: Wang_2024_voriconazole.R (founding example; reference 15 s, exponent -0.875 on voriconazole CL per the final-model equation in Wang 2024 Section 2.2 – much the largest of that model’s four continuous covariate exponents. Time-varying: PT was measured on the day of each blood collection. The negative exponent is the expected direction, a longer prothrombin time marking worse hepatic synthetic function and hence slower clearance of a hepatically metabolised triazole).
• Notes: The unit suffix is deliberate and load-bearing; it is what separates this column from the two other prothrombin-time-derived canonicals, which are on different scales and are NOT interconvertible without information this column does not carry. PTR is the ratio of a subject’s PT to that SAME subject’s own pre-treatment baseline PT, so converting PT_SEC to PTR requires a per-subject baseline that most critical-care cohorts never measured; a model that normalises every subject by one shared cohort constant (as Wang 2024 does, dividing by 15 s) is using a raw value with a centring constant, not a ratio, and must use this column. INR_BASE is the international normalized ratio, a unitless index formed as (PT_patient / PT_normal)^ISI to make PT comparable across thromboplastin reagents; an INR is a number near 1 and can never be 15, so a divisor in the tens is itself sufficient evidence that a paper’s “PT” is seconds rather than INR. Note that Kim_2016_udenafil.R maps a paper labelled “PT” to INR_BASE, which is correct for that paper – its PT genuinely was an INR, cohort median 1.13 – and that precedent must not be extended to papers reporting seconds. NPT_BASE is a protein MASS CONCENTRATION in mg/L, not a clotting time at all. Because prothrombin time in seconds is reagent-dependent (the whole reason INR exists), models using this column should record the reagent or laboratory where the source paper states it; Wang 2024 does not. Departs from the register’s bare-name convention for routine clinical analytes (ALB, CRP, CREAT, HGB, WBC, RBC, HCT, PLT) only because of the PT name collision documented above; the unit-suffixed form follows the existing AT_BL_UDL / CONC_OXA_UM / DOSE_BALCINRENONE_MG precedent, where the unit is the disambiguator. Naming ratified by the maintainers (PMC11206427).

PTR (canonical for prothrombin time ratio (PT relative to baseline))

• Description: Prothrombin time ratio – the ratio of measured prothrombin time to the pre-treatment baseline prothrombin time of the same subject (or trial-arm population mean). Unitless. Used as a time-varying input biomarker that drives downstream pharmacodynamic / event-rate models for direct oral factor-Xa inhibitors and related anticoagulant pharmacology.
• Units: (unitless ratio)
• Type: continuous
• Scope: specific
• Reference category: n/a – continuous. PTR = 1 corresponds to no FXa-inhibitor effect (placebo / pre-dose baseline) and reproduces the placebo log-odds in Yoshioka_2018_FXa_inhibitors_mbma. Must be > 0 because downstream PD equations evaluate log(PTR).
• Source aliases:
  • PTR, PT_RATIO, x – in Yoshioka 2018 Eq. 1 / Eq. 2 the symbol x is used for the PT ratio; document the source-column name per-model in covariateData[[PTR]]$source_name.
• Example models: Yoshioka_2018_FXa_inhibitors_mbma.R (model-based meta-analysis: per-arm population-mean PTR is input; outputs are per-arm event probability of ischemic stroke/SE and major bleeding).
• Notes: Distinct from INR_BASE (a time-fixed baseline INR scalar used as an additive constant in warfarin K-PD models) and from PT_SEC (the RAW prothrombin time in seconds). The PTR / PT_SEC distinction is not cosmetic: PTR’s denominator is the subject’s OWN baseline, so it is unitless and centred at 1 by construction, whereas a paper that divides every subject’s raw PT by one shared cohort constant is using PT_SEC with a centring constant and must not be encoded here. PTR is time-varying and must be supplied externally (typically computed from an upstream popPK -> PT-ratio model for the FXa inhibitor of interest, e.g., Girgis 2014 rivaroxaban, Leil 2014 / Chang 2016 apixaban, Krekels 2016 / Koretsune 2015 edoxaban). Yoshioka 2018 corrects all PT measurements to RecombiplasTin reagent equivalence per Gosselin 2016 before computing the ratio; downstream models that consume PTR should document the reagent-correction convention they assume. Scope: specific until a second model ratifies the canonical name.

VWF (canonical for von Willebrand factor concentration)

• Description: Plasma concentration (or activity) of von Willebrand factor (VWF) – the multimeric carrier protein that binds and protects circulating factor VIII (FVIII) from proteolytic degradation and rapid clearance. Used as a covariate on FVIII (and FVIII-Fc) clearance because the vast majority (>95%) of circulating FVIII is in complex with VWF.
• Units: IU/dL (equivalent to % of pooled normal plasma); document per-model via covariateData[[VWF]]$units. Some sources report VWF:Ag (antigen) versus VWF:RCo (ristocetin cofactor activity); record which assay was used in covariateData[[VWF]]$notes.
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (VWF / ref)^exponent. Reference values observed: 118 IU/dL (Nestorov 2014, study-population median).
• Source aliases: none; VWF is the universal abbreviation. Source papers may write vWF (lowercase v) or specify the assay (VWF:Ag).
• Example models: Nestorov_2014_factorviii.R (reference 118 IU/dL, exponent -0.343 on CL; VWF antigen).
• Notes: Higher VWF protects FVIII from clearance, so the exponent on CL is negative. VWF is time-varying within an individual (acute-phase response, age, blood group, etc.), but most published population PK models use baseline-only VWF when the within-subject dynamics are not characterized; document the per-model convention in covariateData[[VWF]]$notes.

FVIIIRECENT (canonical for most recently measured untreated FVIII:C activity)

• Description: The patient’s most recently measured plasma factor VIII coagulant activity (FVIII:C) in the absence of any treatment effect on the measurement – i.e. the patient’s current endogenous (untreated) FVIII:C level. In the desmopressin (DDAVP) setting it is the value obtained at most 1 day prior to a DDAVP test / treatment administration, used as a per-occasion covariate that indexes the patient’s endogenous FVIII synthetic capacity (the source paper found it more predictive of the DDAVP-triggered FVIII:C response than the alternative FVIII-lowest, ever-lowest measurement). In the FVIII-concentrate replacement setting it is the measured endogenous baseline FVIII:C that sits under the exogenous (drug-derived) FVIII and is added to the model prediction in the observation equation. Distinct from the model’s observed FVIII:C time profile – FVIIIRECENT is the single pre-treatment anchor value.
• Units: IU/mL (1 IU/mL = 100% of pooled normal plasma FVIII activity; equivalent to 100 IU/dL).
• Type: continuous
• Scope: general
• Reference category: n/a. Usage is model-specific: power scaling (FVIIIRECENT / ref)^exponent (Schütte 2018, reference 0.15 IU/mL = study-population median FVIII-recent in Table 1), or an additive endogenous offset on the predicted FVIII:C with no reference value (Preijers 2021 Methods Eq. 1, where severe hemophilia A patients have < 0.01 IU/mL and moderate patients 0.01 to < 0.05 IU/mL).
• Source aliases:
  • FVIII-recent, FVIII_recent, fviii_recent, FVIIIRECENT – the recently-measured FVIII:C column (Schütte 2018).
  • Cbase, C_base, “measured endogenous FVIII level”, baseline FVIII:C – the endogenous baseline offset (Preijers 2021).
  • Document the source-column name per-model in covariateData[[FVIIIRECENT]]$source_name.
• Example models:
  • Schutte_2018_desmopressin.R (reference 0.15 IU/mL; exponents +0.74 on baseline FVIII, -0.61 on V1, -0.73 on CL; Schütte 2018 Table 2 final covariate model).
  • Preijers_2021_factor_viii.R (additive endogenous baseline in the observation equation Cc = (C_pred + FVIIIRECENT) * (1 - 0.30 * FORM_FVIII_BDD); Preijers 2021 Methods Eq. 1). Set to 0 to simulate the drug-attributable FVIII only.
• Notes: General scope since the second model (Preijers 2021) registered it; the two founding models use it in different roles (power-scaled covariate on PK parameters vs. additive observation offset), so always read the per-model covariateData[[FVIIIRECENT]]$notes. Time-fixed per treatment episode (one pre-treatment measurement per desmopressin occasion or per surgical procedure). Schütte 2018 additionally defines a binary missing-data branch with flat correction factors (1.2 / 1.1 / 0.78 on baseline FVIII / V1 / CL) when FVIIIRECENT was unavailable for a fitted subject; per the standing nlmixr2lib pattern, the missing branch is documented in the validation vignette deviations rather than encoded as a separate MISSING_FVIIIRECENT covariate, and simulation users are expected to supply FVIIIRECENT for every simulated patient. Distinct from FVIII-lowest (ever-lowest historical FVIII:C), which Schütte 2018 tested but did NOT retain in the final covariate model.

DDIMER (canonical for plasma D-dimer concentration)

• Description: Plasma D-dimer protein concentration, the fibrin-degradation peptide produced by plasmin-mediated cleavage of cross-linked fibrin. Used in vascular / coagulation-pathology models as a circulating biomarker of fibrin turnover, intra-aneurysmal thrombus burden, or systemic fibrinolytic activity.
• Units: ng/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – used either with log10-transformed proportional scaling log10(DDIMER) / median(log10(DDIMER)) (Sherer 2012) or with categorical strata (Sherer 2012 sensitivity analysis groups: <=150, 151-300, 301-900, >900 ng/mL). Reference values observed: 326 ng/mL (Sherer 2012 cohort median; log10 approx 2.513).
• Source aliases:
  • C^(D-dimer) – used in Sherer_2012_AAA.R (the symbol in Sherer 2012 Methods equation page 2).
• Example models: Sherer_2012_AAA.R (proportional log10-transformed covariate on the baseline AAA growth rate beta1 (e_ddimer_b1 = 0.90 mm/year) and on the first derivative of growth rate with size beta2 (e_ddimer_b2 = 0.37/year)).
• Notes: Specific scope until a second model registers the canonical. Time-fixed (baseline-only) in Sherer 2012 because the HIMS cohort had a single follow-up D-dimer measurement; the source paper flags this as a limitation. Cohort interquartile range 142-785 ng/mL; extrapolation outside this range is not validated by the source. The log10 transformation reflects Sherer 2012’s finding that “differences in AAA growth were predominantly driven by patients with the highest plasma D-dimer concentrations.” Distinct from the time-varying biomarker columns in indirect-response / TMDD models – DDIMER enters Sherer 2012 as a baseline regression covariate, not as a dynamic exposure / response variable.

AAA_DIAM (canonical for baseline abdominal aortic aneurysm diameter)

• Description: Abdominal aortic aneurysm (AAA) maximum infrarenal diameter, ascertained by ultrasound at study entry. Used in vascular disease-progression models as the per-subject baseline severity covariate that anchors the typical-value regression for individual-level growth parameters.
• Units: mm
• Type: continuous
• Scope: specific
• Reference category: n/a – used in proportional form AAA_DIAM / median(AAA_DIAM) so the effect coefficients represent the contribution at the cohort median. Reference value observed: 32.7 mm (Sherer 2012 cohort median; q1 30.8, q3 36.0).
• Source aliases:
  • Y(0) – used in Sherer_2012_AAA.R (the symbol in Sherer 2012 Methods equation page 2; the baseline screening ultrasound diameter).
• Example models: Sherer_2012_AAA.R (proportional covariate on all three individual-level parameters: e_aaadiam_b0 = 32.6 mm on baseline size beta0, e_aaadiam_b1 = 2.03 mm/year on baseline growth rate beta1, and e_aaadiam_b2 = 0.59/year on the first derivative of growth rate with size beta2).
• Notes: Specific scope until a second model registers the canonical. Time-fixed (baseline-only) per subject – the value is the single screening ultrasound diameter; the time-evolving AAA diameter during follow-up is the model’s observation, not the covariate. Sherer 2012 inclusion criterion (HIMS cohort): 30-49 mm small AAA, so extrapolation outside this range to <30 mm (non-aneurysmal aorta) or >=50 mm (surgical-referral threshold) is not validated by the source.

FIB (canonical for plasma fibrinogen concentration)

• Description: Plasma fibrinogen concentration (Clauss assay or equivalent). Acute-phase protein synthesised in the liver; values rise in inflammation, infection, trauma, and pregnancy and fall in disseminated intravascular coagulation, severe hepatic failure, and congenital afibrinogenaemia. Used as a per-subject (or time-varying) covariate that proxies hepatic acute-phase response or inflammatory burden on the clearance of drugs whose elimination correlates with these processes.
• Units: umol/L (Taubert 2016 reports values in umol/L; the more common clinical units are g/L or mg/dL – document per-model via covariateData[[FIB]]$units. Approximate conversion using a molecular weight of 340 kDa: 1 umol/L ~ 0.34 g/L.)
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (FIB / ref)^exponent. Reference values observed: 13.0 umol/L (Taubert 2016 patient-group-1 median; a critically-ill ICU cohort with elevated baseline fibrinogen).
• Source aliases:
  • Fibrinogen – the spelled-out column header in Taubert 2016 Table 1 and supplementary Table S3; same biological quantity, no value transformation.
• Example models: Taubert_2016_linezolid.R (umol/L; reference 13.0; positive power exponent 0.04 on CL; quantitatively a weak effect at the cohort range but selected by stepwise covariate modeling).
• Notes: Distinct from FVIIIRECENT and VWF (specific coagulation-factor activities used in haemophilia / DDAVP-response models) – FIB is the routine plasma-fibrinogen concentration used as an acute-phase / inflammatory marker in popPK analyses. Distinct from CRP, IL6, LDH, and other acute-phase / cell-turnover biomarkers; FIB may correlate with these but is a separate measurement.

Disease severity scores

SCORE_EASI (canonical for Eczema Area and Severity Index)

• Description: Eczema Area and Severity Index score (atopic-dermatitis severity composite; bounded continuous, scale 0-72 with higher values = more severe disease).
• Units: (score)
• Type: continuous
• Scope: general
• Reference category: n/a – healthy volunteers have SCORE_EASI = 0. Effect enters as an additive term in models that pool AD patients with HV (e.g., Tiraboschi_2025_amlitelimab.R).
• Source aliases:
  • EASI – prior canonical name (pre-2026-06-19 standardization audit).
  • BEASI (baseline SCORE_EASI) – used in Tiraboschi_2025_amlitelimab.R.
• Example models: Tiraboschi_2025_amlitelimab.R, Serrano_2026_atopicDermatitis_placebo_mbma.R (STUDY-ARM MEAN baseline EASI in a model-based meta-analysis, centered on the across-trial mean of 29; additive -0.0486 logit-unit effect per EASI point on the maximum EASI-75 placebo response, i.e. sicker arms respond less to placebo).
• Notes: When used as a time-invariant baseline covariate (BEASI), document in covariateData[[SCORE_EASI]]$notes. In MBMA models the column carries an ARM MEAN rather than a subject value; record the grain and any centering constant in the per-model notes, as Serrano_2026_atopicDermatitis_placebo_mbma.R does. Canonical name is SCORE_EASI without the B prefix to match the AGE / WT / ALB pattern where baseline vs time-varying status is recorded in notes rather than the column name.

SCORE_DAS28CRP (canonical for 28-joint Disease Activity Score computed with C-reactive protein (DAS28-CRP))

• Description: DAS28-CRP rheumatoid-arthritis disease-activity composite (28 tender and swollen joint counts, C-reactive protein and patient global health; continuous, roughly 0-10, higher = more active disease). As a covariate it is usually the BASELINE value (per-subject, or the arm mean in an MBMA).
• Units: (score)
• Type: continuous
• Scope: general
• Reference category: n/a – continuous. Leil 2021 centres it at 6.2.
• Source aliases:
  • baseline DAS28 – Leil 2021 Table 2 / Table S3 label (arm mean, DAS28-ESR values converted to the CRP scale).
  • DAS28-ESR (convert, not rename) – Leil 2021 Eq. 1 / Results give SCORE_DAS28CRP = 0.899 * DAS28-ESR - 0.194 at the arm-mean level.
• Example models: Leil_2021_rheumatoidArthritis_das28_mbma.R (STUDY-ARM MEAN baseline; additive 1.60 * log(SCORE_DAS28CRP / 6.2) on the placebo Emax and exponent 1.30 on the placebo ET50 of the DAS28 change from baseline).
• Notes: Distinct from the das28 / das28cfb model OUTPUTS registered in compartment-names.md: this column carries the observed score as an input. The ESR-based variant is a different score (on average ~0.2-0.8 units higher) and should get its own SCORE_DAS28ESR canonical rather than share this column. In MBMA models record the grain (arm mean) and the centring constant in the per-model notes.

SCORE_PPNRS (canonical for Peak Pruritus Numerical Rating Scale score)

• Description: Peak Pruritus Numerical Rating Scale score – an 11-point (0-10) patient-reported measure of worst itch intensity over the previous 24 hours, where 0 is “no itch” and 10 is “worst itch imaginable”. Recorded daily and conventionally averaged over each week before modelling (“weekly average PP-NRS”). Higher values = more severe pruritus. The near-identical Worst Itch NRS (WI-NRS) used in dupilumab trials maps to this canonical; see Source aliases.
• Units: (score)
• Type: continuous
• Scope: general
• Reference category: n/a – used as a per-subject baseline anchor (the initial value of the modelled response) or as a screened covariate, not with a reference-value normalisation.
• Source aliases:
  • PP-NRS / PPNRS – the instrument’s usual printed name.
  • WI-NRS (Worst Itch Numerical Rating Scale) – the dupilumab prurigo-nodularis trials (PRIME, PRIME2) report WI-NRS rather than PP-NRS. Takechi 2025 Discussion treats the two as conceptually equivalent (“the PP-NRS and WI-NRS have similar response scales and recall periods, suggesting conceptual equivalence based on cognitive debriefing and usability testing in US adults with moderate-to-severe PN”) while cautioning that cross-trial interpretation should allow for the difference. No value transformation; record the instrument actually used in the per-model notes.
• Example models: Takechi_2025_nemolizumab_ppnrs.R (per-subject observed baseline, used as the initial value of the weekly average PP-NRS response; median 8.6, range 6.4-10).
• Notes: Member of the SCORE_<instrument> family (SCORE_EASI, SCORE_MADRS, SCORE_HAMD, SCORE_NIHSS, …), registered on the same family-conforming basis. Baseline-vs-time-varying status is recorded in the per-model notes rather than in the column name, matching the SCORE_EASI decision. When a model consumes the score as a per-subject baseline anchor there is deliberately no corresponding estimated baseline parameter in ini().

SCORE_WOMAC_PAIN, SCORE_WOMAC_STIFFNESS, SCORE_WOMAC_FUNCTION (canonical for WOMAC osteoarthritis subscale scores on the 0-10-per-item scale)

• Description: Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) subscale totals: pain (5 items), stiffness (2 items) and physical function (17 items). Each item is expressed on a 0-10 scale (0 = none, 10 = extreme), so the subscale ranges are pain 0-50, stiffness 0-20 and function 0-170. Higher values = worse symptoms. As a covariate it is usually the BASELINE value (per subject, or the arm mean in an MBMA).
• Units: (score)
• Type: continuous
• Scope: general
• Reference category: n/a – continuous. Wen 2022 centres each subscale at the median arm baseline (pain 25, stiffness 10.23, function 83.75).
• Source aliases:
  • BASE – Wen 2022 eMethods 5 NONMEM column (one control stream per subscale).
  • WOMAC Likert 3.1 version (0-4 per item; pain 0-20, stiffness 0-8, function 0-68) – convert, not rename: multiply by 2.5.
  • WOMAC VAS 3.1 version (0-100 mm per item; pain 0-500, stiffness 0-200, function 0-1700) – convert, not rename: divide by 10.
  • A subscale reported as the AVERAGE per item on 0-10 (as in Boucher_2018_naproxen_mbma.R, where WOMAC pain is on a 0-10 scale) – convert: multiply by the item count (5 / 2 / 17).
• Example models: Wen_2022_osteoarthritis_womacpain_placebo_mbma.R, Wen_2022_osteoarthritis_womacstiffness_placebo_mbma.R, Wen_2022_osteoarthritis_womacfunction_placebo_mbma.R (STUDY-ARM MEAN baseline; the starting value of the arm-mean placebo trajectory and a linear covariate on the placebo Emax, Emax * (1 + theta * (SCORE_WOMAC_<SUBSCALE> - median))).
• Notes: Member of the SCORE_<instrument> family; the subscale is a suffix because the three subscales are separate instruments in practice (trials report them separately, and a model uses one at a time). Unlike SCORE_EASI, the instrument has several published item scales, so the canonical FIXES the 0-10-per-item standardization and a dataset on another version must be converted with the factors above. In MBMA models record the grain (arm mean) and the centring constant in the per-model notes. The matching model outputs are womacpain / womacstiffness / womacfunction in compartment-names.md.

SCORE_MGADL (canonical for Myasthenia Gravis Activities of Daily Living score)

• Description: Myasthenia Gravis Activities of Daily Living score – eight-item patient-reported outcome measure (each item 0-3), total 0-24, higher values = greater symptom severity and functional limitation.
• Units: (score)
• Type: continuous
• Scope: general
• Reference category: n/a – healthy participants (no gMG) have SCORE_MGADL = 0 by definition. Effect enters as a baseline covariate on MG-ADL response parameters in gMG cohorts.
• Source aliases:
  • MGADL – prior canonical name (pre-2026-06-19 SCORE_ family standardization).
• Example models: Valenzuela_2025_nipocalimab.R (reference 7 points; power-form effect on IDecplacebo and on the slope between MG-ADL change and IgG reduction).
• Notes: Baseline-only in Valenzuela 2025 (the observation is the absolute change from baseline MG-ADL). When used time-varying (e.g., in pure PD models driven by disease-progression dynamics), document in covariateData[[SCORE_MGADL]]$notes. Canonical name is SCORE_MGADL without a BL prefix to match the SCORE_EASI / AGE / WT / ALB pattern.

SCORE_BCVA (canonical for best-corrected visual acuity)

• Description: Best-corrected visual acuity score measured on the Early Treatment Diabetic Retinopathy Study (ETDRS) chart, expressed as the number of letters read correctly (0-100; higher values = better vision). Used as a baseline severity covariate in ophthalmology PK/PD models of anti-VEGF treatment.
• Units: ETDRS letters (0-100)
• Type: continuous
• Scope: specific
• Reference category: n/a – used as a baseline input to set the initial condition of an indirect-response SCORE_BCVA state or as a power-form effect on response parameters. Reference value observed: 55 letters (Mulyukov 2018 narrative: mean study-population baseline SCORE_BCVA).
• Source aliases:
  • BCVA – prior canonical name (pre-2026-06-19 standardization audit).
  • BVA (baseline visual acuity) – used in Mulyukov_2018_ranibizumab.R.
• Example models: Mulyukov_2018_ranibizumab.R (baseline SCORE_BCVA used as the center for the initial-condition draw g0 = SCORE_BCVA + eta_g0).
• Notes: Ophthalmology-specific. Baseline-only in Mulyukov 2018 (carried once per subject and used only as the starting SCORE_BCVA for the indirect-response model). Canonical name drops the B prefix to match the SCORE_EASI / AGE / WT / ALB pattern (baseline-vs-time-varying status recorded in covariateData[[SCORE_BCVA]]$notes). Scope is specific until a second ophthalmology model ratifies the name; at that point promote to general.

PREV_AE_SCORE (canonical for previous-time-step ordinal adverse-event score)

• Description: Ordinal adverse-event score recorded at the immediately preceding observation time, used as a Markov-state covariate that conditions the current-time logit / probability calculation on the previous outcome. Integer 0..N where N is the maximum AE grade in the source-paper grading scheme; convention is that the value is 0 at the first observation (no prior AE).
• Units: (ordinal score; document per-model the scale in covariateData[[PREV_AE_SCORE]]$notes)
• Type: count
• Scope: general
• Reference category: n/a – typically used as a categorical conditioner (IF (PREV_AE_SCORE == 0)...) or via piecewise-FPS indicator decomposition; the natural reference is PREV_AE_SCORE = 0 (no prior AE).
• Source aliases:
  • PREVSCOR – used in Girard_2012_pimasertib.R (CTCAE 0..3 ocular-AE score).
  • PDV – used in Yoshida_2024_fazpilodemab.R (GIAE grade 0 / 1 / 2-3 combined, carried via the NONMEM IF (EVID.EQ.0.AND.TYPE.EQ.1) PDV=DV idiom in the DTMM control stream).
• Example models: Girard_2012_pimasertib.R (Markov-state covariate that selects per-previous-score logit thresholds b01/b11/b21 and b02/b12/b22 and per-previous-score emax levels; reset to 0 at TIME = 0 per source IF (TIME.EQ.0) PREVSCOR=0), Yoshida_2024_fazpilodemab.R (Markov-state covariate that selects per-previous-grade DTMM intercepts b1_g{0,1,2}_ae / b2b1_g{0,1,2}_ae and per-previous-grade discontinuation intercepts b_dc_g{0,1,2}; pooled grades 2 and 3 into a single PREV_AE_SCORE = 2 category per the source paper Methods page 545).
• Notes: Scope promoted from specific to general on 2026-06-30 alongside the Yoshida 2024 fazpilodemab extraction (the second model to use this canonical). The ordinal scale itself is paper-specific (different AE-grading schemes, different number of categories, different grouping rules – Girard 2012 collapses CTCAE grades 1+2 into a single “1-2” category and treats grades >=3 as a third stratum; Yoshida 2024 pools CTCAE grades 2 and 3 into a single PREV_AE_SCORE = 2 category) and must be documented per-model in covariateData[[PREV_AE_SCORE]]$notes; the COVARIATE CONCEPT (previous-time-step ordinal AE grade as a Markov-state conditioner) is general. When assembling the simulation event table, set PREV_AE_SCORE = 0 at the first observation of every subject and update each subsequent observation to the previous observation’s sampled score – matching the NONMEM IF (TIME.EQ.0) PREVSCOR=0 / PREVSCOR = DV carry-forward idiom. Distinct from PAIN (continuous baseline pain score) and from SCORE_MGADL / SCORE_EASI (continuous severity scores not modelled as Markov states).

SCORE_CDR_SOB (canonical for Clinical Dementia Rating - Sum of Boxes score)

• Description: Clinical Dementia Rating scale - Sum of Boxes (CDR-SOB) score, the unweighted sum of the six CDR-box scores (memory, orientation, judgement and problem solving, community affairs, home and hobbies, personal care; each scored 0 / 0.5 / 1 / 2 / 3). Total ranges 0-18; higher values indicate more severe cognitive and functional impairment. Widely used as a primary efficacy endpoint in Alzheimer’s-disease (AD) clinical trials and as a disease-progression biomarker in mild-cognitive-impairment (MCI) / prodromal-AD populations.
• Units: (CDR-SOB units, 0-18 score)
• Type: continuous
• Scope: general
• Reference category: n/a – the source paper centres covariate effects on dataset medians (e.g., Delor 2013 uses SCORE_CDR_SOB / 2 in the DOT power-form and SCORE_CDR_SOB - 1 in the mixture-logit additive form).
• Source aliases:
  • CDR_SOB – prior canonical name (pre-2026-06-19 standardization audit).
  • CDR_bsl – used in Delor_2013_alzheimer.R (baseline CDR-SOB at study entry).
  • CDR – alternative bare-name often seen in ADNI / CAMD-style NONMEM datasets.
• Example models: Delor_2013_alzheimer.R (time-fixed baseline covariate; enters both the per-subject DOT power form ((SCORE_CDR_SOB / 2)^e_cdr_sob_dot with e_cdr_sob_dot = -0.072) and the per-subject slow-progression mixture-logit additive form (+ e_cdr_sob_slow * (SCORE_CDR_SOB - 1) with e_cdr_sob_slow = -1.27)).
• Notes: Canonical name is SCORE_CDR_SOB without a _BL suffix to match the SCORE_EASI / SCORE_MGADL / SCORE_BCVA pattern (baseline-vs-time-varying status recorded in covariateData[[SCORE_CDR_SOB]]$notes). The CDR sum-of-boxes form is distinct from the global CDR rating (CDR_GLOBAL, a 0 / 0.5 / 1 / 2 / 3 ordinal); the sum-of-boxes is preferred in disease-progression modelling for its finer granularity.

SCORE_ADAS_COG (canonical for ADAS-cog total cognitive subscale score)

• Description: Alzheimer’s Disease Assessment Scale - cognitive subscale (ADAS-cog) total score. The total-11 form ranges 0-70; the modernised total-13 form ranges 0-85. Higher values = more cognitive impairment. The ADAS-cog is the most widely used cognitive endpoint in AD clinical trials and disease-progression modelling.
• Units: (ADAS-cog units; document the form (total-11 / total-13) per-model in covariateData[[SCORE_ADAS_COG]]$units)
• Type: continuous
• Scope: general
• Reference category: n/a – covariate effects typically centred on a dataset median (e.g., Delor 2013 centres on SCORE_ADAS_COG = 12.67).
• Source aliases:
  • ADAS_COG – prior canonical name (pre-2026-06-19 standardization audit).
  • ADAS_bsl – used in Delor_2013_alzheimer.R (baseline ADAS-cog total-11 at study entry).
  • ADAS, ADAS_COG_11, ADAS_COG_13 – alternative bare-name forms seen across ADNI / CAMD datasets.
• Example models: Delor_2013_alzheimer.R (time-fixed baseline covariate; ADAS-cog total-11 form; enters the per-subject DOT power form (SCORE_ADAS_COG / 12.67)^e_adas_cog_dot with e_adas_cog_dot = -0.0439).
• Notes: Canonical name is SCORE_ADAS_COG (no _BL suffix; baseline-vs-time-varying recorded in notes). The total-11 vs total-13 form must be documented per-model in covariateData[[SCORE_ADAS_COG]]$units because the same numeric SCORE_ADAS_COG value has different clinical interpretation across the two forms. Conrado 2014 uses ADAS-cog as the modelled observation (response variable) rather than as a baseline covariate; that model file therefore does not list SCORE_ADAS_COG in its covariateData.

SCORE_MMSE (canonical for Mini Mental State Examination score)

• Description: Mini Mental State Examination total score (0-30; higher values = better cognitive function). A widely used cognitive-screening instrument in AD and MCI populations.
• Units: (SCORE_MMSE units, 0-30 score)
• Type: continuous
• Scope: general
• Reference category: n/a – covariate effects typically centred on a dataset median (e.g., Delor 2013 centres on SCORE_MMSE = 26).
• Source aliases:
  • MMSE – prior canonical name (pre-2026-06-19 standardization audit).
  • MMSE_bsl – used in Delor_2013_alzheimer.R (baseline SCORE_MMSE at study entry).
• Example models: Delor_2013_alzheimer.R (time-fixed baseline covariate; modifies the per-subject disease-progression acceleration parameter alpha via a power form (SCORE_MMSE / 26)^e_mmse_alpha with e_mmse_alpha = -2.01).
• Notes: Canonical name is SCORE_MMSE (no _BL suffix; baseline-vs-time-varying recorded in notes). SCORE_MMSE is the inverse-direction counterpart of SCORE_CDR_SOB / SCORE_ADAS_COG (SCORE_MMSE high = healthy; SCORE_CDR_SOB / SCORE_ADAS_COG high = impaired); covariate-effect coefficient signs are therefore typically opposite to those for SCORE_CDR_SOB / SCORE_ADAS_COG.

SCORE_FAQ (canonical for Functional Assessment Questionnaire score)

• Description: Functional Assessment Questionnaire (Pfeffer SCORE_FAQ) total score: sum of ten functional-activities items (each scored 0 = normal to 3 = dependent), total 0-30; higher values = greater functional impairment. Used as a functional-status covariate alongside cognitive scores in AD and MCI populations.
• Units: (SCORE_FAQ units, 0-30 score)
• Type: continuous
• Scope: general
• Reference category: n/a – covariate effects typically centred on a dataset median (e.g., Delor 2013 centres on SCORE_FAQ = 1).
• Source aliases:
  • FAQ – prior canonical name (pre-2026-06-19 standardization audit).
  • FAQ_bsl – used in Delor_2013_alzheimer.R (baseline SCORE_FAQ at study entry).
• Example models: Delor_2013_alzheimer.R (time-fixed baseline covariate; enters the per-subject slow-progression mixture-logit additive form + e_faq_slow * (SCORE_FAQ - 1) with e_faq_slow = -0.341).
• Notes: Canonical name is SCORE_FAQ (no _BL suffix; baseline-vs-time-varying recorded in notes). Distinct from the cognitive scores (SCORE_CDR_SOB / SCORE_ADAS_COG / SCORE_MMSE): SCORE_FAQ measures instrumental activities of daily living rather than cognitive performance, and adds incremental information about disease-stage severity in MCI cohorts.

SCORE_RHPNM (canonical for normalized hippocampal volume (head-size and age-adjusted))

• Description: Normalised hippocampal volume: the subject’s average left+right hippocampal volume divided by the value expected for a healthy subject of the same age and estimated intracranial volume (head size). 1.0 corresponds to the healthy reference; values below 1.0 indicate hippocampal atrophy. Derived from MRI volumetry and a healthy-subject regression on age and intracranial volume.
• Units: (unitless ratio; 1.0 = healthy reference)
• Type: continuous
• Scope: general
• Reference category: n/a – covariate effects typically centred on SCORE_RHPNM = 1 (the healthy reference, e.g., Delor 2013).
• Source aliases:
  • RHPNM – prior canonical name (pre-2026-06-19 standardization audit).
  • SCORE_RHPNM – used in Delor_2013_alzheimer.R (baseline normalized hippocampal volume; Delor 2013 derivation: RHPNMbsl_i = HIPVbsl_i / HPNMbsl_i where HPNMbsl_i = Age_i * (-26.6268 + EICVbsl_i * 0.0016 + 3340.4395)).
• Example models: Delor_2013_alzheimer.R (time-fixed baseline covariate; enters the per-subject slow-progression mixture-logit additive form + e_rhpnm_slow * (SCORE_RHPNM - 1) with e_rhpnm_slow = 7.5, a strongly positive effect indicating that less atrophic hippocampi (SCORE_RHPNM closer to 1) are associated with a higher probability of being in the slow-progressing subpopulation).
• Notes: Distinct from the raw hippocampal volume (which would be a HIPV canonical not yet registered; raw HIPV is confounded with head size and age, hence the need for the normalisation). The Delor 2013 paper notes that the same effect is only marginally significant with unnormalised HIPV (P = 0.02) but strongly significant with the normalised form (SCORE_RHPNM). The exact age / EICV regression coefficients are paper-specific and any future model adopting this canonical should re-derive the normalisation for its own population or document why the Delor 2013 regression is reused.

SCORE_ROSS (canonical for modified Ross score of paediatric heart-failure severity)

• Description: Modified Ross score, a clinical heart-failure severity instrument for infants and young children. Integer sum of per-domain ordinal items across feeding / exertional, respiratory (tachypnoea, retractions), cardiac auscultation (S3, hepatomegaly), and perfusion domains. Scale 0-12 in the modified Reithmann / Ross / Connolly form used across paediatric heart-failure trials; lower values = milder heart failure, higher values = more severe. The paediatric analogue of the adult-only NYHA functional classification.
• Units: (score; integer 0-12)
• Type: count
• Scope: general
• Reference category: n/a – typically entered as an exponential effect exp(theta * (SCORE_ROSS - ref)) on a PK parameter, centred on a dataset median. Per-model reference value must be documented in covariateData[[SCORE_ROSS]]$notes. The exponential form is required over a power form because the score can legitimately be 0 (per Steichert 2025 Section 2.2.2 rationale).
• Source aliases:
  • Ross score – paper-prose form.
• Example models: Steichert_2025_enalapril_enalaprilat_pediatric.R (reference SCORE_ROSS = 4, the analysed-population weighted median; exponential effect theta = -0.15 (RSE 26.3%) on the apparent volume of distribution of enalaprilat Vd_ENAAT/F).
• Notes: Distinct from RACHS1 (Risk Adjustment for Congenital Heart Surgery, a perioperative surgery-risk category derived from procedure type) and from the adult-only NYHA classification. Count type rather than continuous because the score is an integer sum of per-domain ordinal items. Different modifications of the Ross score exist across paediatric cardiology (Reithmann 1989, Ross 1992 derivation; Connolly 2001, Laer / LENA 2015+ form); the specific modification used must be documented in covariateData[[SCORE_ROSS]]$notes per model.

SCORE_MADRS (canonical for Montgomery-Asberg Depression Rating Scale score)

• Description: Montgomery-Asberg Depression Rating Scale total score: ten clinician-rated items (apparent sadness, reported sadness, inner tension, reduced sleep, reduced appetite, concentration difficulties, lassitude, inability to feel, pessimistic thoughts, suicidal thoughts), each scored 0-6 at interview, total 0-60. Higher values = more severe depression. The standard severity instrument in antidepressant trials; remission is conventionally defined as a total score below 10.
• Units: (SCORE_MADRS units, 0-60 score)
• Type: continuous
• Scope: general
• Reference category: n/a – covariate effects are centred or normalised on a cohort value; record which assessment timepoint the column carries in covariateData[[SCORE_MADRS]]$notes. Reference value observed: 28.64 (the week-1 cohort value in Shigetome 2025).
• Source aliases:
  • MADRS_W1 – NONMEM $INPUT column name in Shigetome 2025 Text S3, carrying the score one week after starting treatment.
  • MADRS_BASE / MADRSW0 – the same instrument at baseline; screened but not retained in Shigetome 2025.
• Example models: Shigetome_2025_paroxetine_madrs.R (week-1 value; enters the time to half-maximal improvement as the added power term et50 + (SCORE_MADRS / 28.64)^2.38, so lower week-1 severity brings the improvement forward).
• Notes: Canonical name is SCORE_MADRS with no timepoint suffix, matching the SCORE_MMSE / SCORE_EASI / AGE / WT convention where baseline-versus-time-varying status and the specific assessment visit are recorded in covariateData[[SCORE_MADRS]]$notes rather than in the column name. Direction of severity is the same as SCORE_CDR_SOB / SCORE_ADAS_COG and the opposite of SCORE_MMSE (high = healthy), so covariate-effect signs are not transferable between them. When a model’s endpoint is derived from MADRS (e.g. an enhancement rate defined as the percentage reduction from the baseline score), the baseline score defines the endpoint and is not a covariate; only a separately-timed score such as the week-1 value belongs in this column. Distinct from the Hamilton Depression Rating Scale (HAM-D, see SCORE_HAMD), Beck Depression Inventory and Quick Inventory of Depressive Symptomatology, which are different instruments on different scales and need their own canonicals.

SCORE_HAMD (canonical for Hamilton Depression Rating Scale (HAM-D) total score)

• Description: Hamilton Depression Rating Scale total score, the oldest and most widely used clinician-rated severity instrument in antidepressant trials. The standard 17-item form (HAM-D-17 / HDRS-17) scores each item 0-2 or 0-4 for a total of 0-52; longer 21-, 24- and 29-item variants exist and score higher, so the item count of the version a study used must be recorded per model. Higher values = more severe depression. Conventional anchors on the 17-item scale: 8-16 mild, 17-23 moderate, 24 and above severe; remission is usually defined as a total of 7 or below. Carried either as a baseline severity covariate or, in meta-analytic dose-response models, as the study-arm mean baseline severity.
• Units: (SCORE_HAMD units, 17-item score 0-52)
• Type: continuous
• Scope: general
• Reference category: n/a – covariate effects are centred or normalised on a cohort value; record the centring constant, the item count of the scale version, and which assessment timepoint the column carries in covariateData[[SCORE_HAMD]]$notes. Centring value observed: 25 (the hard-coded centring constant in Maringwa 2025).
• Source aliases:
  • baseline – column name in the Maringwa 2025 Table S1 study-level dataset, carrying the arm-mean pre-randomization HAMD total.
  • B / B_ij – the paper’s mathematical notation for the same quantity in its Equations 2 and its displayed final equation.
• Example models: Maringwa_2025_venlafaxine_fluoxetine_mbma.R (study-arm mean baseline HAMD in a venlafaxine / fluoxetine MBMA; enters as the shared multiplicative term 1 + 0.0986 * (SCORE_HAMD - 25) on BOTH drug effects, so a more severely depressed arm is predicted to show a larger drug effect, and the term cancels from placebo arms because it multiplies the drug effect only).
• Notes: Canonical name is SCORE_HAMD with no item-count and no timepoint suffix, matching the SCORE_MADRS / SCORE_MMSE / SCORE_EASI convention where the scale version and the assessment visit are recorded in covariateData[[SCORE_HAMD]]$notes rather than in the column name. A separate canonical from SCORE_MADRS because HAM-D and MADRS are different instruments on different ranges (0-52 vs 0-60) with different item content, so neither the values nor a fitted covariate slope transfers between them; the SCORE_MADRS entry’s own notes call for this split. Direction of severity is the same as SCORE_MADRS (high = more severe) and the opposite of SCORE_MMSE. Two cautions for models that centre this covariate. First, a MULTIPLICATIVE centred term of the form 1 + slope * (SCORE_HAMD - c) has a root at SCORE_HAMD = c - 1/slope below which the modelled effect reverses sign, so the calibrated range must be stated (in Maringwa 2025 the root is at 14.9, just below the lowest arm mean of 14.1). Second, in a meta-analytic model the centring constant used in ESTIMATION need not equal the rounded cohort mean quoted alongside published PREDICTIONS – Maringwa 2025 centres at 25 but displays every prediction at 23 – so take the centring constant from the fitting code or control stream, not from a figure or table caption. When a model’s endpoint is derived from HAM-D (a change from baseline, or a percentage reduction), the baseline score defines the endpoint and is not a covariate; a study-arm mean baseline used to modulate the treatment effect, as here, remains a covariate.

ACUTE_MED_DAYS (canonical for baseline number of days/month of acute migraine medication use)

• Description: Baseline number of days per month on which acute migraine medication (triptans or ergot compounds) was used during the 28-day run-in period prior to first dose. Enters as a piecewise-linear shift on baseline migraine or moderate-to-severe headache days in migraine exposure-response models.
• Units: days/month
• Type: continuous
• Scope: specific
• Reference category: n/a – piecewise-linear shift with breakpoint at 5 d/mo: contributes 0 below 5 and slope * (ACUTE_MED_DAYS - 5) above 5 (Fiedler-Kelly 2020). The 5-day breakpoint reflects the clinical guideline for medication-overuse headache.
• Source aliases: “Baseline days/month of acute medications” – used in FiedlerKelly_2020_fremanezumab_em.R and FiedlerKelly_2020_fremanezumab_cm.R.
• Example models: FiedlerKelly_2020_fremanezumab_em.R (slope 0.438 d/d, episodic migraine), FiedlerKelly_2020_fremanezumab_cm.R (slope 0.460 d/d, chronic migraine).
• Notes: Specific scope because the variable is migraine-domain-bound. Time-fixed per subject (baseline-only). When future migraine E-R models register additional aliases or alternative breakpoints, document them per-model and consider promoting to general.

MIGRAINE_DAYS_BL (canonical for baseline number of monthly migraine days)

• Description: Baseline number of migraine days per month, counted over the screening / run-in period before first dose. Time-fixed per subject (no time-varying partner column).
• Units: days/month
• Type: continuous
• Scope: specific
• Reference category: n/a – power model (MIGRAINE_DAYS_BL / ref)^exponent. Reference value observed: 13 days/month (Baker 2020, population median of the pooled healthy + episodic + chronic migraine eptinezumab cohort).
• Source aliases:
  • MDBASE – sponsor NONMEM name in the FDA Clinical Pharmacology Review of BLA 761119 (eptinezumab), Table 12; used in Baker_2020_eptinezumab.R.
  • “baseline MMD” / “monthly migraine days” – the clinical abbreviation; spelled out here following the rule that new canonicals avoid specialty acronyms.
• Example models: Baker_2020_eptinezumab.R (exponent 0.044 on eptinezumab clearance; clearance at 4 vs 28 baseline migraine days differs by about 9%).
• Notes: Follows the <X>_BL baseline pattern and is the spelled-out sibling of ACUTE_MED_DAYS. Healthy participants have no migraine days; record in the model notes what value the source assigned them (Baker 2020 does not say; its model simulates the healthy reference subject at the 13-day reference value).

NEXAC12M (canonical for number of disease exacerbations in the 12 months before study entry)

• Description: Count of clinically significant disease exacerbations experienced in the 12 months immediately preceding study entry, supplied as a per-subject baseline covariate. Standard measure of pre-treatment disease instability in chronic airway disease (asthma, COPD) and other relapsing-remitting conditions; enters exposure-response models as a marker of how brittle the subject’s disease was before randomization. Time-fixed per subject.
• Units: count (events in the prior 12 months)
• Type: count
• Scope: specific
• Reference category: n/a – used with power scaling (NEXAC12M / ref)^exponent. Reference values observed: 1.0 (Zhang 2025, median of the pooled Phase 2b + Phase 3 uncontrolled moderate-to-severe asthma cohort; range 1-50).
• Source aliases:
  • PREEXAC (“number of prior exacerbations in the past year”) – used in Zhang_2025_dupilumab_fev1.R.
• Example models: Zhang_2025_dupilumab_fev1.R (power effect (NEXAC12M / 1.0)^-0.0411 on baseline pre-bronchodilator FEV1; subjects with more prior exacerbations start from a lower baseline lung function, though Zhang 2025 Table 4 shows the effect is not clinically meaningful – only -7% at the 95th-percentile count of 6), Jin_2025_benralizumab_aaer.R (log-linear effect + 0.17 * NEXAC12M inside the exponent of the baseline asthma exacerbation rate, i.e. each additional prior-year exacerbation multiplies the baseline rate by exp(0.17) = 1.19; Jin 2025 Resource 17 row beta_Base, PE).
• Notes: Distinct from [[HOSPRA]], which is a time-varying binary indicator of being hospitalised for a pulmonary exacerbation at the moment a spirometry value is recorded; NEXAC12M is a time-fixed baseline count over a fixed 12-month look-back window. Also distinct from the decomposed-band pattern used for baseline seizure counts (NSP3M_LT2 / NSP3M_7_50 / NSP3M_GT50): those source papers binned the count into categories, whereas Zhang 2025 uses the raw count continuously inside a power function, which cannot be expressed as band indicators. The exacerbation definition is disease- and protocol-specific (Zhang 2025: severe asthma exacerbations per the DRI12544 / EFC13579 protocols); document the source definition in covariateData[[NEXAC12M]]$notes per model. Register a differently-windowed canonical (e.g. a 6-month or 24-month look-back) rather than reusing this name if a future paper changes the window. Scope: specific until a second model ratifies the semantics; promote to general at that point.

DIS_AMYLOID_LOAD (canonical for systemic-amyloidosis whole-body amyloid load grade)

• Description: Ordinal whole-body amyloid load score in patients with systemic amyloidosis. Integer 0-3: 0 = no amyloid (healthy volunteers), 1 = small amyloid load, 2 = moderate amyloid load, 3 = large amyloid load. The grading combines organ-by-organ amyloid presence (liver, spleen, bone, adrenals, gut, heart) and SAP-scintigraphy uptake into a single per-patient severity grade at baseline. Time-fixed per subject.
• Units: (categorical 0-3)
• Type: categorical
• Scope: specific
• Reference category: 0 (no amyloid). Sahota 2015 Eq. 2 also treats category 1 as part of the reference for the V4 effect (categories 0 and 1 share a V4 multiplier of 1); only categories 2 and 3 carry a non-zero effect via the cumulative parameters e_amload2_vp_sap and e_amload2_vp_sap + e_amload3_vp_sap respectively.
• Source aliases:
  • AMLOAD – prior canonical name (pre-2026-06-19 DIS_ prefix standardization) and source-paper column in Sahota_2015_miridesap.R.
• Example models: NA_NA_miridesap.R (DDMODEL00000262; Sahota 2015 Eq. 2 multiplicative effect on SAP peripheral volume V4: V4 = V4_ref * (1 + e_amload2_vp_sap * I(DIS_AMYLOID_LOAD>=2) + e_amload3_vp_sap * I(DIS_AMYLOID_LOAD>=3)); reported effects e_amload2_vp_sap = 6.39 / e_amload3_vp_sap = 26.39 yielding ~7.4x V4 at moderate load and ~33.8x at large load), Sahota_2015_miridesap.R (paper-only extraction of the same Sahota 2015 final model with identical Eq. 2 effect on V4; values 6.39 / 26.39 taken from Table 2).
• Notes: Scope: specific because the grading scheme is amyloidosis-specific (Sahota 2015 Methods: “The whole body amyloid load covariate, AMLOAD, was a categorical score: 0 for no amyloid in healthy volunteers, 1 for small, 2 for moderate, and 3 for large”). The cumulative monotonic parameterisation in Sahota 2015 Eq. 2 encodes a positive-only step at each grade increment. Co-used with DIS_AMYLOID_LIVER for the binary hepatic-involvement modifier.

DIS_AMYLOID_LIVER (canonical for hepatic amyloid involvement indicator)

• Description: Binary indicator for the presence of amyloid in the liver as a separate organ involvement from the overall whole-body amyloid load. 1 = liver amyloid present at baseline; 0 = no liver amyloid. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no liver amyloid).
• Source aliases:
  • AMLIVER – prior canonical name (pre-2026-06-19 DIS_ prefix standardization) and source-paper column in Sahota_2015_miridesap.R.
• Example models: NA_NA_miridesap.R (DDMODEL00000262; Sahota 2015 Eq. 2 multiplicative effect on SAP intercompartmental clearance Q4: Q4 = Q4_ref * (1 + e_amliver_q4 * DIS_AMYLOID_LIVER); reported effect 4.01, yielding ~5x Q4 in patients with hepatic amyloid), Sahota_2015_miridesap.R (paper-only extraction of the same Sahota 2015 final model with identical Eq. 2 effect on Q4; value 4.01 from Table 2).
• Notes: Scope: specific because the covariate is amyloidosis-specific (Sahota 2015 Methods names AMLIVER alongside AMSPLEEN / AMHEART as organ-specific amyloid-involvement binary indicators; only AMLIVER was retained as a covariate in the final model). Used in combination with the global DIS_AMYLOID_LOAD grade so the model can express both general amyloid burden and the specific hepatic-clearance modifier (the SAP-CPHPC complex is cleared by the liver, motivating the hepatic-amyloid-specific Q4 effect).

DIS_ARF (canonical for binary acute-renal-failure indicator)

• Description: Binary indicator that the subject is in acute renal failure (equivalently, acute kidney injury) at the time of the modelled treatment. 1 = acute renal failure; 0 = normal renal function. Used when a population PK analysis pools subjects in acute renal failure with subjects whose renal function is intact and estimates a separate typical clearance in each stratum. Acute renal failure is commonly ascertained from urine output (anuria or oliguria) or from an acute creatinine rise, and in the critical-care popPK literature is frequently reported as a bare cohort yes/no flag with no accompanying creatinine-clearance value – which is exactly the situation this canonical exists to encode.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (normal renal function).
• Source aliases:
  • ARF – the Table 1 row label and the parameter-name suffix (CLbody_arf / CLbody_nrf) used in Zhang_2025_fluconazole.R.
  • URINE – the NONMEM $INPUT column in the Zhang 2025 supplement, a urine-output category whose level 4 denotes normal renal function; DIS_ARF = 1 - (URINE == 4).
• Example models: Zhang_2025_fluconazole.R (selects between two separately estimated residual body-clearance typical values sharing one IIV term: cl <- exp(lcl_arf * DIS_ARF + lcl_nrf * (1 - DIS_ARF) + etalcl), with lcl_arf = log(0.41 L/h) and lcl_nrf = log(1.25 L/h) from Zhang 2025 Table 2, reproducing the supplement’s $PK block CL = THETA(1)*EXP(ETA(1)) overridden by IF (URINE.EQ.4) CL = THETA(2)*EXP(ETA(1))).
• Notes: Distinct from the RENALIMP_MILD / RENALIMP_MOD / RENALIMP_SEV family, which are CHRONIC-impairment severity bands defined on creatinine clearance: those presuppose a measured CrCl and a stable baseline, whereas acute renal failure is an unstable acute state in which CrCl is neither measurable nor meaningful (the founding cohort is largely anuric or oliguric, so CrCl is near zero and carries no gradation). Do not substitute RENALIMP_SEV for DIS_ARF; the acute-versus-chronic distinction changes both the covariate’s ascertainment and the direction in which clearance is expected to recover. Distinct from URINE_FLOW (instantaneous mL/h) and URINE_VOL_24H (24-h residual diuresis), which are the continuous urine-output measurements from which an acute-renal-failure flag may be derived but which are not themselves the binary. Pairs naturally with RRT_CRRT_ACTIVE / RRT_HEMODIAL_ACTIVE and QBL / QEFF, because a cohort in acute renal failure is usually receiving renal replacement therapy and the model must then separate residual body clearance from extracorporeal clearance. Scope: specific; promote to general if a second paper pools acute-renal-failure with normal-renal-function subjects using the same bare-binary semantics. Registered in the DIS_<condition> family alongside the acute / critical-illness siblings DIS_CRITILL, DIS_ARDS, DIS_SEPSIS, and DIS_BURN_RECENT by the maintainers’ decision.

KDIGO_AKI_1 (canonical for KDIGO acute-kidney-injury stage 1 indicator)

• Description: 1 = the subject is in Kidney Disease: Improving Global Outcomes (KDIGO) acute-kidney-injury stage 1 (serum creatinine 1.5-1.9 x baseline or a rise of >= 0.3 mg/dL, or urine output < 0.5 mL/kg/h for 6-12 h) at the time of the modelled dose; 0 = otherwise. Member of the mutually-exclusive KDIGO_AKI_1 / KDIGO_AKI_2 / KDIGO_AKI_3 level-indicator trio that decomposes the ordinal KDIGO AKI stage 0-3; all three = 0 selects the stage-0 (no AKI) reference. Level indicators, not cumulative >= thresholds, following the ISS_II / ISS_III and DIS_CHF_NYHA2 / _NYHA3 / _NYHA4 precedents, so a paper that estimates one coefficient per stage maps one-to-one.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (KDIGO stage 0, when KDIGO_AKI_1, KDIGO_AKI_2 and KDIGO_AKI_3 are all 0).
• Source aliases:
  • KDIGO – the single ordinal 0-3 column of Pressiat_2022_amikacin.R (Pressiat 2022 Table 3 row ‘KDIGO stage 1/CL’); KDIGO_AKI_1 = (KDIGO == 1).
• Example models: Pressiat_2022_amikacin.R (Monolix categorical effects on clearance, cl <- exp(lcl + e_kdigo_aki_1_cl * KDIGO_AKI_1 + e_kdigo_aki_2_cl * KDIGO_AKI_2 + e_kdigo_aki_3_cl * KDIGO_AKI_3 + etalcl), coefficients -0.41 / -0.59 / -0.93).
• Notes: Mutually exclusive with KDIGO_AKI_2, KDIGO_AKI_3. The _AKI_ token disambiguates this ACUTE staging from the KDIGO CHRONIC-kidney-disease G/A staging (see UACR). Distinct from DIS_ARF, a single binary acute-renal-failure flag that would collapse the three stage-specific coefficients, and from the chronic RENALIMP_* severity bands defined on creatinine clearance. The covariate-effect parameter form is e_kdigo_aki_1_<param>.

KDIGO_AKI_2 (canonical for KDIGO acute-kidney-injury stage 2 indicator)

• Description: 1 = the subject is in Kidney Disease: Improving Global Outcomes (KDIGO) acute-kidney-injury stage 2 (serum creatinine 2.0-2.9 x baseline, or urine output < 0.5 mL/kg/h for >= 12 h) at the time of the modelled dose; 0 = otherwise. Member of the mutually-exclusive KDIGO_AKI_1 / KDIGO_AKI_2 / KDIGO_AKI_3 level-indicator trio that decomposes the ordinal KDIGO AKI stage 0-3; all three = 0 selects the stage-0 (no AKI) reference. Level indicators, not cumulative >= thresholds, following the ISS_II / ISS_III and DIS_CHF_NYHA2 / _NYHA3 / _NYHA4 precedents, so a paper that estimates one coefficient per stage maps one-to-one.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (KDIGO stage 0, when KDIGO_AKI_1, KDIGO_AKI_2 and KDIGO_AKI_3 are all 0).
• Source aliases:
  • KDIGO – the single ordinal 0-3 column of Pressiat_2022_amikacin.R (Pressiat 2022 Table 3 row ‘KDIGO stage 2/CL’); KDIGO_AKI_2 = (KDIGO == 2).
• Example models: Pressiat_2022_amikacin.R (Monolix categorical effects on clearance, cl <- exp(lcl + e_kdigo_aki_1_cl * KDIGO_AKI_1 + e_kdigo_aki_2_cl * KDIGO_AKI_2 + e_kdigo_aki_3_cl * KDIGO_AKI_3 + etalcl), coefficients -0.41 / -0.59 / -0.93).
• Notes: Mutually exclusive with KDIGO_AKI_1, KDIGO_AKI_3. The _AKI_ token disambiguates this ACUTE staging from the KDIGO CHRONIC-kidney-disease G/A staging (see UACR). Distinct from DIS_ARF, a single binary acute-renal-failure flag that would collapse the three stage-specific coefficients, and from the chronic RENALIMP_* severity bands defined on creatinine clearance. The covariate-effect parameter form is e_kdigo_aki_2_<param>.

KDIGO_AKI_3 (canonical for KDIGO acute-kidney-injury stage 3 indicator)

• Description: 1 = the subject is in Kidney Disease: Improving Global Outcomes (KDIGO) acute-kidney-injury stage 3 (serum creatinine >= 3.0 x baseline or >= 4.0 mg/dL, initiation of renal replacement therapy, or urine output < 0.3 mL/kg/h for >= 24 h or anuria for >= 12 h) at the time of the modelled dose; 0 = otherwise. Member of the mutually-exclusive KDIGO_AKI_1 / KDIGO_AKI_2 / KDIGO_AKI_3 level-indicator trio that decomposes the ordinal KDIGO AKI stage 0-3; all three = 0 selects the stage-0 (no AKI) reference. Level indicators, not cumulative >= thresholds, following the ISS_II / ISS_III and DIS_CHF_NYHA2 / _NYHA3 / _NYHA4 precedents, so a paper that estimates one coefficient per stage maps one-to-one.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (KDIGO stage 0, when KDIGO_AKI_1, KDIGO_AKI_2 and KDIGO_AKI_3 are all 0).
• Source aliases:
  • KDIGO – the single ordinal 0-3 column of Pressiat_2022_amikacin.R (Pressiat 2022 Table 3 row ‘KDIGO stage 3/CL’); KDIGO_AKI_3 = (KDIGO == 3).
• Example models: Pressiat_2022_amikacin.R (Monolix categorical effects on clearance, cl <- exp(lcl + e_kdigo_aki_1_cl * KDIGO_AKI_1 + e_kdigo_aki_2_cl * KDIGO_AKI_2 + e_kdigo_aki_3_cl * KDIGO_AKI_3 + etalcl), coefficients -0.41 / -0.59 / -0.93).
• Notes: Mutually exclusive with KDIGO_AKI_1, KDIGO_AKI_2. The _AKI_ token disambiguates this ACUTE staging from the KDIGO CHRONIC-kidney-disease G/A staging (see UACR). Distinct from DIS_ARF, a single binary acute-renal-failure flag that would collapse the three stage-specific coefficients, and from the chronic RENALIMP_* severity bands defined on creatinine clearance. The covariate-effect parameter form is e_kdigo_aki_3_<param>.

SCORE_UPDRS_II (canonical for UPDRS subscale 2 (activities of daily living) score)

• Description: Score on subscale 2 (part II, activities of daily living) of the Unified Parkinson’s Disease Rating Scale, a clinician- or patient-reported measure of how far Parkinson disease impairs everyday function (speech, salivation, swallowing, handwriting, cutting food, dressing, hygiene, turning in bed, falling, freezing, walking, tremor, sensory complaints). Higher is worse. Used as a disease-severity covariate on exposure-response model parameters.
• Units: (UPDRS subscale 2 units, score 0-52)
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as the normalised power term (SCORE_UPDRS_II / 17)^exponent. The normalising constant 17 is the cohort median (Knebel 2012 Table I: mean 17.6, median 17, range 1-40).
• Source aliases:
  • UPDS – printed name in Knebel 2012 (individual-parameter equations and Table II row labels, e.g. *(UPDS/17)^theta6).
• Example models: Knebel_2012_istradefylline_offtime.R (power effect on the baseline percentage OFF time E0, on the maximum disease-progression/placebo-response effect EmaxP, and on the maximum istradefylline effect EmaxI; Knebel 2012 Table II thetas 6, 12 and 19).
• Notes: Specific scope. Member of the SCORE_<INSTRUMENT> family (SCORE_ADAS_COG, SCORE_MMSE, SCORE_CDAI, SCORE_MADRS, SCORE_EASI, SCORE_SLEDAI, SCORE_MGADL, SCORE_BCVA, SCORE_FAQ, SCORE_CDR_SOB, SCORE_MAYO_E); SCORE_MADRS is the closest structural analogue, being likewise a psychometric severity score entering a PD model as a normalised power term. The subscale suffix _II is load-bearing: the UPDRS has four parts with very different ranges and clinical meanings (part I mentation/behaviour/mood, part II activities of daily living, part III motor examination, part IV complications of therapy), and papers routinely covary on one part rather than the total. A model using the UPDRS total score, or another individual part, must register a sibling canonical (SCORE_UPDRS_TOTAL, SCORE_UPDRS_III,…) rather than overload this name. Note the distinct case of Lee_2011_parkinson_progression.R, where the UPDRS total score is the model’s endpoint (d_updrs, change from baseline) and therefore not a covariate at all.

OFFTIME_BL (canonical for baseline daily OFF time in Parkinson disease)

• Description: Per-subject baseline daily OFF time, i.e. the absolute number of awake hours per day spent in the OFF state (the state of reduced levodopa benefit, with re-emergent motor symptoms) before treatment. Recorded from patient diaries in Parkinson disease trials of adjunctive therapy in patients with levodopa-related motor response complications. Time-fixed per subject.
• Units: hr (hours per day)
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as the normalised power term (OFFTIME_BL / 6.3)^exponent. The normalising constant 6.3 is the cohort median (Knebel 2012 Table I: mean 6.39, median 6.3, range 0.25-17.8 hr).
• Source aliases:
  • BOFF – printed name in Knebel 2012 (individual-parameter equations and Table II row labels, e.g. *(BOFF/6.3)^theta13).
• Example models: Knebel_2012_istradefylline_offtime.R (power effect on the maximum disease-progression/placebo-response effect EmaxP and on the maximum istradefylline effect EmaxI, but deliberately NOT on the baseline percentage OFF time E0; Knebel 2012 Table II thetas 13 and 20).
• Notes: Specific scope. Member of the <MARKER>_BL baseline family (HAERATE_BL, PLAQUE_BL, BMD_BL, QTC_BL, HGB_BL, GLN_BL, FERRITIN_BL, TRACP5B_BL, PKK_BL, INS_BL); HAERATE_BL is the closest analogue, being likewise a baseline clinical-event burden rather than a laboratory value. Distinct from the endpoint of the founding model, which is the PERCENTAGE of awake time spent OFF rather than an absolute number of hours; the two are related through each subject’s awake-time denominator, which Knebel 2012 does not report, so this covariate cannot be derived from the endpoint or vice versa. Note the resulting asymmetry in the founding model: OFFTIME_BL is a covariate on EmaxP and EmaxI while the baseline of the endpoint itself is the structural parameter E0, and Knebel 2012 observed its strongest graphical covariate signal between the E0 random effect and baseline OFF time yet did not enter the covariate on E0. A model covarying on baseline ON time, or on baseline OFF time expressed as a percentage rather than in hours, should register a sibling canonical rather than overload this name. The baseline-ON-time sibling this note anticipated is now registered as ONTIME_BL, immediately below.

ONTIME_BL (canonical for baseline daily ON time in Parkinson disease)

• Description: Per-subject baseline daily ON time, i.e. the absolute number of hours per day spent in the ON state (the state of adequate levodopa benefit) before treatment. Recorded from patient diaries in Parkinson disease trials of adjunctive therapy in patients with levodopa-related motor response complications. Note that the diary recording window is usually shorter than the waking day and is set by the trial protocol, so the value is an hours-per-window figure and the window must be recorded per model: Loprete 2016 used the 18 h window 0600-2400 of the Hauser diary, averaged over the 2 to 5 recording days preceding the visit, and counted ON time plus ON time with minor dyskinesia (excluding ON time with troublesome dyskinesia). Time-fixed per subject.
• Units: hr (hours per day)
• Type: continuous
• Scope: specific
• Reference category: n/a – entered UNCENTRED in the founding model, as a plain additive offset (ontime = ONTIME_BL + intercept + slope * time) rather than as a normalised power term. This is the structural opposite of its OFFTIME_BL sibling, which enters normalised by a cohort median.
• Source aliases:
  • BL – printed name in Loprete 2016 (equations 5 to 7 and the Table 4 caption, PD = BL + INTPLAC + INTTREAT + SLOPPLAC * TIMEmonths).
• Example models: Loprete_2016_safinamide_ontime.R (additive baseline of the daily-ON-time endpoint; Loprete 2016 equation 7 and Table 4. Founding example).
• Notes: Specific scope. Member of the <MARKER>_BL baseline family (OFFTIME_BL, HAERATE_BL, PLAQUE_BL, BMD_BL, QTC_BL, HGB_BL, GLN_BL, FERRITIN_BL, TRACP5B_BL, PKK_BL, INS_BL); registered as the sibling that OFFTIME_BL’s own notes require, because ON time and OFF time are not interchangeable and are not derivable from one another without the diary window and the asleep-time denominator, neither of which either source paper reports. The load-bearing distinction from OFFTIME_BL is ORIENTATION: a higher ONTIME_BL is a better baseline state, so a covariate or baseline coefficient carries the opposite sign to the same quantity expressed as OFF time, and an encoder who reads the two as the same axis will get every sign backwards. Also distinct in ROLE from OFFTIME_BL in its own founding model: there the baseline OFF time is a covariate on drug-effect parameters while the endpoint has its own estimated baseline E0, whereas here the baseline IS the endpoint’s baseline, supplied per subject as data instead of estimated (Loprete 2016 Table 4 reports the interindividual variance on BL as 0 FIX precisely because the observed individual values are used). A model that estimates a typical baseline ON time as a structural parameter should use the canonical lrbase instead of this covariate; use ONTIME_BL only when the paper feeds observed per-subject baselines in as data. A model using baseline ON time expressed as a PERCENTAGE of the recording window, rather than in hours, should register a sibling canonical rather than overload this name.

SCORE_NIHSS (canonical for National Institutes of Health Stroke Scale total score)

• Description: Total score on the National Institutes of Health Stroke Scale (NIHSS), the standard bedside instrument for quantifying neurological deficit after acute stroke. It sums 11 item scores covering level of consciousness, gaze, visual fields, facial and limb motor function, limb ataxia, sensation, language, dysarthria and extinction/inattention. The scale is reverse-oriented relative to most function scores: 0 is normal and 42 is maximal deficit, so a LOWER value is better. Register the RAW score here. Some NONMEM datasets carry a reversed companion column (cNIH = 42 - NIHSS, cBASE = 42 - BASE) so that the working variable increases with health; that reversed form is a data-preparation device and must NOT be stored under this canonical – convert back to the raw score first.
• Units: (NIHSS points, 0-42)
• Type: continuous
• Scope: general
• Reference category: n/a – entered uncentred in the founding models. Baseline-versus-time-varying status is recorded per model in covariateData[[SCORE_NIHSS]]$notes, following the SCORE_CDR_SOB convention of not carrying a _BL suffix on the canonical name.
• Source aliases:
  • BNH – baseline (admission) NIHSS, the NONMEM $INPUT name in the Karlsson 2010 NIHSS control stream; maps to the BASE column of that paper’s dataset excerpt.
  • BNIH – the same baseline NIHSS carried into the Karlsson 2010 Barthel Index dataset as a cross-scale severity covariate.
  • NIH / NIHSS – the current-occasion raw score.
• Example models: Karlsson_2010_acute_stroke_nihss.R (baseline NIHSS on the logit-scale relative-improvement magnitude, e_score_nihss_impr = -0.0602 per point, i.e. a more severe stroke improves proportionally less at each visit), Karlsson_2010_acute_stroke_bi.R (the SAME baseline NIHSS used cross-scale, on the logit-scale relative-improvement magnitude of the Barthel Index, e_score_nihss_impr = -0.0325 per point; founding examples).
• Notes: Member of the SCORE_<INSTRUMENT> family (SCORE_ADAS_COG, SCORE_MMSE, SCORE_CDR_SOB, SCORE_MADRS, SCORE_EASI, SCORE_UPDRS_II, …). The reverse orientation is the trap: a negative coefficient on SCORE_NIHSS means worse baseline stroke, not better, and the sign of any effect must be read against that. Karlsson 2010’s Table II footnote a makes the point explicitly – “A low NIHSS score is positive while a low BI score is negative for the patient, which is the reason for the opposite influence of the previous observation in the two models” – so a model that pairs NIHSS with an orientation-opposite scale such as the Barthel Index will show mirrored covariate signs by construction. Distinct from SCORE_BI_PREV (Barthel Index, higher is better) and from SCORE_NIHSS_PREV (the previous-occasion NIHSS Markov-state predictor); a model that uses both a fixed baseline NIHSS and a time-varying previous NIHSS needs both canonicals, as Karlsson_2010_acute_stroke_nihss.R does. A model using the Scandinavian Stroke Scale or the modified Rankin Scale should register a sibling canonical rather than overload this name – the ranges, orientations and item sets all differ. Ratified canonically alongside the Karlsson 2010 acute-stroke disease-progression extraction.

SCORE_GCS (canonical for Glasgow Coma Scale total score)

• Description: Total score on the Glasgow Coma Scale (GCS), the bedside instrument for grading impairment of consciousness from the sum of the best eye (1-4), verbal (1-5) and motor (1-6) responses. Range 3-15; 15 is fully conscious and 3 is deep coma, so a HIGHER value is better.
• Units: (score points, 3-15)
• Type: continuous
• Scope: general
• Reference category: n/a – continuous. Baseline-versus-time-varying status is recorded per model in covariateData[[SCORE_GCS]]$notes, following the SCORE_NIHSS convention of not carrying a _BL suffix on the canonical name. Reference value observed: 13 (Svensson 2020 cohort median at baseline).
• Source aliases:
  • GCSB – baseline GCS, the NONMEM $INPUT name in the Svensson 2020 survival control stream.
  • GCS – the current-occasion score.
• Example models: Svensson_2020_rifampicin_survival.R (baseline GCS as a linear factor on the hazard of death in tuberculous meningitis, (1 + e_gcs_haz * (SCORE_GCS - 13)) with e_gcs_haz = -0.256, i.e. each point below 13 raises the hazard by 25.6%).
• Notes: Member of the SCORE_<INSTRUMENT> family (SCORE_NIHSS, SCORE_MMSE, SCORE_CDR_SOB, …). Orientation is opposite to SCORE_NIHSS (there a lower score is better). Some datasets carry the three component subscores (E, V, M) separately; sum them before storing here, and register sibling canonicals if a model uses a subscore on its own. The pediatric GCS shares the 3-15 range and may be stored here with a note.

SCORE_NIHSS_PREV (canonical for previous-occasion NIHSS total score)

• Description: NIHSS total score recorded at the immediately preceding observation occasion, on the RAW 0-42 scale (lower is better). A Markov-state predictor: it conditions the current occasion’s transition probabilities and score reconstruction on the score the subject was last observed at, which is what makes a non-monotonic assessment-scale trajectory tractable as a sequence of discrete transitions. Time-varying per observation.
• Units: (NIHSS points, 0-42)
• Type: continuous
• Scope: general
• Reference category: n/a – entered uncentred in the founding model. At the first modelled occasion set it to the admission (baseline) score, matching the source dataset (Karlsson 2010 Supplement B excerpt: subject 1 has BASE = 17 and PNIH = 17 at day 7, then PNIH = 23 at day 29 after a day-7 score of 23).
• Source aliases:
  • PNH – the NONMEM $INPUT name in the Karlsson 2010 NIHSS control stream.
  • PNIH – the corresponding column label in that paper’s dataset excerpt.
• Example models: Karlsson_2010_acute_stroke_nihss.R (enters the not-maximum-score probability, e_score_nihss_prev_p_notmax = 0.474 per point; the logit-scale relative-decline magnitude, e_score_nihss_prev_decl = 0.0765 per point; and BOTH reconstruction rules nihss_improve = (1 - imz) * (SCORE_NIHSS_PREV + 1) and nihss_decline = dez * (42 - SCORE_NIHSS_PREV) + SCORE_NIHSS_PREV; founding example).
• Notes: Same Markov-state role as PREV_AE_SCORE, but that canonical is for ORDINAL adverse-event grades on a short 0..N toxicity scale used to select per-previous-grade intercepts, whereas this one is a wide continuous clinical assessment score entering linearly and driving a score-reconstruction rule; they are not interchangeable. Store the RAW score, not the reversed cNIH = 42 - NIHSS data-preparation column. When assembling a simulation event table, initialise SCORE_NIHSS_PREV to the baseline score at the first occasion and update each subsequent occasion to the score reconstructed at the previous one – the trajectory is generated recursively, not solved (see the Karlsson_2010_acute_stroke vignette for the loop). A model that instead conditions on the score two occasions back, or on a running minimum/nadir, needs its own canonical. Ratified canonically alongside the Karlsson 2010 acute-stroke disease-progression extraction.

SCORE_BI_PREV (canonical for previous-occasion Barthel Index total score)

• Description: Barthel Index total score recorded at the immediately preceding observation occasion, on the raw 0-100 scale in five-point steps (20 attainable values). The Barthel Index measures independence in activities of daily living (feeding, bathing, grooming, dressing, bowel and bladder continence, toilet use, transfers, mobility, stairs); higher is better, with 100 denoting full independent function. A Markov-state predictor with the same role as SCORE_NIHSS_PREV: it conditions the current occasion’s transition probabilities and score reconstruction on the last observed score. Time-varying per observation.
• Units: (Barthel Index points, 0-100)
• Type: continuous
• Scope: general
• Reference category: n/a – entered uncentred in the founding model.
• Source aliases:
  • PBI – the NONMEM $INPUT name and dataset-excerpt column label in the Karlsson 2010 Barthel Index control stream.
• Example models: Karlsson_2010_acute_stroke_bi.R (enters the improvement probability, e_score_bi_prev_p_improve = 0.0113 per point; the not-maximum-score probability, e_score_bi_prev_p_notmax = -0.0984 per point; the logit-scale relative-improvement magnitude, e_score_bi_prev_impr = 0.0156 per point; and BOTH reconstruction rules bi_improve = SCORE_BI_PREV + imz * (100 - SCORE_BI_PREV) and bi_decline = (1 - dez) * (SCORE_BI_PREV + 1); founding example).
• Notes: Registered in the previous-occasion form only, because the founding paper had no admission Barthel Index to register as a baseline: assessment at admission was not possible “for practical and ethical reasons”, and the authors tested imputed baselines between 0 and 20 (static and randomly drawn) and rejected them for insufficient predictive power, modelling the BI from day 7 onward instead (Karlsson 2010 Data Preparation). A future paper that does report a baseline Barthel Index should register a sibling SCORE_BI in the SCORE_<INSTRUMENT> family rather than overload this name. Orientation is the opposite of SCORE_NIHSS (higher is better here, worse there), which is why covariate signs on the previous score are mirrored between the two Karlsson 2010 models – see the Table II footnote a quoted in the SCORE_NIHSS entry. The coarse five-point granularity is load-bearing rather than cosmetic: the authors attribute the BI model’s weaker parameter precision to the scale offering only 20 attainable values against the NIHSS’s 42. Ratified canonically alongside the Karlsson 2010 acute-stroke disease-progression extraction.

Critical-illness severity

DIS_CRITILL (canonical for binary critical-illness / ICU-admission indicator)

• Description: Binary indicator that the subject is critically ill (equivalently, admitted to an intensive care unit) at the time of the modelled treatment. 1 = critically ill; 0 = not critically ill (ward inpatient or outpatient). Time-fixed per subject in the papers registered so far. Used when a population PK analysis pools critically ill and non-critically-ill subjects in one dataset and tests the contrast as a PK covariate – critical illness perturbs clearance and volume through capillary leak, altered protein binding, augmented or impaired renal clearance, and organ support, but is frequently reported as a bare yes/no cohort flag with no accompanying severity score.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (not critically ill).
• Source aliases:
  • critically ill – the covariate name used in Nguyen_2021_ganciclovir.R (Yang 2023 Table 3 footnote: “critically ill: 1 for critically ill patients and 0 for others”).
  • Health – the categorical covariate name used in VegasComitre_2021_amoxicillin_dog.R (Phoenix NLME; Vegas Comitre 2021 Table 4 footnote codes sick = 1, healthy = 0).
• Notes: Distinct from the continuous ICU severity scores SAPS_II and APACHE_II (which quantify how severely ill a patient is, and presuppose ICU admission), from ORG_FAIL_COUNT (integer count of failing organs), from MECH_VENT (a specific organ support), and from the ICU-admission-etiology indicators ICU_ADM_POLYTRAUMA / ICU_ADM_POSTSURG / ICU_ADM_MEDICAL (which classify why an already-admitted patient is in the ICU and are undefined for a non-ICU subject). DIS_CRITILL is the parent flag those etiology indicators presuppose: a dataset carrying any ICU_ADM_* indicator has DIS_CRITILL = 1 for those subjects. Do not derive DIS_CRITILL from ORG_FAIL_COUNT > 0 or from DIS_SEPSIS – a critically ill patient may have neither organ failure nor sepsis, and the founding paper reports the bare binary with no organ-failure count or sepsis diagnosis behind it. This canonical is only needed where a model contrasts critically ill against non-critically-ill subjects within one dataset; a study whose entire cohort is critically ill (e.g. Li_2021_ganciclovir.R, Krens_2020_ganciclovir.R, Horvatits_2014_ganciclovir.R) records critical illness as a population property in population$disease_state, not as a covariate column. Scope: general since VegasComitre_2021_amoxicillin_dog.R, the second paper to pool critically ill with non-critically-ill subjects using the same bare-binary semantics. The label is species-neutral: in that veterinary model the non-critically-ill group is healthy laboratory dogs, not ward patients.
• Example models: Nguyen_2021_ganciclovir.R (multiplicative power factor on clearance: cl = 2.55 * (WT/11.7)^0.75 * (CRCL/167)^0.763 * 0.806^DIS_CRITILL, so a critically ill child has clearance 19.4% lower than a non-critically-ill child of the same weight and eGFR; Yang 2023 Table 3), BeraldiMagalhaes_2021_ethambutol.R (a STRATUM SELECTOR rather than a multiplicative effect: every structural parameter – CL, Vc, ka on two occasions, F, Q, Vp – and its IIV was estimated separately for 10 mechanically ventilated ICU patients and 20 outpatients with tuberculosis, and DIS_CRITILL picks the _icu or _outpt value. All ICU patients also received the FDC tablet crushed via nasogastric tube while outpatients swallowed it whole, so here the indicator carries route/formulation handling as well as critical illness), VegasComitre_2021_amoxicillin_dog.R (veterinary: critically ill ICU dogs pooled with healthy laboratory beagles; exponential effects on clearance, exp(-0.829 * DIS_CRITILL), and on the superficial intercompartmental clearance, exp(-9.946 * DIS_CRITILL), plus health-specific residual error and sick-dog-only between-occasion variability), Zhao_2021_vancomycin.R (switch on the central volume: vc = 8.02 L non-ICU, 35.7 L ICU, encoded as exp(lvc + e_dis_critill_vc * DIS_CRITILL); Zhao 2021 Equation 2).
• Notes: Distinct from the continuous ICU severity scores SAPS_II and APACHE_II (which quantify how severely ill a patient is, and presuppose ICU admission), from ORG_FAIL_COUNT (integer count of failing organs), from MECH_VENT (a specific organ support), and from the ICU-admission-etiology indicators ICU_ADM_POLYTRAUMA / ICU_ADM_POSTSURG / ICU_ADM_MEDICAL (which classify why an already-admitted patient is in the ICU and are undefined for a non-ICU subject). DIS_CRITILL is the parent flag those etiology indicators presuppose: a dataset carrying any ICU_ADM_* indicator has DIS_CRITILL = 1 for those subjects. Do not derive DIS_CRITILL from ORG_FAIL_COUNT > 0 or from DIS_SEPSIS – a critically ill patient may have neither organ failure nor sepsis, and the founding paper reports the bare binary with no organ-failure count or sepsis diagnosis behind it. This canonical is only needed where a model contrasts critically ill against non-critically-ill subjects within one dataset; a study whose entire cohort is critically ill (e.g. Li_2021_ganciclovir.R, Krens_2020_ganciclovir.R, Horvatits_2014_ganciclovir.R) records critical illness as a population property in population$disease_state, not as a covariate column. Scope: specific; promote to general if a second paper pools critically ill with non-critically-ill subjects using the same bare-binary semantics.

ORG_FAIL_COUNT (canonical for number of organs failing in critically ill patients)

• Description: Integer count of failing organs in a critically ill patient at a given observation day, ascertained per-day and reported as the worst-of-day count. The count is decomposed into mutually exclusive strata 0 / 1 / 2 / 3 / >=4 (Vet 2016 used the strata 0, 1, 2, 3, and 4-or-5) that select per-stratum typical clearance values; the strata are not collapsed onto a single linear or power covariate effect because the underlying organ-failure mechanisms (cardiovascular, pulmonary, renal, hepatic, neurologic, hematologic) impair drug elimination heterogeneously.
• Units: (count)
• Type: categorical
• Scope: specific
• Reference category: 0 (no organs failing). Per-stratum typical CL values are estimated for ORG_FAIL_COUNT = 1, 2, 3, and >=4; ORG_FAIL_COUNT = 0 sets the baseline typical CL (frequently FIXED at the baseline value, as in Vet 2016).
• Source aliases:
  • ORGF – used in Vet_2016_midazolam.R (DDMODEL00000249 NMTRAN $INPUT column; values 0..>=4). Renamed to canonical ORG_FAIL_COUNT when assembling input data for the packaged model.
• Example models: Vet_2016_midazolam.R (per-stratum typical CL values: ORG_FAIL_COUNT=0 fixed at 1.6 L/h for a 5 kg child with CRP=32 mg/L; ORG_FAIL_COUNT=1 -> 1.29 L/h; ORG_FAIL_COUNT=2 -> 0.957 L/h; ORG_FAIL_COUNT=3 -> 0.842 L/h; ORG_FAIL_COUNT>=4 -> 0.678 L/h).
• Notes: Specific scope because the variable is critical-care-population-bound (PICU / ICU). Time-varying within subject (re-evaluated each ICU day). The Vet 2016 organ-failure ascertainment follows the Wilkinson 1987 paediatric multiple organ system failure (MOSF) criteria – the per-paper criteria, once confirmed by the maintainers, should be documented in each model’s covariateData[["ORG_FAIL_COUNT"]]$notes. Decompose inside model() into binary indicators (orgf1 <- (ORG_FAIL_COUNT == 1), orgf2 <- (ORG_FAIL_COUNT == 2), orgf3 <- (ORG_FAIL_COUNT == 3), orgf_ge4 <- (ORG_FAIL_COUNT >= 4)) and select per-stratum CL with mutually-exclusive multiplicative-flag arithmetic.

SAPS_II (canonical for new Simplified Acute Physiology Score II at ICU admission)

• Description: New Simplified Acute Physiology Score II (SAPS II) at intensive-care-unit admission. Validated 17-item ICU severity-of-illness score (Le Gall, Lemeshow & Saulnier, JAMA 1993;270:2957-2963) computed from age, vital signs, laboratory values, type of admission, and chronic-disease history during the first 24 hours after ICU admission; higher scores indicate greater severity and a higher predicted hospital mortality. Range theoretically 0-163; in adult ICU cohorts admission scores typically span ~10-100 with cohort means in the 35-65 range. Time-fixed per ICU stay.
• Units: points
• Type: continuous
• Scope: specific
• Reference category: n/a – used with power scaling (SAPS_II / ref)^exponent. Reference value observed: 50 points (Abboud 2009 typical-subject reference for the septic-shock cohort, mean SAPS II = 64 +/- 23).
• Source aliases:
  • SAPS II (with whitespace, as printed in the source paper’s prose) – used in Abboud_2009_epinephrine.R.
• Example models: Abboud_2009_epinephrine.R (power exponent -0.67 on epinephrine CL with reference 50; higher SAPS II is associated with lower clearance), Polito_2016_fludrocortisone.R (power exponents +0.019 on apparent oral fludrocortisone CL/F and +0.036 on the absorption lag time Tlag, both with reference 53 = the cohort median SAPS II among the 14 patients with detectable plasma concentrations; higher SAPS II is associated with longer absorption lag and faster apparent oral clearance).
• Notes: Specific scope because the column is critical-care-population-bound (adult ICU) and the score’s clinical meaning depends on the SAPS-II derivation rules; future ICU models reusing the score with the same definition can extend Example models rather than registering a new canonical. Should not be confused with SAPS I or SAPS 3 (different scoring rules / item sets) – register those under separate canonicals if a future paper uses them.

APACHE_II (canonical for Acute Physiology and Chronic Health Evaluation II score at ICU admission)

• Description: Acute Physiology and Chronic Health Evaluation II (APACHE II) score at intensive-care-unit admission. ICU severity-of-illness score (Knaus WA, Draper EA, Wagner DP, Zimmerman JE. Crit Care Med 1985;13(10):818-829) computed from the worst values of 12 acute-physiology variables during the first 24 hours after ICU admission, plus points for age and chronic health status. Integer range 0-71; higher scores indicate greater severity and higher predicted hospital mortality. Adult ICU cohorts typically span 5-40 with means in the 15-30 range. Time-fixed per ICU stay.
• Units: points
• Type: continuous
• Scope: specific
• Reference category: n/a – typically used with centred linear or power scaling. Reference values observed: 26 points (Swart 2004 midazolam learning-group mean APACHE II, used as the centring value in Q = 40.8 - (APACHE - 26) * 2.75); 22.5 points (Luo 2023 tigecycline cohort median, the centring value in CL = 11.30 - (APACHE_II - 22.5) * 0.14).
• Source aliases:
  • APACHE II-score on admission to ICU – printed row label in Swart 2004 Table 1; used in Swart_2004_midazolam.R.
  • APACHE – shortened name in Swart 2004 Table 5 formula Q = 40.8 - (APACHE - 26) * 2.75.
  • APACHE II – printed row label in Luo 2023 Table 1 and in the Results eq. 1 covariate expression; used in Luo_2023_tigecycline.R.
• Example models: Swart_2004_midazolam.R (centred linear effect on Q with reference 26: q = 40.8 - (APACHE_II - 26) * 2.75; each 1-point increase above 26 lowers Q by 2.75 L/h, capturing decreased intercompartmental clearance in more severely ill patients), Luo_2023_tigecycline.R (centred linear effect on CL with reference 22.5, the cohort median: cl = 11.30 - (APACHE_II - 22.5) * 0.14; each 1-point increase above 22.5 lowers tigecycline CL by 0.14 L/h. Luo 2023 prints this equation uncentred as CL = (11.30 - 0.14 x APACHE II) x e^0.065; the centring is restored because the uncentred reading contradicts the paper’s own Table 4 AUC0-24 values, its Monte Carlo PTA grid and the observed concentration distribution – see that model’s ini() comments and its validation vignette).
• Notes: Specific scope because the score’s clinical meaning is critical-care-population-bound (adult ICU) and the Knaus 1985 derivation rules must apply. Distinct from SAPS_II (Le Gall 1993, 17-item Simplified Acute Physiology Score II) – APACHE II and SAPS II are different scoring systems with different item weights and calibration; a single ICU popPK paper may report one, the other, or both. Following the precedent set by SAPS_II Notes (“Should not be confused with SAPS I or SAPS 3 … register those under separate canonicals if a future paper uses them”), APACHE II gets its own canonical. Future ICU popPK papers using APACHE II can extend Example models rather than registering a new canonical; scope may be promoted to general after multiple corroborating papers. Ratified canonically on 2026-07-26 alongside the Swart 2004 lorazepam / midazolam extraction.

SOFA (canonical for Sequential Organ Failure Assessment score)

• Description: Sequential Organ Failure Assessment (SOFA) score (Vincent JL et al. Intensive Care Med 1996;22(7):707-710): the sum of six organ-system sub-scores (respiratory, coagulation, hepatic, cardiovascular, central nervous system, renal), each 0-4, giving an integer total 0-24; higher scores indicate more organ dysfunction. Unlike APACHE_II (fixed at ICU admission) SOFA is routinely re-scored daily, so it may be time-varying within subject; record in per-model notes which day(s) the value refers to.
• Units: (score)
• Type: continuous
• Scope: general
• Reference category: n/a – typically enters as a median-centred power or linear effect. Reference values observed: 7 points (Wang 2020 caspofungin, median of the ECMO and non-ECMO groups on the sampling day).
• Source aliases:
  • SOFA / SOFA score – Wang 2020 Table 2 and Table 4 (‘Theta SOFA on Q’); used in Wang_2020_caspofungin.R.
• Example models: Wang_2020_caspofungin.R (power effect on intercompartmental clearance: q = exp(lq) * (SOFA / 7)^1.98; the Q equation is not printed in the source and the power form and centring follow its Methods).
• Notes: Many earlier extractions screened SOFA and did not retain it, recording it under the non-canonical name SOFA in covariatesDataExcluded (e.g. Chen_2024_nirmatrelvir.R, Sime_2019_tazobactam.R, Maseda_2018_micafungin.R, BorsukDeMoor_2018_tigecycline.R); this entry makes that name canonical. General scope because the score is used across adult critical-care populations. Distinct from APACHE_II and SAPS_II (different scoring systems) and from ORG_FAIL_COUNT (a count of failing organs, not a graded score). Paediatric variants (pSOFA) and SOFA sub-scores should get their own canonicals.

APACHE_II_SEV (canonical for elevated-APACHE-II severity stratum indicator)

• Description: 1 = the subject falls in the elevated-severity stratum of the APACHE II score, 0 = otherwise. Binary severity indicator DERIVED from the continuous APACHE_II score, for source papers that enter APACHE II severity as a discrete proportional shift rather than as a continuous scalar. The score threshold that defines the elevated stratum is paper-specific and must be documented in per-model covariateData[[APACHE_II_SEV]]$notes; where the source paper does not state it, say so explicitly rather than inventing one.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (not in the elevated-APACHE-II severity stratum).
• Source aliases:
  • APACHE II score – printed row label in Xie_2025_aztreonam_avibactam.R (Table S3 row APACHE II score on CL_AVI CL) and APACHE in Das 2024 supplementary Table 5 (APACHE effect on CL, CL*(1 + theta16)).
• Example models: Xie_2025_aztreonam_avibactam.R (proportional shift of -0.118 on avibactam clearance; aztreonam carries no APACHE term), Li_2019_avibactam.R (proportional shift of -0.197 on avibactam clearance, Table 2 theta15; the companion Li_2019_ceftazidime.R screened APACHE II and did not retain it, so the asymmetry within a fixed-ratio combination product is reproduced in two independent analyses of this lineage. This is the model that supplies the threshold – see Notes). Franzese_2022_avibactam.R (pediatric + adult refit of Li_2019_avibactam.R: proportional shift of -0.192 on avibactam CL, Table S3 theta15; children have no APACHE score and take 0; Franzese_2022_ceftazidime.R again does not retain it).
• Notes: Sibling of the continuous canonical APACHE_II, in the same relationship RENALIMP_SEV has to CRCL: APACHE_II holds the score in points, APACHE_II_SEV holds the derived stratum. Register both when a model needs the stratum, and do NOT overload APACHE_II with a 0/1 value – its declared type is continuous. Why binary rather than continuous in the founding model: neither Xie 2025 nor its predecessor Das 2024 prints the equation, but Das 2024 supplementary Table 5 writes the row as CL*(1 + theta16), which is the grammar Das uses for every binary population effect in that table (Population effect on Vc (cUTI), Vc*(1 + theta11)) and pointedly not the grammar Das uses for continuous covariates, which always carry the covariate symbol in the expression (AGE/35**theta14 on CL, (CrCL/80)**theta7). A raw continuous score is additionally impossible on arithmetic grounds: at a typical ICU APACHE II of 10, CL*(1 + (-0.118)*10) is negative. The founding model’s own notes record that the threshold is unstated in every available source, and its validation vignette holds the flag at the 0 reference throughout so that no reproduced published value depends on the inference. The threshold is APACHE_II > 10, supplied by Li_2019_avibactam.R – the earliest published analysis in this same ceftazidime-avibactam / aztreonam-avibactam lineage, whose Methods section states it twice: it lists ‘Acute Physiology and Chronic Health Evaluation version II (APACHE II > 10)’ among the screened markers of systemic disturbance, and justifies the cut as ‘predicted mortality rises steeply for scores > 10 (> 10% mortality), and this represents a reasonable cutoff for defining more severely ill patients’. Treat > 10 as the lineage default when a successor paper in this programme is silent, but confirm it per paper rather than assuming it globally – the cut is conventional for this programme, not for APACHE II in general. Note also that a MISSING score takes the 0 reference in Li 2019 rather than being excluded, because APACHE II was collected for cIAI and NP patients only and is absent for the entire cUTI cohort. Scope specific until a second model ratifies it.

SCORE_CCI (canonical for Charlson Comorbidity Index total score)

• Description: Charlson Comorbidity Index (CCI) total score: the weighted sum of a fixed list of chronic comorbid conditions (Charlson ME, Pompei P, Ales KL, MacKenzie CR. J Chronic Dis 1987;40(5):373-383), optionally with the Deyo/Quan ICD-code adaptations used to derive it from administrative or electronic-health-record data. Higher scores indicate a greater burden of chronic comorbidity and a higher predicted 1- and 10-year mortality. Integer, theoretically unbounded above but in practice 0-15 or so; adult inpatient and ICU cohorts typically have means in the 3-7 range. A CHRONIC-burden index and therefore time-fixed per admission, in contrast to the acute-physiology ICU scores that are recomputed from the first 24 hours of an ICU stay.
• Units: (SCORE_CCI units, weighted comorbidity count)
• Type: continuous
• Scope: specific
• Reference category: n/a – used with a ratio scaling (SCORE_CCI / ref) inside an exponential, or with centred linear or power scaling. Reference value observed: 5.62 (Chen 2025, the training-set mean CCI of a MIMIC-IV adult sepsis cohort). Note that Chen 2025’s ratio is UNCENTRED, so the covariate term does not equal 1 at the reference; check the printed equation before assuming a centring.
• Source aliases:
  • CCI – printed row label in Chen 2025 Table 1 and the symbol used in its equation 15; used in Chen_2025_vancomycin.R. Same quantity, no value transformation.
  • Charlson – printed row label in Chen 2025 supplementary Table S1 (Charlson on CL).
• Example models: Chen_2025_vancomycin.R (exponential effect on vancomycin clearance in adult ICU sepsis patients: cl = 3.35 * (CRCL/93)^0.997 * exp(-0.151 * (SCORE_CCI / 5.62)); each additional Charlson point multiplies clearance by exp(-0.151/5.62) = 0.973, so a patient at the cohort mean CCI of 5.62 clears vancomycin 14% more slowly than a comorbidity-free patient).
• Notes: Named with the SCORE_ prefix per the SCORE_<instrument> family established by the 2026-06-19 clinical-instrument standardization audit (SCORE_EASI, SCORE_MADRS, SCORE_NIHSS, SCORE_MMSE, …). The bare acronym CCI is deliberately NOT used because it is already overloaded in this registry with two unrelated meanings – chronic constriction injury, the neuropathic-pain surgical model (Bender_2009_pregabalin_rat_binary.R, Bender_2009_pregabalin_rat_smetab.R), and computer-controlled infusion, the target-controlled-infusion pump mode (Xie_2000_m3g_rat.R) – so an unprefixed column name would be ambiguous on sight. This makes SCORE_CCI differ from the bare-acronym APACHE_II / SAPS_II ICU-score canonicals, which face no such collision; the two conventions coexist in the register and the disambiguating prefix wins here. Distinct from APACHE_II and SAPS_II in substance as well as in name: those score ACUTE physiologic derangement over the first ICU day and predict short-term mortality, whereas the Charlson index scores CHRONIC comorbid disease and is fixed on admission; a single ICU popPK paper may report any combination of the three. Also distinct from ORG_FAIL_COUNT (a count of concurrently failing organ systems, an acute measure). Scope specific until a second paper ratifies the column; future models using the Charlson index with the same derivation should extend Example models rather than register a new canonical, and should document the Deyo/Quan variant and the reference value in covariateData[[SCORE_CCI]]$notes.

SCORE_ISAR (canonical for Identification of Seniors At Risk frailty screening score)

• Description: Identification of Seniors At Risk (ISAR) score: a six-item yes/no screening questionnaire for functional decline and adverse outcomes in older adults presenting to hospital (McCusker J, Bellavance F, Cardin S, et al. J Am Geriatr Soc 1999;47(10):1229-1237). One point per positive item (pre-morbid dependence, recent increase in dependence, recent hospitalisation, impaired vision, impaired memory, polypharmacy of three or more drugs), so the score is an integer 0-6; higher = frailer. Commonly dichotomised at 2/3 (0-2 fit, >= 3 frail).
• Units: (SCORE_ISAR units, 0-6 score)
• Type: continuous
• Scope: specific
• Reference category: n/a – used as an uncentred exponential exp(beta * SCORE_ISAR) in the founding model, so the printed typical value is the value at ISAR = 0. Check the source equation for centring before reuse.
• Source aliases:
  • ISAR – printed symbol in Al-Qurain 2022 (Table 1, Table 3 ‘ISAR effect on Q’); same quantity, no value transformation.
• Example models: AlQurain_2022_tramadol.R (ISAR raises the tramadol inter-compartmental clearance, Q = 42.6 * exp(0.255 * SCORE_ISAR) L/h, in older inpatients with ISAR 1-5).
• Notes: Named per the SCORE_<instrument> family (SCORE_CCI, SCORE_MMSE, …). A frailty SCREEN, not a frailty phenotype (Fried) or a deficit-accumulation frailty index; do not substitute either for it.

LACT (canonical for serum lactate concentration)

• Description: Serum (or plasma / arterial-blood) L-lactate concentration. Routine acid-base / tissue-perfusion marker in critically ill patients; rising values reflect tissue hypoperfusion, anaerobic glycolysis, hepatic dysfunction, or metabolic disturbance. Distinct from LDH (lactate dehydrogenase enzyme activity), which is a related but mechanistically different cell-turnover marker. Time-varying within an ICU stay (often updated several times per day) although source papers may use it as a per-day or per-subject summary – document per-model in covariateData[[LACT]]$notes.
• Units: mmol/L (SI; equivalent to mEq/L; commonly reported as mg/dL = mmol/L * 9.0)
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (LACT / ref)^exponent. Reference values observed: 1.91 mmol/L (Taubert 2016 patient-group-1 day-1 median; a critically-ill ICU cohort spanning 0.53-10.7 mmol/L).
• Source aliases:
  • Lactate – the spelled-out column header in Taubert 2016 Table 1 and supplementary Table S3; same biological quantity, no value transformation.
• Example models: Taubert_2016_linezolid.R (mmol/L; reference 1.91; negative power exponent -0.21 on CL; lower lactate is associated with higher clearance, consistent with reduced cardiac output / impaired tissue perfusion at higher lactate), Morse_2012_ghb_rbc_invitro.R (mmol/L; 0-150 mM L-lactate added to an erythrocyte uptake buffer as a competing monocarboxylate-transporter substrate, driving a sigmoidal Imax inhibition of GHB uptake with IC50 19.1 mM).
• Notes: Distinct from LDH (serum lactate dehydrogenase enzyme activity in U/L); a popular point of confusion because both encode “lactate” in the lab-marker abbreviation. Scope note added 2026-09-02: the column covers L-lactate as a competing transporter substrate deliberately added to an incubation medium, not only as a passively-observed critical-care perfusion marker, and it spans the in vitro buffer and the in vivo plasma matrices. Morse 2012 is why one column must do both – that paper’s conclusion is precisely the comparison of an in vitro IC50 of 19.1 mM against the 2-5 mM plasma L-lactate its rats actually reached, so splitting the buffer and plasma readings into separate canonicals would have made the paper’s own argument inexpressible. Note the very different operating ranges: added in vitro concentrations reach 150 mM, more than an order of magnitude above any physiologic value, so a model reusing this column should state which regime it was characterised in. Critically-ill PK models that include both should keep the two columns separate. Taubert 2016 reports that lactate correlated significantly with the cardiovascular SOFA sub-score (i.e., lactate acts here as a perfusion / cardiac-output proxy more than as a direct biochemical driver of linezolid elimination). General scope because serum lactate is a universally applicable critical-care covariate.

RACHS1 (canonical for Risk Adjustment for Congenital Heart Surgery 1 (RACHS-1) category)

• Description: Integer 1-6 RACHS-1 surgical-risk category, ascertained pre-operatively from the type of congenital heart defect and the planned operation. Higher categories indicate greater perioperative risk of in-hospital mortality (Jenkins 2002 Pediatrics). Time-fixed per subject.
• Units: (categorical; 1-6 integer)
• Type: categorical
• Scope: specific
• Reference category: Paper-specific. Oualha 2014 uses RACHS-1 = 2 as the low-risk reference stratum on SV*SVR_max (categories 1 are absent from the cohort; 3 and 4 are pooled as the higher-risk stratum). Hornik 2019 uses RACHS-1 < 4 as the reference and pools >= 4 as the higher-risk stratum.
• Source aliases:
  • RACHS-1 – the publication’s printed form with a hyphen, not a valid R identifier; renamed to RACHS1 when assembling input data.
• Example models: Oualha_2014_epinephrine.R (decomposed inside model() into a binary indicator rachs1_high <- (RACHS1 >= 3) that selects an additive log-shift on SV*SVR_max from 0.44 to 0.26 for the high-risk pool), Hornik_2019_methylprednisolone_il6.R (decomposed inside model() into rachs1_high <- (RACHS1 >= 4), which multiplies the IL-6 cardiopulmonary-bypass effect CPBE by 2.59; reference stratum RACHS-1 < 4, Hornik 2019 Table 3 and Data S2 RANKN).
• Notes: Specific scope because the variable is paediatric-cardiac-surgery-population-bound and the reference category depends on which RACHS-1 strata the cohort contains (Oualha 2014 has categories 2-4 only; a paper with categories 1-6 would need a different decomposition). Decompose inside model() into mutually exclusive binary indicators matching the source’s pooling (e.g., rachs1_high <- (RACHS1 >= 3)) and document the pooling rule in covariateData[["RACHS1"]]$notes.

CVP (canonical for central venous pressure)

• Description: Central venous pressure (CVP), measured in mmHg through a central venous catheter. Time-varying when monitored continuously; in cardiovascular Emax PD models it enters the mean-arterial-pressure equation (MAP = HR * SV*SVR + CVP) as an additive constant rather than a fitted covariate effect.
• Units: mmHg
• Type: continuous
• Scope: specific
• Reference category: n/a – used additively in the MAP equation. Cohort medians observed: 11 mmHg (Oualha 2014, range 8-15).
• Source aliases: none.
• Example models: Oualha_2014_epinephrine.R (enters Eq. 7 as the additive offset in MAP = HR * SV*SVR + CVP; the vignette defaults to the cohort median 11 mmHg when CVP is not supplied per subject).
• Notes: Specific scope because CVP is meaningful only for haemodynamic-PD models that resolve mean arterial pressure into its component cardiac-output and venous-return terms. The Oualha 2014 final model does not test CVP as a fitted covariate on PK or PD parameters; the column is used only as the structural offset in the MAP equation. Future haemodynamic models that fit a covariate effect on CVP itself can re-use the canonical name.

MORTRISK_HIGH (canonical for high-mortality-risk composite indicator)

• Description: 1 = subject meets the paper-defined “high mortality risk” composite criteria at study entry; 0 = subject is in the low or intermediate risk strata. Subject-level baseline indicator. The defining criteria are paper-specific: in Thuo 2011 (severe malnutrition cohort, following Berkley et al. Arch Dis Child 2003), a child is “high risk” if any one of the following is present: depressed conscious state, bradycardia (heart rate < 80 bpm), evidence of shock (capillary refill time >= 2 s, temperature gradient or weak pulse), or hypoglycaemia (blood glucose < 3 mmol/L); “intermediate risk” is any one of deep acidotic breathing, severe dehydration with diarrhoea, lethargy, hyponatraemia (Na < 125 mmol/L), or hypokalaemia (K < 2.5 mmol/L); “low risk” is none of the above. The canonical column collapses the intermediate and low strata into the 0 reference because the Thuo 2011 model only retains the high-vs-not-high contrast as a covariate on CL.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (low or intermediate mortality risk).
• Source aliases:
  • HIGHRISK / RISK (Thuo 2011 NONMEM notation; paper writes “high risk” in prose and structural equations as [1 + theta3 * (high risk)]) – used in Thuo_2011_ciprofloxacin.R.
• Example models: Thuo_2011_ciprofloxacin.R (multiplicative fractional effect on apparent CL: 1 + (-0.283) * MORTRISK_HIGH, i.e., a 28.3% reduction in apparent oral clearance for high-risk children; the standardised CL falls from 42.7 L/h/70 kg in low/intermediate-risk to 30.6 L/h/70 kg in high-risk; the paper attributes the contrast to delayed gastric emptying / impaired gut absorption in critically ill malnourished children).
• Notes: Specific scope because the underlying definition of “high mortality risk” is paper-defined (Thuo 2011 follows Berkley 2003’s three-stratum risk score for severely malnourished children). Distinct from SAPS_II (continuous adult-ICU severity score), RACHS1 (paediatric cardiac surgery risk category), and ORG_FAIL_COUNT (integer organ-failure count) – each is its own canonical with paper-specific scoring rules. Future paediatric-severe-malnutrition popPK papers that retain the Berkley 2003 three-stratum score (or a close variant) can reuse this canonical; per-model covariateData[[MORTRISK_HIGH]]$notes must document the exact criteria the source paper used.

MECH_VENT (canonical for invasive-mechanical-ventilation status indicator)

• Description: 1 = subject is receiving invasive mechanical ventilation during the modeled period; 0 = not. Treatment-status indicator (not a physiologic measurement) captured at study entry or as a time-fixed per-subject flag for the ICU admission. Captures the bulk fluid-distribution and hemodynamic shifts associated with positive-pressure ventilation – increased plasma renin activity, aldosterone, and antidiuretic hormone – which can change PK distribution volumes in critically ill patients.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (not mechanically ventilated).
• Source aliases:
  • mechanical ventilation – printed phrase in Georges 2009 Table 1 (no/yes count column); renamed to canonical MECH_VENT when assembling input data. Used in Georges_2009_ceftazidime.R.
• Example models: Georges_2009_ceftazidime.R (selector effect on V1 via the multiplicative log-effect form: vc = exp(lvc + e_mech_vent_vc * MECH_VENT + etalvc), with lvc = log(18.9) at MECH_VENT = 0 and e_mech_vent_vc = log(9.02 / 18.9) recovering TVV1 = 9.02 L at MECH_VENT = 1; Discussion attributes the V1 contraction to PPV-induced hormonal fluid shifts), Li_2019_ceftazidime.R, Li_2019_avibactam.R (proportional shifts of +0.297 and +0.175 on the respective central volumes; Li 2019 labels the covariate NPv and defines it as ‘the presence of a ventilator in the hospital room on the day of PK sampling’, i.e. time-fixed at the sampling day. Note this is NOT the HABP-vs-VABP contrast, which those models carry separately as DIS_HABP / DIS_VABP: a hospital-acquired-pneumonia patient who happened to be ventilated on the sampling day has MECH_VENT = 1 and DIS_VABP = 0), Patel_2022_imipenem.R, Patel_2022_relebactam.R (proportional shifts of +0.23 and +0.36 on the respective central volumes, source column VENT2, time-fixed at baseline; the effect is defined only WITHIN pneumonia – Patel 2022 Table 3 footnote g ‘The ventilation effect on participants with pneumonia receiving ventilation. The reference group was participants with pneumonia’ – so model() applies it to DIS_VABP + DIS_HABP * MECH_VENT, i.e. ventilated HABP plus every VABP patient, and never outside pneumonia), Irie_2021_favipiravir.R (TIME-VARYING, re-evaluated as ventilation status changes: IMV defined as WHO ordinal clinical status 7; multiplicative ratio 1.71^MECH_VENT on favipiravir CL/F in COVID-19, control-stream data item TUBE; the baseline-only IMV flag was not significant, the time-varying one was). Franzese_2022_ceftazidime.R / Franzese_2022_avibactam.R (the pediatric + adult refit of the Li 2019 pair; same ‘ventilator on the PK sampling day’ definition, control stream POP5 = 1; proportional shifts on Vc of +0.202, Table S2 theta22, and +0.267, Table S3 theta28).
• Notes: Specific scope because the indicator is ICU-population-bound and the mode-of-ventilation conventions vary across cohorts (Georges 2009 uses a binary IMV/no-IMV partition without distinguishing non-invasive ventilation, tracheostomy delivery, or weaning trial state). Distinct from RRT_HEMODIAL_STATUS / RRT_CRRT_STATUS (renal-replacement-therapy treatment-status flags) and from SAPS_II (continuous severity score). When a future critically-ill popPK paper reuses MECH_VENT with the same definition (binary, time-fixed at ICU admission), extend the Example models list rather than registering a new canonical; promotion to general scope is appropriate after a second corroborating paper. Per-model covariateData[[MECH_VENT]]$notes must document whether the indicator is time-fixed at admission or re-evaluated as ventilation status changes during the ICU stay.

ICU_ADM_POLYTRAUMA (canonical for ICU admission etiology indicator: polytrauma)

• Description: 1 = subject’s ICU admission etiology is polytrauma (severe multi-system blunt or penetrating trauma); 0 = otherwise. One of three mutually-exclusive ICU admission etiology indicators (ICU_ADM_POLYTRAUMA + ICU_ADM_POSTSURG + ICU_ADM_MEDICAL = 1 for every subject). Time-fixed per subject (the admission category is recorded once at ICU entry). Used as a categorical-stratum covariate on PK distribution parameters that differ between trauma-induced fluid shifts and elective surgical or medical patient profiles.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (other admission etiology – ICU_ADM_POSTSURG = 1 or ICU_ADM_MEDICAL = 1).
• Source aliases:
  • Admission for polytrauma – printed row label in Georges 2009 Table 1 (count column); renamed to canonical ICU_ADM_POLYTRAUMA when assembling input data. Used in Georges_2009_ceftazidime.R.
• Example models: Georges_2009_ceftazidime.R (selector effect on V2: TVV2 = theta6 = 57.1 L at ICU_ADM_POLYTRAUMA = 1; multiplicative log-effect form encodes the shift relative to the ICU_ADM_MEDICAL reference stratum as e_icu_adm_polytrauma_vp = log(57.1 / 13.6)).
• Notes: Specific scope because the polytrauma stratum definition is cohort-dependent (Georges 2009 partitions polytrauma vs postsurgical vs medical-reason at the time of ICU admission per Toulouse-cohort criteria; other ICU populations may use different boundaries – e.g., trauma + sepsis combination, isolated head injury, burns separated from polytrauma). The three ICU_ADM_* indicators must be mutually exclusive in the input data; the canonical mapping convention is that exactly one of them is 1 per subject. Distinct from MORTRISK_HIGH (severity score), SAPS_II (continuous severity), and ORG_FAIL_COUNT (organ-failure count) – all of which describe critical-illness severity rather than admission etiology. Future adult-ICU popPK papers that retain the same three-stratum etiology partition can extend the Example models list and the scope may be promoted to general after a second corroborating paper.

ICU_ADM_POSTSURG (canonical for ICU admission etiology indicator: postsurgical)

• Description: 1 = subject’s ICU admission etiology is postsurgical (elective or urgent surgical recovery, not polytrauma); 0 = otherwise. One of three mutually-exclusive ICU admission etiology indicators (see ICU_ADM_POLYTRAUMA for the partition definition). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (other admission etiology – ICU_ADM_POLYTRAUMA = 1 or ICU_ADM_MEDICAL = 1).
• Source aliases:
  • Admission for postsurgical – printed row label in Georges 2009 Table 1 (count column); renamed to canonical ICU_ADM_POSTSURG when assembling input data. Used in Georges_2009_ceftazidime.R.
• Example models: Georges_2009_ceftazidime.R (selector effect on V2: TVV2 = theta7 = 25.7 L at ICU_ADM_POSTSURG = 1; multiplicative log-effect form encodes the shift relative to the ICU_ADM_MEDICAL reference stratum as e_icu_adm_postsurg_vp = log(25.7 / 13.6)).
• Notes: See ICU_ADM_POLYTRAUMA for shared scope rationale and partition rules. Specific scope because the postsurgical stratum definition is cohort-dependent and the partition into the three Georges 2009 categories is not universal across ICU popPK papers (some cohorts merge postsurgical with medical or split it into elective vs urgent).

ICU_ADM_MEDICAL (canonical for ICU admission etiology indicator: medical)

• Description: 1 = subject’s ICU admission etiology is a medical reason (sepsis, pneumonia, cardiac, neurologic, or other non-surgical / non-polytrauma medical condition); 0 = otherwise. One of three mutually-exclusive ICU admission etiology indicators (see ICU_ADM_POLYTRAUMA for the partition definition). Time-fixed per subject. Commonly serves as the structural reference category in models that compare polytrauma / postsurgical strata against the medical baseline, because medical-reason patients in adult ICU cohorts tend to have distribution volumes closer to non-critically-ill individuals (per Georges 2009 Discussion).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (other admission etiology – ICU_ADM_POLYTRAUMA = 1 or ICU_ADM_POSTSURG = 1). When this stratum is chosen as the multiplicative-effect reference (as in Georges 2009), e_icu_adm_medical_<param> = 0 is implicit and only the two non-reference effects are estimated.
• Source aliases:
  • Admission for medical reason – printed row label in Georges 2009 Table 1 (count column); renamed to canonical ICU_ADM_MEDICAL when assembling input data. Used in Georges_2009_ceftazidime.R.
• Example models: Georges_2009_ceftazidime.R (selector effect on V2: TVV2 = theta8 = 13.6 L at ICU_ADM_MEDICAL = 1; this is the structural reference of the multiplicative parameterisation – exp(lvp) recovers 13.6 L when both other indicators are zero).
• Notes: See ICU_ADM_POLYTRAUMA for shared scope rationale. Specific scope. The Georges 2009 Discussion notes ‘Patients with a medical reason for admission presented a volume of distribution in the same order of magnitude as those of healthy subjects’ which motivates the choice of this stratum as the structural reference for V2 in the model file.

HIE_POST (canonical for post hypoxic-ischemic event indicator)

• Description: 1 = subject has undergone a hypoxic-ischemic (HI) event and is in the post-insult state (sustained cerebral injury and / or multi-organ stress that alters drug disposition); 0 = no HI event. Time-varying within a subject as the indicator flips from 0 to 1 at the documented HI insult time and remains 1 thereafter for the modeled observation window. Common in neonatal-encephalopathy / perinatal-asphyxia popPK studies (and potentially in adult cardiac-arrest / traumatic-arrest popPK studies) where the HI event causes persistent reductions in hepatic and renal drug clearance via CYP450 / UGT pathway impairment, decreased cardiac output, and multi-organ dysfunction. The exact definition of “post-HI state” (e.g., calculated from a continuous severity score thresholded to binary, or simply set to 1 from the documented insult time onward) is paper-specific; document per-model in covariateData[[HIE_POST]]$notes. Used as a binary state indicator on PK parameters (most commonly CL).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no HI event; pre-insult or never-insulted).
• Source aliases:
  • AED – used in Ezzati_2014_dexmedetomidine_piglet.R (the paper defines AED as “acute energy depletion”, computed by the time integral of beta-nucleotide-triphosphate (NTP) peak-area changes during hypoxia-ischemia; in the final model the AED severity score is used as a binary 0 / 1 state flag for whether the subject experienced the HI insult, NOT as a continuous covariate – 9 piglets had AED = 1 throughout dosing and 1 control piglet had AED = 0). NOTE: the paper’s AED column-name encodes a calculated NTP-integral severity score that the model dichotomises to a 0 / 1 state flag; downstream users assembling event tables should set HIE_POST = 1 from the HI insult time onward for all HI-exposed subjects.
• Example models: Ezzati_2014_dexmedetomidine_piglet.R (multiplicative factor on dexmedetomidine CL post-insult: CL = CLstd * (Wt/70)^0.75 * (1 + Ftemp * (TEMP - 37)) * (1 + (FAED - 1) * HIE_POST) with FAED = 0.558 – clearance is multiplied by 0.558 in the post-HI state, i.e. CL is reduced TO 55.8% of the pre-insult value, equivalent to a 44.2% reduction relative to the pre-insult typical value).
• Notes: General scope because hypoxic-ischemic injury is a recurring exposure across neonatal-encephalopathy, perinatal-asphyxia, and potentially adult cardiac-arrest popPK studies. Distinct from BURN_RECENT (acute hypermetabolic phase indicator with different physiology and a time-windowed threshold) and from DIS_* disease-state indicators (which describe chronic conditions rather than acute insult states). Distinct from ORG_FAIL_COUNT (integer organ-failure count) and MORTRISK_HIGH (mortality risk score) – HIE_POST is a binary state flag for a specific insult exposure, not a downstream severity or outcome score. Future papers that report a continuous HI severity score (e.g., AED integral, Sarnat staging, MRI biomarker) could either (a) re-use HIE_POST as a binary derived from the continuous score, or (b) register a sibling continuous canonical (e.g., HIE_SEVERITY) – the maintainers decide per paper.

Interferon / biomarker panels

BGENE21 (canonical for 21-gene type I interferon signature score)

• Description: Baseline 21-gene type I interferon signature score – a composite transcriptomic score summarising the expression of 21 interferon-regulated genes in whole blood relative to a healthy-donor reference, used as a biomarker of type I IFN pathway activation in SLE and related autoimmune conditions.
• Units: unitless fold-change score (relative to healthy-donor reference).
• Type: continuous
• Scope: specific
• Reference category: n/a – used with power scaling (BGENE21 / ref)^exponent. Reference values observed: 32 in Narwal 2013 (study-population median was 33), 12.04 in Zheng 2016 (median of the SLE phase IIb cohort, range 0.32-38.59).
• Source aliases: none.
• Example models: Narwal_2013_sifalimumab.R (reference 32, exponent 0.0558 on CL), Zheng_2016_sifalimumab.R (reference 12.04, power effect on CL with exponent 0.09).
• Notes: Specific to drugs whose mechanism targets the type I IFN pathway (e.g., anti-IFN-alpha antibodies like sifalimumab, anifrolumab). Higher BGENE21 indicates stronger target engagement / disease activity and is associated with increased drug clearance via target-mediated mechanisms. The 21-gene panel composition is tied to the MedImmune/AstraZeneca SLE development programme; a different IFN gene signature (e.g., a 4-gene or 5-gene panel) should be registered under its own canonical name (BGENE4, IFN_SIG,…) to avoid conflating panel definitions.

BGENE21_HIGH (canonical for binary high-vs-low IFN-21-gene indicator)

• Description: 1 = subject’s baseline 21-gene type I IFN signature score is at or above the paper-specified high/low cut-off, 0 = below cut-off.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (low).
• Source aliases: none.
• Example models: Almquist_2022_anifrolumab.R (binary high-IFN indicator on CL).
• Notes: Pair with continuous BGENE21 when the paper reports both. The high/low cut-off is paper-specific (commonly the population median) and must be documented in covariateData[[BGENE21_HIGH]]$notes for every model that uses this covariate. The maintainers chose BGENE21_HIGH (not IFNGS_HIGH) so the link to the existing BGENE21 register entry is explicit while the binary nature stays visible in the column name.

MicroRNA / gene-expression biomarkers

Single-analyte microRNA or messenger-RNA expression levels measured by qPCR or sequencing. Name an entry <ANALYTE>_<MATRIX> (analyte first, then the specimen it was measured in), because expression of the same microRNA in urine, plasma and tissue is not interchangeable. Composite multi-gene signature scores belong in the Interferon / biomarker panels section above (for example [[BGENE21]]), not here.

MIR155_URINE (canonical for urinary-pellet miR155-5p relative expression)

• Description: Relative expression of microRNA miR155-5p in the cell pellet of a first-morning urine sample, quantified by qPCR and expressed as 2^-dCq, where dCq is the target Cq minus the reference-control Cq of the same sample. Higher values mean higher expression. miR155 is an inflammatory microRNA expressed by activated T cells and macrophages; in kidney transplantation its urinary expression rises with T-cell-mediated (cellular) rejection.
• Units: (relative expression, 2^-dCq). Dimensionless. Document the reference control in covariateData[[MIR155_URINE]]$notes, because a different normaliser changes the scale.
• Type: continuous
• Scope: specific
• Reference category: n/a. Quintairos 2021 enters it linearly and untransformed on the logit, so no reference value applies and the intercept is the logit at zero expression.
• Source aliases:
  • M155 – Quintairos 2021 control stream and S3 Table dataset column.
  • miR155-5p – Quintairos 2021 Table 2 and Table 5. Table 2 labels the unit ‘dCt’, but the printed values are on the 2^-dCq scale that the Methods define.
• Example models: Quintairos_2021_kidneyTransplantRejection.R (founding example; time-varying, measured at week 1 and months 1, 2, 3 and 6 post-transplant; logit of acute rejection before the next visit = -5.89 + 3.51 * MIR155_URINE; observed range 0.01-2.80 and median 0.08 over the 183 modelled visits).
• Notes: The value spans more than two orders of magnitude and is strongly right-skewed (week-1 median 0.08 against a global mean of 0.39 in Quintairos 2021 Table 2), so a simulated cohort should draw it on the log scale or resample observed values. Store a value on the 2^-dCq scale. If a paper reports dCq itself (a log2-scale quantity where higher means LOWER expression), convert it with 2^(-dCq) rather than storing it here unchanged. Expression of the same microRNA in plasma or serum, or of a different microRNA, needs its own entry (MIR155_PLASMA, MIR142_URINE, …).

Inflammation markers

EOS (canonical for blood eosinophil count)

• Description: Blood eosinophil count (baseline or time-varying).
• Units: cells/uL
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (EOS / ref)^exponent.
• Source aliases:
  • BEOS (baseline EOS) – used in Kotani_2022_astegolimab.R.
• Example models: Kotani_2022_astegolimab.R (reference 180 cells/uL, baseline).
• Notes: Used as a surrogate of inflammatory burden that correlates with protein turnover and therefore mAb clearance. Canonical name drops the B prefix to match the SCORE_EASI / AGE / WT / ALB pattern; baseline-vs-time-varying status is documented in covariateData[[EOS]]$notes.

FENO (canonical for fractional exhaled nitric oxide concentration)

• Description: Fractional concentration of nitric oxide in exhaled breath (FeNO), measured by chemiluminescence or electrochemical analyser at a 50 mL/s expiratory flow rate per the ATS/ERS standard. A non-invasive marker of eosinophilic (type-2) airway inflammation: elevated FeNO reflects IL-4 / IL-13-driven upregulation of inducible nitric oxide synthase in airway epithelium. Standard baseline biomarker covariate in asthma exposure-response models, typically alongside blood eosinophil count [[EOS]].
• Units: ppb (parts per billion)
• Type: continuous
• Scope: specific
• Reference category: n/a – used with power scaling (FENO / ref)^exponent. Reference values observed: 25 ppb (Zhang 2025, median of the pooled Phase 2b + Phase 3 uncontrolled moderate-to-severe asthma cohort; range 3-387 ppb). 25 ppb is also the conventional clinical cut-point separating low from intermediate type-2 inflammation in adults.
• Source aliases:
  • BFENO (baseline FeNO) – used in Zhang_2025_dupilumab_fev1.R.
• Example models: Zhang_2025_dupilumab_fev1.R (power effect (FENO / 25)^0.682 on the dupilumab maximum treatment effect Emax for pre-bronchodilator FEV1; patients with higher baseline FeNO gain substantially more lung function – +154% Emax at the 95th percentile of 98 ppb, -54% at the 5th percentile of 8 ppb).
• Notes: Canonical name drops the B prefix of the common BFENO source column to match the EOS / IGE / ALB pattern; baseline-vs-time-varying status is documented in covariateData[[FENO]]$notes per model. Record the expiratory flow rate in the model notes if the source paper used a non-standard one (values are strongly flow-dependent, so a FeNO measured at 100 or 200 mL/s is not interchangeable with the 50 mL/s standard). Scope: specific until a second model ratifies the semantics; promote to general at that point.

BLBCELL (canonical for baseline CD19+ B cell count)

• Description: Baseline CD19+ B cell count (cells/uL) measured by fluorescence-activated cell sorting (FACS) prior to first dose. Used as a covariate / scaling biomarker for B-cell-targeted antibody PK-PD models (e.g., anti-CD20 mAbs in multiple sclerosis or B cell malignancies).
• Units: cells/uL
• Type: continuous
• Scope: specific
• Reference category: n/a – used with power scaling (BLBCELL / ref)^exponent. Reference value observed: 200 cells/uL (Yu 2022, median of the pooled five-study cohort).
• Source aliases:
  • Bcell0 – used in Yu_2022_ofatumumab.R.
  • BBCC (NHL Phase I/Ib/II convention; values in 10^6 cells/L = cells/uL) – used in Lu_2019_polatuzumab.R.
• Example models: Yu_2022_ofatumumab.R (power effect on the maximum B-cell-lysis stimulatory effect Emax, exponent 0.275, reference 200 cells/uL), Lu_2019_polatuzumab.R (two distinct effects: power on CL_INF with input floored at 1 cell/uL, and a thresholded power on CL_T with the BLBCELL/121-cells/uL ratio floored at 1), Lu_2017_polatuzumab_neuropathy.R (carries forward the Lu 2019 acMMAE PK-side effects via the inlined acMMAE popPK layer; not used directly by the Lu 2017 TTE PD layer), Dave_2019_venetoclax.R (not a covariate effect: the individual observed pre-dose CD19+ B-lymphocyte count, source column MIBTCD19 in 10^6 cells/L = cells/uL, IS the baseline of the Friberg-type B-lymphocyte PD model, circ0 = BLBCELL * exp(etalcirc0), and sets every PD initial condition).
• Notes: Distinct from a time-varying B cell count, which is the PD response variable rather than a covariate. Scope: specific because the clinically relevant baseline depends on the surface marker (CD19, CD20, CD22) and whether the panel reports total B cells or memory/naive subsets – register a new canonical name if a future paper uses a different marker. Both Yu 2022 (anti-CD20 ofatumumab) and Lu 2019 (anti-CD79b polatuzumab vedotin) use CD19+ counts, so the canonical is reused; subtype-specific differences are documented in each model’s covariateData[[BLBCELL]]$notes.

BL_PARP_PBL (canonical for baseline poly(ADP-ribose) polymerase activity in peripheral blood lymphocytes)

• Description: Subject-specific baseline (pre-dose) poly(ADP-ribose) polymerase (PARP) activity in peripheral blood lymphocytes (PBL), measured as picomoles of PAR polymer formed per 10^6 PBL by an enzyme-activity assay. Used as a covariate / scaling biomarker on the maximal-inhibition residual activity (Emin) parameter in PARP-inhibitor PK/PD models.
• Units: pmol/10^6 PBL (document per-model via covariateData[[BL_PARP_PBL]]$units).
• Type: continuous
• Scope: specific
• Reference category: n/a – used with power scaling (BL_PARP_PBL / ref)^exponent. Reference value observed: 90.8 pmol/10^6 PBL (Wang 2015 rucaparib; population typical baseline E0 used as a stand-in for the unreported study-cohort median because the paper reports only the typical E0 and the exponent value, not the numeric BLB_median).
• Source aliases:
  • BLB (Wang 2015’s notation for “baseline level in blood / PBL”) – used in Wang_2015_rucaparib.R.
• Example models: Wang_2015_rucaparib.R (power effect on residual maximum-inhibition parameter Emin; exponent 0.620 with the form Emin = TV(Emin) * (BL_PARP_PBL / BLB_median)^alpha; PBL paired with a separate tumor-tissue PARP activity covariate that is not yet a registered canonical because the units differ – pmol/mg protein for tumor vs pmol/10^6 PBL for blood).
• Notes: Specific scope because the column is meaningful only for PARP-inhibitor PK/PD models (rucaparib, olaparib, niraparib, talazoparib, veliparib, etc.) and the units are tied to the PBL-specific assay format. A future tumor-tissue PARP activity covariate would need a separate canonical because the units differ (pmol/mg protein) and the biology of cellular PARP activity per mg of protein is not numerically interchangeable with PBL-normalized PARP activity. The Wang 2015 model uses BL_PARP_PBL only on Emin (residual maximum inhibition) and not on E0 or IC50; per-paper effects must be documented in each model’s covariateData[[BL_PARP_PBL]]$notes. The paper does not publish the numeric study-cohort median of BLB used to center the covariate; the model file uses 90.8 pmol/10^6 PBL (the population typical baseline E0 reported in Wang 2015 Table 2) as a defensible default reference and documents the assumption in the vignette’s Assumptions and deviations section.

BL_TGN_RBC (canonical for baseline erythrocyte 6-thioguanine-nucleotide concentration)

• Description: Subject-specific baseline (first observed) concentration of 6-thioguanine nucleotides (TGN), the active metabolites of 6-mercaptopurine, inside red blood cells. Reported clinically as “E-TGN”. Used as the initial condition of the rbc_tgn compartment in thiopurine PK/PD models whose observation record begins during ongoing maintenance therapy, when the red-cell pool has already accumulated and cannot be initialised at zero.
• Units: umol/L. Document per-model via covariateData[[BL_TGN_RBC]]$units. Papers frequently report E-TGN as nmol per mmol hemoglobin; convert to umol/L using the assay’s assumed hemoglobin molecular weight and erythrocyte hemoglobin concentration and record the conversion in $notes (Gebhard 2023 uses MW 64458 g/mol and 330 g Hb/L erythrocytes).
• Type: continuous
• Scope: specific
• Reference category: n/a – supplied as an ODE initial condition (rbc_tgn(0) = BL_TGN_RBC), not as a scaling covariate. Cohort median 0.83 umol/L, range 0-7.6 umol/L (Gebhard 2023 Table 1).
• Source aliases:
  • INITGN – used in Gebhard_2023_mercaptopurine.R and Gebhard_2023_mercaptopurine_anc.R (Gebhard 2023: “X_E^6MP(0) = INITGN… with INITGN being the first observation in the data set at time point 0”).
• Example models: Gebhard_2023_mercaptopurine.R, Gebhard_2023_mercaptopurine_anc.R.
• Notes: Distinct from CONMED_MP, which is a concomitant-6-mercaptopurine-medication FLAG, not a measured red-cell concentration. Specific scope because the column is meaningful only for thiopurine models that carry an intracellular red-cell TGN state. A model that can start its estimation at a drug-free time point should initialise rbc_tgn(0) = 0 and not carry this covariate; Gebhard 2023’s own Discussion names that as the preferred remedy, but the published model as estimated requires the observed initial value.

BL_E2 (canonical for baseline serum oestradiol concentration)

• Description: Subject-specific observed baseline (pre-treatment) serum oestradiol (E2, 17-beta-oestradiol) concentration. Distinct from any model-ESTIMATED baseline: this is the measured value carried as a covariate, used in Pohl 2022 to scale the magnitude of the linzagolix drug effect on oestradiol (“the linzagolix drug effect was assumed to be related to observed baseline E2 by an estimated exponent”, Supporting Information section 3.2).
• Units: pg/mL. Document per-model via covariateData[[BL_E2]]$units; pmol/L is the other common clinical unit (1 pg/mL = 3.671 pmol/L for oestradiol, MW 272.38 g/mol).
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as a normalised power term (BL_E2 / 52)^theta. The 52 pg/mL normalisation constant of Pohl 2022 is printed only inside the Supporting Information equation, not in any table. Study medians: 53.0 pg/mL (EDELWEISS endometriosis patients), 44.0 (KLH1204 Japanese endometriosis patients), 25.0 and 20.5 (the two healthy-volunteer studies) – Pohl 2022 Table 3.
• Source aliases:
  • E2BASE – printed name in the Pohl 2022 Supporting Information section 3.2 core PK-E2 equation (E2_BASE).
• Example models: Pohl_2022_linzagolix_e2.R.
• Notes: Member of the existing BL_<biomarker> family (BL_PARP_PBL, BL_TGN_RBC, BL_MTX_RBC), following the same “observed baseline of a measured analyte, carried as a covariate” role. Specific scope because the column is meaningful only for models of ovarian-axis suppression (GnRH agonists and antagonists, aromatase inhibitors, combined oral contraceptives). Note the pharmacological subtlety that makes this column load-bearing rather than cosmetic: in a menses-anchored trial the pre-treatment sample is drawn at the NADIR of the menstrual oestradiol cycle, so BL_E2 systematically understates a subject’s cycle-average oestradiol – which is why Pohl 2022 needed a separate apparent-placebo-rise term (Table 5, “Placebo increase factor”) alongside this covariate. Do not confuse the two: BL_E2 scales the drug effect, the placebo term describes loss of cycle synchronisation over the trial.

BL_MTX_RBC (canonical for baseline erythrocyte methotrexate concentration)

• Description: Subject-specific baseline (first observed) concentration of methotrexate – predominantly as methotrexate polyglutamates – inside red blood cells. Reported clinically as “E-MTX”. Used as the initial condition of the rbc_mtx compartment in low-dose-methotrexate PK models whose observation record begins during ongoing maintenance therapy, when the red-cell pool has already accumulated and cannot be initialised at zero.
• Units: umol/L. Document per-model via covariateData[[BL_MTX_RBC]]$units. Papers frequently report E-MTX as nmol per mmol hemoglobin; convert to umol/L and record the conversion in $notes (Gebhard 2023 uses MW 64458 g/mol and 330 g Hb/L erythrocytes).
• Type: continuous
• Scope: specific
• Reference category: n/a – supplied as an ODE initial condition (rbc_mtx(0) = BL_MTX_RBC), not as a scaling covariate. Cohort median 0.026 umol/L, range 0-0.10 umol/L (Gebhard 2023 Table 1).
• Source aliases:
  • INIMTX – used in Gebhard_2023_methotrexate.R (Gebhard 2023: “X_E^MTX(0) = INIMTX with INIMTX being the first observation in the data set at time point 0”).
• Example models: Gebhard_2023_methotrexate.R.
• Notes: Distinct from CONMED_MTX, which is a concomitant-methotrexate-medication FLAG, not a measured red-cell concentration. Note that the red-cell methotrexate half-life is long (30-40 days in the literature Gebhard 2023 cites), so this initial condition remains influential over a much longer horizon than a typical plasma baseline.

BL_PRU (canonical for observed baseline platelet reactivity in P2Y12 reaction units)

• Description: Subject-specific observed baseline platelet reactivity measured with the VerifyNow P2Y12 assay, in P2Y12 reaction units (PRU), drawn before the first dose of a P2Y12 inhibitor. Carried as a time-fixed covariate that sets the individual PRU baseline of an antiplatelet PK-PD model in place of an estimated baseline.
• Units: PRU (device-standardised VerifyNow P2Y12 reaction units).
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as the individual baseline, rbase = exp(log(BL_PRU) + etalrbase), not as a scaling covariate. The cohort distribution is not reported by Kathman 2022; the PRU register entry quotes a normal baseline range of 180-376 PRU.
• Source aliases:
  • BPRU – Kathman 2022 control stream (Supplementary Material S1 Base=EXP(LOG(BPRU) + ETA(14))) and Table 3 (‘Base = Log(BPRU) baseline PRU’).
• Example models: Kathman_2022_bentracimab.R (ticagrelor + bentracimab PK-PD; baseline ‘fixed in model to observed baseline values’ with a 10% fixed IIV to absorb measurement error).
• Notes: Member of the existing BL_<biomarker> family (BL_E2, BL_PARP_PBL, BL_TGN_RBC, BL_MTX_RBC): the observed baseline of a measured analyte carried as a covariate. The BL_ prefix is load-bearing here rather than cosmetic, because the same models predict PRU as their observation output (PRU / pru in compartment-names.md); a bare PRU covariate column would collide with the endpoint. Distinct from a model-estimated baseline such as PRU(0) = kin / kout in Jung_2024_clopidogrel.R. Record the timing of the baseline draw in the per-model notes – a sample drawn on P2Y12-inhibitor maintenance therapy is not a baseline in this sense.

DOSE_MTX_MGM2 (canonical for administered methotrexate dose per body-surface area)

• Description: The methotrexate dose administered on the current dose record, normalised to body-surface area and expressed in mg/m^2. Per-dose-record covariate; constant within an inter-dose interval and updated when the prescriber alters the dose.
• Units: mg/m^2
• Type: continuous
• Scope: specific
• Reference category: n/a – enters a saturable dose-dependent bioavailability term F = 1 - Imax * DOSE_MTX_MGM2 / (D50 + DOSE_MTX_MGM2) (Gebhard 2023 Supplementary Table S1, Imax = 0.77, D50 = 15.01 mg/m^2, both fixed from the literature source Table S1 cites). Cohort median weekly dose 15.0 mg/m^2, range 1.3-45.0 mg/m^2 (Gebhard 2023 Table 1).
• Source aliases:
  • MTX – used in Gebhard_2023_methotrexate.R (the bare symbol Gebhard 2023 Supplementary Table S1 uses inside the bioavailability formula, annotated “MTX in mg/m2”).
• Example models: Gebhard_2023_methotrexate.R.
• Notes: Well-formed member of the auto-approved DOSE_<DRUG>_<UNITS> family (cf. DOSE_EMPA_MGD, DOSE_PHT_MGKGD). A dedicated column is required rather than the general DOSE canonical because Gebhard_2023_methotrexate.R supplies its amt in umol/m^2 (the unit the model’s concentration states demand) while the published bioavailability formula is calibrated in mg/m^2; carrying both in one DOSE column would silently mix units. Convert with the methotrexate molecular weight 454.44 g/mol: amt [umol/m^2] = DOSE_MTX_MGM2 [mg/m^2] * 1000 / 454.44.

DOSE_CARBOPLATIN_MGKG (canonical for administered carboplatin dose per kilogram of body weight)

• Description: The carboplatin dose administered on the current dose record, normalised to body weight and expressed in mg/kg. Per-dose-record covariate.
• Units: mg/kg
• Type: continuous
• Scope: specific
• Reference category: n/a – enters the dose arm of a sigmoidal Emax exposure-toxicity regression, Y = Emax * DOSE^n / (E50^n + DOSE^n) (Beguin 2024 Eq. 4 and Table 2, Emax = 0.76, E50 = 9.0 mg/kg, n = 17.80). Toxicity-cohort mean 11.5 +/- 2.4 mg/kg.
• Source aliases:
  • Dose (mg/kg) – Beguin 2024 Table 2 column heading. Same orientation and units, no value transformation.
• Example models: Beguin_2024_carboplatin_thrombocytopenia_dog.R.
• Notes: Well-formed member of the auto-approved DOSE_<DRUG>_<UNITS> family (cf. DOSE_MTX_MGM2, DOSE_EMPA_MGD), so no separate ratification was required. A dedicated column rather than the general DOSE canonical, and rather than a podo()-derived amount, for two reasons: the model’s amt is in ug per kg (the unit its ug/L concentration state demands) while the published regression is calibrated in mg/kg, so one shared column would silently mix units by a factor of 1000; and Beguin 2024 fitted the dose arm to the PRESCRIBED dose rather than to an amount reconstructed from the infusion record, so the covariate must remain evaluable when the PK layer is not solved. Convert with amt [ug/kg] = DOSE_CARBOPLATIN_MGKG [mg/kg] * 1000. Note that the paper prescribed on body-surface area, BSA(m2) = 0.1 * WT(kg)^(2/3) (Beguin 2024 Eq. 1), and reports the per-kg equivalent; the paper’s own conclusion is that the per-kg figure is the better toxicity predictor of the two.

CSF1 (canonical for colony-stimulating factor 1 / macrophage-colony-stimulating factor concentration)

• Description: Plasma colony-stimulating factor 1 (CSF-1, also known as macrophage colony-stimulating factor, M-CSF) concentration (baseline or time-varying). The hematopoietic cytokine that signals through CSF-1R to drive monocyte / macrophage differentiation and survival; used as both a target-engagement biomarker (anti-CSF-1R mAbs increase circulating free CSF-1) and a baseline covariate on PK / PD parameters in CSF-1R-pathway PopPK/PD models.
• Units: pg/mL (= ng/L; the two labels are numerically equivalent). Document per-model via covariateData[[CSF1]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – used with power scaling (CSF1 / ref)^exponent. Reference values observed: 549 pg/mL (Yang 2024 axatilimab; pooled-cohort median).
• Source aliases: none formally; informal aliases include BLCSF1 (baseline CSF-1) and BL_CSF1 (model-parameter notation in Monolix / NONMEM control streams).
• Example models: Yang_2024_axatilimab.R (baseline-only covariate on linear clearance CL with power exponent 0.912 and on the model parameter BL_CSF1 with power exponent 0.656; reference 549 pg/mL).
• Notes: Specific scope because the column is meaningful only for CSF-1R-pathway-targeting drugs (axatilimab and future anti-CSF-1R molecules). Distinct from any CSF-1 model state representing time-course dynamics – covariate column is the pre-dose laboratory observation, typically measured by an ELISA assay (Yang 2024 used the R&D Systems Quantikine ELISA). Per-model covariateData[[CSF1]]$notes should document the assay used and any LOQ-related imputation for samples below the assay’s limit of detection.

CRP (canonical for C-reactive protein)

• Description: C-reactive protein concentration. Covers both standard and high-sensitivity (hs-CRP) assays and both baseline and time-varying usages. Each model’s covariateData[[CRP]]$description and notes must state the assay type (standard vs hs-CRP) and whether the column carries a baseline-only or time-varying value, including the paper-specific reference value used for power scaling.
• Units: mg/L (document per-model via covariateData[[CRP]]$units).
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (CRP / ref)^exponent or exponential effects exp(coef * (CRP - ref)). Reference values observed: 4.23 mg/L (Moein 2022, IBD standard assay), 4.31 mg/L (Moein 2022 Table 3 median), 5.21 mg/L (Thakre 2022, baseline hs-CRP), 7.41 mg/L (Chua 2025, baseline standard assay), 14.2 mg/L (Xu 2019, baseline standard assay), 15.7 mg/L (Ma 2020, baseline standard assay), 0.837 mg/dL = 8.37 mg/L (Wang 2020, IBD standard assay; the model carries the source unit mg/dL).
• Source aliases:
  • hsCRP – high-sensitivity CRP (mixed-case preserved from earlier register drafts).
  • HSCRP – all-caps variant.
  • CRPHS – used in Thakre_2022_risankizumab.R (baseline, high-sensitivity assay).
  • BLCRP – baseline CRP; used in Xu_2019_sarilumab.R and Ma_2020_sarilumab_das28crp.R.
  • BCRP – baseline CRP as named in Tsuchiwata 2026; used in Tsuchiwata_2026_tofacitinib.R. Do not confuse this source abbreviation with the BCRP efflux transporter (ABCG2).
• Example models: Thakre_2022_risankizumab.R, Xu_2019_sarilumab.R, Chua_2025_mirikizumab.R, Moein_2022_etrolizumab.R, Ma_2020_sarilumab_das28crp.R, Wang_2020_ontamalimab.R (mg/dL, reference 0.837), Tsuchiwata_2026_tofacitinib.R (mg/dL, baseline, standard assay, reference 0.851 mg/dL = 8.51 mg/L; enters CL/F as (CRP / 0.851)^-0.0185), Yu_2022_crohns_cdai150_mbma.R, Yu_2022_crohns_cdai100_mbma.R, Yu_2022_crohns_cdai70_mbma.R, Yu_2022_crohns_dcrp_mbma.R (mg/dL, study-arm mean baseline standard-assay CRP in a Crohn’s disease MBMA; power form on every drug effect with references 1.688, 1.742, 1.657 and 1.786, the arm means of each endpoint’s analysis dataset. The source Methods state mg/L but its values are on the mg/dL scale).
• Notes: The prior separate hsCRP, BLCRP, and standard-assay CRP canonicals were merged on 2026-04-20 to a single general-scope CRP canonical. Assay type (standard vs hs-CRP), baseline-vs-time-varying status, and the paper-specific reference value all live in each model’s covariateData[[CRP]]$description / notes. Only aggregate values from hs-validated assays as CRP when the downstream analysis relies on low-range sensitivity; for most inflammatory-disease cohorts (IBD, RA/PsA), baseline CRP is well above the hs-sensitivity range and the distinction is moot.

AAG (canonical for alpha-1 acid glycoprotein concentration)

• Description: Serum alpha-1 acid glycoprotein (AAG; orosomucoid; ORM1) concentration, an acute-phase plasma glycoprotein that binds basic and lipophilic drugs (including taxanes such as docetaxel and paclitaxel). Elevated in cancer, inflammation, and infection; influences free-drug fraction and downstream PD effects in cytotoxic-chemotherapy myelosuppression models.
• Units: g/L (= mg/mL; 1 g/L is the conventional clinical-PK reporting unit). Document per-model via covariateData[[AAG]]$units if a different unit is used.
• Type: continuous
• Scope: general
• Reference category: n/a – used in piecewise-linear, power, or exponential effect forms (e.g., the Kloft 2006 cytotoxic-chemotherapy myelosuppression family fits a piecewise-linear effect with breakpoint at the cohort median 1.34 g/L: separate slopes apply for AAG <= 1.34 and AAG > 1.34). Reference values observed: 1.34 g/L (Kloft 2006 / Netterberg 2017, cohort median in mixed adult-cancer cohort).
• Source aliases:
  • AAG – used in Netterberg_2017_docetaxel.R (per the bundle’s NM-TRAN $INPUT block; matching Kloft 2006).
  • AAGI2 – LOCF-imputed time-varying AAG series; NONMEM control-stream column name used in Said_2025_imatinib.R (Said 2025 Data S1 $INPUT, consumed inside $DES by the closed-form binding solution).
  • AGP1 – used in Ozawa_2007_docetaxel.R (Appendix I $INPUT block); reported in mg/dL with conversion to canonical g/L via AAG_g_per_L = AGP_mg_per_dL / 100.
• Example models: Netterberg_2017_docetaxel.R (piecewise-linear effects on baseline ANC with separate low-AAG and high-AAG slopes around median 1.34 g/L; linear effect on the drug-effect slope SL via (1 + theta * (AAG - 1.34))), Ozawa_2007_docetaxel.R (multiplicative power-form effect on the linear drug-effect slope: SLOPE = theta_SLOPE * (AAG / 0.94)^e_aag_slope with e_aag_slope = -1.38; reference value 0.94 g/L from the published NONMEM control stream AGPm = 94 mg/dL), Said_2025_imatinib.R (STRUCTURAL rather than covariate-effect use: AAG has no reference value and no e_aag_* coefficient, and instead sets the binding capacity bmax = 11700 * AAG inside the closed-form saturable-binding solution that yields unbound imatinib from total imatinib, so the unbound fraction that drives elimination, metabolite formation, and distribution falls as AAG rises; time-varying, LOCF-imputed, median 4 measurements per patient).
• Notes: General scope because serum AAG is a routine clinical-laboratory measurement that recurs across cytotoxic-chemotherapy population-PK / PD analyses (Bruno 1996/1998 docetaxel popPK uses AAG as a CL covariate; Kloft 2006 and downstream Friberg-family myelosuppression models use it on baseline ANC and drug-effect slope; Ozawa 2007 uses a power-form effect on the drug-effect slope only). Time-fixed at baseline unless the source paper states otherwise – Said_2025_imatinib.R is the counter-example, carrying a time-varying LOCF-imputed series. Two structurally different uses of this column now coexist and must not be conflated: (a) a covariate effect on a typical value, which needs a reference / breakpoint value (Kloft 2006, Netterberg 2017, Ozawa 2007); and (b) a structural binding-capacity input with no reference value, where AAG multiplies a molar scale factor to give Bmax inside a saturable-binding equation (Said 2025). Papers of type (b) should record no reference value and no e_aag_* coefficient. The Kloft 2006 piecewise-linear breakpoint at 1.34 g/L corresponds to the population median in their pooled cancer cohort; the Ozawa 2007 normalisation reference 0.94 g/L is the cohort median used inside that paper’s NONMEM control stream. Future papers may use different breakpoints / reference values, so document the per-model reference in covariateData[[AAG]]$notes. Distinct from CRP (a different acute-phase reactant with different binding properties).

AAG_MISSING (canonical for alpha-1-acid-glycoprotein-result-missing indicator)

• Description: 1 = baseline alpha-1-acid glycoprotein was not measured for this subject and a population value was imputed; 0 = AAG measured. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (AAG measured). Pairs with AAG, which carries the measured-or-imputed value on the same record.
• Source aliases:
  • A1AGLP_MISS – used in Yin_2024_soticlestat.R (NONMEM $INPUT block).
• Example models: Yin_2024_soticlestat.R (does not shift the typical value; it inflates the between-subject SD of relative bioavailability by (1 + 0.42 * AAG_MISSING), so imputed subjects carry extra unexplained variability).
• Notes: Member of the established <COV>_MISSING missing-indicator family (ADA_MISSING, CYP2C9_MISSING, SNP_CYP2C19_RS3814637_MISSING, SNP_CYP3A4_RS2242480_MISSING): a binary flag that distinguishes “value imputed because unmeasured” from a measured value, so the imputed subjects are not silently treated as exchangeable with measured ones. General scope because whole-trial AAG imputation recurs whenever a pooled analysis spans trials that did not all collect the acute-phase protein panel (Yin 2024: four of eight trials). Note the two distinct modelling roles this family takes – a shift of the typical value (ADA_MISSING on clearance in Suri_2018_brentuximab.R) versus a shift of the between-subject VARIANCE only (AAG_MISSING here); document which applies in covariateData[[AAG_MISSING]]$notes.

PROCALCITONIN (canonical for serum procalcitonin concentration)

• Description: Serum (or plasma) procalcitonin concentration, the 116-amino-acid calcitonin precursor released from parenchymal tissue under bacterial-inflammatory stimulation. The most widely used clinical discriminator of bacterial sepsis from non-infectious inflammation, and a graded severity marker in critical illness. In population-PK analyses of hepatically-metabolized drugs it acts as an inflammation covariate: the acute-phase response downregulates cytochrome-P450 expression, so a rising procalcitonin predicts falling metabolic clearance. Both baseline and time-varying usages are covered; document per-model in covariateData[[PROCALCITONIN]]$notes which applies.
• Units: ug/L (= ng/mL; the two labels are numerically equivalent). Document per-model via covariateData[[PROCALCITONIN]]$units.
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (PROCALCITONIN / ref)^exponent or exponential effects exp(coef * (PROCALCITONIN - ref)). Reference values observed: 0.19 ug/L (Liu 2025, cohort median in an elderly Chinese voriconazole TDM cohort).
• Source aliases:
  • PCT – the near-universal clinical and publication abbreviation for procalcitonin; used in Liu_2025_voriconazole.R (Table 1 / Table 2 and the published final-model equation) and screened but not retained in Peng_2025_meropenem.R. The canonical deliberately spells the analyte out rather than adopting this abbreviation, because a bare PCT column collides with the established _PCT percent suffix carried by BODYFAT_PCT, RACE_ASIAN_PCT, TUMTP_SQUAM_PCT, PS_ECOG_0_PCT, BMBLAST_PCT, CUM_FLUID_BAL_PCT and ADHERENCE_PLACEBO_PCT, and with the unrelated PCT = proximal convoluted tubule used in the renal QSP model Lu_2014_sglt_qsp.R. Map the source column PCT to this canonical and record source_name = "PCT".
  • PROCT, ProCT, PCT_BASE, BLPCT – procalcitonin variants seen in publication tables and NONMEM $INPUT columns (the canonical drops the BL prefix per the IL6 / IL8 / CRP convention, with baseline-vs-time-varying status documented in per-model notes).
• Example models: Liu_2025_voriconazole.R (median-normalized power effect on apparent voriconazole clearance, (PROCALCITONIN / 0.19)^-0.209, so a septic rise in procalcitonin predicts slower clearance and higher voriconazole exposure; ranked fourth by SHAP importance in that paper’s downstream machine-learning ensemble), Peng_2025_meropenem.R (screened as a clearance covariate and not retained; documentation-only entry in covariatesDataExcluded).
• Notes: Registered on the same terms as the established general-scope inflammation-marker canonicals CRP, IL6, IL8 and AAG, and is not an alias of any of them. The name is spelled out in full rather than abbreviated to PCT for the collision reasons given under Source aliases; the matching covariate-effect parameter is therefore e_procalcitonin_<param> (e.g. e_procalcitonin_cl), not e_pct_<param>. Procalcitonin is induced far more selectively by bacterial infection than C-reactive protein is (CRP rises in viral infection, autoimmune flare and tissue injury alike), rises and falls faster than CRP, and is measured on a separate immunoassay – papers routinely fit the two as competing covariates and retain only one, so conflating them would silently merge distinguishable effects. Liu 2025 is a case in point: it reports procalcitonin as a retained clearance covariate while CRP was excluded from the analysis entirely for a >50% missing rate. Unit vigilance is warranted: the clinically meaningful range spans roughly three orders of magnitude (<0.05 ug/L healthy, ~0.5 ug/L bacterial-infection threshold, >10 ug/L septic shock), and ug/L and ng/mL are numerically identical while ng/L is 1000-fold different – an unnoticed ng/L table silently rescales every prediction through the power term. Right-skewed in unselected inpatient cohorts (Liu 2025: median 0.19, IQR 0.13-0.28 ug/L), so median normalization rather than mean centring is the usual encoding. Distinct from WBC and NEUT, which count the responding cell populations rather than measuring the inflammatory mediator.

IL6 (canonical for serum interleukin-6 concentration)

• Description: Serum (or plasma) interleukin-6 (IL-6) concentration. Pro-inflammatory cytokine; elevated in rheumatoid arthritis, Castleman’s disease, sepsis, COVID-19, and other inflammatory conditions. Both baseline and time-varying usages are covered; document per-model in covariateData[[IL6]]$notes whether the column is baseline-only or time-varying.
• Units: pg/mL (= ng/L; the two labels are numerically equivalent).
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (IL6 / ref)^exponent or with the log-transformed form (log(IL6 * 1000) / log(ref * 1000))^exponent that some legacy NONMEM analyses adopt. Reference values observed: 31 pg/mL (Cojutti 2020 darunavir, SARS-CoV-2 cohort median; plain power (IL6 / 31)^-0.23 on CL/F), 20 pg/mL (Frey 2013 baseline; the formula (log(IL-6 * 1000)/9.9)^exponent is the algebraic equivalent of (log(IL-6) / log(20))^exponent after the constant-factor rescaling that ties the reference log to 9.9 = log(20000)).
• Source aliases:
  • BLIL6, bIL6, IL6_BASE – baseline IL-6 (used in some NONMEM control streams; canonical drops the BL prefix per the EOS / SCORE_EASI / AGE convention with baseline-vs-time-varying status documented in per-model notes).
  • IL-6, IL_6 – punctuation variants seen in publication tables and figures.
• Example models: Cojutti_2020_darunavir_covid19.R (IL-6 on the TDM day, reference 31 pg/mL; power effect on darunavir CL/F), Frey_2013_tocilizumab.R (baseline IL-6, reference 20 pg/mL; log-transformed power effects on EC50, BASE, and the DMARD background-effect parameter).
• Notes: Anti-IL-6 / anti-IL-6R drugs (tocilizumab, sarilumab, siltuximab) often see large transient increases in measured IL-6 after dosing – when the column is time-varying for these drugs, the assay typically detects total (free + drug-bound) IL-6 because the antibody complex slows clearance of IL-6. Document the assay type (free vs. total IL-6) and whether the value is pre-dose (baseline-only) or post-dose / time-varying in each model’s covariateData[[IL6]]$notes. Frey 2013’s formula relies on the relative log-IL-6 ratio rather than the linear concentration, so the column units must be pg/mL exactly (not ng/mL) for the published exponents to apply unchanged.

IL8 (canonical for serum interleukin-8 concentration)

• Description: Serum (or plasma) interleukin-8 (IL-8, systematic name CXCL8) concentration. Prototypical neutrophil-recruiting chemokine produced by macrophages, endothelium and epithelium; rises within hours of tissue injury and is one of the earliest and largest cytokine responses in severe traumatic injury, sepsis, ARDS and burn. Both baseline and time-varying usages are covered; document per-model in covariateData[[IL8]]$notes whether the column is baseline-only or time-varying.
• Units: pg/mL (= ng/L; the two labels are numerically equivalent).
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling. Stitt 2026 uses the un-normalized form IL8^exponent (i.e. an implicit reference of 1 pg/mL) rather than the usual (IL8 / ref)^exponent; the cohort median admission value in that paper is 20.3 pg/mL but does NOT appear as a divisor in the published equation. A future model using a normalized form should record its own reference in covariateData[[IL8]]$notes.
• Source aliases:
  • IL-8, IL_8 – punctuation variants seen in publication tables and figures; used as IL-8 in Stitt 2026 Table 1 / Table 2 and Equation 1.
  • CXCL8 – the systematic chemokine nomenclature; the same analyte, no value transformation.
  • BLIL8, BIL8, IL8_BASE – baseline IL-8 in NONMEM $INPUT columns (the canonical drops the B / BL prefix per the IL6 / IL22 / CRP convention, with baseline-vs-time-varying status documented in per-model notes).
• Example models: Stitt_2026_tranexamicAcid.R (median admission serum IL-8 in the TAMPITI adult severe-trauma cohort, 20.3 pg/mL; un-normalized power effect IL8^-0.0873 on TXA clearance, so higher IL-8 predicts slower clearance).
• Notes: Direct sibling of the registered IL6 and IL22 serum-cytokine canonicals and registered on the same terms, but a biologically distinct mediator and NOT an alias of either: IL-8 is a CXC chemokine whose primary action is neutrophil chemotaxis and activation, whereas IL-6 is a pleiotropic pro-inflammatory / acute-phase inducer and IL-22 an epithelial-effector cytokine downstream of IL-23; all three are measured on separate assays and can move independently. The un-normalized encoding in the founding example makes the column units load-bearing in an unusually strict way: because IL8^exponent has no divisor to cancel a scale change, recording values in ng/L-labelled-as-something-else, or in ng/mL, silently rescales every prediction by 1000^exponent. Values against this column must be pg/mL. Any model reusing IL8 should state explicitly in its notes whether the paper’s form is normalized or un-normalized. Distinct from NLR (neutrophil-to-lymphocyte ratio) and NEUT (absolute neutrophil count), which measure the downstream cell population that IL-8 recruits rather than the chemokine itself; the three frequently co-vary in trauma and critical-illness cohorts but are separate measurements.

IL22 (canonical for serum interleukin-22 concentration)

• Description: Serum (or plasma) interleukin-22 (IL-22) concentration. Effector cytokine produced downstream of IL-23 receptor signalling by Th17 / Th22 and innate lymphoid cells; elevated in Crohn’s disease and ulcerative colitis and correlated with disease activity. Both baseline and time-varying usages are covered; document per-model in covariateData[[IL22]]$notes whether the column is baseline-only or time-varying.
• Units: pg/mL (= ng/L; the two labels are numerically equivalent).
• Type: continuous
• Scope: general
• Reference category: n/a – used as the driver variable of a sigmoid Imax / Emax term rather than as a multiplicative covariate. Reference (half-maximal) values observed: 22.8 pg/mL (Zhang 2023 brazikumab IB50 on the CDAI production-rate inhibition).
• Source aliases:
  • BIL22, BLIL22, IL22_BASE – baseline IL-22 (used in NONMEM $INPUT columns; the canonical drops the B / BL prefix per the IL6 / CRP / SCORE_EASI convention, with baseline-vs-time-varying status documented in per-model notes).
  • IL-22, IL_22 – punctuation variants seen in publication tables and figures.
• Example models: Zhang_2023_brazikumab_il22.R (baseline serum IL-22 as the per-subject predictive-biomarker driver of the brazikumab drug effect on the CDAI input rate: idrug = imax * IL22^hill / (ec50^hill + IL22^hill) with imax = 0.297, ec50 (the paper’s IB50) = 22.8 pg/mL, hill (the paper’s gamma) fixed at 20; the effect is gated by ON_TREATMENT because the drug term is zero in the placebo arm).
• Notes: Sibling of IL6, but a biologically distinct cytokine and NOT an alias of it: IL-22 sits downstream of IL-23 whereas IL-6 is an independent pro-inflammatory mediator, and the two are measured on separate assays. Anti-IL-23 antibodies (brazikumab, risankizumab, guselkumab, mirikizumab) block IL-23 signalling and thereby suppress downstream IL-22 production, so for those drugs a post-dose IL-22 column is pharmacodynamically driven rather than a stable baseline covariate – record the assay (Zhang 2023 used the R&D Systems Quantikine ELISA, quantifiable range 10-800 pg/mL in 100% serum) and the baseline-vs-time-varying status in each consuming model’s covariateData[[IL22]]$notes. Distinct from CRP, which Zhang 2023 fits as a parallel alternative predictive biomarker in a separate model file.

IFNG (canonical for total serum interferon-gamma concentration)

• Description: Total (free plus therapeutic-antibody-bound) serum or plasma interferon-gamma (IFN-gamma) concentration. Type II interferon produced by activated T cells and NK cells; the driver cytokine of the hyperinflammatory feedback loop in haemophagocytic lymphohistiocytosis (HLH) and macrophage activation syndrome (MAS), and the pharmacological target of anti-IFN-gamma antibodies. Both baseline and time-varying usages are covered; document per-model in covariateData[[IFNG]]$notes which applies, and whether the assay measures free or total (drug-bound plus free) IFN-gamma.
• Units: pg/mL (= ng/L; the two labels are numerically equivalent).
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (IFNG / ref)^exponent. Reference values observed: 1e6 pg/mL (Brossard 2025 / Brossard & Laveille 2024, the normalisation point at which the target-mediated clearance component CLNL is tabulated).
• Source aliases:
  • IFNg, IFN-gamma, IFNgamma, TIFNG – punctuation and spelling variants seen in publication tables; TIFN denotes total IFN-gamma explicitly in the Jacqmin 2022 primary-HLH control stream.
• Example models: Brossard_2025_emapalumab.R (time-varying total serum IFN-gamma as the driver of the non-linear, target-mediated clearance component: clnl = exp(lclnl) * (IFNG / 1e6)^0.542, normalised to 1e6 pg/mL; the branch is switched off entirely in MAS patients by DIS_MAS).
• Notes: Direct sibling of the existing IL6 and IL22 serum-cytokine canonicals and registered on the same terms, but a biologically distinct mediator and NOT an alias of either: IFN-gamma is the sole type II interferon, signals through IFNGR1/IFNGR2-JAK1/JAK2-STAT1, and is measured on its own assay. As with IL6 under anti-IL-6R therapy, an anti-IFN-gamma antibody (emapalumab) causes measured total IFN-gamma to rise sharply after the first dose because the antibody-cytokine complex clears more slowly than free cytokine – so a post-dose IFN-gamma column for these drugs is a drug-perturbed quantity, not a stable baseline covariate, and the free/total distinction is load-bearing. Above roughly 10,000 pg/mL, total IFN-gamma drives target-mediated disposition of emapalumab; below it, emapalumab clearance is effectively linear. Distinct from BGENE21 / BGENE21_HIGH, which are type I interferon transcriptomic signature scores rather than a measured type II cytokine concentration, and from CONMED_IFNALPHA / CONMED_IFNB1A, which flag coadministration of exogenous interferon products.

Cardiometabolic / target biomarkers

HDLC (canonical for high-density lipoprotein cholesterol)

• Description: Serum high-density lipoprotein cholesterol concentration (baseline or time-varying).
• Units: mg/dL or mmol/L – document the unit used in each model via covariateData[[HDLC]]$units (1 mmol/L ~= 38.67 mg/dL for cholesterol).
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (HDLC / ref)^exponent. Reference value observed: 54 mg/dL (Frey 2010 pooled-cohort median).
• Source aliases:
  • HDL-C – Frey 2010 spelling with hyphen.
  • HDL_C – common alternative spelling.
• Example models: Frey_2010_tocilizumab.R (mg/dL, reference 54; small negative exponent -0.2 on linear CL; the paper interprets the effect as a body-size surrogate rather than a mechanism), Cazaubon_2019_mitotane.R (g/L – 1 g/L = 100 mg/dL; reference 0.65 g/L cohort median; power exponent -0.344 on CL; per-subject median over the observation window; attributed to lipoprotein sequestration of the lipophilic drug).
• Notes: Cardiometabolic lipid-panel covariate. In Frey 2010 it correlates with body size (HDL-C is lower in larger patients) and the small CL effect (-14% to +15% across the observed 23-135 mg/dL range) was retained but not interpreted as mechanistic.

TCHOL (canonical for total serum cholesterol)

• Description: Serum (or plasma) total cholesterol concentration (baseline or time-varying). Sum of HDL, LDL, VLDL, and other lipoprotein-associated cholesterol fractions; distinct from HDLC which captures only the HDL fraction.
• Units: mmol/L or mg/dL – document the unit used in each model via covariateData[[TCHOL]]$units (1 mmol/L ~= 38.67 mg/dL for cholesterol).
• Type: continuous
• Scope: general
• Reference category: n/a – used with linear-deviation forms 1 + theta * (TCHOL - ref) or power scaling (TCHOL / ref)^exponent. Reference value observed: 3 mmol/L (Archary 2018, severely malnourished pediatric LPV cohort, baseline mean 2.7-2.9 mmol/L).
• Source aliases:
  • CHOL – Archary 2018 NONMEM column abbreviation; the universal short form.
  • TC – alternative abbreviation common in lipid-panel literature.
• Example models: Archary_2018_lopinavir.R (mmol/L, reference 3; linear effect on apparent CL/F: 1 + 0.207 * (TCHOL - 3); serves as a surrogate for nutritional / hepatic-function recovery rather than a mechanistic effect).
• Notes: In severe acute malnutrition, total cholesterol tracks lipid-pool repletion and hepatic recovery and may serve as a surrogate covariate when the actual driver of bioavailability or clearance variability is not directly measured. Lopinavir is highly protein-bound (to albumin and AAG) and lipophilic, so circulating cholesterol can correlate with binding capacity for lipophilic drugs. Distinct from HDLC (HDL fraction only) and from CRP / ALB / AAG (separate canonicals for related malnutrition / inflammation markers).

APOC3 (canonical for serum apolipoprotein C-III concentration)

• Description: Serum (or plasma) apolipoprotein C-III concentration (baseline or time-varying). ApoC-III inhibits both lipoprotein-lipase-dependent and LPL-independent triglyceride clearance and is the direct pharmacological target of the APOC3-silencing therapeutic class, so this covariate appears both as a baseline lab value and as the measured PD biomarker in exposure-response models for that class.
• Units: mg/dL – document the unit used in each model via covariateData[[APOC3]]$units. Reported in mg/dL in the FCS / severe-hypertriglyceridemia literature; ug/mL is numerically 10x the mg/dL value should a source use it.
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (APOC3 / ref)^exponent. Reference value observed: 34 mg/dL (Ousey 2026 Table 2 centring for the FCS cohort).
• Source aliases:
  • APOC3 – Ousey 2026 spelling (and the HGNC gene symbol).
  • apoC-III, ApoC3, APO C-III – common prose spellings of the protein.
  • Baseline APOC3 – Ousey 2026 Table 2 covariate row name.
• Example models: Ousey_2026_plozasiran.R (mg/dL, reference 34; power exponent 0.888 on the estimated APOC3 baseline of a cascading K-PD indirect-response model, where serum APOC3 is simultaneously the covariate, an ODE state, and an observed endpoint).
• Notes: Cardiometabolic lipid-panel covariate, alongside TRIG, HDLC, LDLC and TCHOL. Distinct from APOE4_COUNT, which is an APOE-epsilon4 allele count (a genotype, unrelated to this apolipoprotein). Expected to recur across the APOC3-targeting class (plozasiran, olezarsen, volanesorsen), whose papers use the same baseline lab value; assay quantification limits matter because FCS baselines can sit above the ULOQ (Ousey 2026 used LLOQ 0.94 and ULOQ 88 mg/dL, imputing BLQ to LLOQ/2 and above-ULOQ to ULOQ), so per-model imputation handling belongs in covariateData[[APOC3]]$notes.

TRIG (canonical for serum triglyceride concentration)

• Description: Serum triglyceride concentration (baseline or time-varying).
• Units: mmol/L or mg/dL – document the unit used in each model via covariateData[[TRIG]]$units (1 mmol/L ~= 88.5 mg/dL for triglyceride).
• Type: continuous
• Scope: general
• Reference category: n/a – used with linear-deviation form (1 + theta * (TRIG - ref)) or power scaling (TRIG / ref)^exponent. Reference value observed: 5.3 mmol/L (Archary 2019 lamivudine equation centring – the equation centring in the source is reported as the cohort average; cohort median in Archary 2019 Table 1 is 2.2-2.3 mmol/L, see model-file Errata).
• Source aliases: none known.
• Example models: Archary_2019_lamivudine.R (mmol/L, reference 5.3; linear-deviation effect on Vc/F with coefficient -0.13 per mmol/L deviation from the reference; lower triglyceride implies higher apparent central volume), Cazaubon_2019_mitotane.R (g/L – 1 g/L = 100 mg/dL; reference 1.56 g/L cohort median; power exponent -0.526 on CL; per-subject median over the observation window).
• Notes: Cardiometabolic lipid-panel covariate. In Archary 2019 the inverse triglyceride–Vc relationship is interpreted as a nutritional-status / hydrophilic-drug-distribution surrogate: triglycerides rise during nutritional rehabilitation, and lamivudine (a hydrophilic drug) shows decreasing apparent volume as nutritional status improves. The linear-deviation form preserves the source’s published parameterization; per-model covariateData[[TRIG]]$units is load-bearing because the centring reference and slope are unit-specific.

LDLC (canonical for low-density lipoprotein cholesterol)

• Description: Serum low-density lipoprotein cholesterol concentration (baseline or time-varying). The pharmacologically meaningful endpoint for lipid-lowering therapies (statins, PCSK9 inhibitors, ANGPTL3 inhibitors); also serves as a baseline covariate or as the time-varying PD response in indirect-response exposure-response models.
• Units: mg/dL or mmol/L – document the unit used in each model via covariateData[[LDLC]]$units (1 mmol/L ~= 38.67 mg/dL for cholesterol).
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (LDLC / ref)^exponent for the baseline-LDLC covariate role, or with no reference (used directly as the PD response state) when it is the modelled output. Reference values observed: 211 mg/dL (Pu 2021 HoFH typical-patient definition).
• Source aliases:
  • LDL-C – common spelling with hyphen.
  • LDL_C – common alternative spelling.
  • LDLBL (baseline LDL-C) – used in Pu_2021_evinacumab.R (Pu 2021 NM-TRAN $INPUT column for centred baseline LDL-C as a covariate on IC50).
• Example models: Pu_2021_evinacumab.R (mg/dL, baseline reference 211 mg/dL; power exponent -1.17 on IC50, where higher baseline LDL-C predicts a smaller IC50 and therefore greater sensitivity to evinacumab; LDL-C is also the PD output state initialised at the baseline value), Nicolas_2019_alirocumab.R (mg/dL; the dual role again, and in its purest form – LDLC is both the initial condition of the ldl turnover state and, through the drug-free steady state kin = kout * LDLC, the anchor of the zero-order production rate, because Nicolas 2019 tabulates no kin and no typical baseline. The type IV indirect response model is linear in the state, so the PERCENT change from baseline it predicts is exactly independent of LDLC, which is why the paper reports its derived endpoints only as percentages)., Nolain_2022_alirocumab.R (mg/dL; baseline [LDLC]baseline as the initial condition of the ldl turnover state and the anchor of kin = kout(0) * LDLC, where kout(0) folds in the baseline free-PCSK9 inhibition of LDL-C degradation; type II indirect response, so percent change is again independent of LDLC).
• Notes: Cardiometabolic lipid-panel covariate. Distinct from HDLC (high-density lipoprotein cholesterol) and from any total-cholesterol or non-HDL-C derivation. When LDL-C is both the response variable AND a covariate (as in Pu 2021, where baseline LDLC drives IC50 and the time-varying state is the modelled PD), document the dual role in covariateData[[LDLC]]$notes.

LDLC_DELTA (canonical for comparator-adjusted change in LDL cholesterol)

• Description: Comparator-adjusted (placebo-adjusted) absolute change in LDL-C achieved on treatment, i.e. the change from baseline in the active arm minus the change from baseline in the comparator arm. Negative values indicate LDL-C lowering relative to the comparator. Distinct from LDLC, which carries the pre-treatment concentration: this canonical is a treatment response, and in model-based meta-analysis it is the covariate through which achieved lipid lowering (rather than assigned dose or drug) predicts a downstream clinical outcome.
• Units: mg/dL or mmol/L – document the unit used in each model via covariateData[[LDLC_DELTA]]$units (1 mmol/L ~= 38.67 mg/dL for cholesterol). The unit is load-bearing because the coefficient is a per-unit slope.
• Type: continuous
• Scope: general
• Reference category: n/a – entered uncentred as a linear term; LDLC_DELTA = 0 is the no-differential-lowering anchor at which the model intercept is defined.
• Source aliases:
  • dLDLc, delta-LDLc – Volkova 2023 Table 1 / Section 2.2 symbol (set with a Greek capital delta in the source).
  • LDL-C change from baseline (placebo-adjusted) – common prose form.
• Example models: Volkova_2023_lipidLowering_mace_mbma.R (mg/dL; linear coefficients on the log risk ratio of coronary revascularization +0.007, coronary mortality +0.004 and nonfatal myocardial infarction +0.008 per mg/dL, with a renal-disease interaction of -0.005 per mg/dL on the nfMI slope).
• Notes: Cardiometabolic lipid-panel covariate; derived counterpart of LDLC. Volkova 2023 Section 2.2 defines the general form for any lipid biomarker as (Biom - Biom_baseline)_treatment - (Biom - Biom_baseline)_comparator, measured at the latest available time point; record the measurement time point per model in covariateData[[LDLC_DELTA]]$notes because trials differ in when the on-treatment lipid panel was drawn. Should a future source need the analogous comparator-adjusted change in another lipid fraction, register a sibling on the same pattern (HDLC_DELTA, REMC_DELTA, TRIG_DELTA) rather than overloading this entry.

REMC (canonical for remnant cholesterol)

• Description: Serum (or plasma) remnant cholesterol concentration – the cholesterol carried by triglyceride-rich lipoproteins (VLDL and IDL in the fasting state, plus chylomicron remnants in the non-fasting state). Computed rather than assayed in most sources, as total cholesterol minus HDL-C minus LDL-C. An independent atherogenic and heart-failure risk factor distinct from the LDL-C fraction.
• Units: mg/dL or mmol/L – document the unit used in each model via covariateData[[REMC]]$units (1 mmol/L ~= 38.67 mg/dL for cholesterol).
• Type: continuous
• Scope: general
• Reference category: n/a – entered uncentred as a linear term in the founding example; a future model using power scaling should record its reference in covariateData[[REMC]]$notes. Observed baseline population mean: 30.72 mg/dL (Volkova 2023 Supplementary Table 3, across 53 trials).
• Source aliases:
  • remC – Volkova 2023 Table 1 / Section 2.2 symbol.
  • RC, remnant-C, TRL-C (triglyceride-rich lipoprotein cholesterol) – common prose forms.
• Example models: Volkova_2023_lipidLowering_mace_mbma.R (mg/dL, baseline; linear coefficient -0.023 per mg/dL on the log risk ratio of heart failure – higher baseline remnant cholesterol predicts greater heart-failure benefit from lipid-lowering therapy).
• Notes: Cardiometabolic lipid-panel covariate, alongside LDLC, HDLC, TCHOL and TRIG. Almost always derived, not measured: when a source does not report it, the Friedewald relations give TCHOL = LDLC + HDLC + TRIG/5 and hence REMC = TCHOL - HDLC - LDLC (Volkova 2023 Section 2.2). Record per model whether the value was reported or derived, and whether the panel was fasting – the Friedewald derivation is invalid at high triglycerides (conventionally above ~400 mg/dL) and non-fasting samples include chylomicron remnants. Note the arithmetic identity REMC = TCHOL - HDLC - LDLC means REMC is collinear with the other three by construction; a model that carries REMC alongside all of TCHOL, HDLC and LDLC is over-parameterised.

ANGPTL3 (canonical for angiopoietin-like protein 3 concentration)

• Description: Total serum angiopoietin-like protein 3 (ANGPTL3) concentration. ANGPTL3 is the pharmacological target for anti-ANGPTL3 monoclonal antibodies (evinacumab) and antisense oligonucleotides (vupanorsen). Baseline ANGPTL3 acts as a soluble-target biomarker that contributes to target-mediated drug disposition; higher baseline target predicts a higher saturable Vmax.
• Units: mg/L (equivalent to ug/mL). Document per-model via covariateData[[ANGPTL3]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – used with power scaling (ANGPTL3 / ref)^exponent. Reference value observed: 0.08 mg/L (Pu 2021 typical-patient median).
• Source aliases:
  • ANGBL (baseline ANGPTL3) – used in Pu_2021_evinacumab.R (Pu 2021 NM-TRAN $INPUT column for centred baseline ANGPTL3).
  • ANGBL0 – alternative Pu 2021 raw column.
• Example models: Pu_2021_evinacumab.R (mg/L, baseline reference 0.08 mg/L; power exponent +0.405 on Vmax, where higher baseline target predicts a faster saturable elimination – biologically consistent with evinacumab being co-cleared along with bound ANGPTL3), Bihorel_2025_evinacumab.R (the refinement of that model; same power form on Vmax with exponent +0.395, but centred on 0.0908 mg/L rather than 0.08 – confirm the reference value per model, as the two evinacumab entries differ).
• Notes: Specific scope because the column is meaningful only for drugs whose mechanism involves ANGPTL3 (anti-ANGPTL3 mAbs / ASOs). Reusing the name for another anti-ANGPTL3 agent is acceptable (extend the example-models list). The assay in Pu 2021 detects both free and target-bound ANGPTL3 after acid pretreatment of serum; document the assay type (free vs. total) per model in covariateData[[ANGPTL3]]$notes.

FPCSK9 (canonical for free (unbound) proprotein convertase subtilisin/kexin type 9 concentration)

• Description: Free (unbound, non-drug-bound) serum proprotein convertase subtilisin/kexin type 9 (PCSK9) concentration. For anti-PCSK9 monoclonal antibodies (alirocumab, evolocumab, bococizumab) the free-PCSK9 pool is the pharmacologically active target fraction; drug-target binding reduces FPCSK9 relative to total PCSK9.
• Units: ng/mL (document per-model if a paper reports a different unit).
• Type: continuous
• Scope: specific
• Reference category: n/a – used with linear-deviation forms TVPARAM + theta * (FPCSK9 / ref) or power-form (FPCSK9 / ref)^theta. Reference values observed: 72.9 ng/mL (Martinez 2019 time-varying median).
• Source aliases: none known.
• Example models: Martinez_2019_alirocumab.R (time-varying; additive-linear effect on Km with slope -0.541 per (FPCSK9/72.9), reference 72.9 ng/mL).
• Notes: Specific scope because the column is meaningful only for drugs whose mechanism is PCSK9 inhibition; reusing the name for a different anti-PCSK9 agent is acceptable (add to Example models). For non-PCSK9 drugs that use a similar target-concentration biomarker, register a new canonical (e.g., FIL6R, FTNF) rather than overloading FPCSK9. Per-model covariateData[[FPCSK9]]$notes should state whether the value is baseline-only or time-varying and how missing values were imputed (Martinez 2019 used LOCF). Distinct from TPCSK9 / TPCSK9_BASE, which carry the TOTAL (free plus drug-bound) pool.

TPCSK9 (canonical for total (free plus drug-bound) PCSK9 concentration, time-varying)

• Description: Total serum proprotein convertase subtilisin/kexin type 9 (PCSK9) concentration – the assay-reported sum of the free pool and the fraction bound to an anti-PCSK9 therapeutic antibody. Time-varying and, under treatment, strongly treatment-elevated: antibody binding slows clearance of the PCSK9 it captures, so total PCSK9 rises several-fold above its pre-treatment level while free PCSK9 falls. The total pool is therefore an exposure-like marker of accumulated target engagement rather than a measure of pharmacologically active target.
• Units: ng/mL (document per-model in covariateData[[TPCSK9]]$units if a different unit – typically nM – is used; conversion uses a PCSK9 molecular weight of ~72 kDa, so 1 nM ~= 72 ng/mL).
• Type: continuous
• Scope: specific
• Reference category: n/a – used with linear-deviation forms TVPARAM + theta * (TPCSK9 - ref) or power scaling. Reference values observed: 3340 ng/mL (Nicolas 2019 on-treatment median of the pooled data set).
• Source aliases:
  • TPCSK9 – Nicolas 2019 Table 2 footnote c covariate name; casing only.
• Example models: Nicolas_2019_alirocumab.R (ng/mL, time-varying; additive-linear effect on the Emax core of the LDL-C indirect response model, +0.000331 per ng/mL relative to the 3340 ng/mL median, spanning a 93-130% increase in Emax between the 5th (491) and 95th (6340) percentiles).
• Notes: Names the analyte with the register’s established total-analyte T prefix (TPRO, TBILI, TBA) and contrasts with the free-pool FPCSK9 on the same analyte; TPCSK9_BASE is the pre-treatment, time-fixed companion on the TBILI / TBILI_BASE pattern. Specific scope for the same reason as FPCSK9: the column is meaningful only for drugs whose mechanism is PCSK9 inhibition; reusing it for another anti-PCSK9 agent is acceptable (extend Example models). The free/total distinction is load-bearing and the two are not interchangeable – in the Nicolas 2019 pooled data set the pre-treatment medians differ roughly two-fold (free 265, total 644 ng/mL) and the on-treatment total median is 3340 ng/mL, five-fold its own baseline, so substituting one column for the other misstates the covariate by a factor of several. No model in the registry generates this trajectory: it must be supplied as a data column, from measurement or from a mechanistic total-PCSK9 model such as Djebli_2017_alirocumab.R, whose total_target state is the same quantity. Per-model covariateData[[TPCSK9]]$notes should record how missing values were imputed (Nicolas 2019 used LOCF).

TPCSK9_BASE (canonical for per-subject baseline total PCSK9 (time-fixed))

• Description: Total (free plus drug-bound) serum PCSK9 concentration measured before the first dose of an anti-PCSK9 therapeutic, held time-fixed per subject. Pre-treatment there is no therapeutic antibody in circulation, so the “bound” fraction is nil and the column measures the endogenous total pool; it is nonetheless distinct from the free baseline value because the assay formats differ and endogenous PCSK9 also circulates bound to LDL particles.
• Units: ng/mL (document per-model in covariateData[[TPCSK9_BASE]]$units if a different unit is used).
• Type: continuous
• Scope: specific
• Reference category: n/a. Nicolas 2019 enters it UNCENTRED (EC50 = (1.44 + 0.00219 * TBSPCSK9) * 1.21^HDSTATIN), so its model intercept is an extrapolation to zero total PCSK9 that is never realised; a model that centres it should record the reference in covariateData[[TPCSK9_BASE]]$notes. Observed 5th-50th-95th percentiles 355 / 644 / 1130 ng/mL (Nicolas 2019 Sect. 3.5).
• Source aliases:
  • TBSPCSK9 – Nicolas 2019 Table 2 footnote b covariate name; also the Nolain 2022 Eq. 6 name.
  • [Ptot]baseline – Nolain 2022 Table 3 symbol (the same quantity entering the TMDD initial condition).
• Example models: Nicolas_2019_alirocumab.R (ng/mL, uncentred additive effect on the LDL-C EC50, +0.00219 mg/L per ng/mL, giving EC50 2.22-3.91 mg/L across the observed 5th-95th percentile range), Nolain_2022_alirocumab.R (nM; power effect (TPCSK9_BASE / 6.99)^0.930 on the free-PCSK9 IC50 of the LDL-C model, centred on the dataset median 6.99 nM, AND the initial condition / synthesis anchor of the QSS-TMDD total-PCSK9 state: Ptot(0) = TPCSK9_BASE * Vc, ksyn = kdeg * TPCSK9_BASE; the paper defines it as the individual observed pre-dose value).
• Notes: Time-fixed baseline companion to TPCSK9, on the same TBILI / TBILI_BASE pattern the register already uses for total bilirubin – carry both columns whenever a model fits a baseline effect and a time-varying effect of the same analyte on different parameters, as Nicolas 2019 does (baseline total on EC50, time-varying total on Emax). Distinct from PCSK9, which is the baseline free (unbound) value and is the covariate on Emax and on the Hill coefficient in the same model: a data set that populates one from the other shifts the covariate by roughly two-fold. Nicolas 2019 computed the baseline as the mean of all measurements taken before the first dose, including screening (Sect. 2.3); record the per-model baseline definition, since a single pre-dose draw and a screening-inclusive mean are not the same quantity.

SBCMA (canonical for soluble B-cell maturation antigen concentration)

• Description: Baseline serum (or plasma) concentration of soluble B-cell maturation antigen (sBCMA), the shed extracellular domain of the BCMA receptor (TNFRSF17). Serves as a soluble-target biomarker for BCMA-directed therapeutics (anti-BCMA antibody-drug conjugates such as belantamab mafodotin, BCMA-targeted bispecifics, and BCMA CAR-T) – sBCMA is elevated in multiple-myeloma and reflects tumour burden, and contributes to target-mediated drug disposition by sequestering circulating drug.
• Units: ng/mL (equivalent to ug/L; 1 ng/mL = 1 ug/L). Document per-model via covariateData[[SBCMA]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – used with power scaling (SBCMA / ref)^exponent. Reference value observed: 50 ng/mL (Papathanasiou 2025 typical-patient definition).
• Source aliases:
  • SBCMABL (baseline soluble BCMA) – used in Papathanasiou_2025_belantamab.R.
• Example models: Papathanasiou_2025_belantamab.R (ng/mL, reference 50; power exponents on initial CL +0.113, on ADC Vc +0.0401, on Imax +0.160), Collins_2023_belantamab_mprotein.R (ng/mL, reference 100; power exponent -0.414 on the effect-compartment rate constant KEO of the serum M-protein tumour-growth-inhibition model).
• Notes: Specific scope because the column is meaningful only for drugs whose mechanism involves the BCMA receptor (and thus a circulating soluble-target pool). Reusing the name for another anti-BCMA agent is acceptable (extend the example-models list). For other oncology TMDD targets register a new canonical (e.g., HER2_ECD already exists for HER2; an analogous SCD20, SCD38 would follow the same pattern). Multiple myeloma populations show sBCMA spanning roughly 2 to 2,000 ng/mL, so the (SBCMA/50)^exponent form should be evaluated with care over the full clinical range.

SBLYS (canonical for soluble B lymphocyte stimulator concentration)

• Description: Baseline serum (or plasma) concentration of soluble B lymphocyte stimulator (BLyS), also called B-cell activating factor (BAFF) or TNFSF13B. A soluble TNF-superfamily ligand that drives B-cell survival and maturation, is elevated in systemic lupus erythematosus and other autoimmune disease, and is the target of BLyS-directed and BLyS/APRIL dual-directed therapeutics (atacicept, belimumab, telitacicept, ianalumab, tabalumab). As a circulating soluble target it sequesters drug and so contributes to target-mediated drug disposition; in TMDD models it typically scales the baseline total-target pool rather than a clearance term.
• Units: ng/mL (equivalent to ug/L; 1 ng/mL = 1 ug/L). Document per-model via covariateData[[SBLYS]]$units.
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (SBLYS / ref)^exponent. Reference value observed: 2.56 ng/mL (Pitsiu 2023; the median of the QUANTIFIABLE values, deliberately not the overall cohort median of 1.79 ng/mL).
• Source aliases:
  • BLYS / BLYSB (baseline BLyS, and its BLQ-imputed twin) – used in Pitsiu_2023_atacicept.R.
  • BAFF, sBAFF, TNFSF13B – the equivalent immunology / assay-kit spellings; map them to SBLYS.
• Example models: Pitsiu_2023_atacicept.R (ng/mL, reference 2.56; power exponent +0.176 on the baseline total-target concentration Rmax of a two-compartment QSS-TMDD model, Pitsiu 2023 Table 1).
• Notes: General scope because serum BLyS is a routine, paper-independent cytokine measurement and an autoimmune disease-activity biomarker, not a cohort definition tied to one analysis – this differs from SBCMA (specific), whose meaning is confined to BCMA-directed oncology programs. Named SBLYS following the SBCMA soluble-target-biomarker pattern, which the SBCMA entry’s own Notes sanction for new soluble TMDD targets; the leading S marks the soluble (shed / circulating) species. BLyS rather than BAFF is used as the canonical stem because the founding paper uses that name exclusively, with the BAFF spellings recorded above as aliases. Assay LLOQ handling is load-bearing and belongs in the per-model notes: BLyS is frequently BLQ in real cohorts (44% of subjects in Pitsiu 2023), and the imputation rule chosen for those subjects materially changes predictions – imputing the founding model’s BLQ subjects at the LLOQ (1.56 ng/mL) instead of the paper’s LLOQ/2 (0.78 ng/mL) overpredicts steady-state Cmax by up to 9%. Atacicept’s second soluble target, APRIL / TNFSF13, has deliberately NOT been registered: Pitsiu 2023 screened it without retaining it, so it lives in that model’s covariatesDataExcluded under the documentation-only key SAPRIL. A future extraction that actually retains an APRIL effect should ratify SAPRIL at that point (note the unit differs – APRIL is reported in pg/mL).

HBA1C (canonical for glycated hemoglobin)

• Description: Glycated hemoglobin (HbA1c, %). Routine clinical laboratory measurement on whole blood reflecting average glycemia over the preceding ~2-3 months; reported in National Glycohemoglobin Standardization Program (NGSP) units.
• Units: % (NGSP). Document the unit standard (NGSP vs IFCC mmol/mol) per-model via covariateData[[HBA1C]]$units when a paper reports IFCC units (IFCC mmol/mol = 10.93 * NGSP% - 23.50).
• Type: continuous
• Scope: general
• Reference category: n/a – used with a power form (HBA1C / ref)^exponent or linear-deviation form. Reference value observed: 5.88% (Oniki 2018 NAFLD-risk dataset baseline mean).
• Source aliases:
  • HbA1c – used in Oniki_2018_nafld_risk.R (dataset column for glycated hemoglobin in %).
• Example models: Oniki_2018_nafld_risk.R (%, reference 5.88; power exponent -3.34 for (HBA1C / 5.88) on the (BMI50 - 17) half-saturation offset of the sigmoidal logit-of-NAFLD function, Oniki 2018 Eq. 4 / Figure 2c).
• Notes: General scope because HbA1c is a routine, paper-independent glycemic-control laboratory measurement. Companion glycemic covariate to FPG (baseline fasting plasma glucose), which is routinely reported alongside HbA1c in T2DM / metabolic-syndrome populations. Distinct from GLU (time-varying within-subject glucose regressor).

Bone biomarkers

TRACP5B_BL (canonical for baseline serum tartrate-resistant acid phosphatase 5b concentration)

• Description: Pre-treatment serum tartrate-resistant acid phosphatase 5b (TRACP-5b) concentration, an osteoclast-derived bone-resorption biomarker. Time-fixed per subject. Used as the per-subject anchor for the steady-state marker pool in indirect-response bone-turnover models (Kin = TRACP5B_BL * Kout, marker(0) <- TRACP5B_BL) and as a continuous covariate on disease-progression and BMD-coupling parameters (power-model standardisation to the cohort median).
• Units: mU/dL (clinical reporting convention used in Japanese osteoporosis cohorts and the Osteolinks TRACP-5b kit; document per-model via covariateData[[TRACP5B_BL]]$units). Some clinical contexts report TRACP-5b in U/L; 1 U/L = 100 mU/dL only if the assays are calibrated equivalently – per-model unit documentation is load-bearing.
• Type: continuous
• Scope: specific
• Reference category: n/a – used with power scaling (TRACP5B_BL / ref)^exponent. Reference value observed: 400 mU/dL (Mori 2018 cohort – mean baseline 401.1 mU/dL across N = 306 Japanese primary-osteoporosis patients; the paper says continuous covariates were standardised to their cohort median but the numeric median is not published, so the rounded cohort mean serves as the documented reference).
• Source aliases:
  • TRACP-5b – text-form of the marker name used in Mori 2018; transliterated to TRACP5B_BL for the per-subject baseline canonical name.
• Example models: Mori_2018_zoledronicAcid.R (mU/dL, reference 400; power exponents -1.534 on EKD50, -1.350 on Slope, -1.319 on T50; and -1.112 on Scale in the active-treatment arm only – the Scale covariate is gated by ON_TREATMENT).
• Notes: Specific scope because the cohort-median reference (400 mU/dL) is tied to the Mori 2018 Japanese primary-osteoporosis cohort and the Osteolinks assay; future bone-resorption-marker extractions using a different assay (different antibodies, different normalisation) should ratify the same canonical and document the per-model assay / reference in covariateData[[TRACP5B_BL]]$notes. Distinct from any time-varying serum TRACP-5b state output (which would be a model output rather than a covariate column).

BMD_BL (canonical for baseline lumbar-spine bone mineral density)

• Description: Pre-treatment lumbar-spine (vertebral, typically L1-L4 or L2-L4) bone mineral density measured by dual-energy X-ray absorptiometry (DXA). Time-fixed per analysis unit – per subject for individual-level fits, per trial arm for aggregate-data (MBMA) fits. Used as the initial condition for a BMD state compartment (bmd(0) <- BMD_BL, BMD_LS(0) <- BMD_BL), as the deviation reference inside a BMD effect-compartment ODE (d/dt(bmd) <- Ke0 * [Scale * (marker - Marker0) - (bmd - BMD_BL)]), and as the anchor for the zero-order BMD synthesis rate of an indirect-response BMD model (kin <- BMD_BL * kout, which makes the drug-free steady state equal the baseline).
• Units: g/cm^2 (the standard DXA areal-BMD unit; document per-model via covariateData[[BMD_BL]]$units). Some clinical contexts report BMD as the dimensionless T-score; the canonical column stores the raw areal density.
• Type: continuous
• Scope: general
• Reference category: n/a – used as a per-subject (or per-arm) anchor; no power-form covariate effect in either example model.
• Source aliases:
  • BMD0 – in-text mathematical symbol used by Mori 2018 for the baseline BMD value entering the BMD ODE; the covariate column name is BMD_BL.
  • BASE – in-text mathematical symbol used by Wang 2023 (Equations 3 and 4, R(0) = KS / KD = BASE) for the baseline lumbar-spine BMD anchoring the K-PD indirect-response model; tabulated per trial as “Baseline of vertebral BMD” (Wang 2023 Table 1).
• Example models: Mori_2018_zoledronicAcid.R (g/cm^2, used as the initial condition bmd(0) <- BMD_BL and the deviation reference inside the BMD ODE; per-subject median 0.677 g/cm^2 across the Mori 2018 cohort), Wang_2023_zoledronicAcid_mbma.R (g/cm^2, per trial arm rather than per subject, used as BMD_LS(0) <- BMD_BL and to set kin <- BMD_BL * kout; per-study values 0.64-1.06 g/cm^2 across the 10 digitised trials).
• Notes: The anatomical site (lumbar spine) and the modality (DXA) are part of the column’s definition; BMD at other sites belongs in a sibling canonical (BMD_FN_BL femoral neck, BMD_TH_BL total hip) rather than overloading this column. Scope was specific while Mori 2018 was the only user and the L2-L4 / Hologic detail of the ZONE protocol was inseparable from it; promoted to general when Wang 2023 arrived as an independent second paper using the identical construct at the identical site with a different instrument mix and a different analysis unit (trial arm rather than subject), which is exactly the generalisation the column needed to survive. The DXA vendor and the precise vertebral levels vary between contributing trials and are model-level detail for covariateData[[BMD_BL]]$notes, not a reason to fork the canonical. Companion bone biomarker to TRACP5B_BL.

Kallikrein-kinin system biomarkers and hereditary-angioedema disease-activity covariates

PKK (canonical for plasma prekallikrein concentration as a time-varying PD driver)

• Description: Plasma prekallikrein concentration used as a time-varying regressor / exposure driver in downstream exposure-response models for hereditary-angioedema (HAE) prophylactic therapies (e.g., prekallikrein-lowering antisense oligonucleotides such as donidalorsen). Averaging convention is per-model (e.g., per-4-week average PKKavg,4W in Singh 2025); the canonical column stores the appropriate exposure summary at each observation row and is consumed with LOCF (piecewise-constant) or linear interpolation semantics as documented per model.
• Units: mg/L (document per-model via covariateData[[PKK]]$units).
• Type: continuous
• Scope: specific
• Reference category: n/a – enters the sigmoidal Emax attack-rate model as PKK^Hill / (EC50^Hill + PKK^Hill) (Singh 2025 page 3 equation). Set to PKK_BL at the pre-dose baseline row and to model-predicted values thereafter (either from a companion popPK/PD model in the library like Diep_2026_donidalorsen or from observed data for placebo arms).
• Source aliases:
  • PKKavg,4W – Singh 2025 Section 2.2 and Table 1 (per-4-week average plasma prekallikrein).
• Example models: Singh_2025_donidalorsen.R (per-4-week average PKK reads at each observation row; drives the sigmoidal Emax term).
• Notes: Specific scope because PKK is meaningful only for HAE / kallikrein-kinin-pathway PD models that consume prekallikrein as an exogenous exposure driver. Companion baseline canonical PKK_BL. The companion popPK/PD backbone for donidalorsen simulations is Diep_2026_donidalorsen (indirect-response model with donidalorsen-driven inhibition of PKK production).

PKK_BL (canonical for per-subject baseline plasma prekallikrein concentration)

• Description: Per-subject baseline (pre-dose) plasma prekallikrein concentration, time-fixed per subject. For donidalorsen-treated cohorts the value is typically the initial condition of an upstream popPK/PD model (e.g., Diep_2026_donidalorsen’s baseline PKK = rbase, derived from lrbase + etalrbase with HAE-status adjustment); for placebo arms the observed baseline PKK is used directly. Enters downstream exposure-response models as a covariate on EC50 via a power form.
• Units: mg/L (document per-model via covariateData[[PKK_BL]]$units).
• Type: continuous
• Scope: specific
• Reference category: n/a – used with power scaling (PKK_BL / ref)^exponent. Reference value observed: 122 mg/L (Singh 2025 page 3 equation (BLPKK/122)^bPKK; matches the Section 2.3 simulation baseline PKK).
• Source aliases:
  • BLPKK – Singh 2025 page 3 equation and Table 1 abbreviation.
• Example models: Singh_2025_donidalorsen.R (per-subject baseline PKK reference 122 mg/L; power exponent e_pkk_bl_ec50 = 0.13 on the EC50^Hill denominator term of the sigmoidal Emax attack-rate model).
• Notes: Specific scope for the same reasons as PKK (HAE / kallikrein-kinin-pathway-bound). Sister to PKK (time-varying). Distinct from the general-purpose _BL baseline-biomarker family (TRACP5B_BL, BMD_BL, HGB_BL, INS_BL, FERRITIN_BL) whose members serve broader clinical contexts; PKK_BL is narrowly the prekallikrein-pathway baseline.

HAERATE_BL (canonical for baseline per-4-week normalized hereditary-angioedema attack rate)

• Description: Per-subject baseline HAE attack rate normalised to a 4-week window, computed from the number of investigator-confirmed HAE attacks during a defined screening / run-in period divided by the contributed days and multiplied by 28 days. Time-fixed per subject. Used as a per-subject covariate on Emax in HAE exposure-response models (higher baseline attack burden -> proportionally higher on-treatment maximum attack rate).
• Units: attacks per 4 weeks (document per-model via covariateData[[HAERATE_BL]]$units).
• Type: continuous
• Scope: specific
• Reference category: n/a – used with power scaling (HAERATE_BL / ref)^exponent. Reference value observed: 3 attacks per 4 weeks (Singh 2025 page 3 equation (BLRATE/3)^bHAE; matches the Section 2.3 simulation baseline attack rate).
• Source aliases:
  • BLRATE – Singh 2025 page 3 equation and Table 1 abbreviation (“baseline per-4-week normalized HAE attack rate”).
• Example models: Singh_2025_donidalorsen.R (per-subject baseline attack rate reference 3 attacks/4W; power exponent e_haerate_bl_emax = 1.03 on Emax of the sigmoidal Emax attack-rate model).
• Notes: Specific scope because the variable is HAE-domain-bound. Time-fixed per subject (baseline-only). Conceptually a specific-domain analogue of ACUTE_MED_DAYS (baseline migraine acute-medication days per month, migraine E-R models); both encode a baseline symptom / event rate as a power-law scaling covariate on the on-treatment response. If future HAE E-R models use a different screening window or normalisation window (e.g., per-day, per-month), document the per-model window in covariateData[[HAERATE_BL]]$notes.

LSR28_BL (canonical for observed baseline natural log-transformed 28-day seizure rate)

• Description: Per-subject observed seizure rate during the baseline (pre-randomisation) period, normalised to a 28-day window and natural log-transformed with no offset: LSR28_BL = log(28 * seizures / baseline days). Time-fixed per subject. Used as a linear covariate on the on-treatment log seizure rate in antiepileptic exposure-response models.
• Units: log(seizures per 28 days) (natural log; document per-model via covariateData[[LSR28_BL]]$units).
• Type: continuous
• Scope: general
• Reference category: n/a – enters linearly without centring in the founding model (slope * LSR28_BL). Observed medians 3.00 (children 4-16 years) and 2.40 (adults) in Chan 2021 Table 3 footnote a.
• Source aliases:
  • Baseline, Baseline_i – Chan 2021 Appendix Equations II (“the observed log-transformed baseline 28-day seizure rate for the ith individual”).
• Example models: Chan_2021_pregabalin_lsr28.R (population-specific slopes 0.945 (adults) and 1.03 (children) on the on-treatment LSR28 of an Emax exposure-response model).
• Notes: The seizure type counted (all focal onset seizures in the founding model) is part of the endpoint definition; record it in covariateData[[LSR28_BL]]$notes. Not interchangeable with a transform that adds an offset before the log – Gidal_2018_eslicarbazepine_ssf.R works on ln(SSF + 0.33) and estimates its baseline rather than reading it as a covariate – so a model on an offset transform needs its own column. LSR28 is an abbreviation (log seizure rate per 28 days) kept because it is the endpoint name used across the pregabalin program.

Drug exposure metrics

CAV (canonical for average drug plasma concentration over a dosing interval)

• Description: Average plasma concentration of the modelled drug over a dosing interval (Cav = AUC_tau / tau). Used as the time-varying or per-period exposure metric in exposure-response models that feed individual empirical-Bayes PK predictions from a previously published population PK model into a downstream PD model.
• Units: ug/mL (document per-model via covariateData[[CAV]]$units).
• Type: continuous
• Scope: specific
• Reference category: n/a – used in Emax/EC50 (e.g., Emax * CAV / (EC50 + CAV)) or power (e.g., (CAV / CavMedian)^exponent) drug-effect terms. Set to 0 for placebo periods.
• Source aliases:
  • CAV, Cav, CAVG, Cav,W (Svensson 2017 weekly-average bedaquiline plasma concentration; same orientation as the canonical, in mg/L).
  • METRIC_TASPO_C – Li 2015 (per-arm taspoglutide average plasma concentration over weeks 2-4, in pmol/L; the model’s source_name is “Cavg.2-4w (Li 2015 Section 3.2)”). This descriptive column name maps onto CAV via this alias rather than being a separate canonical, consistent with the MBMA usage already documented in this entry’s Notes.
• Example models: FiedlerKelly_2020_fremanezumab_em.R, FiedlerKelly_2020_fremanezumab_cm.R, Schoemaker_2018_levetiracetam.R (DDMODEL00000239; LEV plasma concentration in mg/L), Svensson_2017_bedaquiline.R (weekly-average bedaquiline concentration in mg/L driving an Emax effect on the mycobacterial-load half-life; EC50 = 1.42 mg/L, Emax fixed at -100%; placebo subjects use CAV = 0), Li_2015_taspoglutide_mbma.R (MBMA study-arm-level Cavg of taspoglutide between weeks 2 and 4 of QW dosing, in pmol/L; 0 / 59.85 / 119.7 pmol/L for placebo / 10 mg / 20 mg arms; drives an additive Emax response on body-weight change), Lacy_2018_cabozantinib_tumor.R (individual predicted daily-average cabozantinib plasma concentration in ng/mL from the upstream Lacy 2018 popPK; drives a saturable Cavg/(EC50 + Cavg) Hill-1 term on a time-attenuating decay rate of tumor SOD; EC50 = 251 ng/mL; per-cohort steady-state Cavg 375 / 750 / 1125 ng/mL for 20 / 40 / 60 mg/day starting doses; dose-hold periods set CAV = 0), Lacy_2018_cabozantinib_dose_modification.R (same upstream-popPK-derived Cavg in ng/mL; drives a log-linear effect with coefficient theta_drug = 0.000807 per ng/mL on the active-dose log hazard for repeated dose modifications; dose-hold periods set CAV = 0 and switch to the dose-hold baseline log hazard), Sano_2023_fesoterodine_mcc.R (individual 5-HMT average steady-state concentration Cavg,ss in ng/mL from the companion Sano 2023 population PK model, computed as F * DOSE / (CL/F * tau) with tau = 24 h; drives an Emax model on maximum cystometric capacity with EC50 = 6.22 ng/mL; set to 0 on the baseline occasion per Sano 2023 Online Resource 8b, which collapses the Emax term so the baseline prediction is exactly BASE), Yin_2021_pexidartinib.R (running average pexidartinib plasma concentration up to the current tumor-measurement time, in mg/L, from the upstream Yin 2020 popPK modellib('Yin_2020_pexidartinib'); drives a saturable 1 - exp(-kdrug * CAV) drug-effect term in the longitudinal RECIST tumor-size PD model with kdrug typical value 0.196 (mg/L)^-1; placebo periods carry CAV = 0), Panday_2025_sert_tremor_cavg_mbma.R (MBMA study-arm-level steady-state TOTAL average plasma concentration in ng/mL for each of 20 SERT inhibitors at the dose for which tremor incidence was reported; converted to a molar unbound brain concentration via Eqs. S1 and S2b and then to a SERT target coverage driving an Emax model of tremor incidence – note this is a TOTAL, not unbound, concentration, with the unbound correction applied inside model()), Feng_2019_ipilimumab.R (Cavg1, the time-averaged ipilimumab serum concentration over the interval after the first dose, in ug/mL, from an upstream ipilimumab population PK analysis; drives a linear multiplier (1 + CAV * -0.00342) on the linear tumor-growth rate of a three-subpopulation mixture tumor-growth-dynamics model. Unlike the other entries here the averaging window is a single first-dose interval, not a running or steady-state average, and the source paper does not state the concentration unit – see that model’s covariateData[[CAV]]$notes for the dimensional-feasibility argument pinning it to ug/mL), Huh_2024_ritlecitinib.R (average ritlecitinib plasma concentration in ng/mL over the interval between adjacent Severity of Alopecia Tool records, from the empirical-Bayes estimates of the upstream Wojciechowski 2023 population PK model modellib('Wojciechowski_2023_ritlecitinib_final') combined with the patient’s own dosing diary; drives an emax * CAV / (ec50 + CAV) stimulation of a latent-variable indirect-response chain with EC50 = 53.6 ng/mL, essentially the 52 ng/mL Cavg of the 50 mg QD regimen. The averaging window is the OBSERVATION interval rather than a dosing interval, chosen by Huh 2024 Section 4.1 specifically so that treatment interruptions are reflected in the exposure metric; placebo periods and interruptions carry CAV = 0), Patel_2025_radamts13_count_thrombocytopenia.R, Patel_2025_radamts13_count_ldh.R (mean average plasma ADAMTS13 activity over a dosing interval, in IU/mL, averaged over a roughly 6-month prophylaxis period, from the individual post hoc parameters of the companion modellib('Patel_2025_radamts13') popPK; one value per patient per period rather than a time-varying column, driving a sigmoid Emax reduction of the Poisson thrombocytopenia count with EC50 = 0.0149 IU/mL and a log-linear slope of -5.12 per IU/mL on the elevated-LDH count), Patel_2025_radamts13_rtte_thrombocytopenia.R, Patel_2025_radamts13_rtte_ldh.R (the same IU/mL ADAMTS13 activity metric driving a sigmoid Emax term on the log hazard of a repeated time-to-event model, EC50 = 0.0113 and 0.0133 IU/mL respectively; nominally time-varying daily Cave per Patel 2025 Methods, but the paper’s own published simulations hold it constant at the treatment-arm median Cave, and untreated periods carry CAV = 0), Qin_2025_ropeginterferon_chr_week24.R, Qin_2025_ropeginterferon_chr_week52.R, Qin_2025_ropeginterferon_jak2_week24.R, Qin_2025_ropeginterferon_jak2_week52.R, Qin_2025_ropeginterferon_alt_increase.R, Qin_2025_ropeginterferon_ast_increase.R (individual average TOTAL serum ropeginterferon alfa-2b concentration in ng/mL, from the empirical-Bayes parameters of the companion modellib('Qin_2025_ropeginterferon') quasi-equilibrium TMDD population PK model; all six enter the regression LINEARLY and UNCENTRED. This paper is the clearest illustration in the register that the AVERAGING WINDOW is part of the column’s meaning and that one paper may need several mutually non-interchangeable CAV columns: weeks 0-24 from first dose for the week-24 CHR and JAK2 V617F models, weeks 0-52 for their week-52 siblings, and a STUDY-DEPENDENT dose-titration-phase window – weeks 0-16 in A19-201 but weeks 0-6 in A20-202 – for the transaminase safety models. None of these is a steady-state dosing-interval average; each runs from the first dose to its landmark and so is diluted by the titration period), Babel_2026_telisotuzumab_orr.R, Babel_2026_telisotuzumab_neuropathy.R, Babel_2026_telisotuzumab_corneal.R (individual average serum concentration of the telisotuzumab vedotin CONJUGATE in ug/mL, from post hoc estimates of the companion modellib('Babel_2026_telisotuzumab') population PK model using the ACTUAL doses each patient received and averaged up to the time of the event or to the end of treatment; enters on the natural-log scale and uncentred), Babel_2026_telisotuzumab_teae.R (the same paper’s fourth endpoint, but the PAYLOAD metric: individual average plasma concentration of unconjugated monomethyl auristatin E in ng/mL, from modellib('Babel_2026_telisotuzumab_mmae'), entering LINEARLY and uncentred. This paper is the register’s clearest warning that one CAV column name can carry two different ANALYTES with two different UNITS inside a single publication – ug/mL of conjugate for three endpoints, ng/mL of payload for the fourth – so a consumer must read each model’s covariateData[[CAV]]$units rather than assume the family is homogeneous), Qin_2025_ropeginterferon_anemia.R (individual average TOTAL serum ropeginterferon alfa-2b concentration in ng/mL, from the empirical-Bayes parameters of the companion modellib('Qin_2025_ropeginterferon') quasi-equilibrium TMDD population PK model; all seven enter the regression LINEARLY and UNCENTRED. This paper is the clearest illustration in the register that the AVERAGING WINDOW is part of the column’s meaning and that one paper may need several mutually non-interchangeable CAV columns: weeks 0-24 from first dose for the week-24 CHR and JAK2 V617F models, weeks 0-52 for their week-52 siblings, a STUDY-DEPENDENT dose-titration-phase window – weeks 0-16 in A19-201 but weeks 0-6 in A20-202 – for the transaminase safety models, and that window’s COMPLEMENT, the dose-maintenance phase running from week 16 (A19-201) or week 6 (A20-202) to the end of treatment, for the anemia safety model. The first four are not steady-state dosing-interval averages; each runs from the first dose to its landmark and so is diluted by the titration period. The maintenance-phase column is the one metric of the seven that excludes titration, and is therefore systematically the highest), Gidal_2018_eslicarbazepine_ssf.R, Gidal_2018_eslicarbazepine_response.R, Gidal_2018_eslicarbazepine_weekly_seizures.R (the three eslicarbazepine efficacy models of one paper, all driven by the same empirical-Bayes Cav-ss in ng/mL but through three different functional forms – an Emax with EC50 3,530 ng/mL on log-transformed 28-day seizure frequency, a power term (CAV / 10205)^0.609 on the responder logit, and an Emax with EC50 9,450 ng/mL on a weekly Poisson count. The two EC50 estimates differ by a factor of 2.7 between models fitted to the same trials, which is a useful caution against treating a Cav-ss-driven EC50 as a property of the drug rather than of the endpoint), Bi_2018_testosteroneCypionate_sperm.R (18-week BACKWARD-ROLLING average total testosterone concentration in ng/mL, driving a sigmoid Emax inhibition of sperm production with Cavg50 = 8.68 ng/mL and a Hill coefficient of 11.3. Two features make this entry unusual. First, the window length was itself a model-selection decision – Bi 2018 tabulated candidate windows from 3 to 40 weeks and kept 18 weeks on objective-function value – so the number is a fitted structural choice, not a protocol constant. Second, and unlike every other example here, the column is NOT zero before treatment: the analyte is endogenous, so the pre-dose value is the subject’s own baseline testosterone average (around 5-6 ng/mL) and the model’s unsuppressed initial condition depends on that value sitting well below Cavg50), Chan_2021_pregabalin_lsr28.R (individual predicted average steady-state pregabalin concentration Cav,ss in ug/mL from the final population PK model modellib('Chan_2021_pregabalin'), i.e. daily dose / (24 * individual CL/F); drives an Emax term on the natural log-transformed 28-day seizure rate with a common EC50 of 4.69 ug/mL in children and adults; placebo patients carry CAV = 0), Brendel_2021_irinotecan_liposomal_diarrhea_irinotecan.R, Brendel_2021_irinotecan_liposomal_diarrhea_sn38.R, Brendel_2021_irinotecan_liposomal_neutropenia_irinotecan.R, Brendel_2021_irinotecan_liposomal_neutropenia_sn38.R (steady-state Cavg,ss = AUCss,tau / tau from post hoc estimates of the companion Brendel 2021 popPK model, entering univariable logistic exposure-safety models on the log10 scale; total irinotecan in ug/mL for the _irinotecan pair and the METABOLITE SN-38 in ng/mL for the _sn38 pair, so the unit differs between the two pairs), Lalovic_2020_lemborexant_somnolence.R, Lalovic_2020_lemborexant_nasopharyngitis.R, Lalovic_2020_lemborexant_uti.R, Lalovic_2020_lemborexant_influenza.R, Lalovic_2020_lemborexant_urti.R, Lalovic_2020_lemborexant_headache.R (individual average steady-state lemborexant concentration Cav,ss in ng/mL = daily dose / (24 h x CL/F) from the companion Lalovic_2020_lemborexant popPK; linear on the logit of each treatment-emergent adverse-event probability, with placebo selected by the PLACEBO indicator rather than CAV = 0), Shulgin_2020_ici_trae34_mbma.R, Shulgin_2020_ici_trae34_covariate_mbma.R, Shulgin_2020_ici_gi_imae34_mbma.R, Shulgin_2020_ici_hepatic_imae34_mbma.R (study-arm-level MBMA use in nM: the simulated steady-state average concentration of the anti-CTLA-4 antibody over doses 3-4 of the cohort regimen, divided by IC50_CTLA4; 0 for regimens without an anti-CTLA-4 antibody), Zhang_2022_ici_irae_mbma.R, Zhang_2022_ici_irae_grade3_mbma.R (study-arm-level MBMA use in ng/mL: the steady-state average plasma concentration of the anti-CTLA-4 antibody at the cohort regimen, divided by IC50_CTLA4 in the same unit; the paper does not state the body weight or averaging window, and the normalized exposures it fitted on are listed per cohort in its supplementary dataset, so those models document the per-regimen values to supply), Bihorel_2021_BMS986166_nddhr.R, Bihorel_2021_BMS986166_nalc_sd.R, Bihorel_2021_BMS986166_nalc_md.R (individual average blood concentration of the ACTIVE METABOLITE BMS-986166-P in ng/mL, from the maximum a posteriori estimates of the companion modellib('Bihorel_2021_BMS986166') population PK model: over the first 24 h after the first dose for the day-1 heart-rate nadir and single-dose lymphocyte nadir models, and over the 28th once-daily dosing interval for the repeated-dose lymphocyte nadir model, which is therefore about 85% of steady state rather than a steady-state average; drives inhibitory sigmoid Emax terms with IC50 1.26, 2.693 and 1.72 ng/mL; placebo recipients carry CAV = 0).
• Notes: Specific scope because the value is intrinsically tied to the modelled drug – there is no shared meaning across drugs or studies. Each model’s covariateData[[CAV]]$notes should state how the Cav values are derived (e.g., empirical-Bayes from a referenced population PK model) and that the column is set to 0 for placebo periods. The averaging window is also model-specific (per-dosing-interval Cav = AUC_tau / tau in Schoemaker 2018 / Fiedler-Kelly 2020, but weekly-rolling-mean Cav_W in Svensson 2017 – where the bedaquiline once-daily loading + thrice-weekly maintenance schedule makes “per dosing interval” ambiguous; weeks 2-4 Cavg in Li 2015 carried forward for the entire 8-52 week follow-up); document the averaging convention in each model’s covariateData[[CAV]]$notes. MBMA usage (Li 2015) treats CAV as a study-arm-level (not individual-level) exposure metric – the meaning is the same (period-averaged plasma concentration of the modelled drug) so a separate canonical is not warranted.

DOSE_EMPA_MGD (canonical for daily empagliflozin dose)

• Description: Patient’s own once-daily empagliflozin dose, in mg/day. Per-dose-record covariate; constant within an inter-dose interval and updated when the prescriber alters the daily dose. Set to 0 mg/day for placebo arms.
• Units: mg/day
• Type: continuous
• Scope: specific
• Reference category: n/a – enters into the steady-state AUC computation AUCss [nM*h] = DOSE_EMPA_MGD * 1e6 / 450.91 / cl that drives the Emax stimulation of FPG elimination in the indirect-response PD model.
• Source aliases:
  • DOSE – used in Baron_2016_empagliflozin.R (Baron 2016 Methods; studied doses 1, 5, 10, 25, 50 mg QD per Table S2 with 10 and 25 mg dominating; 1129 / 40.9% and 1269 / 46.0% of patients respectively).
• Example models: Baron_2016_empagliflozin.R (drives AUCss feeding Gmax * AUCss / (AUC50 + AUCss) – AUC50 = 703 nMh), Riggs_2014_empagliflozin.R (the predecessor exposure-response analysis by the same group; drives the STIM function that lowers FPG and raises urinary glucose excretion, with AUC50 = 626 nMh for the pooled studies A + B + D and MW 450.9 rather than 450.91 g/mol as printed in that paper), Sato_2024_sglt2_hba1c_mbma.R (study-arm mean daily dose rather than an individual’s dose; divided by the drug’s UGE reference dose of 11.8 mg/day to give the normalized dose driving a class-level sigmoid Emax on HbA1c change).
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family (e.g., DOSE_PHT_MGKGD for phenytoin). Distinct from DOSE_PHT_MGKGD (per-kg phenytoin daily-dose) – empagliflozin is dosed in flat mg/day (no per-kg adjustment in label). The sibling once-daily SGLT2-inhibitor canonicals anticipated by this entry now exist: DOSE_CANA_MGD, DOSE_DAPA_MGD, DOSE_IPRA_MGD, DOSE_LUSEO_MGD and DOSE_TOFO_MGD below. A further SGLT2-inhibitor extraction should register its own sibling rather than reuse this name. The column carries both individual-level (Baron 2016, Riggs 2014) and study-arm-level (Sato 2024 MBMA) daily doses; the quantity is the same mg/day dose either way, so a separate arm-level canonical is not warranted – document the aggregation level in each model’s covariateData[[DOSE_EMPA_MGD]]$notes.

DOSE_CANA_MGD (canonical for daily canagliflozin dose)

• Description: Total daily canagliflozin dose, in mg/day. Set to 0 mg/day for placebo arms and for arms given a different SGLT2 inhibitor.
• Units: mg/day
• Type: continuous
• Scope: specific
• Reference category: n/a – divided by the drug’s UGE reference dose to form the dimensionless normalized dose that drives the class-level sigmoid Emax.
• Source aliases:
  • Dose (mg) on rows with Drug = canagliflozin – used in Sato_2024_sglt2_hba1c_mbma.R (Sato 2024 Supplementary Table S2; pooled trial doses 50-300 mg/day).
• Example models: Sato_2024_sglt2_hba1c_mbma.R (study-arm mean daily dose divided by the 97.7 mg/day UGE reference dose; the column additionally selects the canagliflozin-specific 1.33-fold Emax potentiation via the derived indicator DOSE_CANA_MGD > 0).
• Notes: Sibling of DOSE_EMPA_MGD in the DOSE_<DRUG>_<UNITS> auto-approve family, registered as that entry’s Notes anticipated. Canagliflozin is dosed in flat mg/day. In Sato_2024_sglt2_hba1c_mbma.R this is a study-arm mean rather than an individual’s dose; exactly one of the six SGLT2-inhibitor dose columns is non-zero on any active arm, and a placebo arm sets all six to 0.

DOSE_DAPA_MGD (canonical for daily dapagliflozin dose)

• Description: Total daily dapagliflozin dose, in mg/day. Set to 0 mg/day for placebo arms and for arms given a different SGLT2 inhibitor.
• Units: mg/day
• Type: continuous
• Scope: specific
• Reference category: n/a – divided by the drug’s UGE reference dose to form the dimensionless normalized dose.
• Source aliases:
  • Dose (mg) on rows with Drug = dapagliflozin – used in Sato_2024_sglt2_hba1c_mbma.R (Sato 2024 Supplementary Table S2; pooled trial doses 1-50 mg/day).
• Example models: Sato_2024_sglt2_hba1c_mbma.R (study-arm mean daily dose divided by the 7.05 mg/day UGE reference dose).
• Notes: Sibling of DOSE_EMPA_MGD in the DOSE_<DRUG>_<UNITS> auto-approve family, registered as that entry’s Notes anticipated. Distinct from the Yao_2023_dapagliflozin_mbma.R PK model, which consumes real rxode2 dose events rather than a dose covariate column.

DOSE_IPRA_MGD (canonical for daily ipragliflozin dose)

• Description: Total daily ipragliflozin dose, in mg/day. Set to 0 mg/day for placebo arms and for arms given a different SGLT2 inhibitor.
• Units: mg/day
• Type: continuous
• Scope: specific
• Reference category: n/a – divided by the drug’s UGE reference dose to form the dimensionless normalized dose.
• Source aliases:
  • Dose (mg) on rows with Drug = ipragliflozin – used in Sato_2024_sglt2_hba1c_mbma.R (Sato 2024 Supplementary Table S2; pooled trial doses 12.5-300 mg/day).
• Example models: Sato_2024_sglt2_hba1c_mbma.R (study-arm mean daily dose divided by the 79.0 mg/day UGE reference dose – the largest reference dose of the six, reflecting ipragliflozin’s comparatively shallow dose-UGE curve).
• Notes: Sibling of DOSE_EMPA_MGD in the DOSE_<DRUG>_<UNITS> auto-approve family. Ipragliflozin is approved only in Japan and South Korea, so its trials are almost all Japanese.

DOSE_LUSEO_MGD (canonical for daily luseogliflozin dose)

• Description: Total daily luseogliflozin dose, in mg/day. Set to 0 mg/day for placebo arms and for arms given a different SGLT2 inhibitor.
• Units: mg/day
• Type: continuous
• Scope: specific
• Reference category: n/a – divided by the drug’s UGE reference dose to form the dimensionless normalized dose.
• Source aliases:
  • Dose (mg) on rows with Drug = luseogliflozin – used in Sato_2024_sglt2_hba1c_mbma.R (Sato 2024 Supplementary Table S2; pooled trial doses 0.5-10 mg/day).
• Example models: Sato_2024_sglt2_hba1c_mbma.R (study-arm mean daily dose divided by the 7.38 mg/day UGE reference dose; luseogliflozin’s 2.5 and 5 mg clinical doses are the lowest normalized doses in the pool, about 0.34 and 0.68).
• Notes: Sibling of DOSE_EMPA_MGD in the DOSE_<DRUG>_<UNITS> auto-approve family. Luseogliflozin is a CYP-metabolised SGLT2 inhibitor approved in Japan, Thailand and Malaysia.

DOSE_TOFO_MGD (canonical for daily tofogliflozin dose)

• Description: Total daily tofogliflozin dose, in mg/day. Set to 0 mg/day for placebo arms and for arms given a different SGLT2 inhibitor.
• Units: mg/day
• Type: continuous
• Scope: specific
• Reference category: n/a – divided by the drug’s UGE reference dose to form the dimensionless normalized dose.
• Source aliases:
  • Dose (mg) on rows with Drug = tofogliflozin – used in Sato_2024_sglt2_hba1c_mbma.R (Sato 2024 Supplementary Table S2; pooled trial doses 2.5-40 mg/day).
• Example models: Sato_2024_sglt2_hba1c_mbma.R (study-arm mean daily dose divided by the 15.0 mg/day UGE reference dose; tofogliflozin’s phase I programme ran to 640 mg, giving it by far the widest normalized-dose span of the six drugs and the narrowest observed UGE range).
• Notes: Sibling of DOSE_EMPA_MGD in the DOSE_<DRUG>_<UNITS> auto-approve family. Tofogliflozin is a CYP-metabolised SGLT2 inhibitor approved in Japan, with the shortest elimination half-life of the six drugs in Sato 2024 Table 1 (5.29 h).

DOSE_EE_UG (canonical for daily ethinyl estradiol dose in a combined oral contraceptive)

• Description: Daily dose of ethinyl estradiol (EE), the estrogen component of a combined oral contraceptive (COC), in ug. Constant within a treatment arm / subject for the duration of a conventional 21/7 monophasic regimen. Supplied as a covariate column rather than as a dosing event for models that describe a COC bleeding or efficacy endpoint without instantiating an EE PK compartment.
• Units: ug (per day)
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as a power effect on a structural parameter, normalized to a reference dose. Reference values observed: 30 ug in Chen_2024_combinedOralContraceptives_btb_mbma.R (Chen 2024 Eq. 6, B_i = theta_B,pop * (DOSE_EE_UG / 30)^e_dose_ee_ug_b, with e_dose_ee_ug_b = -2.45; the NEGATIVE exponent means lower EE doses raise the slow-phase breakthrough-bleeding intercept, giving 2.70x at 20 ug and 5.46x at 15 ug relative to 30 ug).
• Source aliases:
  • EE dose (mcg) – used in Chen_2024_combinedOralContraceptives_btb_mbma.R (Chen 2024 Table 1 / Table S2 column header; observed levels 15, 20, 30 and 35 ug).
• Example models: Chen_2024_combinedOralContraceptives_btb_mbma.R (model-based meta-analysis of breakthrough bleeding across four progestin/EE combinations; EE dose is the only covariate retained on the second-phase intercept and is the paper’s central dose-response finding).
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family (siblings: [[DOSE_EMPA_MGD]], [[DOSE_LOR_MGD]], [[DOSE_TPM_MGD]], [[DOSE_SEMAGLUTIDE_MG]]). Ethinyl estradiol is dosed in flat ug/day in every marketed COC, so the ug unit rather than mg is the label-native scale and the one every source table prints. Dose-as-covariate rather than an rxode2 amt / EVID = 1 event because the founding model is purely algebraic with no PK compartment, following the [[DOSE_AGT_UG]] / [[DOSE_ISOPROTERENOL_UG]] precedent. Pairs with [[DOSE_PROGESTIN_UMOL]], which carries the progestin component of the same regimen; a COC arm is fully described by the two columns together. Distinct from [[CONMED_BIRTHCONTROL]], which is a binary “is this subject on hormonal contraception” flag used when contraception is a co-medication rather than the modelled drug. A future model needing the EE dose of a non-oral route (patch, vaginal ring) should register a route-qualified sibling rather than overload this name, since the delivered daily EE differs from the nominal tablet strength.

DOSE_REVEFENACIN_UG (canonical for the nominal nebulized revefenacin dose in ug)

• Description: Nominal revefenacin inhalation-solution dose loaded in the nebulizer for the current once-daily administration, in ug. Carried as a covariate because the bioavailability of the nebulized dose depends on the dose level; the same value is also given as amt on the dosing record.
• Units: ug
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as a power effect on relative bioavailability, F1 = TV x (DOSE_REVEFENACIN_UG / 175)^0.0987 in Lo_2021_revefenacin.R. The paper normalizes continuous covariates to the population median but does not print the median dose; 175 ug (the approved dose, and the dose level at which the dose-level distribution of the pooled data crosses its median) is the reference the maintainers adopted, and it reproduces the paper’s simulated mean steady-state AUC0-24 of both analytes to about 2%.
• Source aliases:
  • Dose – Lo 2021 Table 2 row ‘Dose effect on F1’; observed levels 22, 44, 88, 175, 350 and 700 ug.
• Example models: Lo_2021_revefenacin.R (175 ug is absorbed 7% more than 88 ug, the paper’s own Discussion figure).
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family (siblings [[DOSE_EE_UG]], [[DOSE_AGT_UG]]). Distinct from [[CONMED_REVEFENACIN_DOSE]] in the COPD MBMA family, which is a per-arm total daily dose of revefenacin as one of 23 compared maintenance agents rather than the dose of a modelled PK input. In rxode2 5.1.8 place this column AFTER the event columns (amt, evid, cmt, dur) in the simulation data frame.

DOSE_CEFOTIAM_MG (canonical for the administered cefotiam dose in milligrams)

• Description: Milligram dose of cefotiam administered in the current dosing interval, carried as a covariate so that a cumulative-urinary-recovery endpoint can be expressed as a PERCENT OF DOSE rather than as an amount. Time-varying across dosing intervals in a multiple-dose regimen; constant in the single-dose study that founds the name.
• Units: mg
• Type: continuous
• Scope: specific
• Reference category: n/a – enters only as the denominator of the urinary-recovery endpoint, urinePct = 100 * urine / DOSE_CEFOTIAM_MG. Reference value observed: 1027.5 mg in Shah_2019_cefotiam.R (the single dose given to every subject in the study).
• Source aliases:
  • Dose – the variable name in the Berkeley Madonna Monte Carlo listing of the Shah 2019 Supplementary Materials (set to 1000 mg there because the simulated regimens used 1000 mg every 8 h rather than the study’s 1027.5 mg single dose).
• Example models: Shah_2019_cefotiam.R (three-compartment unbound-referenced popPK in cystic-fibrosis patients and healthy volunteers, with total plasma concentration and percent of dose excreted unchanged in urine fitted simultaneously).
• Notes: Well-formed member of the auto-approved DOSE_<DRUG>_<UNITS> family (cf. [[DOSE_MTX_MGM2]], [[DOSE_CARBOPLATIN_MGKG]], [[DOSE_ELA_MG]]). A dedicated dose column rather than a podo()-derived amount because the endpoint is a CUMULATIVE recovery across the whole 24 h collection, so it must normalize to the total dose administered rather than to the most recent dose event; podo() would silently give the wrong denominator the moment a second dose is given. The alternative encoding – keeping the urinary endpoint as a cumulative amount in mg, as Bulitta_2011_cefpirome.R does with urineAmt – needs no dose column, but it cannot carry Shah 2019’s published additive residual SD of 0.384% without first multiplying it by a study-specific dose, which would hard-code 1027.5 mg into the residual model. Models whose urinary endpoint is an amount should continue to use the urineAmt form and should NOT introduce this column.

DOSE_ELA_MG (canonical for the administered elafibranor dose in milligrams)

• Description: Milligram dose of elafibranor on the dose record, carried as a covariate because relative bioavailability rises with dose through a sigmoidal Emax function of the dose itself. Time-varying across study periods in dose-escalation trials.
• Units: mg
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as 1 + Emax * DOSE_ELA_MG^gamma / (ED50^gamma + DOSE_ELA_MG^gamma). Reference values observed: Emax = 0.800, ED50 = 11.0 mg, gamma = 2.31 (Ooi 2026 Table S3), so the factor is 1.11 at 5 mg, 1.78 at 50 mg and 1.79 at 80 mg – it saturates near 1.8 above roughly 50 mg, which reproduces the less-than-proportional exposure the authors report below 50 mg.
• Source aliases:
  • DOSEN – the NONMEM $INPUT column in Ooi 2026 Supplementary Datafile S1, used in $PK as DOSE_FREL.
• Example models: Ooi_2026_elafibranor.R (doses 5-360 mg pooled over 17 studies, Table S1).
• Notes: Member of the DOSE_<drug>_<units> family. MUST carry the MILLIGRAM dose even when event-table amounts are in umol, because ED50 is reported on the milligram scale. Only the parent model carries this covariate; the GFT1007 model has no dose effect on bioavailability.

DOSE_BEROTRALSTAT_MG (canonical for the administered berotralstat oral dose in milligrams)

• Description: Milligram berotralstat dose on the dose record, carried as a covariate because relative bioavailability rises with dose as a power function of the dose itself. Set it to the dose record’s amt; time-varying across study periods in dose-escalation designs.
• Units: mg
• Type: continuous
• Scope: specific
• Reference category: 300 mg (the median dose in the pooled analysis), where relative bioavailability is 1. Enters as (DOSE_BEROTRALSTAT_MG / 300)^0.497 (Mathis 2022 Equation 1), so the factor is 0.709 at the commercial 150 mg dose and 1.29 at 500 mg.
• Source aliases:
  • DOSE – the NONMEM $INPUT column in Mathis 2022 Supporting Information Text S2, used in $PK as F1=1*(DOSE/CDOSE)**THETA(11)*(1+FOODONF) with CDOSE = 300.
• Example models: Mathis_2022_berotralstat.R (single doses 10-900 mg and once-daily doses 110-450 mg pooled over 13 studies).
• Notes: Member of the DOSE_<drug>_<units> family. In the source studies API-in-capsule doses are expressed as the HCl salt and blend-in-capsule (commercial) doses as the free base (Mathis 2022 Supporting Information Table S1 note); the two formulations were bioequivalent at those nominal doses.

DOSE_PROGESTIN_UMOL (canonical for daily progestin dose on a molar basis)

• Description: Daily dose of the progestin component of a combined oral contraceptive expressed on a MOLAR basis, in umol. Constant within a treatment arm / subject for a conventional 21/7 monophasic regimen. The molar scale is what makes a single column meaningful across chemically distinct progestins: it is the column that carries progestin identity in models that do not fit a separate per-progestin effect.
• Units: umol (per day)
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as a power effect on a structural parameter, normalized to a reference molar dose. Reference values observed: 0.48 umol in Chen_2024_combinedOralContraceptives_btb_mbma.R (Chen 2024 Eq. 5, alpha_i = theta_alpha,pop * (DOSE_PROGESTIN_UMOL / 0.48)^e_dose_progestin_umol_alpha, with e_dose_progestin_umol_alpha = 0.576); 0.48 umol is exactly the levonorgestrel 150 ug daily dose.
• Source aliases:
  • Progestin dose by MW – used in Chen_2024_combinedOralContraceptives_btb_mbma.R (Chen 2024 Table 2 row label; the paper’s Methods state “Given that progestins differ in molecular weight, molar doses were used in the analysis to mitigate collinearity issues”).
• Example models: Chen_2024_combinedOralContraceptives_btb_mbma.R (model-based meta-analysis of breakthrough bleeding; the molar progestin dose is the sole descriptor distinguishing desogestrel, drospirenone, gestodene and levonorgestrel arms, entering as a power effect on the rapid initial decline rate).
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family. The “drug” token is the CLASS (PROGESTIN) rather than a single INN, which is the point of the canonical: the column is only well defined because the molar conversion puts every progestin on one scale. Populate it by dividing the daily progestin dose in ug by the progestin’s molecular weight in g/mol – desogestrel 310.48, drospirenone 366.50, gestodene 310.43, levonorgestrel 312.45 – giving DSG 150 ug = 0.483 umol, DRSP 3000 ug = 8.186 umol, GSD 60 ug = 0.193 umol, GSD 75 ug = 0.242 umol, LNG 100 ug = 0.320 umol, LNG 150 ug = 0.480 umol. Record the molecular weights used in the model’s covariateData[[DOSE_PROGESTIN_UMOL]]$notes, because a source paper reporting only mass doses leaves the conversion implicit. Class-level naming matches [[CONMED_NUC]], CONMED_AZOLE, CONMED_ABX and CONMED_AED, which are likewise drug classes rather than single INNs. Pairs with [[DOSE_EE_UG]]. Note that a model using this column has NO separate progestin-type covariate by construction; if a future extraction fits progestin-specific effects it should carry per-progestin indicators alongside (not instead of) this column, since the two encode different things. Progestin-only-pill regimens are outside the founding model’s domain (Chen 2024 excluded them from model development), so do not use DOSE_EE_UG = 0 with this column to represent a POP.

DOSE_LOR_MGD (canonical for daily lorlatinib dose)

• Description: Patient’s own total daily lorlatinib dose, in mg/day. Per-dose-record covariate; constant within an inter-dose interval and updated when the prescriber alters the daily dose (e.g., following a dose reduction for tolerability). For q.d. regimens the value equals the single-dose amount (e.g., 100 mg q.d. -> DOSE_LOR_MGD = 100); for b.i.d. regimens the value is the sum across the day (e.g., 75 mg b.i.d. -> DOSE_LOR_MGD = 150). Set to 0 mg/day during off-treatment periods (e.g., planned drug holidays).
• Units: mg/day
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as a linear centered effect 1 + e_dose_lor_cl * (DOSE_LOR_MGD - 100) on total lorlatinib clearance (both the single-dose CLI and the steady-state induced CLMX arm), with reference 100 mg/day (the phase II / labelled recommended dose). Effect coefficient observed: e_dose_lor_cl = 0.001 per mg/day in Chen_2021_lorlatinib.R (Chen 2021 Table 4 theta_TDOSE_on_CL); 200 mg/day gives a 10% CL increase, 10 mg/day gives a 9% CL reduction relative to the 100 mg/day reference.
• Source aliases:
  • TDOSE – used in Chen_2021_lorlatinib.R (Chen 2021 Methods “Inclusion of covariates”; total daily lorlatinib dose across the 10-200 mg/day range spanning phase I/II B7461001 escalation cohorts and healthy-participant single-dose studies).
• Example models: Chen_2021_lorlatinib.R (drives dose-nonlinearity of CL via a linear centered term shared between CLI and CLMX; captures increased auto-induction potency at higher doses).
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family (siblings: DOSE_EMPA_MGD for empagliflozin, DOSE_PHT_MGKGD for phenytoin, DOSE_CIPARGAMIN_MG for cipargamin). Lorlatinib is dosed in flat mg/day (no per-kg adjustment in label), matching the DOSE_EMPA_MGD pattern. Distinct from the rxode2/nlmixr2 event column amt (which carries the administered dose at each dose event); DOSE_LOR_MGD is a per-record covariate carrying the current daily-dose LEVEL used by the covariate model, updated at each dose-schedule change. When simulating typical q.d. regimens without dose changes it can be set as a per-subject constant equal to the assigned dose level.

DOSE_TAC_MGD (canonical for daily tacrolimus dose)

• Description: The patient’s own total daily tacrolimus dose, in mg/day. Per-record patient-level covariate; constant within a dosing interval and updated whenever the prescriber alters the dose, which in a therapeutic-drug-monitoring cohort happens often. For the usual twice-daily immediate-release regimen the value is the sum across the day (0.75 mg b.i.d. -> DOSE_TAC_MGD = 1.5), matching the [[DOSE_LOR_MGD]] and [[DOSE_TPM_MGD]] convention.
• Units: mg/day
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as an UNCENTERED exponential effect exp(e_dose_tac_cl * DOSE_TAC_MGD) on apparent oral clearance, with no normalising reference dose. Effect coefficient observed: e_dose_tac_cl = 0.148 per mg/day in Wang_2019a_tacrolimus.R (Wang 2019 Table IV theta TAMT), i.e. +16.0% CL/F per additional mg/day; across the observed 1.0-4.0 mg/day span that is a 1.6-fold range in CL/F. Also used as a median-normalised power (DOSE_TAC_MGD / 5)^e_dose_tac_cl with e_dose_tac_cl = 0.351 in Chen_2021b_tacrolimus.R (Chen 2021 Table 2 DOSE; reference 5 mg/day = pooled cohort median).
• Source aliases:
  • TAMT – used in Wang_2019a_tacrolimus.R (Wang 2019 Methods ‘Patients and data collection’ defines TAMT as the daily dose of tacrolimus; Table I gives 1.62 +/- 0.75 mg/day, median 1.5, range 1.0-4.0).
  • DOSE – used in Chen_2021b_tacrolimus.R (Chen 2021 Results: ‘DOSE is the daily dose of tacrolimus’; Table 1 median 5 mg/day, range 1-11).
• Example models: Wang_2019a_tacrolimus.R (uncentered exponential effect on CL/F in Chinese children with refractory nephrotic syndrome, alongside AGE and CYSC; the only covariate the paper retains that is not a patient characteristic), Chen_2021b_tacrolimus.R (median-normalised power effect on CL/F in Chinese adult kidney transplant recipients, retained alongside CYP3A5 genotype – so in that cohort the dose effect persists even with genotype in the model).
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family (siblings: [[DOSE_LOR_MGD]] for lorlatinib, [[DOSE_TPM_MGD]] for topiramate, [[DOSE_EMPA_MGD]] for empagliflozin). A dose-on-clearance covariate in a therapeutic-drug-monitoring dataset must be read with care and is not by itself evidence of pharmacokinetic nonlinearity. In a TDM cohort the dose is titrated to a trough target, so patients who clear the drug faster are prescribed more of it, and daily dose becomes a surrogate for whatever unmeasured characteristic drives the fast clearance – the association runs from clearance to dose, not the other way around. Wang 2019 says so explicitly, attributing its positive coefficient to CYP3A5 genotype, which was not measured in that cohort (“the effect of TAMT on CL/F may be primarily derived from CYP3A5 gene polymorphisms … at present, CYP3A5 genotyping is not routinely performed in Chinese patients with PRNS”). A future tacrolimus extraction that has genotype available should prefer CYP3A5_EXPR (or the paired CYP3A5_STAR1_HET / CYP3A5_STAR1_HOM indicators) over this column. Distinct from the rxode2 / nlmixr2 event column amt, which carries the amount of each individual administration; DOSE_TAC_MGD is a per-record covariate carrying the current daily-dose LEVEL consumed by the covariate model. The two must be kept consistent when simulating: for a b.i.d. regimen amt is half of DOSE_TAC_MGD, and a simulation that varies the dose without updating this column silently loses the covariate effect. A future paediatric tacrolimus model dosing in mg/kg/day should register a sibling DOSE_TAC_MGKGD rather than reuse this name.

DOSE_VORI_MGD (canonical for daily voriconazole maintenance dose)

• Description: The patient’s own total daily maintenance voriconazole dose, in mg/day (the sum of all doses in a day, e.g. 200 mg twice daily -> 400). Per-record covariate, updated whenever the prescriber changes the maintenance dose.
• Units: mg/day
• Type: continuous
• Scope: specific
• Reference category: n/a – it is the dose-rate input of a steady-state Michaelis-Menten equation, Css = km * F * DOSE_VORI_MGD / (vmax - F * DOSE_VORI_MGD), not a covariate effect on a parameter. The equation is only defined for F * DOSE_VORI_MGD < vmax.
• Source aliases:
  • Daily dose – used in Suetsugu_2021_voriconazole.R (Suetsugu 2021 Eq. 1: ‘daily dose is the amount of voriconazole administered daily’).
• Example models: Suetsugu_2021_voriconazole.R (algebraic steady-state trough model for adult allo-HSCT recipients; oral and intravenous doses treated alike with F fixed to 1).
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family (siblings [[DOSE_TAC_MGD]], [[DOSE_LOR_MGD]], [[DOSE_TPM_MGD]]); the VORI abbreviation matches [[CONC_VORI_NGML]] and [[ORAL_VORI]]. Because the model has no time dimension, the column replaces dosing events entirely: a consumer simulates by putting the dose level on observation rows, not by supplying amt. Distinct from the event column amt and from [[CONMED_VORICONAZOLE]] (a perpetrator indicator in models of other drugs).

DOSE_TPM_MGD (canonical for daily topiramate dose)

• Description: Total daily topiramate dose, in mg/day. Usually a per-record patient-level covariate; constant within a dosing interval and updated when the prescriber alters the daily dose. For a twice-daily regimen the value is the sum across the day (89.2 mg b.i.d. -> DOSE_TPM_MGD = 178.4), matching the [[DOSE_LOR_MGD]] convention. Set to 0 mg/day during off-treatment periods (and in the placebo arms of a model-based meta-analysis). In an MBMA the same column carries the dose ASSIGNED TO A STUDY ARM rather than taken by an individual patient – the quantity is identical, only the aggregation grain differs; record the grain in the model file’s covariateData[[DOSE_TPM_MGD]]$notes.
• Units: mg/day
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as a power effect (DOSE_TPM_MGD / 100)^e_dose_tpm_cl on apparent clearance, normalized at 100 mg/day. Effect coefficient observed: e_dose_tpm_cl = 0.193 in Lee_2024_topiramate.R (Lee 2024 Table S1 theta9); 200 mg/day raises CL/F by 14.3% and 50 mg/day lowers it by 12.5% relative to the 100 mg/day reference.
• Source aliases:
  • DOSE – used in Lee_2024_topiramate.R (Lee 2024 Results CL/F equation and Table S1; cohort mean 178.4 +/- 117.9 mg/day).
  • dose – used in Boucher_2016_topiramate_mbma.R (Boucher 2016 Supplementary Materials Table 2 and the $INPUT ID DOSE NTOT DV line of its NONMEM control stream; study-arm assigned doses of 0, 50, 100 and 200 mg/day).
• Example models: Lee_2024_topiramate.R (mild positive dose-dependence of topiramate apparent clearance, multiplying the additive enzyme-inducing-comedication clearance sum alongside the [[CRCL]] power term), Boucher_2016_topiramate_mbma.R (study-ARM-level assigned dose in a landmark MBMA, where the column is the dose-response driver itself – it enters an Emax term emax * DOSE_TPM_MGD / (ed50 + DOSE_TPM_MGD) on the logit of the arm’s paresthesia rate rather than acting on a structural PK parameter, and a value of 0 recovers the placebo intercept exactly).
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family (siblings: [[DOSE_EMPA_MGD]] for empagliflozin, [[DOSE_LOR_MGD]] for lorlatinib, [[DOSE_PHT_MGKGD]] for phenytoin). Topiramate is dosed in flat mg/day for adults (no per-kg adjustment in the adult label), matching the DOSE_EMPA_MGD / DOSE_LOR_MGD pattern rather than the per-kg DOSE_PHT_MGKGD pattern; a future paediatric topiramate extraction dosing in mg/kg/day should register a sibling DOSE_TPM_MGKGD rather than reuse this name. Distinct from the rxode2 / nlmixr2 event column amt, which carries the amount of each individual administration; DOSE_TPM_MGD is a per-record covariate carrying the current daily-dose LEVEL used by the covariate model. Because the covariate is the daily dose while amt is the per-administration amount, the two must be kept consistent when simulating: for a b.i.d. regimen amt is half of DOSE_TPM_MGD. A dose-level covariate on clearance encodes a mild PK nonlinearity as a covariate rather than as a saturable elimination term, so simulations that vary the dose must update this column or the nonlinearity will be silently lost.

DOSE_ELAGOLIX_MGD (canonical for total daily elagolix dose)

• Description: Total daily elagolix dose, in mg/day, summed across the day’s administrations (150 mg once daily -> 150; 200 mg twice daily -> 400). Per-record, time-varying covariate: set to 0 mg/day for placebo and for off-treatment follow-up, and updated when a treatment sequence switches regimen (e.g. placebo followed by 150 mg once daily in an extension study).
• Units: mg/day
• Type: continuous
• Scope: specific
• Reference category: n/a – enters the shifted/scaled-logit dose-estradiol model e2 = exp(lE2min) + (exp(lE2max) - exp(lE2min)) / (1 + exp(e2slope * DOSE_ELAGOLIX_MGD)) of Stodtmann_2021_elagolix.R (Stodtmann 2021 equation 3, Table 2; slope 0.00894 per mg/day), which returns the average estradiol at the steady state of the current daily dose. A value of 0 gives the untreated level (exp(5.20) + exp(2.14)) / 2 = 94.9 pg/mL.
• Source aliases:
  • DailyDose – used in Stodtmann_2021_elagolix.R (Stodtmann 2021 equation 3: “DailyDose is the total daily dose for elagolix”; fitted range 0-600 mg/day, Figure 1).
• Example models: Stodtmann_2021_elagolix.R (drives average estradiol, which in turn drives the estrogen fraction of the Riggs 2012 bone QSP model and thereby lumbar spine BMD, CTX and P1NP).
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family (siblings: [[DOSE_TPM_MGD]], [[DOSE_LOR_MGD]], [[DOSE_EMPA_MGD]]). The model has no elagolix PK: estradiol is a direct steady-state function of the daily dose, so the column is the only drug input and no dose event rows are needed. Distinct from the rxode2 / nlmixr2 event column amt.

DOSE_GLPG1690_MGD (canonical for total daily GLPG1690 dose)

• Description: Total daily GLPG1690 (ziritaxestat) dose, in mg/day, summed across the day’s administrations (600 mg once daily -> 600; 150 mg twice daily -> 300). For a single-dose record it is the administered dose. Time-fixed per subject within a regimen.
• Units: mg/day
• Type: continuous
• Scope: specific
• Reference category: n/a – enters the linear dose-dependent clearance cl <- exp(lcl + etalcl) * (1 - e_dose_glpg1690_mgd_cl * DOSE_GLPG1690_MGD) of Taneja_2019_glpg1690.R (Taneja 2019 Sect. 3.2 CL_i(DOSE) = TVCL_i * (1 - CLSLP * DOSE), Table 3 slope 0.000415 per mg). A value of 0 gives the dose-independent typical CL/F of 23.3 L/h.
• Source aliases:
  • DOSE – Taneja 2019 Sect. 3.2 and Fig. 3.
• Example models: Taneja_2019_glpg1690.R (apparent clearance falls about 4% per 100 mg/day; studied range 20-1500 mg/day; the linear form reaches CL = 0 at about 2410 mg/day and must not be extrapolated past the studied range).
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family (siblings: [[DOSE_ELAGOLIX_MGD]], [[DOSE_LOR_MGD]], [[DOSE_TPM_MGD]]). The total DAILY dose rather than the per-administration amount is identified from Taneja 2019 Table 4, which prints identical steady-state AUC and identical 95% CI for the QD and BID rows of every total daily dose; that holds only if both regimens share one clearance. Distinct from the rxode2 / nlmixr2 event column amt.

DOSE_SEMAGLUTIDE_MG (canonical for per-subject maintenance semaglutide dose)

• Description: Per-subject maintenance dose of semaglutide, in mg per once-weekly SC injection (the target maintenance dose reached after the dose-escalation phase). Constant per subject in the phase III SUSTAIN maintenance analysis. Values used in the founding example (Carlsson Petri 2018): 0.5 or 1.0 mg once weekly. A binary indicator such as DOSE_LOW_MAINT = as.integer(DOSE_SEMAGLUTIDE_MG < 0.75) can be derived inline in model when the source paper tests a binary low-vs-high maintenance-dose contrast.
• Units: mg (per once-weekly injection)
• Type: continuous
• Scope: specific
• Reference category: n/a – used with a binary indicator derived from the continuous value (e.g., theta^DOSE_LOW_MAINT in Carlsson Petri 2018 with 1.0 mg as the reference).
• Source aliases:
  • DOSE (per-subject maintenance dose column) – used in CarlssonPetri_2018_semaglutide.R (Carlsson Petri 2018 Methods ‘For semaglutide, there are two maintenance doses (0.5 and 1.0 mg), which were included as covariates to assess the dose dependency of semaglutide exposure’).
• Example models: CarlssonPetri_2018_semaglutide.R (binary 0.5-vs-1.0 mg maintenance-dose indicator derived inline in model(); CL/F ratio 1.00 per Table S3 confirming dose proportionality; the coefficient is fixed at 1.00 in the model file to preserve fidelity to the published estimate).
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family. Distinct from the rxode2/nlmixr2 event column amt (which carries the administered dose at dose events; DOSE_SEMAGLUTIDE_MG is a per-subject fixed covariate carrying the assigned MAINTENANCE dose target, not the per-record actual dose which varies during the dose-escalation phase). Future once-weekly-semaglutide extractions with the same 0.5-vs-1.0 mg contrast (or an extension to 2.0 mg / oral 3-14 mg tablets) can reuse this canonical; other GLP-1 receptor agonists dosed in mg (dulaglutide, tirzepatide) should register sibling canonicals (e.g. DOSE_DULAGLUTIDE_MG, DOSE_TIRZEPATIDE_MG) rather than overload this name because the numeric coefficient value is drug-specific. Founded alongside the CarlssonPetri_2018_semaglutide extraction.

DOSE_EFAVIRENZ_MG (canonical for daily dose of coadministered efavirenz)

• Description: Daily dose of efavirenz, in mg, taken by a subject as a coadministered CYP3A4 perpetrator (not as the analyte). 0 = no efavirenz. Carried in addition to the binary CONMED_EFV flag when a source simulates more than one efavirenz dose level and reports a materially different victim-drug exposure for each, because efavirenz autoinduces its own clearance via CYP2B6 and its CYP3A4 induction effect is therefore dose-dependent.
• Units: mg (per once-daily dose)
• Type: continuous
• Scope: specific
• Reference category: 0 (no efavirenz). Used as a level selector, not as a continuous dose-response driver: the founding example derives efv_hi <- CONMED_EFV * (DOSE_EFAVIRENZ_MG >= 500) inline in model() so that the coefficient for the higher published dose level applies at 600 mg and the lower-level coefficient applies at any smaller nonzero dose.
• Source aliases:
  • efavirenz dose – used in Li_2018_guanfacine.R (Li 2018 Section 2.5 and Table 3 simulate efavirenz 400 mg and 600 mg once daily, giving guanfacine AUC ratios of 0.58 and 0.33 respectively).
• Example models: Li_2018_guanfacine.R (selects between two back-solved relative-CYP3A4-activity coefficients, 1.777 at 400 mg and 2.858 at 600 mg, each derived from that dose level’s published guanfacine AUC ratio).
• Notes: Member of the DOSE_<drug>_<units> auto-approve family. Do not interpolate between the published levels – sources that simulate two efavirenz doses give no dose-response function linking them, and the 400 mg perpetrator model is typically built by re-estimating efavirenz oral clearance rather than by scaling the 600 mg one (Li 2018 adjusted efavirenz CL_oral to 17.7 L/h for the 400 mg model). Distinct from the rxode2/nlmixr2 event column amt, which carries the analyte’s dose. Pair with CONMED_EFV rather than replacing it, so a model that only needs the presence/absence contrast can use the binary flag alone.

DOSE_DELAMANID_MG (canonical for administered delamanid dose amount)

• Description: Amount of delamanid, in mg, given on this dose record. A per-dose-record covariate used where relative bioavailability depends on the dose level.
• Units: mg (per dose)
• Type: continuous
• Scope: specific
• Reference category: n/a – used as a level selector. In Sasaki_2022_delamanid.R doses above 50 mg are the reference (F1 multiplier 1) and doses of 50 mg or less have F1 multiplied by 1.580.
• Source aliases:
  • Dose – Sasaki 2022 Table 2 row ‘Dose on F1: increase dose = 50 mg and lower’ (coefficient fixed from the adult population PK model over 50-400 mg).
• Example models: Sasaki_2022_delamanid.R (pediatric delamanid + DM-6705; f_dose <- 1 + 0.580 * (DOSE_DELAMANID_MG <= 50)), Wang_2020_delamanid.R (relative bioavailability 0.760 for 200 mg and 0.580 for 250 or 300 mg vs 100 mg).
• Notes: Member of the DOSE_<drug>_<units> auto-approve family. Carried as a data column rather than read from the dose amount inside model() so it is available at the dose record when f(depot) is evaluated. Delamanid’s bioavailability falls with increasing dose because of its poor solubility.

DOSE_LIRAGLUTIDE_MG (canonical for per-arm assigned liraglutide dose)

• Description: Assigned liraglutide dose in mg per once-daily subcutaneous injection; 0 when the arm/subject did not receive liraglutide. In the founding model-based meta-analysis this is a study-ARM-level covariate carrying the arm’s assigned dose level, and it acts only as a presence indicator because no liraglutide dose-response was estimable.
• Units: mg (per once-daily injection)
• Type: continuous
• Scope: specific
• Reference category: n/a – 0 means “this arm did not receive liraglutide”. In Guo_2025_glp1ReceptorAgonists_mbma.R the model consumes it as (DOSE_LIRAGLUTIDE_MG > 0), so any non-zero value selects the flat liraglutide Emax of -4.25 kg.
• Source aliases:
  • Liraglutide dose – Guo 2025 Results 3.1 and Supplementary Table S3 (per-trial dosing regimen column). The source NMTRAN column name is not separately reported.
• Example models: Guo_2025_glp1ReceptorAgonists_mbma.R (presence indicator only; Guo 2025 Discussion reports “no significant change in weight reduction effect was noted within this range” over liraglutide 1.2-3 mg, so the dose-response was not estimable and the tabulated Emax applies flat).
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family, registered as a sibling of DOSE_SEMAGLUTIDE_MG exactly as that entry’s Notes direct (“other GLP-1 receptor agonists dosed in mg … should register sibling canonicals … rather than overload this name because the numeric coefficient value is drug-specific”). Distinct from the rxode2/nlmixr2 event column amt: an MBMA does not consume dose events, so the assigned dose level must reach model() as a covariate column. Founded alongside the Guo 2025 GLP-1RA weight-reduction MBMA extraction, together with eleven sibling DOSE_<GLP1RA>_MG columns for the other drugs in that analysis.

DOSE_SEMAGLUTIDE_INJ_MG (canonical for per-arm assigned injectable semaglutide dose)

• Description: Assigned subcutaneous (injectable) semaglutide dose in mg per once-weekly injection; 0 when the arm/subject did not receive injectable semaglutide. Route-qualified because injectable and oral semaglutide are separate drugs in a dose-response analysis: they have different bioavailability, different mg dose ranges (0.05-2.4 mg injected vs 1-40 mg oral) and, in the founding model, different estimated Emax values.
• Units: mg (per once-weekly injection)
• Type: continuous
• Scope: specific
• Reference category: n/a – 0 means “this arm did not receive injectable semaglutide”. Enters an Emax-in-dose term with ED50 = 0.384 mg in the founding model.
• Source aliases:
  • Semaglutide (INJ) dose – Guo 2025 Equation 6 and Supplementary Table S6 row “thetaDose on Emax_Semaglutide(INJ)”.
• Example models: Guo_2025_glp1ReceptorAgonists_mbma.R (Emax-in-dose, Emax_dose = -11.7 * Dose / (0.384 + Dose); the paper’s worked example at 1.0 mg reproduces its printed 26-week and 52-week effects of 5.77 and 7.57 kg).
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family, with a route qualifier between the drug and unit tokens on the precedent of DOSE_BPN_SL_MG and DOSE_EFP_MAX_MG. Deliberately NOT merged with the older DOSE_SEMAGLUTIDE_MG, which is scoped to the per-subject maintenance dose target (0.5 vs 1.0 mg) in an individual-level popPK analysis of the SUSTAIN trials; that column is a subject-level maintenance target used to derive a binary contrast, whereas this one is an arm-level assigned dose entering a continuous Emax-in-dose term. A future extraction needing the arm-level injectable dose should reuse this name; one needing the SUSTAIN maintenance-dose contrast should reuse DOSE_SEMAGLUTIDE_MG. Founded alongside the Guo 2025 GLP-1RA weight-reduction MBMA extraction.

DOSE_SEMAGLUTIDE_PO_MG (canonical for per-arm assigned oral semaglutide dose)

• Description: Assigned oral semaglutide dose in mg per once-daily tablet; 0 when the arm/subject did not receive oral semaglutide. Route-qualified for the same reason as DOSE_SEMAGLUTIDE_INJ_MG: the oral tablet’s mg dose range is more than an order of magnitude above the injectable’s, so a shared unqualified column would make the two formulations’ dose-response terms incomparable.
• Units: mg (per once-daily tablet)
• Type: continuous
• Scope: specific
• Reference category: n/a – 0 means “this arm did not receive oral semaglutide”. In Guo_2025_glp1ReceptorAgonists_mbma.R the model consumes it as (DOSE_SEMAGLUTIDE_PO_MG > 0), so any non-zero value selects the flat oral-semaglutide Emax of -5.36 kg.
• Source aliases:
  • Semaglutide (P.O) dose – Guo 2025 Results 3.1 and Supplementary Table S3.
• Example models: Guo_2025_glp1ReceptorAgonists_mbma.R (presence indicator only; no oral-semaglutide dose-response was estimable over the studied 1-40 mg range).
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family with a route qualifier. See DOSE_SEMAGLUTIDE_INJ_MG for why the injectable and oral formulations carry separate canonicals and why neither overloads DOSE_SEMAGLUTIDE_MG. Founded alongside the Guo 2025 GLP-1RA weight-reduction MBMA extraction.

DOSE_EXENATIDE_MG (canonical for per-arm assigned exenatide dose)

• Description: Assigned exenatide dose in mg per administration; 0 when the arm/subject did not receive exenatide. In the founding analysis the column spans both the twice-daily immediate-release and the once-weekly extended-release products without distinguishing them, which is why the unit is per-administration rather than a daily or weekly rate.
• Units: mg (per administration)
• Type: continuous
• Scope: specific
• Reference category: n/a – 0 means “this arm did not receive exenatide”. Consumed as (DOSE_EXENATIDE_MG > 0) in the founding model, selecting the flat exenatide Emax of -6.05 kg.
• Source aliases:
  • Exenatide dose – Guo 2025 Results 3.1 and Supplementary Table S3.
• Example models: Guo_2025_glp1ReceptorAgonists_mbma.R (presence indicator only; no exenatide dose-response was estimable over the studied 0.01-2 mg range).
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family. A future extraction that needs to separate the immediate-release from the extended-release formulation should add a FORM_<...> indicator alongside this column (the FORM_<drug>_<formulation> auto-approve family) rather than split this canonical, because the mg dose itself is the same quantity in both products. Founded alongside the Guo 2025 GLP-1RA weight-reduction MBMA extraction.

DOSE_DANUGLIPRON_MG (canonical for per-arm assigned danuglipron dose)

• Description: Assigned danuglipron (PF-06882961, an oral small-molecule GLP-1 receptor agonist) dose in mg per administration; 0 when the arm/subject did not receive danuglipron.
• Units: mg (per administration)
• Type: continuous
• Scope: specific
• Reference category: n/a – 0 means “this arm did not receive danuglipron”. Enters an Emax-in-dose term with ED50 = 80 mg in the founding model.
• Source aliases:
  • Danuglipron dose – Guo 2025 Equation 2 and Supplementary Table S6 row “thetaDose on Emax_Danuglipron”.
• Example models: Guo_2025_glp1ReceptorAgonists_mbma.R (Emax-in-dose, Emax_dose = -9.29 * Dose / (80 + Dose); the maximum administered dose of 200 mg reaches 66.4 percent of Emax per Guo 2025 Discussion).
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family. Guo 2025 Results 3.1 lists danuglipron under both the mono-agonist and the dual-agonist headings; the compound is a GLP-1 mono-agonist and the duplicate listing is a transcription slip in the source, which does not affect this column. Founded alongside the Guo 2025 GLP-1RA weight-reduction MBMA extraction.

DOSE_LINVENCORVIR_MG (canonical for the nominal linvencorvir dose per administration)

• Description: Nominal oral linvencorvir (RO7049389, RG7907) dose per administration of the subject’s treatment arm, in mg, carried as a covariate because the plasma clearance of the parent falls with dose through a power function of the dose itself. Constant within a treatment arm; the same value on every record of a subject in that arm.
• Units: mg (per administration)
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as (DOSE_LINVENCORVIR_MG / 200)^e_dose_cl on plasma clearance, with exponent -0.684 in the founding model, so the clearance is 71.7 L/h at 200 mg and about 12.7 L/h at 2,500 mg in non-Asian males (Cosson 2021 Results).
• Source aliases:
  • TRT – the NONMEM $INPUT column in the Cosson 2021 control stream (Supplementary Material S5), used in $PK as NDOS = DOSE / (200 * BIO) with DOSE = TRT * BIO, so the bioavailability cancels and only the nominal dose enters.
• Example models: Cosson_2021_linvencorvir.R (single doses 150-2,500 mg and multiple doses of 200-1,000 mg q.d. or 200-800 mg b.i.d.).
• Notes: Member of the auto-approved DOSE_<DRUG>_<UNITS> family. A dedicated column rather than podo() because the stream reads the treatment-arm column TRT, not the dose record’s AMT. The per-administration reading (not the daily total) follows the paper’s definition of the reference value as “CL when dose equals 200 mg” and its Figure 3, whose dose axis spans the single-dose range up to 2,500 mg. rxode2 can drop a dose-named covariate that precedes amt in the event data, so place this column after the event columns.

DOSE_RIFAMPICIN_MG (canonical for the administered daily rifampicin dose in milligrams)

• Description: Daily rifampicin dose in mg, carried as a covariate in victim-drug models where the magnitude of rifampicin-mediated enzyme induction is dose-graded rather than all-or-none. 0 when rifampicin is not coadministered. Paired with the binary CONMED_RIFAMPICIN indicator, which gates whether the column is read at all.
• Units: mg (per day)
• Type: continuous
• Scope: general
• Reference category: n/a – enters as a linear deviation from a stated reference dose. In Ezuruike_2018_ethinylestradiol.R the reference is 600 mg once daily and the term is e_dose_rifampicin_cyp3a4 * (DOSE_RIFAMPICIN_MG - 600) / 300 on the log-activity scale, so the two simulated levels (300 and 600 mg once daily) are reproduced exactly and intermediate doses interpolate.
• Source aliases:
  • Rifampicin 300 mg QD / Rifampicin 600 mg QD – Ezuruike 2018 Table 1, perpetrator dosing column.
• Example models: Ezuruike_2018_ethinylestradiol.R (the two simulated rifampicin regimens imply relative CYP3A4 activities of 4.35 at 300 mg and 4.85 at 600 mg once daily, back-solved from the published ethinylestradiol AUC ratios of 0.40 and 0.36; the column selects between them).
• Notes: Member of the DOSE_<drug>_<units> auto-approve family. Distinct from the rxode2/nlmixr2 event column amt, which carries the VICTIM drug’s dose; this column carries the PERPETRATOR’s dose and never appears on a dose record of its own. Distinct also from CP_RIF_UM (a time-varying rifampicin concentration feeding a mechanistic induction Emax term): use CP_RIF_UM when the source models the perpetrator’s own pharmacokinetics and DOSE_RIFAMPICIN_MG when only the dose level is available and the induction magnitude is read off published exposure ratios. Because the number is a daily dose, a model consuming a divided regimen (e.g. 300 mg twice daily) must supply 600, not 300, and should say so in its covariateData note.

DOSE_ORFORGLIPRON_MG (canonical for per-arm assigned orforglipron dose)

• Description: Assigned orforglipron (LY3502970, an oral non-peptide GLP-1 receptor agonist) dose in mg per once-daily tablet; 0 when the arm/subject did not receive orforglipron.
• Units: mg (per once-daily tablet)
• Type: continuous
• Scope: specific
• Reference category: n/a – 0 means “this arm did not receive orforglipron”. Enters an Emax-in-dose term with ED50 = 14.6 mg in the founding model.
• Source aliases:
  • Orforglipron dose – Guo 2025 Equation 5 and Supplementary Table S6 row “thetaDose on Emax_Orforglipron”.
• Example models: Guo_2025_glp1ReceptorAgonists_mbma.R (Emax-in-dose, Emax_dose = -14.7 * Dose / (14.6 + Dose); the 24 mg simulation tier gives 8.66 kg at 52 weeks, the largest mono-agonist effect in the analysis).
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family. Founded alongside the Guo 2025 GLP-1RA weight-reduction MBMA extraction.

DOSE_TIRZEPATIDE_MG (canonical for per-arm assigned tirzepatide dose)

• Description: Assigned tirzepatide (a GLP-1/GIP dual agonist) dose in mg per once-weekly subcutaneous injection; 0 when the arm/subject did not receive tirzepatide.
• Units: mg (per once-weekly injection)
• Type: continuous
• Scope: specific
• Reference category: n/a – 0 means “this arm did not receive tirzepatide”. Consumed as (DOSE_TIRZEPATIDE_MG > 0) in the founding model, selecting the flat tirzepatide Emax of -12.9 kg.
• Source aliases:
  • Tirzepatide dose – Guo 2025 Results 3.1 and Supplementary Table S3.
• Example models: Guo_2025_glp1ReceptorAgonists_mbma.R (presence indicator only; no tirzepatide dose-response was estimable over the studied 5-15 mg range).
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family, and is the sibling canonical that the DOSE_SEMAGLUTIDE_MG Notes anticipated by name. Founded alongside the Guo 2025 GLP-1RA weight-reduction MBMA extraction.

DOSE_PRETOMANID_MG (canonical for administered pretomanid oral dose in mg)

• Description: Nominal pretomanid oral dose level on the dose record, in milligrams. Enters popPK models as a power-function covariate on absorption and bioavailability parameters – pretomanid shows less-than-proportional exposure with increasing dose under fasting conditions – so it is a dose-record-level covariate rather than only an event-table amt.
• Units: mg
• Type: continuous
• Scope: general
• Reference category: n/a – used as a normalised power term (DOSE_PRETOMANID_MG / 200)^exponent, where 200 mg is the recommended clinical dose.
• Source aliases:
  • DOSE – Salinger 2019 supplemental Table S3 $INPUT column, in mg.
• Example models: Salinger_2019_pretomanid.R (four power-function effects normalised to 200 mg: on relative bioavailability in fasted subjects only (e_dose_fdepot = -0.264, carrying its own random effect), on the first-order absorption rate constant for all subjects (e_dose_ka = -0.128), on the transit mean transit time in fasted subjects only (e_dose_mtt = -0.155), and on the apparent central volume (e_dose_vc = 0.111). A separate indicator term e_dose_1000mg_fdepot = -0.00302 applies only to the 1000 mg FED records, which Salinger 2019 added to keep an odd high-dose fed cohort from distorting the fit near the 200 mg clinical dose), Zou_2022_pretomanid.R (one power-function effect on relative bioavailability, e_dose_fdepot = 0.0822 normalised to 200 mg, fitted to fed-only data over 10-200 mg; the fed-only design is why the exponent is positive here while Salinger 2019’s fasted exponent is negative).
• Notes: Well-formed member of the auto-approved DOSE_<drug>_<units> canonical family. The dose dependence is genuinely a bioavailability / absorption phenomenon rather than saturable elimination: Salinger 2019’s model is linear in clearance at any given dose, and the dose covariate does not touch CL at all. Because several of the effects are gated on prandial state, a data set must carry FED alongside this column for the model to be evaluable. Keep this column consistent with the event table’s amt on dose rows; a mismatch silently rescales bioavailability without changing the administered amount. Sibling to DOSE_TBAJ587_MG and DOSE_BEDAQUILINE_MG in the antitubercular set.

DOSE_CONTEZOLID_MG (canonical for administered oral contezolid dose in mg)

• Description: Nominal oral contezolid dose level on the dose record, in milligrams. Selects a dose-level-specific relative bioavailability, so it is a dose-record-level covariate rather than only an event-table amt.
• Units: mg
• Type: continuous
• Scope: general
• Reference category: 800 mg (the reference dose of Bulitta 2024 Table 2, relative bioavailability fixed to 1 with food).
• Source aliases: none; Bulitta 2024 tabulates the relative bioavailabilities by dose level and fed state (F_rel,400mg,fed, F_rel,1200mg,fasting, …).
• Example models: Bulitta_2024_contezolid.R (categorical use: records below 600 mg take the 400 mg row, 600-1000 mg the 800 mg row and above 1000 mg the 1200 mg row of the dose-level x fed-state relative-bioavailability table, each with its own fixed 0.1 BSV).
• Notes: Well-formed member of the auto-approved DOSE_<drug>_<units> canonical family. Applies to oral contezolid only; IV contezolid acefosamil doses do not read it, so an IV-only record may carry any value (800 is the neutral choice).

DOSE_TBAJ587_MG (canonical for administered TBAJ-587 oral dose in mg)

• Description: Administered oral dose of the second-in-class diarylquinoline antitubercular TBAJ-587, in mg, referenced to 200 mg. Used as a continuous covariate on absorption, clearance and metabolite-formation parameters.
• Units: mg
• Type: continuous
• Scope: specific
• Reference category: n/a – continuous, referenced to 200 mg (the dose level at which every Leding 2026 Table 1 typical value is reported, and the only level studied under fed conditions).
• Source aliases:
  • DOSE – Leding 2026 Supporting Information Code S1 $INPUT (DOSE ; dose in mg), kept distinct there from AMT ; dose in nmol.
• Example models: Leding_2026_tbaj587.R (founding example; five retained effects, all referenced to 200 mg – a power function on apparent parent clearance (exponent 0.298), on both relative fractions metabolised (fm_m3 -0.418, fm_m2 -0.373) and on apparent M2 clearance (0.146), plus an exponential function on the absorption rate constant, ka * exp(-0.000462 * (DOSE - 200)). Studied levels 25, 50, 100, 200, 400 and 800 mg).
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family. This column is not the dosing amount and cannot be derived from it. Leding 2026 fitted natural-log-transformed molar concentrations and dosed AMT in nmol, while the dose covariates act on the mg dose, so a model carrying them needs both quantities on the record; the paper reports no molecular weight, so neither can be computed from the other. Set it on every record of a subject to that subject’s assigned mg dose level. For the multiple-dose simulations of Leding 2026 Methods 2.4 it is the daily mg dose, taking the loading-dose value during a loading period and the maintenance value thereafter – which is what makes the dose dependence of apparent clearance change partway through a loading regimen. The dose non-linearity is real but mechanistically unexplained (Leding 2026 Discussion: “The mechanism of the dose non-linearity is not known but describes likely multiple different processes”); note it raises apparent oral clearance with increasing dose, and the authors report that omitting it or replacing it with a dose-dependent relative bioavailability worsened the fit.

DOSE_VOXELOTOR_MG (canonical for the nominal voxelotor dose of the subject’s regimen)

• Description: Nominal oral voxelotor dose level of the subject’s treatment arm, in mg, carried as a covariate because the whole-blood-to-plasma concentration ratio falls with dose. Constant within an arm.
• Units: mg
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as (DOSE_VOXELOTOR_MG / 900)^e_dose_bpr on the whole-blood-to-plasma ratio Rbp, with exponent -0.37 in the founding model, so Rbp is 17.2% lower at 1500 mg than at 900 mg at the same hematocrit.
• Source aliases:
  • dose – Savic 2022 Table 2 row label ‘Nominal dose on Rbp, (dose/900)^TH’.
• Example models: Savic_2022_voxelotor.R (nominal doses 500-1500 mg once daily in sickle cell disease; Figure 2c).
• Notes: Member of the auto-approved DOSE_<DRUG>_<UNITS> family. A dedicated column rather than podo() because the source uses the NOMINAL dose of the regimen, not the amount of the most recent dose record. Savic 2022 does not say how its 500 mg twice-daily arm was coded (per administration or per day); every other arm was once daily, so the two readings coincide for them.

DOSE_TAF_MG (canonical for total tenofovir alafenamide dose delivered in vivo by a subdermal implant)

• Description: Total mass of tenofovir alafenamide actually released in vivo by the subdermal implant(s) over the whole insertion period, in mg, estimated from the residual drug assayed in the used implants after removal. A per-subject constant describing the entire implant course, referenced to 17.3 mg.
• Units: mg
• Type: continuous
• Scope: specific
• Reference category: n/a – continuous, referenced to 17.3 mg (the cohort median delivered dose, which Beliveau 2026 uses to centre the covariate).
• Source aliases:
  • Dose – Beliveau 2026 Table 4, row “Dose effect on TAF volume”, written as (V/F)TAF x (Dose/17.3)^0.618.
  • Estimated implant dose delivered (mg) – Beliveau 2026 Table 1 baseline-characteristics row name.
• Example models: Beliveau_2026_tenofovir_alafenamide_implant.R (founding example; a power effect with exponent 0.618 on apparent central TAF volume, the only covariate retained in the final model, significant at P < 0.001), Beliveau_2026_tenofovir_alafenamide_implant_weibull.R.
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family. This column is the mass delivered, not the implant payload and not the dosing amount on a record, and the three differ by more than an order of magnitude here: each implant was manufactured with 110 +/- 10 mg of drug, but the estimated delivered dose ranged from 0.05 to 109 mg (Table 1), because most implants released only a small fraction of their payload. In the zero-order-release model the record-level dose is a release RATE in ug/h, so this cumulative mg quantity cannot be derived from it and must be carried on every record of a subject. Beliveau 2026 reads the effect as saturable absorption from the subcutaneous space rather than as a property of the device: doubling the delivered dose from 17.3 to 34.6 mg raises apparent volume by 54% (2^0.618 = 1.535), equivalent to a 35% fall in apparent bioavailability, and halving it to 8.65 mg lowers apparent volume by 36%. Use the generic DOSE_TAF_MG spelling, not an implant-specific one: the column names the drug and its unit, and a future oral or topical tenofovir-alafenamide model with a mg dose covariate belongs under the same canonical.

DOSE_COTADUTIDE_MG (canonical for per-arm assigned cotadutide dose)

• Description: Assigned cotadutide (MEDI0382, a GLP-1/glucagon dual agonist) dose in mg per once-daily subcutaneous injection; 0 when the arm/subject did not receive cotadutide.
• Units: mg (per once-daily injection)
• Type: continuous
• Scope: specific
• Reference category: n/a – 0 means “this arm did not receive cotadutide”. Enters an Emax-in-dose term with ED50 = 0.219 mg in the founding model.
• Source aliases:
  • Cotadutide dose – Guo 2025 Equation 1 and Supplementary Table S6 row “thetaDose on Emax_Cotadutide”.
• Example models: Guo_2025_glp1ReceptorAgonists_mbma.R (Emax-in-dose, Emax_dose = -10.5 * Dose / (0.219 + Dose); the maximum administered dose of 0.6 mg reaches 70.8 percent of Emax per Guo 2025 Discussion).
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family. A cotadutide QSP model already exists in this package (Bosch_2024_cotadutide_qsp.R) but drives its effect from simulated exposure rather than an assigned dose column, so it does not use this covariate. Founded alongside the Guo 2025 GLP-1RA weight-reduction MBMA extraction.

DOSE_MAZDUTIDE_MG (canonical for per-arm assigned mazdutide dose)

• Description: Assigned mazdutide (IBI362 / LY3305677, a GLP-1/glucagon dual agonist) dose in mg per once-weekly subcutaneous injection; 0 when the arm/subject did not receive mazdutide.
• Units: mg (per once-weekly injection)
• Type: continuous
• Scope: specific
• Reference category: n/a – 0 means “this arm did not receive mazdutide”. Consumed as (DOSE_MAZDUTIDE_MG > 0) in the founding model, selecting the flat mazdutide Emax of -7.75 kg.
• Source aliases:
  • Mazdutide dose – Guo 2025 Results 3.1 and Supplementary Table S3.
• Example models: Guo_2025_glp1ReceptorAgonists_mbma.R (presence indicator only; no mazdutide dose-response was estimable over the studied 3-10 mg range).
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family. Founded alongside the Guo 2025 GLP-1RA weight-reduction MBMA extraction.

DOSE_BI456906_MG (canonical for per-arm assigned BI 456906 (survodutide) dose)

• Description: Assigned BI 456906 (survodutide, a GLP-1/glucagon dual agonist) dose in mg per once-weekly subcutaneous injection; 0 when the arm/subject did not receive BI 456906.
• Units: mg (per once-weekly injection)
• Type: continuous
• Scope: specific
• Reference category: n/a – 0 means “this arm did not receive BI 456906”. Consumed as (DOSE_BI456906_MG > 0) in the founding model, selecting the flat BI 456906 Emax of -13.5 kg.
• Source aliases:
  • BI 456906 dose – Guo 2025 Results 3.1 and Supplementary Table S3.
• Example models: Guo_2025_glp1ReceptorAgonists_mbma.R (presence indicator only; no dose-response was estimable over the studied 1.8-4.8 mg range, yet BI 456906 has the largest dual-agonist effect in the analysis at 12.8 kg over 52 weeks).
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family. The DEVELOPMENT CODE is retained in the canonical name rather than the INN survodutide, because Guo 2025 uses “BI 456906” throughout (the INN was assigned after the analysis) and the register’s job is to be findable from the source paper’s own vocabulary. A future extraction of a paper that uses the INN should reuse THIS name and record “survodutide dose” as a source alias rather than register DOSE_SURVODUTIDE_MG; the two are the same compound and a second canonical would silently fragment the drug. Founded alongside the Guo 2025 GLP-1RA weight-reduction MBMA extraction.

DOSE_JNJ64565111_MG (canonical for per-arm assigned JNJ-64565111 dose)

• Description: Assigned JNJ-64565111 (a GLP-1/glucagon dual agonist) dose in mg per once-weekly subcutaneous injection; 0 when the arm/subject did not receive JNJ-64565111.
• Units: mg (per once-weekly injection)
• Type: continuous
• Scope: specific
• Reference category: n/a – 0 means “this arm did not receive JNJ-64565111”. Enters an Emax-in-dose term with ED50 = 6.73 mg in the founding model.
• Source aliases:
  • JNJ-64565111 dose – Guo 2025 Equation 3 and Supplementary Table S6 row “thetaDose on Emax_JNJ-64565111”.
  • JNJ-6456111 and JNJ-65465111 – two digit-transposed misspellings of the same compound that appear in Guo 2025 (in the Equation 3 label and in the Table 1 adverse-event row respectively); the correct code JNJ-64565111 is the one used in Results 3.1, 3.2 and Supplementary Table S6.
• Example models: Guo_2025_glp1ReceptorAgonists_mbma.R (Emax-in-dose, Emax_dose = -18.6 * Dose / (6.73 + Dose); the maximum administered dose of 10 mg reaches only 65.8 percent of Emax, and Guo 2025 Discussion singles this drug out as the one whose effect could be materially improved by dose escalation).
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family, with the hyphen dropped from the development code so the column name stays a valid R identifier. The two source misspellings are recorded above deliberately: a future extractor grepping the source PDF for “JNJ-6456111” would otherwise fail to connect it to this canonical. Founded alongside the Guo 2025 GLP-1RA weight-reduction MBMA extraction.

DOSE_RETATRUTIDE_MG (canonical for per-arm assigned retatrutide dose)

• Description: Assigned retatrutide (LY3437943, a GLP-1/GIP/glucagon TRIPLE agonist) dose in mg per once-weekly subcutaneous injection; 0 when the arm/subject did not receive retatrutide.
• Units: mg (per once-weekly injection)
• Type: continuous
• Scope: specific
• Reference category: n/a – 0 means “this arm did not receive retatrutide”. Enters an Emax-in-dose term with ED50 = 4 mg in the founding model.
• Source aliases:
  • Retatrutide dose – Guo 2025 Equation 4 and Supplementary Table S6 row “thetaDose on Emax_Retatrutide”.
• Example models: Guo_2025_glp1ReceptorAgonists_mbma.R (Emax-in-dose, Emax_dose = -22.6 * Dose / (4 + Dose); retatrutide carries the largest Emax of the 12 drugs analysed and is the only tri-agonist).
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family. Founded alongside the Guo 2025 GLP-1RA weight-reduction MBMA extraction.

DOSE_BPN_SL_MG (canonical for administered sublingual buprenorphine dose)

• Description: Sublingual (SL) buprenorphine dose in mg, supplied as a data column so a dose-dependent SL bioavailability can be evaluated inside model. Needed because rxode2 model code cannot read the amt of the dose record it is scaling, and because a multi-route buprenorphine model must scale only the SL depots by the dose-dependent term while leaving the intravenous and subcutaneous-depot routes at their own (dose-proportional) bioavailability. Route-qualified in the name (_SL_) for exactly that reason: the same model carries intravenous, sublingual, and two subcutaneous depot routes, so an unqualified dose column would be ambiguous.
• Units: mg (per sublingual administration)
• Type: continuous
• Scope: specific
• Reference category: n/a – used with power scaling normalised to a 16 mg reference dose, the typical daily SL maintenance dose: F_SL = 0.14 * (DOSE_BPN_SL_MG / 16)^-0.371.
• Source aliases:
  • Dose – Bjornsson 2023 Table 3 row “Dose dependency on FSL” and the Results 3.1.3 equation FSL = 0.14 x (Dose/16)^-0.371; the source NMTRAN column name is not separately reported.
• Example models: Bjornsson_2023_buprenorphine.R (power-form effect on the logit-transformed SL bioavailability F_SL, exponent -0.371, reference 16 mg; the paper verifies the equation against its own point values of 18.1, 14.0, and 12.0 percent at 8, 16, and 24 mg, the dose-dependence arising from the first-pass metabolism of the swallowed fraction of a sublingual tablet. Set to the SL dose on records in an SL treatment period; the value does not affect intravenous or CAM2038 depot simulations because it scales only the two SL depot compartments), Jones_2021_buprenorphine.R (threshold form rather than power form: the SL bioavailability relative to BUP-XR is multiplied by F1DOSE = 0.765 when DOSE_BPN_SL_MG >= 16, Jones 2021 Section 3.1 and Table 3; scales only the SL depot, not the BUP-XR subcutaneous depots).
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family, with a route qualifier between the drug and unit tokens on the precedent of DOSE_EFP_MAX_MG. BPN is the source paper’s own abbreviation for buprenorphine. Distinct from the rxode2/nlmixr2 event column amt (which carries the administered dose at dose events but is not readable from model), from the generic DOSE canonical (which is unqualified and therefore ambiguous in a model carrying four routes of administration), and from a CAM2038 depot dose column (no such covariate is needed – Bjornsson 2023 Results 3.1.3 found buprenorphine PK dose-proportional for CAM2038 weekly 8-32 mg and monthly 64-192 mg). A future extraction of a different buprenorphine formulation whose own parameters depend on the administered amount should register a sibling (e.g. DOSE_BPN_SC_MG) rather than overload this name. Founded alongside the Bjornsson 2023 buprenorphine / CAM2038 extraction.

DOSE_APOMORPHINE_SL_MG (canonical for administered apomorphine sublingual film dose)

• Description: Apomorphine sublingual film (APL, KYNMOBI) dose in mg of the current administration, supplied as a data column so the dose-dependent sublingual bioavailability can be evaluated inside model. rxode2 model code cannot read the amt of the dose record it scales, and a model that carries both the sublingual film and subcutaneous apomorphine must scale only the sublingual depot, which is why the name is route-qualified.
• Units: mg (per sublingual administration)
• Type: continuous
• Scope: specific
• Reference category: n/a – used with power scaling normalised to a 20 mg reference dose: F1 = Biosl * Biorsc * (DOSE_APOMORPHINE_SL_MG / 20)^-0.206.
• Source aliases:
  • DOSE – Agbo 2021 Table 3 row “Dose of sublingual apomorphine on F1”; the source NMTRAN column name is not reported.
• Example models: Agbo_2021_apomorphine.R (power-form effect on the sublingual bioavailability F1, exponent -0.206, reference 20 mg; the Discussion’s “relative bioavailability of a 30-mg dose … is ~80%” of a 10-mg dose is 3^-0.206 = 0.797. Set to the sublingual film dose on sublingual records; the value on subcutaneous records has no effect because only f(depot) uses it).
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family, with a route qualifier between the drug and unit tokens on the precedent of DOSE_BPN_SL_MG. Distinct from the generic DOSE canonical, which is unqualified and therefore ambiguous in a model that carries two routes. Founded alongside the Agbo 2021 apomorphine extraction.

DOSE_EFP_MAX_MG (canonical for per-subject maximum administered efaproxiral dose)

• Description: Per-subject maximum administered single-dose efaproxiral dose, in mg. Time-fixed per subject; defined as the largest single-administration dose the subject received during the trial. Distinct from a per-administration DOSE column because the covariate enters the model as a static per-subject scalar (a power-model effect on the RBC:plasma proportionality SLPRBC) rather than a per-dose-record exposure regressor.
• Units: mg
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as a power-model effect (DOSE_EFP_MAX_MG / 6800)^theta_SLPRBC~MDOS on the RBC:plasma proportionality SLPRBC. Reference value 6800 mg per Equation 1 of Gastonguay 2005 (close to the median 100 mg/kg x ~72 kg derived from the cohort).
• Source aliases:
  • MDOS – used in Gastonguay_2005_efaproxiral.R (Gastonguay 2005 Data section; per-subject maximum administered dose value used as a power-model covariate on SLPRBC).
• Example models: Gastonguay_2005_efaproxiral.R (efaproxiral; theta_SLPRBC~MDOS = -0.125, 95% CI -0.220 to -0.0222, classified NCI = not clinically important).
• Notes: Specific scope because the absolute reference value (6800 mg) is tied to the efaproxiral cohort and the per-administration MDOS abstraction is uncommon outside hemoglobin-modifier infusion programs. Follows the DOSE_<DRUG>_<MODIFIER>_<UNITS> auto-approve family (e.g., DOSE_PHT_MGKGD for phenytoin, DOSE_EMPA_MGD for empagliflozin) – the _MAX_ token disambiguates from a per-administration DOSE_EFP_MG column that a future extraction may need.

DOSE_SECUKINUMAB_MGWK (canonical for study-arm average weekly secukinumab dose)

• Description: Average weekly secukinumab dose of a study arm over the first 12 weeks of treatment, in mg/week: the total dose given during the 12 weeks divided by 12 weeks. Trial-arm-level regressor of a dose-based MBMA; one value per arm. Set to 0 for placebo arms.
• Units: mg/week
• Type: continuous
• Scope: specific
• Reference category: n/a – the regressor x of the sigmoid Emax trend line E0 + x^hill * (Emax - E0) / (x^hill + E50^hill) for the week-12 placebo-adjusted PASI75 / PASI90 responder rate (Ayyar 2022 Eq. 1, Figure 2).
• Source aliases:
  • x, Dose (mg/week) – Ayyar 2022 Eq. 1 and the Figure 2 x-axis.
• Example models: Ayyar_2022_secukinumab_mbma.R (founding example; E50 about 17 mg/week for PASI75 and 37 mg/week for PASI90).
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family, with MGWK for mg/week by analogy with MGD for mg/day. Differs from the per-administration CONMED_SECUKINUMAB_DOSE of the He 2021 and Checchio 2017 MBMAs, which is the maintenance dose per injection: the loading phase changes this average but not the per-injection dose. In the founding paper the arm x-positions plotted in Figure 2 sit about 25% below the stated definition, so recompute the regressor from the regimen rather than reading it off that figure.

DOSE_IXEKIZUMAB_MGWK (canonical for study-arm average weekly ixekizumab dose)

• Description: Average weekly ixekizumab dose of a study arm over the first 12 weeks of treatment, in mg/week: the total dose given during the 12 weeks divided by 12 weeks. Trial-arm-level regressor of a dose-based MBMA; one value per arm. Set to 0 for placebo arms.
• Units: mg/week
• Type: continuous
• Scope: specific
• Reference category: n/a – the regressor x of the sigmoid Emax trend line for the week-12 placebo-adjusted PASI75 / PASI90 responder rate (Ayyar 2022 Eq. 1, Figure 2).
• Source aliases:
  • x, Dose (mg/week) – Ayyar 2022 Eq. 1 and the Figure 2 x-axis.
• Example models: Ayyar_2022_ixekizumab_mbma.R (founding example; E50 about 2.6 mg/week for PASI75 and 2.9 mg/week for PASI90).
• Notes: Sibling of DOSE_SECUKINUMAB_MGWK; see that entry for how it differs from the per-administration CONMED_IXEKIZUMAB_DOSE. The founding paper’s Figure 2 plots the arms about 20% below the stated definition.

DOSE_CIPARGAMIN_MG (canonical for administered cipargamin single-dose amount)

• Description: Administered single oral dose of the spiroindolone antimalarial cipargamin (formerly KAE609), in mg. Time-fixed per subject in the founding single-dose study (each patient receives one dose on day 1); a per-dose-record covariate in principle if the design ever ran a multi-dose regimen. Not a PK covariate – the amount already appears on the dose record via amt. Used inside model as the regressor in the dose-dependent Emax equation Emax_i = TVEmax * (DOSE_CIPARGAMIN_MG / 10)^COVdose_Emax (Hien 2017 equation 7 in the supplemental text).
• Units: mg
• Type: continuous
• Scope: specific
• Reference category: n/a – enters via (DOSE_CIPARGAMIN_MG / 10)^e_dose_emax with reference dose 10 mg (the smallest study cohort dose in Hien 2017).
• Source aliases:
  • DOSE – used in Hien_2017_cipargamin.R (Hien 2017 Methods ‘Study design’; studied doses 10, 15, 20, 21 [one patient administered 21 mg in error], and 30 mg).
• Example models: Hien_2017_cipargamin.R (drives the dose-dependent typical Emax in the two-population parasite clearance PD model: emax = emax_typical * (DOSE_CIPARGAMIN_MG / 10)^e_dose_emax with e_dose_emax = 0.0463 per Hien 2017 Table 3).
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family (siblings: DOSE_EMPA_MGD, DOSE_PHT_MGKGD). Distinct from those two because cipargamin is a single-dose administration (mg, not mg/day) in the founding study, so the units suffix is MG. Sibling canonicals may be registered for other single-dose antimalarial extractions using the same auto-approve pattern.

DOSE_CABAMIQUINE_MG (canonical for administered cabamiquine single-dose amount)

• Description: Administered single oral dose of the Plasmodium eEF2 inhibitor cabamiquine (formerly DDD107498 / M5717), in milligrams of free base. Time-fixed per subject in the founding studies (each participant receives one dose). Not a PK covariate in the usual sense – the amount already appears on the dose record via amt – but it is required as a data column because the model uses it as the regressor in an empirical power effect of dose on the apparent central volume, V2/F = 2363 * (WT/70)^1 * DOSE_CABAMIQUINE_MG^e_dose_vc with e_dose_vc = -0.50 (Courlet 2023 Supplementary Material 1). The authors introduced the term to describe greater-than-dose-proportional exposure that they could not explain mechanistically: “it was ultimately necessary to include an empirical covariate effect of dose on V2/F in the model”.
• Units: mg (free base)
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as an uncentred power term, DOSE_CABAMIQUINE_MG^e_dose_vc, i.e. an implicit reference dose of 1 mg. Courlet 2023 never reports a centering dose for the term; the uncentred reading is the only one compatible with the reported 146-193 h terminal half-life, because centering on a typical study dose would leave V2/F = 2363 L and a steady-state volume above the Vz implied by that half-life.
• Source aliases:
  • DOSE – used in Courlet_2023_cabamiquine_pk.R and Courlet_2023_cabamiquine.R (Courlet 2023 Results ‘PK/PD data’; 14 dose levels pooled across the two phase I studies).
• Example models: Courlet_2023_cabamiquine_pk.R, Courlet_2023_cabamiquine.R (drives the empirical dose effect on the apparent central volume of the three-compartment transit-absorption / enterohepatic-recycling PK model).
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family (siblings: DOSE_CIPARGAMIN_MG, DOSE_EMPA_MGD, DOSE_TBPPI_MG). The free-base basis is load-bearing: Courlet 2023 study 1 (IBSM) administered cabamiquine as the succinate salt and study 2 (SpzCh) as free base, with 1 mg succinate salt = 0.797 mg free base (Methods ‘Data set’), so the 150 / 400 / 800 mg salt IBSM cohorts are 119.6 / 318.8 / 637.6 mg free base. Populate this column – and the dose-record amt – on the free-base scale. The vignette shows that the free-base reading reproduces the published per-dose recrudescence rates (all three predictions inside the observed Clopper-Pearson 95% CIs) whereas the salt reading does not. Distinct from DOSE_CIPARGAMIN_MG (a different antimalarial, and a covariate on Emax rather than on volume).

DOSE_ANV_MG (canonical for administered ainuovirine dose level)

• Description: Administered once-daily oral dose level of the third-generation non-nucleoside reverse transcriptase inhibitor ainuovirine (ANV), in mg. Not a PK covariate in the usual sense – the amount already appears on the dose record via amt – but it is required as a data column because the model uses it to select which of three piecewise relative-bioavailability levels applies. Han 2024 found ainuovirine to be nonlinear in dose while its elimination half-life was essentially constant across dose levels, and therefore carried the whole nonlinearity in relative bioavailability rather than in a saturable elimination term: F = 1 at 75 mg (reference), 0.716 at 150 mg, 0.410 at 300 mg (Han 2024 Table 3, footnote “F 150 mg is the bioavailability of 150 mg relative to 75 mg”). Time-fixed per subject in both founding trials; a per-dose-record covariate in principle.
• Units: mg
• Type: continuous
• Scope: specific
• Reference category: 75 mg – the dose level at which relative bioavailability is anchored to 1. The column is not a continuous regressor: model() bands it at the arithmetic midpoints of the studied levels (<= 112.5 -> 75 mg, 112.5-225 -> 150 mg, > 225 -> 300 mg) and selects one of three estimated F values. Reference values observed: 75, 150 and 300 mg (Han 2024 phase 1 ADYY-ACC007-103); 150 mg is the recommended clinical dose and the only level in the phase 3 trial ADYY-ACC007-301.
• Source aliases:
  • Dose group (75 mg / 150 mg / 300 mg) – Han 2024 Table 1 / Table 2 column headers; the paper does not print the NONMEM $INPUT column name.
• Example models: Han_2024_ainuovirine.R (selects the piecewise relative bioavailability fdepot = 1 * is_dose75 + 0.716 * is_dose150 + 0.410 * is_dose300 in the two-compartment population PK model – founding example).
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family (siblings: DOSE_CIPARGAMIN_MG, DOSE_CABAMIQUINE_MG, DOSE_TBPPI_MG, DOSE_TAK071_MG). ANV is the paper’s own abbreviation for ainuovirine. The piecewise reading distinguishes this entry from most of the family, which enter as continuous power terms (DOSE_x / ref)^e_dose_param: Han 2024 estimated two independent bioavailability levels rather than one exponent, so a power form would misstate the model, and interpolation between the studied levels carries no support from the source data. A future ainuovirine model that fits a continuous dose-exponent form should record that in its own covariateData[[DOSE_ANV_MG]]$notes rather than re-registering the column. Distinct from MULTI_DOSE_PT, which the same model carries alongside this one: DOSE_ANV_MG selects the bioavailability level (a between-dose-group contrast) while MULTI_DOSE_PT selects the clearance level (a within-subject first-dose-versus-steady-state contrast); the two are orthogonal and both apply to the phase 1 records.

DOSE_TICLOPIDINE_MG (canonical for the ticlopidine dose currently being administered)

• Description: Amount of ticlopidine, in mg, currently being administered, carried on every record of a ticlopidine dosing period and 0 outside it (placebo phase, and the washout that follows the last ticlopidine dose). TIME-VARYING by construction: the fall to 0 at the end of dosing is load-bearing.
• Units: mg
• Type: continuous
• Scope: specific
• Reference category: n/a – the column enters an additive mechanistic term, not a centred covariate effect. In Ashraf_2018_s_ketamine_ticlopidine.R it supplies the Fahmi 2008 gut-input contribution to the portal-vein inhibitor concentration that drives mechanism-based CYP2B6 inactivation: ipv = fu_b,TIC * (A_portal / V_portal + ka_TIC * DOSE_TICLOPIDINE_MG / QH). Ashraf 2018 Study III dosed 250 mg twice daily for 6 days.
• Source aliases:
  • AMT – the reserved NONMEM data item, read directly inside the $DES block of the Ashraf 2018 deposited control stream (Supplementary Information S3): TCE = TFUB*(A(9)+((F7*TKA*AMT)/QH)).
• Example models: Ashraf_2018_s_ketamine_ticlopidine.R (founding example; doi:10.1002/psp4.12346).
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family (siblings: DOSE_VONOPRAZAN_MG, DOSE_ANV_MG, DOSE_CIPARGAMIN_MG, DOSE_TBPPI_MG). Required as an explicit column because rxode2 model code cannot read the amt of a dose record, and because the quantity wanted is the ticlopidine dose specifically – in a joint victim-plus-perpetrator model the most recent dose record is frequently the victim’s, so neither podo() nor NONMEM’s bare AMT is unambiguous. podo() is additionally wrong here because it retains the last dose amount indefinitely, which would hold CYP2B6 permanently inactivated instead of letting activity recover over the 4-5 days the source reports (Ashraf 2018 Figure 3b). Distinct from CONMED_TICLOPIDINE, which marks the study phase and stays 1 through washout.

DOSE_VONOPRAZAN_MG (canonical for administered vonoprazan dose level)

• Description: Administered oral dose level of the potassium-competitive acid blocker vonoprazan, in mg, carried on every record of the dosing interval (not only on dose records). Required as a data column because the model uses it as a continuous power-model regressor on relative oral bioavailability and rxode2 model code cannot read the amt of the dose record it is scaling. Time-fixed per subject in the founding studies, but a per-dose-record covariate in principle – a titration or crossover design would carry different values within one subject.
• Units: mg
• Type: continuous
• Scope: specific
• Reference category: 20 mg – the centring dose at which the effect equals 1, stated in Mulford 2026 Methods 2.2.2 (“we implemented continuous covariate effects as power models centered at 20 mg for dose, 70 kg for weight, and 18 years for age”) and confirmed by the deposited control stream line TVFREL = TVFREL * (DOSE/20)**THETA(10). 20 mg is the highest approved adult vonoprazan dose and one of the two pediatric dose levels studied. Dose levels observed across the ten pooled studies span 1 mg to 120 mg (Table S1), so the power term is supported over that range only.
• Source aliases:
  • DOSE – the $INPUT column name in the Mulford 2026 Supporting Information Model Code.
• Example models: Mulford_2026_vonoprazan.R, Mulford_2026_vonoprazan_maturation.R (continuous power term (DOSE_VONOPRAZAN_MG / 20)^e_dose_fdepot with e_dose_fdepot = 0.290, encoding the slightly over-proportional exposure Mulford 2026 Results 3.1 describes; founding examples).
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family (siblings: DOSE_ANV_MG, DOSE_CIPARGAMIN_MG, DOSE_CABAMIQUINE_MG, DOSE_TBPPI_MG, DOSE_TAK071_MG). Unlike DOSE_ANV_MG, which selects one of three piecewise bioavailability levels, this column is a genuine continuous regressor – the source estimates a single exponent, so interpolation between studied dose levels is supported. The exponent is POSITIVE, so higher doses carry higher relative bioavailability; a consumer who treats the dose non-linearity as saturable elimination will get the direction wrong. In the founding models this column is orthogonal to DAY2, which carries the within-subject day-1-versus-later contrast on absorption rate, clearance and central volume.

DOSE_DNDI6148_MGKG (canonical for administered DNDI-6148 weight-normalised dose level)

• Description: Administered oral dose of the benzoxaborole antileishmanial / antichagasic DNDI-6148, weight-normalised to mg/kg, carried on each dose record. Time-fixed per subject in both founding studies (each mouse stays on one dose level for the whole 10-day course in Henninger 2026; Assmus 2026 is a single-ascending-dose first-in-human study, so each participant receives exactly one dose). Required as a data column because both papers make the weight-normalised dose level a covariate on relative oral bioavailability – Henninger 2026 as a power term, F = (DOSE_DNDI6148_MGKG / 6.25)^theta_dose with theta_dose = -0.34 (Table 1 footnote c), and Assmus 2026 as an exponential term on both F and apparent clearance (see Reference category) – and rxode2 model code cannot read the amt of the dose record it is scaling. The unit token is MGKG rather than MG because both papers’ covariate terms are written on the mg/kg scale and their reference values (6.25 mg/kg and 1.7 mg/kg respectively) are mg/kg quantities; the absolute dose the amt column carries is DOSE_DNDI6148_MGKG * WT * 1000 in ug (Henninger, murine) or DOSE_DNDI6148_MGKG * WT in mg (Assmus, human), which is a different number per subject.
• Units: mg/kg. Henninger 2026 expresses the murine dose levels as free base; Assmus 2026 expresses the human dose levels as free acid equivalent (Methods, ‘Study design’), the drug having been administered as the arginine monohydrate salt. The distinction is a property of each study’s dosing statement, not of the column, which always carries whatever mg/kg quantity its model’s covariate term was estimated against.
• Type: continuous
• Scope: specific
• Reference category: n/a in both founding models, but the two centre the column differently and each is centred so that its effect is exactly 1 at its own reference dose.
  • Henninger 2026 normalises to the lowest studied murine dose level, 6.25 mg/kg: F = (DOSE_DNDI6148_MGKG / 6.25)^e_dose_fdepot, so F = 1 by construction in the 6.25 mg/kg group (Table 1 F_rel = 1.0, fixed). The centring is stated explicitly in the Table 1 footnote and is corroborated by the Results, which report relative bioavailability falling to 79.1 / 62.6 / 49.5 percent at 12.5 / 25 / 50 mg/kg – reproduced to three significant figures by 2^-0.34 = 0.790, 4^-0.34 = 0.625 and 8^-0.34 = 0.494.
  • Assmus 2026 centres on the median dose of its first-in-human cohort, 1.7 mg/kg, with an exponential rather than a power form, and applies it to apparent clearance as well as to bioavailability: F_i = F * exp(e_dose_fdepot * (DOSE_DNDI6148_MGKG - 1.7)) and CL_i/F_i = CL/F * exp(e_dose_cl * (DOSE_DNDI6148_MGKG - 1.7)), with e_dose_fdepot = -0.123 and e_dose_cl = -0.150 (Table 2 footnote c; S1 Code $PK, COV = EXP(THETA(5)*(DOSE_P_KG - 1.7)) and COV2 = EXP(THETA(6)*(DOSE_P_KG - 1.7))). Because the term is a difference in mg/kg rather than a ratio, the coefficient’s unit is kg/mg, not unitless as in the power form. Both effects equal 1 at 1.7 mg/kg. The Results corroborate the F arm directly: F = 121% at the lowest dose (0.12 mg/kg), 100% at the median and 62% at the highest (5.56 mg/kg), reproduced by exp(-0.123 * (0.12 - 1.7)) = 1.21 and exp(-0.123 * (5.56 - 1.7)) = 0.622.
• Source aliases:
  • dose – Henninger 2026 Table 1 footnote c, F = (dose / 6.25 mg/kg)^theta_dose; the source NMTRAN column name is not separately reported.
  • DOSE_P_KG – Assmus 2026 S1 Code $PK; the NMTRAN column name is given verbatim in the published control stream.
• Example models: Henninger_2026_dndi6148_mouse.R (power effect on depot bioavailability, exponent -0.34, reference 6.25 mg/kg; the dose-dependent decrease is attributed by the authors to solubility-limited absorption at the higher dose levels).
  • Henninger_2026_dndi6148_mouse.R (power effect on depot bioavailability, exponent -0.34, reference 6.25 mg/kg; the dose-dependent decrease is attributed by the authors to solubility-limited absorption at the higher dose levels).
  • Assmus_2026_dndi6148.R (exponential effects on depot bioavailability and apparent clearance, coefficients -0.123 and -0.150 kg/mg, centred on the median dose 1.7 mg/kg; the same solubility-limited-absorption mechanism is invoked for the F arm, while the clearance arm is attributed to saturation of a non-renal, probably biliary, elimination pathway).
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family (siblings: DOSE_CABAMIQUINE_MG, DOSE_CIPARGAMIN_MG, DOSE_TBPPI_MG). DNDI6148 drops the hyphen from the sponsor’s DNDI-6148 development code so the column name stays a valid R identifier; DNDI-6148 has no INN at the time of extraction. The dose-dependent bioavailability seen in mice is not carried into Henninger 2026’s own human projection – its Methods 2.2.5 excludes it, along with saturable clearance, from the allometrically scaled human model, so the sibling Henninger_2026_dndi6148_human.R does not use this covariate. Assmus 2026, fitting the actual first-in-human data, does find a dose effect on bioavailability in humans, which retrospectively supports the murine finding while estimating a different functional form on a different dose range; the two models are independent fits and their coefficients are not interconvertible. Distinct from the rxode2/nlmixr2 event column amt, which carries the absolute administered amount but is not readable from model, and from the generic DOSE canonical, which etTrans consumes and does not expose to model. Founded alongside the Henninger 2026 DNDI-6148 extraction; second usage added with the Assmus 2026 extraction.

DOSE_TBPPI_MG (canonical for administered tebipenem pivoxil hydrobromide dose level)

• Description: Administered dose of the oral carbapenem pro-drug tebipenem pivoxil hydrobromide (TBP-PI-HBr) carried on each record, in milligrams of pro-drug (not of the active moiety tebipenem). Time-fixed per subject in fixed-dose designs and time-varying in crossover designs where the same subject receives more than one dose level.
• Units: mg
• Type: continuous
• Scope: specific
• Reference category: n/a – used as a normalized linear term (DOSE_TBPPI_MG / 1200) inside a gated covariate effect, where 1200 mg is the high dose of the study SPR994-104 crossover.
• Source aliases:
  • DOSEMG – the column name printed in Ganesan 2023 Eq. 6.
• Example models: Ganesan_2023_tebipenem.R (dose effect on the absorption rate constant, gated to a single study: ka * (1 + e_dose_ka * (DOSE_TBPPI_MG / 1200) * STUDY_SPR994_104) with e_dose_ka = -0.478).
• Notes: A drug-specific member of the DOSE_<drug>_<units> family, preferred here over the general DOSE canonical for two reasons. First, the dose recorded is pro-drug mass, so every apparent PK parameter estimated against it is conditioned on the pro-drug-to-active-moiety mass ratio and on the (unmeasured, presumed near-complete) conversion fraction – a drug-specific caveat worth carrying in the column name. Second, rxode2’s event-table translator (etTrans) consumes a column literally named DOSE and does not expose it to model as a covariate, so a model that reads the dose level as a covariate must use a distinguishable name or supply it through params.

DOSE_TAK071_MG (canonical for administered TAK-071 dose level)

• Description: Administered dose of TAK-071 (a muscarinic M1 positive allosteric modulator) carried on each dose record, in milligrams. Time-fixed per subject in the fixed-dose Phase 2 cohorts and time-varying in the Phase 1 study, where the same participant received a single dose and then a multiple-dose regimen at a different level, and in the 3-way crossover relative-bioavailability arm.
• Units: mg
• Type: continuous
• Scope: specific
• Reference category: n/a – used as a normalized power term (DOSE_TAK071_MG / 5)^exponent, where 5 mg is the reference dose stated in Jia 2025 Methods (“dose to 5 mg”) and in the Table 1 footnote b (“Typical value for a 5 mg dose with formulation DIC”).
• Source aliases:
  • DOSE – the NONMEM $INPUT column name in the Jia 2025 Supplemental Code control stream.
• Example models: Jia_2025_tak_071.R (two simultaneous power effects on the same column: relative bioavailability Frel = (DOSE_TAK071_MG / 5)^-0.0965 applied as f(depot), and the slow first-order absorption rate ka_slow *= (DOSE_TAK071_MG / 5)^-0.885, which the parallel quick route inherits because the quick rate is parameterised as a ratio to the slow rate. Both exposures and absorption rate therefore fall as dose rises – less-than-dose-proportional exposure for a Biopharmaceutics Classification System Class II compound).
• Notes: A drug-specific member of the auto-approved DOSE_<drug>_<units> family, required here rather than the general DOSE canonical because rxode2’s event-table translator (etTrans) consumes a column literally named DOSE on the rxUi / readModelDb solve path and never exposes it to model, producing “The following parameter(s) are required for solving: DOSE” at solve time. Same rationale as DOSE_TBPPI_MG.

DOSE_UTRELOXASTAT_MG (canonical for administered utreloxastat dose level)

• Description: Administered dose of utreloxastat (PTC857, a 15-lipoxygenase inhibitor in development for amyotrophic lateral sclerosis) carried on each dose record, in milligrams of a single administration (not the daily total; the first-in-human study dosed both QD and BID, so a 250 mg BID subject carries 250, not 500). Time-fixed per subject within each study part of the Hu 2026 FIH analysis (SAD 100-1000 mg single dose, MAD 150-500 mg BID for 14 days, food-effect 500 mg crossover).
• Units: mg
• Type: continuous
• Scope: specific
• Reference category: n/a – used as a normalized power term (DOSE_UTRELOXASTAT_MG / 250)^0.21 on relative bioavailability. Hu 2026 states the power form and the exponent (theta9 = 0.21, Table 3) but never prints the normalizing dose; 250 mg is adopted because it is the reference dose of the paper’s own covariate forest plot (Figure 4 and Table 5, “Reference: 250 mg/BID/fasted”) and the median dose of both the SAD and the MAD cohorts (Table 1). The choice cancels out of every exposure ratio the paper reports, so it is not falsifiable against the publication; it does set the absolute concentration scale, and is recorded here and in the model file so a downstream user can change it deliberately.
• Source aliases:
  • Dose levels – the Hu 2026 Table 3 row label for the theta9 effect (“9 Dose levels on Frel”).
• Example models: Hu_2026_utreloxastat.R (power effect on relative bioavailability applied at f(depot): Frel = (1 + 0.36 * FED_LOWFAT + 0.57 * FED_HIGHFAT) * (DOSE_UTRELOXASTAT_MG / 250)^0.21. The positive exponent makes absorption more-than-dose-proportional – total exposure scales as Dose^1.21 – which reproduces the paper’s Table 5 steady-state exposure ratios of 0.53 at 150 mg and 2.28 at 500 mg relative to 250 mg).
• Notes: A drug-specific member of the auto-approved DOSE_<drug>_<units> family, required here rather than the general DOSE canonical for the same solve-path reason as DOSE_TBPPI_MG and DOSE_TAK071_MG: rxode2’s event-table translator (etTrans) consumes a column literally named DOSE and never exposes it to model. Note the direction is opposite to DOSE_TAK071_MG, whose exponent is negative (less-than-dose-proportional); Hu 2026 attributes the positive exponent to gut-level saturation of efflux transport or intestinal metabolism at higher doses (Discussion), explicitly not to saturable systemic elimination, which the semi-logarithmic terminal-phase parallelism across dose levels rules out.

DOSE_ABROCITINIB_MG (canonical for the administered abrocitinib amount per administration)

• Description: Amount of abrocitinib given at one administration, in mg. Per-dose-record covariate that must equal the amt of the dose records it accompanies.
• Units: mg
• Type: continuous
• Scope: specific
• Reference category: n/a. Observed range 3-800 mg.
• Source aliases:
  • DOSE – Wojciechowski 2022 Online Resource 2 (Dose amount associated with record (mg)).
• Example models: Wojciechowski_2022_abrocitinib.R (founding example; sets the first-order fraction min(1, amt_fo / DOSE_ABROCITINIB_MG) of the parallel zero-/first-order absorption, and switches on a -0.778 logit-bioavailability shift at exactly 800 mg).
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family. Needed as a column, not read from amt, because the model splits each administration into two dose records and evaluates the split at the dose. Distinct from [[DOSE_ABROCITINIB_MGD]], the daily total.

DOSE_ABROCITINIB_MGD (canonical for the randomized total daily abrocitinib dose)

• Description: Total daily abrocitinib dose of the subject’s regimen, in mg/day (for example 400 for 200 mg twice daily). Constant within a regimen.
• Units: mg/day
• Type: continuous
• Scope: specific
• Reference category: 200 mg/day, the reference of the clearance power term.
• Source aliases:
  • DOSR – Wojciechowski 2022 Online Resource 2 (Randomized total daily dose (mg)).
• Example models: Wojciechowski_2022_abrocitinib.R (founding example; clearance scales as (F * DOSE_ABROCITINIB_MGD / 200)^-0.169, with F the individual absolute bioavailability, so the dose term acts on the effective amount reaching the circulation).
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family (sibling of DOSE_LOR_MGD). Distinct from [[DOSE_ABROCITINIB_MG]], the per-administration amount.

DOSE_MANNAC_MG (canonical for administered N-acetylmannosamine (ManNAc) dose level)

• Description: Administered oral dose of N-acetylmannosamine (ManNAc, an orphan-drug substrate-replacement therapy for GNE myopathy) carried on each dose record, in milligrams of a single administration (not the daily total: a 4 g TID subject carries 4000, a 6 g BID subject 6000). Time-varying within subject in the Van Wart 2021 analysis, where the same participants received 3 g or 6 g BID, then 6 g BID, then 4 g TID.
• Units: mg
• Type: continuous
• Scope: specific
• Reference category: n/a – used as a normalized power term (DOSE_MANNAC_MG / 6000)^-0.405 on relative bioavailability, with F = 1 at the 6 g reference dose (Van Wart 2021 Table 4, “F for 6 g dose = 1, Fixed”). Records with the column at 0 leave F at 1, reproducing the control stream’s IF(DOSEMG.GT.0) guard; set it to the dose amount on every dose record.
• Source aliases:
  • DOSEMG – the NONMEM $INPUT column in the Van Wart 2021 Online Resource 6 control stream (F1=1*(DOSEMG/6000)**THETA(11)).
• Example models: VanWart_2021_mannac.R (power effect on relative bioavailability applied at f(depot); F = 1.32 at 3 g and 0.81 at 10 g, the range printed in the Results. The negative exponent makes exposure less than dose-proportional, which the authors attribute to dose-limiting solubility or incomplete absorption at higher doses, and it is why 4 g Q8H gives more plasma Neu5Ac than 6 g Q12H at the same 12 g/day).
• Notes: A drug-specific member of the auto-approved DOSE_<drug>_<units> family, required here rather than the general DOSE canonical for the same solve-path reason as DOSE_TBPPI_MG: rxode2’s event-table translator (etTrans) consumes a column literally named DOSE and never exposes it to model. Place the column after id, time, amt, evid, cmt and dvid in an event table, or rxode2 can drop it.

DOSE_CAPIVASERTIB_MG (canonical for administered capivasertib dose level)

• Description: Planned dose of capivasertib (a pan-AKT inhibitor) carried on each dose record, in milligrams of a single administration and NOT the daily total – every patient in the Fernandez-Teruel 2024 analysis dosed twice daily, so a 400 mg b.i.d. patient carries 400, not 800. Time-fixed per patient in the pooled phase I / II dataset, where the planned dose level ranges over 80, 160, 240, 320, 360, 400, 480, 560, 600, 640 and 800 mg (Fernandez-Teruel 2024 Table 2).
• Units: mg
• Type: continuous
• Scope: specific
• Reference category: n/a – used as a linear-deviation term centred on 480 mg, (1 + (DOSE_CAPIVASERTIB_MG - 480) * e_dose_cl_time_max). The 480 mg centring constant is printed literally in the paper’s covariate equation (f) and is also the modal planned dose of the pooled cohort (175 of 441 patients, 39.7%).
• Source aliases:
  • DOSE – the covariate column name in Fernandez-Teruel 2024 equation (f).
• Example models: FernandezTeruel_2024_capivasertib.R (linear-deviation effect on the LOG-scale magnitude of the sigmoidal time-dependent auto-inhibition of apparent clearance: lcl_time_max * (1 + (DOSE_CAPIVASERTIB_MG - 480) * (-0.00183)). Because the log-scale magnitude is itself negative, the negative slope makes the multiplier fall with dose and the inhibition therefore DEEPEN with dose – the maximal reduction in CL/F is 17% at 400 mg, 21% at 480 mg and 53% at 800 mg, reproducing the 18% / 22% / 54% the paper reports in Sect. 3.5. The mechanism the authors propose is time-dependent auto-inhibition of CYP3A4, of which capivasertib is both a substrate and a weak inhibitor), FernandezTeruel_2025_capivasertib.R (the successor fit of the same model by the same first author, adding the Phase III CAPItello-291 and China PK studies; identical linear-deviation form on the log-scale magnitude with the slope re-estimated to -0.00213 and the reference dose unchanged at 480 mg, lcl_time_max * (1 + (DOSE_CAPIVASERTIB_MG - 480) * (-0.00213)), which reproduces the two reductions that paper prints in Sect. 3.2 – 11.2% at 400 mg and 29.1% against a reported 29.2% at 640 mg. Note the maximal reductions are markedly SHALLOWER than the 2024 fit at the same doses despite a deeper Imax, because the two coefficients moved together).
• Notes: A drug-specific member of the auto-approved DOSE_<drug>_<units> family, required here rather than the general DOSE canonical for the same solve-path reason as DOSE_TBPPI_MG, DOSE_TAK071_MG and DOSE_UTRELOXASTAT_MG: rxode2’s event-table translator (etTrans) consumes a column literally named DOSE and never exposes it to model. Unlike those three, the effect here is on a time-dependent-clearance parameter rather than on absorption or bioavailability, so the column drives the shape of the CL/F time course rather than the extent of absorption; the paper found capivasertib exposure to be dose-proportional over 80-480 mg and more than proportional beyond 480 mg purely as a consequence of this term.

DOSE_GHI_MLKG (canonical for administered guhong injection volume dose per kg body weight)

• Description: Volume dose of guhong injection (GHI), a Chinese herbal compound preparation of N-acetyl-L-glutamine plus an aqueous safflower (Carthamus tinctorius L.) extract, administered as a single intravenous injection and recorded in mL of preparation per kg body weight. The preparation is dosed by volume rather than by the mass of any one constituent, so the volume dose is the natural per-record quantity: the dose of each individual quantified constituent is this column multiplied by that constituent’s content in GHI (Chen 2024 Results 3.2: 30.0 mg/mL N-acetyl-L-glutamine, 1 mg/mL hydroxysafflor yellow A, 12.9 ug/mL chlorogenic acid, 78.3 ug/mL p-coumaric acid, 11.7 ug/mL rutin, 3.4 ug/mL hyperoside, 65.7 ug/mL kaempferol-3-O-rutinoside, 8.7 ug/mL kaempferol-3-O-glucoside).
• Units: mL/kg
• Type: continuous
• Scope: specific
• Reference category: n/a – used as a dose-group selector, not as a normalized continuous term. Chen 2024 fitted an independent two-compartment disposition model and an independent set of sigmoid-Emax parameters in each of its three dose groups (GHI-L 2.5 mL/kg, GHI-M 5 mL/kg, GHI-H 10 mL/kg) rather than estimating a dose-covariate function, so models that carry this column select a parameter set by dose level. The Chen_2024_*_rat.R family assigns GHI-L below 3.75 mL/kg, GHI-M from 3.75 to below 7.5 mL/kg, and GHI-H at or above 7.5 mL/kg – the midpoints of the studied levels. Values outside 2.5-10 mL/kg extrapolate the nearest fitted group and are not supported by the source data.
• Source aliases:
  • GHI dose (mL/kg) – the wording used in Chen 2024 Section 2.3 (“GHI-administered groups were respectively injected with GHI (2.5, 5, and 10 mL/kg, i.v.)”); the paper prints no data-column name.
• Example models: Chen_2024_chlorogenicAcid_rat.R, Chen_2024_hydroxysafflorYellowA_rat.R, Chen_2024_hyperoside_rat.R, Chen_2024_kaempferol3OGlucoside_rat.R, Chen_2024_kaempferol3ORutinoside_rat.R, Chen_2024_nAcetylglutamine_rat.R, Chen_2024_pCoumaricAcid_rat.R, Chen_2024_rutin_rat.R (all eight select V1, V2, CL1 and Q – and, where a PK/PD equation was fitted, Emax, ED50 and gamma – from this column).
• Notes: A drug-specific member of the DOSE_<drug>_<units> family. The general DOSE canonical is unusable here for the mechanical reason documented under DOSE_TBPPI_MG: rxode2’s etTrans consumes a column literally named DOSE and never exposes it to model. Because GHI is a multi-constituent botanical preparation with no single active moiety, the volume dose – not a mass dose – is what the study randomized and what indexes the fitted parameter sets; a per-constituent DOSE_<constituent>_UGKG column would have to be derived separately for each of the eight analytes and would lose the fact that they are dosed together in fixed proportion. Distinct from the general DOSE_HIGH high-dose-cohort indicator, which is binary; this column preserves the actual administered volume so a downstream user can recover each constituent’s dose.

DOSE_TV46000_ML (canonical for administered TV-46000 subcutaneous injection volume)

• Description: Volume of the TV-46000 risperidone long-acting subcutaneous antipsychotic (LASCA) suspension delivered at a given injection, in millilitres of suspension. TV-46000 is a fixed-strength copolymer-based risperidone suspension (approximately 360 mg risperidone per mL), so the injected volume and the milligram dose are proportional and the paper’s own extrinsic-factor list names the covariate “injection volume (dose)”. The volume is nonetheless the quantity the model reads, because the mechanism the covariate encodes is physical: a larger bolus of suspension deposited subcutaneously has a smaller surface-area-to-volume ratio and releases drug more slowly by the direct route. Per-dose-record (a subject who switches dose strength between injections carries a different value on each dose record).
• Units: mL
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as a median-normalized power term on the fast direct-release rate constant, ka1 * (DOSE_TV46000_ML / ref)^e_dose_ka1 (Perlstein 2025 Methods, continuous-covariate equation Pi = P * (COVi / COVmedian)^theta_i). Reference values observed: Perlstein 2025 reports the pooled cohort mean injection volume as 0.303 mL (Table 2, Overall column, 3287 injections) but never prints the median that the covariate equation actually centres on; Perlstein_2025_risperidone_tv46000.R therefore uses 0.303 mL and documents the substitution. Volumes observed in the source programme: 0.035 mL (12.5 mg), 0.070 mL (25 mg), 0.139 mL (50 mg) and 0.278 mL (100 mg) are printed in the paper; the remaining strengths follow the same ~360 mg/mL proportionality (75 mg ~ 0.208 mL, 125 mg ~ 0.347 mL, 150 mg ~ 0.417 mL, 200 mg ~ 0.556 mL, 250 mg ~ 0.695 mL). Must be strictly positive; a zero or NA value makes the power term undefined.
• Source aliases:
  • INJV – the parameter-name stem printed in Perlstein 2025 Table 3 (KA1INJV1, “injection volume effect on KA1”) and the row label “Injection volume, mL, mean (SD)” in Table 2. The paper does not disclose the NONMEM $INPUT column name.
• Example models: Perlstein_2025_risperidone_tv46000.R (power effect on the fast direct-release absorption rate constant: ka1 * (DOSE_TV46000_ML / 0.303)^(-0.384), so a larger injected volume slows the direct release; Perlstein 2025 Table 3, KA1INJV1 = -0.384).
• Notes: A drug-specific member of the DOSE_<drug>_<units> family, in the same vein as DOSE_GHI_MLKG (a volume rather than a mass dose) but recorded as an absolute volume rather than per kilogram, because the TV-46000 model normalizes against a cohort median volume and not against body size. The general DOSE canonical is unusable for the mechanical reason documented under DOSE_TBPPI_MG: rxode2’s etTrans() consumes a column literally named DOSE and never exposes it to model(). Distinct from INJSITE_ARM, which records where the same volume was injected, and from the binary DOSE_HIGH high-dose-cohort indicator.

DOSE_TV44749_MG (canonical for administered TV-44749 subcutaneous injection dose)

• Description: Olanzapine milligram dose delivered at a given TV-44749 injection. TV-44749 is an investigational copolymer-based extended-release subcutaneous olanzapine suspension; the copolymer precipitates on injection and forms a depot that releases olanzapine over a monthly interval. The dose the model reads is the olanzapine mass, not a suspension volume – Perlstein 2026 reports the covariate and every simulated regimen in milligrams, and describes the simulated dose grid as “approximately 0.1 mL increments [35.4 mg] of dosing volume”, i.e. milligrams are the primary unit and volume the derived one. Per-dose-record: a subject who changes dose strength between injections carries a different value on each dose record.
• Units: mg
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as a linear ADDITIVE effect on the second-process Weibull release time, TD1 = TD1_0 + TD1_1 * DOSE (Perlstein 2026 Table 1 footnote: “TD_0 and TD1_0, the parameter values at the dose = 0; TD_1 and TD1_1, dose dependent changes”), so the reference is a nominal zero dose rather than a cohort centring value and no median is needed. In Perlstein_2026_olanzapine_lai.R the release processes are parameterised by their rate scalers, so the effect is applied on the reciprocal: ra2 = 1 / (1 / ra2_0 + e_dose_ra2 * DOSE_TV44749_MG), which is algebraically identical. Doses observed in the source study: 70 and 105 mg (healthy participants, subtherapeutic), 283, 318, 425, 531 and 566 mg (patients); the paper simulated 150-708 mg. Must be supplied on observation records as well as dose records, because it is read inside model.
• Source aliases:
  • DOSE – the covariate is listed simply as “dose” in Perlstein 2026 Methods (Covariate Analysis); the paper does not disclose the NONMEM $INPUT column name. The bare name DOSE is unusable for the mechanical reason documented under DOSE_TBPPI_MG: rxode2’s etTrans consumes a column literally named DOSE and never exposes it to model.
• Example models: Perlstein_2026_olanzapine_lai.R (linear additive effect on the sustained-release phase: TD1 = 323 + 0.0939 * DOSE_TV44749_MG h, so a 531 mg injection releases its second phase over 373 h against 323 h at a nominal zero dose; Perlstein 2026 Table 1, TD1_1 = 0.0939).
• Notes: A drug-specific member of the DOSE_<drug>_<units> family. Sibling to DOSE_TV46000_ML, the corresponding covariate for the same sponsor’s risperidone long-acting subcutaneous product, and a useful contrast: TV-46000’s covariate is registered as a VOLUME because that paper’s effect is physically about the injected bolus geometry and is centred on a cohort median volume, whereas TV-44749’s is a MASS entering an uncentred additive term. Both encode the same underlying idea – a bigger subcutaneous depot releases more slowly – and in both models the effect is statistically significant but judged not clinically relevant by the authors (Perlstein 2026 Discussion: “the rate of absorption only marginally decreased with the increase of the dose, without any effect on the extent of absorption”).

DOSE_UFH_UH (canonical for concomitant continuous-infusion unfractionated heparin dose rate)

• Description: Patient’s concurrent continuous intravenous infusion rate of unfractionated heparin (UFH), in absolute units per hour. Absolute (not weight-normalized) infusion rate is preserved at the column level because the source model parameterizes the UFH-on-clearance effect against a cohort-median absolute rate rather than a per-kg rate. Time-varying per observation as the UFH infusion is titrated. Set to 0 units/h for patients not receiving concurrent UFH.
• Units: units/h (absolute infusion rate; document per-model via covariateData[[DOSE_UFH_UH]]$units if a different unit is reported).
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as an additive linear increment on clearance in the source paper’s final model: CL = CL_base * (WT / 70)^0.75 + e_ufh_cl * (DOSE_UFH_UH / ref). Reference values observed: 173 units/h (Moffett 2017 cohort median absolute UFH infusion rate). The additive form (rather than multiplicative) reflects the observed clinical mechanism – UFH binds to and potentiates antithrombin activity, so higher UFH infusion rates deplete circulating antithrombin at a constant additive rate on top of the baseline weight-scaled clearance.
• Source aliases:
  • UFH – used in Moffett_2017_antithrombin.R (Moffett 2017 Table 3 covariate; the paper’s absolute UFH infusion-rate column in units/h). The cohort mean per-kg UFH dose was 34.1 +/- 22.7 units/kg/h (Moffett 2017 Results and Table 5); the per-record absolute rate is the value that entered the (UFH/173) covariate ratio.
• Example models: Moffett_2017_antithrombin.R (linear-additive effect on AT clearance: CL_dL_h = 0.917 * (WT/70)^0.75 + 0.129 * (DOSE_UFH_UH / 173); reference 173 units/h from Moffett 2017 Table 3).
• Notes: Specific scope until a second antithrombin / anticoagulation-covariate model ratifies the canonical. UFH is a heterogeneous polymeric anticoagulant that acts by potentiating endogenous antithrombin; concurrent UFH therefore accelerates depletion of dosed exogenous antithrombin, motivating its inclusion as a covariate on AT clearance. Future models parameterizing the same covariate on a per-kg basis (units/kg/h) should either add a source alias here with the value transformation DOSE_UFH_UKGH = DOSE_UFH_UH / WT documented, or register a sibling canonical DOSE_UFH_UKGH if the effect coefficient’s meaning changes. Distinct from the AT-related coagulation biomarker AT_BL_UDL (baseline antithrombin activity, used as a covariate on VD in the same Moffett 2017 model). Set to 0 for patients not receiving concurrent UFH; the additive covariate term collapses to zero and the clearance reduces to the baseline weight-scaled arm.

DOSE_PHT_MGKGD (canonical for daily phenytoin dose per kg body weight)

• Description: Patient’s own total daily dose of phenytoin (mg) divided by current body weight (kg), expressed as mg/kg/d. Per-dose-record covariate; constant within an inter-dose interval and updated when the prescriber alters the daily dose.
• Units: mg/kg/d
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as a self-dose-rate regressor in the dose-dependent powder bioavailability formula F_powder = 1 - exp(-theta / DOSE_PHT_MGKGD). Has no effect when paired with FORM_POWDER = 0 (tablet); a non-NA non-zero placeholder must still be supplied.
• Source aliases:
  • Dij – used in Yukawa_1990_phenytoin.R (paper’s per-record daily-dose-per-weight regressor, mg/kg/d, in the powder bioavailability equation 4 of Yukawa 1990).
• Example models: Yukawa_1990_phenytoin.R (powder bioavailability F_powder = 1 - exp(-9.92 / DOSE_PHT_MGKGD); F approaches 1 below ~2 mg/kg/d and decreases monotonically as the daily dose increases, reflecting the lower wettability of the Aleviatin brand phenytoin powder formulation).
• Notes: Specific scope because the value is intrinsically tied to phenytoin (PHT) and the Yukawa 1990 powder-vs-tablet bioavailability contrast. Drug-self-dose covariates for other drugs should register sibling canonicals (e.g., DOSE_<DRUG>_MGKGD) rather than reuse this name – the absolute coefficient (theta_BA2 = 9.92 in Yukawa 1990) is not transferable across drugs. Computed as the total daily dose summed across the 2-3 daily phenytoin doses (mg/d) divided by the patient’s body weight at the dose record (kg).

DOSE_VPA_MGKGD (canonical for daily valproic acid dose per kg body weight)

• Description: Patient’s own total daily dose of valproic acid (mg), summed across the 1-3 daily administrations, divided by the current body weight (kg). Per-dose-record covariate; constant within an inter-dose interval and updated when the prescriber alters the daily dose.
• Units: mg/kg/d
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as a self-dose-rate power regressor on apparent clearance, cl = CLp/F * (DOSE_VPA_MGKGD / 25)^k, with 25 mg/kg/d the rounded cohort mean daily dose. Must be strictly positive; a zero or NA value makes the power term undefined.
• Source aliases:
  • DD – used in Zhang_2023_valproic_acid_exponent.R (Zhang 2023 Eq. 7, the “simple exponent model”; the paper’s abbreviation list defines DD as daily dose in mg/day and DDW as daily dose in mg/kg/day, but Eq. 7’s reference value of 25 and the reported CLp/F of 0.331 L/h against a base-model CL/F of 0.311 L/h only reconcile when DD is read on the mg/kg/day scale – see that model’s vignette Errata).
• Example models: Zhang_2023_valproic_acid_exponent.R (power effect on CL/F with exponent 0.658 and reference 25 mg/kg/d; observed range 8.70-57.69 mg/kg/d in a paediatric epilepsy therapeutic-drug-monitoring cohort).
• Notes: Specific scope because the value is intrinsically tied to valproic acid and to the saturable-plasma-protein-binding non-linearity that makes valproate clearance rise with dose; the exponent is not transferable across drugs. Sibling canonical of DOSE_PHT_MGKGD (same drug-self-daily-dose-per-weight concept for phenytoin) – register a further DOSE_<DRUG>_MGKGD sibling rather than reusing either name. Note the modelling caveat the source authors themselves raise: because the daily dose is the quantity a therapeutic-drug-monitoring model is meant to predict, using it as a clearance covariate is circular, so models built on this covariate reproduce a published comparison rather than recommend a dosing strategy.

DOSE_EMODEPSIDE_MGKGD (canonical for daily emodepside dose per kg body weight)

• Description: Total daily oral dose of emodepside (mg), summed across the once- or twice-daily administrations, divided by body weight (kg). Per-dose-record covariate; constant within a dosing regimen.
• Units: mg/kg/day
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as a linear effect on the absorption mean transit time centred at 0.08 mg/kg/day, mtt = MTT_typ * (1 + 1.05 * (DOSE_EMODEPSIDE_MGKGD - 0.08)).
• Source aliases:
  • DOSE_KG – the Assmus 2022 S1 Code NONMEM column (COV4 = 1+THETA(16)*(DOSE_KG - 0.08)); the paper’s text calls it “dose per kg body weight per day”.
• Example models: Assmus_2022_emodepside.R (+105% in the mean transit time per mg/kg/day; observed range 0.012-0.667 mg/kg/day across 1-40 mg single doses and 5-20 mg/day multiple doses in healthy men).
• Notes: Member of the DOSE_<DRUG>_MGKGD family (siblings DOSE_PHT_MGKGD, DOSE_VPA_MGKGD). Specific scope because the coefficient is intrinsic to emodepside absorption. Twice-daily 10 mg in a 75 kg adult is 20/75 = 0.267 mg/kg/day. The effect is linear and unbounded, so values far above the observed 0.667 mg/kg/day extrapolate the mean transit time linearly.

DOSE_VPA_MGD (canonical for total daily valproic acid dose)

• Description: Patient’s own total daily dose of valproic acid (mg), summed across the 1-3 daily administrations and NOT normalised by body weight. Per-dose-record covariate; constant within an inter-dose interval and updated when the prescriber alters the daily dose.
• Units: mg/d
• Type: continuous
• Scope: general
• Reference category: n/a – the functional form is per-paper and is not a property of this column. In Zhang_2023_valproic_acid_ddemax.R it drives a sigmoid-Emax term on apparent clearance, cl = CLp/F * (1 + Emax * DOSE_VPA_MGD^gamma / (DD50^gamma + DOSE_VPA_MGD^gamma)); in Ogusu_2014_valproic_acid.R it enters as a power of the 1000 mg/d-normalised ratio, (DOSE_VPA_MGD/1000)^theta, on Vd/F, on CL/F and on the slope of a logistic PD model. Must be strictly positive in both.
• Source aliases:
  • DD – used in Zhang_2023_valproic_acid_ddemax.R (Zhang 2023 Eq. 5, the “dose-dependent maximum effect model”).
  • Dose – used in Ogusu_2014_valproic_acid.R (Ogusu 2014 Eqs. 6, 7 and 9, defined in the text below Eq. 8 as “Dose is the daily VPA dose (mg/day)”).
  • VA – used in Damnjanovic_2023_valproic_acid.R (the covariate behind Damnjanovic 2023 Table 2(c) beta_Cl_VA; the table footnote does not define the abbreviation, but the Results narrative for valproic acid identifies the retained clearance covariate as “daily dose on Cl” and the Discussion states that coadministration of the other two study antiepileptics was not significant, so VA is valproate’s own daily dose rather than a comedication flag).
• Example models: Zhang_2023_valproic_acid_ddemax.R (sigmoid-Emax effect on CL/F with Emax 2.8, Hill 1.68 and DD50 37.4 mg/d, all fixed from Ding 2015; observed range 60-1250 mg/d, median 480, in a paediatric epilepsy therapeutic-drug-monitoring cohort), Ogusu_2014_valproic_acid.R (power-of-normalised-dose terms (DOSE_VPA_MGD/1000)^1.52 on Vd/F, (DOSE_VPA_MGD/1000)^0.596 on CL/F and (DOSE_VPA_MGD/1000)^1.55 as the exposure slope of a logistic gamma-GT-elevation model; observed range 100-2600 mg/d, mean 934 mg/d, in a Japanese epilepsy therapeutic-drug-monitoring cohort spanning ages 2.2-52.2 years), Nakashima_2015_valproic_acid.R, Nakashima_2015_valproic_acid_le18y.R, Nakashima_2015_valproic_acid_ge19y.R (power-form effect on BOTH apparent clearance and apparent volume in the fixed upstream PK layer: vc *= (DOSE_VPA_MGD/1000)^1.52 and cl *= (DOSE_VPA_MGD/1000)^0.596, referenced to 1000 mg/d; Ogusu 2014 Eqs 6-7 and Table 2; observed range 50-3200 mg/d, mean 1120, in a Japanese epilepsy cohort on sustained-release valproate), Damnjanovic_2023_valproic_acid.R (exponential-linear effect on CL/F, exp(0.0012 * DOSE_VPA_MGD), untransformed rather than centred because the source reports no cohort mean daily dose).
• Notes: Scope promoted from specific to general on 2026-09-04 alongside the Ogusu 2014 valproic acid extraction – a second independent paper, a different cohort (Japanese, ages 2.2-52.2) and a different functional form (power-of-normalised-dose rather than sigmoid-Emax), demonstrating that the column definition generalizes beyond the Zhang 2023 usage. The mass-scale sibling of DOSE_VPA_MGKGD (same drug, same self-daily-dose concept, per-patient rather than per-kilogram). Both are needed because Zhang 2023 uses the per-kilogram scale in its Eq. 7 simple exponent model and the per-patient scale in its Eq. 5 dose-dependent maximum effect model – do not merge them or convert one into the other inside model(), since the reference values (25 mg/kg/d and 37.4 mg/d respectively) are not interconvertible without a weight. The scale for Eq. 5 is a reading ratified by the maintainers (PMC10587682): Eq. 5’s text defines DD in mg/kg/day and Ding 2015 tabulates DD50 = 37.4 against a mg/kg/day daily dose, but only the mg/day reading reproduces Zhang’s own reported clearance (0.307 L/h against a base-model 0.311 L/h) and its near-zero MDPE; the mg/kg/day reading gives 0.153 L/h and a typical concentration roughly 2.7-fold the observed median. The conflicting evidence – Supplementary Figure S5A favours mg/kg/day – is recorded in that model’s vignette Errata. Never name a dose covariate column bare DOSE: rxode2::etTrans() consumes a column of that name (any casing) before model() sees it.
• Register a further DOSE_<DRUG>_MGD sibling rather than reusing this name for another drug.

DOSE_LTG_MGD (canonical for total daily lamotrigine dose)

• Description: Patient’s own total daily dose of lamotrigine (mg), summed across the day’s administrations (for a twice-daily regimen the value is the sum of both administrations) and NOT normalised by body weight. Per-dose-record covariate; constant within an inter-dose interval and updated when the prescriber alters the daily dose.
• Units: mg/d
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as an exponential-linear term on apparent clearance, cl = CL/F * exp(e_dose_ltg_cl * DOSE_LTG_MGD), applied untransformed rather than centred. Effect coefficient observed: e_dose_ltg_cl = 0.0056 per mg/day in Damnjanovic_2023_lamotrigine.R, i.e. a 0.56% rise in CL/F per mg/day. Must be non-negative; a value of 0 makes the term unity.
• Source aliases:
  • DailyDose – used in Damnjanovic_2023_lamotrigine.R (Damnjanovic 2023 Table 2(b) beta_Cl_DailyDose, whose footnote defines it as the “factor for the relationship between Cl and lamotrigine daily dose”).
• Example models: Damnjanovic_2023_lamotrigine.R (paediatric epilepsy therapeutic-drug-monitoring cohort on dual antiepileptic therapy; the daily-dose term multiplies a binary [[CONMED_VPA]] term on the same clearance).
• Notes: Named-drug member of the auto-approved DOSE_<DRUG>_<UNITS> family (siblings: [[DOSE_VPA_MGD]] for valproate, [[DOSE_TPM_MGD]] for topiramate, [[DOSE_VGB_MGD]] for vigabatrin). Distinct from [[CONMED_LAMOTRIGINE]], the binary indicator for lamotrigine as a comedication in a model of some other drug – DOSE_LTG_MGD is the modelled drug’s own dose level. Distinct also from the rxode2 / nlmixr2 event column amt, which carries the amount of each individual administration; for a b.i.d. regimen amt is half of DOSE_LTG_MGD, and the two must be kept consistent when simulating or the dose-dependence is silently lost. Register a further DOSE_LTG_MGKGD sibling rather than reusing this name if a future source doses lamotrigine on a per-kilogram scale. As with the valproate siblings, a self-daily-dose covariate on clearance is circular in a therapeutic-drug-monitoring setting – the daily dose is the quantity such a model is meant to recommend – so models built on it reproduce a published fit rather than derive a dosing strategy. Never name a dose covariate column bare DOSE: rxode2::etTrans consumes a column of that name (any casing) before model sees it.

DOSE_ESL_MGD (canonical for total daily eslicarbazepine acetate dose)

• Description: Total daily oral dose of the antiepileptic prodrug eslicarbazepine acetate (ESL) in milligrams, supplied as a data column so a dose-level effect can be evaluated inside model independently of the dose records. In the registering models the value is specifically the STARTING daily dose during the first week of the titration period, which is not the same thing as the randomised maintenance dose – patients randomised to 800 or 1,200 mg titrated up from 400 or 800 mg over two weeks.
• Units: mg/day
• Type: continuous
• Scope: specific
• Reference category: 0 mg/day (placebo). Values observed: 0, 400 and 800 mg/day as a starting dose; the randomised maintenance doses in the same trials were 400, 800 and 1,200 mg/day.
• Source aliases:
  • 400 mg_i and 800 mg_i – the two mutually exclusive indicator variables Gidal 2018 Appendix S1 Equations E-3, E-4 and E-5 carry. The canonical column holds the dose LEVEL and the consuming model derives the indicators as (DOSE_ESL_MGD == 400) and (DOSE_ESL_MGD == 800), which keeps one column instead of two and makes the placebo group explicit as the value 0.
• Example models: Gidal_2018_eslicarbazepine_dizziness.R, Gidal_2018_eslicarbazepine_headache.R, Gidal_2018_eslicarbazepine_somnolence.R (the single strongest predictor in all three adverse-event models, with logit shifts versus placebo of 1.97 / 3.62, 1.42 / 2.64 and 2.47 / 4.22 for the 400 and 800 mg starting doses).
• Notes: Member of the DOSE_<drug>_<units> family. The distinction between starting and maintenance dose is load-bearing and must be restated in covariateData[[DOSE_ESL_MGD]]$notes by any model reusing this column: the practical conclusion of the founding paper is that starting at 400 rather than 800 mg improves the risk-benefit profile, which is a statement about the titration schedule and not about the maintenance dose. A model that needs the maintenance dose instead should say so explicitly rather than reusing the column silently.

DOSE_OMEPRAZOLE_MGD (canonical for total daily concomitant omeprazole dose)

• Description: Total daily dose of concomitant omeprazole in milligrams (0 when omeprazole is not co-administered), supplied as a data column so a dose-threshold interaction can be evaluated inside model. Omeprazole inhibits CYP2C19 (and to a lesser extent CYP3A4) in a dose-dependent way, so some models retain an effect only above a dose threshold rather than for any omeprazole use.
• Units: mg/day
• Type: continuous
• Scope: specific
• Reference category: 0 mg/day (no concomitant omeprazole). In the founding model any dose below 40 mg/day is equivalent to the reference.
• Source aliases:
  • OME – Khan-asa 2020 indicator, 1 for omeprazole >= 40 mg/day and 0 for 20 mg/day or less; recovered in model() as DOSE_OMEPRAZOLE_MGD >= 40.
• Example models: KhanAsa_2020_voriconazole.R (fractional effect (1 - 0.306 * ome40) on voriconazole CL/F, where ome40 = 1 when DOSE_OMEPRAZOLE_MGD >= 40).
• Notes: Member of the DOSE_<drug>_<units> family and the dose-resolved counterpart of the binary CONMED_OMEPRAZOLE. Use this column rather than CONMED_OMEPRAZOLE when the paper distinguishes omeprazole doses; in the founding paper omeprazole 20 mg/day was tested and not retained, so a presence indicator would wrongly apply the effect to 20 mg/day patients.

DOSE_CBZ_MGD (canonical for total daily concomitant carbamazepine dose)

• Description: Total daily oral dose of concomitant carbamazepine in milligrams, supplied as a data column so a dose-dependent enzyme-induction effect can be evaluated inside model. Carbamazepine induces the UGT enzymes responsible for glucuronidating several co-administered antiepileptic drugs, so the magnitude of the interaction depends on its dose rather than merely on its presence.
• Units: mg/day
• Type: continuous
• Scope: specific
• Reference category: 0 mg/day (no concomitant carbamazepine). A power term of the form (DOSE_CBZ_MGD / ref)^exponent is then exactly 0, so the increment vanishes and no separate carbamazepine presence indicator is needed alongside it. Reference values observed: 800 mg/day (Gidal 2018, carbamazepine 400 mg twice daily, the modal regimen); the characterised range is 400 mg/day (200 mg BID) to 1,200 mg/day (400 mg TID).
• Source aliases:
  • dose_carbamazepine_j – Gidal 2018 Appendix S1 Equation E-1.
• Example models: Gidal_2018_eslicarbazepine.R (adds 1.08 * (DOSE_CBZ_MGD / 800)^0.411 L/h to eslicarbazepine apparent oral clearance, i.e. 33.4% to 52.5% above the no-carbamazepine value across the characterised dose range, with a corresponding 25.1% to 34.4% fall in AUC at steady state).
• Notes: Member of the DOSE_<drug>_<units> family and the dose-resolved counterpart of the binary CONMED_CBZ. The two are NOT interchangeable and a model may need both – the founding paper uses the dose column in its PK model, where the induction is graded, and the binary indicator in its exposure-response models, where baseline carbamazepine use acts as a pharmacodynamic covariate. Note also that a carbamazepine column of either kind is usually kept OUT of a composite enzyme-inducing-antiepileptic-drug flag such as CONMED_EIAED when it already carries its own term; folding it in would double-count the induction.

DOSE_BATEFENTEROL_UGD (canonical for total daily inhaled batefenterol dose)

• Description: Patient’s TOTAL daily inhaled dose of batefenterol (GSK961081), the bifunctional muscarinic-antagonist / beta2-agonist (MABA), in ug/day, summed across both administrations for the twice-daily regimens and NOT normalised by body weight. Set to 0 ug/day for placebo. Study-arm-level and constant over the treatment period.
• Units: ug/day
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as the dose regressor of a standard Emax term on the day-29 trough FEV1, emax * DOSE_BATEFENTEROL_UGD / (DOSE_BATEFENTEROL_UGD + ed50) with ed50 = 152 ug/day. Must be non-negative; the value 0 makes the term exactly 0, which is the intended placebo behaviour.
• Source aliases:
  • Total Dose – used in Ambery_2015_batefenterol_fev1.R (Ambery 2015 Eq. 1; the x-axis of that paper’s Fig. 3d and Fig. 4).
• Example models: Ambery_2015_batefenterol_fev1.R (dose regressor of the day-29 trough-FEV1 Emax dose-response in moderate-to-severe COPD; the five observed levels are 0, 100, 200, 400 and 800 ug/day, pooling once- and twice-daily arms because Ambery 2015 Sect. 4 found “no apparent influence of dosing regimen on the PD model”).
• Notes: Named-drug member of the auto-approved DOSE_<DRUG>_<UNITS> family; the microgram-scale sibling of the milligram-scale entries [[DOSE_VPA_MGD]], [[DOSE_LTG_MGD]] and [[DOSE_SIPOGLITAZAR_MGD]], and the per-drug counterpart of [[DOSE_IND]] (per-arm inhaled indacaterol dose, the same clinical setting and the same Emax-on-trough-FEV1 role but registered before the <UNITS> suffix convention settled). Microgram units are load-bearing here: an inhaled bronchodilator’s whole dose range sits below 1 mg, so a milligram column would carry fractional values and the published ED50 of 152 would be misread by three orders of magnitude. Distinct from the rxode2 / nlmixr2 event column amt, which carries the amount of each individual administration; for the twice-daily arms amt is half of DOSE_BATEFENTEROL_UGD, and the two must be kept consistent when the companion PK model Ambery_2015_batefenterol.R is simulated alongside this one. Never name a dose covariate column bare DOSE: rxode2::etTrans() consumes a column of that name (any casing) before model() sees it. Scope: specific because the value is intrinsically tied to batefenterol.

DOSE_SIPOGLITAZAR_MGD (canonical for total daily sipoglitazar dose)

• Description: Patient’s own TOTAL daily dose of the PPAR alpha/delta/gamma agonist sipoglitazar, in mg/day, summed across both administrations for the twice-daily regimens and NOT normalised by body weight. Set to 0 mg/day for placebo and active-comparator (rosiglitazone) subjects.
• Units: mg/day
• Type: continuous
• Scope: specific
• Reference category: n/a – enters the steady-state exposure computation AUC [mg*day/L] = DOSE_SIPOGLITAZAR_MGD / (cl * 24), which drives the Emax stimulation of the FPG elimination rate constant KoutG in the indirect-response PD model (DEF = Emax * AUC / (AUC50 + AUC), AUC50 = 1.15 mg*day/L).
• Source aliases:
  • dose – Stringer 2014 Methods The individual PK parameters derived from a previous analysis were used to calculate individual exposure, AUC (AUC = dose/CL) over the dose interval at steady state. Studied regimens are 8 mg QD, 16 mg QD, 16 mg BID, 32 mg QD, 32 mg BID and 64 mg QD, i.e. total daily doses of 8, 16, 32, 32, 64 and 64 mg/day; the genotype-based dose-assignment designs of Table 1 additionally simulate 96 mg/day.
• Example models: Stringer_2014_sipoglitazar.R (drives Emax * AUC / (AUC50 + AUC) on KoutG, with the per-genotype clearance selected by UGT2B15_STAR2_HET / UGT2B15_STAR2_HOM / UGT2B15_MISSING).
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family and specifically the DOSE_VPA_MGD / DOSE_MPA_MGD total-daily-dose siblings; register a further DOSE_<DRUG>_MGD sibling rather than reusing this name for another drug. The TOTAL-daily-dose basis (rather than a per-administration amount) is fixed by the paper’s own definition of the exposure driver – AUC50 is the AUC0-24h at steady state achieving half the maximal response – and is confirmed numerically, since it is the only basis under which the Stringer 2014 Figure 2a steady-state FPG asymptotes and the reported 6-month FPG changes are reproduced. Stringer 2014 tested but did not retain a separate AUC50 between the QD and BID regimens, so a 32 mg BID subject and a 64 mg QD subject carry identical drug effect at identical clearance; the 32 mg BID arm is nonetheless distinguishable through the TRT cohort indicator, which selects that arm’s separately estimated FPG baseline. Never name a dose covariate column bare DOSE: rxode2::etTrans() consumes a column of that name (any casing) before model() sees it.

DOSE_MPA_MGD (canonical for total daily mycophenolic acid dose)

• Description: Patient’s own total daily dose of mycophenolic acid (MPA), summed across the two daily administrations and NOT normalised by body weight. Expressed on a single MPA-equivalent mass scale so that enteric-coated mycophenolate sodium (EC-MPS) and mycophenolate mofetil (MMF) daily doses are directly comparable. Per-subject / per-occasion covariate; constant within a dosing interval and updated when the prescriber alters the daily dose.
• Units: mg/d
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as a power term on apparent clearance normalised to a cohort-mean daily dose, cl = CL/F * (DOSE_MPA_MGD / 1500)^e_dose_mpa_mgd_cl. Must be strictly positive.
• Source aliases:
  • TDD (total daily dose, mg) – used in Tsyplakova_2025_mycophenolateSodium.R and Tsyplakova_2025_mycophenolateMofetil.R (Tsyplakova 2025 Eq. 8 and Eq. 11; both models normalise to the same 1500 mg/day constant).
• Example models: Tsyplakova_2025_mycophenolateSodium.R (power effect on CL/F with exponent 0.77 against a 1500 mg/day reference; EC-MPS cohort), Tsyplakova_2025_mycophenolateMofetil.R (same covariate, exponent 1.27 against the same 1500 mg/day reference; MMF cohort).
• Notes: A DOSE_<DRUG>_MGD sibling of DOSE_VPA_MGD, registered rather than reusing that entry per its own instruction. The effect direction is POSITIVE (higher daily dose associates with higher apparent clearance), which in a stable, therapeutically-monitored transplant cohort is an indication-by-confounding artefact rather than a mechanistic dose-dependency: Tsyplakova 2025 Discussion states that “patients receiving higher total daily doses may have required these regimens to compensate for greater clearance”, and additionally notes the covariate may partly proxy for body weight, which was unavailable in that dataset. Treat the coefficient as descriptive of the source cohort, not as a causal dose-dependency to extrapolate outside the observed dose range. The MPA-equivalent mass scale matters: Tsyplakova 2025 converted MMF doses with a 0.72 factor (MMF 500-2000 mg/day maps onto EC-MPS 360-1440 mg/day) so that both formulations share one covariate scale, and both of its models normalise to the same 1500 mg/day constant – when populating this column for a new MPA model, state the formulation basis in covariateData[[DOSE_MPA_MGD]]$notes. Distinct from DOSE (the generic administered-dose-level column) and from the per-administration amount carried on the event table’s amt: this is the summed 24-hour dose used as a typical-value covariate. Never name a dose covariate column bare DOSE: rxode2::etTrans consumes a column of that name (any casing) before model sees it.

DOSE_MMF_MGKG (canonical for mycophenolate mofetil dose per administration per kilogram of body weight)

• Description: The mycophenolate mofetil (MMF) dose given at a single administration, divided by body weight, on the MMF mass scale (NOT converted to MPA-equivalent mass). Per-dose-record covariate; constant within a dosing interval and updated when the prescriber changes the dose.
• Units: mg/kg
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as a power term on MPA apparent clearance normalised to the cohort median, cl = CL/F * (DOSE_MMF_MGKG / 11.16)^e_dose_cl in Wei_2022_mycophenolic_acid.R (Wei 2022 Eq. 1, e_dose_cl = 0.452). Must be strictly positive.
• Source aliases:
  • DOSE – Wei 2022 Eq. 1 (“DOSE was the MMFdt administered dose”). Same orientation; the paper prints no units, and mg/kg per administration is read from the normaliser 11.16 matching the Table 1 median of 11.2 mg/kg/dose.
• Example models: Wei_2022_mycophenolic_acid.R (pediatric liver transplant, q12h dispersible tablets).
• Notes: A well-formed member of the auto-approved DOSE_<DRUG>_<UNITS> family, distinct from DOSE_MPA_MGD (total daily dose, MPA-equivalent mass, not weight-normalised): a q12h MMF regimen of x mg/kg per dose in a WT kg patient gives DOSE_MPA_MGD = 2 * 0.739 * x * WT only if the MPA-equivalent factor is applied, so do not interconvert without stating that factor. As with DOSE_MPA_MGD, the positive dose effect on CL/F in a therapeutically monitored cohort is descriptive of that cohort (Wei 2022 Discussion attributes it to saturable absorption or to prescribers giving larger doses to faster clearers), not a causal dose-dependency to extrapolate beyond the observed 8.9-61.5 mg/kg range. Never name a dose covariate column bare DOSE: rxode2::etTrans() consumes a column of that name (any casing) before model() sees it.

DOSE_DIGOXIN_MGD (canonical for total daily oral digoxin dose)

• Description: The patient’s own total daily maintenance dose of oral digoxin, i.e. the administered amount divided by the dosing interval, NOT normalised by body weight. A DOSE_<DRUG>_MGD sibling of DOSE_VPA_MGD, registered rather than reusing that entry per its own instruction. Unlike most members of the family, this column carries the modelled drug’s OWN dose as a model input rather than a co-medication’s dose intensity, because the founding model has no compartment to dose into (see Notes).
• Units: mg/day
• Type: continuous
• Scope: specific
• Reference category: n/a – not centred or normalised. In Komatsu_2015_digoxin.R it enters the algebraic steady-state identity directly as the numerator, Cc = 1000 * DOSE_DIGOXIN_MGD / (24 * cl). Must be strictly positive.
• Source aliases:
  • Dosage – Komatsu 2015 Table 1 row group, which tabulates the five observed regimens by number of observations rather than as a named data column.
  • D / D_ij – the symbol in Komatsu 2015 Methods (Pharmacokinetics model) for the administered amount in the steady-state identity Css_ij = D_ij / (CL_ij * tau_ij). Divide by the paper’s tau_ij to obtain the canonical daily dose.
• Example models: Komatsu_2015_digoxin.R (founding example; observed regimens 0.125 mg every 3 days = 0.04167 mg/d, 0.125 mg every 2 days = 0.0625 mg/d, 0.0625 mg/d, 0.125 mg/d and 0.25 mg/d across 287 trough observations in 192 Japanese cardiology patients, with 0.1875 mg/d additionally appearing in the paper’s dosing nomogram).
• Notes: Collapsing the administered amount D and the dosing interval tau into a single daily-dose column is algebraically identical to carrying them separately, because a steady-state model can only ever use them as the ratio D/tau; carrying one column avoids a second register entry for an interval that is never used on its own. The reason the dose is a covariate column at all, rather than an amt on an event record, is structural: Komatsu 2015 fit only steady-state trough samples and therefore fit no absorption or disposition process, so the model has no ODE state and no dosing event – the dose is an algebraic input. A digoxin model that DOES carry disposition structure (Zhou_2010_digoxin.R, Jelliffe_2014_digoxin.R, Gu_2025_digoxin_pbpk.R) must dose through the event table’s amt instead and must not use this column. Never name a dose covariate column bare DOSE: rxode2::etTrans consumes a column of that name (any casing) before model sees it. Promote to general if a second paper ratifies.

DOSE_LITHIUM_CARBONATE_MGD (canonical for total daily lithium carbonate dose)

• Description: The patient’s own total daily dose of oral lithium carbonate, in mg of the SALT (Li2CO3) per day, summed over the daily administrations and NOT normalised by body weight. A DOSE_<DRUG>_MGD sibling of DOSE_VPA_MGD. The salt is spelled out in the name because lithium models dose in mmol of lithium ion while the clinical record (and the covariate equation) uses mg of lithium carbonate: 1 mg Li2CO3 = 2 / 73.89 = 0.02707 mmol Li, so a 600 mg/day regimen is 16.24 mmol Li/day. The column carries the salt mass; the event-table amt carries the converted lithium-ion amount, and the two must describe the same regimen.
• Units: mg/d
• Type: continuous
• Scope: specific
• Reference category: n/a – the functional form is per-paper. In Jin_2022_lithium.R it enters CL/F as the median-centred power (DOSE_LITHIUM_CARBONATE_MGD / 600)^0.354. Must be strictly positive.
• Source aliases:
  • TDD – ‘total daily dose’, Jin 2022 Methods 2.1, Table 1 and Equation 8.
  • DD – ‘DD on CL’, the row label of the supplementary Word copy of Jin 2022 Table 2 (the typeset Table 2 prints ‘TDD on CL’).
• Example models: Jin_2022_lithium.R (founding example; observed range 150-1500 mg/day, median 600, mean 720 +/- 236, in 268 Chinese patients with bipolar disorder sampled at trough).
• Notes: Do not reuse for a dose expressed in mmol of lithium or mEq (1 mmol Li = 1 mEq): register a DOSE_LITHIUM_MMOLD sibling instead, since the coefficient of a centred power is not affected by the unit but the centring value is. Lithium citrate doses are also not interchangeable with this column. Never name a dose covariate column bare DOSE (see DOSE_DIGOXIN_MGD). Promote to general if a second paper ratifies.

DOSE_RBV_MGD (canonical for total daily oral ribavirin dose)

• Description: Total daily dose of concomitant oral ribavirin, summed across the day’s administrations and NOT normalised by body weight. Ribavirin is the standard-of-care backbone in chronic-hepatitis-C regimens, so this column records the co-medication’s dose intensity rather than the modelled drug’s own dose. A DOSE_<DRUG>_MGD sibling of DOSE_VPA_MGD and DOSE_MPA_MGD, registered rather than reusing those per their own instruction.
• Units: mg/day
• Type: continuous
• Scope: general
• Reference category: n/a – enters as a power term on a normalised daily dose. Reference value observed: 1200 mg/day (Riggs 2012, the standard adult chronic-HCV daily dose and the reference individual’s value).
• Source aliases:
  • RIBA (# of 200 mg tabs/day) – Riggs 2012 expresses the covariate as the number of 200 mg tablets per day, normalised to 6 tablets/day; multiply the tablet count by 200 to populate this column in mg/day.
• Example models: Riggs_2012_albinterferon.R (power effect on apparent clearance with exponent 0.0436 against a 1200 mg/day reference, per Table I row “RBV present”. The effect is a SPLIT term rather than a plain power function: when the daily dose is non-zero the multiplier is (DOSE_RBV_MGD / 1200)^0.0436, and when it is zero the multiplier switches to a separate estimated constant 1.1, per Table I row “RBV not present”. A plain power term cannot express this because 0^0.0436 is 0 rather than the estimated no-ribavirin factor, so the model carries an explicit indicator).
• Notes: Distinct from CSS_RBV, which is an individual post-hoc ribavirin steady-state plasma CONCENTRATION from an upstream PK fit rather than an administered dose, and from K_RBV. Also distinct from the generic DOSE column, which rxode2::etTrans consumes before model sees it – never name a dose covariate bare DOSE. Papers that report ribavirin as a weight-banded regimen (e.g. 1000 mg/day below 75 kg, 1200 mg/day at or above 75 kg) should populate the per-subject administered daily dose, not the band label, and record the banding rule in covariateData[[DOSE_RBV_MGD]]$notes. When a source reports ribavirin only as present/absent with no dose, use a CONMED_RIBAVIRIN binary instead of forcing a nominal dose into this column.

PRED_DOSE (canonical for concomitant oral prednisolone daily dose)

• Description: Concomitant oral prednisolone (or prednisolone-equivalent glucocorticoid) daily dose. Time-varying across the dosing period as the post-transplant conmed_steroid taper progresses.
• Units: mg/day. Document per-model via covariateData[[PRED_DOSE]]$units when a paper uses a different unit (mg/kg/day) or a different glucocorticoid (methylprednisolone, dexamethasone, hydrocortisone) – in the latter case convert to prednisolone-equivalent mg/day before populating the column and record the conversion factor in covariateData[[PRED_DOSE]]$notes.
• Type: continuous
• Scope: general
• Reference category: n/a – continuous, with 0 mg/day (no prednisolone) the natural reference value. Effect forms in the registered example models: sigmoid-Emax fractional reduction (1 - Pred_max * PRED_DOSE / (Pred_50 + PRED_DOSE)) on bioavailability (Storset 2014, Pred_max = 0.67, Pred_50 = 35 mg/day, Hill = 1); threshold-form binary multiplier (1 + e * (PRED_DOSE >= 20)) on intrinsic clearance (ter Heine 2018, e = 0.31 for the >= 20 mg/day high-dose contrast). Document the per-model functional form in covariateData[[PRED_DOSE]]$notes.
• Source aliases:
  • Prednisolone dose – used in Storset_2014_tacrolimus.R (mg/day).
  • Prednisolone dose (total daily dose, mg/day) – used in TerHeine_2018_everolimus.R (mg/day; collapsed to a binary high-dose indicator at the >= 20 mg/day threshold inside model).
• Example models: Storset_2014_tacrolimus.R (Emax-style fractional reduction in tacrolimus oral bioavailability via prednisolone-driven induction of intestinal CYP3A / P-glycoprotein; Storset 2014 Methods Equations 4 + 6 with Hill = 1), TerHeine_2018_everolimus.R (threshold-form binary high-dose indicator at >= 20 mg/day driving a multiplicative +31% increase in apparent intrinsic clearance for everolimus via prednisolone-driven CYP3A4 induction; ter Heine 2018 Table 2 ‘Final model’).
• Notes: Distinct from PRED_CMAX_FREE (free prednisolone Cmax co-medication exposure) – PRED_CMAX_FREE is the modelled-from-data peak free concentration, whereas PRED_DOSE is the administered daily-dose level supplied directly from the dosing record. Both can coexist in a future model that simultaneously tests dose-driven and exposure-driven effects of prednisolone on a tacrolimus PK parameter. Distinct from CONMED_STEROID (binary baseline / concomitant corticosteroid use indicator) and PRICORT (binary prior corticosteroid use indicator) – PRED_DOSE carries the daily dose value, not just an on / off flag. Time-varying because tacrolimus / everolimus PK depend on the conmed_steroid dose at the time of each observation; the conmed_steroid taper schedule must be supplied as a per-time-row covariate column. The corresponding methylprednisolone single-dose induction-bolus indicator (Storset 2014 binary covariate, not retained in the final model) would warrant a separate canonical (e.g. MPRED_BOLUS) if a future model retains it. Threshold-form binary indicators (e.g. PRED_DOSE >= 20 mg/day in ter Heine 2018) are derived inside model from the continuous PRED_DOSE column rather than as separate registered canonicals so the underlying continuous dose value remains available for sensitivity analyses. Scope promoted from specific to general on 2026-05-24 with the ter Heine 2018 everolimus extraction (second model ratifying the canonical, this time as a CYP3A4-induction covariate on intrinsic clearance rather than a CYP3A / P-gp-induction covariate on oral bioavailability).

PRED_CMAX_FREE (canonical for free prednisolone Cmax co-medication exposure)

• Description: Maximum free (unbound, ultrafiltrable) plasma prednisolone concentration over a co-medication dosing interval, used as a co-medication-exposure covariate on a primary modelled drug (e.g., tacrolimus) when concomitant prednisolone is suspected to alter the primary drug’s PK via membrane-permeability or fluid-balance effects. Time-fixed per subject in the source paper (one Cmax value per subject, derived from limited-sampling concentrations).
• Units: nmol/L
• Type: continuous
• Scope: specific
• Reference category: n/a – used in centred-deviation form (1 + e * (PRED_CMAX_FREE - ref)). Reference value observed: 155.5 nmol/L (Bergmann 2014 study median).
• Source aliases:
  • PredCmax,free / PREDCFR – used in Bergmann_2014_tacrolimus.R (Bergmann 2014 Table 2 footnote; 162 nmol/L median per Table 1, 155.5 nmol/L centring value per Table 2 equation).
• Example models: Bergmann_2014_tacrolimus.R (linear deviation effect on tacrolimus apparent central volume V1/F: every 1 nmol/L increase from 155.5 nmol/L decreases V1/F by 0.28%).
• Notes: Specific scope because the value is intrinsically tied to a particular co-medication (oral prednisolone) and a particular reference population (kidney transplant recipients on a tapering conmed_steroid regimen). Future popPK models that test free prednisolone Cmax as a covariate on a different primary drug should reuse this canonical; if a future model uses a different prednisolone exposure metric (e.g., PRED_AUC0_12_FREE for free AUC 0-12 h, or PRED_CMAX_TOTAL for total Cmax), register parallel canonicals rather than overload PRED_CMAX_FREE. The source-paper Cmax is derived from limited-sampling concentrations at 1 / 2 / 4 hours postdose per Bergmann 2014 Methods (validated against earlier full-profile data from the same cohort). Distinct from CAV (average concentration of the modelled drug) and CP_MGL (instantaneous concentration of the modelled drug as a time-varying PD driver) – PRED_CMAX_FREE is the maximum concentration of a co-medication, used as a per-subject covariate.

CU_ASP8232 (canonical for time-varying unbound ASP8232 plasma concentration input to a PD-only exposure-response model)

• Description: Time-varying unbound (not bound to soluble or membrane-bound VAP-1) plasma concentration of the vascular adhesion protein-1 (VAP-1) inhibitor ASP8232 (Cu), used as the exposure driver for a PD-only exposure-response model whose PK layer is defined in a separate paper. Per-observation covariate; supplied by the user in the event table because the modelling paper (Hoefman 2021, PD/ER) does not embed a PK compartment. The exposure time-course is produced by the companion multi-target TMDD PK-PD paper (Snelder 2021, doi:10.1007/s10928-020-09717-w) and consumed by the PD-layer paper (Hoefman 2021, doi:10.1007/s10928-020-09716-x).
• Units: nM
• Type: continuous
• Scope: specific
• Reference category: n/a – 0 nM for placebo subjects at every record; the paper reports Cu at steady state of 125.58 nM for 40 mg qd oral ASP8232 in a typical DKD subject (Hoefman 2021 Methods Simulations).
• Source aliases:
  • Cu – Hoefman 2021 Methods Sampling-and-measurements subsection and Table 1 caption: Cu model-predicted unbound ASP8232 plasma concentration in nM; identical orientation, no value transformation.
• Example models: Hoefman_2021_asp8232.R (drives three drug-effect terms: (a) acute additive eGFR CysC decline theta_4 * Cu; (b) sigmoid Imax albuminuria-lowering on AER via logCu = log(1 + 1000 * Cu), theta_9 * logCu^theta_24 / (theta_23^theta_24 + logCu^theta_24) with theta_24 = Hill = 10 FIXED; (c) proportional Emax creatinine-transporter-inhibition on sCr theta_13 * Cu / (theta_31 + Cu) with theta_31 = 52.9 nM FIXED from a separate multi-trial PopPK model per Discussion / “data on file”).
• Notes: Specific scope because the value is intrinsically tied to a particular drug (ASP8232) and a particular upstream PK model; the concept “time-varying unbound driver-drug concentration supplied as external data because the modelling paper does not embed a PK layer” is not currently generalizable across drugs. Future PD-only exposure-response models for other drugs that inherit their PK from a companion paper should register sibling canonicals (e.g., CU_<DRUG>) rather than reuse this name. The log(1 + 1000 * Cu) linearization stated in Table 1’s caption is a nM-to-per-unit unit-shift plus a log(1 + x) transform for numerical stability at Cu = 0 – the 1 +... is what allows the transform to remain finite in the placebo condition. Distinct from CAV (period-averaged concentration of the modelled drug from an in-file PK model) and PRED_CMAX_FREE (co-medication free concentration): CU_ASP8232 is the concentration of the modelled drug, but supplied externally rather than computed in the model file.

CU_DABIGATRAN_UM (canonical for time-varying unbound sum dabigatran plasma concentration as a PD driver)

• Description: Time-varying unbound sum dabigatran plasma concentration: unbound dabigatran plus unbound dabigatran acyl-glucuronide, i.e. the pharmacologically active moiety bound neither to plasma proteins nor to the antidote idarucizumab. Per-observation covariate supplied by the user because the source’s PK layer (a PK-Sim / MoBi whole-body PBPK model) is not reproducible in nlmixr2lib.
• Units: umol/L (dabigatran MW 471.5 g/mol, so 1 umol/L = 471.5 ng/mL; divide an unbound concentration in ng/mL by 471.5).
• Type: continuous
• Scope: specific
• Reference category: n/a – 0 when no dabigatran is present (every coagulation time then returns its baseline). The founding fit spans 0 to about 0.47 umol/L (0 to about 220 ng/mL; Fuhr 2020 Figure 6).
• Source aliases:
  • [DAB] / DABunb sum – Fuhr 2020 ESM Equations 3 and 16-19 (‘[DAB] = unbound sum dabigatran plasma concentrations’; Equation 3 gives the unit as umol/L); identical orientation, no value transformation.
• Example models: Fuhr_2020_dabigatran.R (drives aPTT and TT as combined linear plus Emax functions and dTT and ECT as linear functions, ESM Equations 16-19).
• Notes: Specific scope; dabigatran-specific. A family-conforming sibling of CU_ASP8232, whose Notes prescribe a drug-specific CU_<DRUG> canonical for this pattern; the _UM unit suffix follows the CP_<DRUG>_<UNITS> convention (CP_RIF_UM, CP_GDC_UM). Unbound, not total: dabigatran plasma protein binding is about 35% (fu about 0.65, Fuhr 2020 Introduction), so supplying a total concentration overstates the driver by about 1.5-fold. Sum of parent and glucuronide: the assays in the founding studies hydrolysed the acyl-glucuronide before measurement, and the glucuronide (10-24% of total dabigatran) is assumed equipotent. Idarucizumab-bound drug is excluded, so after an idarucizumab dose this column should fall to near zero. Distinct from the total-concentration Cc output of the dabigatran PK models in the registry (Liesenfeld_2013_dabigatran.R, Yang_2024_dabigatran.R, in ng/mL). Founding example: Fuhr_2020_dabigatran.R.

AUC_CARBO (canonical for per-cycle average AUC of carboplatin)

• Description: Per-cycle average AUC of carboplatin used as the time-varying drug-exposure covariate driving cytotoxic tumour-death rates in tumour-size-dynamics models of platinum-based chemotherapy. The value is held step-wise constant over each chemotherapy cycle and resets at the start of the next cycle.
• Units: carboplatin AUC units (typically mg*min/mL); document per-model via covariateData[[AUC_CARBO]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – set to 0 in cycles where carboplatin is not administered (e.g., post-discontinuation or non-platinum arms).
• Source aliases:
  • CB (NONMEM $INPUT column in DDMODEL00000217 / DDMODEL00000218) – used in Zecchin_2016_tumorovarian.R and Zecchin_2016_survival.R. The DDMORE bundles ship the simulated datasets with the column re-labelled AUC0; downstream consumers should map AUC0 -> AUC_CARBO.
• Example models: Zecchin_2016_tumorovarian.R (Zecchin 2016 SLD model for advanced ovarian cancer, DDMODEL00000217), Zecchin_2016_survival.R (Zecchin 2016 OS model, DDMODEL00000218; the OS model integrates the same SLD ODE inline, with the prior IPP-fit subject-level KG/KD0/KD1/IBASE supplied via the dataset).
• Notes: Specific scope because the column meaning is tied to a particular cytotoxic agent (carboplatin) and a particular per-cycle averaging convention. Reusing the same column for another platinum analogue (cisplatin, oxaliplatin) is not appropriate – register a sibling canonical (AUC_CISPLATIN, AUC_OXALIPLATIN) when needed. The Zecchin 2016 SLD and OS models use the value directly in the death-rate term kd0 * AUC_CARBO * tumorSize, with an internal /1000 numerical scaling carried verbatim from the source $DES block.

AUC_APREMILAST (canonical for steady-state dosing-interval AUC of apremilast)

• Description: Per-subject (time-fixed) area under the apremilast plasma concentration-time curve over the 12 h twice-daily dosing interval at steady state (AUCtau,ss), predicted from a population PK model and supplied as a static exposure column to a sequentially fitted exposure-response model that carries no PK compartments.
• Units: ng*h/mL. Document per-model via covariateData[[AUC_APREMILAST]]$units. Must be on the same scale as the auc50 the consuming model carries.
• Type: continuous
• Scope: specific
• Reference category: n/a – enters through an Emax term emax * AUC / (auc50 + AUC). Set to 0 for placebo; the drug term then vanishes exactly. Okubo 2021 median week-16 values: 1065.4 (10 mg BID), 2030.6 (20 mg BID) and 3169.2 (30 mg BID) ng*h/mL.
• Source aliases:
  • AUCtau,ss / AUCss – Okubo 2021 notation (Methods 2.5, Results 3.3, Table 5).
• Example models: Okubo_2021_apremilast_pasi75.R, Okubo_2021_apremilast_pasi50.R, Okubo_2021_apremilast_spga.R (longitudinal logistic PASI-75 / PASI-50 / sPGA exposure-response models in moderate to severe plaque psoriasis).
• Notes: Member of the AUC_<DRUG> family. Compute it from modellib('Okubo_2021_apremilast') as individual dose / cl (dose in mg times 1000 for ng*h/mL with CL in L/h), or by integrating Cc over one 12 h interval at steady state. A once-daily-equivalent (24 h) apremilast AUC is twice this value and should not be supplied under this name.

AUC_ATEZO (canonical for Cycle-1 atezolizumab serum AUC over the day 0 to day 21 dosing interval)

• Description: Per-subject, time-fixed model-predicted atezolizumab serum AUC from day 0 to day 21 following the first dose (Cycle 1), used as the exposure regressor in exposure-response analyses. Not an observed quantity: it is derived from the individual empirical-Bayes parameters of a population PK model, which is precisely why it exists as a canonical column – Chan 2025 used it as a co-primary endpoint of a registrational phase III study, the first population-PK-derived measure to serve that role in a pivotal PK non-inferiority trial. The 21-day window is the Q3W dosing interval; do not substitute a steady-state AUC (AUCss) or an AUC over a different interval, because the exposure-response coefficients are calibrated to the Cycle-1 value and the two differ by roughly a factor of two under Q3W accumulation.
• Units: ug*day/mL (equivalently mgday/L). Document per-model via covariateData[[AUC_ATEZO]]$units. Note that the source exposure-response tables rescale it by 1000 when reporting coefficients (“Increase by 1000 ugday/mL”); that rescaling belongs in model(), not in the column.
• Type: continuous
• Scope: specific
• Reference category: n/a – a continuous exposure regressor entering linearly and uncentred, so the companion model intercept is the value at AUC_ATEZO = 0, outside the observed range. Reference distribution (Chan 2025 Table S5B, subcutaneous 1875 mg Q3W, N = 246): geometric mean 2907 ug*day/mL (geoCV 35.9 percent), median 2974, range 666-6572. The intravenous 1200 mg Q3W comparator arm (N = 122) gives geometric mean 3328 (geoCV 19.4 percent), i.e. a subcutaneous-to-intravenous geometric mean ratio of 0.87 (90 percent CI 0.83-0.92).
• Source aliases:
  • AUC0-21d, AUC 0-21d – printed name in Chan 2025 Tables S10C, S10F and S10G.
• Example models: Chan_2025_atezolizumab_aeg35.R (logistic, grade 3-5 adverse events, log-odds 0.0879 per 1000 ugday/mL), Chan_2025_atezolizumab_isr.R (logistic, injection site reactions, log-odds 0.581 per 1000 ugday/mL), Chan_2025_atezolizumab_os.R (Cox relative hazard for overall survival, log-hazard 0.134 per 1000 ugday/mL), Chan_2025_atezolizumab.R, Shemesh_2020_atezolizumab_orr_auc.R, Shemesh_2020_atezolizumab_aeg34_auc.R, Shemesh_2020_atezolizumab_aesi_auc.R (logistic exposure-response models for objective response, grade 3/4 adverse events and any-grade adverse events of special interest in high-tumour-mutational-burden patients on intravenous atezolizumab 1200 mg Q3W; the Cycle-1 AUC is an empirical-Bayes prediction from the Stroh 2017 intravenous popPK model – the fixed intravenous layer of Chan_2025_atezolizumab.R – and enters LINEARLY and uncentred with slopes of -0.0924, -0.0649 and 0.299 per 1000 ugday/mL, all non-significant; the coefficients were recovered from the plotted fitted curves because Shemesh 2020 prints none).
• Notes: Specific scope because the column meaning is tied to a particular drug and a particular Cycle-1 21-day averaging window. Sibling exposure metrics from the same analysis use the general canonicals [[CMAX]] and [[CTROUGH]] rather than drug-suffixed columns, because those two are already registered as generic user-supplied exposure regressors; AUC_ATEZO is drug-suffixed instead because it follows the established AUC_<DRUG> family (AUC_ADU, AUC_LEC, AUC_LEN, AUC_IBRU, …) and there is no generic bare AUC canonical. Chan 2025 selected, per endpoint, whichever exposure metric gave the lowest p-value, so a single paper legitimately carries all three columns across its companion models; every one of the resulting exposure coefficients was non-significant.

AUC_BAST_FW (canonical for first-week AUC in the BAST PTTE 2017 teaching dataset)

• Description: Area under the plasma concentration-time curve over the first week of treatment for the unspecified hypothetical drug used in the BAST PTTE 2017 teaching guiding-document (DDMODEL00000243). Per-subject, time-fixed (a single early-exposure summary value carried as a baseline covariate). Used to drive the Event 2 Gompertz hazard via centred-deviation form exp(coef * (AUC_BAST_FW - 3065.5)).
• Units: ug*h/L (per BAST PTTE guiding document Section 2.2.2)
• Type: continuous
• Scope: specific
• Reference category: n/a – used in centred-deviation form. Reference value observed: 3065.5 ug*h/L (BAST PTTE 2017 simulated cohort median; runEV2_105 base hazard).
• Source aliases:
  • AUC – verbatim NM-TRAN $INPUT column name in DDMODEL00000243’s executable.mod files and in the bundle’s Simulated_event_data.csv. Renamed to AUC_BAST_FW in the canonical register so that future models using a generic AUC column with different semantics will not silently collide.
• Example models: NA_NA_tte_gompertz_ev2.R (BAST PTTE 2017 / DDMODEL00000243 Event 2 hazard model; centred at 3065.5 ug*h/L, coefficient 3.09e-4 on the NONMEM rescaled scale, equivalent to exp(3.09e-4 * (AUC_BAST_FW - 3065.5)) on the hazard).
• Notes: Specific scope because the value is intrinsically tied to an unspecified hypothetical drug and a fictional simulated population in the BAST PTTE 2017 guiding document – the column has no shared meaning across drugs or studies. Future models should register a sibling canonical (e.g., AUC_<DRUG>) with explicit drug semantics rather than overload this name. The BAST guiding document Section 2.2.2 defines this as “AUC of drug treatment given within the first week (ug*h/L).”

AUC_GEM (canonical for per-cycle average AUC of gemcitabine)

• Description: Per-cycle average AUC of gemcitabine (sum of parent and active metabolite exposure, per Zecchin 2016 Methods) used as the time-varying drug-exposure covariate driving cytotoxic tumour-death rates in tumour-size-dynamics models of gemcitabine-containing chemotherapy.
• Units: gemcitabine AUC units (typically mg*h/L or the paper’s mol*day / 10^6 cells scaling for the parent-plus-active-metabolite composite); document per-model via covariateData[[AUC_GEM]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – set to 0 in cycles where gemcitabine is not administered (e.g., carboplatin-monotherapy arms).
• Source aliases:
  • G (NONMEM $INPUT column in DDMODEL00000217 / DDMODEL00000218) – used in Zecchin_2016_tumorovarian.R and Zecchin_2016_survival.R. The DDMORE bundles ship the simulated datasets with the column re-labelled AUC1; downstream consumers should map AUC1 -> AUC_GEM.
• Example models: Zecchin_2016_tumorovarian.R (Zecchin 2016 SLD model for advanced ovarian cancer, DDMODEL00000217), Zecchin_2016_survival.R (Zecchin 2016 OS model, DDMODEL00000218).
• Notes: Specific scope. The Zecchin 2016 SLD and OS models use the value directly in the death-rate term kd1 * AUC_GEM * tumorSize, with an internal /100 numerical scaling carried verbatim from the source $DES block.

AUC_ELA (canonical for steady-state dosing-interval AUC of elafibranor)

• Description: Per-subject (time-fixed) area under the elafibranor plasma concentration-time curve over the once-daily dosing interval at steady state, supplied as a static drug-exposure column to the sequentially fitted exposure-response model, which carries no PK compartments. Ooi 2026 Supplementary Datafile S6 $PK computes it as F * dose_mg / CL * 1e6 / 384.49 (CL in mL/h, elafibranor molecular weight 384.49 g/mol).
• Units: umol*h/L. Document per-model via covariateData[[AUC_ELA]]$units. Must be on the same scale as the AUC50 the consuming model carries.
• Type: continuous
• Scope: specific
• Reference category: n/a – enters only through the sum AUC_ELA + AUC_GFT1007, which drives an inhibitory Emax term for alkaline phosphatase and a linear term for total bilirubin. Set to 0 for placebo arms; both drug terms then vanish exactly. Reference value observed: AUC50 = 24.1 umolh/L for the summed metric (Ooi 2026 Table 2); the summed metric had median 32.3 umolh/L on 80 mg/day and 39.3 umol*h/L on 120 mg/day (Figure 5 legend).
• Source aliases:
  • AUCSSP – the variable name in Ooi 2026 Supplementary Datafile S6 $PK; printed in the paper as AUC tau,ss of elafibranor. Same units and orientation as the canonical.
• Example models: Ooi_2026_elafibranor_alptb.R (joint ALP / total-bilirubin indirect-response model).
• Notes: Member of the AUC_<DRUG> family; sibling of AUC_GFT1007, the active metabolite. Compute it from modellib('Ooi_2026_elafibranor') as individual fdepot * dose_umol / cl, or equivalently from the milligram dose with the 384.49 g/mol conversion. Elafibranor contributes only about one sixth of the summed exposure metric at 80 mg/day.

AUC_GFT1007 (canonical for steady-state dosing-interval AUC of GFT1007, the active elafibranor metabolite)

• Description: Per-subject (time-fixed) area under the GFT1007 plasma concentration-time curve over the once-daily elafibranor dosing interval at steady state, supplied as a static drug-exposure column to the sequentially fitted exposure-response model. Ooi 2026 Supplementary Datafile S6 $PK computes it as F * dose_mg / CL * 1e6 / 386.51 (CL in mL/h, GFT1007 molecular weight 386.51 g/mol).
• Units: umol*h/L. Document per-model via covariateData[[AUC_GFT1007]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – enters only through the sum AUC_ELA + AUC_GFT1007. Set to 0 for placebo arms.
• Source aliases:
  • AUCSSM – the variable name in Ooi 2026 Supplementary Datafile S6 $PK.
• Example models: Ooi_2026_elafibranor_alptb.R (joint ALP / total-bilirubin indirect-response model).
• Notes: Member of the AUC_<DRUG> family; sibling of AUC_ELA. The two enter only as a sum because elafibranor and GFT1007 were shown to be equipotent (Ooi 2026 Methods 2.3.2). Compute it from modellib('Ooi_2026_elafibranor_gft1007'). Note that the source converts the milligram dose to moles with the METABOLITE molecular weight here and with the PARENT molecular weight for AUC_ELA, a 0.5% difference that is reproduced as published rather than harmonised.

AUC_ESL (canonical for dosing-interval AUC of eslicarbazepine)

• Description: Per-subject area under the eslicarbazepine plasma concentration-time curve over the 24-hour once-daily dosing interval (AUC0-24), supplied as a static drug-exposure column to sequentially fitted exposure-response models that carry no PK compartments of their own. Eslicarbazepine is the primary active metabolite of the antiepileptic prodrug eslicarbazepine acetate (ESL) and accounts, with its glucuronides, for 94% of oral systemic exposure, so it is the analyte every ESL exposure-response analysis uses. At steady state under once-daily dosing the value is simply dose / (CL/F).
• Units: ng*h/mL. Document per-model via covariateData[[AUC_ESL]]$units. Must be on the same scale as the slope or AUC50 the consuming model carries.
• Type: continuous
• Scope: specific
• Reference category: n/a – enters linearly (a log-odds slope in the safety logistic models, a (mmol/L)/(ngh/mL) slope in the serum-sodium model). Set to 0 for placebo arms; every consuming term then vanishes exactly and the model’s intercept is the placebo value. NOT centred and NOT scaled in any model registering this column. Reference values observed: 164,600 ngh/mL on ESL 400 mg/day and 329,200 ng*h/mL on 800 mg/day for a subject with the typical CL/F of 2.43 L/h (Gidal 2018).
• Source aliases:
  • AUC – the variable name in Gidal 2018 Appendix S1 Equations E-3 and E-4, printed in the paper as AUC0-24 of eslicarbazepine. Same units and orientation as the canonical.
  • AUC_0-24_ij – the per-visit form in Gidal 2018 Equation E-6, where the column is time-varying across study visits rather than time-fixed.
• Example models: Gidal_2018_eslicarbazepine_dizziness.R, Gidal_2018_eslicarbazepine_headache.R (log-odds slopes of -0.00000835 and -0.00000632 per ngh/mL; both NEGATIVE, a published result the paper calls unexpected and explains by the models counting only the FIRST occurrence of each adverse event, which falls in the low-exposure titration period), Gidal_2018_eslicarbazepine_serum_sodium.R (slope -0.0000041 (mmol/L)/(ngh/mL) on the change from baseline in serum sodium).
• Notes: Member of the AUC_<DRUG> family; siblings include AUC_ELA, AUC_GFT1007, AUC_CARBO and AUC_GCV. The units are pinned unusually firmly for this member: Gidal 2018 Table S-6 prints its serum-sodium slope units explicitly as (mmol/L)/(ng x h/mL), and the main-text worked example – a 400 mg dose increase raises AUC0-24 by 165 ugh/mL and lowers serum sodium by 0.68 mmol/L – reproduces only on the ngh/mL scale. Compute it from modellib('Gidal_2018_eslicarbazepine') as individual dose / cl. Distinct from CAV and CMAX, the interval-average and peak metrics from the same PK model: Gidal 2018 screened all three and retained different ones for different endpoints, so do not substitute one for another when reusing these models.

AUC_GCV (canonical for per-q12h-interval AUC of ganciclovir)

• Description: Time-varying ganciclovir AUC over a q12h dosing interval (AUC_0-12), used as the drug-exposure input to indirect-response viral-turnover PK/PD models of cytomegalovirus (CMV) viral load decline under (val)ganciclovir treatment. The Koloskoff 2025 source computes individual AUC_0-12 from an upstream popPK model (Franck 2021) and feeds it to the PD model as a Monolix “varying input”; the PD model itself does not integrate a PK ODE, so AUC_GCV is supplied to nlmixr2 as a time-varying data column.
• Units: mg*h/L (document per-model via covariateData[[AUC_GCV]]$units if a different exposure unit is reported).
• Type: continuous
• Scope: specific
• Reference category: n/a – set to 0 in pre-treatment / off-treatment records so the drug-stimulation term Emax * AUC_GCV / (EC50 + AUC_GCV) vanishes and the viral load returns to the kin / kout steady-state baseline.
• Source aliases:
  • AUC_0-12 – the printed variable name in Koloskoff 2025 (Methods Section 2.3, Eq. 1, and Table 1). Q24h dosing intervals are entered as AUC_0-24 / 2 so all data live in the q12h framework (Koloskoff 2025 Methods Section 2.1).
• Example models: Koloskoff_2025_ganciclovir.R (Koloskoff 2025 indirect viral turnover model for CMV viral load in pediatric SOT / HSCT recipients; AUC_GCV enters the ODE via kout * (1 + Emax * AUC_GCV / (EC50 + AUC_GCV)) * viral_load).
• Notes: Specific scope – the column meaning is tied to ganciclovir as the drug and to a q12h interval-averaging convention. Sibling drug-specific AUC canonicals (AUC_CARBO, AUC_GEM, AUC_BAST_FW, AUC_PAZO) follow the same AUC_<DRUG> naming pattern; a future PK/PD model that uses a different exposure metric for ganciclovir (e.g., trough concentration, instantaneous concentration) should register a parallel canonical rather than overload AUC_GCV. Koloskoff 2025 Monte Carlo simulations are reported under AUC_0-24 (Tables 3 and 4) assuming AUC_0-24 = 2 x AUC_0-12 at steady state; nlmixr2 simulations should set AUC_GCV to the q12h-interval value (i.e., AUC_0-24 / 2).

AUC_LCM (canonical for daily AUC of lacosamide at steady state)

• Description: Daily area under the plasma concentration-time curve of lacosamide (LCM) at steady state (mg*h/L), used as the drug-exposure covariate on the seizure hazard in time-to-seizure models of LCM anti-epileptic therapy. Dose-step-varying (the value tracks the patient’s currently assigned LCM target dose level: 100 -> 200 -> 400 -> 600 mg/day). Set to 0 for subjects not on lacosamide so the centred-deviation covariate contribution is gated by a treatment-arm indicator (CONMED_LCM) inside the model rather than by the covariate value itself.
• Units: mg*h/L (document per-model via covariateData[[AUC_LCM]]$units if a different exposure unit is reported).
• Type: continuous
• Scope: specific
• Reference category: n/a – enters via centred-deviation form slope * (AUC_LCM - AUC_ref). Reference value observed: 104 mgh/L (Lindauer 2017 SP0993 typical daily AUC at the first LCM target dose 200 mg/day; Lindauer 2017 Table 3 note c). Approximate scaling at higher doses: ~208 mgh/L at 400 mg/day, ~312 mg*h/L at 600 mg/day.
• Source aliases:
  • AUC_LCM – printed name in Lindauer 2017 Table 3 (rows AUC_LCM * k1 and AUC_LCM * k2); the paper computes per-subject daily AUC from an empirical-Bayes clearance estimate of a previously published lacosamide popPK model (Lindauer 2017 Section 2.3).
• Example models: Lindauer_2017_lacosamide_seizure.R (centred-deviation form on both first-seizure and subsequent-seizure Weibull scale parameters: slopes -0.00917 (1st) and -0.00751 (2nd+); gated by CONMED_LCM = 1).
• Notes: Specific scope; lacosamide-specific. Follows the AUC_<DRUG> sibling family (AUC_CARBO, AUC_GEM, AUC_BAST_FW, AUC_PAZO, AUC_GCV, AUC_RTV). Downstream users should compute the per-subject daily AUC from a lacosamide popPK model (e.g., a Cawello 2013 / 2015 popPK) or from the actual LCM dose regimen if a compartmental model is being coupled to the TTE hazard.

AUC_CBZ (canonical for daily AUC of carbamazepine at steady state)

• Description: Daily area under the plasma concentration-time curve of carbamazepine (CBZ; typically as controlled-release CBZ-CR) at steady state (mg*h/L), used as the drug-exposure covariate on the seizure hazard in time-to-seizure models of CBZ anti-epileptic therapy. Dose-step-varying (the value tracks the patient’s currently assigned CBZ-CR target dose level: 200 -> 400 -> 800 -> 1200 mg/day in the Lindauer 2017 SP0993 design). Set to 0 for subjects not on carbamazepine so the centred-deviation covariate contribution is gated by a treatment-arm indicator (CONMED_LCM in Lindauer 2017; equivalent to 1 - CONMED_CBZ for that design) inside the model rather than by the covariate value itself.
• Units: mg*h/L (document per-model via covariateData[[AUC_CBZ]]$units if a different exposure unit is reported).
• Type: continuous
• Scope: specific
• Reference category: n/a – enters via centred-deviation form slope * (AUC_CBZ - AUC_ref). Reference value observed: 132 mg*h/L (Lindauer 2017 SP0993 typical daily AUC at the first CBZ-CR target dose 400 mg/day for a 70-kg patient; Lindauer 2017 Section 3.4 first paragraph).
• Source aliases:
  • AUC_CBZ – printed name in Lindauer 2017 Table 3 (rows AUC_CBZ * k1 and AUC_CBZ * k2); the paper computes per-subject daily AUC from an empirical-Bayes clearance estimate of a previously published carbamazepine popPK model (Lindauer 2017 Section 2.3).
• Example models: Lindauer_2017_lacosamide_seizure.R (centred-deviation form on both first-seizure and subsequent-seizure Weibull scale parameters: slopes -0.00658 (1st) and -0.0153 (2nd+); gated by CONMED_LCM = 0 – i.e., the CBZ-CR arm).
• Notes: Specific scope; carbamazepine-specific. Distinct from CONMED_CBZ – CONMED_CBZ is the binary carbamazepine coadministration indicator (used in models where the effect of CBZ enters as a categorical on / off shift, e.g., Schoemaker 2017 brivaracetam), while AUC_CBZ carries the magnitude of CBZ exposure and is used when the effect scales with dose. Follows the AUC_<DRUG> sibling family.

AUC_PAZO (canonical for per-period mean AUC of pazopanib)

• Description: Per-period (per-dose-group in preclinical xenograft studies; per-subject mean dose-adjusted in clinical studies) mean AUC of pazopanib used as the drug-exposure covariate driving the antiangiogenic and cytotoxic effect rates in semi-mechanistic tumour-growth / angiogenesis-inhibition (TGI) models of pazopanib in renal-cell carcinoma. Time-varying step-wise (held constant within a treatment period and resetting when dose level changes or treatment ends).
• Units: ug*h/mL (= mg*h/L). Document per-model via covariateData[[AUC_PAZO]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – enters via power form a = a0 * AUC_PAZO^e_auc_pazo_a and c = c0 * AUC_PAZO^e_auc_pazo_c. Set to 0 in periods where pazopanib is not administered; the model rate terms reduce to baseline drug-off (a, c -> 0 when AUC_PAZO -> 0) under that convention.
• Source aliases:
  • AUC – used in Ouerdani_2015_pazopanib_mouse.R and Ouerdani_2015_pazopanib.R (Ouerdani 2015 Methods Equations 2-3; preclinical values 220.2, 656.8, 1140.8 ugh/mL for the 10, 30, 100 mg/kg mouse dose groups; clinical mean 771.6 ugh/mL for 800 mg QD pazopanib in RCC patients, with per-subject values 629.4-802.4 ug*h/mL derived from an Emax fit to mean AUCs at the patient’s dose history).
• Example models: Ouerdani_2015_pazopanib_mouse.R (preclinical TGI in CAKI-2 xenograft mice; AUC enters as c = c0 * AUC_PAZO^0.332 only – the cytotoxic exponent e_auc_pazo_a was fixed to 0 because the in-mouse cytotoxic effect did not vary with exposure across the 10-100 mg/kg range), Ouerdani_2015_pazopanib.R (clinical TGI in RCC patients; AUC enters as both a = a0 * AUC_PAZO^0.125 and c = c0 * AUC_PAZO^0.142).
• Notes: Specific scope because the column meaning is tied to a particular drug (pazopanib) and the power-form parameterisation that the Ouerdani 2015 model uses. Reusing the name for a different tyrosine-kinase inhibitor is not appropriate – register a sibling canonical (AUC_SORAF for sorafenib, AUC_SUNI for sunitinib, etc.) when needed. The Ouerdani 2015 paper reports the preclinical AUCs in ug*h/mL from a separate preclinical PK study (cited as the FDA Pharmacology Review for pazopanib NDA 022465); the clinical AUCs come from an Emax fit (Equation derived from Methods) to pooled mean AUCs at varying daily doses (5 mg to 2000 mg) across five prior pazopanib trials.

AUC_RTV (canonical for ritonavir AUC over the 0-24 h dosing interval)

• Description: Per-subject (time-fixed within an evaluated regimen) ritonavir AUC over the 0-24 h once-daily dosing interval, used as a co-medication exposure covariate driving boosted-protease-inhibitor (atazanavir, lopinavir, darunavir, etc.) clearance in popPK models that account for ritonavir’s CYP3A4-inhibition effect. In Dickinson 2009 the value is computed by non-compartmental methods on the observed ritonavir concentration-time profile (WinNonlin 5.2) and feeds the atazanavir CL/F power form cl = exp(lcl) * (AUC_RTV / 7.52)^e_aucrtv_cl.
• Units: mg*h/L (document per-model via covariateData[[AUC_RTV]]$units if a different exposure unit is reported).
• Type: continuous
• Scope: specific
• Reference category: n/a – enters via centred power form. Reference value observed: 7.52 mg*h/L (Dickinson 2009 cohort median across HIV-infected and healthy volunteers; healthy 7.36, HIV 7.59 per Table 1).
• Source aliases:
  • RTVAUC – printed name in Dickinson 2009 Table 1 / Table 3 (the paper writes RTVAUC0-24 with the dosing-interval subscript; the column name is registered without the subscript to fit standard data-column character constraints).
• Example models: Dickinson_2009_atazanavir.R (atazanavir CL/F power-function dependence on RTVAUC0-24, centred at 7.52 mg*h/L with exponent -0.8).
• Notes: Specific scope because the column meaning is tied to ritonavir as the booster drug and to the 0-24 h once-daily dosing-interval AUC convention. Sibling drug-specific AUC canonicals (AUC_CARBO, AUC_GEM, AUC_BAST_FW, AUC_PAZO, AUC_GCV) follow the same AUC_<DRUG> naming pattern. A future PK model that uses a different ritonavir exposure metric (trough concentration, q12h-interval AUC for BID ritonavir regimens) should register a parallel canonical rather than overload AUC_RTV. For simulation users without observed ritonavir AUC, the Dickinson 2009 cohort median 7.52 mg*h/L reproduces typical-value behaviour (the centring point of the covariate effect).

AUC_VERUB (canonical for verubecestat AUC over the 24 h dosing interval at steady state)

• Description: Time-varying verubecestat plasma AUC over the once-daily 24 h dosing interval at steady state, used as the driver of the inhibitory Emax sigmoid on the amyloid plaque formation rate Kin in the van Maanen 2025 amyloid plaque turnover model (paper Eq 4: Inh_verub = Imax * AUC_VERUB / (AUC_VERUB + AUC50)). The van Maanen 2025 analysis derives individual AUC_VERUB from the upstream Dockendorf 2022 verubecestat population PK model (cited in Table S1).
• Units: uM*h (micromolar * hour). Must be in the same units as the model’s AUC50 parameter (uM*h) so the sigmoid AUC/(AUC + AUC50) is dimensionless. Document per-model via covariateData[[AUC_VERUB]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – enters via Emax-sigmoid form Imax * AUC / (AUC + AUC50). Set to 0 in periods where verubecestat is not administered so the inhibition term vanishes and Kin returns to its typical (undrugged) value. Typical AUC values reverse-computed from the paper’s Table S5 stable-plaque predictions and AUC50 = 0.392 uMh: 40 mg daily -> approximately 4.4 uMh (91.8% inhibition); 11.9 mg -> approximately 1.44 uMh (78.7%); 2.9 mg -> approximately 0.35 uMh (47.2%); 1.7 mg -> approximately 0.21 uM*h (34.5%).
• Source aliases:
  • AUC_verub – printed name in van Maanen 2025 Eq 4 and Figure 1.
• Example models: vanMaanen_2025_amyloid.R (verubecestat inhibition of plaque formation via Imax = 1 fixed and AUC50 = 0.392 uM*h).
• Notes: Specific scope because the value is intrinsically tied to verubecestat (BACE1 inhibitor) and the once-daily 24 h dosing-interval AUC convention. Sibling drug-specific AUC canonicals (AUC_CARBO, AUC_GEM, AUC_BAST_FW, AUC_PAZO, AUC_GCV, AUC_LCM, AUC_CBZ, AUC_RTV, AUC_EMPA, AUC_ADU, AUC_DON, AUC_GAN, AUC_LEC) follow the same AUC_<DRUG> naming pattern.

AUC_ADU (canonical for aducanumab serum AUC over the 4-week dosing interval at steady state)

• Description: Time-varying aducanumab serum AUC over the 4-week Q4W dosing interval at steady state, used as the driver of the linear stimulation term on the amyloid plaque elimination rate Kout in the van Maanen 2025 amyloid plaque turnover model (paper Eq 6: Stim_ADU = slope_adu * (AUC_ADU / MW_ADU)). The van Maanen 2025 analysis derives aducanumab AUC from the upstream Sevigny 2016 phase 1b popPK study (Table S1) with linear dose scaling where the target dose differs from the source PK study dose.
• Units: mg*day/L (milligrams * day per litre). Must be in mgday/L (not mgh/mL, not ugh/mL) so that AUC_ADU / MW_ADU_g_per_mol evaluates in mmolday/L = mMday and the paper’s slope units (mM^-1 day^-1) multiply to a dimensionless stimulation term. Conversion from mg*h/L: divide by 24. Document per-model via covariateData[[AUC_ADU]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – enters via linear form slope_adu * AUC_ADU / MW_ADU. Set to 0 in periods where aducanumab is not administered so the stimulation term vanishes and Kout returns to its typical (undrugged) value. Reference regimen (Table S3): EMERGE/ENGAGE high-dose titration -> target 10 mg/kg IV Q4W, giving a typical steady-state 4-week AUC of approximately 1944 mg*day/L for a 70 kg subject at CL approximately 0.36 L/day.
• Source aliases:
  • AUC_mAb (for aducanumab) – printed name in van Maanen 2025 Eq 6 and Figure 1.
• Example models: vanMaanen_2025_amyloid.R (aducanumab stimulation of plaque elimination via slope_adu = 719 mM^-1 day^-1 and MW_ADU = 145912 g/mol).
• Notes: Specific scope. Sibling anti-A-beta mAb AUC canonicals AUC_DON (donanemab), AUC_GAN (gantenerumab), AUC_LEC (lecanemab) follow the same convention and are consumed by the same van Maanen 2025 model.

AUC_DON (canonical for donanemab serum AUC over the 4-week dosing interval at steady state)

• Description: Time-varying donanemab serum AUC over the 4-week Q4W dosing interval at steady state, used as the driver of the linear stimulation term on the amyloid plaque elimination rate Kout in the van Maanen 2025 amyloid plaque turnover model (paper Eq 6: Stim_DON = slope_don * (AUC_DON / MW_DON)). The van Maanen 2025 analysis derives donanemab AUC from the Lowe 2021 phase 1b SAD/MAD popPK data (Table S1).
• Units: mg*day/L. Same unit convention as AUC_ADU (see notes there). Document per-model via covariateData[[AUC_DON]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – set to 0 in periods where donanemab is not administered. Reference regimen (Table S3): TRAILBLAZER-ALZ 1/2/4 titration to 1400 mg IV Q4W after 3 loading doses of 10 mg/kg IV Q4W.
• Source aliases:
  • AUC_mAb (for donanemab) – printed name in van Maanen 2025 Eq 6 and Figure 1.
• Example models: vanMaanen_2025_amyloid.R (donanemab stimulation of plaque elimination via slope_don = 1120 mM^-1 day^-1 and MW_DON = 145087 g/mol).
• Notes: Specific scope. Sibling anti-A-beta mAb AUC canonicals AUC_ADU, AUC_GAN, AUC_LEC follow the same convention.

AUC_GAN (canonical for gantenerumab serum AUC over the 4-week dosing interval at steady state)

• Description: Time-varying gantenerumab serum AUC over the 4-week dosing interval at steady state (delivered as SC Q4W or IV titration in the source trials, expressed here as a 4-week interval-AUC to match the paper’s Kout stimulation term), used as the driver of the linear stimulation term on the amyloid plaque elimination rate Kout in the van Maanen 2025 amyloid plaque turnover model (paper Eq 6: Stim_GAN = slope_gan * (AUC_GAN / MW_GAN)). The van Maanen 2025 analysis derives gantenerumab AUC from the Portron 2020 phase 1 single active dose PK data (Table S1).
• Units: mg*day/L. Same unit convention as AUC_ADU. Document per-model via covariateData[[AUC_GAN]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – set to 0 in periods where gantenerumab is not administered. Reference regimen (Table S3): GRADUATE 1/2 titration to 1020 mg every 4 weeks (delivered as 510 mg SC Q2W) after a 9-month step-up sequence of 120 -> 255 -> 510 -> 1020 mg per 4 weeks.
• Source aliases:
  • AUC_mAb (for gantenerumab) – printed name in van Maanen 2025 Eq 6 and Figure 1.
• Example models: vanMaanen_2025_amyloid.R (gantenerumab stimulation of plaque elimination via slope_gan = 397 mM^-1 day^-1 and MW_GAN = 146300 g/mol).
• Notes: Specific scope. Sibling anti-A-beta mAb AUC canonicals AUC_ADU, AUC_DON, AUC_LEC follow the same convention.

AUC_LEC (canonical for lecanemab serum AUC over the 4-week dosing interval at steady state)

• Description: Time-varying lecanemab serum AUC over the 4-week dosing interval at steady state (delivered as IV Q2W in Clarity AD, expressed here as the equivalent 4-week interval-AUC = 2 * Q2W-interval AUC), used as the driver of the linear stimulation term on the amyloid plaque elimination rate Kout in the van Maanen 2025 amyloid plaque turnover model (paper Eq 6: Stim_LEC = slope_lec * (AUC_LEC / MW_LEC)). The van Maanen 2025 analysis derives lecanemab AUC from the Logovinsky 2016 phase 1 SAD/MAD PK data (Table S1).
• Units: mg*day/L. Same unit convention as AUC_ADU. Document per-model via covariateData[[AUC_LEC]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – set to 0 in periods where lecanemab is not administered. Reference regimen (Table S3): Clarity AD 10 mg/kg IV Q2W (equivalent to approximately 20 mg/kg per 4 weeks).
• Source aliases:
  • AUC_mAb (for lecanemab) – printed name in van Maanen 2025 Eq 6 and Figure 1.
• Example models: vanMaanen_2025_amyloid.R (lecanemab stimulation of plaque elimination via slope_lec = 606 mM^-1 day^-1 and MW_LEC = 150000 g/mol).
• Notes: Specific scope. Sibling anti-A-beta mAb AUC canonicals AUC_ADU, AUC_DON, AUC_GAN follow the same convention.

AUC_IBRU (canonical for daily AUC(0-24) of ibrutinib)

• Description: Daily 0-24 h area under the plasma concentration-time curve of ibrutinib, used as the systemic-exposure driver in ibrutinib exposure-response models. Dose-step-varying: the value tracks the patient’s currently assigned daily ibrutinib dose level (420, 280 or 140 mg/day in the de-escalation schedules simulated by Ibrahim 2023; 420 and 840 mg/day in the PCYC-1102 study itself) and drops to 0 during treatment interruptions. Ibrahim 2023 selected AUC(0-24) rather than a concentration because ibrutinib’s half-life is short (4-13 h) and daily AUC has been proposed as the surrogate exposure metric for its therapeutic drug monitoring.
• Units: h*ng/mL (document per-model via covariateData[[AUC_IBRU]]$units if a different exposure unit is reported).
• Type: continuous
• Scope: specific
• Reference category: n/a – enters saturable Emax / Imax forms of the shape AUC_IBRU / (AUC50 + AUC_IBRU) rather than a centred-deviation or power form. Reference (half-maximal) values observed: 34.1 hng/mL for inhibition of phosphorylated-Btk production (Ibrahim 2023 Table 1, IAUC50,pBTK); 91.7 hng/mL for stimulation of systolic blood pressure and 63.1 h*ng/mL for diastolic blood pressure (Ibrahim 2023 Table 2, AUC50,sBP / AUC50,dBP).
• Source aliases:
  • DAILYAUC – printed name in the Ibrahim 2023 nlmixr control streams (Supporting Information S4, auc <- DAILYAUC in run100 / run200 / run300).
• Example models: Ibrahim_2023_ibrutinib_leukocyte_spd.R (Imax inhibition of pBtk production, IAUC50 = 34.1), Ibrahim_2023_ibrutinib_sbp.R (Emax stimulation of sBP production, AUC50 = 91.7), Ibrahim_2023_ibrutinib_dbp.R (Emax stimulation of dBP production, AUC50 = 63.1), Ibrahim_2025_ibrutinib_cll.R (Imax inhibition of pBtk production, IAUC50 = 28.4), Ibrahim_2025_ibrutinib_bp.R (Emax stimulation of sBP and dBP production through a single shared AUC50,BP = 62.3), Ibrahim_2025_ibrutinib_venetoclax.R (same Imax term as the 2025 efficacy model, alongside the venetoclax concentration covariate [[CONC_VEN_MGL]]).
• Notes: Specific scope; ibrutinib-specific. Follows the AUC_<DRUG> sibling family (AUC_CARBO, AUC_GEM, AUC_LCM, AUC_CBZ, AUC_PAZO, AUC_GCV, AUC_RTV, AUC_EMPA). As with AUC_LCM, downstream users must compute the per-subject daily AUC from an ibrutinib popPK model rather than from this register – Ibrahim 2023 integrated individual post hoc profiles of the two-compartment ibrutinib popPK model of Marostica et al. (Cancer Chemother Pharmacol. 2015;75(1):111-121), which is not currently part of nlmixr2lib. Because all three Ibrahim 2023 models use the same exposure column with different half-maximal constants, the ratio of those constants is itself interpretable: AUC50,sBP and AUC50,dBP are about three and two times IAUC50,pBTK, which is the quantitative basis for the paper’s finding that dose de-escalation sheds hypertension risk faster than it sheds efficacy.

AUC_FENEBRUTINIB (canonical for steady-state daily AUC(0-24) of fenebrutinib)

• Description: Individual steady-state daily (0-24 h) area under the plasma concentration-time curve of fenebrutinib (GDC-0853), a noncovalent Bruton’s tyrosine kinase inhibitor, used as the time-fixed exposure driver of fenebrutinib exposure-response models. 0 for placebo-treated subjects.
• Units: h*ng/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – enters saturable Emax forms AUC_FENEBRUTINIB / (AUC50 + AUC_FENEBRUTINIB). Half-maximal values: 2650 hng/mL for the ACR20/50/70 logit drug effect, 293 hng/mL for the DAS28 (CRP) drug effect (Chan 2020 Tables S4 and S5).
• Source aliases:
  • AUC – Chan 2020 Model S2 / Model S3 NONMEM $PRED column name. No value transformation.
• Example models: Chan_2020_fenebrutinib_acr.R, Chan_2020_fenebrutinib_das28.R.
• Notes: Specific scope; fenebrutinib-specific. Follows the AUC_<DRUG> sibling family (AUC_IBRU, AUC_EMPA, AUC_LCM, …). Chan 2020 derived it as the empirical Bayes estimate of the companion popPK model Chan_2020_fenebrutinib.R at the NOMINAL dose (dose intensity >= 99.3% per arm); users can regenerate it from that model as the 0-24 h AUC at steady state. Phase 2 medians (Chan 2020 Figure S3): about 1200, 3400 and 7000 h*ng/mL for 50 mg QD, 150 mg QD and 200 mg BID.

AUC_LZGX (canonical for daily AUC(0-24) of linzagolix)

• Description: Linzagolix area under the plasma concentration-time curve over the 24 h once-daily dosing interval, used as the systemic-exposure driver of the oestradiol-suppression model in Pohl 2022. Dose-varying: the value tracks the subject’s currently assigned daily linzagolix dose and is 0 on placebo, which zeroes the inhibitory term. Pohl 2022 selected daily AUC rather than a concentration because oestradiol is sampled at weekly-to-monthly visits with no relation to the linzagolix dosing clock, while linzagolix itself has a short (~8 h) half-life and reaches steady state within about two days.
• Units: ng*h/mL (document per-model via covariateData[[AUC_LZGX]]$units if a different exposure unit is reported).
• Type: continuous
• Scope: specific
• Reference category: n/a – enters a sigmoid Imax form of the shape AUC_LZGX^hill / (AUC50^hill + AUC_LZGX^hill) rather than a centred-deviation or power form. Reference (half-maximal) value observed: 1.68e5 ng*h/mL for suppression of serum oestradiol (Pohl 2022 Table 5, AUC50).
• Source aliases:
  • AUC – printed name in the Pohl 2022 Supporting Information section 3.2 core PK-E2 equation (AUC_ij).
• Example models: Pohl_2022_linzagolix_e2.R (sigmoid Imax suppression of serum oestradiol, AUC50 = 1.68e5 ng*h/mL, Hill exponent 1.78).
• Notes: Specific scope; linzagolix-specific. Follows the AUC_<DRUG> sibling family (AUC_CARBO, AUC_GEM, AUC_LCM, AUC_CBZ, AUC_PAZO, AUC_GCV, AUC_RTV, AUC_EMPA, AUC_IBRU, AUC_LEN). Unlike AUC_IBRU and AUC_LCM, whose upstream popPK models are not in nlmixr2lib, the popPK model that generates this column IS distributed here as Pohl_2022_linzagolix.R (same publication, Table 4), so a downstream user can compute the column end-to-end: for linear PK under once-daily dosing the steady-state daily AUC is Dose / (CL/F), and the companion model’s concentrations are in ug/mL, so the ugh/mL integral must be multiplied by 1000 to match this column’s ngh/mL unit. Pohl 2022 Figure 2 shows the resulting AUCss histograms for the 50, 75, 100 and 200 mg EDELWEISS dose groups.

AUC_AMPH (canonical for per-24-h-interval AUC of amphenmulin in an in-vitro dynamic model)

• Description: Amphenmulin area under the concentration-time curve over the current 24 h dosing interval (AUC_0-24h), measured in the medium of an in-vitro dynamic (dilution) apparatus rather than in plasma, used as the drug-exposure input to antibacterial PK/PD-index models. Wang 2024 samples the reaction chamber of a peristaltic-pump-driven three-necked-flask model by HPLC-MS/MS and derives the per-interval value non-compartmentally in Phoenix; because that source publishes no structural PK model for the apparatus, the PD model consumes AUC_AMPH directly as a piecewise-constant time-varying covariate (one value per 24 h interval) rather than integrating a PK ODE. The model forms the PK/PD index AUC24h/MIC by dividing this covariate by the challenge strain’s MIC.
• Units: h*ug/mL (equivalently mg*h/L). Document per-model via covariateData[[AUC_AMPH]]$units if a different exposure unit is reported.
• Type: continuous
• Scope: specific
• Reference category: n/a – set to 0 for untreated control records so the sigmoid term (E0 - Emax) * (AUC_AMPH/MIC)^N / (EC50^N + (AUC_AMPH/MIC)^N) vanishes and the predicted 24 h change in mycoplasma count reduces to the control value E0.
• Source aliases:
  • AUC24h / AUC 24 h – printed names in Wang 2024 Materials and Methods (Integration and modeling of pharmacokinetics/pharmacodynamics, defining the AUC24h/MIC index) and Table 3. Wang 2024 Results (In vitro pharmacokinetics and the effects on M. gallisepticum) reports the per-interval ranges across the six regimens: 0.34-6.19 hug/mL over 0-24 h, 0.57-7.69 hug/mL over 24-48 h and 0.62-8.23 h*ug/mL over 48-72 h.
• Example models: Wang_2024_amphenmulin_pkpd_index.R (Wang 2024 inhibitory sigmoid Emax PK/PD integration against Mycoplasma gallisepticum strain S6 in an in-vitro dynamic model; AUC_AMPH / mic drives the per-interval signed change in log10 CFU/mL).
• Notes: Specific scope – amphenmulin-specific, and tied to a 24 h interval-AUC convention. Member of the AUC_<DRUG> family; AUC_TILM is the closest structural analogue, being likewise a per-24-h-interval antibacterial exposure handed to a sigmoid PK/PD-index model in place of a PK ODE, and originating from the same laboratory (Ding H, South China Agricultural University). Distinct from the CONC_<DRUG>_<UNITS> family, which carries an instantaneous in-vitro concentration driving a mechanism-based kill-rate model: the sibling model Wang_2024_amphenmulin_killrate.R needs no such covariate because it integrates the apparatus PK as an ODE state. A future amphenmulin model using the Cmax/MIC metric, which Wang 2024 Table 3 also reports, should register a parallel canonical rather than overload AUC_AMPH.

AUC_APTM (canonical for plasma AUC of APTM, the pleuromutilin derivative 14-O-[(4-amino-6-hydroxy-pyrimidine-2-yl) thioacetyl] mutilin)

• Description: APTM area under the plasma concentration-time curve after a single oral dose, used as the numerator of the AUC/MIC PK/PD index that drives an in vivo antibacterial exposure-response model. Yang 2026 sampled plasma from Mycoplasma gallisepticum-infected chickens by LC-MS/MS, derived the exposure non-compartmentally in Phoenix WinNonlin, and regressed it on dose so the index could be extrapolated to the eight doses of the efficacy arm; because that source publishes no structural compartmental PK model, the PD model consumes AUC_APTM directly as a covariate rather than integrating a PK ODE. The model forms the PK/PD index AUC/MIC by dividing this covariate by the challenge strain’s MIC.
• Units: h*ug/mL (equivalently mg*h/L). Note that Yang 2026 Table 2 tabulates h*ng/mL, so divide the tabulated value by 1000; this keeps the AUC/MIC ratio consistent with a mic expressed in ug/mL. Document per-model via covariateData[[AUC_APTM]]$units if a different exposure unit is reported.
• Type: continuous
• Scope: specific
• Reference category: n/a – set to 0 for untreated control records so the sigmoid term Imax * (AUC_APTM/MIC)^gamma / (IC50^gamma + (AUC_APTM/MIC)^gamma) vanishes and the predicted 72 h change in lung mycoplasma load reduces to the control value E0.
• Source aliases:
  • AUC0-24h – the printed name in Yang 2026 Materials and methods (‘Pharmacokinetic, pharmacodynamic, and statistical analysis’, defining the AUC0-24h/MIC index), Figure 3B and Table 3. Numerically this is the AUCINF column of Table 2, not a separately tabulated 0-24 h exposure, which the paper never prints. Two independent checks establish the identification: least-squares regression of the Table 2 AUCINF values (2382.53, 8551.66 and 30671.54 h*ng/mL at 5, 15 and 40 mg/kg) on dose returns R^2 = 0.99480, matching the paper’s stated 0.9948 to five significant figures; and inverting that regression at the Table 3 index targets reproduces the paper’s own back-calculated daily doses of 12.22 and 21.83 mg/kg to within 0.2%.
  • AUCINF – the Table 2 row the values actually come from.
• Example models: Yang_2026_APTM_aucmic.R (Yang 2026 inhibitory sigmoid Emax PK/PD integration against Mycoplasma gallisepticum strain S6 in an intratracheal chicken infection model; AUC_APTM / mic drives the signed 72 h change in lung log10 CFU/mL).
• Notes: Specific scope – APTM-specific. Member of the AUC_<DRUG> family; AUC_AMPH is the closest structural analogue, being likewise the plasma/medium exposure of a novel pleuromutilin derivative tested against M. gallisepticum strain S6 and handed to a sigmoid PK/PD-index model in place of a PK ODE, and originating from the same laboratory (Ding H, South China Agricultural University). Follows the AUCMIC_TYLO entry’s rule for choosing between the two families: Yang 2026 does report the challenge strain’s MIC, so the absolute-AUC form plus a model mic parameter is used rather than the ratio-carrying AUCMIC_<DRUG> form, which lets a user re-target the model to an isolate of different susceptibility. Unlike AUC_AMPH, AUC_TILM and AUC_CEFQ, this column is not a per-24-h-interval exposure updated across a multi-day course: Yang 2026’s in vivo readout is a single cross-sectional count at 72 h, so the covariate carries one single-dose exposure per dose group and is time-invariant. The paired Cmax/MIC index of the same paper is registered separately as CMAX_APTM.

AUC_CEFQ (canonical for per-24-h-interval serum AUC of cefquinome)

• Description: Cefquinome area under the serum concentration-time curve over the current 24 h dosing interval (AUC_0-24h), used as the drug-exposure input to ex vivo antibacterial PK/PD-index models. Gao 2025 dosed foals 1 mg/kg intramuscularly, sampled serum by HPLC, derived AUC non-compartmentally in WinNonlin 5.2.1, and then incubated E. coli in those same serum samples to build ex vivo time-kill curves; because that source publishes no structural compartmental PK model, the PD model consumes AUC_CEFQ directly as a piecewise-constant time-varying covariate (one value per 24 h interval) rather than integrating a PK ODE. Unlike the other members of this family, the index Gao 2025 forms is AUC_CEFQ / MIC / 24 h – the AUC/MIC ratio divided by 24 h, which the paper adopts deliberately so the index is dimensionless and can be substituted straight into its dose equation. Divide by the MIC and by 24 when reproducing that paper’s targets.
• Units: h*ug/mL (equivalently mg*h/L). Document per-model via covariateData[[AUC_CEFQ]]$units if a different exposure unit is reported.
• Type: continuous
• Scope: specific
• Reference category: n/a – set to 0 for drug-free control records so the sigmoid term vanishes and the predicted 24 h change in bacterial count reduces to the no-drug value.
• Source aliases:
  • AUC0-24h – printed name in Gao 2025 Section 2.9 (PK/PD Integration and Modeling, defining the AUC0-24h/MIC divided by 24 h index), Section 2.10 (Monte Carlo Simulation, where it is the log-normally distributed PK input) and Section 2.11 (Estimation of Dosage Regimen). Gao 2025 Table 1 reports the closely related truncated statistic AUC0-last as 12.33 +/- 0.69 hug/mL intravenously and 5.41 +/- 0.81 hug/mL intramuscularly after 1 mg/kg; because the last quantifiable sample is at 12 h (Figure 2) and the terminal half-life is 2.35-4.16 h, AUC0-last and AUC0-24h are numerically interchangeable in that study.
• Example models: Gao_2025_cefquinome_pkpd_index.R (Gao 2025 inhibitory sigmoid Emax PK/PD integration against Escherichia coli in foal serum; AUC_CEFQ / mic / 24 drives the per-interval signed change in log10 CFU/mL).
• Notes: Specific scope – cefquinome-specific, and tied to a 24 h interval-AUC convention. Member of the AUC_<DRUG> family; AUC_AMPH and AUC_TILM are the closest structural analogues, being likewise per-24-h-interval antibacterial exposures handed to a sigmoid PK/PD-index model in place of a PK ODE. It differs from both in carrying a serum (ex vivo) rather than an in-vitro-apparatus or tissue-cage exposure, and in that the companion model Gao_2025_cefquinome_foal.R does supply a compartmental PK model, so a user who wants a closed PK-PD loop can simulate the AUC rather than supply it; the covariate is retained because Gao 2025’s own PD analysis is defined per 24 h interval on non-compartmental AUC. Distinct from the CONC_<DRUG>_<UNITS> family, which carries an instantaneous concentration driving a mechanism-based kill-rate model.

AUC_CEFTIOFUR (canonical for the 24 h ex vivo exposure AUC of ceftiofur, as desfuroylceftiofur)

• Description: Ceftiofur area under the concentration-time curve over a 24 h exposure window (AUC24h), measured as its active metabolite desfuroylceftiofur (DFC), used as the numerator of the AUC24h/MIC index that drives an ex vivo inhibitory sigmoid Imax PK/PD-integration model. In Mi 2022’s ex vivo design each plasma or bronchoalveolar lavage fluid sample, taken after a single 5 mg/kg intramuscular dose to Pasteurella multocida-infected pigs, holds one fixed concentration C for the whole 24 h incubation, so AUC24h = 24 h * C.
• Units: h*ug/mL (equivalently mg*h/L). Document per-model via covariateData[[AUC_CEFTIOFUR]]$units if a different exposure unit is reported.
• Type: continuous
• Scope: specific
• Reference category: n/a – set to 0 for drug-free control records so the sigmoid term vanishes and the predicted 24 h change in log10 CFU/mL reduces to the control value E0.
• Source aliases:
  • AUC24 h – numerator of the PK/PD index AUC24 h/MIC printed in Mi 2022 Section 4.6 (as INDEX), Section 4.7 (dose equation) and Table 3 (target rows).
• Example models: Mi_2022_ceftiofur_pkpd_plasma.R, Mi_2022_ceftiofur_pkpd_balf.R (Mi 2022 ex vivo sigmoid Imax PK/PD integration against P. multocida HB13 in plasma and BALF; AUC_CEFTIOFUR / mic drives the signed 24 h change in log10 CFU/mL).
• Notes: Specific scope – ceftiofur-specific. Member of the AUC_<DRUG> family; the full INN is spelled out (as in AUC_FENEBRUTINIB) because every short form (CEF, CEFT, TIO, XNL) collides with, or is not recognisable beside, the abbreviations of other cephalosporins. Follows the AUCMIC_TYLO entry’s rule for choosing between the AUC_<DRUG> and ratio-carrying AUCMIC_<DRUG> families: Mi 2022 reports the challenge strain’s MIC (0.06 ug/mL in vitro and ex vivo), and that MIC reconstructs the paper’s ex vivo kill at the plasma sample concentrations it lists (Figure 2A), so the absolute-AUC form plus a model mic parameter is used. Unlike AUC_CEFQ, the index is not divided by 24 h.

AUC_LEN (canonical for lenvatinib steady-state daily AUC)

• Description: Lenvatinib area under the plasma concentration-time curve over the once-daily 24 h dosing interval at steady state, used as the drug-exposure driver of PK/PD models in which lenvatinib exposure acts on serum biomarker turnover or on tumor growth inhibition. Time-varying step-wise: held constant between assessments and updated when the dose level changes (or when treatment is interrupted or stopped). Majid 2024 derives it inside the upstream population PK run as AUC = 1000 * F1 * DGRP / CL – 1000 x relative bioavailability x current daily dose (mg) divided by the individual apparent clearance (L/h) – and carries it into the two downstream PK/PD runs as a data column. Two closely related conventions occur in that paper and are pooled onto this single canonical: the value at the time of the biomarker assessment (LENAUC, biomarker model) and the average over the interval between two tumor assessments computed from the average dose in that interval (LEVAVAUC, tumor model); record which one applies via covariateData[[AUC_LEN]]$notes.
• Units: ng*h/mL (equivalently ug*h/L). Must be in the same units as the model’s EC50 so the Emax term is dimensionless. Note that the Majid 2024 tumor control stream (Text S3) rescales the column to ug*h/mL with AUC = LEVAVAUC/1000 and divides its carried EC50 values by 1000 to match; the register keeps the untransformed ng*h/mL scale so the column is interchangeable across the biomarker and tumor models. Document per-model via covariateData[[AUC_LEN]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – enters via Emax / sigmoid-Emax forms Emax * AUC^gamma / (EC50^gamma + AUC^gamma) (biomarker turnover) and Emax * AUC / (AUC + EC50 * exp(lambda*t)) (tumor growth inhibition). Set to 0 for placebo subjects and for off-treatment periods; both forms then vanish exactly, leaving the disease-progression / pure-growth dynamics. Reference values observed: the common biomarker EC50 is 930 ngh/mL and the tumor EC50 is 1420 ngh/mL (Majid 2024 Tables 2 and 3); records above 10 000 ng*h/mL were excluded from the biomarker fit (Text S2 $DATA IGNORE).
• Source aliases:
  • LENAUC (Majid 2024 Text S2 biomarker control stream; lenvatinib AUC at the time of the biomarker assessment).
  • LEVAVAUC (Majid 2024 Text S3 tumor control stream; lenvatinib average AUC based on the average dose between two tumor assessments).
  • AUC / DNAUC (Majid 2024 Text S1 population PK $TABLE output columns; DNAUC is the dose-normalised form AUC * 24 / DGRP and is NOT this canonical).
• Example models: Majid_2024_lenvatinib_biomarkers.R (drives the sigmoid Emax effect on Kout for VEGF and FGF-23 and on Kin for Tie-2 and Ang-2, with a common EC50 of 930 ngh/mL), Majid_2024_lenvatinib_tumor.R (drives both the tumor-shrinkage Emax term with EC50 1420 ngh/mL and the re-integrated Tie-2 / Ang-2 turnover sub-models), Majid_2024_lenvatinib_tumor_asdeposited.R (same, but rescales the column internally to ug*h/mL as the deposited control stream does, which is what exposes the stream’s unscaled Tie-2 EC50), Tamai_2017_lenvatinib_teae_dosemod.R (the exposure driver of a landmark logistic exposure-safety model rather than of a turnover or tumour-growth model: logit = -4.71 + 1.82 * AUC_LEN / 1000 for a cycle-1 adverse event forcing lenvatinib withdrawal or dose reduction. Uses the same once-daily steady-state 24 h AUC definition as Majid 2024, but FROZEN at the starting dose rather than tracked step-wise across dose changes – necessary because the modelled event is itself the first dose change).
• Notes: Specific scope because the column meaning is tied to lenvatinib and to the once-daily steady-state 24 h AUC convention. Member of the AUC_<DRUG> family (AUC_CARBO, AUC_GEM, AUC_GCV, AUC_PAZO, AUC_RTV, AUC_VERUB, AUC_ADU, AUC_DON, AUC_GAN, AUC_LEC, AUC_IBRU, AUC_LCM, AUC_CBZ, AUC_AMPH); AUC_PAZO is the closest structural analogue, being likewise a per-period mean AUC of a tyrosine-kinase inhibitor driving a tumour-growth / angiogenesis model in place of a PK ODE, and the AUC_PAZO entry explicitly directs that a different TKI be given its own sibling canonical rather than reusing that name. A future lenvatinib model using a different exposure metric (Cmax, Ctrough, or a cumulative rather than per-interval AUC) should register a parallel canonical rather than overload AUC_LEN.

AUC_NALD (canonical for naldemedine steady-state daily AUC)

• Description: Naldemedine area under the plasma concentration-time curve over the once-daily 24 h dosing interval at steady state (AUCss), used as the exposure driver of the landmark logistic exposure-response models for spontaneous-bowel-movement response and for gastrointestinal adverse events in opioid-induced constipation. Time-fixed per subject. Kubota 2018 obtained it by empirical Bayes estimation from the companion population PK model; because naldemedine PK is linear, it is equal to Dose / (CL/F) for the individual’s apparent clearance.
• Units: ng*h/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – enters the logit untransformed and uncentred, so the regression intercept is the log-odds at AUCss = 0.
• Source aliases:
  • AUCss – Kubota 2018 Tables V and VI and Supplemental Table S4.
• Example models: Kubota_2018_naldemedine_sbm_phase2b.R, Kubota_2018_naldemedine_sbm_phase3.R (efficacy: probability of an SBM responder; slopes 0.0191 and 0.0194 per ng*h/mL), Kubota_2018_naldemedine_gi_mild_phase2b.R, Kubota_2018_naldemedine_gi_moderate_phase2b.R, Kubota_2018_naldemedine_gi_mild_phase3.R, Kubota_2018_naldemedine_gi_moderate_phase3.R, Kubota_2018_naldemedine_gi_severe_phase3.R (safety: probability of a gastrointestinal adverse event at three cumulative severity thresholds), Kubota_2018_naldemedine.R.
• Notes: Specific scope because the column meaning is tied to naldemedine and to the once-daily steady-state 24 h AUC convention. Member of the AUC_<DRUG> family (AUC_LEN, AUC_ANV, AUC_MBV_SS, AUC_VAREN, …), so it auto-approves on first use under the family-conforming-covariate rule. Two conventions in the founding paper are load-bearing for anyone assembling data for these models. First, placebo subjects carry AUC_NALD = 0 and are part of the fit – the exposure-response population is “subjects of the population PK analysis OR of the placebo group”, so the intercept is a real placebo-arm log-odds rather than an extrapolation off the low end of the observed exposure range. Second, the probabilities Kubota 2018 tabulates in Table VI for the 0.1 mg and 0.4 mg arms were not evaluated at those arms’ observed mean exposures: the Table VI footnotes state that the 0.1 mg AUC was “assumed to be half” and the 0.4 mg AUC “double” the 0.2 mg value, giving 11.06 and 44.22 in phase 2b where Supplemental Table S4 reports observed means of 10.50 and 43.76. Reproducing Table VI requires the assumed values; characterising the fitted population requires the observed ones. Observed exposures by arm (Supplemental Table S4, empirical Bayes estimates, ng*h/mL): phase 2b 0.1 mg mean 10.50 / median 10.38 / range 7.02-13.49; 0.2 mg mean 22.11 / median 20.25 / range 12.66-51.36; 0.4 mg mean 43.76 / median 38.18 / range 25.78-74.93; phase 3 0.2 mg mean 27.50 / median 24.47 / range 5.669-92.1. The phase 3 mean exceeds the phase 2b mean at the same 0.2 mg dose because the phase 3 population differs in the covariates that drive CL/F.

AUC_ANV (canonical for ainuovirine steady-state AUC over the 24 h dosing interval)

• Description: Ainuovirine (ANV) area under the plasma concentration-time curve over the once-daily 24 h dosing interval at steady state (AUCtau), used as the exposure driver of landmark logistic exposure-response regressions. Per-subject and time-fixed within the analysis: Han 2024 derives it by Bayesian post-hoc estimation from the companion population PK model reported in the same paper (Methods: “Individual steady-state trough concentration (Ctrough) and area under the steady-state curve (AUCtau) were exposure factors evaluated by Bayesian analysis”) and carries it into the exposure-response runs as a data column. Because the regimen is once daily, AUCtau is AUC(0-24) at steady state and equals AUCss numerically. The closed form under the companion PK model is F * dose / (CL/F) with the steady-state clearance, i.e. 0.716 * 150 mg / 15.96 L/h for the phase 3 regimen.
• Units: ng*h/mL (equivalently ug*h/L). Document per model via covariateData[[AUC_ANV]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – a continuous exposure regressor entering linearly and uncentred on the logit scale, beta0 + beta1 * AUC_ANV. Han 2024 states the linear form outright (“Logistic regression indicated a linear effect of Ctrough, AUCtau, and curative effect”), so the coefficient is per ng*h/mL and not per e-fold; the intercept is consequently the logit at zero exposure, an extrapolated anchor rather than a reference-patient probability. Reference values observed: Han 2024 tabulates no AUCtau distribution and does not print the quartile boundaries used for its Figures 4D and 5D, so the only anchor is the model-predicted 0.716 * 150000 ug / 15.96 L/h = about 6700 ng*h/mL for the 150 mg once-daily regimen.
• Source aliases:
  • AUCtau / AUC tau – Han 2024 Methods, Tables 4 and 5, and Figures 4B/4D/4F and 5B/5D/5F.
• Example models: Han_2024_ainuovirine_adr_auctau.R (drives the significant exposure-safety logistic regression logit(p) = -0.27 + 0.00014 * AUC_ANV for the incidence of adverse drug reactions, Han 2024 Table 5 – founding example), Han_2024_ainuovirine_virologic_auctau.R (drives the parallel, non-significant exposure-efficacy regression logit(p) = 1.44 + 0.000012 * AUC_ANV for HIV-RNA < 50 copies/mL at week 48, Han 2024 Table 4).
• Notes: Specific scope because the column meaning is tied to ainuovirine and to the once-daily steady-state 24 h AUC convention. Member of the AUC_<DRUG> family (AUC_LEN, AUC_VAREN, AUC_EMPA, AUC_RTV, AUC_VERUB, AUC_LCM, AUC_CBZ, …); AUC_RTV is the closest structural analogue, being likewise the 0-24 h interval AUC of an orally dosed antiretroviral. The units are load-bearing and Han 2024 never states them on Tables 4 or 5, so they were established from the paper’s own unit system: Figure 1D plots dose-normalised AUCinf on an axis labelled h*ng*mL^-1*mg^-1, so the un-normalised AUC is in ng*h/mL, and every ainuovirine concentration in the paper (Figure 1A/1B y-axes, the Table 3 additive residual of 8.89 ng/mL) is in ng/mL. Two internal cross-checks agree: the ratio of the paper’s Ctrough slope to its AUCtau slope on the ADR endpoint is 0.40e-2 / 0.14e-3 = 28.6 h, the right order of magnitude for an AUCtau/Ctrough ratio over a 24 h interval and not the 28600 h a 1000-fold unit error would imply; and at the exposure the companion PK model predicts for the phase 3 regimen the AUCtau and Ctrough parameterisations of the ADR endpoint agree to about one percentage point. A future ainuovirine model using a different exposure metric (Cmax, or a cumulative rather than per-interval AUC) should register a parallel canonical rather than overload AUC_ANV; the trough metric already has one in the general CTROUGH canonical.

AUC_ALIS (canonical for the time-averaged daily AUC of alisertib)

• Description: Alisertib (MLN8237) area under the plasma concentration-time curve per day, time-averaged over the treatment course up to the event, used as the exposure driver of landmark logistic exposure-safety regressions. Per-subject and time-fixed within the analysis. Zhou 2018 (Methods, “Exposure-safety analyses”) computes it for each patient “from the start of alisertib dosing up to the start of the worst grade of the toxicity of interest while on treatment”, using the actually administered doses – so dose reductions and interruptions are folded in – together with that patient’s individual empirical-Bayes CL/F from the companion population PK model. This makes it a treatment-course average, materially different from a steady-state AUC(0-tau): a patient who was dose-reduced early carries a lower AUC_ALIS than their on-label steady-state exposure, and the metric is partly a consequence of how the course unfolded.
• Units: umol*h/L/day (equivalently uM*h/day, the axis label of Zhou 2018 Figure 7). Alisertib concentrations in the Zhou 2018 analysis are molar throughout – the residual error is reported in nmol/L and the exposure tables in nM.h and uM.h – so this column is molar too. Multiply by the alisertib molar mass (518.92 g/mol) to obtain ug*h/mL/day. Document per model via covariateData[[AUC_ALIS]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – enters the logit through its natural logarithm and uncentred, logit(p) = beta0 + beta1 * log(AUC_ALIS), per Methods (“relationships between the log-transformed time-averaged alisertib AUC and incidence”). The intercept is therefore the logit at AUC_ALIS = 1 umol*h/L/day, an extrapolated anchor and not a reference-patient probability. Reference values printed in Zhou 2018 Results: 15.63 (Western patients at the 50 mg twice-daily Western MTD), 23.72 (East Asian patients at 50 mg twice daily) and 14.23 (East Asian patients at the 30 mg twice-daily regional recommended phase 2 dose). Figure 7 plots the observed distribution out to roughly 160-200 umol*h/L/day.
• Source aliases:
  • Time-averaged AUC (uM x h day^-1) – Zhou 2018 Figure 7 x-axis label.
  • AUC per day – Zhou 2018 Methods prose.
• Example models: Zhou_2018_alisertib_neutropenia.R (grade >= 3 neutropenia, logit(p) = -2.316 + 0.675 * log(AUC_ALIS), n = 591 – founding example), Zhou_2018_alisertib_stomatitis.R (grade >= 2 stomatitis, -6.485 + 1.434 * log(AUC_ALIS), n = 593; the steepest of the three), Zhou_2018_alisertib_diarrhea.R (grade >= 2 diarrhoea, -3.752 + 0.677 * log(AUC_ALIS), n = 594), Zhou_2018_alisertib.R.
• Notes: Specific scope because the column meaning is tied to alisertib and to the treatment-course-averaged daily convention. Member of the AUC_<DRUG> family (AUC_LEN, AUC_ANV, AUC_VAREN, AUC_IBRU, AUC_PAZO, …); the closest structural analogue is AUC_MBV_DAY, which is likewise a per-calendar-day rather than steady-state exposure driving a panel of logistic adverse-event regressions, and which the Sun_2025_maribavir_* family explicitly distinguishes from its own steady-state sibling AUC_MBV_SS. The same distinction applies here: do not substitute a steady-state AUC(0-tau) for this column. The three Zhou 2018 coefficient pairs are FIGURE-DERIVED (digitised from Figure 7, which is the only place the fitted relationships appear – the paper tabulates no logistic-regression parameters), and each reproduces the paper’s own printed model-estimated incidences to within the printed rounding; a future alisertib model that obtains tabulated coefficients should keep this canonical and update the example models. A future alisertib model using a different exposure metric (Cmax, Ctrough, or a steady-state AUC(0-tau)) should register a parallel canonical rather than overload AUC_ALIS.

CSS_ALIS (canonical for alisertib average steady-state plasma concentration from an upstream PK fit)

• Description: Individual average steady-state plasma concentration (Css,avg) of alisertib (MLN8237), carried as a per-subject exposure metric that drives downstream landmark logistic exposure-safety models. Not an observation. Zhou 2022 computes it per patient as Css,avg (uM) = Dose (mg) * 1000 / (CL/F (L/h) * 518.92 * tau), where Dose is the administered starting dose (not the actually administered course, so dose reductions are NOT folded in), tau is the dosing interval (24 h once daily, 12 h twice daily) and CL/F is the individual post hoc apparent clearance from the companion paediatric population PK model, including the enteric-coated tablet relative bioavailability of 0.671. Member of the CSS_<DRUG> family alongside CSS_RBV, CSS_ALBIFN, CSS_NEMTA and CSS_DFO.
• Units: umol/L (uM). Alisertib concentrations in the Zhou analyses are molar; CSS_ALIS * 518.92 / 1000 gives ug/mL (molar mass 518.92 g/mol). Document per model via covariateData[[CSS_ALIS]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – enters each logit LINEARLY and UNCENTRED, logit(p) = logit_ref + e_css_logit * CSS_ALIS, so the intercept is the logit at zero exposure. Reference values printed in Zhou 2022 Results: geometric mean 2.218 uM at the paediatric recommended dose of 80 mg/m^2 once daily (enteric-coated tablet) and 1.66 uM at the level -1 dose of 60 mg/m^2. Figure 6 plots the analysis range from about 0.7 to 24 uM.
• Source aliases:
  • Css,avg – Zhou 2022 Methods equation and Results.
  • Alisertib mean concentration (uM) – Zhou 2022 Figure 6 x-axis label.
• Example models: Zhou_2022_alisertib_stomatitis.R (grade >= 2 stomatitis, logit(p) = -2.162 + 0.2238 * CSS_ALIS, back-solved from the printed predictions – founding example), Zhou_2022_alisertib_febrile_neutropenia.R (febrile neutropenia, logit(p) = -2.194 + 0.1364 * CSS_ALIS, recovered from the Figure 6B vector curve), Zhou_2022_alisertib.R (the paediatric population PK model that generates the column).
• Notes: Registered as the parallel canonical that the AUC_ALIS entry reserves for a future alisertib model with a different exposure metric. Do NOT substitute AUC_ALIS / 24: AUC_ALIS is a treatment-course average built from actually administered doses up to the toxicity onset (Zhou 2018, adults), and enters its logits through log(), whereas this is a steady-state starting-dose metric entering linearly. The Zhou 2022 Discussion warns that, because only 12 patients received twice-daily dosing, the relationships are empirical and should not be extrapolated across regimens.

AUC_HEMO (canonical for hemoporfin AUC over the 0-30 minute photodynamic treatment window)

• Description: Hemoporfin area under the plasma concentration-time curve from time zero (start of the 20-minute intravenous infusion) to 30 minutes, AUC(0-30min), used as the exposure driver of landmark logistic exposure-response regressions for photodynamic-therapy efficacy in port-wine stain. Per-subject and time-fixed within the analysis: Chen 2025 derives it as an empirical Bayes estimate from the companion population PK model reported in the same paper and carries it into the exposure-response runs as a data column. Unusually for an AUC_<DRUG> member, the integration window is not a dosing interval – it is the photoactivation window. Chen 2025 Methods 2.8 states the rationale explicitly: hemoporfin is inert until light is applied, the 530 nm irradiation runs from 10 to 30 minutes after the start of infusion, and “the treatment time window for hemoporfin overlaps with the early phase of drug disposition”, giving “an overall treatment time window of 0-30 min”. The exposure accumulated after 30 minutes is therefore pharmacologically irrelevant to efficacy, which is why a conventional AUC(0-inf) or interval AUC is not the right driver for this drug.
• Units: h*ug/mL (equivalently mg*h/L). Note this is not the ng*h/mL convention of most AUC_<DRUG> members, and the difference is load-bearing: the coefficients are per h*ug/mL and a 1000-fold unit error would move the fitted probabilities from the published values to essentially 0 or 1. Chen 2025 states the units on Tables S1 and S2 and in Results 3.4 (mean pediatric AUC(0-30min) 12.4 h*ug/mL); the regression slopes carry the reciprocal unit tag mL*ug^-1*h^-1. Note the companion population PK model predicts concentrations in ng/mL, so a simulated concentration-time profile must be divided by 1000 before integrating to this column. Document per model via covariateData[[AUC_HEMO]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – a continuous exposure regressor entering linearly and uncentred on the logit scale, beta0 + beta1 * AUC_HEMO (Chen 2025 Equation 3). The intercept is consequently the logit at zero exposure, an extrapolated anchor rather than a reference-patient probability. Reference values observed: 12.4 h*ug/mL (cohort mean at the empirical 5 mg/kg pediatric dose), 14.9 (standardised 6 mg/kg), 15.5 (weight-stratified 7/6/5 mg/kg) and 15.7 (standardised 190 mg/m^2 BSA-based dosing) – Chen 2025 Table S2.
• Source aliases:
  • AUC 0-30min / AUC(0-30min) – Chen 2025 Methods 2.8, Results 3.4 and 3.5, Tables 3, S1 and S2, and Figures 2b, 3d-f and 4.
• Example models: Chen_2025_hemoporfin_investigator_rating.R (drives the significant exposure-efficacy regression logit(p) = -10.3 + 0.794 * AUC_HEMO for a ‘good’ or ‘excellent’ investigator rating, Chen 2025 Table S1 – founding example), Chen_2025_hemoporfin_patient_rating.R (the parallel significant patient-rating regression logit(p) = -12.9 + 0.981 * AUC_HEMO), Chen_2025_hemoporfin_almost_cured.R, Chen_2025_hemoporfin_significant_improvement.R, Chen_2025_hemoporfin_any_improvement.R, Chen_2025_hemoporfin.R.
• Notes: Specific scope because the column meaning is tied to hemoporfin and to the photoactivation-window integration limit. Member of the AUC_<DRUG> family (AUC_ANV, AUC_LEN, AUC_VAREN, AUC_RTV, AUC_VERUB, AUC_LCM, AUC_CBZ, …); AUC_ANV is the closest structural analogue, being likewise a Bayesian post-hoc AUC from a same-paper companion PK model entering an uncentred linear landmark logistic regression, with both a significant and a non-significant endpoint sharing the one column. The published Table S2 predictions are an exact arithmetic check on any implementation of this column: expit(-10.3 + 0.794 * 12.4) = 38.9% and expit(-12.9 + 0.981 * 12.4) = 32.4%, both matching Table S2 to the printed digit. A future hemoporfin model using Cmax (which Chen 2025 reports is highly linearly correlated with AUC(0-30min) and therefore dropped from the exposure-response screen) or a full AUC(0-inf) should register a parallel canonical rather than overload AUC_HEMO.

AUC_PTX (canonical for single-cycle AUC of total paclitaxel)

• Description: Individual area under the total plasma paclitaxel concentration-time curve for one treatment cycle (one 3-h infusion), used as the exposure driver of a landmark logistic exposure-safety regression. Per-subject and time-fixed within the analysis. Zhou 2020 (Methods ‘Exposure-Safety Analysis’) computed it from “the individual post hoc PK parameters and dosage”; for the linear companion population PK model this is Dose / CL_i. “Total” means liposome-encapsulated plus released (free and protein-bound) paclitaxel, the analyte measured after paclitaxel liposome (Lipusu), because the encapsulated fraction could not be separated analytically.
• Units: mg*h/L (equivalently ug*h/mL), the Figure 5 x-axis label of Zhou 2020. Document per model via covariateData[[AUC_PTX]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – enters the logit linearly and uncentred, a + b * AUC_PTX, so the intercept is the logit at zero exposure. Observed range in Zhou 2020 Figure 5 about 7.7-17 mg*h/L; typical value 11.1 mg*h/L for a 240 mg dose at the population clearance of 21.55 L/h.
• Source aliases:
  • AUC – Zhou 2020 Methods, Results ‘Exposure-Safety Analysis’ equation and Figure 5 axis (“AUC (mg*hr/L)“).
• Example models: Zhou_2020_paclitaxel_liposomal_neutropenia.R (grade >= 2 neutropenia, logit(p) = -3.5008 + 0.372 * AUC_PTX, n = 45 – founding example), Zhou_2020_paclitaxel_liposomal.R (companion PK model from which the column is computed).
• Notes: Member of the AUC_<DRUG> family. Specific to total paclitaxel after a liposomal formulation and to a single-dose (one-cycle) AUC. Do not substitute the unbound paclitaxel AUC or a time-above-threshold metric (e.g. the T > 0.05 umol/L driver of conventional-paclitaxel neutropenia models) – those are different exposure metrics and should get their own canonicals.

AUC_VAREN (canonical for varenicline steady-state daily AUC)

• Description: Varenicline area under the plasma concentration-time curve over the 24 h dosing interval at steady state, AUC(0-24)ss, used as the drug-exposure driver of exposure-response models for smoking-cessation efficacy and for nausea tolerability. Ravva 2010 Methods derives it per subject as total daily dose divided by the individual apparent clearance – “Individual 24-h daily exposure, measured as AUC(0-24)ss, was calculated as dose divided by CL/F_i (apparent clearance); the individual empirical Bayes estimate of apparent clearance was predicted from the final population PK model and parameters obtained from a pooled analysis in adult smokers” – and carries it into the exposure-response runs as a data column. Time-fixed per subject per dose level.
• Units: ng*h/mL. Document per-model via covariateData[[AUC_VAREN]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – enters the logit of the response probability additively and linearly, theta2 * AUC_VAREN, alongside a baseline intercept that the demographic covariates multiply. Set to 0 for placebo subjects (Ravva 2010 Figures 3a and 4a captions: “exposure was set to zero for the placebo group”); the drug term then vanishes exactly, leaving the covariate-adjusted baseline probability. Reference values observed: the reported reference-population quit probability of 0.562 at 1 mg b.i.d. (Ravva 2010 Figure 3b) implies a typical AUC(0-24)ss of about 186 ngh/mL at 1 mg b.i.d. and about 93 ngh/mL at 0.5 mg b.i.d.; the paper’s own exposure bins are 50-142, >142-184 and >184-408 ngh/mL, and the semiparametric fit is described as reliable only up to about 300 ngh/mL.
• Source aliases:
  • AUC(0-24)ss – the printed name in Ravva 2010 Table 3, Equation 2 and Figures 1-4. Same units and orientation as the canonical, so no value transformation.
• Example models: Ravva_2010_varenicline_car_w9_12.R (linear logit slope 0.00813 per ngh/mL on the weeks 9-12 continuous abstinence rate), Ravva_2010_varenicline_nausea.R (linear logit slope 0.00738 per ngh/mL on nausea incidence), Ravva_2010_varenicline_car_w4_7_study1.R, Ravva_2010_varenicline_car_w4_7_study2.R (the preliminary per-study weeks 4-7 fits, slopes 0.00623 and 0.00543, where exposure is the only predictor), Fediuk_2021_varenicline_nausea.R (adolescent smokers; linear logit slope 0.00911 per ng*h/mL on nausea-or-vomiting incidence, computed from the companion adolescent popPK Fediuk_2021_varenicline.R), Fediuk_2021_varenicline_car_w9_12.R (adolescent weeks 9-12 continuous abstinence; non-significant slope digitized from the source figure).
• Notes: Specific scope because the column meaning is tied to varenicline and to the 24 h steady-state daily-AUC convention. Member of the AUC_<DRUG> family (AUC_CARBO, AUC_GEM, AUC_GCV, AUC_PAZO, AUC_RTV, AUC_VERUB, AUC_ADU, AUC_DON, AUC_GAN, AUC_LEC, AUC_IBRU, AUC_LCM, AUC_CBZ, AUC_AMPH, AUC_LEN); AUC_LCM and AUC_CBZ are the closest structural analogues, being likewise per-subject daily AUCs computed from an empirical-Bayes clearance estimate of a separately published population PK model and carried into a downstream PD fit as a data column rather than solved from a PK ODE. Downstream users should compute the per-subject value from the companion Ravva_2009_varenicline.R population PK model as total daily dose / (CL/F); that model makes apparent clearance a function of creatinine clearance and race, which is what generates the between-subject exposure spread the exposure-response models consume. A future varenicline model using a different exposure metric (Cmax, Ctrough, or a cumulative rather than per-interval AUC) should register a parallel canonical rather than overload AUC_VAREN.

AUC_MPAG (canonical for dose-normalised AUC of mycophenolic acid 7-O-glucuronide (MPAG))

• Description: Per-subject, dose-normalised area under the plasma concentration-time curve of mycophenolic acid 7-O-glucuronide (MPAG), the inactive phenolic glucuronide metabolite of mycophenolic acid (MPA) formed by UGT1A9. Used as a metabolite-exposure covariate on the apparent clearance of the parent MPA. MPAG accumulates in renal impairment and displaces MPA from albumin, so a subject’s MPAG burden is a marker of both renal function and free-MPA fraction.
• Units: mg*h/L per g MMF (dose-normalised to the administered mycophenolate mofetil dose). Document per-model via covariateData[[AUC_MPAG]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as the numerator of the dose-normalised MPAG:MPA AUC ratio, which is raised to a power exponent on CL/F. Reference value observed: 588.8 mg*h/L/g (Rong 2019 as transcribed in the Maizaud 2025 S1 File mrgsolve @covariates defaults), which paired with AUC_MPA = 53 gives a ratio of 11.11.
• Source aliases:
  • AUCMPAG – Maizaud 2025 S1 File mrgsolve [PARAM] @annotated @covariates column name for the Rong 2019 model.
• Example models: Rong_2019_mycophenolic_acid.R (numerator of the (AUC_MPAG / AUC_MPA)^0.68 power term on MPA apparent clearance).
• Notes: Rong 2019 prints a single covariate – the MPAG:MPA AUC ratio – rather than the two component AUCs. The ratio is deliberately NOT registered as its own canonical: the register has no *_RATIO covariate anywhere, and founding a one-off ratio pattern for a single model was rejected in favour of joining the existing 17-member AUC_<DRUG> family (AUC_CARBO, AUC_GEM, AUC_GCV, AUC_LCM, AUC_CBZ, AUC_PAZO, AUC_RTV, AUC_ADU,…). Consuming models must form the ratio inside model() from the two members. Because both AUCs are dose-normalised by the same MMF dose, the normalisation cancels in the ratio and the internally-computed value reproduces the paper’s printed covariate exactly. The two-column form is also the more informative data contract: MPAG and MPA AUCs are separately measurable, and a future MPA/MPAG parent-metabolite model can reuse either alone. Companion canonicals AUC_MPA and ACMPAG_CC registered in the same pass.

AUC_MPA (canonical for dose-normalised AUC of mycophenolic acid (MPA))

• Description: Per-subject, dose-normalised area under the plasma concentration-time curve of mycophenolic acid (MPA), the active moiety of mycophenolate mofetil (MMF). Supplied as a measured per-subject historical-exposure covariate, not as a model output. Serves as the denominator of the dose-normalised MPAG:MPA AUC ratio used as a covariate on MPA apparent clearance.
• Units: mg*h/L per g MMF (dose-normalised to the administered mycophenolate mofetil dose). Document per-model via covariateData[[AUC_MPA]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as the denominator of the dose-normalised MPAG:MPA AUC ratio, which is raised to a power exponent on CL/F. Reference value observed: 53 mg*h/L/g (Rong 2019 as transcribed in the Maizaud 2025 S1 File mrgsolve @covariates defaults).
• Source aliases:
  • AUCMPA – Maizaud 2025 S1 File mrgsolve [PARAM] @annotated @covariates column name for the Rong 2019 model.
• Example models: Rong_2019_mycophenolic_acid.R (denominator of the (AUC_MPAG / AUC_MPA)^0.68 power term on MPA apparent clearance).
• Notes: See the AUC_MPAG entry for the rationale behind registering the two component AUCs rather than a single ratio canonical, and for the requirement that consuming models form the ratio inside model(). Note the unusual role: AUC_MPA is the exposure of the very analyte the model predicts, entering as a covariate on that analyte’s own clearance. This is legitimate only because it is a measured historical per-subject quantity in the Rong 2019 dataset, not a quantity the model computes – a model that tried to self-consistently derive it would be circular.

AUC_LUSP (canonical for luspatercept steady-state AUC over the 3-week dosing interval)

• Description: Luspatercept area under the serum concentration-time curve over the 21-day (once every 3 weeks) dosing interval at steady state, used as the exposure driver of landmark exposure-response models. Chen 2021 derives it outside the population PK run as dose / individual CL/F (empirical Bayes estimate), which is exact for the linear one-compartment luspatercept model. Time-fixed per subject in the founding models, which all use the STARTING dose (AUCss); the same paper’s time-averaged AUCavg (cumulative dose / CL/F / treatment days x 21) and event-interval AUC_TEAE metrics coincide with it only while a patient is on the starting dose.
• Units: ug*day/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – enters the regressions untransformed and uncentred, so each intercept is an extrapolation to zero exposure (placebo patients were excluded from the founding fits). Reference values observed: mean 129 ugday/mL after 1 mg/kg (Chen 2021 Results); plotted range about 13-301 ugday/mL.
• Source aliases:
  • AUCss – Chen 2021 Figures 2A, 2B, 4A and 4B.
• Example models: Chen_2021_luspatercept_hb_change.R (linear hemoglobin change over weeks 1-3), Chen_2021_luspatercept_hb_response.R (probability of a >= 1 g/dL day-21 hemoglobin increase), Chen_2021_luspatercept_teae_grade3.R and Chen_2021_luspatercept_bone_pain.R (cycle 1-2 adverse-event probabilities); the companion PK model Chen_2021_luspatercept.R generates the column as dose / CL.
• Notes: Specific scope because the column meaning is tied to luspatercept and its q3w dosing interval. Member of the AUC_<DRUG> family (AUC_NALD, AUC_LEN, …). A future luspatercept model using the time-averaged AUCavg or a cumulative AUC should register a parallel canonical rather than overload this one, because under titration-to-response dosing those metrics diverge from the starting-dose value.

AUC_FEXO (canonical for the fexofenadine AUC(2-6 h) P-glycoprotein phenotyping index)

• Description: Per-subject area under the plasma fexofenadine concentration-time curve from 2 to 6 h after an oral probe dose, used as a continuous individual P-glycoprotein (ABCB1) phenotypic-activity index. Fexofenadine is a P-gp substrate that is essentially not metabolised, so its short-window exposure is an in-vivo readout of intestinal and hepatic P-gp efflux capacity. The orientation is INVERSE: a HIGHER AUC_FEXO means LOWER P-gp activity. An encoder who reads the column as “P-gp activity” and expects it to rise with activity will get the sign of every effect backwards. The index is measured as part of a phenotyping cocktail (the Geneva cocktail of Bosilkovska 2014, which co-administers low-dose midazolam, fexofenadine, caffeine, omeprazole, dextromethorphan and losartan), so a model carrying this column will frequently also carry the cocktail’s CYP probes – most often CYP3A4, from the midazolam metabolic ratio drawn in the same session.
• Units: mg*h/L. Document per-model via covariateData[[AUC_FEXO]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – enters via a centred power form (AUC_FEXO / ref)^exponent. Reference value observed: 205 mgh/L (Gaspar 2023, the constant printed in the Table 2 final-model equation). Geneva-cocktail phenotype-category means (mean +/- SD, Bosilkovska 2014 as quoted by Gaspar 2023 Methods “Phenotyping”): 285.5 +/- 67.1 mgh/L for reduced-activity poor metabolizers, 100.1 +/- 47.5 for normal metabolizers and 50.4 +/- 15.3 for increased-activity ultra-rapid metabolizers. Those three values are the natural inputs for a simulation that wants to sweep P-gp phenotype without individual probe data.
• Source aliases:
  • P-gp AUC2-6 / AUC2-6 – the printed names in Gaspar 2023 Methods “Phenotyping”, Results “Covariate analyses” and the Table 2 note. Same units and orientation as the canonical, so no value transformation. The canonical name omits the 2-6 window subscript to fit standard data-column character constraints, following the precedent of AUC_RTV (registered without the 0-24 subscript the source prints).
• Example models: Gaspar_2023_apixaban.R (founding example; power effect (AUC_FEXO / 205)^-0.17 on the apparent oral clearance of apixaban in hospitalized adults, alongside (CRCL / 100)^0.47. P-gp phenotype explained a further 17% of the between-patient clearance variance after creatinine clearance had explained 41%, and moves clearance by -16% for a poor and +12% for an ultra-rapid metabolizer relative to a normal one).
• Notes: Registered as a member of the existing AUC_<DRUG> family (AUC_CARBO, AUC_GEM, AUC_GCV, AUC_RTV, AUC_MPA, AUC_MPAG, …) rather than as a new transporter-activity canonical such as ABCB1. Two reasons. First, the number this column carries is literally a drug AUC in mg*h/L over a stated window, and the family’s own precedent (AUC_MPAG Notes) is to join the family rather than found a one-off pattern. Second, naming the column for the transporter would hide the inverse orientation and, worse, would falsely imply that P-gp activity indices from different probes are interchangeable – they are not. A fexofenadine AUC(2-6) is not numerically comparable to a digoxin renal clearance, a talinolol AUC, or an ABCB1 mRNA-expression score, and each would need its own canonical. This is the transporter analogue of the situation the CYP3A4 and CYP2D6 continuous-activity-score entries handle by documenting the probe per model; the difference is that those entries are named for the ENZYME and carry paper-specific units, whereas this one is named for the PROBE and fixes the units, which is the safer contract when the probe determines the scale. A future paper that phenotypes P-gp with a different probe should register a parallel AUC_<PROBE> (or an appropriately named clearance canonical) rather than overload this column. Do NOT confuse with SNP_ABCB1_RS1045642, SNP_ABCB1_RS1128503 and their siblings (used in Gu_2025_rivaroxaban.R), which are germline ABCB1 GENOTYPE indicators, nor with CONMED_PGP_INH, a binary concomitant-P-gp-inhibitor flag: this column is a measured in-vivo phenotype and integrates genotype, comedication and inflammation into one number. Gaspar 2023 exploits exactly that – it evaluates no drug-drug interactions separately “as they are quantitatively integrated within the CYP3A4/5 and P-gp phenotypic activities” (Methods “Covariate model”).

AUC_OSELCARB (canonical for steady-state 24 h AUC of oseltamivir carboxylate)

• Description: Per-subject steady-state area under the plasma concentration-time curve from 0 to 24 h of oseltamivir carboxylate (OC), the active neuraminidase-inhibiting metabolite of the prodrug oseltamivir. Supplied as a static (time-fixed) per-subject exposure column to exposure-response models that consume an upstream population PK analysis without instantiating a PK ODE. Named for the METABOLITE rather than the administered prodrug (as AUC_HTBZ is), because OC and not oseltamivir carries the antiviral activity and is the exposure metric the source exposure-response analyses select; a future model driven by prodrug exposure would need a separate AUC_OSEL sibling.
• Units: ng*h/mL. Document per-model via covariateData[[AUC_OSELCARB]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a for a continuous exposure driver, but note that the founding models consume the column only through printed 3-group cutoffs rather than as a continuous slope, and that placebo records are assigned exactly 0. Reference values observed: the optimally-determined 3-group cutoff pairs 1,495 / 14,497, 1,568 / 13,638 and 0 / 14,180 ngh/mL for the three Rayner 2013 efficacy endpoints; approximately 6,000 ngh/mL as the average AUC0-24 at the approved 75 mg BID adult regimen (Rayner 2013 Discussion).
• Source aliases:
  • AUC0-24 / AUC 0-24 – the printed variable name in Rayner 2013 Tables 2, 3, 5 and 6 (written OC AUC0-24 where the analyte needs naming); same orientation, no value transformation.
• Example models: Rayner_2013_oseltamivir_symptomscore.R (linear regression on the 9-day composite influenza symptom score AUC), Rayner_2013_oseltamivir_symptomalleviation.R, Rayner_2013_oseltamivir_shedding.R (Cox proportional-hazards models on time to symptom alleviation and time to cessation of viral shedding).
• Notes: Specific scope because the column meaning is tied to this metabolite and to the steady-state q24h AUC convention. Member of the AUC_<DRUG> family (AUC_CARBO, AUC_GEM, AUC_GCV, AUC_PAZO, AUC_RTV, AUC_VERUB, AUC_ADU, AUC_DON, AUC_GAN, AUC_LEC, AUC_IBRU, AUC_LCM, AUC_CBZ, AUC_AMPH, AUC_LEN, AUC_EMPA, AUC_HTBZ), which AUC_EMPA’s Notes declare extensible; AUC_HTBZ is the closest structural analogue (also a metabolite-named steady-state interval AUC carried out of an upstream population PK run as a data column). The upstream model that generates this column is modellib('Kamal_2013_oseltamivir'). Rayner 2013 also derived OC Cmax and Cmin, but they correlated with AUC0-24 at Spearman rho above 0.96 and were not carried into the analyses; a future model that uses one of those metrics should register a CMAX / CMIN-family entry or an AUC-family sibling rather than overloading this canonical. Companion to the oselcarb metabolite suffix used by the central_oselcarb compartment in Kamal_2013_oseltamivir.R.

AUC_INH (canonical for steady-state isoniazid AUC over the 24 h dosing interval)

• Description: Per-subject steady-state area under the plasma isoniazid concentration-time curve from 0 to 24 h under once-daily first-line tuberculosis therapy. Supplied as a static (time-fixed) per-subject exposure column to a pharmacodynamic model that consumes an upstream population PK analysis without instantiating a PK ODE; in the founding model it is the individual empirical Bayes AUC0-24 from the companion isoniazid popPK model of the same cohort.
• Units: umol*h/L in the founding model (the Kloprogge 2020 PK data were molar). Convert from mg*h/L with the isoniazid molecular weight 137.14 g/mol (AUC_umol = AUC_mg * 1000 / 137.14). Document per-model via covariateData[[AUC_INH]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – used as an exponential effect centred on a cohort median, exp(theta * (AUC_INH - median)). Reference value observed: 137.3 umolh/L (Kloprogge 2020 PK-cohort median 18.83 mgh/L).
• Source aliases:
  • AUCinh – Kloprogge 2020 Supplementary S2 Table row LAM_AUCinh (‘effect of AUC isoniazid of LAM’); same orientation, no value transformation.
• Example models: Kloprogge_2020_hrze_bacterialload.R (exponential effect on the sputum bacterial kill rate LAM, coefficient 0.0005 per umol*h/L).
• Notes: Member of the AUC_<DRUG> family, which AUC_EMPA’s Notes declare extensible; sibling of AUC_RIF from the same model. The upstream model that generates this column is modellib('Kloprogge_2020_isoniazid') (dose / CL at steady state). Distinct from an AUC/MIC ratio: Kloprogge 2020 also tested isoniazid AUC0-24/MIC in the MIC subset, which should get a separate canonical if a model ever retains it.

AUC_RIF (canonical for rifampicin plasma AUC over a 24 h dosing interval)

• Description: Per-subject (time-fixed) rifampicin plasma AUC over a 0-24 h once-daily dosing interval, used as an exposure covariate in exposure-response models. In Svensson 2020 it is the individual AUC0-24 on day 2 +/- 1 of treatment (before autoinduction is complete), predicted from the empirical Bayes estimates of the companion popPK model, and it reduces the hazard of death through an Emax function.
• Units: mg*h/L (Svensson 2020). Kloprogge 2020 carries the column in umol*h/L because its PK data were molar; convert with the rifampicin molecular weight 822.94 g/mol (AUC_umol = AUC_mg * 1000 / 822.94). Each model’s covariateData[["AUC_RIF"]]$units states which.
• Type: continuous
• Scope: specific
• Reference category: n/a – enters through an Emax function 1 - AUC_RIF / (AUC50 + AUC_RIF) with no centring value. In Kloprogge 2020 it is an exponential effect centred on the PK-cohort median, exp(theta * (AUC_RIF - median)), with 35.36 umolh/L (29.10 mgh/L).
• Source aliases:
  • AUC – NONMEM $INPUT name in the Svensson 2020 survival control stream.
  • AUCrif – Kloprogge 2020 Supplementary S2 Table row T12_AUCrif (‘effect of AUC rifampicin on start time of delayed bacillary clearance’); same orientation, no value transformation.
• Example models: Svensson_2020_rifampicin_survival.R (Emax reduction of the tuberculous-meningitis mortality hazard with AUC50 = 171 mgh/L; compute the value from modellib('Svensson_2020_rifampicin') at OCC = 1 over the second daily dose), Kloprogge_2020_hrze_bacterialload.R (exponential effect on the half-time T1/2 of the fall in sputum bacterial kill rate, coefficient 0.0335 per umolh/L).
• Notes: Member of the AUC_<DRUG> family (AUC_RTV, AUC_LCM, AUC_GCV, …). Specific scope because the sampling day matters for rifampicin: autoinduction raises clearance over the first 1-2 weeks, so a day-2 AUC0-24 and a steady-state AUC0-24 for the same dose differ materially. A model that uses a steady-state or induced-state rifampicin AUC should record that in covariateData[[AUC_RIF]]$notes, or register a sibling canonical if both appear in one model. Kloprogge 2020 derives the column from modellib('Kloprogge_2020_rifampicin') (dose / CL at steady state); its sibling there is AUC_INH.

AUCU_VALE (canonical for unbound valemetostat AUC over the 24 h dosing interval at steady state)

• Description: Per-subject unbound (free, non-protein-bound) valemetostat plasma area under the concentration-time curve over the once-daily 24 h dosing interval at steady state (unbound AUCss). Supplied as a per-subject data column, not computed by the consuming model: the Fukae 2024 exposure-response analyses are static logistic regressions with no PK layer, and the individual exposure values are empirical-Bayes predictions from the companion population PK model (Fukae M, Baron K, Tachibana M, et al. “Population pharmacokinetics of total and unbound valemetostat and platelet dynamics in healthy volunteers and patients with non-Hodgkin lymphoma.” CPT Pharmacometrics Syst Pharmacol 2024, doi:10.1002/psp4.13201, PMC11494923). Valemetostat is an oral EZH1/EZH2 dual inhibitor; the unbound metric was chosen over total exposure because free drug is theoretically the moiety driving efficacy and toxicity (Fukae 2024 Discussion).
• Units: ng*h/mL. Document per-model via covariateData[[AUCU_VALE]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – enters centred and scaled as (AUCU_VALE - 375) / 250, where 375 ngh/mL is the approximate population median (the Fukae 2024 reference patient’s exposure, corresponding to the typical value at the approved 200 mg once-daily dose) and 250 ngh/mL is the approximate population standard deviation (the increment every published odds ratio is expressed per). Observed distribution: 5th-95th percentile 184-887 ngh/mL across the 102-patient R/R NHL safety analysis set; the exposure categories used for the Figure 1 graphical check are low (midpoint 175, interval 0-350), medium (450, 350-550) and high (1075, 550-1600) ngh/mL.
• Source aliases:
  • unbound AUCSS / unbound valemetostat AUCSS – Fukae 2024 Methods, Tables 2-3 and Figures 1-4. No value transformation.
  • AUCU – short NONMEM-style spelling for the same quantity.
• Example models: Fukae_2024_valemetostat_orr_central.R, Fukae_2024_valemetostat_orr_investigator.R, Fukae_2024_valemetostat_anemia.R, Fukae_2024_valemetostat_anc_decrease.R, Fukae_2024_valemetostat_plt_decrease.R, Fukae_2024_valemetostat_teae_grade3.R, Fukae_2024_valemetostat_dose_interruption.R, Fukae_2024_valemetostat_dose_reduction.R.
• Notes: Specific scope; valemetostat-specific and tied to the once-daily steady-state 24 h dosing-interval convention. Member of the AUC_<DRUG> family (AUC_CARBO, AUC_GEM, AUC_RTV, AUC_VERUB, AUC_BTP, AUC_LEN, AUC_IBRU,…), with the unboundness marked in the stem (AUCU_ rather than AUC_) by the maintainers’ decision on 2026-08-28. That choice deliberately leaves the plain AUC_VALE name free for a future total-valemetostat-exposure extraction from the companion popPK paper, honouring the AUC_EMPA entry’s standing instruction to register a parallel canonical rather than overload an existing one; it also makes the free-vs-total distinction visible at every use site, the same load-bearing distinction that separates IGE_FREE from IGE and PRED_CMAX_FREE from a total-concentration metric. A consuming model must not mix the two: the Fukae 2024 coefficients are calibrated to unbound exposure and would be badly mis-scaled if fed a total AUCss.

ACMPAG_CC (canonical for acyl mycophenolic acid glucuronide (AcMPAG) plasma concentration)

• Description: Plasma concentration of acyl mycophenolic acid glucuronide (AcMPAG), the acyl-linked glucuronide conjugate of mycophenolic acid (MPA) formed by UGT2B7. Distinct from MPAG, the inactive 7-O-phenolic glucuronide formed by UGT1A9: AcMPAG is pharmacologically active (it inhibits inosine monophosphate dehydrogenase and has been implicated in the gastrointestinal toxicity of mycophenolate mofetil). Used as a metabolite-exposure covariate on the apparent clearance of the parent MPA.
• Units: mg/L. Document per-model via covariateData[[ACMPAG_CC]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as an uncentered power term ACMPAG_CC^exponent on CL/F. Reference value observed: 0.54 mg/L (Rong 2019 as transcribed in the Maizaud 2025 S1 File mrgsolve @covariates defaults).
• Source aliases:
  • AcMPAG – Maizaud 2025 S1 File mrgsolve [PARAM] @annotated @covariates column name for the Rong 2019 model. The S1 annotation text for that column reads “Mycophenolic acid glucuronide C0” (i.e. MPAG) while both the parameter name and Maizaud 2025 Table 1 identify the acyl glucuronide; the parameter name and Table 1 are authoritative.
• Example models: Rong_2019_mycophenolic_acid.R (uncentered power term ACMPAG_CC^(-0.09) on MPA apparent clearance; because this term is uncentered, the model’s lcl is an intercept rather than a typical clearance – with the source covariate defaults the realised CL/F is 15.60 L/h, not 2.87 L/h).
• Notes: The _CC suffix follows the concentration-column convention already used by CONMED_SILDENAFIL_NMETAB_CC. Unlike that canonical, which carries the metabolite of a co-medication, ACMPAG_CC carries the study drug’s own metabolite acting on the parent’s clearance, so it is not a member of the CONMED_* family. Consuming models must state whether the value is a trough (C0) or a time-averaged concentration; Rong 2019 as transcribed does not disambiguate, and the S1 annotation’s “C0” wording is the only hint. Companion canonicals AUC_MPAG and AUC_MPA registered in the same pass.

PLAQUE_BL (canonical for per-subject baseline amyloid plaque burden (initial-condition use))

• Description: Per-subject pre-treatment amyloid plaque burden on the Centiloid scale, used as the initial condition for a plaque state variable in indirect-response amyloid plaque turnover models. Parallel to HGB_BL (per-subject baseline haemoglobin as an initial condition for haemoglobin turnover models): PLAQUE_BL is a static per-subject baseline that supplies the plaque state’s initial value at t = 0 so the trajectory is anchored to the observed baseline before any drug effect starts perturbing it. van Maanen 2025 Supplement Section ‘Further details on the exposure-response model’ specifies this pattern: ‘Individual participant “observed” baseline amyloid plaque burden (before the first dose) was an initial condition at time 0. No additional interindividual variability was included.’
• Units: CL (Centiloid; 0 = amyloid-negative, 100 = typical AD plaque burden per Klunk 2015 Centiloid consensus). Document per-model via covariateData[[PLAQUE_BL]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – subject-level baseline supplied as a covariate column. Reference values observed: 70-104 CL across the van Maanen 2025 modelling cohorts (Table S2); 71.2 CL median in the APECS individual analysis dataset. Amyloid-positivity threshold used in the van Maanen 2025 ADNI external validation was > 24 CL.
• Source aliases:
  • Baseline_Plaque – printed name in van Maanen 2025 Eq 5 (initial condition line Plaque_{t=0} = Baseline_Plaque).
• Example models: vanMaanen_2025_amyloid.R (van Maanen 2025 indirect-response plaque turnover model; sets plaque(0) <- PLAQUE_BL per subject).
• Notes: Specific scope because the initial-condition idiom is paper-defined (van Maanen 2025 specifies per-subject baseline rather than estimating a typical plaque baseline). Promote to general if a second paper ratifies the same baseline-plaque-as-initial-condition pattern (e.g., an amyloid-plaque or biomarker turnover model in an independent Alzheimer’s disease programme).

CLI (canonical for individual posthoc clearance from an upstream popPK fit)

• Description: Subject-specific empirical-Bayes (posthoc) total plasma clearance from a separately published population PK model that the current PD model treats as a fixed input. Used as a per-subject (time-fixed) covariate in PD-only models that derive a per-cycle exposure metric (e.g., AUC = DOSE / CLI) without instantiating a PK ODE.
• Units: L/h (document per-model via covariateData[[CLI]]$units).
• Type: continuous
• Scope: specific
• Reference category: n/a – appears directly in derived-exposure expressions; not a covariate effect coefficient.
• Source aliases:
  • CL – used in the Hansson 2013 sunitinib biomarker / TGI / fatigue / OS PD-model family (DDMODEL00000197 and siblings, including DDMODEL00000222 and the paper-derived Hansson_2013_sunitinib_svegfr3_os.R) and in Schindler_2016_sunitinib.R (DDMODEL00000221) as the posthoc CL column from the paper’s upstream 2-compartment popPK fit.
• Example models: Hansson_2013a_sunitinib.R (DDMODEL00000197; typical-value reference 32.819 L/h, drawn from the bundle’s simulated dataset for subject 1 – broadly consistent with Houk et al. 2010 typical sunitinib CL), Hansson_2013b_sunitinib.R (DDMODEL00000198; tumor growth inhibition variant; same per-subject CL column as Hansson 2013a/c), Hansson_2013c_sunitinib.R (DDMODEL00000222; uses a per-record CL column with subject-specific values 30-43 L/h in the bundle’s three-subject simulated dataset), Hansson_2013_sunitinib_svegfr3_os.R (parametric overall-survival Weibull TTE from Hansson 2013 e84 paper text; same per-subject CL covariate fed in alongside DOSE for the per-cycle exposure summary auc = DOSE / CLI), Schindler_2016_sunitinib.R (DDMODEL00000221; per-subject post-hoc CL fed in as the CL column, vignette uses 50 L/h literature-typical sunitinib CL/F per Houk 2010), Hansson_2013_sunitinib_myelosuppression.R (Hansson 2013 e85 paper-direct extraction; uses the same upstream Houk 2009 sunitinib popPK as the source of CL / CLI per Methods), Hansson_2013_sunitinib_dbp.R (Hansson 2013 e85; indirect-response dBP sub-model), Hansson_2013_sunitinib_hfs.R (Hansson 2013 e85; hand-foot syndrome Markov + PO sub-model paralleling Hansson_2013c fatigue), Schindler_2017_sunitinib_fatigue.R (Schindler 2017 mCTMM fatigue model; consumes the same per-subject upstream sunitinib CL as the Hansson 2013 family), Schindler_2017_sunitinib_hfs.R (Schindler 2017 mCTMM HFS model; consumes the same per-subject upstream sunitinib CL as the Hansson 2013 family), Centanni_2025_sunitinib_thrombocytopenia.R (Centanni 2025 absolute-thrombocyte-count Friberg model newly developed as an addition to the Hansson 2013 sunitinib framework; consumes the same per-subject upstream sunitinib CL and forms the identical auc = DOSE / CLI daily-exposure driver).
• Notes: Specific scope because the values are intrinsically tied to a specific upstream popPK fit (sunitinib in this case; another model adopting CLI would carry its own upstream-PK lineage). Each model’s covariateData[[CLI]]$notes should cite the upstream popPK source (paper or DDMORE ID) and explain how to populate the column for new simulations (typically: simulate the upstream popPK first to obtain individual CL, or set every subject to the typical-value CL for typical-trajectory simulations). Distinct from DOSE (current administered dose level) – the two columns jointly carry a per-cycle drug-exposure summary (AUC = DOSE / CLI) for PD-only models that consume posthoc PK from an upstream popPK fit instead of instantiating their own PK ODE.

CL_INDIV (canonical for per-subject empirical-Bayes drug clearance)

• Description: Individual point estimate of the modelled drug’s clearance, supplied per-subject as a fixed data column. Used in sequential PK->PD models where the PK structure has been fixed from a previously-published population PK analysis and the per-subject empirical-Bayes (POSTHOC) clearances are passed through to drive the PD layer rather than being re-estimated alongside the PD parameters.
• Units: L/h (document per-model via covariateData[[CL_INDIV]]$units).
• Type: continuous
• Scope: specific
• Reference category: n/a – used directly inside model as CL_INDIV in place of an estimated cl <- exp(lcl + etalcl).
• Source aliases: CLI – used in Friberg_2002_paclitaxel.R (NM-TRAN data column for per-subject paclitaxel CL).
• Example models: Friberg_2002_paclitaxel.R, Wahlby_2004_paclitaxel_myelosuppression.R (per-subject paclitaxel CL EBE supplied as a covariate column following the Friberg 2002 convention; the Wahlby 2004 PD model layers BIL and DBIL covariate effects on top of the Friberg-Karlsson chain).
• Notes: Specific scope because the value is intrinsically tied to the modelled drug – there is no shared meaning across drugs. Each model’s covariateData[[CL_INDIV]]$notes should state which upstream popPK source the EBE values come from (e.g., Henningsson 2001 paclitaxel popPK, fixed in the DDMORE encoding) and whether placebo periods are present. Companion volumes are registered as VC_INDIV and VP_INDIV.

CMAX_M1 (canonical for maximum drug plasma concentration during month 1)

• Description: Empirical-Bayes maximum plasma concentration (Cmax) reached during the first month (or the first cycle) of dosing for the modelled drug, used as a per-subject early-exposure covariate in PD / safety models that condition later-time outcomes on early exposure. Continuous, time-invariant per subject (set once from the month-1 / cycle-1 PK simulation).
• Units: ng/mL (document per-model via covariateData[[CMAX_M1]]$units if a different concentration unit is reported).
• Type: continuous
• Scope: specific
• Reference category: n/a – used as an additive logit shift (CMAX_M1 - CMAX_M1_REF) * theta (Girard 2012) or as a power scaling depending on the source. Reference value observed: 0 ng/mL (Girard 2012 sets MED17 = 0 as the centering reference).
• Source aliases:
  • CMAXM1 – used in Girard_2012_pimasertib.R.
• Example models: Girard_2012_pimasertib.R (additive logit shift on the cumulative-logit AE-score model: theta_cmaxm1 * CMAXM1).
• Notes: Specific scope because the value depends on the upstream drug-specific population-PK model used to derive the empirical-Bayes Cmax. Sibling of the existing CAV (average dosing-interval concentration); both are derived exposure metrics fed into downstream PD / safety models. Document the upstream PK model in covariateData[[CMAX_M1]]$notes for any future user.

CMAX_APTM (canonical for peak plasma concentration of APTM, the pleuromutilin derivative 14-O-[(4-amino-6-hydroxy-pyrimidine-2-yl) thioacetyl] mutilin)

• Description: APTM maximum (peak) plasma concentration after a single oral dose, used as the numerator of the Cmax/MIC PK/PD index that drives an in vivo antibacterial exposure-response model. Yang 2026 sampled plasma from Mycoplasma gallisepticum-infected chickens by LC-MS/MS, read Cmax non-compartmentally in Phoenix WinNonlin, and regressed it on dose so the index could be extrapolated to the eight doses of the efficacy arm; because that source publishes no structural compartmental PK model, the PD model consumes CMAX_APTM directly as a covariate rather than integrating a PK ODE. The model forms the PK/PD index Cmax/MIC by dividing this covariate by the challenge strain’s MIC.
• Units: ug/mL (equivalently mg/L). Note that Yang 2026 Table 2 tabulates ng/mL, so divide the tabulated value by 1000; this keeps the Cmax/MIC ratio consistent with a mic expressed in ug/mL. Document per-model via covariateData[[CMAX_APTM]]$units if a different concentration unit is reported.
• Type: continuous
• Scope: specific
• Reference category: n/a – set to 0 for untreated control records so the sigmoid term Imax * (CMAX_APTM/MIC)^gamma / (IC50^gamma + (CMAX_APTM/MIC)^gamma) vanishes and the predicted 72 h change in lung mycoplasma load reduces to the control value E0.
• Source aliases:
  • Cmax / C max – printed name in Yang 2026 Table 2 (1352.50, 4739.75 and 10882.50 ng/mL at 5, 15 and 40 mg/kg), Figure 3C, Materials and methods (‘Pharmacokinetic, pharmacodynamic, and statistical analysis’, defining the Cmax/MIC index) and Table 3. Least-squares regression of the three tabulated values on dose returns R^2 = 0.994059, reproducing the paper’s stated 0.9940 to the four decimal places it prints (the computed value rounds to 0.9941, so the paper truncated rather than rounded); inverting that regression at the Table 3 index targets reproduces the paper’s own back-calculated daily doses of 14.13 and 23.38 mg/kg to within 0.3%.
• Example models: Yang_2026_APTM_cmaxmic.R (Yang 2026 inhibitory sigmoid Emax PK/PD integration against Mycoplasma gallisepticum strain S6 in an intratracheal chicken infection model; CMAX_APTM / mic drives the signed 72 h change in lung log10 CFU/mL).
• Notes: Specific scope – APTM-specific. This is the first drug-specific member of a CMAX_<DRUG> family and is the parallel canonical that the AUC_AMPH and AUCMIC_TYLO entries direct a future Cmax/MIC-indexed antibacterial model to register rather than overloading an AUC_<DRUG> name; it is the peak-concentration twin of AUC_APTM, and the two drive the sibling models Yang_2026_APTM_cmaxmic.R and Yang_2026_APTM_aucmic.R respectively. Distinct from CMAX_M1 (an empirical-Bayes month-1 peak derived from an upstream population PK model, not tied to a named drug) and from PRED_CMAX_FREE (the peak of a co-medication rather than of the modelled drug). Distinct from the CONC_<DRUG>_<UNITS> family, which carries an instantaneous concentration driving a mechanism-based kill-rate model: a peak concentration is a summary exposure metric, not a time course. A future APTM model using a different exposure metric (Ctrough, T>MIC, or a per-interval steady-state AUC) should register a parallel canonical rather than overload this name.

CMIN_ARI (canonical for model-predicted minimum aripiprazole plasma concentration proximate to a clinical event)

• Description: Individual model-predicted minimum (trough) aripiprazole plasma concentration at the time proximate to an impending-relapse event or to censoring, generated by applying a population PK model to the subject’s recorded dosing history together with that subject’s empirical-Bayes PK parameters. Continuous; time-invariant per subject (one value, read at the event / censoring time). Used as the exposure driver of a downstream time-to-event exposure-response model.
• Units: ng/mL (document per-model via covariateData[[CMIN_ARI]]$units if a different concentration unit is reported).
• Type: continuous
• Scope: specific
• Reference category: n/a – consumed in Wang_2022_aripiprazole_relapse.R through the threshold indicator (CMIN_ARI >= 95), which shifts the constant relapse log-hazard; a model that fitted it continuously would use it directly.
• Source aliases:
  • predicted aripiprazole Cmin proximate to the event – the spelled-out name in Wang 2022 Table 2 and Methods, ‘Definition of PK Exposures’.
  • Cmin – the printed symbol in Wang 2022 Figures 5 and 6.
• Example models: Wang_2022_aripiprazole_relapse.R (exponential time-to-event model of impending relapse in schizophrenia; the dichotomy at 95 ng/mL multiplies the hazard by exp(-1.484) = 1/4.41, and the upstream population PK model that generates the column is modellib('Wang_2022_aripiprazole')).
• Notes: Specific scope because the value is intrinsically tied to aripiprazole and to the upstream population PK model used to derive it; document that upstream model in each model’s covariateData[[CMIN_ARI]]$notes, along with the window over which the trough is read (Wang 2022 uses 24 h after the preceding oral dose when the event falls in the first AOM dose period and 672 h after the preceding AOM dose thereafter). This is the first member of a CMIN_<DRUG> family, registered rather than overloading a neighbouring canonical exactly as the AUC_HTBZ Notes (“a future model that uses one of those metrics should register a CMAX / CMIN-family entry or an AUC-family sibling rather than overloading this canonical”) and the CMAX_APTM Notes (“a future APTM model using a different exposure metric (Ctrough, T>MIC, or a per-interval steady-state AUC) should register a parallel canonical”) direct. It is the trough twin of the registered CMAX_<DRUG> family and a sibling of CAV (dosing-interval average), CMAX_M1 (empirical-Bayes month-1 peak) and the CSS_<DRUG> family (steady-state average from an upstream PK fit); the distinguishing feature is that a Cmin is the extreme of the interval rather than its average, which is what makes it the natural driver when efficacy is expected to fail at the end of a long dosing interval. Distinct from the CONC_<DRUG>_<UNITS> family, which carries an instantaneous concentration time course into a mechanism-based model rather than a summary exposure metric. A placebo record is NOT automatically 0: Wang 2022 predicted nonzero troughs for placebo subjects because washout of pre-randomisation aripiprazole was incomplete (minimum 0.000329 ng/mL), so a model reusing this canonical must state whether placebo records carry a predicted or a zeroed value.

VC_INDIV (canonical for per-subject empirical-Bayes central volume of distribution)

• Description: Individual point estimate of the modelled drug’s central volume of distribution, supplied per-subject as a fixed data column. Companion to CL_INDIV in sequential PK->PD encodings.
• Units: L (document per-model via covariateData[[VC_INDIV]]$units).
• Type: continuous
• Scope: specific
• Reference category: n/a – used directly inside model as VC_INDIV in place of an estimated vc <- exp(lvc + etalvc).
• Source aliases: V1I – used in Friberg_2002_paclitaxel.R (NM-TRAN data column for per-subject paclitaxel V1).
• Example models: Friberg_2002_paclitaxel.R, Wahlby_2004_paclitaxel_myelosuppression.R (per-subject paclitaxel V1 EBE following the Friberg 2002 convention).
• Notes: See CL_INDIV notes for the broader convention.

VP_INDIV (canonical for per-subject empirical-Bayes peripheral volume of distribution)

• Description: Individual point estimate of the modelled drug’s first peripheral volume of distribution, supplied per-subject as a fixed data column. Companion to CL_INDIV and VC_INDIV in sequential PK->PD encodings.
• Units: L (document per-model via covariateData[[VP_INDIV]]$units).
• Type: continuous
• Scope: specific
• Reference category: n/a – used directly inside model as VP_INDIV in place of an estimated vp <- exp(lvp + etalvp).
• Source aliases: V2I – used in Friberg_2002_paclitaxel.R (NM-TRAN data column for per-subject paclitaxel V2).
• Example models: Friberg_2002_paclitaxel.R, Wahlby_2004_paclitaxel_myelosuppression.R (per-subject paclitaxel V2 EBE following the Friberg 2002 convention).
• Notes: See CL_INDIV notes for the broader convention. For models requiring a second peripheral compartment, register VP2_INDIV (and add a follow-on entry to this register) when a second model legitimately needs it.

BAS_SVEGFR3 (canonical for individual posthoc baseline soluble VEGFR-3 concentration from an upstream PD fit)

• Description: Subject-specific empirical-Bayes (posthoc) baseline plasma sVEGFR-3 (soluble vascular endothelial growth factor receptor 3) concentration from a separately published indirect-response biomarker model that the current downstream PD model treats as a fixed input. Used as a per-subject (time-fixed) covariate in fatigue / adverse-event PD models that consume the upstream biomarker dynamics as data covariates without instantiating the biomarker ODE.
• Units: pg/mL (document per-model via covariateData[[BAS_SVEGFR3]]$units).
• Type: continuous
• Scope: specific
• Reference category: n/a – appears directly inside model as the initial condition svegfr3(0) <- BAS_SVEGFR3 and inside the relative-change driver bm = (svegfr3 - BAS_SVEGFR3) / BAS_SVEGFR3.
• Source aliases:
  • BAS3 – used in Hansson_2013b_sunitinib.R (DDMODEL00000198), Hansson_2013c_sunitinib.R (DDMODEL00000222), and Hansson_2013_sunitinib_svegfr3_os.R as the posthoc sVEGFR-3 baseline column from the paper’s upstream Hansson 2013a biomarker indirect-response fit (DDMODEL00000197).
• Example models: Hansson_2013b_sunitinib.R (DDMODEL00000198; tumor growth inhibition with sVEGFR-3 driven shrinkage), Hansson_2013c_sunitinib.R (DDMODEL00000222; bundle’s three-subject simulated dataset reports BAS_SVEGFR3 values 42554-57365 pg/mL), Hansson_2013_sunitinib_svegfr3_os.R (parametric overall-survival Weibull TTE from Hansson 2013 e84 paper text; sVEGFR-3 dynamics simulated inline using the upstream biomarker-PD per-subject parameters), Hansson_2013_sunitinib_myelosuppression.R (Hansson 2013 e85 paper-direct extraction; simulates sVEGFR-3 turnover in-model to drive the Emax effect on neutrophil proliferation), Hansson_2013_sunitinib_hfs.R (Hansson 2013 e85; hand-foot syndrome Markov + PO sub-model driven by sVEGFR-3 relative change through an effect compartment), Schindler_2017_sunitinib_fatigue.R (Schindler 2017 mCTMM fatigue model; sVEGFR-3 baseline consumed as the upstream Hansson 2013a per-subject BAS_SVEGFR3 posthoc), Schindler_2017_sunitinib_hfs.R (Schindler 2017 mCTMM HFS model; sVEGFR-3 baseline consumed as the upstream Hansson 2013a per-subject BAS_SVEGFR3 posthoc).
• Notes: Specific scope because the value is intrinsically tied to a specific upstream biomarker model (sVEGFR-3 indirect response under sunitinib in this case). The downstream fatigue model only consumes individual posthoc baseline / MRT / EC50 of the upstream biomarker; it does not re-fit them. Each model’s covariateData[[BAS_SVEGFR3]]$notes should cite the upstream biomarker-PD source (paper or DDMORE ID) and explain how to populate the column for new simulations (typically: simulate from the upstream biomarker model to obtain individual posthoc baselines, or set every subject to the typical-value baseline for typical-trajectory simulations).

MRT_SVEGFR3 (canonical for individual posthoc mean residence time of soluble VEGFR-3 from an upstream PD fit)

• Description: Subject-specific empirical-Bayes (posthoc) mean residence time of plasma sVEGFR-3 from a separately published indirect-response biomarker model that the current downstream PD model treats as a fixed input. Used as a per-subject (time-fixed) covariate in fatigue / adverse-event PD models that consume the upstream biomarker dynamics as data covariates without instantiating the biomarker ODE; appears as kout3 = 1 / MRT_SVEGFR3 inside model.
• Units: h (hours) – document per-model via covariateData[[MRT_SVEGFR3]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – appears directly in derived rate-constant expressions; not a covariate effect coefficient.
• Source aliases:
  • MRT3 – used in Hansson_2013b_sunitinib.R (DDMODEL00000198), Hansson_2013c_sunitinib.R (DDMODEL00000222), and Hansson_2013_sunitinib_svegfr3_os.R as the posthoc sVEGFR-3 MRT column from the paper’s upstream Hansson 2013a biomarker indirect-response fit (DDMODEL00000197).
• Example models: Hansson_2013b_sunitinib.R (DDMODEL00000198), Hansson_2013c_sunitinib.R (DDMODEL00000222; bundle’s three-subject simulated dataset reports MRT_SVEGFR3 values 313-408 h, broadly consistent with the Hansson 2013a typical sVEGFR-3 MRT of 401 h), Hansson_2013_sunitinib_svegfr3_os.R (parametric overall-survival Weibull TTE; sVEGFR-3 dynamics simulated inline using the upstream biomarker-PD per-subject MRT), Hansson_2013_sunitinib_myelosuppression.R (Hansson 2013 e85 paper-direct extraction), Hansson_2013_sunitinib_hfs.R (Hansson 2013 e85 paper-direct extraction), Schindler_2017_sunitinib_fatigue.R (Schindler 2017 mCTMM fatigue; sVEGFR-3 MRT consumed as the upstream Hansson 2013a per-subject MRT_SVEGFR3 posthoc), Schindler_2017_sunitinib_hfs.R (Schindler 2017 mCTMM HFS; sVEGFR-3 MRT consumed as the upstream Hansson 2013a per-subject MRT_SVEGFR3 posthoc).
• Notes: Specific scope; same upstream-biomarker dependency rationale as BAS_SVEGFR3. The downstream fatigue model consumes the upstream MRT directly without re-fitting it.

EC50_SVEGFR3 (canonical for individual posthoc drug-effect EC50 on soluble VEGFR-3 from an upstream PD fit)

• Description: Subject-specific empirical-Bayes (posthoc) half-maximum-effect concentration of the modelled drug on the production of sVEGFR-3 (or, depending on the upstream model parameterization, on the analogous Imax inhibition pathway) from a separately published indirect-response biomarker model that the current downstream PD model treats as a fixed input. Used as a per-subject (time-fixed) covariate in fatigue / adverse-event PD models that consume the upstream biomarker dynamics as data covariates without instantiating the biomarker ODE.
• Units: mgh/L (when the per-cycle drug-exposure summary is auc = DOSE / CLI in mgh/L, EC50_SVEGFR3 carries the same units; document per-model via covariateData[[EC50_SVEGFR3]]$units).
• Type: continuous
• Scope: specific
• Reference category: n/a – appears directly inside model in the simple-Imax drug-effect term eff3 = auc / (EC50_SVEGFR3 + auc).
• Source aliases:
  • EC53 – used in Hansson_2013b_sunitinib.R (DDMODEL00000198), Hansson_2013c_sunitinib.R (DDMODEL00000222), and Hansson_2013_sunitinib_svegfr3_os.R as the posthoc sVEGFR-3 EC50 column from the paper’s upstream Hansson 2013a biomarker indirect-response fit (DDMODEL00000197).
• Example models: Hansson_2013b_sunitinib.R (DDMODEL00000198), Hansson_2013c_sunitinib.R (DDMODEL00000222; bundle’s three-subject simulated dataset reports EC50_SVEGFR3 values 1.0-2.8 mgh/L, consistent with the Hansson 2013a typical sVEGFR-3 IC50 typical value of 1.0 mgh/L), Hansson_2013_sunitinib_svegfr3_os.R (parametric overall-survival Weibull TTE; sVEGFR-3 dynamics simulated inline using the upstream biomarker-PD per-subject EC50), Hansson_2013_sunitinib_myelosuppression.R (Hansson 2013 e85 paper-direct extraction), Hansson_2013_sunitinib_hfs.R (Hansson 2013 e85 paper-direct extraction), Schindler_2017_sunitinib_fatigue.R (Schindler 2017 mCTMM fatigue; sVEGFR-3 EC50 consumed as the upstream Hansson 2013a per-subject EC50_SVEGFR3 posthoc), Schindler_2017_sunitinib_hfs.R (Schindler 2017 mCTMM HFS; sVEGFR-3 EC50 consumed as the upstream Hansson 2013a per-subject EC50_SVEGFR3 posthoc).
• Notes: Specific scope; same upstream-biomarker dependency rationale as BAS_SVEGFR3. The downstream fatigue model consumes the upstream EC50 directly without re-fitting it.

BAS_SKIT (canonical for individual posthoc baseline soluble KIT concentration from an upstream PD fit)

• Description: Subject-specific empirical-Bayes (posthoc) baseline plasma sKIT (soluble stem cell factor receptor) concentration from a separately published indirect-response biomarker model that the current downstream PD model treats as a fixed input. Used as a per-subject (time-fixed) covariate in tumor-growth-inhibition / fatigue / adverse-event PD models that consume the upstream sKIT dynamics as data covariates without instantiating the biomarker ODE for sKIT in isolation.
• Units: pg/mL (document per-model via covariateData[[BAS_SKIT]]$units).
• Type: continuous
• Scope: specific
• Reference category: n/a – appears directly inside model as the initial condition for treated and placebo sKIT compartments and inside the relative-change driver (skit_pla - skit_drug) / skit_pla (or the analogous BAS_SKIT-denominated form when only one sKIT compartment is simulated).
• Source aliases:
  • SBAS – used in Hansson_2013b_sunitinib.R (DDMODEL00000198) as the posthoc sKIT baseline column from the paper’s upstream Hansson 2013a biomarker indirect-response fit (DDMODEL00000197).
• Example models: Hansson_2013b_sunitinib.R (DDMODEL00000198; the Hansson 2013 e84 paper Table 2 reports a typical sKIT baseline of 39200 pg/mL with ~50% CV, matching Hansson_2013a_sunitinib’s typical value).
• Notes: Specific scope because the value is intrinsically tied to a specific upstream biomarker model (sKIT indirect response under sunitinib in this case). The downstream tumor-growth-inhibition model only consumes individual posthoc baseline / MRT / EC50 / DP-slope of the upstream biomarker; it does not re-fit them. Each model’s covariateData[[BAS_SKIT]]$notes should cite the upstream biomarker-PD source (paper or DDMORE ID) and explain how to populate the column for new simulations (typically: simulate from the upstream biomarker model to obtain individual posthoc baselines, or set every subject to the typical-value baseline for typical-trajectory simulations). Sister covariates: MRT_SKIT, EC50_SKIT, SLOPE_SKIT (companions for the same upstream biomarker fit).

MRT_SKIT (canonical for individual posthoc mean residence time of soluble KIT from an upstream PD fit)

• Description: Subject-specific empirical-Bayes (posthoc) mean residence time of plasma sKIT from a separately published indirect-response biomarker model that the current downstream PD model treats as a fixed input. Used as a per-subject (time-fixed) covariate in tumor-growth-inhibition / fatigue / adverse-event PD models that consume the upstream sKIT dynamics as data covariates without instantiating the biomarker ODE for sKIT in isolation; appears as kout_skit = 1 / MRT_SKIT inside model.
• Units: h (hours) – document per-model via covariateData[[MRT_SKIT]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – appears directly in derived rate-constant expressions; not a covariate effect coefficient.
• Source aliases:
  • SMRT – used in Hansson_2013b_sunitinib.R (DDMODEL00000198) as the posthoc sKIT MRT column from the paper’s upstream Hansson 2013a biomarker indirect-response fit (DDMODEL00000197).
• Example models: Hansson_2013b_sunitinib.R (DDMODEL00000198; the Hansson 2013 e84 paper Table 2 reports a typical sKIT MRT of 101 days = 2424 h, matching Hansson_2013a_sunitinib’s typical value of 2430 h).
• Notes: Specific scope; same upstream-biomarker dependency rationale as BAS_SKIT. The downstream tumor-growth-inhibition model consumes the upstream MRT directly without re-fitting it.

EC50_SKIT (canonical for individual posthoc drug-effect EC50 on soluble KIT from an upstream PD fit)

• Description: Subject-specific empirical-Bayes (posthoc) half-maximum-effect concentration of the modelled drug on the production of sKIT (or, depending on the upstream model parameterization, on the analogous Imax inhibition pathway) from a separately published indirect-response biomarker model that the current downstream PD model treats as a fixed input. Used as a per-subject (time-fixed) covariate in tumor-growth-inhibition / fatigue / adverse-event PD models that consume the upstream sKIT dynamics as data covariates without instantiating the biomarker ODE for sKIT in isolation.
• Units: mgh/L (when the per-cycle drug-exposure summary is auc = DOSE / CLI in mgh/L, EC50_SKIT carries the same units; document per-model via covariateData[[EC50_SKIT]]$units).
• Type: continuous
• Scope: specific
• Reference category: n/a – appears directly inside model in the simple-Imax drug-effect term eff_skit = auc / (EC50_SKIT + auc).
• Source aliases:
  • SEC5 – used in Hansson_2013b_sunitinib.R (DDMODEL00000198) as the posthoc sKIT EC50 column from the paper’s upstream Hansson 2013a biomarker indirect-response fit (DDMODEL00000197).
• Example models: Hansson_2013b_sunitinib.R (DDMODEL00000198; the Hansson 2013 e84 paper Table 2 reports a typical (common across all four biomarkers) IC50 of 1.0 mg*h/L, matching Hansson_2013a_sunitinib’s shared typical value).
• Notes: Specific scope; same upstream-biomarker dependency rationale as BAS_SKIT. The downstream tumor-growth-inhibition model consumes the upstream EC50 directly without re-fitting it.

SLOPE_SKIT (canonical for individual posthoc linear disease-progression slope on soluble KIT from an upstream PD fit)

• Description: Subject-specific empirical-Bayes (posthoc) linear disease-progression slope on the placebo / untreated sKIT compartment from a separately published indirect-response biomarker model that the current downstream PD model treats as a fixed input. Used as a per-subject (time-fixed) covariate in tumor-growth-inhibition models that need to simulate the placebo-arm sKIT trajectory as a comparator for the drug-arm sKIT compartment; appears as dps_skit = BAS_SKIT * (1 + SLOPE_SKIT * t) and kin_skit = dps_skit * kout_skit inside model.
• Units: 1/h (document per-model via covariateData[[SLOPE_SKIT]]$units).
• Type: continuous
• Scope: specific
• Reference category: n/a – appears directly in the placebo-arm Kin expression for sKIT; not a covariate effect coefficient on a structural rate.
• Source aliases:
  • SLO – used in Hansson_2013b_sunitinib.R (DDMODEL00000198) as the posthoc sKIT linear disease-progression slope column from the paper’s upstream Hansson 2013a biomarker indirect-response fit (DDMODEL00000197).
• Example models: Hansson_2013b_sunitinib.R (DDMODEL00000198; the Hansson 2013 e84 paper Table 2 reports a typical disease-progression slope of 0.0261/month shared between VEGF and sKIT, which equals approximately 3.5e-5/h, matching Hansson_2013a_sunitinib’s typical value).
• Notes: Specific scope; same upstream-biomarker dependency rationale as BAS_SKIT. Sign convention follows the upstream biomarker fit – positive slope means the placebo / natural-history sKIT trajectory drifts upward over time (capturing disease progression).

CSS_RBV (canonical for individual posthoc ribavirin steady-state plasma concentration from an upstream PK fit)

• Description: Subject-specific empirical-Bayes (posthoc) modal estimate of the ribavirin (RBV) steady-state plasma trough concentration from a separately fitted population PK model (e.g., the Laouenan 2015 upstream RBV popPK fit) that the current downstream PD model treats as a fixed input. Used together with K_RBV inside model to reconstruct the individual ribavirin concentration time-course analytically as riba(t) = CSS_RBV * (1 - exp(-K_RBV * t)).
• Units: ng/mL (document per-model via covariateData[[CSS_RBV]]$units).
• Type: continuous
• Scope: specific
• Reference category: n/a – appears directly inside model in the analytical RBV-concentration expression riba = CSS_RBV * (1 - exp(-K_RBV * t)) that drives the inhibition term riba / (riba + ec50).
• Source aliases:
  • css_mode (modal posterior estimate of individual ribavirin Css from the upstream popPK fit, in ng/mL) – used in Laouenan_2015_ribavirin.R (DDMODEL00000285). Rename css_mode -> CSS_RBV before passing the dataset to rxSolve.
• Example models: Laouenan_2015_ribavirin.R (DDMODEL00000285; the bundle’s Simulated_Laouenant_2015_CPTPSP_hb_RBV.txt carries css_mode values 2,400-4,000 ng/mL in the 15-subject ANRS-CO20-CUPIC cohort).
• Notes: Specific scope because the value is intrinsically tied to a specific upstream popPK fit (ribavirin in HCV-cirrhotic patients on triple therapy in this case) and to the analytical Css*(1-exp(-k*t)) parameterization; another drug or another popPK parameterization would carry its own canonical. Each model’s covariateData[[CSS_RBV]]$notes should cite the upstream popPK source (paper or DDMORE ID) and explain how to populate the column for new simulations (typically: simulate the upstream popPK first to obtain individual Css and approach-rate, or set every subject to the typical Css for typical-trajectory simulations). Companion column: K_RBV. Distinct from CAV (dosing-interval-averaged concentration used in Emax / EC50 PD models with a single per-period exposure number) – CSS_RBV carries the asymptotic steady-state value paired with the approach-rate constant, supporting the full time-course reconstruction.

K_RBV (canonical for individual posthoc ribavirin approach-to-steady-state rate constant from an upstream PK fit)

• Description: Subject-specific empirical-Bayes (posthoc) modal estimate of the first-order rate constant governing the exponential approach of ribavirin plasma trough concentrations to steady state, from the same upstream popPK fit that supplies CSS_RBV. Used together with CSS_RBV inside model to reconstruct the individual ribavirin concentration time-course analytically as riba(t) = CSS_RBV * (1 - exp(-K_RBV * t)).
• Units: 1/day (document per-model via covariateData[[K_RBV]]$units).
• Type: continuous
• Scope: specific
• Reference category: n/a – appears directly inside the analytical RBV-concentration expression alongside CSS_RBV.
• Source aliases:
  • k_mode (modal posterior estimate of individual ribavirin approach-to-Css rate constant from the upstream popPK fit, in 1/day) – used in Laouenan_2015_ribavirin.R (DDMODEL00000285). Rename k_mode -> K_RBV before passing the dataset to rxSolve.
• Example models: Laouenan_2015_ribavirin.R (DDMODEL00000285; the bundle’s Simulated_Laouenant_2015_CPTPSP_hb_RBV.txt carries k_mode values 0.013-0.47 day^-1 across the 15-subject cohort, corresponding to approach-to-Css half-lives of 1.5-55 days).
• Notes: Specific scope; same upstream-PK-dependency rationale as CSS_RBV. Distinct from any structural kel PK parameter – K_RBV is an apparent approach-to-Css rate from a lumped exponential parameterization of the trough time-course, not the elimination-rate constant of a one-compartment IV model (the lumped form absorbs absorption, distribution, and elimination into a single first-order rate). Companion column: CSS_RBV.

CSS_ALBIFN (canonical for albinterferon alfa-2b average steady-state serum concentration from an upstream PK fit)

• Description: Individual average steady-state serum concentration (Cavg) of albinterferon alfa-2b, carried as a per-subject exposure metric that drives a downstream exposure-response model. Not an observation: it is derived from each patient’s Bayesian maximum a posteriori PK parameter estimates as the cumulative AUC over the patient’s entire dosing history divided by the treatment duration, so it reflects dose reductions and early discontinuation. Member of the CSS_<DRUG> family alongside CSS_RBV and CSS_DFO.
• Units: ng/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – enters an exposure-response model as a centred linear term on the logit scale, e_css_albifn_logitsvr * (CSS_ALBIFN - 75). Reference value observed: 75 ng/mL (Riggs 2012, the hypothetical reference individual of Table II footnote a).
• Source aliases:
  • Cavg – Riggs 2012 notation throughout Methods, Results, Table II, and Figure 3.
• Example models: Riggs_2012_albinterferon_svr_gt1.R, Riggs_2012_albinterferon_svr_gt23.R (the sole exposure term of each genotype stratum’s sustained-virologic-response logistic regression; logit slopes 0.0063627 and 0.0098258 per ng/mL respectively), Riggs_2012_albinterferon.R.
• Notes: Riggs 2012 also computed a first-dose Cmax per patient but deliberately EXCLUDED it from the SVR exposure-response, because albinterferon is an indirect antiviral acting through receptor binding over time, which makes a time-averaged metric the mechanistically relevant one (Methods p. 478); Cmax was retained only for the tabular safety exposure-response. Do not substitute a Cmax column here. Observed medians of the individual Cavg estimates were 62.7 ng/mL at 900 ug every 2 weeks and 83.0 ng/mL at 1200 ug every 2 weeks, and the whole-cohort quartile boundaries used for the safety tabulation were 22.4 / 54.4 / 68.4 / 85.9 / 176 ng/mL. Specific scope because the concentration scale is drug-specific; the CSS_<DRUG> family exists precisely so that each drug’s steady-state exposure metric keeps its own column rather than several drugs sharing an ambiguous bare CSS.

CSS_NEMTA (canonical for nemtabrutinib average on-treatment plasma concentration from an upstream PK fit)

• Description: Individual average on-treatment plasma concentration (Cavg) of nemtabrutinib, carried as a per-subject exposure metric that drives a downstream exposure-response model. Not an observation: Kemal 2026 simulated each participant’s concentration-time profile from the companion population PK model using that participant’s actual dosing history and their individual post hoc PK parameter estimates, then computed Cavg as the cumulative on-treatment AUC divided by the treatment duration – so it reflects dose interruptions, dose reductions and early discontinuation, and it is an on-treatment average rather than a strict steady-state value. Member of the CSS_<DRUG> family alongside CSS_RBV, CSS_ALBIFN and CSS_DFO.
• Units: ng/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – enters the exposure-response models as an UNCENTRED linear term on the logit scale, e_css_nemta_logitbor * CSS_NEMTA, so that each model’s intercept is the raw glm intercept printed (or digitized) for it. This differs deliberately from CSS_ALBIFN, which is centred on a reference individual’s exposure because its source tabulates fitted probabilities rather than coefficients. Observed distribution (Kemal 2026 Figures 3 and 4 reference boxplots): median about 600 ng/mL at 45 mg daily, about 750 ng/mL at 65 mg and about 850 ng/mL at 80 mg, with the pooled range extending to about 2160 ng/mL.
• Source aliases:
  • Cavg – Kemal 2026 notation throughout Methods 2.3 and 2.4, Results, Table S4, and the Figure 2/3/4 axis labels.
• Example models: Kemal_2026_nemtabrutinib_bor.R (the exposure term of the best-overall-response logistic regression in CLL/SLL; logit slope 0.0019 per ng/mL, Table S4), Kemal_2026_nemtabrutinib_ae.R (any-grade drug-related adverse events; logit slope 0.001003 per ng/mL, digitized from Figure 4), Kemal_2026_nemtabrutinib_hypertension.R (any-grade hypertension; logit slope 0.0007645 per ng/mL, digitized from Figure 4), Kemal_2026_nemtabrutinib.R (the population PK model that generates the column).
• Notes: Kemal 2026 also computed a per-patient Cmax over the whole treatment course and reports that the two metrics were significantly correlated, but selected Cavg as the predictor for efficacy, safety, tolerability and the covariate forest-plot simulations (Results 3.4). Do not substitute a Cmax column here. Note that the same column carries a different slope for each of the three endpoints, and that time on treatment accompanies it in the efficacy model (through a saturable term with ET50 = 200 days) but was screened and rejected for both safety endpoints – so the exposure metric is shared across the sister models while nothing else is. Specific scope because the concentration scale is drug-specific; the CSS_<DRUG> family exists precisely so that each drug’s exposure metric keeps its own column rather than several drugs sharing an ambiguous bare CSS.

CSSU_VALE (canonical for unbound valemetostat average on-treatment plasma concentration up to event from an upstream PK fit)

• Description: Individual average unbound (free, non-protein-bound) plasma concentration of valemetostat up to the event of interest – the paper’s Cavgtte, “unbound Cavg up to the time-to-event”. Carried as a per-subject exposure metric that drives a downstream exposure-response model; it is not an observation. Inoue 2025 simulated each participant’s unbound concentration-time profile from the companion population PK model using that participant’s actual dosing history and individual post hoc PK parameter estimates, then averaged over the interval from first dose to the endpoint event (or to the end of treatment for subjects without an event), so the value reflects dose interruptions, dose reductions and differing follow-up. Because the averaging window ends at the event, the window differs between endpoints for the same subject: the Table S5 column is Cavg up to anemia, the Table S9 column is Cavg up to dose reduction, and so on. Valemetostat is an oral EZH1/EZH2 dual inhibitor. Member of the CSS_<DRUG> family alongside CSS_RBV, CSS_NEMTA, CSS_ALBIFN, CSS_DFO and CSS_LEC, with the unboundness marked in the stem.
• Units: ng/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – in six of the seven models it enters centred and scaled as (CSSU_VALE - m) / s, with constants that differ between the efficacy and safety analysis sets because they are different cohorts: efficacy (N = 119 PTCL patients) m = 13.9, s = 12.2 ng/mL; safety (N = 251 pooled PTCL + ATLL) m = 17.7, s = 13.7 ng/mL. The efficacy centre is the reference subject exposure printed in the Figure S6 caption and the efficacy scale is printed in the Results narrative (“every 12.2 ng/mL increase”); the Figure S8 caption prints no reference subject, so the safety constants were recovered by exact algebra from the forest panels. One exception: the grade >= 3 neutropenia model (Table S6) does not standardize the exposure at all. It uses the saturating variable E = C / (ED50 + C) with ED50 = 39.7 ng/mL, both for the main exposure effect (coefficient Emax) and as the multiplier of every interaction term, so E = 0.313 at the reference exposure and its interaction terms do not vanish at the reference subject. See Inoue_2025_valemetostat_neutropenia.R, whose ini() records the scale-free binary-row evidence for that reading.
• Source aliases:
  • Cavgtte – Inoue 2025 Tables S4-S10 and Figures S6/S8 axis labels. No value transformation.
  • unbound Cavg up to response / up to anemia / up to dose reduction etc. – the same column named for each endpoint’s own event window.
• Example models: Inoue_2025_valemetostat_orr_bicr.R, Inoue_2025_valemetostat_anemia.R, Inoue_2025_valemetostat_neutropenia.R, Inoue_2025_valemetostat_thrombocytopenia.R, Inoue_2025_valemetostat_teae_grade3.R, Inoue_2025_valemetostat_dose_reduction.R, Inoue_2025_valemetostat_dose_interruption.R, Inoue_2025_valemetostat.R.
• Notes: Specific scope; the concentration scale is drug-specific and the CSS_<DRUG> family exists precisely so each drug keeps its own column. The unboundness is marked in the stem (CSSU_ rather than CSS_) following the maintainers’ decision of 2026-08-28 that established AUCU_VALE for the same drug, which deliberately left the plain AUC_VALE free for a future total-exposure metric; CSS_VALE is likewise reserved here for a total-concentration metric. A consuming model must not mix the two – the Inoue 2025 coefficients are calibrated to unbound concentrations, which are roughly 2.6% of total at the reference AAG, so feeding a total-concentration column would mis-scale every exposure effect by more than an order of magnitude. Distinct from AUCU_VALE, which is the Fukae 2024 metric: that one is an unbound AUC over the 24 h dosing interval at steady state in ng*h/mL, whereas this is an unbound average concentration over the on-treatment window up to the event in ng/mL. The two are different quantities on different scales from different papers and are not interchangeable; the numerical relationship Cavg = AUCtau / tau holds only for a subject at true steady state with no interruptions.

CSS_DFO (canonical for deferoxamine average steady-state plasma concentration from an upstream PK fit)

• Description: Subject-specific (or time-varying) average steady-state plasma concentration of deferoxamine (DFO; iron chelator). In Bellanti 2015 it is generated externally by a literature-derived two-compartment, zero-order-absorption (8-h SC infusion), first-order-elimination PK model (CL/F 19.3 L/h, Q/F 17.6 L/h, V/F 77.4 L, Vp/F 238 L at adult 70-kg reference; allometric exponents 0.75 on clearances and 1.00 on volumes) and fed into the downstream ferritin disease model as the drug-exposure driver in the linear concentration-effect term DFO = SLP * CSS_DFO on the ferritin degradation rate. Treated as time-varying so that drug holidays / partial compliance are captured by setting CSS_DFO = 0 (or a reduced value) over the affected interval – the paper’s compliance-corrected effective concentration TCss_AV = SCss_AV * (1 - CMPL) collapses to a CSS_DFO scaling in this implementation.
• Units: ug/mL (equivalently mg/L; the reporting convention used in Bellanti 2015 Fig 1 and Fig 3). Document per-model via covariateData[[CSS_DFO]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – enters directly into the drug-effect expression DFO = SLP * CSS_DFO, where SLP has units of 1/(ug/mL). Set to 0 to disable the chelation effect (drug holidays, untreated baseline disease-progression simulations). Reference values observed: ~3.5 ug/mL for 30 mg/kg/day, ~5.5 ug/mL for 45 mg/kg/day, ~7.5 ug/mL for 60 mg/kg/day at 45 kg body weight on the 5-days-per-week 8-h SC infusion schedule (Bellanti 2015 Fig 3).
• Source aliases:
  • SCssAV (Bellanti 2015 paper symbol; “simulated steady-state concentration, average”) – the per-subject population-PRED value before compliance correction.
  • TCssAV (Bellanti 2015 paper symbol; “true steady-state concentration, average”) – the post-compliance value SCssAV * (1 - CMPL); collapse into a single time-varying CSS_DFO column for nlmixr2lib by precomputing the (1 - CMPL) reduction in the input data.
• Example models: Bellanti_2015_deferoxamine.R (ug/mL; time-varying input on the linear DFO = SLP * CSS_DFO ferritin-degradation effect; 27 transfusion-dependent beta-thalassaemia major paediatric / adolescent patients on 20-60 mg/kg/day DFO 5 days per week).
• Notes: Specific scope because the column is intrinsically tied to deferoxamine PK and to the linear slope * CssAV effect form used in this paper. Sibling drug-specific Css canonical: CSS_RBV (ribavirin). The companion K_RBV approach-to-Css rate constant is not needed here because Bellanti 2015 treats CssAV as a population-typical steady-state value rather than reconstructing the rise-to-Css trajectory. For new simulations: a user supplies CSS_DFO directly (either a constant typical-Css value, or a time-varying column that switches to 0 during drug holidays); the vignette walks through computing CssAV analytically from a desired dose schedule via CssAV = (dose_per_week * F) / (CL_i * 168 h) with CL_i = 19.3 * (WT/70)^0.75 L/h for allometric scaling to paediatric / adolescent body weights.

CSS_CVC (canonical for cenicriviroc average steady-state plasma concentration from a noncompartmental analysis)

• Description: Individual average steady-state plasma concentration (Css) of cenicriviroc, the dual CCR5 / CCR2 antagonist developed as TBR-652 / TAK-652, carried as a per-subject exposure metric that drives a downstream static exposure-response model. Member of the CSS_<DRUG> family alongside CSS_RBV, CSS_ALBIFN and CSS_DFO. Distinguished from its three siblings by its provenance: those are all posthoc Bayesian or model-simulated quantities from an upstream population PK fit, whereas CSS_CVC is a noncompartmental quantity, computed as AUC0-24 / 24 h from an intensive steady-state profile. There is no compartmental PK model of cenicriviroc in the source paper to generate it with. Set to 0 for placebo subjects; both models that use the column are intercept-free, so a zero exposure returns exactly zero effect, which is how the placebo arm enters the published fits.
• Units: ng/mL (document per-model via covariateData[[CSS_CVC]]$units).
• Type: continuous
• Scope: specific
• Reference category: n/a – enters directly inside model() as the driver of the hyperbolic Emax term emax * CSS_CVC / (ic50 + CSS_CVC). Reference values observed (Marier 2011 Table 2, cohort means with CV%): 15.92 (52.8%), 51.89 (40.4%), 79.82 (30.6%), 110.8 (52.8%) and 303.0 (36.8%) ng/mL at 25, 50, 75, 100 and 150 mg QD; individual values span roughly 0-490 ng/mL. Compare against the drug’s in-vitro-estimated therapeutic concentration of 2 ng/mL, and against the fitted IC50s of 13.1 ng/mL (HIV-1 RNA) and 29.8 ng/mL (MCP-1).
• Source aliases:
  • Css / CSS (“average concentrations under steady-state conditions”, Marier 2011 Methods ‘Pharmacokinetics’ and the Table 2 column header). Rename to CSS_CVC before passing the dataset to rxSolve.
• Example models: Marier_2011_cenicriviroc_hiv1rna.R, Marier_2011_cenicriviroc_mcp1.R (both ng/mL; the shared exposure driver of the paper’s two independently fitted Emax exposure-response models, in 54 HIV-1-infected treatment-experienced adults dosed 25-150 mg QD for 10 days).
• Notes: Total, not unbound. Specific scope because the concentration scale and the NCA derivation are drug- and study-specific; the CSS_<DRUG> family exists precisely so each drug’s steady-state exposure metric keeps its own column rather than several drugs sharing an ambiguous bare CSS. Populate it for new simulations either by supplying Css directly or via Css = Dose / (CL/F * 24 h), but note that CL/F is not constant over the studied dose range – Marier 2011 Table 2 reports cohort mean CL/F of 81.31, 49.26, 42.17, 48.06 and 22.95 L/h at 25, 50, 75, 100 and 150 mg QD, exposure rose more than dose-proportionally at 150 mg, and two subjects dosed at the Argentine site showed more than 2-fold higher exposure than the rest of their cohort (a suspected high-fat food effect). A single clearance value will therefore misstate Css outside the 50-100 mg range. Distinct from CAV (a study-arm-level dosing-interval-averaged concentration used in MBMA-style Emax models) in being a per-subject value, and from CP_<DRUG>_<UNITS> (an instantaneous, time-varying plasma concentration fed per event record) in being time-fixed and exposure-summary-valued.

CSS_ANACETRAPIB (canonical for population-predicted anacetrapib steady-state trough plasma concentration from an upstream PK fit)

• Description: Population-predicted steady-state trough (24 h post-dose, C24h) plasma concentration of anacetrapib, carried as a per-subject exposure metric that drives a downstream exposure-response model. Not an observation and not an individual post hoc value: it is generated by solving a separately fitted anacetrapib population PK model at the regimen, formulation and prandial state of interest with the inter-individual random effects zeroed, then reading the pre-dose concentration at steady state. Member of the CSS_<DRUG> family alongside CSS_RBV, CSS_ALBIFN and CSS_DFO.
• Units: ng/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – enters a proportional Emax exposure-response term, Emax * CSS_ANACETRAPIB / (EC50 + CSS_ANACETRAPIB). Set to 0 for placebo periods. Reference values observed: 61.8 ng/mL at 50 mg once daily fasted and 276 ng/mL at 100 mg once daily with a patient-selected meal, both as the hot-melt-extruded tablet (reconstructed from Krishna_2011_anacetrapib.R); Krishna 2011 observed troughs above 400 ng/mL only in six normal volunteers given 400 mg with a high-fat meal, where the HDL-C model shows some lack of fit.
• Source aliases:
  • C_trough, Ctrough,i, C24h – Krishna 2011 notation in Eqs. 1, 2, 8 and 9, in Tables II and III, and throughout the Results and Discussion.
• Example models: Krishna_2011_anacetrapib_hdlc.R (sole driver of the proportional Emax HDL-C model, EC50 135 ng/mL), Krishna_2011_anacetrapib_ldlc.R (proportional Emax LDL-C model, EC50 237 ng/mL, in interaction with CONMED_ATORVASTATIN).
• Notes: Specific scope for the same reason as the rest of the family – the concentration scale is drug-specific, and the CSS_<DRUG> family exists precisely so each drug’s steady-state exposure metric keeps its own column rather than several drugs sharing an ambiguous bare CSS. Use the POPULATION-predicted value, not an individual post hoc one. Krishna 2011 Table III reports the LDL-C model fitted against both, notes the estimates “do not differ meaningfully”, and selects the population-predicted fit “since this allows for prediction of future lipid effects, given the treatment population and fed state, in subjects who may or may not have PK data available”. The fitted PD between-subject variances were therefore estimated against a driver carrying no PK variability; propagating PK inter-individual variability into this column would double-count it. Krishna 2011 selected the trough over five other candidate metrics (trough and 24 h average CETP inhibition, trough and 24 h average CETP activity, and the anacetrapib daily average concentration) and abandoned the mechanistic CETP-activity cascade entirely, because CETP activity rebounds with repeated dosing while the lipid effects do not – so do not substitute a CETP-activity or a daily-average-concentration column here. Distinct from CAV (dosing-interval average) and CMAX (peak): the register’s CMAX entry directs a model needing a different exposure metric to register a parallel canonical rather than overload an existing one, which is what this entry does for the trough.

CP0_MAT_NGML (canonical for maternal plasma concentration at delivery used to initialise a neonate’s central compartment)

• Description: Observed maternal plasma concentration of the modelled drug in a sample drawn at (or within minutes of) delivery, supplied as a static per-subject covariate on the neonate and used to set the neonate’s central-compartment initial condition at t = 0. This is the standard way a mother-infant dyad analysis carries in-utero (transplacental) exposure into a neonatal PK model when the umbilical cord is not modelled as an intermediate compartment: the neonate is born with drug already on board, and the maternal sample is the only measurement of it. Distinct from the CP_<DRUG>_<UNITS> family, which is an instantaneous, time-varying concentration of an upstream drug fed per event record into a PD or perpetrator model – CP0_MAT_NGML is time-fixed, one value per subject, of the same drug the model itself describes, and drives a state initial condition rather than an equation. Parallel in role to PLAQUE_BL and HGB_BL (per-subject observed baseline supplying a state’s t = 0 value), differing in that the baseline was measured in a different individual (the mother) than the one being modelled.
• Units: ng/mL. Document per-model via covariateData[[CP0_MAT_NGML]]$units; register a sibling only if a paper reports maternal concentrations in units that cannot be converted (the _NGML suffix is part of the canonical name, so a mg/L analysis should convert rather than add a name).
• Type: continuous
• Scope: general
• Reference category: n/a – static per-subject concentration supplied as a covariate column. Set to 0 for a neonate with no detectable transplacental transfer (the value is an amount-defining input, so 0 correctly means “born with no drug on board”). Reference values observed: mean (SD) 28.2 (23.5) ng/mL for ondansetron across 29 maternal samples in Lam 2025 (Table 1).
• Source aliases:
  • C0 – printed column name in the Lam 2025 Appendix S1 NONMEM control stream ($INPUT... DV C0 MDV CMT; $PK line A_0(2) = C0 * S2) – used in Lam_2025_ondansetron.R.
• Example models: Lam_2025_ondansetron.R (Lam 2025 neonatal ondansetron popPK; sets central(0) <- CP0_MAT_NGML * vc / 1000 so that Cc(0) equals the maternal concentration).
• Notes: Drug-agnostic (general scope) on 2026-08-14: maternal concentration at delivery is a general transplacental-exposure concept and the next drug to use it should not need a second canonical, so the drug is implicitly “the drug this model describes” (the same convention CP_MGL uses). Note the unit bookkeeping this covariate implies: the model must convert the concentration to an amount using the central volume before assigning the initial condition, and the conversion factor depends on the model’s dose and volume units (in Lam 2025, central in mg and vc in L give CP0_MAT_NGML * vc / 1000). Founding example: Lam_2025_ondansetron.R.

CP_MGL (canonical for instantaneous drug plasma concentration as a time-varying PD driver)

• Description: Instantaneous (per-event-record) plasma concentration of the modeled drug, supplied directly as a time-varying covariate column rather than computed from a coupled PK model. Used in PD-only myelosuppression / toxicity / response models that consume an upstream PK trajectory as input – typically because the source analysis was done as a sequential PK-then-PD fit, with the PK model fixed from a previously published popPK analysis (e.g., a Bruno-style docetaxel popPK feeding a Friberg / Kloft myelosuppression PD model).
• Units: mg/L (= ug/mL; the two labels are numerically equivalent and the canonical reporting convention for cytotoxic-chemotherapy concentrations in Kloft 2006 / Friberg 2002 family analyses). Document per-model via covariateData[[CP_MGL]]$units if a different unit is used.
• Type: continuous
• Scope: specific
• Reference category: n/a – typically enters as a linear term in the drug-effect expression (e.g., drug = SL * CP_MGL, where SL has units of 1/(mg/L)). Set to 0 for placebo periods or any time outside the drug-exposure window. Reference values observed: docetaxel typical Cmax ~3 mg/L after a 100 mg/m^2 1-hour IV infusion (Kloft 2006 / Netterberg 2017 simulated dataset).
• Source aliases:
  • CP (Kloft 2006 / Netterberg 2017 NM-TRAN $INPUT convention for “predicted drug concentration”; values in mg/L) – used in Netterberg_2017_docetaxel.R.
• Example models: Netterberg_2017_docetaxel.R (linear drug effect on the proliferation rate of the Friberg myelosuppression chain: (1 - SL * CP_MGL) with SL = 19.27 (mg/L)^-1 after Kloft 2006’s THETA(3)/808*1000 MW-808 conversion; CP_MGL supplied per event row from an upstream docetaxel popPK simulation), Puisset_2007_docetaxel.R (linear drug effect Slope * CP_MGL on the proliferating compartment of the Friberg myelosuppression chain; the upstream docetaxel PK in Puisset 2007 was Baille 1997 / Bruno 1996 individual Bayesian posthoc profiles supplied externally and not encoded in the model file).
• Notes: Specific scope because the covariate’s mechanistic meaning is bound to the modeled drug and the source paper’s chosen PK input (e.g., a Bruno 1996/1998 docetaxel popPK trajectory for Netterberg 2017). Distinct from CAV (dosing-interval-averaged exposure used in Emax / EC50 PD models) – CP_MGL is the instantaneous concentration, sampled at every PD event time. When a future paper requires the same time-varying-PK-as-PD-input pattern for a different drug, register a drug-specific canonical (e.g., CP_PACL_MGL for paclitaxel) rather than overloading this name; CP_MGL retains the implicit “drug = the modeled drug under the PD analysis” semantics. When a paper supplies the time-varying PK as separate per-subject empirical-Bayes PK parameters (e.g., CL_INDIV, VC_INDIV, VP_INDIV), use those columns in a coupled PK-PD ODE model (see Friberg_2002_paclitaxel.R) rather than reducing to CP_MGL. The choice between PK-as-covariate (this canonical) and PK-as-EBE-parameters depends on whether the source paper’s NM-TRAN dataset shipped Cp directly or shipped the upstream individual PK parameters.

CP_OXA_MGL (canonical for instantaneous oxaliplatin plasma concentration as a time-varying PD driver)

• Description: Instantaneous plasma concentration of oxaliplatin supplied directly as a time-varying covariate column rather than computed from a coupled PK model. Used in PD-only platelet-dynamics / myelosuppression models that consume an upstream oxaliplatin popPK trajectory as input.
• Units: mg/L (= ug/mL; oxaliplatin MW = 397.29 g/mol, so 1 mg/L = 2.52 umol/L).
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as a power-function term in the platelet-dynamics drug-effect expression E_drug = alpha * CP_OXA_MGL^beta. Set to 0 outside the drug-exposure window (e.g., pre-HIO or CRS-alone patients). Reference values observed: hyperthermic intraperitoneal oxaliplatin (HIO) administered at 200 mg/L initial peritoneal concentration for 30 min gives typical plasma Cmax ~1 mg/L with half-life ~14 h (Perez-Ruixo 2013 Cancer Chemother Pharmacol 71:693-704).
• Source aliases:
  • Cp (Perez-Ruixo 2015 The AAPS Journal Methods and Eq. 8 symbol for the oxaliplatin plasma concentration driving E_drug) – used in PerezRuixo_2015_oxaliplatin_platelet_dynamics.R.
• Example models: PerezRuixo_2015_oxaliplatin_platelet_dynamics.R (drug effect E_drug = alpha * CP_OXA_MGL^beta on the megakaryocyte-progenitor proliferation rate ktr, with alpha = 0.881 L/mg and beta = 2.63 per Perez-Ruixo 2015 Table I; CP_OXA_MGL supplied per event row from an upstream oxaliplatin popPK simulation not packaged in nlmixr2lib at extraction time – the vignette shows a placeholder monoexponential Cp trajectory tuned to typical HIO Cmax ~1 mg/L and half-life ~14 h).
• Notes: Specific scope; oxaliplatin-specific. The drug-specific naming follows the CP_<drug>_<units> precedent from CP_MGL (general docetaxel-family form), CP_OXY_NGML, CP_FBX_NGML, CP_LSN_NGML, CP_MORPH_NGML, CP_RIF_UM, CP_GDC_UM, CP_REM_NGML, and CP_GLASDEGIB_NGML. The paper uses Cp as the paper-symbol; renamed to CP_OXA_MGL in the register per the naming pattern. The natural input source is a coupled oxaliplatin popPK simulation (Perez-Ruixo 2013 Cancer Chemother Pharmacol; not available when this model was built); users can either digitise Cp trajectories from Figure 5 of Perez-Ruixo 2015 or simulate a placeholder monoexponential decline with typical HIO parameters.

CP_OXY_NGML (canonical for instantaneous oxypurinol plasma concentration as a time-varying PD driver)

• Description: Instantaneous plasma concentration of oxypurinol (the active metabolite of allopurinol; xanthine oxidase inhibitor) supplied directly as a time-varying covariate column rather than computed from a coupled PK model. Used in semi-mechanistic uric-acid disposition models that take XOI exposure as input to the production-inhibition equation.
• Units: ng/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as a Hill term in the production-inhibition expression 1 - Rmax * CP_OXY_NGML / (CP_OXY_NGML + p50). Set to 0 outside the drug-exposure window or for non-XOI scenarios. Reference values observed: mean daily concentration on 300 mg/day allopurinol is approximately 10,000 ng/mL (Aksenov 2018, Eq. 13).
• Source aliases:
  • [P]_PIN (oxypurinol) – the symbol used in Aksenov 2018 Eq. 9 for the production-inhibitor concentration when the inhibitor is oxypurinol.
• Example models: Aksenov_2018_uricAcid.R (Hill-type production inhibition with rmax_oxy = 0.84 and p50_oxy = 14000 ng/mL per Aksenov 2018 Table 1).
• Notes: Specific scope because the canonical name is bound to oxypurinol; allopurinol’s PK is conventionally summarized via the active metabolite oxypurinol (Day et al. 2007). Distinct from CP_FBX_NGML (febuxostat) and CP_LSN_NGML (lesinurad). When the source paper supplies an upstream popPK for oxypurinol (e.g., Wright et al. 2013, Anzai & Endou 2012), the user simulates that PK to populate this column; otherwise a steady-state value can be used.

CP_FBX_NGML (canonical for instantaneous febuxostat plasma concentration as a time-varying PD driver)

• Description: Instantaneous plasma concentration of febuxostat (xanthine oxidase inhibitor) supplied directly as a time-varying covariate column rather than computed from a coupled PK model. Used in semi-mechanistic uric-acid disposition models that take XOI exposure as input to the production-inhibition equation.
• Units: ng/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as a Hill term in the production-inhibition expression 1 - Rmax * CP_FBX_NGML / (CP_FBX_NGML + p50). Set to 0 outside the drug-exposure window or for non-XOI scenarios. Reference values observed: mean daily concentration on 40 mg/day febuxostat is approximately 1000-2000 ng/mL (Aksenov 2018, Bhattaram & Gobburu 2017 regulatory review).
• Source aliases:
  • [P]_PIN (febuxostat) – the symbol used in Aksenov 2018 Eq. 9 for the production-inhibitor concentration when the inhibitor is febuxostat.
• Example models: Aksenov_2018_uricAcid.R (Hill-type production inhibition with rmax_fbx = 1 (fixed) and p50_fbx = 120 ng/mL for hyperuricemic subjects (or 87 ng/mL for normouricemic subjects) per Aksenov 2018 Table 1).
• Notes: Specific scope; febuxostat-specific. The p50 parameter differs between hyperuricemic and normouricemic populations in Aksenov 2018; Rmax is fixed at 1 per Bhattaram & Gobburu 2017. Distinct from CP_OXY_NGML (oxypurinol) and CP_LSN_NGML (lesinurad).

CP_LSN_NGML (canonical for instantaneous lesinurad plasma concentration as a time-varying PD driver)

• Description: Instantaneous plasma concentration of lesinurad (uricosuric URAT1 inhibitor) supplied directly as a time-varying covariate column rather than computed from a coupled PK model. Used in semi-mechanistic uric-acid disposition models that take uricosuric exposure as input to the fractional-excretion-increase equation.
• Units: ng/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as a Hill term in the fractional-excretion expression FE = FE0 + Fmax * CP_LSN_NGML / (CP_LSN_NGML + p50). Set to 0 outside the drug-exposure window. Reference values observed: peak plasma concentration after single dose 200 mg lesinurad is approximately 6,000-9,000 ng/mL (Fleischmann et al. 2014; Shen et al. 2015).
• Source aliases:
  • [P]_RIN (lesinurad) – the symbol used in Aksenov 2018 Eq. 10 for the reabsorption-inhibitor concentration when the inhibitor is lesinurad.
• Example models: Aksenov_2018_uricAcid.R (Hill-type increase in fractional excretion with fmax_lsn = 0.56 (fixed) and p50_lsn = 23000 ng/mL for hyperuricemic subjects (or 11000 ng/mL for normouricemic subjects) per Aksenov 2018 Table 1).
• Notes: Specific scope; lesinurad-specific. The p50 parameter differs between hyperuricemic and normouricemic populations in Aksenov 2018; Fmax was fixed during estimation. Distinct from CP_OXY_NGML (oxypurinol) and CP_FBX_NGML (febuxostat).

CP_MORPH_NGML (canonical for instantaneous morphine plasma concentration as a time-varying PD driver)

• Description: Instantaneous plasma concentration of morphine supplied directly as a time-varying covariate column rather than computed from a coupled PK model. Used in PD-only IRT / latent-pain models that take morphine exposure as an external input to a concentration-effect equation.
• Units: ng/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – enters linearly into the latent-pain equation pain = pain_state - e_morph_pain * CP_MORPH_NGML + e_time_pain * time (Valitalo 2017). Reference values observed: most individual predicted concentrations in Valitalo 2017 were within 0-60 ng/mL (Figure 2a); the IRT linear morphine-effect slope is 0.0091 (ng/mL)^-1, so a 20 ng/mL morphine exposure reduces the latent pain by ~0.18 latent-variable units.
• Source aliases:
  • CP (Valitalo 2017 NM-TRAN $INPUT convention for “morphine plasma concentration”; values in ng/mL) – used in Valitalo_2017_morphine.R.
• Example models: Valitalo_2017_morphine.R (linear morphine concentration-effect on the IRT latent pain variable; CP_MORPH_NGML supplied per event row from an upstream morphine popPK simulation, typically Knibbe_2009_morphine.R).
• Notes: Specific scope; morphine-specific. The drug-specific naming follows the existing CP_OXY_NGML / CP_FBX_NGML / CP_LSN_NGML precedent established with Aksenov 2018. Distinct from the broader CP_MGL (mg/L PD-driver convention used in Netterberg 2017 docetaxel myelosuppression and similar) because the IRT PD models in this family use ng/mL natively. When a future morphine PD analysis uses mg/L, the conversion is CP_MORPH_NGML = CP_MORPH_MGL * 1000.

CP_BPN_NGML (canonical for instantaneous buprenorphine plasma concentration as a time-varying PD driver)

• Description: Instantaneous plasma concentration of buprenorphine (BPN) supplied directly as a time-varying covariate column rather than computed from a coupled PK model. Used in PD-only opioid-blockade models that take buprenorphine exposure as an external input to an Imax concentration-effect equation for abuse-liability and withdrawal endpoints.
• Units: ng/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as the Hill term of an Imax inhibition expression, e.g. 1 - imax * CP_BPN_NGML^hill / (ic50^hill + CP_BPN_NGML^hill) (Walsh 2024). Reference values observed: the observed BPN plasma concentration range across the Walsh 2024 phase 2 study was 0.636-12.3 ng/mL, and the paper’s own COWS application simulation swept 0-10 ng/mL. Set to 0 for the pre-treatment (unblocked) condition, which returns the model to its baseline.
• Source aliases:
  • Cp (Walsh 2024 notation, “the time-matched BPN concentration”; values in ng/mL) – used in Walsh_2024_buprenorphine_drugLiking.R, Walsh_2024_buprenorphine_desireToUse.R and Walsh_2024_buprenorphine_cows.R.
• Example models: Walsh_2024_buprenorphine_drugLiking.R (Imax blockade of period-corrected drug liking Emax VAS after an intramuscular hydromorphone 18 mg challenge; IC50 0.075 ng/mL), Walsh_2024_buprenorphine_desireToUse.R (Imax suppression of desire to use VAS with an exponential onset delay; IC50 0.0129 ng/mL), Walsh_2024_buprenorphine_cows.R (bounded-integer Imax suppression of the COWS total score; IC90 0.109 ng/mL).
• Notes: Specific scope; buprenorphine-specific. The drug-specific naming follows the established CP_<drug>_<units> precedent (CP_OXY_NGML, CP_FBX_NGML, CP_LSN_NGML, CP_MORPH_NGML, CP_RIF_UM, CP_REM_NGML, CP_GLASDEGIB_NGML). Walsh 2024 sampled one BPN plasma concentration per challenge day, approximately 60 min before each hydromorphone administration, and time-matched it to the PD score; the analysis is therefore driven by daily concentrations rather than a dense profile. The upstream CAM2038 (subcutaneous long-acting buprenorphine depot) popPK model is NOT reported in Walsh 2024 and no buprenorphine popPK model exists in the nlmixr2lib registry, so downstream users must supply the concentration trajectory externally – either from their own popPK source or by digitising Supplementary Figure S2 of Walsh 2024.

CP_EIDD_NGML (canonical for instantaneous EIDD-1931 plasma concentration as a time-varying antiviral PD driver)

• Description: Instantaneous plasma concentration of EIDD-1931, the pharmacologically active nucleoside-analogue metabolite of the antiviral prodrug molnupiravir, supplied directly as a time-varying covariate column rather than computed from a coupled PK model. Used in within-host viral-dynamics / QSP models that consume an external molnupiravir exposure profile as the driver of an Imax inhibition of viable-virus production.
• Units: ng/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – enters the Imax inhibition function 1 - Imax * CP_EIDD_NGML^nhill / (IC50_av^nhill + CP_EIDD_NGML^nhill); set to 0 for placebo arms, for neutralizing-antibody-only arms, and for any time outside the drug-exposure window. Reference values observed: Rao 2023 Supplementary Figure 16 (digitized from Painter et al. 2021, doi:10.1128/AAC.02428-20) shows single-dose peaks of roughly 300 ng/mL (50 mg) to 3000 ng/mL (800 mg) within 1-2 h, falling below ~100 ng/mL by 12 h; the fitted antiviral IC50 is 1086 ng/mL.
• Source aliases:
  • Cav / Cantiviral (Rao 2023 Supplementary Methods and src/covid19_dxdt.m, where the concentration enters as an interp1 lookup on the digitized profile) – used in Rao_2023_covid19_qsp.R.
• Example models: Rao_2023_covid19_qsp.R (Imax inhibition of viable-virus production from infected cells in the COVID-19 QSP model; the digitized 0-12 h profiles for the 50, 100, 200, 300, 400, 600 and 800 mg single doses ship with the authors’ code release as data/molnupiravir_PK.csv).
• Notes: Specific scope; bound to EIDD-1931 and to the plasma (rather than intracellular) compartment. Rao 2023 states explicitly that the intracellular active triphosphate was not modelled and that the IC50 was optimized against the reported plasma EIDD-1931 concentration, so this canonical must not be reused for an intracellular-metabolite driver. Named for the active metabolite rather than the prodrug because that is the moiety whose concentration drives the effect; a future model that drives an effect from molnupiravir parent concentrations should register a separate CP_MOLNUPIRAVIR_NGML. Follows the drug-specific CP_<DRUG>_NGML precedent established with Aksenov 2018 (CP_OXY_NGML, CP_FBX_NGML, CP_LSN_NGML) and Valitalo 2017 (CP_MORPH_NGML).

CP_RIF_UM (canonical for instantaneous rifampicin plasma concentration as a time-varying OATP1B-perpetrator covariate)

• Description: Instantaneous plasma concentration of rifampicin supplied directly as a time-varying covariate column rather than computed from a coupled PK model. Used as the perpetrator-drug input to a competitive OATP1B inhibition term in DDI popPK models for OATP1B substrates and endogenous OATP1B biomarkers (e.g., coproporphyrin I, rosuvastatin). Set to 0 outside the rifampicin co-administration window so the inhibition term collapses to the baseline form.
• Units: umol/L (= uM; rifampicin MW = 822.94 g/mol so 1 umol/L = 0.823 mg/L = 823 ng/mL).
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as the denominator of a Michaelis-Menten-style competitive-inhibition term cl_b_eff = cl_b / (1 + CP_RIF_UM / ki) where ki is the OATP1B inhibition constant in umol/L. Reference peak observed: a single 600 mg oral rifampicin dose in the Barnett 2018 cohort produces a typical Cmax of approximately 29 umol/L in the modellib(‘Barnett_2018_rifampicin’) typical-value simulation.
• Source aliases:
  • CRIF (Barnett 2018 Eq. 4 and the analogous RSV inhibition equation; values reported in umol/L).
• Example models: Barnett_2018_coproporphyrin_I.R (drives competitive OATP1B inhibition of biliary CPI clearance: cl_b_eff = cl_b / (1 + CP_RIF_UM / ki) with ki = 1.15 umol/L total / 0.13 umol/L unbound), Barnett_2018_rosuvastatin.R (analogous form with ki = 2.23 umol/L total / 0.25 umol/L unbound), Yoshida_2018_coproporphyrin_I_rifampin.R (drives competitive OATP1B inhibition of the hepatic component of CPI clearance in Yoshida 2018’s one-compartment fNH-parameterised model: kdeg_eff = kdeg * (fnh + (1 - fnh) / (1 + CP_RIF_UM / kiu)) with kiu = 0.0203 umol/L unbound), Almond_2016_rifampicin_invitro.R (nominal rifampicin concentration in hepatocyte CULTURE MEDIUM, 0.03-30 uM, driving a CYP3A4 induction Emax curve), Almond_2016_rifampicin_invivo.R (OPERATIONAL UNBOUND rifampicin concentration at the site of CYP3A4 interaction, driving the refined in vivo induction Emax curve with ec50 = 0.32 umol/L).
• Notes: Specific scope; rifampicin-specific. The drug-specific naming follows the established CP_<drug>_<units> precedent (CP_OXY_NGML, CP_FBX_NGML, CP_LSN_NGML, CP_MORPH_NGML). The natural input source is a coupled rifampicin popPK simulation; in the Barnett 2018 extraction package, users typically simulate modellib('Barnett_2018_rifampicin') first (which returns rifampicin Cc in umol/L after the in-model MW conversion) and feed its central-compartment output as the CP_RIF_UM column on the CPI or RSV event table. Distinct from the binary indicator CONMED_RIF (which captures period-level effects like the V1 / V2 / Q binary covariate shifts and does not carry magnitude information). The Yoshida 2018 rifampin-CPI extraction used the Simcyp v16r1 default single-dose rifampin model output for portal-vein unbound concentration; that PBPK profile is not reproducible from the available sources, and the paper itself documents a ~5x sensitivity of the estimated Ki,u to the choice of perpetrator-PK model, so downstream users must supply their own CP_RIF_UM profile and treat the resulting CPI excursion as conditional on that choice. The Almond 2016 CYP3A4-induction extraction reuses this column for two further senses of “rifampicin concentration” that are the same quantity in the same units but are NOT interchangeable with each other or with a total plasma concentration: an in-vitro hepatocyte CULTURE-MEDIUM concentration (Almond_2016_rifampicin_invitro.R) and an OPERATIONAL UNBOUND concentration at the site of enzyme interaction (Almond_2016_rifampicin_invivo.R, where Almond 2016 Eq. 3 uses fu,B * I_liver / (Kp / B:P) for the liver and Eq. 4 uses the enterocyte concentration for the gut). Supplying a total plasma concentration to the in-vivo induction model overstates the driver by roughly 1 / fu, and rifampicin fu = 0.15. The same reuse-for-a-non-plasma-matrix pattern is established by CP_TUVUSERTIB_NGML (patch-clamp bath) and CP_GDC_UM (unbound portal vein); in every case the per-model covariateData entry states which sense is required.

CP_AMITRIPTYLINE_UM (canonical for instantaneous total amitriptyline plasma concentration as a time-varying PD driver)

• Description: Instantaneous TOTAL (not unbound) plasma concentration of amitriptyline supplied directly as a time-varying covariate column rather than computed from a coupled PK model. Used as the exposure driver of cardiac-electrophysiology PD models for amitriptyline – founding use is the sigmoid Emax relationship between amitriptyline exposure and the R-R interval length – where the source analysis fixed its PK from a separate physiologically based model that the PD publication does not restate.
• Units: umol/L (= uM; amitriptyline MW = 277.4 g/mol, so 1 umol/L = 277.4 ng/mL).
• Type: continuous
• Scope: specific
• Reference category: n/a – enters a sigmoid Emax term e0 + (rmax - e0) * C^hill / (ec50^hill + C^hill). Set to 0 for the drug-free reference, at which the term collapses to the baseline e0. Reference values observed: the Tylutki 2018 Figure 2 fitting dataset spans roughly 0.1-17.8 umol/L (28-4900 ng/mL); therapeutic trough concentrations of 50-300 ng/mL are 0.18-1.08 umol/L, and concentrations above about 2 umol/L come almost entirely from acute-overdose case reports.
• Source aliases:
  • C (Tylutki 2018 Eq. 4 glossary, “AT total plasma concentration [uM]”).
• Example models: Tylutki_2018_amitriptyline_RR.R (drives the sigmoid Emax shortening of the R-R interval, rr = e0 + (rmax - e0) * CP_AMITRIPTYLINE_UM^hill / (ec50^hill + CP_AMITRIPTYLINE_UM^hill) with e0 = 995.3 ms, rmax = 500.8 ms, ec50 = 0.4 umol/L and hill = 1.5).
• Notes: Specific scope; amitriptyline-specific. The drug-specific naming follows the established CP_<drug>_<units> precedent (CP_RIF_UM, CP_CARBAMAZEPINE_UM, CP_PHENOBARBITAL_UM, CP_OXY_NGML), as CP_MGL’s own Notes prescribe for a new drug. The column carries TOTAL plasma drug and must not be fed a free cardiac-tissue concentration: Tylutki 2018 Eq. 5 converts between them as Cfree,cardiac = Ctotal,plasma * Kp_ht * fu_ht with Kp_ht = 11.77 and fu_ht = 0.0012 for amitriptyline (and Kp_ht = 35.63, fu_ht = 0.001 for its active metabolite nortriptyline), a factor of about 1/71, and the same paper uses the free cardiac concentration for its QT endpoint while using total plasma for the R-R endpoint. A concentration reported in ng/mL must be divided by 277.4 before being supplied here. The natural input source is the full-PBPK amitriptyline model of Tylutki 2018 J Pharm Sci 107:1167-1177 (doi:10.1016/j.xphs.2017.11.012), which the founding paper imports unchanged and does not restate; no amitriptyline popPK model is in the nlmixr2lib registry, so users must supply their own trajectory. A future extraction driving a nortriptyline PD endpoint should register the sibling CP_NORTRIPTYLINE_UM (nortriptyline MW = 263.384 g/mol) rather than overloading this column, since the two analytes circulate together after amitriptyline dosing and are not interchangeable.

CP_CARBAMAZEPINE_UM (canonical for instantaneous carbamazepine concentration as a time-varying CYP-perpetrator covariate)

• Description: Instantaneous concentration of carbamazepine supplied directly as a time-varying covariate column rather than computed from a coupled PK model, used as the perpetrator-drug input to a CYP3A4 induction Emax term. Founding use is an in-vitro hepatocyte culture-medium concentration; the column is defined generally so that a plasma or unbound site-of-interaction concentration can use the same name, with the per-model covariateData entry stating which sense is required.
• Units: umol/L (= uM; carbamazepine MW = 236.3 g/mol so 1 umol/L = 0.2363 mg/L).
• Type: continuous
• Scope: specific
• Reference category: n/a – enters a three-parameter Emax induction term fold = 1 + emax * CP_CARBAMAZEPINE_UM / (ec50 + CP_CARBAMAZEPINE_UM). Set to 0 for the vehicle-control / no-perpetrator condition, at which the term collapses to fold = 1.
• Source aliases:
  • concentration of the inducer (Almond 2016 Table 1; values reported in uM).
  • It (Almond 2016 Eqs. 3 and 4, the perpetrator concentration at time t in liver or gut).
• Example models: Almond_2016_carbamazepine_invitro.R (nominal culture-medium concentrations 1-1000 uM driving fold induction of CYP3A4 mRNA and activity in cryopreserved human hepatocytes; ec50 58.7 uM on mRNA and 59.1 uM on activity).
• Notes: Specific scope; carbamazepine-specific. Follows the established CP_<drug>_<units> family (CP_RIF_UM, CP_GDC_UM, CP_TUVUSERTIB_NGML). Carbamazepine autoinduces CYP3A4 and its 10,11-epoxide metabolite is assumed equipotent as an inducer (Almond 2016 Supplemental Table 2), so a plasma-driven use of this column over multiple days should account for the perpetrator’s own rising clearance; the founding in-vitro model holds the concentration static and is unaffected.

CP_PHENOBARBITAL_UM (canonical for instantaneous phenobarbital concentration as a time-varying CYP-perpetrator covariate)

• Description: Instantaneous concentration of phenobarbital supplied directly as a time-varying covariate column rather than computed from a coupled PK model, used as the perpetrator-drug input to a CYP3A4 induction Emax term. Founding use is an in-vitro hepatocyte culture-medium concentration; the per-model covariateData entry states which sense (medium, plasma, unbound site-of-interaction) is required.
• Units: umol/L (= uM; phenobarbital MW = 232 g/mol so 1 umol/L = 0.232 mg/L).
• Type: continuous
• Scope: specific
• Reference category: n/a – enters a three-parameter Emax induction term fold = 1 + emax * CP_PHENOBARBITAL_UM / (ec50 + CP_PHENOBARBITAL_UM). Set to 0 for the vehicle-control / no-perpetrator condition, at which the term collapses to fold = 1.
• Source aliases:
  • concentration of the inducer (Almond 2016 Table 1; values reported in uM).
  • It (Almond 2016 Eqs. 3 and 4, the perpetrator concentration at time t in liver or gut).
• Example models: Almond_2016_phenobarbital_invitro.R (nominal culture-medium concentrations 10-3000 uM driving fold induction of CYP3A4 mRNA and activity in cryopreserved human hepatocytes; ec50 743 uM on mRNA and 473 uM on activity).
• Notes: Specific scope; phenobarbital-specific. Follows the established CP_<drug>_<units> family. Phenobarbital is by far the least potent of the six inducers Almond 2016 characterised, which is why its founding concentration ladder reaches 3000 uM. Orientation for clinically driven use: the clinical DDI study of Almond 2016 Table 4 maintained plasma phenobarbital at 10-20 ug/mL, i.e. roughly 43-86 umol/L, which is an order of magnitude BELOW the in-vitro ec50 – a plasma-driven induction term will therefore sit far down the rising limb of the curve, not on its plateau.

CP_PHENYTOIN_UM (canonical for instantaneous phenytoin concentration as a time-varying CYP-perpetrator covariate)

• Description: Instantaneous concentration of phenytoin supplied directly as a time-varying covariate column rather than computed from a coupled PK model, used as the perpetrator-drug input to a CYP3A4 induction Emax term. Founding use is an in-vitro hepatocyte culture-medium concentration; the per-model covariateData entry states which sense is required.
• Units: umol/L (= uM; phenytoin MW = 252.28 g/mol so 1 umol/L = 0.2523 mg/L).
• Type: continuous
• Scope: specific
• Reference category: n/a – enters a three-parameter Emax induction term fold = 1 + emax * CP_PHENYTOIN_UM / (ec50 + CP_PHENYTOIN_UM). Set to 0 for the vehicle-control / no-perpetrator condition, at which the term collapses to fold = 1.
• Source aliases:
  • concentration of the inducer (Almond 2016 Table 1; values reported in uM).
  • It (Almond 2016 Eqs. 3 and 4, the perpetrator concentration at time t in liver or gut).
• Example models: Almond_2016_phenytoin_invitro.R (nominal culture-medium concentrations 1-1000 uM driving fold induction of CYP3A4 mRNA and activity in cryopreserved human hepatocytes; ec50 123 uM on mRNA and 51.3 uM on activity).
• Notes: Specific scope; phenytoin-specific. Follows the established CP_<drug>_<units> family. Orientation for clinically driven use: the clinical DDI study of Almond 2016 Table 4 adjusted the phenytoin dose to maintain plasma concentrations of 10-20 ug/mL, i.e. roughly 40-79 umol/L, which straddles the in-vitro activity ec50 of 51.3 umol/L. Note that Almond 2016 Supplemental Table 2 also carries phenytoin induction parameters for CYP2C9 and CYP2B6, but those are third-party literature values cited as Simcyp compound-file inputs rather than that paper’s own bench work and are not encoded in any model.

CP_EFAVIRENZ_UM (canonical for instantaneous efavirenz concentration as a time-varying CYP-perpetrator covariate)

• Description: Instantaneous concentration of efavirenz supplied directly as a time-varying covariate column rather than computed from a coupled PK model, used as the perpetrator-drug input to a CYP3A4 induction Emax term. Founding use is an in-vitro hepatocyte culture-medium concentration; the per-model covariateData entry states which sense is required.
• Units: umol/L (= uM). No efavirenz molecular weight is reported in the founding source, so no mass-unit equivalence is asserted here.
• Type: continuous
• Scope: specific
• Reference category: n/a – enters a three-parameter Emax induction term fold = 1 + emax * CP_EFAVIRENZ_UM / (ec50 + CP_EFAVIRENZ_UM). Set to 0 for the vehicle-control / no-perpetrator condition, at which the term collapses to fold = 1.
• Source aliases:
  • concentration of the inducer (Almond 2016 Table 1; values reported in uM).
• Example models: Almond_2016_efavirenz_invitro.R (nominal culture-medium concentrations 0.1-30 uM driving fold induction of CYP3A4 mRNA and activity in cryopreserved human hepatocytes; ec50 8.4 uM on mRNA and 4.9 uM on activity).
• Notes: Specific scope; efavirenz-specific. Follows the established CP_<drug>_<units> family. Efavirenz is the second most potent of the six inducers Almond 2016 characterised, behind rifampicin. Caveat carried from the founding model: efavirenz was characterised in vitro only in that paper and appears in no clinical DDI simulation there, so the founding ec50/emax have no in-vivo verification within their source.

CP_NIFEDIPINE_UM (canonical for instantaneous nifedipine concentration as a time-varying CYP-perpetrator covariate)

• Description: Instantaneous concentration of nifedipine supplied directly as a time-varying covariate column rather than computed from a coupled PK model, used as the perpetrator-drug input to a CYP3A4 induction Emax term. Founding use is an in-vitro hepatocyte culture-medium concentration; the per-model covariateData entry states which sense is required.
• Units: umol/L (= uM; nifedipine MW = 346.3 g/mol so 1 umol/L = 0.3463 mg/L).
• Type: continuous
• Scope: specific
• Reference category: n/a – enters a three-parameter Emax induction term fold = 1 + emax * CP_NIFEDIPINE_UM / (ec50 + CP_NIFEDIPINE_UM). Set to 0 for the vehicle-control / no-perpetrator condition, at which the term collapses to fold = 1.
• Source aliases:
  • concentration of the inducer (Almond 2016 Table 1; values reported in uM).
• Example models: Almond_2016_nifedipine_invitro.R (nominal culture-medium concentrations 0.03-100 uM driving fold induction of CYP3A4 mRNA and activity in cryopreserved human hepatocytes; ec50 13.0 uM on mRNA and 4.0 uM on activity).
• Notes: Specific scope; nifedipine-specific. Follows the established CP_<drug>_<units> family. This column carries nifedipine in its PERPETRATOR role. Nifedipine is dual-role in the founding source – a CYP3A4 substrate (victim) in the clinical rifampicin and phenobarbital DDI simulations, and separately a CYP3A4 inducer characterised in vitro – and the founding induction parameters have the widest between-donor spread of any inducer in that study (donor-level CV about 76 percent on emax for both endpoints, n = 4 donors), so they are a weakly determined central estimate.

CP_GDC_UM (canonical for instantaneous GDC-0810 portal-vein unbound concentration as a time-varying OATP1B-perpetrator covariate)

• Description: Instantaneous unbound portal-vein concentration of GDC-0810 (an orally bioavailable selective estrogen receptor downregulator and weak OATP1B inhibitor) supplied directly as a time-varying covariate column rather than computed from a coupled PBPK model. Used as the perpetrator-drug input to a competitive OATP1B inhibition term in DDI models for OATP1B substrates and endogenous OATP1B biomarkers (coproporphyrin I). Set to 0 outside the GDC-0810 co-administration window so the inhibition term collapses to the baseline (no-inhibition) form.
• Units: umol/L (= uM)
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as the denominator of a Michaelis-Menten-style competitive-inhibition term kdeg_eff = kdeg * (fnh + (1 - fnh) / (1 + CP_GDC_UM / kiu)) where kiu is the OATP1B unbound inhibition constant in umol/L (Yoshida 2018 estimated kiu = 0.00174 umol/L for GDC-0810).
• Source aliases:
  • CGDC (Yoshida 2018 NM-TRAN convention for the per-record GDC-0810 portal-vein unbound concentration; values in umol/L). The Yoshida 2018 paper itself reports values from the in-house Y. Chen et al. PBPK model (referenced as personal communication; not available).
• Example models: Yoshida_2018_coproporphyrin_I_GDC0810.R (drives competitive OATP1B inhibition of the hepatic component of CPI clearance in Yoshida 2018’s one-compartment fNH-parameterised model).
• Notes: Specific scope; GDC-0810-specific. The drug-specific naming follows the established CP_<drug>_<units> precedent (CP_OXY_NGML, CP_FBX_NGML, CP_LSN_NGML, CP_MORPH_NGML, CP_RIF_UM). The natural input source in Yoshida 2018 was an in-house PBPK model for GDC-0810 (Y. Chen et al., personal communication); that source is not available and no GDC-0810 PK model exists in the nlmixr2lib registry, so downstream users must supply CP_GDC_UM externally. The paper notes (Discussion) that observed GDC-0810 plasma AUC IIV was approximately 20%, so the IIV reported on Ki,u (30.1% CV) partially includes per-subject variability in portal-vein exposure rather than purely the intrinsic inhibition-constant variability.

CP_PROPRANOLOL_NGML (canonical for instantaneous propranolol plasma concentration as a time-varying PD driver)

• Description: Instantaneous plasma concentration of propranolol (a non-selective beta-adrenoceptor antagonist) supplied directly as a time-varying covariate column rather than computed from a coupled PK model. Used in cardiovascular-systems (CVS) hemodynamic models that take antihypertensive exposure as an external input to an Emax inhibition of the cardiac-output production rate.
• Units: ng/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as the Hill term of an Emax inhibition expression EFF = emax * CP_PROPRANOLOL_NGML / (ic50 + CP_PROPRANOLOL_NGML) subtracted inside the CO production term (Snelder 2013 Equations 4 and 6). Set to 0 outside the exposure window. Reference values observed: the fitted IC50 in conscious spontaneously hypertensive rats is 9.82 ng/mL (Snelder 2013 Table 6) under continuous exposure at 1 mg/mL in drinking water.
• Source aliases:
  • C (Snelder 2013 Equation 6; values in ng/mL).
• Example models: Snelder_2013_cardiovascular_rat.R (Emax inhibition of the CO production rate; Emax and IC50 both estimated, 0.335 and 9.82 ng/mL).
• Notes: Specific scope; propranolol-specific. Follows the established CP_<drug>_<units> precedent. Snelder 2013 Table 3 used the three-compartment propranolol model of van Steeg et al. 2010 (Wistar-Kyoto rats) with absorption represented as an infusion of fixed 24 h duration (matching the drinking-water route) and the absorption rate constant estimated; no PK parameter values are published in Snelder 2013 and no propranolol PK model exists in the nlmixr2lib registry, so downstream users must supply this column externally. Propranolol is one of only two compounds in the Snelder 2013 training set acting primarily on cardiac output, and the only one of the two for which CO itself was not measured – the paper attributes the residual kout_TPR / FB2 correlation (-0.984) partly to that gap.

STIM_QUININE_MM (canonical for applied quinine HCl dihydrate stimulus concentration in a brief-access taste aversion experiment)

• Description: Applied sipper-tube concentration of quinine HCl dihydrate (mM) directly presented to the rat during an 8-second brief-access taste aversion (BATA) trial. This is the experimental stimulus level that contacts the taste receptors in solution – not a systemic plasma concentration of an absorbed drug. Drives the sigmoid-Emax effect on the logistic mixing probability in Sheng 2016’s two-generalized-Poisson mixture model for bimodal lick-count data.
• Units: mM
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as a Hill term in the logistic-Emax expression E = E0 + Emax * STIM_QUININE_MM^c / (RIC50^c + STIM_QUININE_MM^c) (Sheng 2016, Methods, “Model for the drug effect”). Set to 0 for the water (control) presentation. Reference values observed: Sheng 2016 presented seven concentrations 0 / 0.01 / 0.03 / 0.1 / 0.3 / 1 / 3 mM (Table 1); the final-model RIC50 is 0.0423 mM.
• Source aliases:
  • QUININE (Sheng 2016 NM-TRAN $INPUT column name for the per-record applied stimulus concentration; values in mM).
• Example models: Sheng_2016_quinine_rat.R (sigmoid Emax on the logistic mixing probability between a low-count generalized-Poisson distribution and a right-truncated high-count generalized-Poisson distribution; STIM_QUININE_MM is the only model covariate).
• Notes: Specific scope because the canonical name is bound to quinine HCl dihydrate as the bitter stimulus. Distinct from CAV (systemic average drug plasma concentration over a dosing interval) and the CP_* family (instantaneous plasma concentration as time-varying PD driver) – STIM_QUININE_MM is the applied stimulus concentration in solution that contacts the taste receptors directly, with no PK absorption / distribution involved. Future BATA-style extractions with a different bitter compound (caffeine, denatonium, etc.) should register a parallel canonical (e.g. STIM_CAFFEINE_MM) rather than overload this name; the STIM_<drug>_<units> pattern mirrors the established CP_<drug>_<units> precedent.

STIM_ARTESUNATE_NM (canonical for applied in-vitro artesunate concentration driving a P. falciparum hypoxanthine-uptake-inhibition model)

• Description: Applied artesunate concentration in the in vitro hypoxanthine-uptake-inhibition assay well (nM). Per-record covariate (the applied well concentration that contacts the parasite culture directly; the model has no PK and no time evolution). Drives the sigmoid Emax inhibition of normalised hypoxanthine uptake by clinical Plasmodium falciparum isolates.
• Units: nM
• Type: continuous
• Scope: specific
• Reference category: n/a – enters the sigmoid Emax inhibition expression as the concentration argument; set to 0 for the drug-free control well. Reference values observed: Simpson 2013 used a doubling-dilution series 0.044 to 87.0 nM plus a drug-free control well; the WT-reference (Genotype 1) population EC50 is 2.3 nM (Table 3).
• Source aliases:
  • C – used in Simpson_2013_artesunate.R (per-record applied well concentration in nM).
• Example models: Simpson_2013_artesunate.R (the only model covariate that varies per record; drives the sigmoid Emax inhibition with the pfmdr1 genotype effects entering on EC50).
• Notes: Specific scope because the canonical name is bound to artesunate as the applied antimalarial stimulus. Member of the in-vitro applied-drug-concentration STIM_<drug>_<units> family (siblings STIM_CHLOROQUINE_NM, STIM_LUMEFANTRINE_NM, STIM_MEFLOQUINE_NM, and the bitter-stimulus STIM_QUININE_MM) – the applied stimulus / well concentration that contacts the target directly, distinct from CAV (systemic average plasma concentration), the CP_<drug> plasma-PD-driver family, and the CONC_<drug>_MGL in-vitro antibacterial family (which is reported in mg/L).

STIM_CHLOROQUINE_NM (canonical for applied in-vitro chloroquine concentration driving a P. falciparum hypoxanthine-uptake-inhibition model)

• Description: Applied chloroquine concentration in the in vitro hypoxanthine-uptake-inhibition assay well (nM). Per-record covariate (the applied well concentration that contacts the parasite culture directly; the model has no PK and no time evolution). Drives the sigmoid Emax inhibition of normalised hypoxanthine uptake by clinical Plasmodium falciparum isolates.
• Units: nM
• Type: continuous
• Scope: specific
• Reference category: n/a – enters the sigmoid Emax inhibition expression as the concentration argument; set to 0 for the drug-free control well. Reference values observed: Simpson 2013 used a doubling-dilution series 10.02 to 10255.9 nM plus a drug-free control well; the WT-reference (Genotype 1) population EC50 is 242 nM (Table 3).
• Source aliases:
  • C – used in Simpson_2013_chloroquine.R (per-record applied well concentration in nM).
• Example models: Simpson_2013_chloroquine.R (the only model covariate that varies per record; drives the sigmoid Emax inhibition with the pfmdr1 genotype effects entering on EC50).
• Notes: Specific scope because the canonical name is bound to chloroquine as the applied antimalarial stimulus. Member of the in-vitro applied-drug-concentration STIM_<drug>_<units> family (siblings STIM_ARTESUNATE_NM, STIM_LUMEFANTRINE_NM, STIM_MEFLOQUINE_NM, and the bitter-stimulus STIM_QUININE_MM) – the applied stimulus / well concentration that contacts the target directly, distinct from CAV, the CP_<drug> family, and the CONC_<drug>_MGL in-vitro antibacterial family.

STIM_LUMEFANTRINE_NM (canonical for applied in-vitro lumefantrine concentration driving a P. falciparum hypoxanthine-uptake-inhibition model)

• Description: Applied lumefantrine concentration in the in vitro hypoxanthine-uptake-inhibition assay well (nM). Per-record covariate (the applied well concentration that contacts the parasite culture directly; the model has no PK and no time evolution). Drives the sigmoid Emax inhibition of normalised hypoxanthine uptake by clinical Plasmodium falciparum isolates.
• Units: nM
• Type: continuous
• Scope: specific
• Reference category: n/a – enters the sigmoid Emax inhibition expression as the concentration argument; set to 0 for the drug-free control well. Reference values observed: Simpson 2013 used a doubling-dilution series 2.40 to 235.8 nM plus a drug-free control well; the WT-reference (Genotype 1) population EC50 is 35.7 nM (Table 3).
• Source aliases:
  • C – used in Simpson_2013_lumefantrine.R (per-record applied well concentration in nM).
• Example models: Simpson_2013_lumefantrine.R (the only model covariate that varies per record; drives the sigmoid Emax inhibition with the pfmdr1 genotype effects entering on EC50).
• Notes: Specific scope because the canonical name is bound to lumefantrine as the applied antimalarial stimulus. Member of the in-vitro applied-drug-concentration STIM_<drug>_<units> family (siblings STIM_ARTESUNATE_NM, STIM_CHLOROQUINE_NM, STIM_MEFLOQUINE_NM, and the bitter-stimulus STIM_QUININE_MM) – the applied stimulus / well concentration that contacts the target directly, distinct from CAV, the CP_<drug> family, and the CONC_<drug>_MGL in-vitro antibacterial family.

STIM_MEFLOQUINE_NM (canonical for applied in-vitro mefloquine concentration driving a P. falciparum hypoxanthine-uptake-inhibition model)

• Description: Applied mefloquine concentration in the in vitro hypoxanthine-uptake-inhibition assay well (nM). Per-record covariate (the applied well concentration that contacts the parasite culture directly; the model has no PK and no time evolution). Drives the sigmoid Emax inhibition of normalised hypoxanthine uptake by clinical Plasmodium falciparum isolates.
• Units: nM
• Type: continuous
• Scope: specific
• Reference category: n/a – enters the sigmoid Emax inhibition expression as the concentration argument; set to 0 for the drug-free control well. Reference values observed: Simpson 2013 used a doubling-dilution series 1.62 to 1646.6 nM plus a drug-free control well; the WT-reference (Genotype 1) population EC50 is 53.0 nM (Table 3).
• Source aliases:
  • C – used in Simpson_2013_mefloquine.R (per-record applied well concentration in nM).
• Example models: Simpson_2013_mefloquine.R (the only model covariate that varies per record; drives the sigmoid Emax inhibition with the pfmdr1 genotype effects entering on EC50).
• Notes: Specific scope because the canonical name is bound to mefloquine as the applied antimalarial stimulus. Member of the in-vitro applied-drug-concentration STIM_<drug>_<units> family (siblings STIM_ARTESUNATE_NM, STIM_CHLOROQUINE_NM, STIM_LUMEFANTRINE_NM, and the bitter-stimulus STIM_QUININE_MM) – the applied stimulus / well concentration that contacts the target directly, distinct from CAV, the CP_<drug> family, and the CONC_<drug>_MGL in-vitro antibacterial family.

STIM_OXYGEN_UM (canonical for applied oxygen concentration driving a hypoxia-activated-prodrug model)

• Description: Oxygen concentration in the medium, buffer or tissue immediately surrounding the cells (uM). The applied environmental stimulus level that contacts the target directly, on the same footing as the applied-drug members of the STIM_ family – not a systemic blood-gas measurement and not a haemoglobin-saturation percentage. Drives the oxygen inhibition of one-electron bioreductive activation in hypoxia-activated-prodrug (HAP) models, where the activating nitro radical anion is back-oxidised by O2 so that prodrug turnover is near-maximal under anoxia and effectively switched off in well-oxygenated cells. Typically enters as the hyperbolic factor ko2 / (ko2 + STIM_OXYGEN_UM) multiplying a maximal (anoxic) metabolic rate constant, and may reappear in an exposure-response term when cell kill is taken to be proportional to the rate of reductive metabolism.
• Units: uM
• Type: continuous
• Scope: specific
• Reference category: n/a – enters a hyperbolic inhibition term as the concentration argument; set to 0 for the anoxic arm. Reference values observed: 0 uM = anoxia (95% N2 / 5% CO2 gas phase); 180 uM = humidified 20% O2 in the gas phase at 37 degrees C (Hong 2019 Supplementary Methods, citing Wenger 2015); < 1 uM is Hong 2019’s operational definition of the tumor hypoxic fraction; the fitted half-inhibition constant for evofosfamide is 0.27 uM (0.2 mmHg).
• Source aliases:
  • [O2] – used in Hong_2019_evofosfamide_cellular.R (main-text equations 3 and 5).
  • O2, pO2 – notation variants; note that a value quoted in mmHg must be converted, as this column is in uM (Hong 2019 prints both, 0.2 mmHg = 0.27 uM).
• Example models: Hong_2019_evofosfamide_cellular.R (oxygen inhibition of TH-302 bioreduction via kmet = ko2 / (ko2 + STIM_OXYGEN_UM) * kmet0, and the same factor scaling the ‘no bystander’ cell-kill model of equation 5).
• Notes: Specific scope pending a second HAP paper; the concept generalises directly to tirapazamine, PR-104A, SN30000 and other bioreductive prodrugs, which share this functional form. Member of the applied-stimulus STIM_<agent>_<units> family, whose non-drug members already include the bitter-taste stimulus STIM_QUININE_MM and the ligand input nodes STIM_EGF_NORM / STIM_IGF1_NORM / STIM_IL6_NORM / STIM_TNFA_NORM / STIM_DHT_NORM; oxygen is an applied environmental stimulus rather than an administered drug, which is why it belongs here rather than in CONC_<drug>_MGL or the CP_<drug> plasma-driver family. Distinct from the three oxygen-adjacent columns already registered, none of which is interchangeable with it: STO2 is a regional tissue haemoglobin saturation in percent, PACO2 is an arterial blood-gas partial pressure of a different gas, and OXYSUP_LOW / OXYSUP_HIGH are binary indicators of supplemental-oxygen therapy. A future model needing oxygen as a partial pressure rather than a concentration should register a sibling (e.g. STIM_OXYGEN_MMHG) rather than aliasing onto this one, because the conversion is temperature- and solubility-dependent.

ETSEVO (canonical for end-tidal sevoflurane concentration in the breathing circuit)

• Description: End-tidal sevoflurane concentration (vol %) recorded continuously by an anesthesia gas monitor during sevoflurane general anesthesia. Distinct from a systemic plasma concentration – ETSEVO is the alveolar / breathing-circuit concentration in vol %, used directly as the PD driver in sigmoid-Emax models of consciousness / recovery endpoints. Per-record covariate (per observation epoch); decreases during emergence from approximately the maintenance MAC value (e.g. 1 MAC, 1.5-1.7 vol % in school-aged children with 50% N2O) to 0.
• Units: vol % (volume percent in the breathing circuit; equivalent to fractional inspired/expired sevoflurane x 100)
• Type: continuous
• Scope: specific
• Reference category: n/a – enters directly as the concentration argument C in a sigmoid Emax probability expression P(ROC) = C50^gamma / (C50^gamma + C^gamma) (Shin 2014 Methods). Reference values observed: typical C50 for return of consciousness 0.37 vol % (mentally intact) and 0.19 vol % (mentally disabled) in Shin 2014 Table 2; literature MAC-awake for sevoflurane is 0.6-0.78 vol % in healthy adults / children.
• Source aliases:
  • DOSE – used in Shin_2014_sevoflurane.R (Shin 2014 Appendix 1 $INPUT column name; the NONMEM column is labelled DOSE but holds the per-record end-tidal concentration, not an administered dose – the model uses it directly in PROB = 1 - DOSE**GAM/(CE50**GAM + DOSE**GAM)).
• Example models: Shin_2014_sevoflurane.R (drives the sigmoid-Emax probability of return of consciousness in pediatric dental-surgery patients during emergence from sevoflurane / N2O general anesthesia).
• Notes: Specific scope because the canonical name is bound to sevoflurane as the volatile anesthetic. Future emergence-from-anesthesia PD extractions for a different volatile (isoflurane, desflurane) should register a parallel canonical (e.g. ETISO, ETDES) rather than overload this name; the ET<agent> pattern parallels the CP_<drug>_<units> precedent for IV PD drivers. Distinct from CP_* (systemic plasma concentration), CAV (steady-state average plasma exposure), and STIM_* (applied non-systemic stimulus concentration): ETSEVO is the alveolar / circuit concentration that equilibrates with brain tissue during inhalation anesthesia, sampled at the exhalation peak.

CEFFECT (canonical for time-varying effect-site (biophase) concentration of an IV anesthetic / sedative supplied as a PD driver)

• Description: Instantaneous (per-event-record) effect-site (biophase) concentration of an IV anesthetic, sedative, or other agent whose pharmacodynamics are commonly modelled against the effect-site Ce rather than plasma Cp. Supplied directly as a time-varying covariate column rather than computed from a coupled PK + effect-site model inside the consuming PD model file. Used in sigmoid Emax / probability-of-event PD-only analyses (e.g., probability of recovery of consciousness, probability of LOC, BIS-response models) where the source paper fixed the upstream PK + effect-site from an external target-controlled-infusion (TCI) controller (Schnider 1998 / 1999 for propofol; Minto 1997 for remifentanil; etc.) and fit only the downstream PD parameters. The PD model consumes CEFFECT directly in model (typically conc <- CEFFECT followed by a sigmoid Emax of conc); the user is responsible for generating the CEFFECT trajectory upstream (TCI-controller simulation, a registered propofol PK model chained with an external Keo effect-site link, or observed plasma concentrations equilibrated with an effect-site rate constant).
• Units: drug-specific; document the actual unit per-model via covariateData[[CEFFECT]]$units. Reference values observed: propofol effect-site Ce in Koo_2012_propofol.R is reported in ug/mL.
• Type: continuous
• Scope: general (the abstract concept “effect-site PD driver” applies to any drug whose PD is fit on Ce; the drug-specific identity is conveyed by the surrounding model file, not by an embedded drug suffix on the column name). Drug-specific override aliases are documented per-model in covariateData[[CEFFECT]]$notes when the source paper uses a different column name.
• Reference category: n/a – enters directly as the concentration argument C in a sigmoid Emax expression such as P(ROC) = C50^lambda / (C50^lambda + C^lambda). Set to 0 for the drug-free reference (e.g., before infusion start or after full washout). Reference values observed: propofol effect-site Ce range approximately 4.4 +/- 1.1 ug/mL at LOC, 3.2 +/- 1.0 ug/mL at end of surgery, and 1.1 +/- 0.3 ug/mL at ROC in Koo 2012’s adult population (Results).
• Source aliases:
  • Ce (Koo 2012 Methods; the per-record propofol effect-site concentration predicted by the Schnider TCI controller, ug/mL). Not stored as a single named NONMEM column in the source $TABLE – the TCI controller computes it and feeds the NONMEM dataset directly.
  • Cv(t) (Crass 2025 Data S1; the per-record individual-predicted pegcetacoplan concentration in the vitreous humour of an intravitreally dosed eye, ug/mL, generated by an external population PK model and supplied to the PD model as data).
• Example models: Koo_2012_propofol.R (drives the sigmoid Emax probability of return of consciousness in adult eye / ENT patients during emergence from propofol-remifentanil TCI anesthesia; CEFFECT carries propofol effect-site Ce in ug/mL), Crass_2025_pegcetacoplan_ga_exposureresponse.R (drives the linear-in-log-concentration reduction of the geographic-atrophy lesion growth rate, (1 + e_ceffect_slope_study * log(CEFFECT + 1)); CEFFECT carries vitreous-humour pegcetacoplan concentration in ug/mL, computed upstream from the intravitreal dosing history and an assumed 4 mL vitreous volume). Sang_2021_doxorubicin_human_qsp.R (heart-tissue doxorubicin concentration in ug/mL, integrated in-model to the cumulative heart AUC that drives the bioenergy-production effect; Sang 2021 obtained it from the He 2018 PBPK model).
• Notes: General-scope canonical for the abstract “effect-site PD driver” concept; the per-model units field tells the user which drug and which scale the trajectory must use. Distinct from CP_* (systemic plasma concentration, mass-balance from the modeled-drug central compartment), CAV (dosing-interval-averaged steady-state plasma exposure), STIM_* (applied in-vitro / sipper-tube stimulus concentration), ETSEVO / ETISO / ETDES (alveolar end-tidal volatile-anesthetic concentration in vol %), and L_OPIOID_pM / L_ANTAGONIST_pM (Mann 2022 multi-ligand competitive-binding effect-site slots in pM, where the SLOT identity is the abstraction rather than the drug identity). The CEFFECT canonical is the name the maintainers ratified for the “effect-site PD driver” family, so a new effect-site driver takes it without a fresh ruling. Future PD extractions that consume an effect-site Ce as a PD driver should reuse this canonical with the per-model units / notes recording the drug and the scale; if a future extraction requires simultaneous occupation of TWO separate effect-site slots (a competitive / interaction model akin to Mann 2022’s two ligand slots), register sibling canonicals (e.g., CEFFECT_AGONIST, CEFFECT_ANTAGONIST) rather than overloading CEFFECT with both.

FPG (canonical for baseline fasting plasma glucose)

• Description: Fasting plasma glucose concentration at baseline (or time-varying baseline-style observation; document per-model). Distinct from GLU (time-varying plasma glucose regressor input for mechanistic glucose-kinetics models).
• Units: mmol/L (or mg/dL – 1 mmol/L glucose is approximately 18.02 mg/dL). Document per-model via covariateData[[FPG]]$units.
• Type: continuous
• Scope: general
• Reference category: n/a – used with linear-deviation form 1 + theta * (FPG - ref). Reference values observed: 8.90 mmol/L (Retlich 2015 popPK/PD linagliptin median fasting glucose at baseline).
• Source aliases: none known.
• Example models: Retlich_2015_linagliptin.R (mmol/L, reference 8.90; linear-deviation effect on baseline DPP-4 activity BSL with coefficient 1.46 % per mmol/L deviation).
• Notes: Glycemic-control covariate (baseline FPG); routinely reported alongside HbA1c in T2DM populations. Distinct from GLU which is a time-varying within-subject glucose regressor for mechanistic glucose-kinetics models (Bizzotto 2016).

DPP4_BL_RFU (canonical for baseline plasma dipeptidyl peptidase-4 activity in relative fluorescence units)

• Description: Baseline plasma dipeptidyl peptidase-4 (DPP-4) enzymatic activity, measured by relative fluorescence units (RFU). DPP-4 is the pharmacological target of the gliptin (DPP-4-inhibitor) drug class; baseline activity correlates with circulating DPP-4 protein concentration in the central compartment and serves as a covariate on the central-compartment binding-site concentration in TMDD models for gliptins.
• Units: RFU (assay-specific; document the assay in covariateData[[DPP4_BL_RFU]]$notes since RFU values are not directly comparable across assays).
• Type: continuous
• Scope: specific
• Reference category: n/a – used with linear-deviation form 1 + theta * (DPP4_BL_RFU - ref). Reference values observed: 12,497 RFU (Retlich 2015 popPK linagliptin median baseline), 11,600 RFU (Retlich 2015 popPK/PD linagliptin median baseline, applied to the individual-predicted BSL_i parameter on EC50).
• Source aliases: none known.
• Example models: Retlich_2015_linagliptin.R (RFU, reference 12,497; linear-deviation effect on the central-compartment binding-site concentration Bmax,C with coefficient 0.00332 % per RFU deviation – captures the inter-individual correlation between baseline DPP-4 protein concentration and the apparent saturable-binding amplitude).
• Notes: Specific scope because DPP-4-activity values in RFU are assay-specific and not directly transferable between studies / instruments. Future gliptin extractions reporting DPP-4 activity in the same assay can reuse this canonical; extractions in absolute enzymatic-rate units (pmol AMC per minute) or normalised units should register a sibling canonical (e.g., DPP4_BL_PMOL_MIN). See GST_BL_NMOL_MIN_ML for the first sibling registered under the <ENZYME>_BL_<RATE-UNITS> pattern these notes prescribe.

GST_BL_NMOL_MIN_ML (canonical for baseline glutathione S-transferase enzyme activity in nmol/min/mL)

• Description: Baseline whole-blood glutathione S-transferase (GST) enzymatic activity in absolute rate units. GST catalyses conjugation of glutathione to electrophilic substrates and is the principal metabolic route for busulfan and other alkylating agents, so baseline GST activity is a mechanistic covariate on glutathione-dependent metabolic capacity. Measured in the founding example by micro-quartz colorimetry, following GST-catalysed conjugation of glutathione to 1-chloro-2,4-dinitrobenzene at 340 nm.
• Units: nmol/min/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – used with power scaling (GST_BL_NMOL_MIN_ML / ref)^exponent. Reference values observed: 9.2 nmol/min/mL (Cao 2025 model-development cohort median; observed range 0.9-20.7, with 3.7-12.4 in the held-out evaluation set).
• Source aliases:
  • GST – the $INPUT column name in the Cao 2025 Supplementary Text S2 control stream and the Table 1 / Table 3 column label.
• Example models: Cao_2025_busulfan.R (power effect e_gst_sdep_gsh = 0.28 on the glutathione-depletion scaling factor sdep_gsh, centered on the 9.2 nmol/min/mL cohort median).
• Notes: The name follows the <ENZYME>_BL_<RATE-UNITS> pattern that DPP4_BL_RFU’s own Notes prescribe for extractions reporting an absolute enzymatic rate; the unit token is load-bearing because colorimetric GST assays are not calibrated to a common scale, so a future study reporting U/L or umol/min/L must register a sibling (GST_BL_U_L) rather than reuse this canonical on an incomparable scale. Distinct from GST genotype covariates: the GSTA1 haplotype is a SNP_<GENE>_RS<rsid>-family categorical, whereas this is the measured phenotype, and Cao 2025 argues the phenotype is the better covariate precisely because GST expression varies with age and sex beyond genotype. Also distinct from GGT (gamma-glutamyl transferase), an unrelated enzyme whose abbreviation is a near-miss in fuzzy name lookups. The _BL_ token is retained because activity was assayed once, before the first dose; a future time-varying GST-activity model should register a separate canonical rather than redefine this one.

GLU (canonical for plasma glucose time-course regressor)

• Description: Plasma glucose concentration as a time-varying regressor input that drives a mechanistic glucose-kinetics model. Not a covariate that modifies a parameter; the model integrates GLU directly through a smoothing filter into a site-of-action glucose variable.
• Units: mmol/L
• Type: continuous
• Scope: specific
• Reference category: n/a – used as a time-varying regressor. The model declares linear(GLU) so rxode2 linearly interpolates GLU between dataset rows.
• Source aliases:
  • iglu (glucose at the current row time) – used in the DDMORE bundle’s Simulated_glucoseKinetics.csv for DDMODEL00000227. Rename iglu -> GLU before passing to rxSolve.
• Example models: Bizzotto_2016_glucose.R (driving regressor for the glucose-at-site-of-action delay), Lu_2014_sglt_qsp.R (drives the rate of glucose entry into PCT1 by glomerular filtration in the SGLT renal-glucose-reabsorption QSP model: filtered glucose load = GFR * GLU mmol/h), Bosch_2025_glp1ra_hba1c.R (drives the IGRH HbA1c sub-model: time-varying average daily glucose -> glycation flux KGGLU_mgdl and glucose-dependent RBC life span (GLU_mgdl/149)^gamma_ls; internal conversion mmol/L -> mg/dL via factor 18.016), Mondick_2018_empagliflozin_t1dm.R, Mondick_2018_empagliflozin_t2dm.R (the interpolated plasma glucose “iPG” of the Mondick 2018 renal-threshold model: drives BOTH the glomerular filtration term eGFR * iPG and the saturable tubular reabsorption term Gmax * iPG / (KM + iPG), whose difference is the urinary glucose excretion rate; internal conversion mmol/L -> mg/dL via factor 18.016 because the paper works in mg/dL), Larizza_2018_insulinMinimalModel.R (the IVGTT glucose profile driving the second-phase provisionary factor of the Magni 2004 insulin minimal model: -alpha * (provision - beta * (GLU - gthresh) * (GLU > gthresh)), i.e. only the excess above the estimated threshold gthresh stimulates secretion. Supplied in mg/dL, not mmol/L, with no internal conversion – the estimated threshold of 89.0 mg/dL and the subject’s basal 87 mg/dL are on that scale; see the per-model covariateData[[GLU]]$notes for how the source’s own “pmol/L” label was shown to be a slip), Sokolov_2020_sglt_qsp.R (arm-level mean daily plasma glucose held constant as the plasma glucose that is filtered into the proximal tubule: filtered load = GFR GLU mmol/h, in the dapagliflozin / empagliflozin / canagliflozin renal SGLT1/2 QSP model).
• Notes: Specific scope because GLU is meaningful only for glucose-kinetics or glucose-PD models that take plasma glucose as an exogenous regressor. The DDMORE bundle’s hand-rolled piecewise-linear interpolation (GL = (t-T1)/(TOBS-T1)*(GLU-GLU1)+GLU1 with bracketing columns iglu / glun / td / tn) is replaced in nlmixr2 by linear(GLU) declared in model; the bracketing columns are not required.

INS (canonical for plasma insulin time-course regressor)

• Description: Plasma insulin concentration as a time-varying regressor input that drives a mechanistic glucose-kinetics model. Not a covariate that modifies a parameter; the model integrates INS directly through a smoothing filter into a site-of-action insulin variable.
• Units: pmol/L
• Type: continuous
• Scope: specific
• Reference category: n/a – used as a time-varying regressor. The model declares linear(INS) so rxode2 linearly interpolates INS between dataset rows.
• Source aliases:
  • iins (insulin at the current row time) – used in the DDMORE bundle’s Simulated_glucoseKinetics.csv for DDMODEL00000227. Rename iins -> INS before passing to rxSolve.
  • INSU – used in the DDMORE bundle’s Simulated_ddmoremockdata2.txt for DDMODEL00000228. Rename INSU -> INS before passing to rxSolve.
• Example models: Bizzotto_2016_glucose.R (driving regressor for the insulin-at-site-of-action delay), NA_NA_paracetamol.R (DDMODEL00000228 OGTT model: drives the insulin-on-glucose-elimination first-order effect compartment via kie * (INS / 6.945 - effect_ins)), Denti_2010_glucoseMinimal.R (Bergman minimal-model insulin forcing function entering the insulin-action ODE as p2 * si * (INS - INS_BL); units pmol/L; supplied via linear(INS)).
• Notes: Specific scope because INS is meaningful only for glucose-kinetics or insulin-PD models that take plasma insulin as an exogenous regressor. For drugs that modify circulating insulin as a downstream effect, use a different mechanism-specific name. The DDMORE bundle’s hand-rolled piecewise-linear interpolation (I = (t-T1)/(TOBS-T1)*(INS-INS1)+INS1 with bracketing columns iins / insn / td / tn) is replaced in nlmixr2 by linear(INS) declared in model; the bracketing columns are not required.

INS_BL (canonical for baseline (fasting) plasma insulin concentration)

• Description: Baseline (fasting) plasma insulin concentration, time-fixed per subject. Used as the per-subject anchor for steady-state insulin-driven processes (e.g., initial condition of an insulin-on-elimination effect compartment) and as the per-subject baseline-state insulin input to a baseline-glucose-production rate calculation.
• Units: pmol/L (or uU/mL; document per-model via covariateData[[INS_BL]]$units). The example model rescales via INS_BL / 6.945 to convert pmol/L to uU/mL.
• Type: continuous
• Scope: specific
• Reference category: n/a.
• Source aliases:
  • BASI (baseline insulin) – used in the DDMORE bundle’s Simulated_ddmoremockdata2.txt for DDMODEL00000228. Rename BASI -> INS_BL before passing to rxSolve.
• Example models: NA_NA_paracetamol.R (DDMODEL00000228 OGTT model: initialises the insulin-on-elimination effect compartment effect_ins(0) = INS_BL / 6.945 and feeds the steady-state baseline-glucose-production rate gpro = gss * (kg + kgi * INS_BL / 6.945) * vg * 180 / 1000), Hong_2013_glucose_insulin_HGC.R (anchors the dynamic-state insulin(0) = ICss * VI initial condition in mU plus the baseline insulin secretion ICss * CLI, no rescaling), Hong_2013_glucose_insulin_MTT.R (same anchor as the HGC companion), Denti_2010_glucoseMinimal.R (Bergman minimal-model basal-insulin anchor; covariate effect on lsi (-0.0282 per pmol/L) and lp2 (-0.0150 per pmol/L), and Ib in the insulin-action ODE p2 * si * (INS - INS_BL); units pmol/L despite the source paper’s apparent ‘pmol/ml’ table label, which is documented as an apparent typo), Gao_2012_exenatide_glucose_insulin_rat.R (rat glucose-insulin feedback model; anchors insulin(0) = Ib and the zero-order secretion rate kinI = koutI * Ib, and is the reference level in the insulin-on-glucose-disposal term (1 + SIns * (insulin - Ib)); supplied in pmol/L to match the source’s Fig. 5 / Fig. 8 axes and divided by 1000 inside model() because SIns is reported in 1/nM – the per-arm values are DIGITISED from Fig. 5, not published).
• Notes: Distinct from INS (time-varying regressor); INS_BL is a per-subject baseline-state anchor used in initial conditions and steady-state derived quantities, not the dynamic regressor itself. Specific scope because the conversion factor (1/6.945) and the rescaled-units interpretation are paper-specific; future extractions that report baseline insulin in mIU/L or pmol/L directly without rescaling can ratify the same canonical and document the per-model units / conversion in covariateData[[INS_BL]]$units / notes. Companion concept to FPG (baseline fasting plasma glucose).

PSA_BL (canonical for baseline serum prostate-specific antigen concentration)

• Description: Subject’s serum prostate-specific antigen (PSA) concentration at the baseline / randomisation visit, before study treatment, carried into a model as a time-fixed per-subject covariate. Used in benign prostatic hyperplasia (BPH) disease-progression and outcome models, where baseline PSA is a surrogate for prostate size and a known risk factor for acute urinary retention and BPH-related surgery.
• Units: ng/mL
• Type: continuous
• Scope: general
• Reference category: n/a – enters as a centred term. Reference value observed: 3.4 ng/mL (D’Agate 2021 pooled-population median).
• Source aliases:
  • B_PSA – D’Agate 2021 Supporting Information Figure S3 label (‘PSA concentration at baseline’).
• Example models: DAgate_2021_bph_aurs_mbma.R (founding example; log-linear effect on the hazard of first acute urinary retention or BPH-related surgery, exp(log(1.08) * (PSA_BL - 3.4))), Zou_2020_leuprorelin.R (merged from a duplicate register entry during merge dedup).
• Notes: Member of the <X>_BL per-subject-baseline family (INS_BL, FERRITIN_BL, D25OH_BL). The measured, unadjusted pre-treatment value: 5-alpha-reductase inhibitors such as dutasteride roughly halve PSA on treatment, and a model of on-treatment PSA as a time course must use a separate time-varying column or an output state, never this one. Distinct from PSA as a modelled tumour-marker output in prostate-cancer models (e.g. the PSA state of vanHasselt_2015_eribulin.R).

PROSTATE_VOL_BL (canonical for baseline prostate volume)

• Description: Subject’s total prostate volume at the baseline visit, usually by transrectal ultrasound, carried into a model as a time-fixed per-subject covariate. Used in BPH disease-progression and outcome models.
• Units: mL (equivalently cm^3; document per-model)
• Type: continuous
• Scope: general
• Reference category: n/a – enters as a centred term. Reference value observed: 48.5 mL (D’Agate 2021 pooled-population median).
• Source aliases:
  • B_PV – D’Agate 2021 Supporting Information Figure S3 label (‘prostate volume at baseline’).
  • PV – D’Agate 2021 main-text abbreviation.
• Example models: DAgate_2021_bph_aurs_mbma.R (founding example; log-linear effect on the hazard of first acute urinary retention or BPH-related surgery, exp(log(1.01) * (PROSTATE_VOL_BL - 48.5))).
• Notes: Member of the <X>_BL per-subject-baseline family. Total gland volume; a transition-zone volume or an ultrasound-estimated prostate weight is a different quantity and needs its own canonical. The organ is spelled out (rather than PV_BL) because PV is an ambiguous pharmacometric abbreviation (plasma volume, portal vein).

QMAX_BL (canonical for baseline maximum urinary flow rate)

• Description: Subject’s maximum urinary flow rate (Qmax) from uroflowmetry at the baseline visit, carried into a model as a time-fixed per-subject covariate. Lower values indicate more severe bladder outlet obstruction. Used in BPH disease-progression and outcome models.
• Units: mL/s
• Type: continuous
• Scope: general
• Reference category: n/a – enters as a centred term. Reference value observed: 10.2 mL/s (D’Agate 2021 pooled-population median).
• Source aliases:
  • Qmax, Q max – D’Agate 2021 main text and Table 2 (‘Baseline maximum urinary flow’).
• Example models: DAgate_2021_bph_aurs_mbma.R (founding example; log-linear effect on the hazard of first acute urinary retention or BPH-related surgery, exp(log(0.91) * (QMAX_BL - 10.2))).
• Notes: Member of the <X>_BL per-subject-baseline family. A peak flow RATE from a voiding study, not a volume. Distinct from URINE_FLOW (instantaneous urine production rate, mL/min, in renal-excretion models), and from the urine-volume collection canonicals URINE_VOL_24H / URINE_VOL_INTERVAL.

CINH (canonical for plasma SGLT-inhibitor concentration time-course regressor)

• Description: Plasma SGLT-inhibitor (e.g., dapagliflozin, canagliflozin) concentration as a time-varying regressor input that drives the rate of unbound-drug entry into the proximal tubule via glomerular filtration in renal-glucose-reabsorption QSP models. Not a covariate that modifies a parameter; the model integrates CINH directly through linear(CINH) and multiplies by the unbound fraction fup inside model.
• Units: nmol/L (matching the units of the inhibitor’s affinity constants Ki1 and Ki2 in the SGLT MM kinetics).
• Type: continuous
• Scope: specific
• Reference category: n/a – used as a time-varying regressor. The model declares linear(CINH) so rxode2 linearly interpolates CINH between dataset rows.
• Source aliases: none known.
• Example models: Lu_2014_sglt_qsp.R (drives the rate of unbound inhibitor entering PCT1 by glomerular filtration: filtered drug load = GFR * fup * CINH nmol/h; set CINH = 0 for baseline / no-inhibitor simulations).
• Notes: Specific scope because CINH is meaningful only for renal-glucose-reabsorption or other SGLT-mediated models that take plasma SGLT-inhibitor exposure as an exogenous regressor. The clinical reporting unit is ng/mL; convert by 1 ng/mL = (1000 / MW_drug) nmol/L (dapagliflozin MW = 409 g/mol so 1 ng/mL = 2.44 nmol/L; canagliflozin MW = 454 g/mol so 1 ng/mL = 2.20 nmol/L). The Lu 2014 paper feeds the model an interpolated dapagliflozin observed mean profile (DeFronzo et al. 2013) or a fitted two-compartment canagliflozin PK profile (Devineni et al. 2013) – the SGLT-mechanism model itself does not include an internal PK sub-model for the inhibitor. Companion concept to GLU (the plasma-glucose regressor that drives the same filtration arm).

IGE (canonical for serum total immunoglobulin E concentration)

• Description: Baseline serum total immunoglobulin E concentration (free IgE plus, in patients on anti-IgE therapy, omalizumab-IgE complex). For anti-IgE monoclonal antibodies (omalizumab, ligelizumab) IgE is the pharmacologic target; baseline IgE sets the magnitude of the target sink and modifies free-IgE clearance and the rate of IgE production in mechanism-based binding/turnover models.
• Units: ng/mL (typical clinical-PK convention). Pretreatment values reported in IU/mL are converted via 1 IU/mL = 2.42 ng/mL (Hayashi 2007 Methods). Document per-model via covariateData[[IGE]]$units.
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (IGE / ref)^exponent. Reference value observed: 482.4 ng/mL (Hayashi 2007 Japanese atopic-asthma cohort).
• Source aliases:
  • IgE0 (baseline IgE concentration) – used in Hayashi_2007_omalizumab.R.
• Example models: Hayashi_2007_omalizumab.R (ng/mL, reference 482.4; power exponents -0.281 on apparent CL of free IgE and +0.657 on apparent IgE production rate; also used as the initial value for the total-IgE state at t = 0).
• Notes: General scope because baseline serum total IgE is a routine clinical-laboratory measurement, not a target tied to one drug. In mechanism-based anti-IgE binding/turnover models the in-model IgE state is a separate dynamic variable (X_TE, in nmol or nmol/L) – IGE is the per-subject baseline column used for covariate scaling and (when applicable) state initialization, not the dynamic state itself. For models that use the alternative reporting unit IU/mL, multiply by 2.42 before applying the canonical-units (ng/mL) reference value, or document the per-model unit choice in covariateData[[IGE]]$units so downstream tooling can interpret the values correctly. Distinct from IGE_FREE (the time-varying serum free IgE concentration used as the PD-driver input in PD-only IgE-FEV1 models).

IGE_FREE (canonical for time-varying serum free immunoglobulin E concentration as PD-driver input)

• Description: Time-varying serum free immunoglobulin E concentration (the unbound fraction not in drug-IgE complex) used as the exogenous PD-driver input in PD-only IgE-FEV1 (and other anti-IgE biomarker) models that consume free IgE as an externally-supplied trajectory rather than computing it from a coupled PK / binding ODE. Per-event-record (one value per PD observation time). Cross-reference: distinct from IGE (per-subject baseline total IgE used for power scaling in mechanism-based binding / turnover popPK models like Hayashi 2007). For sequential PK-PD use, this column is populated from an upstream omalizumab popPK/IgE-binding model (e.g. modellib("Hayashi_2007_omalizumab") emits a freeIgE observable in ng/mL); for standalone fits the column is interpolated from observed free-IgE assays.
• Units: ng/mL (matches the Zhu 2023 paper’s reported values and the Hayashi 2007 freeIgE output convention). Document per-model via covariateData[[IGE_FREE]]$units.
• Type: continuous
• Scope: general (anti-IgE PD modeling – omalizumab today, plausibly ligelizumab or other anti-IgE agents in future).
• Reference category: n/a – enters directly into the inhibition Hill term imax * IGE_FREE^hill / (ec50^hill + IGE_FREE^hill). Set to the subject’s baseline total IgE at pretreatment / placebo records (per Zhu 2023 Methods: free IgE equals total IgE before omalizumab treatment) so the inhibition term evaluates at the subject’s natural free-IgE level. Reference values observed: 16-960 ng/mL range across placebo + omalizumab observations in pediatric Study IA05 (Zhu 2023 Results paragraph).
• Source aliases:
  • C_IgE,i(t) (Zhu 2023 paper symbol in the structural-model equation, PDF page 2). The NM-TRAN dataset column name is not disclosed in the paper text.
• Example models: Zhu_2023_omalizumab_pediatric.R (founding example; ng/mL, per-event-record free IgE driving the IDR-Type-IV FEV1 percent predicted response with imax = 0.0717, ec50 = 39.4 ng/mL, hill = 9 fixed).
• Notes: Distinct from IGE (baseline total IgE for power-scaling). The free-IgE-vs-total-IgE distinction is load-bearing for anti-IgE drugs: free IgE is the pharmacologically active fraction and the relevant PD driver, while total IgE is the assay-reported sum of free + drug-bound IgE that does NOT decline under omalizumab treatment (Zhu 2023 Methods ‘Data Used in the Population Pediatric IgE-FEV1 Model’). Sequential PK-PD compatibility: Hayashi_2007_omalizumab.R emits a freeIgE observable in ng/mL; downstream users simulate that model and populate the IGE_FREE column in the PD dataset from the simulated trajectory. For placebo subjects (no anti-IgE drug), the paper used the subject’s average total IgE level across the steroid-stable period as the IGE_FREE input, because free IgE equals total IgE in the absence of anti-IgE drug (Zhu 2023 Methods).

ESAD (canonical for prior erythropoiesis-stimulating-agent (ESA) dose at baseline)

• Description: Subject’s recorded prior erythropoiesis-stimulating-agent (ESA) dose at study baseline, in epoetin-equivalent activity units per week. Used by ESA-switch population PK/PD models for hemoglobin response to a new ESA when prior endogenous-erythropoietin levels were not measured – the prior-ESA dose acts as a surrogate for the residual hematopoietic stimulation present at randomization. Time-fixed per subject (records the patient’s stable maintenance dose immediately before the new-ESA start).
• Units: units/week (epoetin-equivalent activity units per week). Document per-model via covariateData[[ESAD]]$units when the source paper uses a different per-time unit (units/month, units/day) or converts darbepoetin / methoxy-polyethylene-glycol epoetin beta doses to epoetin equivalents.
• Type: continuous
• Scope: specific
• Reference category: n/a – used with a paper-specific effect on a PD parameter: a log-scale slope on baseline hemoglobin (HgbBL) in Naik_2013_peginesatide.R, a normalized power term on SC50 in Chanu_2010_methoxyPolyethyleneGlycolEpoetinBeta.R. Reference values observed: 7996 units/week in Naik_2013_peginesatide.R (Naik 2013 eq 16; paper-reported population median) and 7000 units/week in Chanu_2010_methoxyPolyethyleneGlycolEpoetinBeta.R (size-weighted mean of the two maintenance-study medians in Chanu 2010 Table I; the paper does not print the pooled median it centred on). The covariate effect is gated by an ESADF indicator (1 if ESAD > 0, else 0) so that subjects with missing or unrecorded prior ESA dose (encoded as ESAD = 0) carry no covariate adjustment, matching Naik 2013’s “no effect of ESAD was incorporated for subjects whose ESAD dose information was not available.”
• Source aliases:
  • ESAD – used in Naik_2013_peginesatide.R (Naik 2013 paper notation; prior epoetin alfa / darbepoetin alfa weekly dose in units/week for CKD hemodialysis subjects enrolling on peginesatide).
  • DEPO – used in Chanu_2010_methoxyPolyethyleneGlycolEpoetinBeta.R (Chanu 2010 notation; “previous weekly EPO dose at the start of C.E.R.A. maintenance treatment”, IU/week, for CKD dialysis subjects switching from epoetin to methoxy polyethylene glycol-epoetin beta).
• Example models: Naik_2013_peginesatide.R (Naik 2013 eq 16; exponential covariate on the baseline-hemoglobin parameter: hgbbl = exp(lhgbbl + etalhgbbl + e_esad_lhgbbl * (ESAD - 7996) * ESADF) with e_esad_lhgbbl = -4.49e-7 1/(units/week); effect is small in magnitude but retained as the only PD-side statistically significant covariate per backward-elimination at P < 0.005), Chanu_2010_methoxyPolyethyleneGlycolEpoetinBeta.R (Chanu 2010 Table IV; normalized power covariate on SC50, ec50 = exp(lec50 + etalec50) * (CRP / 5)^0.319 * esad_ratio^0.303 where esad_ratio collapses to 1 for ESA-naive subjects; a 5-fold rise in prior epoetin dose raises SC50 by 63%, the only PD covariate the paper judges clinically relevant, and the basis for setting the C.E.R.A. starting dose from the prior ESA dose).
• Notes: Specific scope because the values are intrinsically tied to ESA-switch popPK/PD studies and the units/week reporting convention is paper-specific (different ESAs have different specific activities; epoetin-equivalent conversion factors must be documented per-model). The ESADF indicator (built inline in the model() block as (ESAD > 0)) handles the “no prior ESA dose available” data-quality case used by Naik 2013; Chanu_2010_methoxyPolyethyleneGlycolEpoetinBeta.R reuses the same indicator for a stronger purpose, as the ESA-naive / ESA-treated switch that also selects the hemoglobin initial condition and turns the prior-ESA-withdrawal term on or off, so in that model ESAD = 0 means genuinely ESA-naive rather than merely unrecorded. Future papers that need to distinguish prior-ESA-naive from prior-ESA-treated-with-unrecorded-dose within one dataset may still register a parallel canonical (e.g., ESA_NAIVE).

DOSE_IND (canonical for per-arm once-daily inhaled indacaterol dose)

• Description: Per-study-arm once-daily inhaled indacaterol dose (ug/day; 0 for placebo arms). Study-arm-level (not individual-level) drug-exposure covariate used as the dose regressor in a model-based meta-analysis (MBMA) of bronchodilator dose-response.
• Units: ug/day
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as the dose regressor in the MBMA (Emax-type) bronchodilator dose-response; set to 0 for placebo arms.
• Source aliases:
  • DOSE_IND – used in Renard_2011_indacaterol.R (per-arm indacaterol dose, ug/day; Renard 2011 Table 1).
• Example models: Renard_2011_indacaterol.R (per-arm once-daily inhaled indacaterol dose driving the MBMA dose-response; dose range 18.75-600 ug/day across 11 trials, with the six discrete reported doses 18.75, 37.5, 75, 150, 300, and 600 ug).
• Notes: Specific scope because the value is intrinsically tied to indacaterol and the Renard 2011 MBMA dose-response. Drug-specific dose canonical paralleling DOSE_PHT_MGKGD (phenytoin); future MBMA / dose-response models for other drugs should register a sibling DOSE_<DRUG> canonical rather than overloading this name.

CELLS_INTACT (canonical for the intact-versus-lysed cell-preparation flag in an in-vitro target-binding assay)

• Description: Binary flag recording which cell preparation an in-vitro target-binding observation came from: 1 = intact (whole) bacterial cells, in which the outer membrane is present and drug must penetrate it to reach a periplasmic or intracellular target; 0 = lysed cells, i.e. isolated target-containing membrane fractions, in which the outer-membrane barrier has been removed and drug is applied in vast excess. The flag is a property of the sample preparation, not of the bioanalytical measurement method – it is therefore distinct from the ASSAY_<METHOD> bioanalytical-method family.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 = lysed cells (isolated membranes). Structural, not a coefficient multiplier: the flag gates both the outer-membrane influx term (rate_influx = CELLS_INTACT * rate_influx_scaled * CONC_<DRUG>_MGL) and the periplasmic initial condition (periplasm(0) = (1 - CELLS_INTACT) * n_peri_lysed).
• Source aliases:
  • INTACT – used in the Lopez-Arguello 2023 S-ADAPT-TRAN estimation code (Fig. S8 line 66, IF (INTACT.EQ.1) THEN).
• Example models: LopezArguello_2023_<drug>_qsp.R.
• Notes: General scope because the intact-versus-lysed contrast is a standard design in Gram-negative target-site-penetration work and is not specific to beta-lactams or to PBPs; a new model using the same design should reuse this name rather than register a sibling. Because the flag changes model structure rather than scaling a parameter, it has no associated e_<cov>_<param> covariate-effect coefficient.

BUFFER_CA_FREE (canonical for the calcium-free-buffer (open-tight-junction) flag in a sandwich-cultured hepatocyte transport assay)

• Description: Binary flag recording which incubation buffer a sandwich-cultured hepatocyte (SCHH) observation came from: 1 = Ca2+/Mg2+-free buffer containing EGTA, which disrupts the tight junctions so that the contents of the bile canalicular networks wash out into the medium and the measured cell lysate is Cells alone; 0 = standard Ca2+-containing buffer, which maintains the tight junctions so that the canaliculi retain drug and the measured cell lysate is Cells + Bile. This is the B-CLEAR paired-incubation design that makes biliary clearance identifiable in sandwich-cultured hepatocytes, and it is a property of the buffer the well was incubated in, not of the bioanalytical method – it is therefore distinct from the ASSAY_<METHOD> family. It is also distinct from CELLS_INTACT, which records whether cells were lysed at all; here the cells are intact under both levels and only the junctions between them differ.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 = standard Ca2+-containing buffer (tight junctions intact). Structural, not a coefficient multiplier: the flag routes the biliary efflux flux, so it has no associated e_<cov>_<param> coefficient. In the founding example it selects between d/dt(buffer) <- clbl * ctCells + kflux * bile -... with a separate accumulating bile state (level 0) and d/dt(buffer) <- (clbl + clbile) * ctCells -... with bile held at zero (level 1).
• Source aliases:
  • + / - superscripts on the buffer and cell states – Guo 2016 eqs. 1-5, where “‘+’ and ‘-’ refer to Ca2+-containing (standard HBSS) and Ca2+-free HBSS, respectively”. Note the inverted orientation: the paper’s + superscript is the canonical’s 0 level.
• Example models: Guo_2016_taurocholate_schh.R (both levels; the paired standard / Ca2+-free incubations were fitted simultaneously to resolve CLBile from CLBL).
• Notes: General scope because the paired standard-versus-calcium-free incubation is the defining design of the sandwich-cultured hepatocyte biliary-clearance assay and is not specific to bile acids or to any one transporter; a new SCHH extraction using the same design should reuse this name. Buffer identity (Hanks’ balanced salt solution in the founding example) is not encoded in the name because the calcium-free manipulation, not the base salt solution, is what the flag captures. Where a paper reports a biliary excretion index rather than fitting the paired incubations mechanistically, no covariate is needed and this column should not be introduced.

WASH_ACTIVE (canonical for the buffer-wash-in-progress indicator in an in-vitro cell-monolayer assay)

• Description: Binary, time-varying indicator: 1 while the incubation buffer is actively being washed off an in-vitro cell monolayer, 0 otherwise. Gates a first-order wash-out term on the buffer / medium state so that the drug remaining in the medium is cleared over the wash window. The column is the data-side form of the if-then statement such papers describe in prose, which keeps the wash window a property of the experimental design the user supplies rather than hard-coding one paper’s uptake duration into the model.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 = no wash in progress. Structural rather than a coefficient multiplier: it switches a rate term on and off, so it has no associated e_<cov>_<param> coefficient. Enters as - WASH_ACTIVE * kwash * <medium state>, where kwash is a fixed rate constant large enough to empty the state over the wash window.
• Source aliases:
  • KWash “activated for 1 minute at the end of the 20-minute uptake phase using an if-then statement” – Guo 2016 Materials and Methods, the prose form of the same switch.
• Example models: Guo_2016_taurocholate_schh.R (1 over the single 1-minute wash between the 20-minute uptake and the 15-minute efflux phase; kwash fixed at 1e4 per minute, which the paper reports “was sufficient to eliminate the d8-TCA from the buffer compartment based on simulations”).
• Notes: General scope because a wash step between an uptake phase and an efflux phase is standard across cell-monolayer transport assays (sandwich-cultured hepatocytes, Caco-2, transfected-cell uptake) and the indicator carries the same meaning in all of them. Record the wash duration and the paired kwash value per model, because the pair – not the flag alone – determines how completely the medium is emptied. Do not use this column for a model that only covers an uptake phase and never washes; introducing an all-zero column is noise. Distinct from a washout period in a crossover clinical trial, which is a study-design interval rather than a physical rinse and is handled through the event table.

PROTEIN_MG (canonical for the cellular protein content of an in-vitro well or incubation)

• Description: Total cellular protein in the incubated well, dish or tube of an in-vitro experiment, in milligrams. In-vitro transport and metabolism papers report system constants per unit protein – cellular volume in uL/mg protein, clearances in uL/min/mg protein, enzyme activities in pmol/min/mg protein – while the medium volume and the applied dose are absolute per well. PROTEIN_MG is the conversion factor between the two bases, so a model whose ODE states carry absolute amounts needs it to scale the per-protein constants up to the well. It is an experimental input the user supplies per preparation, not a fitted parameter: protein content varies between hepatocyte lots and platings, and papers commonly measure it per preparation (bicinchoninic acid assay) without tabulating it.
• Units: mg
• Type: continuous
• Scope: general
• Reference category: n/a – a multiplicative scaling factor, not a contrast. PROTEIN_MG = 1 is the natural normalisation when a paper’s figures are plotted per mg protein, because the ODE states in absolute amount units are then numerically equal to the published per-mg-protein axis.
• Source aliases:
  • protein content of each preparation – Guo 2016 Materials and Methods, where VCells “was calculated and fixed using the protein content of each preparation and a value of 7.4 uL/mg protein”.
• Example models: Guo_2016_taurocholate_schh.R (scales the 7.4 uL/mg protein cellular volume and the three Table 2 clearances to the well; the value is not reported by the paper, and the vignette recovers 0.4 mg/well by inverting the paper’s own Figure 2 fold-change claims and shows that the Table 4 fold changes are insensitive to the choice across 0.1-2 mg/well).
• Notes: General scope because cellular protein content is the standard normalising basis for in-vitro transport, hepatocyte and microsomal work and means the same thing in every such experiment. Distinct from ALB and TP, which are plasma protein concentrations in a subject rather than the protein mass of an in-vitro preparation, and from HUMAN_SERUM_PCT, which is the protein supplementation of a growth medium. Where a paper reports protein per million cells together with a seeding density, derive the per-well value and record the derivation in covariateData[[PROTEIN_MG]]$notes; where it reports neither, say so explicitly rather than substituting a class-typical protein-per-cell factor, and show the model’s sensitivity to the choice.

CONC_DOR_MGL (canonical for static in-vitro doripenem concentration driving a receptor-binding or antibacterial PD model)

• Description: Static (time-invariant) doripenem concentration applied to the medium of an in-vitro bacterial experiment, supplied as an exogenous covariate that drives target-receptor binding or bacterial kill. Distinct from a state-derived plasma concentration (Cc) and from the CP_<DRUG> plasma-PD-driver family: this is an applied experimental concentration in the in-vitro matrix.
• Units: mg/L
• Type: continuous
• Scope: specific
• Reference category: n/a – scales the rate of net influx and PBP access in the whole-cell penicillin-binding-protein (PBP) binding QSP model (Eq 1: Rate_Influx/access = Rate_Influx/access,scaled x CONC_DOR_MGL). The lysed-cell arm of that assay sets CELLS_INTACT = 0, which zeroes the influx term, so the covariate has no effect there.
• Source aliases: none standardized (Lopez-Arguello 2023 writes C_drug in Eq 1 and CDRUG in the Fig. S8 estimation code).
• Example models: LopezArguello_2023_doripenem_qsp.R (static 2 mg/L, 2x the MIC of 1 mg/L; Lopez-Arguello 2023 Table 1).
• Notes: Specific scope because the value is bound to doripenem and to the in-vitro assay design. Member of the in-vitro applied-drug-concentration CONC_<DRUG>_MGL family (siblings CONC_RIF_MGL, CONC_INH_MGL, CONC_EMB_MGL, CONC_IPM_MGL, CONC_TOB_MGL).

CONC_MEM_MGL (canonical for static in-vitro meropenem concentration driving a receptor-binding or antibacterial PD model)

• Description: Static (time-invariant) meropenem concentration applied to the medium of an in-vitro bacterial experiment, supplied as an exogenous covariate that drives target-receptor binding or bacterial kill. Distinct from a state-derived plasma concentration (Cc) and from the CP_<DRUG> plasma-PD-driver family: this is an applied experimental concentration in the in-vitro matrix.
• Units: mg/L
• Type: continuous
• Scope: specific
• Reference category: n/a – scales the rate of net influx and PBP access in the whole-cell penicillin-binding-protein (PBP) binding QSP model (Eq 1: Rate_Influx/access = Rate_Influx/access,scaled x CONC_MEM_MGL). The lysed-cell arm of that assay sets CELLS_INTACT = 0, which zeroes the influx term, so the covariate has no effect there.
• Source aliases: none standardized (Lopez-Arguello 2023 writes C_drug in Eq 1 and CDRUG in the Fig. S8 estimation code).
• Example models: LopezArguello_2023_meropenem_qsp.R (static 1 mg/L, 2x the MIC of 0.5 mg/L; Lopez-Arguello 2023 Table 1).
• Notes: Specific scope because the value is bound to meropenem and to the in-vitro assay design. Member of the in-vitro applied-drug-concentration CONC_<DRUG>_MGL family (siblings CONC_RIF_MGL, CONC_INH_MGL, CONC_EMB_MGL, CONC_IPM_MGL, CONC_TOB_MGL).

CONC_ETP_MGL (canonical for static in-vitro ertapenem concentration driving a receptor-binding or antibacterial PD model)

• Description: Static (time-invariant) ertapenem concentration applied to the medium of an in-vitro bacterial experiment, supplied as an exogenous covariate that drives target-receptor binding or bacterial kill. Distinct from a state-derived plasma concentration (Cc) and from the CP_<DRUG> plasma-PD-driver family: this is an applied experimental concentration in the in-vitro matrix.
• Units: mg/L
• Type: continuous
• Scope: specific
• Reference category: n/a – scales the rate of net influx and PBP access in the whole-cell penicillin-binding-protein (PBP) binding QSP model (Eq 1: Rate_Influx/access = Rate_Influx/access,scaled x CONC_ETP_MGL). The lysed-cell arm of that assay sets CELLS_INTACT = 0, which zeroes the influx term, so the covariate has no effect there.
• Source aliases: none standardized (Lopez-Arguello 2023 writes C_drug in Eq 1 and CDRUG in the Fig. S8 estimation code).
• Example models: LopezArguello_2023_ertapenem_qsp.R (static 8 mg/L, 2x the MIC of 4 mg/L; Lopez-Arguello 2023 Table 1).
• Notes: Specific scope because the value is bound to ertapenem and to the in-vitro assay design. Member of the in-vitro applied-drug-concentration CONC_<DRUG>_MGL family (siblings CONC_RIF_MGL, CONC_INH_MGL, CONC_EMB_MGL, CONC_IPM_MGL, CONC_TOB_MGL).

CONC_CAZ_MGL (canonical for static in-vitro ceftazidime concentration driving a receptor-binding or antibacterial PD model)

• Description: Static (time-invariant) ceftazidime concentration applied to the medium of an in-vitro bacterial experiment, supplied as an exogenous covariate that drives target-receptor binding or bacterial kill. Distinct from a state-derived plasma concentration (Cc) and from the CP_<DRUG> plasma-PD-driver family: this is an applied experimental concentration in the in-vitro matrix.
• Units: mg/L
• Type: continuous
• Scope: specific
• Reference category: n/a – scales the rate of net influx and PBP access in the whole-cell penicillin-binding-protein (PBP) binding QSP model (Eq 1: Rate_Influx/access = Rate_Influx/access,scaled x CONC_CAZ_MGL). The lysed-cell arm of that assay sets CELLS_INTACT = 0, which zeroes the influx term, so the covariate has no effect there.
• Source aliases: none standardized (Lopez-Arguello 2023 writes C_drug in Eq 1 and CDRUG in the Fig. S8 estimation code).
• Example models: LopezArguello_2023_ceftazidime_qsp.R (static 2 mg/L, 2x the MIC of 1 mg/L; Lopez-Arguello 2023 Table 1), Kroemer_2024_ceftazidime_avibactam_fosfomycin_tkc.R (static time-kill concentrations 0.002-128 mg/L against a clinical MDR Escherichia coli; Kroemer 2024 Fig. 1).
• Notes: Specific scope because the value is bound to ceftazidime and to the in-vitro assay design. Member of the in-vitro applied-drug-concentration CONC_<DRUG>_MGL family (siblings CONC_RIF_MGL, CONC_INH_MGL, CONC_EMB_MGL, CONC_IPM_MGL, CONC_TOB_MGL).

CONC_FEP_MGL (canonical for static in-vitro cefepime concentration driving a receptor-binding or antibacterial PD model)

• Description: Static (time-invariant) cefepime concentration applied to the medium of an in-vitro bacterial experiment, supplied as an exogenous covariate that drives target-receptor binding or bacterial kill. Distinct from a state-derived plasma concentration (Cc) and from the CP_<DRUG> plasma-PD-driver family: this is an applied experimental concentration in the in-vitro matrix.
• Units: mg/L
• Type: continuous
• Scope: specific
• Reference category: n/a – scales the rate of net influx and PBP access in the whole-cell penicillin-binding-protein (PBP) binding QSP model (Eq 1: Rate_Influx/access = Rate_Influx/access,scaled x CONC_FEP_MGL). The lysed-cell arm of that assay sets CELLS_INTACT = 0, which zeroes the influx term, so the covariate has no effect there.
• Source aliases: none standardized (Lopez-Arguello 2023 writes C_drug in Eq 1 and CDRUG in the Fig. S8 estimation code).
• Example models: LopezArguello_2023_cefepime_qsp.R (static 2 mg/L, 2x the MIC of 1 mg/L; Lopez-Arguello 2023 Table 1).
• Notes: Specific scope because the value is bound to cefepime and to the in-vitro assay design. Member of the in-vitro applied-drug-concentration CONC_<DRUG>_MGL family (siblings CONC_RIF_MGL, CONC_INH_MGL, CONC_EMB_MGL, CONC_IPM_MGL, CONC_TOB_MGL).

CONC_FOF_MGL (canonical for static in-vitro fosfomycin concentration driving a receptor-binding or antibacterial PD model)

• Description: Static (time-invariant) fosfomycin concentration applied to the medium of an in-vitro bacterial experiment, supplied as an exogenous covariate that drives target-receptor binding or bacterial kill. Distinct from a state-derived plasma concentration (Cc) and from the CP_<DRUG> plasma-PD-driver family: this is an applied experimental concentration in the in-vitro matrix.
• Units: mg/L
• Type: continuous
• Scope: specific
• Reference category: n/a – 0 mg/L is the drug-free growth control.
• Source aliases: FOF (Kroemer 2024).
• Example models: Kroemer_2024_ceftazidime_avibactam_fosfomycin_tkc.R (static time-kill concentrations 2-16 mg/L against a clinical MDR Escherichia coli; Kroemer 2024 Fig. 1).
• Notes: Specific scope because the value is bound to fosfomycin and to the in-vitro assay design. Member of the in-vitro applied-drug-concentration CONC_<DRUG>_MGL family (siblings CONC_CAZ_MGL, CONC_AVI_MGL, CONC_RIF_MGL, CONC_INH_MGL, CONC_EMB_MGL, CONC_IPM_MGL, CONC_TOB_MGL). Fosfomycin susceptibility testing and time-kill media are supplemented with 25 mg/L glucose-6-phosphate per EUCAST recommendations; the covariate carries the fosfomycin concentration only, not the glucose-6-phosphate. The dynamic sibling model Kroemer_2024_ceftazidime_avibactam_fosfomycin_hfim.R carries fosfomycin as a dosable ODE state (conc_fof) instead, because the hollow fiber concentrations are time-varying.

CONC_FOX_MGL (canonical for static in-vitro cefoxitin concentration driving a receptor-binding or antibacterial PD model)

• Description: Static (time-invariant) cefoxitin concentration applied to the medium of an in-vitro bacterial experiment, supplied as an exogenous covariate that drives target-receptor binding or bacterial kill. Distinct from a state-derived plasma concentration (Cc) and from the CP_<DRUG> plasma-PD-driver family: this is an applied experimental concentration in the in-vitro matrix.
• Units: mg/L
• Type: continuous
• Scope: specific
• Reference category: n/a – scales the rate of net influx and PBP access in the whole-cell penicillin-binding-protein (PBP) binding QSP model (Eq 1: Rate_Influx/access = Rate_Influx/access,scaled x CONC_FOX_MGL). The lysed-cell arm of that assay sets CELLS_INTACT = 0, which zeroes the influx term, so the covariate has no effect there.
• Source aliases: none standardized (Lopez-Arguello 2023 writes C_drug in Eq 1 and CDRUG in the Fig. S8 estimation code).
• Example models: LopezArguello_2023_cefoxitin_qsp.R (static 2,048 mg/L, 2x the MIC of 1,024 mg/L; Lopez-Arguello 2023 Table 1).
• Notes: Specific scope because the value is bound to cefoxitin and to the in-vitro assay design. Member of the in-vitro applied-drug-concentration CONC_<DRUG>_MGL family (siblings CONC_RIF_MGL, CONC_INH_MGL, CONC_EMB_MGL, CONC_IPM_MGL, CONC_TOB_MGL).

CONC_ATM_MGL (canonical for static in-vitro aztreonam concentration driving a receptor-binding or antibacterial PD model)

• Description: Static (time-invariant) aztreonam concentration applied to the medium of an in-vitro bacterial experiment, supplied as an exogenous covariate that drives target-receptor binding or bacterial kill. Distinct from a state-derived plasma concentration (Cc) and from the CP_<DRUG> plasma-PD-driver family: this is an applied experimental concentration in the in-vitro matrix.
• Units: mg/L
• Type: continuous
• Scope: specific
• Reference category: n/a – scales the rate of net influx and PBP access in the whole-cell penicillin-binding-protein (PBP) binding QSP model (Eq 1: Rate_Influx/access = Rate_Influx/access,scaled x CONC_ATM_MGL). The lysed-cell arm of that assay sets CELLS_INTACT = 0, which zeroes the influx term, so the covariate has no effect there.
• Source aliases: none standardized (Lopez-Arguello 2023 writes C_drug in Eq 1 and CDRUG in the Fig. S8 estimation code).
• Example models: LopezArguello_2023_aztreonam_qsp.R (static 8 mg/L, 2x the MIC of 4 mg/L; Lopez-Arguello 2023 Table 1).
• Notes: Specific scope because the value is bound to aztreonam and to the in-vitro assay design. Member of the in-vitro applied-drug-concentration CONC_<DRUG>_MGL family (siblings CONC_RIF_MGL, CONC_INH_MGL, CONC_EMB_MGL, CONC_IPM_MGL, CONC_TOB_MGL).

CONC_PIP_MGL (canonical for static in-vitro piperacillin concentration driving a receptor-binding or antibacterial PD model)

• Description: Static (time-invariant) piperacillin concentration applied to the medium of an in-vitro bacterial experiment, supplied as an exogenous covariate that drives target-receptor binding or bacterial kill. Distinct from a state-derived plasma concentration (Cc) and from the CP_<DRUG> plasma-PD-driver family: this is an applied experimental concentration in the in-vitro matrix.
• Units: mg/L
• Type: continuous
• Scope: specific
• Reference category: n/a – scales the rate of net influx and PBP access in the whole-cell penicillin-binding-protein (PBP) binding QSP model (Eq 1: Rate_Influx/access = Rate_Influx/access,scaled x CONC_PIP_MGL). The lysed-cell arm of that assay sets CELLS_INTACT = 0, which zeroes the influx term, so the covariate has no effect there.
• Source aliases: none standardized (Lopez-Arguello 2023 writes C_drug in Eq 1 and CDRUG in the Fig. S8 estimation code).
• Example models: LopezArguello_2023_piperacillin_qsp.R (static 8 mg/L, 2x the MIC of 4 mg/L; Lopez-Arguello 2023 Table 1).
• Notes: Specific scope because the value is bound to piperacillin and to the in-vitro assay design. Member of the in-vitro applied-drug-concentration CONC_<DRUG>_MGL family (siblings CONC_RIF_MGL, CONC_INH_MGL, CONC_EMB_MGL, CONC_IPM_MGL, CONC_TOB_MGL).

CONC_CAR_MGL (canonical for static in-vitro carbenicillin concentration driving a receptor-binding or antibacterial PD model)

• Description: Static (time-invariant) carbenicillin concentration applied to the medium of an in-vitro bacterial experiment, supplied as an exogenous covariate that drives target-receptor binding or bacterial kill. Distinct from a state-derived plasma concentration (Cc) and from the CP_<DRUG> plasma-PD-driver family: this is an applied experimental concentration in the in-vitro matrix.
• Units: mg/L
• Type: continuous
• Scope: specific
• Reference category: n/a – scales the rate of net influx and PBP access in the whole-cell penicillin-binding-protein (PBP) binding QSP model (Eq 1: Rate_Influx/access = Rate_Influx/access,scaled x CONC_CAR_MGL). The lysed-cell arm of that assay sets CELLS_INTACT = 0, which zeroes the influx term, so the covariate has no effect there.
• Source aliases: none standardized (Lopez-Arguello 2023 writes C_drug in Eq 1 and CDRUG in the Fig. S8 estimation code).
• Example models: LopezArguello_2023_carbenicillin_qsp.R (static 96 mg/L, 2x the MIC of 48 mg/L; Lopez-Arguello 2023 Table 1).
• Notes: Specific scope because the value is bound to carbenicillin and to the in-vitro assay design. Member of the in-vitro applied-drug-concentration CONC_<DRUG>_MGL family (siblings CONC_RIF_MGL, CONC_INH_MGL, CONC_EMB_MGL, CONC_IPM_MGL, CONC_TOB_MGL).

CONC_TIC_MGL (canonical for static in-vitro ticarcillin concentration driving a receptor-binding or antibacterial PD model)

• Description: Static (time-invariant) ticarcillin concentration applied to the medium of an in-vitro bacterial experiment, supplied as an exogenous covariate that drives target-receptor binding or bacterial kill. Distinct from a state-derived plasma concentration (Cc) and from the CP_<DRUG> plasma-PD-driver family: this is an applied experimental concentration in the in-vitro matrix.
• Units: mg/L
• Type: continuous
• Scope: specific
• Reference category: n/a – scales the rate of net influx and PBP access in the whole-cell penicillin-binding-protein (PBP) binding QSP model (Eq 1: Rate_Influx/access = Rate_Influx/access,scaled x CONC_TIC_MGL). The lysed-cell arm of that assay sets CELLS_INTACT = 0, which zeroes the influx term, so the covariate has no effect there.
• Source aliases: none standardized (Lopez-Arguello 2023 writes C_drug in Eq 1 and CDRUG in the Fig. S8 estimation code).
• Example models: LopezArguello_2023_ticarcillin_qsp.R (static 48 mg/L, 2x the MIC of 24 mg/L; Lopez-Arguello 2023 Table 1).
• Notes: Specific scope because the value is bound to ticarcillin and to the in-vitro assay design. Member of the in-vitro applied-drug-concentration CONC_<DRUG>_MGL family (siblings CONC_RIF_MGL, CONC_INH_MGL, CONC_EMB_MGL, CONC_IPM_MGL, CONC_TOB_MGL).

CONC_AVI_MGL (canonical for static in-vitro avibactam concentration driving a receptor-binding or antibacterial PD model)

• Description: Static (time-invariant) avibactam concentration applied to the medium of an in-vitro bacterial experiment, supplied as an exogenous covariate that drives target-receptor binding or bacterial kill. Distinct from a state-derived plasma concentration (Cc) and from the CP_<DRUG> plasma-PD-driver family: this is an applied experimental concentration in the in-vitro matrix.
• Units: mg/L
• Type: continuous
• Scope: specific
• Reference category: n/a – scales the rate of net influx and PBP access in the whole-cell penicillin-binding-protein (PBP) binding QSP model (Eq 1: Rate_Influx/access = Rate_Influx/access,scaled x CONC_AVI_MGL). The lysed-cell arm of that assay sets CELLS_INTACT = 0, which zeroes the influx term, so the covariate has no effect there.
• Source aliases: none standardized (Lopez-Arguello 2023 writes C_drug in Eq 1 and CDRUG in the Fig. S8 estimation code).
• Example models: LopezArguello_2023_avibactam_qsp.R (static 4 mg/L, a fixed concentration within the clinically relevant range; MIC not determined for beta-lactamase inhibitors; Lopez-Arguello 2023 Table 1), Kroemer_2024_ceftazidime_avibactam_fosfomycin_tkc.R (static time-kill concentrations up to ~64 mg/L against a clinical MDR Escherichia coli; Kroemer 2024 Fig. 1).
• Notes: Specific scope because the value is bound to avibactam and to the in-vitro assay design. Member of the in-vitro applied-drug-concentration CONC_<DRUG>_MGL family (siblings CONC_RIF_MGL, CONC_INH_MGL, CONC_EMB_MGL, CONC_IPM_MGL, CONC_TOB_MGL).

CONC_REL_MGL (canonical for static in-vitro relebactam concentration driving a receptor-binding or antibacterial PD model)

• Description: Static (time-invariant) relebactam concentration applied to the medium of an in-vitro bacterial experiment, supplied as an exogenous covariate that drives target-receptor binding or bacterial kill. Distinct from a state-derived plasma concentration (Cc) and from the CP_<DRUG> plasma-PD-driver family: this is an applied experimental concentration in the in-vitro matrix.
• Units: mg/L
• Type: continuous
• Scope: specific
• Reference category: n/a – scales the rate of net influx and PBP access in the whole-cell penicillin-binding-protein (PBP) binding QSP model (Eq 1: Rate_Influx/access = Rate_Influx/access,scaled x CONC_REL_MGL). The lysed-cell arm of that assay sets CELLS_INTACT = 0, which zeroes the influx term, so the covariate has no effect there.
• Source aliases: none standardized (Lopez-Arguello 2023 writes C_drug in Eq 1 and CDRUG in the Fig. S8 estimation code).
• Example models: LopezArguello_2023_relebactam_qsp.R (static 4 mg/L, a fixed concentration within the clinically relevant range; MIC not determined for beta-lactamase inhibitors; Lopez-Arguello 2023 Table 1).
• Notes: Specific scope because the value is bound to relebactam and to the in-vitro assay design. Member of the in-vitro applied-drug-concentration CONC_<DRUG>_MGL family (siblings CONC_RIF_MGL, CONC_INH_MGL, CONC_EMB_MGL, CONC_IPM_MGL, CONC_TOB_MGL).

CONC_SUL_MGL (canonical for static in-vitro sulbactam concentration driving a receptor-binding or antibacterial PD model)

• Description: Static (time-invariant) sulbactam concentration applied to the medium of an in-vitro bacterial experiment, supplied as an exogenous covariate that drives target-receptor binding or bacterial kill. Distinct from a state-derived plasma concentration (Cc) and from the CP_<DRUG> plasma-PD-driver family: this is an applied experimental concentration in the in-vitro matrix.
• Units: mg/L
• Type: continuous
• Scope: specific
• Reference category: n/a – scales the rate of net influx and PBP access in the whole-cell penicillin-binding-protein (PBP) binding QSP model (Eq 1: Rate_Influx/access = Rate_Influx/access,scaled x CONC_SUL_MGL). The lysed-cell arm of that assay sets CELLS_INTACT = 0, which zeroes the influx term, so the covariate has no effect there.
• Source aliases: none standardized (Lopez-Arguello 2023 writes C_drug in Eq 1 and CDRUG in the Fig. S8 estimation code).
• Example models: LopezArguello_2023_sulbactam_qsp.R (static 4 mg/L, a fixed concentration within the clinically relevant range; MIC not determined for beta-lactamase inhibitors; Lopez-Arguello 2023 Table 1).
• Notes: Specific scope because the value is bound to sulbactam and to the in-vitro assay design. Member of the in-vitro applied-drug-concentration CONC_<DRUG>_MGL family (siblings CONC_RIF_MGL, CONC_INH_MGL, CONC_EMB_MGL, CONC_IPM_MGL, CONC_TOB_MGL).

CONC_TZB_MGL (canonical for static in-vitro tazobactam concentration driving a receptor-binding or antibacterial PD model)

• Description: Static (time-invariant) tazobactam concentration applied to the medium of an in-vitro bacterial experiment, supplied as an exogenous covariate that drives target-receptor binding or bacterial kill. Distinct from a state-derived plasma concentration (Cc) and from the CP_<DRUG> plasma-PD-driver family: this is an applied experimental concentration in the in-vitro matrix.
• Units: mg/L
• Type: continuous
• Scope: specific
• Reference category: n/a – scales the rate of net influx and PBP access in the whole-cell penicillin-binding-protein (PBP) binding QSP model (Eq 1: Rate_Influx/access = Rate_Influx/access,scaled x CONC_TZB_MGL). The lysed-cell arm of that assay sets CELLS_INTACT = 0, which zeroes the influx term, so the covariate has no effect there.
• Source aliases: none standardized (Lopez-Arguello 2023 writes C_drug in Eq 1 and CDRUG in the Fig. S8 estimation code).
• Example models: LopezArguello_2023_tazobactam_qsp.R (static 4 mg/L, a fixed concentration within the clinically relevant range; MIC not determined for beta-lactamase inhibitors; Lopez-Arguello 2023 Table 1).
• Notes: Specific scope because the value is bound to tazobactam and to the in-vitro assay design. Member of the in-vitro applied-drug-concentration CONC_<DRUG>_MGL family (siblings CONC_RIF_MGL, CONC_INH_MGL, CONC_EMB_MGL, CONC_IPM_MGL, CONC_TOB_MGL).

CONC_RIF_MGL (canonical for static in-vitro rifampicin concentration driving an antibacterial PD model)

• Description: Static (time-invariant) rifampicin concentration in the growth medium of an in-vitro antibacterial time-kill experiment, supplied as an exogenous covariate that drives the bacterial-kill PD effect. Distinct from a state-derived plasma concentration (Cc) and from the CP_<DRUG> plasma-PD-driver family: this is an applied experimental concentration in the in-vitro matrix.
• Units: mg/L
• Type: continuous
• Scope: specific
• Reference category: n/a – enters the sigmoidal kill function; set to 0 for regimens without rifampicin.
• Source aliases:
  • CRIF – used in Clewe_2018_TB_MTP_GPDI_invitro.R (Clewe 2018 Materials and methods; static rifampicin concentration).
• Example models: Clewe_2018_TB_MTP_GPDI_invitro.R (static rifampicin concentration driving the multistate-TB kill effects; tested concentrations 0.002, 0.008, 0.03, 0.125, 0.5, 8 mg/L per Figure 1).
• Notes: Specific scope because the value is bound to rifampicin and the in-vitro experimental design. Member of the in-vitro applied-drug-concentration CONC_<DRUG>_MGL family (siblings CONC_INH_MGL, CONC_EMB_MGL, CONC_IPM_MGL, CONC_TOB_MGL) – use this family for exogenous static or time-varying drug concentrations in in-vitro time-kill / hollow-fiber PD models, as distinct from the CP_<DRUG> plasma-concentration PD-driver family.

CONC_INH_MGL (canonical for static in-vitro isoniazid concentration driving an antibacterial PD model)

• Description: Static (time-invariant) isoniazid concentration in the growth medium of an in-vitro antibacterial time-kill experiment, supplied as an exogenous covariate that drives the bacterial-kill PD effect and the adaptive-resistance transition. Applied experimental concentration in the in-vitro matrix; distinct from Cc and the CP_<DRUG> plasma-PD-driver family.
• Units: mg/L
• Type: continuous
• Scope: specific
• Reference category: n/a – drives the kill effects on the fast- and slow-growing sub-states and the adaptive-resistance AR_on/AR_off transition rate; set to 0 for regimens without isoniazid.
• Source aliases:
  • CINH – used in Clewe_2018_TB_MTP_GPDI_invitro.R (Clewe 2018 Materials and methods; static isoniazid concentration).
• Example models: Clewe_2018_TB_MTP_GPDI_invitro.R (static isoniazid concentration driving the kill effects on the F and S sub-states and the adaptive-resistance transition kon * CONC_INH_MGL; tested concentrations 0.01, 0.039, 0.156, 0.625, 2.5, 10, 40 mg/L per Figure 1).
• Notes: Specific scope because the value is bound to isoniazid and the in-vitro experimental design. Member of the in-vitro applied-drug-concentration CONC_<DRUG>_MGL family (siblings CONC_RIF_MGL, CONC_EMB_MGL, CONC_IPM_MGL, CONC_TOB_MGL).

CONC_EMB_MGL (canonical for static in-vitro ethambutol concentration driving an antibacterial PD model)

• Description: Static (time-invariant) ethambutol concentration in the growth medium of an in-vitro antibacterial time-kill experiment, supplied as an exogenous covariate that drives the bacterial-kill PD effect. Applied experimental concentration in the in-vitro matrix; distinct from Cc and the CP_<DRUG> plasma-PD-driver family.
• Units: mg/L
• Type: continuous
• Scope: specific
• Reference category: n/a – enters the kill function; set to 0 for regimens without ethambutol.
• Source aliases:
  • CEMB – used in Clewe_2018_TB_MTP_GPDI_invitro.R (Clewe 2018 Materials and methods; static ethambutol concentration).
• Example models: Clewe_2018_TB_MTP_GPDI_invitro.R (static ethambutol concentration driving the multistate-TB kill effects; tested concentrations 0.0078, 0.031, 0.125, 0.5, 2, 8, 32 mg/L per Figure 1).
• Notes: Specific scope because the value is bound to ethambutol and the in-vitro experimental design. Member of the in-vitro applied-drug-concentration CONC_<DRUG>_MGL family (siblings CONC_RIF_MGL, CONC_INH_MGL, CONC_IPM_MGL, CONC_TOB_MGL).

CONC_VAN_MGL (canonical for unbound vancomycin concentration driving an antibacterial PD model)

• Description: Unbound vancomycin concentration driving a bacterial-kill PD effect and, where the model carries one, an adaptive-resistance transition. Supplied as an exogenous covariate rather than derived from a PK state. Distinct from a state-derived plasma concentration (Cc) and from the CP_<DRUG> plasma-PD-driver family: this is an applied or externally-predicted concentration presented to an in-vitro-derived PD model.
• Units: mg/L
• Type: continuous
• Scope: specific
• Reference category: n/a – enters the sigmoidal kill function and the adaptive-resistance activation rate kon * CONC_VAN_MGL; set to 0 for the untreated growth control.
• Source aliases:
  • Conc – used in Olivo_2025_vancomycin_invitro.R (Olivo 2025 Supplementary Equations S4-S5, adaptive-resistance activation).
  • C – used in Olivo_2025_vancomycin_invitro.R (Olivo 2025 Equation 3, sigmoidal Emax kill term; Figure 1 legend defines it as “the unbound drug concentration in interstitial compartment”).
• Example models: Olivo_2025_vancomycin_invitro.R (drives the sigmoidal Emax kill of the active MRSA state and the kon * CONC_VAN_MGL adaptive-resistance switch; static bath concentrations 0, 0.5, 1, 2, 4, 8, 12 and 16 mg/L reproduce the time-kill arms of Figure 3, and a PBPK-predicted unbound interstitial profile reproduces the coupled tissue simulations of Figures 4 and 5).
• Notes: Specific scope because the value is bound to vancomycin and to a PD model parameterised on unbound concentration. Member of the in-vitro applied-drug-concentration CONC_<DRUG>_MGL family (siblings CONC_RIF_MGL, CONC_INH_MGL, CONC_EMB_MGL, CONC_IPM_MGL, CONC_TOB_MGL). This member is exercised both statically and time-varying: the family is explicitly for “exogenous static or time-varying drug concentrations”, and Olivo 2025 uses the same PD block in both modes – static for the time-kill fit, time-varying when driven by tissue concentration-time profiles from a PK-Sim PBPK model that is not itself reproduced in the library. Because the driver is unbound, a caller supplying a total plasma concentration must apply the fraction unbound before passing it in.

CONC_BAI_UM (canonical for static in-vitro baicalein concentration driving an inflammatory-mediator PD model)

• Description: Static (time-invariant) baicalein concentration in the cell-culture medium of an in-vitro LPS-stimulated-macrophage experiment, supplied as an exogenous covariate that drives the anti-inflammatory PD effect. Applied experimental concentration in the in-vitro matrix; distinct from a state-derived plasma concentration (Cc) and from the CP_<drug> plasma-PD-driver family.
• Units: uM
• Type: continuous
• Scope: specific
• Reference category: n/a – enters the log-linear inhibition term f(Bai) = alpha * log(CONC_BAI_UM + 1) (the +1 shift encodes the control with CONC_BAI_UM = 0 -> f = 0 without an undefined ln(0)); set to 0 for the control well. Reference values observed: Xiang 2018 tested 0 (control), 10, 20, and 40 uM (Materials and Methods; Figures 3-4).
• Source aliases: none – the source uses C_Bai (Xiang 2018 Eq 2) in prose; the model column is the canonical CONC_BAI_UM.
• Example models: Xiang_2018_baicalein.R (time-invariant baicalein concentration driving the log-linear inhibition of LPS-stimulated TNF-alpha production in RAW264.7 macrophages, propagating downstream to IL-6, iNOS, and NO).
• Notes: Specific scope because the value is bound to baicalein and the in-vitro experimental design. Member of the in-vitro applied-drug-concentration CONC_<drug>_<units> family; here the unit is uM (sibling concentration covariates such as CONC_RIF_MGL are reported in mg/L, so the <units> suffix is load-bearing). Distinct from the STIM_<drug>_<units> antimalarial-well family and the CP_<drug> plasma-PD-driver family.

CONC_OXA_UM (canonical for static in-vitro oxaliplatin concentration driving a tumour-organoid cytotoxicity PD model)

• Description: Static (time-invariant) oxaliplatin concentration applied to the culture medium of a patient-derived tumour-organoid (PDTO) drug-sensitivity assay, supplied as an exogenous covariate that drives the sigmoidal Emax killing term. Applied experimental concentration in the in-vitro matrix; distinct from a state-derived plasma concentration (Cc) and from the CP_<drug> plasma-PD-driver family.
• Units: uM
• Type: continuous
• Scope: specific
• Reference category: n/a – enters the Hill killing term Emax * C^hill / (EC50^hill + C^hill); set to 0 for the vehicle-control organoid. Zhu 2023 does not tabulate the tested concentration grid; Figure 2A spans roughly 1-1000 umol/L, bracketing the estimated EC50 range of 246-622 umol/L (Table 2).
• Source aliases: none – the source writes C in Equation (2); the model column is the canonical CONC_OXA_UM.
• Example models: Zhu_2023_oxaliplatin_organoid.R (96 h static oxaliplatin exposure of colorectal-cancer PDTOs; cell viability read as the treated-to-vehicle-control organoid volume ratio).
• Notes: Specific scope because the value is bound to oxaliplatin and to the in-vitro organoid assay design. Member of the in-vitro applied-drug-concentration CONC_<drug>_<units> family; the unit suffix is load-bearing because sibling entries such as CONC_RIF_MGL are reported in mg/L while the organoid assays are reported in umol/L.

CONC_SN38_UM (canonical for static in-vitro SN-38 concentration driving a tumour-organoid cytotoxicity PD model)

• Description: Static (time-invariant) SN-38 concentration applied to the culture medium of a patient-derived tumour-organoid (PDTO) drug-sensitivity assay, supplied as an exogenous covariate that drives the sigmoidal Emax killing term. SN-38 is the active metabolite of irinotecan. Applied experimental concentration in the in-vitro matrix; distinct from a state-derived plasma concentration (Cc_sn38) and from the CP_<drug> plasma-PD-driver family.
• Units: uM
• Type: continuous
• Scope: specific
• Reference category: n/a – enters the Hill killing term Emax * C^hill / (EC50^hill + C^hill); set to 0 for the vehicle-control organoid. Zhu 2023 does not tabulate the tested concentration grid; Figure 2B spans roughly 0.01-100 umol/L, bracketing the estimated EC50 range of 4.17-15.5 umol/L (Table 2).
• Source aliases: none – the source writes C in Equation (2); the model column is the canonical CONC_SN38_UM.
• Example models: Zhu_2023_sn38_organoid.R (96 h static SN-38 exposure of colorectal-cancer PDTOs; cell viability read as the treated-to-vehicle-control organoid volume ratio).
• Notes: Specific scope because the value is bound to SN-38 and to the in-vitro organoid assay design. Member of the in-vitro applied-drug-concentration CONC_<drug>_<units> family; SN38 matches the registered metabolite suffix sn38 used for the paired compartment and parameter names.

CONC_FOY251_NM (canonical for static in-vitro FOY-251 concentration driving a TMPRSS2 covalent-inhibition PD model)

• Description: Applied (time-invariant during incubation) FOY-251 concentration in the well of an in-vitro TMPRSS2-inhibition / SARS-CoV-2 viral-entry assay, supplied as an exogenous covariate that drives the reversible covalent inhibition of TMPRSS2. FOY-251 is the active metabolite of camostat mesylate. Applied experimental concentration in the in-vitro matrix; distinct from a state-derived plasma concentration (Cc) and from the CP_<drug> plasma-PD-driver family.
• Units: nM
• Type: continuous
• Scope: specific
• Reference category: n/a – enters the covalent-binding rate kcat * SP * C / (C + ki); set to 0 for the drug-free control well.
• Source aliases: none – Kosinsky 2022 writes C in Equation (3); the model column is the canonical CONC_FOY251_NM.
• Example models: Kosinsky_2022_camostat_invitro.R (recombinant TMPRSS2 activity after 1 h incubation and SARS-2-S pseudovirus entry after 2 h incubation; data of Hoffmann et al. 2020 digitised in Kosinsky 2022 Figure 2).
• Notes: Specific scope because the value is bound to FOY-251 and to the in-vitro TMPRSS2 assay design. Member of the in-vitro applied-drug-concentration CONC_<drug>_<units> family; the unit suffix is load-bearing because sibling entries such as CONC_RIF_MGL are reported in mg/L while this assay is reported in nM.

CONC_IPM_MGL (canonical for in-vitro imipenem concentration driving an antibacterial PD or receptor-binding model)

• Description: Unbound imipenem concentration applied to an in-vitro bacterial system, supplied externally as an exogenous covariate. Applied experimental concentration in the in-vitro matrix; distinct from Cc and the CP_<DRUG> plasma-PD-driver family. Used both time-varying (hollow-fiber infection model growth medium, driving the bacterial-kill PD effect) and static (60-min whole-cell penicillin-binding-protein binding assay, scaling the rate of net influx and PBP access).
• Units: mg/L
• Type: continuous
• Scope: specific
• Reference category: n/a – enters the sigmoidal (Hill) imipenem kill function; set to 0 for tobramycin-monotherapy or control arms.
• Source aliases: none standardized (the model column uses the canonical name directly; Landersdorfer 2018 labels it “imipenem concentration” in the Methods and Fig. 1 legend).
• Example models: Landersdorfer_2018_imipenem_tobramycin.R (externally-supplied time-varying unbound imipenem concentration driving the Hill kill function; the HFIM used continuous infusion targeting the 5th-percentile 7.6, median 13.4, and 95th-percentile 23.3 mg/L unbound concentrations from imipenem 4 g/day continuous infusion in critically ill patients), LopezArguello_2023_imipenem_qsp.R (static 2 mg/L = 2x the MIC of 1 mg/L, scaling the rate of net influx and PBP access in the whole-cell PBP-binding QSP model; Lopez-Arguello 2023 Table 1).
• Notes: Specific scope because the value is bound to imipenem and to an in-vitro assay design. Member of the in-vitro applied-drug-concentration CONC_<DRUG>_MGL family (siblings CONC_RIF_MGL, CONC_INH_MGL, CONC_EMB_MGL, CONC_TOB_MGL, and the fourteen CONC_<DRUG>_MGL entries ratified with the Lopez-Arguello 2023 extraction).

CONC_VORI_NGML (canonical for in-vivo voriconazole whole-blood concentration driving a CYP3A drug-drug-interaction term)

• Description: Voriconazole whole-blood concentration supplied externally as a (time-varying) covariate that drives a reversible CYP3A-inhibition term inside a victim-drug PBPK model. Distinct from the victim drug’s own Cc: this is the perpetrator’s concentration, provided as data rather than simulated, for models whose perpetrator sub-model is not reproducible from the source. Set to 0 for a victim-drug-alone arm, which collapses the inhibition ratio to 1 (no inhibition).
• Units: ng/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – enters the reversible-inhibition ratio DDIRE = 1 / (1 + CONC_VORI_NGML * fub_vori / KI); 0 gives DDIRE = 1.
• Source aliases:
  • Cb,vor – symbol used in Pei 2023 supplement Eq 1 (“Cb,vor = whole blood concentration of voriconazole”).
• Example models: Pei_2023_tacrolimus_pbpk.R (reversible CYP3A inhibition applied to the CYP3A-mediated fraction of hepatic tacrolimus clearance, with fub_vori = 0.42 and KI = 8.70 ng/mL from Pei 2023 Table S6; Pei 2023 supplement Eqs 1-2), Zhao_2024_tacrolimus.R (empirical popPK rather than PBPK: the measured voriconazole concentration enters as exp(-0.28 * C_VRC) on tacrolimus CL/F and exp(-0.20 * C_VRC) on V/F, with C_VRC in ug/mL, so the model divides the ng/mL column by 1000; the exponential rather than power form is forced by a cohort median of exactly 0 ug/mL).
• Notes: Member of the applied-drug-concentration CONC_<DRUG>_<UNITS> family (siblings CONC_RIF_MGL, CONC_INH_MGL, CONC_TOB_MGL,…), extended here from the in-vitro applied-concentration case to an in-vivo perpetrator concentration; the unit token is NGML rather than MGL because the founding source reports voriconazole in ng/mL and the inhibition constant KI is tabulated on the same scale. The column exists because Pei 2023 generated Cb,vor from a companion voriconazole PBPK model whose tissue-to-plasma partition coefficients, absorption rate and absorbed fraction are not tabulated anywhere in the paper or supplement (Table S6 gives only MW, pKa, LogP, fup, BPR, KI and CLint); exposing the perpetrator concentration as data encodes the published inhibition equation faithfully without substituting distribution parameters from another paper, which the PBPK / QSP sourcing rule forbids. Distinct from the binary [[CONMED_VORICONAZOLE]], which is the right column when a popPK model estimates an on/off coefficient rather than a concentration-driven inhibition term.

CONC_VEN_MGL (canonical for in-vivo venetoclax plasma concentration driving a BCL-2-inhibitor killing effect on leukemic cells)

• Description: Venetoclax plasma concentration supplied externally as a time-varying covariate that drives saturable Emax killing terms on chronic lymphocytic leukemia (CLL) cell populations. Venetoclax has no PK compartment in the consuming model: the concentration is provided as data rather than simulated, exactly as the companion ibrutinib exposure is provided through [[AUC_IBRU]]. Set to 0 for an ibrutinib-monotherapy arm, which collapses every venetoclax term to zero (and the peripheral-blood term to the untreated death rate constant), so the combination model reduces exactly to its monotherapy sibling.
• Units: ug/mL (equivalently mg/L)
• Type: continuous
• Scope: specific
• Reference category: n/a – enters saturable Emax forms of the shape CONC_VEN_MGL / (EC50 + CONC_VEN_MGL) rather than a centred-deviation or power form. Two half-maximal values are used simultaneously in the founding model, and their ratio is the point of the parameterisation: EC50,blood = 0.04 ug/mL versus EC50,tissue = 2.24 ug/mL, a 56-fold potency difference encoding the clinical observation that venetoclax clears circulating CLL cells far more readily than lymph-node disease.
• Source aliases:
  • Ct,venetoclax – symbol used in Ibrahim 2025 Table S2 footnote a.
• Example models: Ibrahim_2025_ibrutinib_venetoclax.R (additive death rate kd,bld * Emax,i * C/(2.24 + C) on the three lymphoid-tissue CLL subpopulations with Emax,1&2 = 0.63 and Emax,3 = 3.15, and multiplicative enhancement kd,bld * (1 + 3465 * C/(0.04 + C)) of the peripheral-blood CLL death rate; Ibrahim 2025 Table S2).
• Notes: Member of the applied-drug-concentration CONC_<DRUG>_<UNITS> family (siblings CONC_RIF_MGL, CONC_INH_MGL, CONC_TOB_MGL, CONC_VORI_NGML,…); like CONC_VORI_NGML this is an in-vivo rather than an in-vitro applied concentration. The unit token is MGL because the founding source tabulates both EC50 values in ug/mL, which is numerically identical to mg/L. The column exists because Ibrahim 2025 generated venetoclax concentration-time profiles from the two-compartment population PK model of Jones et al. (AAPS J. 2016;18(5):1192-1202), which is not open access, is not reproduced anywhere in the paper or its supplement, and is not part of nlmixr2lib; exposing the venetoclax concentration as data encodes the published killing equations faithfully without substituting distribution parameters from another source, which the QSP sourcing rule forbids. Downstream users must therefore supply the concentrations from that model or from observed data.

CONC_TFVDP_FMOLMC (canonical for mucosal-mononuclear-cell tenofovir-diphosphate concentration driving an ex vivo HIV-1 viral-dynamics PK/PD model)

• Description: Tenofovir-diphosphate (TFVdp) concentration in rectal mucosal mononuclear cells (MMCs), measured 30 minutes postdose and supplied externally as a per-explant exogenous covariate that drives the linear drug-effect term of an ex vivo HIV-1 viral-dynamics model. TFVdp has no PK compartment in the consuming PD model: the concentration is provided as data (from a companion PK model) rather than computed inside the model. Applied concentration in the cellular matrix; distinct from the modelled state and from the CP_<DRUG> plasma-PD-driver family.
• Units: fmol/million cells
• Type: continuous
• Reference category: n/a – enters the linear drug effect E = slope * CONC_TFVDP_FMOLMC * exp(-kdeg * t), where t is the ex vivo assay time; set to 0 for a baseline (no-treatment) explant, giving E = 0.
• Scope: specific
• Source aliases:
  • CP – MMC TFVdp concentration column in the Jayachandran 2021 Supplementary Model Code 2 NONMEM control stream.
• Example models: Jayachandran_2021_tenofovir_p24.R (linear inhibition of HIV-1 viral growth in an ex vivo rectal-tissue explant challenge; the MMC TFVdp concentration is generated by the companion multicompartment PK model Jayachandran_2021_tenofovir.R and swept from 0 to 11,000 fmol/million cells to find the ~9,000 fmol/million cells target for apparent viral suppression to 1%).
• Notes: Member of the applied-drug-concentration CONC_<DRUG>_<UNITS> family (siblings CONC_RIF_MGL, CONC_VORI_NGML, CONC_VEN_MGL,…); like CONC_VEN_MGL the consuming model has no PK compartment for the driving drug and takes the concentration as data from a companion PK model, exposing it as a covariate so the published PD equations are encoded faithfully. The unit token is FMOLMC (femtomoles per million cells) because intracellular nucleotide anabolites are assayed per cell count rather than per volume; this is the first CONC_ member reported on that basis, and a future model measuring an intracellular anabolite in fmol/million cells should reuse the token. Distinct from a plasma or in-vitro-medium concentration: the value is an intracellular, cell-count-normalised anabolite concentration in a specific mucosal cell population.

CONMED_RTV_AUC (canonical for ritonavir AUC over the 0-24 h dosing interval)

• Description: Per-subject (time-fixed within an evaluated regimen) ritonavir AUC over the 0-24 h once-daily dosing interval, used as a co-medication exposure covariate driving boosted-protease-inhibitor (atazanavir, lopinavir, darunavir, etc.) clearance in popPK models that account for ritonavir’s CYP3A4-inhibition effect. In Dickinson 2009 the value is computed by non-compartmental methods on the observed ritonavir concentration-time profile (WinNonlin 5.2) and feeds the atazanavir CL/F power form cl = exp(lcl) * (CONMED_RTV_AUC / 7.52)^e_aucrtv_cl.
• Units: mg*h/L (document per-model via covariateData[[CONMED_RTV_AUC]]$units if a different exposure unit is reported).
• Type: continuous
• Scope: specific
• Reference category: n/a – enters via centred power form. Reference value observed: 7.52 mg*h/L (Dickinson 2009 cohort median across HIV-infected and healthy volunteers; healthy 7.36, HIV 7.59 per Table 1).
• Source aliases:
  • RTVAUC – printed name in Dickinson 2009 Table 1 / Table 3 (the paper writes RTVAUC0-24 with the dosing-interval subscript; the column name is registered without the subscript to fit standard data-column character constraints).
• Example models: Dickinson_2009_atazanavir.R (atazanavir CL/F power-function dependence on RTVAUC0-24, centred at 7.52 mgh/L with exponent -0.8), Kengo_2025_atazanavir.R (screened on atazanavir clearance as a centred linear-deviation effect, RTVAUC_CL = 1 + THETA(11)*(AUC_RTV2 - 7.0923), but NOT retained – THETA(11) is fixed at zero in the final control stream, so the covariate is carried in covariatesDataExcluded. The DERIVE cohort median of 7.0923 mgh/L is close to the Dickinson 2009 centring value of 7.52. Kengo 2025 Results 3.4 and its Discussion explain the failure: atazanavir and ritonavir share a clearance pathway and were given at a fixed dose ratio in every participant, so ritonavir exposure and atazanavir clearance are collinear and causality cannot be separated from correlation within that design – a warning that this covariate is only identifiable when the booster dose is varied independently of the substrate).
• Notes: Specific scope because the column meaning is tied to ritonavir as the booster drug and to the 0-24 h once-daily dosing-interval AUC convention. A future PK model that uses a different ritonavir exposure metric (trough concentration, q12h-interval AUC for BID ritonavir regimens) should register a parallel canonical rather than overload CONMED_RTV_AUC. For simulation users without observed ritonavir AUC, the Dickinson 2009 cohort median 7.52 mg*h/L reproduces typical-value behaviour (the centring point of the covariate effect). Distinct from CONMED_RTV (binary boost-status flag used in Colombo 2006); the two canonicals coexist for the binary-indicator and continuous-exposure parameterisations of the same co-medication.

DOSE_RTV_MGKG (canonical for concomitant ritonavir per-administration dose per kg body weight)

• Description: Concomitant ritonavir dose given per administration in mg per kg of body weight (mg/kg). Per-dose-record covariate carried on the lopinavir (or other RTV-boosted antiretroviral) dose row so the parent drug’s bioavailability or clearance equation can read the co-administered ritonavir dose at the moment of dosing. Time-varying across cohort / regimen changes; constant within an inter-dose interval at steady state. Distinct from the per-administration mg dose of ritonavir on the RTV dose-event row – this covariate is the value mg-divided-by-kg, available on every record so the bioavailability formula can be evaluated without back-computing dose from the event table.
• Units: mg/kg
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as a linear-shift regressor in the LPV bioavailability formula F_LPV = 1 - RIF * CONMED_RIF + SLP * (DOSE_RTV_MGKG - DOSE_RTV_STD) of Zhang 2012 Equation 4, with DOSE_RTV_STD = 3 mg/kg (the median ritonavir dose in the no-rifampicin standard-LPV/r arm). The standard 4:1 LPV/r without rifampicin reference uses DOSE_RTV_MGKG = 3 so the linear-shift term vanishes; super-boosted 1:1 LPV/r at median LPV 14 mg/kg gives DOSE_RTV_MGKG = 14; double-dose 4:1 LPV/r at median LPV 23 mg/kg gives DOSE_RTV_MGKG = 5.75.
• Source aliases:
  • DoseRTV – used in Zhang_2012_lopinavir_ritonavir.R (paper Equation 4 regressor; in the paper’s NONMEM dataset this is the per-dose individual ritonavir mg/kg carried as a per-dose covariate column rather than a derived state).
• Example models: Zhang_2012_lopinavir_ritonavir.R (linear-shift on LPV bioavailability: F_LPV = 1 - 0.832 * CONMED_RIF + 0.021 * (DOSE_RTV_MGKG - 3); the linear approximation is valid only inside the cohort-tested ritonavir-dose range of 2.9-14 mg/kg per Zhang 2012 Discussion paragraph 3).
• Notes: Specific scope because the absolute slope coefficient (0.021 / mg/kg in Zhang 2012) and the reference dose (3 mg/kg) are intrinsically tied to the lopinavir-ritonavir-rifampicin pediatric integrated model. Drug-self-dose covariates for other drugs should register sibling canonicals (e.g., DOSE_<DRUG>_MGKG for per-administration or DOSE_<DRUG>_MGKGD for daily) rather than reuse this name. Carrying ritonavir dose as an explicit covariate rather than reading it from the RTV dose-event amount avoids tightly coupling the LPV bioavailability formula to the rxode2 event-table structure and lets a downstream simulation user supply a synthetic per-administration ritonavir dose without constructing matching event records.

DOSE_RIV_MGKG (canonical for per-administration rivaroxaban dose per kg body weight)

• Description: Per-administration rivaroxaban dose in mg per kg of body weight (mg/kg). Per-dose-record covariate carried on the rivaroxaban dose row so the dose-dependent relative oral bioavailability function can read the per-event mg/kg value at the moment of dosing. Time-varying across study phase / regimen changes (single dose, b.i.d., t.i.d.) and weight-band re-titrations; constant within an inter-dose interval at steady state.
• Units: mg/kg
• Type: continuous
• Scope: specific
• Reference category: n/a – enters non-linearly as the input to the exponential decay function tf1 = f1min + (f1max - f1min) * exp(-log(2)/d50 * DOSE_RIV_MGKG) of Willmann 2021 supplement S2 NONMEM $PK block, with f1max = 1.25, f1min = 0.59, d50 = 14.4/82.48 = 0.1746 mg/kg, anchored to F1 = 1.0 at DOSE_RIV_MGKG = 10/82.48 = 0.1213 mg/kg (the median adult dose per the integrated adult popPK analysis in Willmann 2021 reference 19). Reference values observed: 0.1213 mg/kg (adult-anchor F1 = 1.0); 0.30 mg/kg (F1 = 0.791); 0.50 mg/kg (F1 = 0.681) – Willmann 2021 Results p. 1199. Bodyweight-normalised single rivaroxaban doses ranged 0.1-0.5 mg/kg across the EINSTEIN-Jr pediatric program (Willmann 2021 Results p. 1199 and Figure 1).
• Source aliases:
  • DW (= DOSE/WGHT in supplement S2 $PK block) – used in Willmann_2021_rivaroxaban.R (supplement S2 NONMEM $PK block DW = DOSE/WGHT; TF1 = F1MIN + (F1MAX - F1MIN) * EXP(-LOG(2)/D50 * DW)).
• Example models: Willmann_2021_rivaroxaban.R (input to the exponential-decay relative-bioavailability function for the pediatric popPK model of rivaroxaban; tf1 = 0.59 + 0.66 * exp(-log(2)/0.1746 * DOSE_RIV_MGKG) with f1 = tf1 * exp(etalfdepot) and f(depot) <- f1).
• Notes: Specific scope because the absolute constants f1max, f1min, and d50 are intrinsically tied to the dose-dependency function carried over from the integrated adult popPK analysis in Willmann 2021 reference 19 and re-anchored to F1 = 1.0 at 10 mg/82.48 kg for the pediatric extrapolation. Carrying rivaroxaban dose per kg as an explicit covariate column (rather than back-computing from the rxode2 event-table amt and a body-weight column) lets the bioavailability formula be evaluated on every record including non-dose observation records. Mirrors the sibling drug-self-dose canonicals DOSE_PHT_MGKGD (Yukawa 1990 daily phenytoin dose per kg) and DOSE_RTV_MGKG (Zhang 2012 per-administration ritonavir dose per kg); pattern DOSE_<DRUG>_MGKG (per-administration) vs DOSE_<DRUG>_MGKGD (daily).

DOSE_BZN_MGKG (canonical for per-administration benznidazole dose per kg body weight)

• Description: Per-administration oral benznidazole dose in mg per kg of body weight (mg/kg). Carried on every record (not only dose rows) so the power covariate function on the absorption rate constant can be evaluated at observation times as well. Time-varying across dose levels; constant within a dose level. In a mouse study the physical dose amount is individually adjusted to each animal’s weight, so the mg/kg dose level – not the mg amount – is the quantity the covariate model is written against.
• Units: mg/kg
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as a power term normalised to the median dose of the satellite PK study, (DOSE_BZN_MGKG / 30)^e_dose_ka. Assmus 2025 studied 10, 30 and 100 mg/kg in the satellite PK cohort (reference 30 mg/kg) and 10, 20, 30, 50 and 100 mg/kg per administration across the ten efficacy regimens.
• Source aliases:
  • DOSE_PER_KG – the S1 Code $PK regressor in Assmus_2025_benznidazole_mouse.R (COV1 = (DOSE_PER_KG/30)**THETA(5)).
• Example models: Assmus_2025_benznidazole_mouse.R (power effect on the absorption rate constant: ka = 2.18 * (DOSE_BZN_MGKG/30)^-0.775, Assmus 2025 Table 2 theta_Dose = -0.775, giving KA = 5.11, 2.18 and 0.86 1/h at 10, 30 and 100 mg/kg).
• Notes: Specific scope until a second benznidazole model ratifies the canonical. Member of the DOSE_<DRUG>_<UNITS> drug-self-dose family alongside DOSE_RIV_MGKG, DOSE_RTV_MGKG, DOSE_GLN_GKG, DOSE_PHT_MGKGD and DOSE_VPA_MGKGD; pattern DOSE_<DRUG>_MGKG (per-administration) vs DOSE_<DRUG>_MGKGD (daily). Must NOT be named the bare DOSE: rxode2’s etTrans() consumes an event-table column literally named DOSE (any casing) and never exposes it to model(), so the solve fails with “The following parameter(s) are required for solving: DOSE”.

DOSE_GLN_GKG (canonical for per-administration L-glutamine dose per kg body weight)

• Description: Per-administration oral L-glutamine dose in grams per kg of body weight (g/kg). Carried on every record (not only dose rows) so a dose-dependent covariate term can be evaluated at observation times as well. Time-varying across dose-escalation periods; constant within a dose level.
• Units: g/kg
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as a power term normalised to the lowest studied dose level, (DOSE_GLN_GKG / 0.1)^theta. Sadaf 2024 studied 0.1, 0.3, and 0.6 g/kg, with 0.1 g/kg as the normalisation reference.
• Source aliases:
  • DOSE per kg – used in Sadaf_2024_glutamine.R (Sadaf 2024 Table 2 Vi = theta3 * (WT/70) * (DOSE per kg/0.1)^theta4).
• Example models: Sadaf_2024_glutamine.R (power effect on apparent central volume: vc = 52.3 * (WT/70)^1 * (DOSE_GLN_GKG/0.1)^0.27, Sadaf 2024 Table 2 theta4 = 0.27).
• Notes: Specific scope until a second L-glutamine model ratifies the canonical. Member of the DOSE_<DRUG>_<UNITS> drug-self-dose family alongside DOSE_RIV_MGKG, DOSE_RTV_MGKG, DOSE_PHT_MGKGD, and DOSE_VPA_MGKGD; the GKG unit token is used rather than MGKG because oral amino-acid supplementation is dosed in grams per kg, and silently rescaling to mg/kg would make the published normalisation reference (0.1) read as 100. Distinct from GLN_BL, the endogenous baseline glutamine concentration, which is the other Sadaf 2024 covariate.

CP_PRB_MGL (canonical for instantaneous probenecid plasma concentration as a time-varying renal-tubular-perpetrator covariate)

• Description: Instantaneous plasma concentration of probenecid supplied directly as a time-varying covariate column rather than computed from a coupled PK model. Used as the perpetrator-drug input to a competitive inhibition term on saturable (Michaelis-Menten) renal tubular secretion of a co-administered substrate in DDI popPK models. Set to 0 outside the probenecid co-administration window so the apparent-Km inhibition term collapses to the baseline form km_apparent = km.
• Units: mg/L (probenecid MW = 285.34 g/mol; 1 mg/L = 3.505 umol/L).
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as the multiplier on the saturable-secretion apparent-Km term km_apparent = km * (1 + CP_PRB_MGL / ki) where ki is the probenecid competitive inhibition constant at the renal tubular transporter in mg/L. Reference peak observed: the Landersdorfer 2009 8-dose probenecid regimen (total 4.5 g over 70 h) produces typical plasma probenecid concentrations in the range ~30 to 60 mg/L during the gemifloxacin sampling window (Landersdorfer 2009 Fig. 1C).
• Source aliases:
  • [P] (Landersdorfer 2009 Methods / Table 1, “plasma probenecid concentration” symbol entering the competitive-inhibition formula km_apparent = km * (1 + [P] / kic)).
• Example models: Landersdorfer_2009_gemifloxacin.R (drives competitive inhibition of the renal-tubular-secretion arm of gemifloxacin clearance: cl_tubular_secretion = vmax * Cc / (km * (1 + CP_PRB_MGL / ki) + Cc) with ki = 69.3 mg/L; gemifloxacin filtration clearance (fu * GFR ~ 2 L/h, fixed) and non-renal clearance are unaffected by this term but the static-treatment-arm effect on non-renal clearance is encoded separately via CONMED_PROBENECID).
• Notes: Specific scope; probenecid-specific. The drug-specific naming follows the established CP_<drug>_<units> precedent (CP_OXY_NGML, CP_FBX_NGML, CP_LSN_NGML, CP_MORPH_NGML, CP_RIF_UM). Distinct from the binary indicator CONMED_PROBENECID (which captures treatment-arm-level static effects – e.g., on absorption or non-renal clearance – and does not carry plasma-concentration magnitude information). The Landersdorfer 2009 paper does NOT report a probenecid PK sub-model parameter table (the Methods describe the structural form – 1-compartment + lag + parallel first-order + mixed-order elimination – but no parameter table), so users simulating the with-probenecid arm must supply CP_PRB_MGL from an external probenecid popPK source (e.g., literature digitisation of Landersdorfer 2009 Fig. 1C, or an unrelated published probenecid popPK model).

CONC_AGONIST_M (canonical for applied ex-vivo agonist concentration in the Grzesk 2016 vascular-reactivity sigmoidal Emax CRC model)

• Description: Applied concentration (mol/L) of the vasoactive agonist whose identity is selected by the companion AGONIST_CODE covariate, supplied as a per-record exogenous covariate to the static sigmoidal Emax concentration-response model of perfusion pressure in the isolated, perfused male Wistar rat tail artery. Polymorphic in agonist identity: a single CONC_AGONIST_M column carries phenylephrine concentrations on phenylephrine records, arg-vasopressin concentrations on arg-vasopressin records, mastoparan-7 concentrations on mastoparan-7 records, and Bay K8644 concentrations on Bay K8644 records, with the per-agonist EC50 / Emax pair selected inside model by AGONIST_CODE. The model has no PK and no time dynamics; CONC_AGONIST_M is the dose-response x-axis.
• Units: M (mol/L)
• Type: continuous
• Scope: specific
• Reference category: n/a – enters the sigmoidal Emax expression PP = emax * CONC_AGONIST_M^hill / (ec50^hill + CONC_AGONIST_M^hill); CONC_AGONIST_M = 0 yields PP = 0 (no agonist contraction).
• Source aliases: none – Grzesk 2016 Table I reports per-agonist EC50 in M/L directly; the model column is the canonical CONC_AGONIST_M.
• Example models: Grzesk_2016_m3M3FBS.R (Grzesk 2016 Table I CRC data: per-agonist EC50 values 7.50e-8 M (PHE), 1.84e-8 M (AVP), 4.48e-8 M (mastoparan-7), 1.96e-6 M (Bay K8644); m-3M3FBS-shifted EC50 values 6.45e-8, 1.42e-8, 2.55e-8, 2.05e-6 M respectively).
• Notes: Specific scope because the polymorphic agonist-identity-selected-by-companion-column semantics are tied to the Grzesk 2016 four-agonist CRC design. Distinct from the STIM_<drug>_<units> and CONC_<DRUG>_<UNITS> in-vitro-drug-concentration families (where the drug identity is fixed in the canonical name and the unit suffix is mandatory) because Grzesk 2016 deliberately tests four mechanistically distinct vasoactive agonists in a shared experimental scaffold with a shared concentration axis – baking the agonist identity into the canonical name would force four parallel columns rather than the paper’s one-column-plus-selector encoding. M (mol/L) chosen as the unit because Grzesk 2016 Table I reports EC50 values in mol/L and the four agonists span 1e-8 to 1e-6 M (a uniform molar axis is the only practical choice); the _M suffix preserves the CONC_<...>_<UNITS> family’s unit-in-name discipline. Future ex-vivo / in-vitro CRC papers from the Grzesk-group series (Biomed Rep 2 / 2014; Mol Med Report 5 / 2012; Exp Ther Med 4 / 2012; etc.) that test multiple agonists on the same vascular-reactivity scaffold should reuse this canonical with an extended AGONIST_CODE mapping.

CONC_INHIBITOR_MEDIUM_UNBOUND_UM (canonical for the unbound concentration of a transporter or enzyme inhibitor in an in-vitro incubation medium)

• Description: Unbound (free) concentration of the co-incubated inhibitor in the medium bathing an in-vitro cell preparation, supplied as a per-record exogenous covariate. Drives competitive inhibition of transport processes located on the membrane facing the medium – in a hepatocyte assay, the sinusoidal uptake transporters – through a term of the form cl_eff = cl / (1 + CONC_INHIBITOR_MEDIUM_UNBOUND_UM / ic50). Polymorphic in inhibitor identity in the same way as CONC_AGONIST_M: one column carries whichever inhibitor a given record was incubated with, and the matching potency is selected inside model() by a companion CONMED_<INN> indicator. Set to 0 for the no-inhibitor condition, which collapses the term to its uninhibited form.
• Units: umol/L (= uM)
• Type: continuous
• Scope: general
• Reference category: n/a – enters as the numerator of a competitive-inhibition ratio; 0 is the uninhibited reference.
• Source aliases:
  • [I]u,med – Guo 2016 eq. 7 and Table 3; same orientation and units as the canonical.
• Example models: Guo_2016_taurocholate_schh.R (Table 3 measurements in 4 percent BSA: telmisartan 0.012 uM at a 1 uM dosing concentration and 0.20 uM at 10 uM; bosentan 0.031 uM at 0.8 uM and 0.45 uM at 8 uM; drives the NTCP / OATP1B1 split of taurocholate uptake clearance).
• Notes: General scope because a medium unbound inhibitor concentration means the same thing in any in-vitro inhibition experiment. The unbound rather than the total medium concentration is named in the canonical because the free-drug hypothesis makes the unbound concentration the mechanistically correct driver at a membrane facing the medium, and because in-vitro systems containing physiologic albumin have total and unbound concentrations that differ by one to two orders of magnitude; a model must never silently supply one where the other is meant. A member of the applied-concentration CONC_<...>_<UNITS> family, role-named rather than drug-named after the CONC_AGONIST_M precedent, because the whole point of these designs is to run several mechanistically distinct inhibitors through one experimental scaffold. Pairs with CONC_INHIBITOR_CELL_UM, the intracellular counterpart. Distinct from CP_<drug>_<units>, which is an in-vivo plasma perpetrator concentration.

CONC_INHIBITOR_CELL_UM (canonical for the intracellular concentration of a transporter or enzyme inhibitor in an in-vitro cell preparation)

• Description: Concentration of the co-incubated inhibitor inside the cells of an in-vitro preparation, supplied as a per-record exogenous covariate. Drives competitive inhibition of processes whose inhibitor-binding site faces the cell interior – in a hepatocyte assay, the canalicular and basolateral efflux transporters – through a term of the form cl_eff = cl / (1 + CONC_INHIBITOR_CELL_UM / ic50). Polymorphic in inhibitor identity, with the matching potency selected inside model() by a companion CONMED_<INN> indicator. Set to 0 for the no-inhibitor condition.
• Units: umol/L (= uM)
• Type: continuous
• Scope: general
• Reference category: n/a – enters as the numerator of a competitive-inhibition ratio; 0 is the uninhibited reference.
• Source aliases:
  • [I]cell – Guo 2016 eqs. 8-10, the generic slot into which the paper substitutes each of its four measured intracellular quantities in turn.
  • [I]t,cell, [I]u,cell, [I]t,cyt, [I]u,cyt – Guo 2016 Table 3: cellular total, cellular unbound, cytosolic total and cytosolic unbound. All four are alternative values for this one column, NOT four separate canonicals, because the paper’s central question is which of them predicts best; the model structure is identical under all four.
• Example models: Guo_2016_taurocholate_schh.R (Table 3, in the order [I]t,cell / [I]u,cell / [I]t,cyt / [I]u,cyt: telmisartan 1 uM = 16 / 2.1 / 16 / 0.85 uM and 10 uM = 40 / 3.7 / 35 / 2.8 uM; bosentan 0.8 uM = 1.9 / 0.79 / 1.7 / 0.21 uM and 8 uM = 17 / 3.8 / 14 / not available).
• Notes: General scope, for the same reason as its medium counterpart. Unlike CONC_INHIBITOR_MEDIUM_UNBOUND_UM, the canonical name does NOT commit to total or unbound, and does not commit to whole-cell or cytosol, because the choice among those is itself the modelling question this covariate exists to support. Every model using it must therefore state in covariateData[[CONC_INHIBITOR_CELL_UM]]$notes which quantity its values are; the predicted effect differs by roughly the intracellular unbound fraction, which was 0.09-0.13 for telmisartan and 0.22-0.41 for bosentan in the founding example, so the distinction is load-bearing. Where a paper reports only the total cellular concentration and no unbound fraction, record that the values are total rather than converting them.

CONC_HCN_PPM (canonical for inhaled hydrogen cyanide gas concentration in ambient air)

• Description: Hydrogen cyanide (HCN) concentration in the ambient air a subject is breathing, supplied as an exogenous covariate that is the sole exposure driver of an inhalation toxicokinetic PBPK model. Enters the pulmonary mass balance as the air-side driving force of alveolar exchange, q_alv * (ca - plasma / ppa), where ca = CONC_HCN_PPM / 24 converts the gas-phase ppm reading to umol/L and ppa is the plasma:air partition coefficient. There is no dosing event and no administered amount: the entire exposure is specified by this column, so a constant value represents a steady atmosphere from t = 0 and a time-varying column represents a fluctuating fire-scene profile. In the founding example the value is not measured but estimated per decedent by grid search (0-18,000 ppm in 12 ppm steps) against paired post-mortem left and right cardiac blood cyanide concentrations.
• Units: ppm
• Type: continuous
• Scope: specific
• Reference category: n/a – CONC_HCN_PPM = 0 is clean air and yields no cyanide uptake, so zero is the natural unexposed reference.
• Source aliases:
  • Ca – the variable name used in the Harada 2025 Supplemental Material 1 Python script, where it carries the same quantity already converted to umol/L. A source data set expressed in umol/L must be multiplied by 24 to reach the canonical ppm column, or the model’s ppm_per_umol conversion factor overridden; the canonical is ppm because ppm is the unit both the paper’s text and its Table 2 report.
• Example models: Harada_2025_cyanide_pbpk.R (inhaled HCN air concentration driving alveolar cyanide uptake in the Stamyr 2015 six-state cyanide PBPK model; the paper’s 13 reconstructed exposures span 84-16,632 ppm).
• Notes: Specific scope because the value is bound to hydrogen cyanide and to an inhalation-exposure design. Member of the applied-concentration CONC_<drug>_<units> family, which was already not restricted to in-vitro media – CONC_VORI_NGML carries an in-vivo whole-blood concentration – and which this entry extends to the inhaled gas phase. The distinguishing feature against every sibling is the medium: CONC_HCN_PPM is a concentration in ambient air outside the body, whereas the CONC_<drug>_MGL / _UM siblings are concentrations in an in-vitro aqueous matrix and CONC_VORI_NGML is in whole blood. Because the medium is air, the covariate cannot be compared to a blood or plasma concentration without the partition coefficient, and the unit suffix is a gas-phase mixing ratio rather than a mass-per-volume. Distinct from ETSEVO (end-tidal sevoflurane), which is an exhaled alveolar concentration measured at the subject’s airway and therefore already equilibrated with tissue, whereas CONC_HCN_PPM is the inspired atmosphere and is an input rather than a readout. Distinct from CARDIAC_OUTPUT, the other canonical pre-registered for inhalation-model work. Future inhalation toxicokinetic extractions for a different gas should register a sibling (CONC_CO_PPM, CONC_HCL_PPM,…) following this pattern rather than overloading this name; combustion-toxicology papers commonly report carbon monoxide alongside HCN, and Harada 2025 measures carboxyhaemoglobin but does not model carbon monoxide kinetics, so no CONC_CO_PPM is registered yet.

CP_REM_NGML (canonical for instantaneous remikiren plasma concentration as a time-varying PD driver)

• Description: Instantaneous plasma concentration of remikiren (Ro 42-5892; orally active renin inhibitor) supplied directly as a time-varying covariate column rather than computed from a coupled PK model. Used in PD-only Hill / Emax models that take remikiren exposure as an external input to a renin-inhibition concentration-effect equation.
• Units: ng/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as a Hill / Emax term in the inhibition-of-angiotensin-I-production-rate expression APR = Imax * CP_REM_NGML^hill / (IC50^hill + CP_REM_NGML^hill) (Weber 1993). Set to 0 outside the drug-exposure window (the inhibition term then collapses to 0). Reference values observed: mean Cmax was 4-6 ng/mL after 200 mg oral, 23-27 ng/mL after 300 mg oral, 47-83 ng/mL after 600-800 mg oral, with substantial intersubject variability (Weber 1993 Table 1).
• Source aliases:
  • Cp (Weber 1993 Methods ‘Concentration-effect modelling’ Hill equation; values in ng/mL) – the observed plasma remikiren concentration measured by HPLC with fluorescence or coulometric electrochemical detection or by a protein-binding assay (Weber 1993 Methods, ‘Remikiren concentrations’ section).
• Example models: Weber_1993_remikiren.R (PD-only Hill / Emax model for inhibition of the angiotensin I production rate in hypertensive patients; CP_REM_NGML is the observed or simulated plasma concentration of remikiren supplied per event row).
• Notes: Specific scope; remikiren-specific. The drug-specific naming follows the established CP_<drug>_<units> precedent (CP_OXY_NGML, CP_FBX_NGML, CP_LSN_NGML, CP_MORPH_NGML, CP_RIF_UM). Weber 1993 fit the PD model directly to observed plasma concentrations because no structural population PK model was estimated – the paper characterised remikiren PK with model-independent NCA only. Users wishing to drive the PD model from a simulated PK source must supply their own concentration trajectory (the published single-dose Weber-group papers from the same series report compartmental PK in healthy volunteers but are not represented in nlmixr2lib).

CP_RITUXIMAB_UGML (canonical for instantaneous rituximab plasma concentration as a time-varying PD driver)

• Description: Instantaneous plasma concentration of rituximab (chimeric anti-CD20 monoclonal antibody), or of one of its biosimilars, supplied directly as a time-varying covariate column rather than computed from a coupled PK model. Used in PD-only exposure-response models that take rituximab exposure as an external input to a concentration-effect equation (e.g., Emax on DAS28 change from baseline).
• Units: ug/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as an Emax-style concentration-effect term fC = Emax * CP_RITUXIMAB_UGML / (CP_RITUXIMAB_UGML + ec50). Set to 0 outside the drug-exposure window (the concentration-effect term then collapses to 0). Reference values observed: typical steady-state Cmax after the two-dose loading course (1000 mg IV on days 1 and 15) reaches the low hundreds of ug/mL and washes out to <1 ug/mL by week 24 (Williams 2016 Supplemental Figure S1 VPC).
• Source aliases:
  • CONC – Williams 2016 popPK/PD analysis; the paper’s popPK model produced individual predicted concentrations Cij that were passed to the DAS28cfb PD model as an input variable.
• Example models: Williams_2016_rituximab_das28cfb.R (PD-only Emax-style exposure-response model for the DAS28 change from baseline; the source rituximab popPK model was two-compartment with baseline-BSA and sex covariates on CL and Vc, but Williams 2016 does not tabulate the popPK parameter estimates, so users must supply their own concentration trajectory), Bender_2024_mosunetuzumab.R (baseline usage: the patient’s observed predose rituximab concentration seeds the initial condition of a ritux state that then decays at a fixed 24-day terminal half-life, supplying the competing-ligand term of the CD20 receptor-occupancy observable).
• Notes: Specific scope; rituximab-specific. Same covariate is used for the biosimilar candidate PF-05280586 in the Williams 2016 trial (PF-05280586 was shown to be PK-similar to reference rituximab). Drug-product distinction is captured by the separate TRT covariate rather than by separate concentration columns. Two usages coexist under this one name: a genuinely time-varying driver supplied per event row (Williams 2016) and a baseline predose value read once at t = 0 to seed a decaying state (Bender 2024). Both are the same physical quantity in the same units, so they share the column; the model file’s covariateData notes records which usage applies. Bender 2024 reports the concentration as floored at the 0.5 ug/mL assay LOQ, so a “0.500” value there means “at or below the limit of quantification”, not a measured level.

CP_OBINUTUZUMAB_UGML (canonical for instantaneous obinutuzumab plasma concentration as a time-varying PD driver)

• Description: Instantaneous plasma concentration of obinutuzumab (GA101; glycoengineered type II humanized anti-CD20 monoclonal antibody) supplied directly as a covariate column rather than computed from a coupled PK model. Used both as a time-varying exposure-response driver and, at baseline, as the predose residual concentration remaining from a prior anti-CD20 treatment course.
• Units: ug/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as a competing-ligand concentration. Set to 0 when the patient has had no prior obinutuzumab exposure (the competing term then collapses to 0), which is the majority case: Bender 2024 Table 1 reports a median of 0 ug/mL with a maximum of 305 ug/mL across 439 patients, with only 35 patients carrying detectable residual drug.
• Source aliases:
  • BLOBIN – Bender 2024 NONMEM control stream (Table S3); baseline obinutuzumab concentration in ng/mL, used both to seed A_0(5) and, via the aCD20 composite, to carry the covariate effect on baseline clearance.
• Example models: Bender_2024_mosunetuzumab.R (the observed predose obinutuzumab concentration seeds the initial condition of an obin state that decays at a fixed 28-day terminal half-life; the state supplies the obinutuzumab competing-ligand term (kd_mosun / kd_obin) * obin of the CD20 receptor-occupancy observable).
• Notes: Specific scope; obinutuzumab-specific. Follows the established CP_<drug>_<units> pattern (CP_RITUXIMAB_UGML, CP_SUTIMLIMAB_UGML, CP_GLASDEGIB_NGML, …); ug/mL matches the anti-CD20 monoclonal-antibody sibling CP_RITUXIMAB_UGML and is the unit Bender 2024 Table 1 reports. Do not confuse with the obinutuzumab popPK model Gibiansky_2014_obinutuzumab.R, which models obinutuzumab as the analyte in central / peripheral1 rather than taking its concentration as an input column. A user who wants a mechanistically-generated trajectory instead of a decaying-baseline approximation can drive this column from that model.

CP_ACD20_UGML (canonical for the composite baseline anti-CD20 drug concentration)

• Description: Composite baseline plasma concentration of residual anti-CD20 monoclonal antibody carried over from a patient’s prior treatment, defined as the maximum of the patient’s predose rituximab and obinutuzumab concentrations: CP_ACD20_UGML = max(CP_RITUXIMAB_UGML, CP_OBINUTUZUMAB_UGML). It exists because a patient may have received both drugs in prior lines of therapy and therefore carry detectable residual levels of both; the composite collapses the two correlated columns into the single quantity a clearance covariate effect can act on.

• Units: ug/mL

• Type: continuous

• Scope: specific

• Reference category: 0.5 ug/mL (= 500 ng/mL), the rituximab assay LOQ, which is also the value the majority of patients take. Bender 2024 Figure 2 defines the typical patient as having “a baseline anti-CD20 (aCD20) drug concentration of 0.5 ug/mL”; the 95th percentile of the distribution is 55.91 ug/mL.

• Source aliases:

  • ACD20 – Bender 2024 NONMEM control stream (Table S3), in ng/mL, derived as IF (BLOBIN.GT.BLRITUX2) ACD20=BLOBIN / IF (BLRITUX2.GT.BLOBIN) ACD20=BLRITUX2.

• Example models: Bender_2024_mosunetuzumab.R (covariate on the baseline clearance CLbase of the time-dependent-clearance model).

• Notes: Specific scope; anti-CD20 drug class. The token is a drug class composite rather than a single INN, which is a deliberate widening of the CP_<drug>_<units> family shape (CP_EIDD_NGML already carries a non-INN token); it is kept in the family so that all three anti-CD20 concentration columns read consistently.

  This column is derived, not measured – populate it, do not assume a model will compute it. Bender 2024 applies the covariate effect to an imputed aCD20 while seeding the ODE initial conditions from the raw per-drug columns, so max() taken inside model() would not reproduce the paper for the 4.6% of patients whose rituximab value is missing. The control stream imputes a missing rituximab concentration by NHL type before taking the maximum: aggressive NHL (NHLTYPE == 1) or unknown -> 2105 ng/mL (2.105 ug/mL); indolent NHL (NHLTYPE == 2) -> 500 ng/mL (0.5 ug/mL). Users reproducing the published fit should apply that imputation when deriving the column.

  The covariate enters as a ratio of logarithms, not as a concentration ratio, and is therefore not scale-invariant: Bender 2024 writes (LOG(ACD20)/LOG(500))**theta with ACD20 in ng/mL. A model file carrying this column in ug/mL must convert inside the logarithm (log(CP_ACD20_UGML * 1000) / log(500)). The naive concentration-ratio reading (ACD20/500)^theta is falsified by the paper’s own statement that every covariate except albumin changes the typical parameter by no more than 31% at the distribution extremes: at the 95th percentile it predicts a 93% reduction in CLbase, versus 28% for the ratio-of-logs form. The term additionally requires CP_ACD20_UGML > 0.001 (1 ng/mL) to keep the logarithm positive; in practice the rituximab LOQ floor of 0.5 ug/mL keeps every patient well above it.

CP_SUTIMLIMAB_UGML (canonical for instantaneous sutimlimab serum concentration as a time-varying PD driver)

• Description: Instantaneous serum concentration of sutimlimab (BIVV009 / TNT009; humanized monoclonal IgG4 antibody against complement factor C1s) supplied directly as a time-varying covariate column rather than computed from a coupled PK model. Used in PD-only exposure-response models that take sutimlimab exposure as an external input to a concentration-effect equation (e.g., sigmoidal Imax knockdown of classical complement pathway activity).
• Units: ug/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as the driving concentration of the sigmoidal Imax expression CPactivity = e0 - imax * CP_SUTIMLIMAB_UGML^hill / (ic50^hill + CP_SUTIMLIMAB_UGML^hill) (Bartko 2018). Set to 0 outside the drug-exposure window (the inhibition term then collapses to 0 and CP activity returns to e0). Reference values observed (Bartko 2018 Table 2): mean Cmax rose from 40 ug/mL at 3 mg/kg to 2036 ug/mL at 100 mg/kg after a single 60-minute i.v. infusion, against an IC50 of 6.2 ug/mL and an IC90 of 15.5 ug/mL; serum concentrations were below the limit of quantification at the 0.3 and 1 mg/kg dose levels.
• Source aliases: none known; Bartko 2018 refers to it only as the serum sutimlimab (BIVV009) concentration measured by a validated immunoassay and supplies no column name.
• Example models: Bartko_2018_sutimlimab.R (PD-only sigmoidal Imax model of classical complement pathway knockdown in healthy volunteers; CP_SUTIMLIMAB_UGML is the observed or simulated serum concentration supplied per event row).
• Notes: Specific scope; sutimlimab-specific. The drug-specific naming follows the established CP_<drug>_<units> precedent (CP_OXY_NGML, CP_FBX_NGML, CP_LSN_NGML, CP_MORPH_NGML, CP_REM_NGML, CP_GLASDEGIB_NGML, CP_RITUXIMAB_UGML); ug/mL matches the monoclonal-antibody sibling CP_RITUXIMAB_UGML and is the unit Bartko 2018 reports throughout. Bartko 2018 fit the PD model directly to observed serum concentrations because no structural population PK model was estimated – the paper characterised sutimlimab PK with model-independent NCA only (Table 2) and explicitly reports nonlinear, apparently target-mediated elimination below about 100 ug/mL, which a linear compartmental model would not capture. No sutimlimab population PK model exists in the nlmixr2lib registry, so users wishing to drive this PD model from a simulated PK source must supply their own concentration trajectory.

CP_GLASDEGIB_NGML (canonical for instantaneous glasdegib plasma concentration as a time-varying PD driver)

• Description: Instantaneous plasma concentration of glasdegib (PF-04449913; orally active Hedgehog pathway / Smoothened inhibitor) supplied directly as a time-varying covariate column rather than computed from a coupled PK model. Used in PD-only linear exposure-response models that take glasdegib exposure as an external input to a concentration-effect equation (e.g., concentration-QTc).
• Units: ng/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as a linear-slope term in the concentration-QTc expression QTcF = e0 + slope * (CP_GLASDEGIB_NGML / 1000) (Fostvedt 2021; the in-model / 1000 converts the canonical ng/mL covariate to the paper’s reported microgram-per-mL slope scaling). Set to 0 outside the drug-exposure window (the concentration-slope term then collapses to 0). Reference values observed: geometric-mean steady-state Cmax was 1137 ng/mL at 100 mg QD (therapeutic dose) and 2445 ng/mL at 200 mg QD (a 2-fold supratherapeutic exposure used in Fostvedt 2021’s bootstrap predictions; see Fostvedt 2021 Methods 2.3 and Table 4).
• Source aliases:
  • CONC (Fostvedt 2021 Methods 2.2 and 2.3 ‘concentration-QTc modelling’ equations; values in ng/mL but model fit with the slope reported per microgram-per-mL after a CONC / 1000 rescaling internal to the estimation – “The estimates are reported on the microgram scale as the scaling helped the estimation procedure”, Fostvedt 2021 Table 3 footnote).
• Example models: Fostvedt_2021_glasdegib_QTcF.R, Fostvedt_2021_glasdegib_QTcS.R (PD-only linear mixed-effects E-R model for QTcF / QTcS prolongation in patients with cancer; CP_GLASDEGIB_NGML is the observed or simulated plasma concentration supplied per event row).
• Notes: Specific scope; glasdegib-specific. The drug-specific naming follows the established CP_<drug>_<units> precedent (CP_OXY_NGML, CP_FBX_NGML, CP_LSN_NGML, CP_MORPH_NGML, CP_RIF_UM, CP_REM_NGML). Fostvedt 2021 fit the PD model directly to observed plasma concentrations because no structural population PK model is published in the source paper – the geometric-mean Cmax values used for the parametric-bootstrap QTcF predictions (Fostvedt 2021 Section 3.4; cited to reference [15] of the paper) come from a separate Pfizer phase 2 study in patients with hematologic malignancies, not from a parameterised popPK model included in this paper. Users wishing to drive the PD model from a simulated PK source must supply their own concentration trajectory; no glasdegib popPK model exists in the nlmixr2lib registry.

CP_ORISMILAST_NGML (canonical for instantaneous orismilast plasma concentration as a time-varying PD driver)

• Description: Instantaneous plasma concentration of orismilast (UNION Therapeutics oral selective phosphodiesterase 4B/4D inhibitor) supplied directly as a time-varying covariate column rather than computed from a coupled PK model. Used in PD-only exposure-response models that take orismilast exposure as an external input to a concentration-effect equation (e.g., sigmoidal Imax inhibition of whole-blood IL-13 production).
• Units: ng/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as the driving concentration of the sigmoidal Imax expression il13Inhibition = imax * CP_ORISMILAST_NGML^hill / (ec50^hill + CP_ORISMILAST_NGML^hill) (Warren 2025). Set to 0 outside the drug-exposure window (the inhibition term then collapses to 0). Reference values observed: model-predicted steady-state exposures in atopic dermatitis patients read from Warren 2025 Fig. 1A span Caverage-ss ~38-63 ng/mL and Cmax-ss ~90-189 ng/mL across 20-40 mg twice daily, against a whole-blood IL-13 IC50 of 4.08 ng/mL and IC90 of 25.0 ng/mL.
• Source aliases: none known; Warren 2025 refers to it only as the predicted orismilast plasma concentration (Fig. 1A, 1B) and supplies no column name.
• Example models: Warren_2025_orismilast.R (PD-only sigmoidal Imax model of orismilast inhibition of LPS-stimulated whole-blood IL-13 production; CP_ORISMILAST_NGML is the observed or simulated plasma concentration supplied per event row).
• Notes: Specific scope; orismilast-specific. The drug-specific naming follows the established CP_<drug>_<units> precedent (CP_OXY_NGML, CP_FBX_NGML, CP_LSN_NGML, CP_MORPH_NGML, CP_REM_NGML, CP_GLASDEGIB_NGML). Warren 2025 measured the IL-13 potency in whole blood, so the paper maps plasma concentrations onto the potency curve directly with no protein-binding or cellular-penetration correction (“As the in vitro IL-13 potency was assessed in whole blood, adjustments for protein binding and cellular penetrations were not considered”, Warren 2025 Methods). No orismilast population PK model exists in the nlmixr2lib registry and none can be built from Warren 2025: the paper’s orismilast popPK model is a company-internal two-compartment-with-lag analysis of 366 subjects whose parameter values are not published in the paper, its reference list, or any supplement. Users wishing to drive this PD model must therefore supply their own concentration trajectory.

CP_RACSOTALOL_UGML (canonical for instantaneous rac-sotalol plasma concentration as a time-varying PD driver)

• Description: Instantaneous plasma concentration of rac-sotalol (racemic 4-[1-hydroxy-2-(propan-2-ylamino)ethyl]-N-methyl-benzene-sulfonamide; class III antiarrhythmic with non-selective beta-adrenoceptor blocking activity) supplied directly as a time-varying covariate column rather than computed from a coupled PK model. Used in PD-only linear exposure-response models that take rac-sotalol exposure as an external input to a concentration-DeltaQTc equation.
• Units: ug/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as a linear-slope term in the concentration-DeltaQTc expression DeltaQTcI = e0 + slope * CP_RACSOTALOL_UGML +... (Darpo 2014 Table 1 and Figure 6 captions; the slope is reported in ms per ug/mL, so no in-model unit rescaling is needed). Set to 0 outside the drug-exposure window (the concentration-slope term then collapses to 0). Reference values observed: Tmax 2.8 h (men) / 2.9 h (women); Cmax 1.4 ug/mL (men, range 0.9-1.9) / 1.8 ug/mL (women, range 1.1-2.8); AUC_inf 14.6 (men) / 17.1 (women) ug/mL*h after a single 160 mg oral dose (Darpo 2014 Results ‘Rac-sotalol pharmacokinetic profile’ and Figure 1).
• Source aliases:
  • rac-sotalol plasma concentration (Darpo 2014 Methods ‘Study outline’ and Table 1; values in ug/mL throughout, as quantitated by a validated HPLC assay with LLOQ 10 ng/mL and inter-day CV < 8.4%).
• Example models: Darpo_2014_racSotalol_QTcI.R, Darpo_2014_racSotalol_QTcF.R (PD-only linear mixed-effects E-R model for DeltaQTcI / DeltaQTcF in healthy adults after single 160 mg oral rac-sotalol; CP_RACSOTALOL_UGML is the observed or simulated plasma concentration supplied per event row).
• Notes: Specific scope; rac-sotalol-specific. The drug-specific naming follows the established CP_<drug>_<units> precedent (CP_OXY_NGML, CP_FBX_NGML, CP_LSN_NGML, CP_MORPH_NGML, CP_RIF_UM, CP_REM_NGML, CP_GLASDEGIB_NGML). The “rac” prefix is preserved from the source paper because Darpo 2014 consistently uses “rac-sotalol” to distinguish the dosed racemate (Betapace) from d-sotalol (the dextro-isomer studied separately in the SWORD trial cited in the Darpo Discussion). Darpo 2014 fit the PD model directly to observed plasma concentrations because no structural population PK model was estimated – the paper characterised rac-sotalol PK with NCA only (descriptive Cmax / Tmax / AUC). Users wishing to drive the PD model from a simulated PK source must supply their own concentration trajectory; no rac-sotalol popPK model exists in the nlmixr2lib registry.

CP_TUVUSERTIB_NGML (canonical for instantaneous tuvusertib plasma concentration as a time-varying PD driver)

• Description: Instantaneous total plasma concentration of tuvusertib (M1774; an investigational orally administered inhibitor of ataxia telangiectasia and Rad3-related [ATR] protein kinase), supplied directly as a time-varying covariate column rather than computed from a coupled PK model. Used in PD-only linear mixed-effects exposure-response models that take tuvusertib exposure as an external input to a concentration-DeltaQTcF or concentration-DeltaHR equation, and as the bath concentration of an in-vitro hERG concentration-response model.
• Units: ng/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as a linear-slope term in the concentration-DeltaQTc expression DeltaQTcF = e0 + slope * CP_TUVUSERTIB_NGML +... (Mukker 2026 Equation 1 and Table 2; the slope is reported in ms per ng/mL, so no in-model unit rescaling is needed), and as the Hill-function argument of the in-vitro hERG inhibition model. Set to 0 outside the drug-exposure window (the concentration-slope term then collapses to 0). Reference values observed: geometric mean steady-state Cmax 1410 ng/mL at 180 mg QD (the recommended dose for expansion); maximum observed concentration in the C-QTc dataset 3290 ng/mL; median 524 ng/mL, P90 1732 ng/mL, P95 2252 ng/mL (Mukker 2026 Table 3 and Table S2). Unbound steady-state Cmax at 180 mg QD is 415.95 ng/mL (Mukker 2026 Results 3.1).
• Source aliases:
  • tuvusertib plasma concentration / C (Mukker 2026 Equation 1, Table 2 and Table S1; measured by a validated LC/MS method with a lower limit of quantification of 0.5 ng/mL, reported in ng/mL throughout, no value transformation).
• Example models: Mukker_2026_tuvusertib_QTcF.R, Mukker_2026_tuvusertib_HR.R (PD-only Garnett linear mixed-effects E-R models for DeltaQTcF / DeltaHR in patients with advanced solid tumors), Mukker_2026_tuvusertib_hERG.R (in-vitro hERG tail-current inhibition, where the column carries the nominal bath concentration rather than a plasma concentration).
• Notes: Specific scope; tuvusertib-specific. The drug-specific naming follows the established CP_<drug>_<units> precedent (CP_OXY_NGML, CP_FBX_NGML, CP_GLASDEGIB_NGML, CP_RACSOTALOL_UGML). Mukker 2026 fit the C-QTc models directly to observed, time-matched plasma concentrations; the paper does not develop a population PK model of its own (the PK is reported in the companion first-in-human publications). Users wishing to drive these PD models from a simulated PK source must supply their own concentration trajectory. The same column is reused for the in-vitro hERG model because the quantity is dimensionally and semantically identical (a tuvusertib concentration in ng/mL driving a concentration-response expression); the in-vitro-versus-plasma distinction is documented in the per-model covariateData[[CP_TUVUSERTIB_NGML]]$notes. The paper reports the hERG IC50 in both molar and mass units (2.83 uM = 1048 ng/mL), which implies a molecular weight of 370.3 g/mol – consistent with tuvusertib and a useful internal unit check.

CP_MBQ167_NGML (canonical for instantaneous MBQ-167 plasma concentration as a time-varying PD driver)

• Description: Instantaneous total (bound + unbound) plasma concentration of MBQ-167, a small-molecule dual Rac/Cdc42 GTPase inhibitor in preclinical development (development code; no INN), supplied directly as a time-varying covariate column rather than computed from a coupled PK model. Used as the drug input of Simeoni tumor-growth-inhibition models whose source PBPK model cannot be ported.
• Units: ng/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – converted in-model to umol/L with the molecular weight 338.414 g/mol and entered as the argument of a sigmoid Emax kill rate kmax * C^hill / (ic50^hill + C^hill) acting on the cycling tumor cells. Set to 0 for vehicle controls. Reference values: the Reig-Lopez 2020 Simcyp-predicted typical profile after a single 10 mg/kg IP dose in a 20 g mouse has Cmax 833.31 ng/mL at 0.26 h, AUC0-12h 1549.1 ng*h/mL and terminal half-life 2.98 h (Table 3 and Discussion).
• Source aliases:
  • Cp(t) (Reig-Lopez 2020 Figure 1; Table 2 row ‘Drug input’ = ‘Total plasma concentration’; no value transformation).
• Example models: ReigLopez_2020_mbq167_her2_mouse_tgi.R, ReigLopez_2020_mbq167_tnbc_mouse_tgi.R (Simeoni TGI in HER2+ MDA-MB-435 and triple-negative MDA-MB-231 orthotopic xenografts in athymic nude mice).
• Notes: Specific scope; MBQ-167-specific. Follows the established CP_<drug>_<units> family (CP_TUVUSERTIB_NGML, CP_FBX_NGML, CP_BMS911543_UM); the name keeps the development code because no INN exists. Reig-Lopez 2020 generated the trajectory with a Simcyp Animal V19 whole-body PBPK model whose modified mouse physiology is not published, so users must supply it. With the published Hill coefficient of 0.5, the kill rate is sensitive to the low-concentration tail of each dosing interval; supply the full profile on a fine grid with linear covariate interpolation.

CP_DELAMANID_NGML (canonical for instantaneous delamanid plasma concentration as a time-varying PD driver)

• Description: Instantaneous total plasma concentration of delamanid, supplied directly as a time-varying covariate column rather than computed from a coupled PK model. Used in PD-only linear mixed-effects concentration-QTc models.
• Units: ng/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as a linear-slope term DeltaQTc = e0 + slope * CP_DELAMANID_NGML + ... with the slope in ms per ng/mL. Set to 0 to recover the drug-free prediction.
• Source aliases:
  • C (delamanid concentration) – Sasaki 2022 Methods equation and Table S4 (UPLC-MS/MS, LLOQ 1.00 ng/mL).
• Example models: Sasaki_2022_delamanid_QTc_parent.R (pediatric concentration-DeltaQTcB model; slope 0.00792 ms per ng/mL, not significant).
• Notes: Follows the CP_<drug>_<units> precedent (CP_TUVUSERTIB_NGML, CP_OXY_NGML). Can be fed from the Cc output of Sasaki_2022_delamanid.R.

CP_DM6705_NGML (canonical for instantaneous DM-6705 plasma concentration as a time-varying PD driver)

• Description: Instantaneous total plasma concentration of DM-6705, the major (albumin-mediated) metabolite of delamanid, supplied directly as a time-varying covariate column rather than computed from a coupled PK model. Used in PD-only linear mixed-effects concentration-QTc models.
• Units: ng/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as a linear-slope term DeltaQTc = e0 + slope * CP_DM6705_NGML + ... with the slope in ms per ng/mL. Set to 0 to recover the drug-free prediction.
• Source aliases:
  • C (DM-6705 concentration) – Sasaki 2022 Methods equation and Table S5 (UPLC-MS/MS, LLOQ 1.00 ng/mL).
• Example models: Sasaki_2022_delamanid_QTc_dm6705.R (pediatric concentration-DeltaQTcB model; slope 0.0613 ms per ng/mL, the significant relationship of that analysis).
• Notes: Follows the CP_<drug>_<units> precedent. The metabolite token matches the registered dm6705 metabolite suffix. Can be fed from the Cc_dm6705 output of Sasaki_2022_delamanid.R. DM-6705 is the most potent hERG inhibitor among delamanid and its metabolites (IC50 0.04 ug/mL; Sasaki 2022 Discussion).

CP_ABE_NGML (canonical for instantaneous abemaciclib plasma concentration as a time-varying target-engagement driver)

• Description: Instantaneous total (not free) plasma concentration of abemaciclib (ABE), the CDK4/6 inhibitor, supplied directly as a time-varying covariate column rather than computed from a coupled PK model. Used in QSP target-engagement models that consume an external abemaciclib exposure profile as the driver of reversible drug-CDK4 / drug-CDK6 complex formation. The free concentration that actually drives binding is computed inside model as CP_ABE_NGML * fu_abe, where fu_abe is the albumin-scaled fraction unbound.
• Units: ng/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – after conversion to nmol/L it enters the association limb of the target-engagement ODE d/dt(complex) = koff/Kd * CDK_free * C_free - koff * complex. Set to 0 for untreated / pre-dose periods and for placebo arms, which returns every occupancy to 0 and the biomarker chain to its untreated baseline. Reference values observed (Zhang 2025 Table 2, predicted column, Patnaik 2016 regimens, steady state): Cmin 144 ng/mL and Cmax 198 ng/mL at 100 mg BID; Cmin 174 / Cmax 250 at 150 mg BID; Cmin 231 / Cmax 333 at 200 mg BID.
• Source aliases:
  • Cdrug (Zhang 2025 Eq 1, defined as “the free drug concentration in plasma or the CSF”; this register entry carries the TOTAL concentration and applies the fraction unbound in-model, because a total concentration is what a PK model returns) – used in Zhang_2025_abemaciclib_qsp.R.
• Example models: Zhang_2025_abemaciclib_qsp.R (drives the shared-pool CDK4 and CDK6 occupancy limbs in plasma and in CSF; the CSF free concentration is KCSF,p * fu * CP_ABE_NGML with KCSF,p = 0.68).
• Notes: Specific scope; abemaciclib-specific, and bound to the plasma (rather than CSF or brain-interstitial) matrix – the CSF concentration is derived in-model from the published CSF-to-plasma free-concentration ratio, so it must not be supplied through this column. Follows the established CP_<drug>_<units> precedent (CP_OXY_NGML, CP_FBX_NGML, CP_LSN_NGML, CP_MORPH_NGML, CP_EIDD_NGML, CP_REM_NGML, CP_GLASDEGIB_NGML). In the source study this profile is produced by a whole-body PK-Sim 9.1 PBPK model whose system-specific parameters came from the built-in PK-Sim database with no organ ODEs, volumes or blood flows published and no .pksim5 project deposited; that layer is not reproducible from the available sources and is deliberately not extracted, following the Liang_2024_osimertinib_qsp precedent. Downstream users must therefore supply their own abemaciclib exposure trajectory and treat the resulting occupancy as conditional on it.

CP_M2_NGML (canonical for instantaneous abemaciclib metabolite M2 plasma concentration as a time-varying target-engagement driver)

• Description: Instantaneous total plasma concentration of M2 (LSN3106726), one of the three pharmacologically active abemaciclib metabolites, supplied directly as a time-varying covariate column. Used alongside CP_ABE_NGML, CP_M18_NGML and CP_M20_NGML in QSP target-engagement models where parent and metabolites compete for a shared CDK4 pool and a shared CDK6 pool. The free driving concentration is CP_M2_NGML * fu_m2 in-model.
• Units: ng/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – enters the same association limb as CP_ABE_NGML, with M2’s own Kd (1.2 nM for CDK4, 1.3 nM for CDK6), fraction unbound (0.093) and CSF-to-plasma ratio (0.14). Set to 0 to study the parent alone. Reference value observed: Zhang 2025 Introduction (citing the FDA multidiscipline review, ref 14) reports M2 as contributing 13% of the total plasma mass in vivo.
• Source aliases:
  • M2 (Zhang 2025 Table 1 column header and throughout the Methods) – used in Zhang_2025_abemaciclib_qsp.R.
• Example models: Zhang_2025_abemaciclib_qsp.R.
• Notes: Specific scope; plasma matrix only (see CP_ABE_NGML). M2 is the metabolite with the highest fraction unbound (0.093, roughly 2.4x the parent’s 0.039) but the lowest CSF-to-plasma ratio (0.14), so it contributes proportionally much more to plasma occupancy than to CSF occupancy. Named for the metabolite code the source paper uses rather than an INN because abemaciclib’s active metabolites are identified only by the M-numbering in both the paper and the FDA review.

CP_M18_NGML (canonical for instantaneous abemaciclib metabolite M18 plasma concentration as a time-varying target-engagement driver)

• Description: Instantaneous total plasma concentration of M18 (LSN3106728), one of the three pharmacologically active abemaciclib metabolites, supplied directly as a time-varying covariate column. Free driving concentration is CP_M18_NGML * fu_m18 in-model.
• Units: ng/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – enters the same association limb as CP_ABE_NGML, with M18’s own Kd (1.2 nM for CDK4, 2.7 nM for CDK6), fraction unbound (0.046) and CSF-to-plasma ratio (assumed 1.0). Set to 0 to study the parent alone. Reference value observed: Zhang 2025 Introduction (ref 14) reports M18 as contributing 5% of the total plasma mass in vivo, the smallest of the three metabolites.
• Source aliases:
  • M18 (Zhang 2025 Table 1 column header and throughout the Methods) – used in Zhang_2025_abemaciclib_qsp.R.
• Example models: Zhang_2025_abemaciclib_qsp.R.
• Notes: Specific scope; plasma matrix only (see CP_ABE_NGML). M18 is the analyte for which no CSF data existed: Zhang 2025 Methods 2.1 states that “as no data were available for M18, KCSF,p was assumed to be 1.0 for this metabolite”, which makes it the only analyte whose CSF free concentration equals its plasma free concentration and therefore the one whose CSF contribution is most uncertain.

CP_M20_NGML (canonical for instantaneous abemaciclib metabolite M20 plasma concentration as a time-varying target-engagement driver)

• Description: Instantaneous total plasma concentration of M20 (LSN3106729), the largest-contributing pharmacologically active abemaciclib metabolite, supplied directly as a time-varying covariate column. Free driving concentration is CP_M20_NGML * fu_m20 in-model.
• Units: ng/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – enters the same association limb as CP_ABE_NGML, with M20’s own Kd (1.5 nM for CDK4, 1.9 nM for CDK6), fraction unbound (0.030) and CSF-to-plasma ratio (0.91). Set to 0 to study the parent alone. Reference value observed: Zhang 2025 Introduction (ref 14) reports M20 as contributing 26% of the total plasma mass in vivo, twice M2’s share and five times M18’s.
• Source aliases:
  • M20 (Zhang 2025 Table 1 column header and throughout the Methods) – used in Zhang_2025_abemaciclib_qsp.R.
• Example models: Zhang_2025_abemaciclib_qsp.R.
• Notes: Specific scope; plasma matrix only (see CP_ABE_NGML). M20 combines the largest plasma mass share (26%) with a high CSF-to-plasma ratio (0.91) and a CDK6 affinity roughly 4x the parent’s (Kd 1.9 vs 8.2 nM), so it is the dominant contributor to CSF CDK6 occupancy in the founding model – a simulation that omits it will substantially understate the intracranial engagement the paper reports.

CP_RIB_NGML (canonical for instantaneous ribociclib plasma concentration as a time-varying target-engagement driver)

• Description: Instantaneous total (not free) plasma concentration of ribociclib (RIB), the CDK4/6 inhibitor, supplied directly as a time-varying covariate column rather than computed from a coupled PK model. Used in QSP target-engagement models that consume an external ribociclib exposure profile as the driver of reversible drug-CDK4 / drug-CDK6 complex formation. The free CSF concentration that actually drives binding is computed inside model as CP_RIB_NGML * fup * kpuu, where fup is the fraction unbound in plasma and kpuu is the CSF-to-plasma unbound concentration ratio.
• Units: ng/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – after conversion to nmol/L it enters the association limb of the target-engagement ODE d/dt(complex) = koff/Kd * CDK_free * C_free - koff * complex. Set to 0 for untreated / pre-dose periods and for placebo arms, which returns both occupancies to 0. Reference values observed (Zhang 2026 Table 2, predicted column, Infante 2016 regimens, Day 18 steady state): Cmin 524 ng/mL and Cmax 1291 ng/mL at 400 mg OD; Cmin 797 / Cmax 2237 at 600 mg OD; Cmin 1456 / Cmax 3720 at 900 mg OD.
• Source aliases:
  • C_RIB (Zhang 2026 Eq 1, defined as “the concentration of free RIB present in CSF”; this register entry carries the TOTAL PLASMA concentration and applies both the fraction unbound and the CSF-to-plasma unbound ratio in-model, because a total plasma concentration is what a PK model returns) – used in Zhang_2026_ribociclib_qsp.R.
• Example models: Zhang_2026_ribociclib_qsp.R (drives the CDK4 and CDK6 occupancy limbs in CSF; the free CSF concentration is kpuu * fup * CP_RIB_NGML with fup = 0.12 and kpuu = 1.03 at the 600 mg standard regimen).
• Notes: Specific scope; ribociclib-specific, and bound to the plasma (rather than CSF or brain-interstitial) matrix – the CSF concentration is derived in-model from the published Kp,uu, so it must not be supplied through this column. Follows the established CP_<drug>_<units> precedent (CP_ABE_NGML, CP_OXY_NGML, CP_FBX_NGML, CP_LSN_NGML, CP_MORPH_NGML, CP_EIDD_NGML, CP_REM_NGML, CP_GLASDEGIB_NGML). In the source study this profile is produced by a whole-body PK-Sim 9.1 PBPK model whose system-specific parameters came from the built-in PK-Sim database with no organ ODEs, volumes, blood flows or partition coefficients published and no .pksim5 project deposited; that layer is not reproducible from the available sources and is deliberately not extracted, following the Liang_2024_osimertinib_qsp and Zhang_2025_abemaciclib_qsp precedents. Downstream users must therefore supply their own ribociclib exposure trajectory and treat the resulting occupancy as conditional on it; Richardson_2025_ribociclib.R is an in-library plasma popPK model that can serve that role, though it originates from a different publication. Unlike the abemaciclib sibling, ribociclib has no pharmacologically active metabolites in the source model, so a single parent column is sufficient.

DOSE_OBILTOXAXIMAB_MGKG (canonical for administered obiltoxaximab dose level in mg/kg)

• Description: Per-subject obiltoxaximab dose level, in mg/kg of body weight, carried as a covariate so a dose-response model can consume it directly. Constant per subject in single-dose trigger-to-treat designs. Set to 0 for placebo subjects, which reduces a dose-response term to its untreated baseline.
• Units: mg/kg
• Type: continuous
• Scope: specific
• Reference category: n/a – drives the Emax dose-response on the logit cure fraction, Emax * DOSE_OBILTOXAXIMAB_MGKG / (ED50 + DOSE_OBILTOXAXIMAB_MGKG), with ED50 = 1.64 mg/kg in the founding example.
• Source aliases:
  • DOSE – used in Nagy_2017_obiltoxaximab_survival.R (Nagy 2017 Results, “Animal survival modeling”; dose levels 0 / 1 / 4 / 8 / 16 / 32 mg/kg i.v. across rabbit Study 2 and macaque Studies 2-5 per Table 1).
• Example models: Nagy_2017_obiltoxaximab_survival.R (Weibull cure-rate survival model; the paper reports ED50 = 1.64 mg/kg (95% CI 0.515-5.22) and ED90 = 14.8 mg/kg (95% CI 4.6-47.0) from this term, and selects 16 mg/kg as the efficacious dose because it exceeds the ED90).
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family (siblings: DOSE_EMPA_MGD, DOSE_LOR_MGD, DOSE_SEMAGLUTIDE_MG, DOSE_PHT_MGKGD). The _MGKG token records that obiltoxaximab is dosed per kilogram of body weight rather than as a flat dose, which is load-bearing here: the whole animal-to-human translation argument rests on the same 16 mg/kg per-kg dose producing comparable Cmax and superior AUC in humans versus animals. Distinct from the rxode2/nlmixr2 event column amt – this is a per-subject covariate carrying the assigned dose LEVEL for a model that has no PK compartment, whereas amt carries an administered amount at a dose event. The companion population PK model Nagy_2017_obiltoxaximab.R does not use this covariate; it takes the dose through the ordinary event table in mg.

DOSE_APRAMYCIN_MGKG (canonical for administered apramycin dose level in mg/kg)

• Description: Administered apramycin dose level, in mg/kg of body weight, carried as a covariate because the subcutaneous absorption rate constant in the preclinical apramycin population PK model is a power function of the dose LEVEL rather than of the administered amount. Constant within a dosing occasion; a study that escalates the dose between occasions must carry it as a per-record (time-varying) column.
• Units: mg/kg
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as the power term ka = ka30 * (DOSE_APRAMYCIN_MGKG / 30)^pow with ka30 = 2.17 /h and pow = -0.160 in the founding example, so the reference dose level is 30 mg/kg and absorption slows as the dose rises.
• Source aliases:
  • DOSE – used in HernandezLozano_2025_apramycin_mouse.R; the symbol is Dose in Sou 2021 Eq. 3, where the upstream mouse PK model was estimated.
• Example models: HernandezLozano_2025_apramycin_mouse.R (mouse complicated-urinary-tract-infection PKPD; dose levels 0.03-51.2 mg/kg subcutaneously twice daily across three studies, with the dose-dependent absorption inherited from Sou T et al., Clin Pharmacol Ther 2021;109:1063-73, Table 1 mouse column and Eq. 3).
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family (siblings: DOSE_OBILTOXAXIMAB_MGKG, DOSE_RTV_MGKG, DOSE_RIV_MGKG). Distinct from the rxode2/nlmixr2 event column amt, which carries the administered amount in mg (= dose level x body weight); both are needed in the founding example, because the amount drives the mass balance while the level drives the absorption rate constant. The dose-dependency was specific to the mouse in Sou 2021 (dOFV = -49.6); the rat and guinea-pig fits used a plain first-order ka, so a future apramycin model in another species may carry this column without using it, in which case it belongs in covariatesDataExcluded rather than covariateData.

DOSE_TECOVIRIMAT_MGKG (canonical for administered tecovirimat dose level in mg/kg)

• Description: Administered tecovirimat (development code ST-246) dose level, in mg/kg of body weight, carried as a covariate because three parameters of the cynomolgus-monkey population PK model – the absorption rate constant, the apparent oral clearance and the apparent central volume – are power functions of the dose LEVEL rather than of the administered amount. Constant within a dosing regimen; a study that escalates the dose between occasions must carry it as a per-record (time-varying) column.
• Units: mg/kg
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as the power terms (DOSE_TECOVIRIMAT_MGKG / 10)^theta with exponents 0.160 on Ka, 0.093 on CL/F and 0.623 on Vc/F in the founding example, so the reference dose level is 10 mg/kg.
• Source aliases:
  • dose – used in Leeds_2013_tecovirimat_cyno.R; Leeds 2013 Table 1 footnote a defines the symbol as “Dose, dose level in mg/kg”.
• Example models: Leeds_2013_tecovirimat_cyno.R (cynomolgus-monkey smallpox / monkeypox animal-rule population PK; dose levels 0.3-300 mg/kg orally once daily across six studies, of which the population PK dataset spans 0.3-30 mg/kg uninfected and 3-20 mg/kg infected).
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family (siblings: DOSE_APRAMYCIN_MGKG, DOSE_OBILTOXAXIMAB_MGKG, DOSE_RTV_MGKG, DOSE_RIV_MGKG, DOSE_BZN_MGKG, DOSE_DNDI6148_MGKG). Distinct from the rxode2/nlmixr2 event column amt, which carries the administered amount in mg (= dose level x body weight); both are needed in the founding example, because the amount drives the mass balance while the level drives Ka, CL/F and Vc/F. Structurally the closest sibling is DOSE_APRAMYCIN_MGKG, which likewise modulates a preclinical absorption rate constant by a power of the dose level, but note the sign contrast: apramycin absorption slows as the dose level rises (exponent -0.160) whereas tecovirimat absorption speeds up (exponent +0.160). The dose-level dependency was specific to the monkey in Leeds 2013 – the companion human model Leeds_2013_tecovirimat_human.R retains no dose covariate (Results “Human POP PK model development”: “The only differences between the final NHP and human POP PK models were an effect of gender on Vc/F in humans and a modest effect of dose in NHPs”), so a future tecovirimat model in another species may carry this column without using it, in which case it belongs in covariatesDataExcluded rather than covariateData. The canonical uses the INN TECOVIRIMAT rather than the development code ST246 that the source paper uses throughout; record ST-246 in covariateData[[DOSE_TECOVIRIMAT_MGKG]]$source_name when a source paper writes the code.

AUC_SAF189S (canonical for SAF-189s steady-state daily AUC)

• Description: SAF-189s area under the plasma concentration-time curve over the once-daily 24 h dosing interval at steady state, per subject, used as the exposure driver of exposure-safety models that carry no PK layer of their own. Liu 2024 computes it from the individual post hoc parameters of the companion population PK model (modellib('Liu_2024_saf189s')) at the patient’s FIRST dose level, which for a drug whose apparent clearance decays stepwise over the first few days means the AUCss of the dose the patient started on rather than of the dose they may later have been reduced to.
• Units: ng*h/mL. Document per model via covariateData[[AUC_SAF189S]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – a continuous exposure regressor entering on the NATURAL LOG scale in every founding model, so the fitted intercept absorbs the concentration unit and supplying ug*h/mL instead of ng*h/mL would shift each logit by the slope times log(1000). Reference values observed (Liu 2024): pooled median 2,233 ngh/mL, geometric mean at the recommended 160 mg once-daily dose 2,374 ngh/mL, and the exposure-safety analysis set spans 174.0-8,338.1 ng*h/mL (Figure 10 axis limits).
• Source aliases:
  • AUCss, AUC_ss, AUCSS – Liu 2024 Results and Figures 9-10 spellings.
  • AUCss estimated by first dose – the Figure 9 / Figure 10 axis label, naming the dose level the metric is computed at.
• Example models: Liu_2024_saf189s_hyperglycemia.R (odds ratio 3.521 per e-fold for any-grade hyperglycemia), Liu_2024_saf189s_hyperglycemia_grade2.R (odds ratio 7.662 per e-fold for grade >= 2 hyperglycemia), Liu_2024_saf189s_proteinuria.R (odds ratio 2.031 per e-fold for any-grade proteinuria).
• Notes: Specific scope because the column meaning is tied to SAF-189s and to the once-daily steady-state 24 h AUC convention. Member of the AUC_<DRUG> family (AUC_CARBO, AUC_GEM, AUC_GCV, AUC_PAZO, AUC_RTV, AUC_LEN, AUC_IBRU, AUC_VAREN, AUCU_VALE); AUC_LEN and AUC_IBRU are the closest structural analogues, being likewise per-interval steady-state AUCs of an oral tyrosine-kinase inhibitor carried into a downstream exposure-response fit as a data column instead of solved from a PK ODE. Liu 2024 uses a DIFFERENT exposure metric for its efficacy models than for its safety models – Cmin,ss (canonical CTROUGH) for overall response and for the Cox PFS / DOR fits, AUCss for the adverse-event fits – and the two are additionally computed at different dose levels (most prevalent dose vs first dose). Do not substitute one for the other. A future SAF-189s model using a further exposure metric (Cmax,ss, time above a threshold) should register a parallel canonical rather than overload this name.

AUCMIC_TYLO (canonical for the tylosin AUC24h/MIC PK/PD index, carried as a ratio)

• Description: Tylosin PK/PD index formed as the plasma area under the concentration-time curve over 24 h divided by the MIC of the challenge isolate (AUC24h/MIC), supplied to an antibacterial sigmoid Emax PK/PD-integration model as an already-formed RATIO. Founding member of the AUCMIC_<DRUG> family, which is explicitly distinct from the AUC_<DRUG> family (AUC_CARBO, AUC_GEM, AUC_GCV, AUC_LCM, AUC_CBZ, AUC_PAZO, AUC_RTV, AUC_VERUB, AUC_IBRU, AUC_AMPH, AUC_EMPA,…): an AUC_<DRUG> column carries an ABSOLUTE interval-integrated exposure in concentration-time units, which the model then divides by a separate mic parameter, whereas an AUCMIC_<DRUG> column carries the DIMENSIONLESS-numerator ratio itself, in units of time. Use the AUCMIC_<DRUG> form when the source paper defines and reports the index only as a ratio and does not report an MIC that would let the ratio be split into numerator and denominator; use AUC_<DRUG> plus a model mic parameter whenever the paper does report the challenge strain’s MIC, because that form lets a user re-target the model to an isolate of different susceptibility.
• Units: h (equivalently h*ug/mL divided by ug/mL). Document per-model via covariateData[[AUCMIC_TYLO]]$units if a different exposure or susceptibility unit is reported.
• Type: continuous
• Scope: specific
• Reference category: n/a – set to 0 for drug-free control records so the sigmoid term Emax * AUCMIC_TYLO^gamma / (AUCMIC_TYLO^gamma + EC50^gamma) vanishes and the predicted 24 h change in log10 CFU/mL reduces to the control value E0.
• Source aliases:
  • AUC24 h/MIC / AUC/MIC – printed names in Lee 2023 Section 2.8 (defining the sigmoid Emax driver C), Section 2.10 (the dose equation Dose = (AUC24h/MIC x MIC x Cl) / (F x fu)) and Table 3 (rows AUC 24 h /MIC for bacteriostatic activity and AUC 24 h /MIC for bactericidal activity). Lee 2023 reports index values of 0.98-1.12 h for bacteriostatic activity (E = 0) and 1.97-2.54 h for bactericidal activity (E = -3) across the four pathogen x health-state fits, with EC50 estimates of 1.06-1.33 h.
• Example models: Lee_2023_tylosin_Apleuropneumoniae_healthy.R, Lee_2023_tylosin_Apleuropneumoniae_infected.R, Lee_2023_tylosin_Pmultocida_healthy.R, Lee_2023_tylosin_Pmultocida_infected.R (Lee 2023 ex vivo inhibitory sigmoid Emax PK/PD integration for intramuscular tylosin in healthy and co-infected pigs; AUCMIC_TYLO drives the signed 24 h change in log10 CFU/mL).
• Notes: Specific scope – tylosin-specific, and tied to a 24 h interval-AUC convention. The ratio-carrying form was chosen because Lee 2023 never reports the MIC of either challenge isolate (Sections 2.4 and 3.1 state only that strains “with MIC values similar to the MIC90” were selected: BA2000013 for A. pleuropneumoniae, BA1700127 for P. multocida), so there is no sourceable MIC with which to reuse the AUC_<DRUG> plus mic shape. Substituting the collection MIC90s (16 ug/mL for A. pleuropneumoniae, 32 ug/mL for P. multocida) would bake in a value the paper’s own Table 3 contradicts three ways: T > MIC is reported as 13.08 +/- 6.15 h for A. pleuropneumoniae in healthy pigs although Cmax (5.79 ug/mL, Table 2) never reaches 16 ug/mL, so T > MIC must be 0; the AUC/MIC and Cmax/MIC rows are identical for both pathogens although their MIC90s differ; and those two rows imply mutually inconsistent MICs of 10.3 and 19.3 ug/mL. Carrying the ratio introduces no inferred value and matches the paper’s printed equation literally. The closest structural analogues in the register are AUC_AMPH and the unregistered AUC_TILM (Chen_2023_tilmicosin.R), both of which are veterinary per-24-h-interval antibacterial exposures handed to a sigmoid PK/PD-index model in place of a PK ODE, but both of which take the absolute-AUC form because their source papers do report the challenge strain MIC. A future tylosin model that does report its isolate MIC should register AUC_TYLO rather than overload this name; a future model using the Cmax/MIC index should register a parallel canonical. Family and name ratified by the maintainers (PMC10556534) alongside the Lee 2023 tylosin extraction.

AUCMIC_TILM (canonical for the tilmicosin AUC0-24/MIC PK/PD index, carried as a ratio)

• Description: Tilmicosin PK/PD index formed as the area under the concentration-time curve over a 24 h dosing interval divided by the MIC of the challenge isolate (AUC0-24/MIC), supplied to an antibacterial sigmoid Imax PK/PD-integration model as an already-formed RATIO. Member of the AUCMIC_<DRUG> family founded by AUCMIC_TYLO; see that entry for the family definition and for the rule distinguishing it from the absolute-exposure AUC_<DRUG> family. Carried in two matrices by the founding models – plasma and pulmonary interstitial fluid (PIF) – which are fitted separately and therefore ship as separate model files rather than as two columns.
• Units: h (equivalently h*ug/mL divided by ug/mL). Document per-model via covariateData[[AUCMIC_TILM]]$units if a different exposure or susceptibility unit is reported.
• Type: continuous
• Scope: specific
• Reference category: n/a – set to 0 for drug-free control records so the sigmoid term Imax * AUCMIC_TILM^gamma / (AUCMIC_TILM^gamma + IC50^gamma) vanishes and the predicted 24 h change in log10 CFU/mL reduces to the control value E0.
• Source aliases:
  • AUC24/MIC / AUC0-24/MIC / AUC/MIC – printed names in Sun 2026 eq 8 (defining the sigmoid Imax driver C), eq 9 (the dose equation dose = (AUC24h/MIC) x MIC90 x CL / (fu x F)), Table 3 (rows AUC24/MIC for bacteriostatic effect, ... bactericidal effect, ... eradication effect) and Supporting Information Table S10 (Estimation of PK-PD Parameters (AUC - MIC), one AUC/MIC column per matrix per dose per time point). Sun 2026 reports index values of 7.17 / 46.54 / 78.66 h in plasma and 12.13 / 28.48 / 51.08 h in PIF for the bacteriostatic (E = 0), bactericidal (E = -3) and eradication (E = -4) end points.
• Example models: Sun_2026_tilmicosin_pkpd_plasma.R, Sun_2026_tilmicosin_pkpd_pif.R (Sun 2026 hollow-fiber-infection-model sigmoid Imax PK/PD integration for oral tilmicosin against Pasteurella multocida in swine; AUCMIC_TILM drives the signed 24 h change in log10 CFU/mL in each matrix).
• Notes: Specific scope – tilmicosin-specific, and tied to a 24 h interval-AUC convention. This entry is the exception the AUCMIC_TYLO rule anticipates but does not cover. That rule says to use AUC_<DRUG> plus a model mic parameter “whenever the paper does report the challenge strain’s MIC”, and Sun 2026 does report one (Supporting Information “MIC and MBC determination”: P. multocida isolate ZJWZ-A, MIC 8 ug/mL). The precondition is nonetheless falsified here, because that MIC provably does not split the paper’s own index: back-solving MIC = AUC / index against the verified PBPK across Table S10 returns 0.743 ug/mL from the plasma column and 1.082 ug/mL from the PIF column – mutually inconsistent, and roughly 8-10x away from the reported 8. Adopting AUC_TILM with mic = 8 would ship a model contradicting the paper’s headline conclusion, giving a PIF index near 23 h at the recommended 40 mg/kg against a 28.48 h bactericidal target, where Table S10 reports 172.68 h and the paper concludes eradication. Carrying the ratio reproduces all six published target values exactly and introduces no inferred value. The general lesson, worth applying before the AUC_<DRUG>-vs-AUCMIC_<DRUG> rule is next invoked: a reported MIC is not automatically a splittable MIC – test whether it actually reconstructs the paper’s own index column before treating the rule’s precondition as met. Distinct from the unregistered AUC_TILM used by Chen_2023_tilmicosin.R, which carries an absolute tissue-cage exposure; a future tilmicosin model whose reported MIC does reconstruct its index should register AUC_TILM properly rather than overload this name. Name ratified by the maintainers (PMC12903901, 2026-08-28) alongside the Sun 2026 tilmicosin extraction.

AUCMIC_CEFQ (canonical for the cefquinome AUC/MIC PK/PD index, carried as a ratio)

• Description: Cefquinome PK/PD index formed as an area under the concentration-time curve divided by the MIC of the challenge isolate, supplied to an antibacterial inhibitory sigmoid Emax PK/PD-index model as an ALREADY-FORMED RATIO. Member of the AUCMIC_<DRUG> family (siblings AUCMIC_TYLO, AUCMIC_TILM); see the AUCMIC_TYLO entry for the family’s ratio-versus-split rule. Distinct from AUC_CEFQ, which carries an ABSOLUTE per-24-h-interval cefquinome serum exposure that the consuming model then divides by its own mic parameter. Both cefquinome forms are registered because the two source papers differ in whether a single strain MIC can stand as the denominator: Gao 2025 reports one serum MIC for one isolate and so uses AUC_CEFQ plus mic, whereas Wei 2026’s denominator is MIC(combine), which is not a fixed strain property (see Notes).
• Units: h (equivalently h*ug/mL divided by ug/mL). Document per-model via covariateData[[AUCMIC_CEFQ]]$units if a different exposure interval or susceptibility unit is reported; the founding example uses a 72 h cumulative AUC rather than a single dosing interval.
• Type: continuous
• Scope: specific
• Reference category: n/a – set to 0 for the untreated blank-control record so the sigmoid term vanishes and the predicted change reduces to the zero-exposure value e0.
• Source aliases:
  • AUC 72h /MIC (combine) – printed name in Wei 2026 Table 5 (tabulated for all six combination arms in both the plasma and lung blocks), Table 6 (the fitted EC50 and the 3 log10 CFU/mL target row) and Materials and methods, ‘Analysis of antibacterial effects and fitting of PK/PD’. Wei 2026 reports plasma values of 412.98-6024.97 h and lung values of 1421.61-13202.27 h across the six arms, with EC50 estimates of 441.48 h (plasma) and 1052.067 h (lung).
• Example models: Wei_2026_cefquinome_enrofloxacin_plasma_aucmic.R, Wei_2026_cefquinome_enrofloxacin_lung_aucmic.R (Wei 2026 inhibitory sigmoid Emax PK/PD integration for cefquinome plus a fixed 20 mg/kg enrofloxacin background against Klebsiella pneumoniae CLS2 in a chick pneumonia model; the index drives the control-referenced 72 h change in lung log10 CFU/mL).
• Notes: Specific scope – cefquinome-specific, and in the founding example tied to a 72 h cumulative-AUC convention rather than the 24 h interval convention of AUC_CEFQ, AUCMIC_TYLO and AUCMIC_TILM. The ratio-carrying form was chosen because Wei 2026’s denominator is MIC(combine) – the MIC of cefquinome measured by CAMH broth microdilution in the presence of the average free enrofloxacin concentration achieved over 72 h in that matrix – which is a property of the enrofloxacin co-exposure and dosing schedule rather than of the isolate, and which is not constant across the arms Wei 2026 fitted. Reconstructing Table 5 from the non-compartmental Tables 3 and 4 recovers the denominator exactly: plasma MIC(combine) is 0.031 ug/mL for all six arms (matching the paper’s own statement that it was 0.031 for both dosing regimens), but lung MIC(combine) is 0.0168 ug/mL for the three single-dose arms and 0.0085 ug/mL for the split-dose arms. Wei 2026 prints the lung pair rounded as 0.017 and 0.008, and neither printed value reproduces its own Table 5. Splitting the ratio into an absolute AUC_CEFQ divided by one fixed mic would therefore bake in a denominator that is wrong for half the lung arms.

CMAXMIC_CEFQ (canonical for the cefquinome Cmax/MIC PK/PD index, carried as a ratio)

• Description: Cefquinome PK/PD index formed as the peak concentration divided by the MIC of the challenge isolate, supplied to an antibacterial inhibitory sigmoid Emax PK/PD-index model as an ALREADY-FORMED RATIO. Founding member of the CMAXMIC_<DRUG> family, the peak-concentration counterpart of AUCMIC_<DRUG> (AUCMIC_TYLO, AUCMIC_TILM, AUCMIC_CEFQ) and FTMIC_<DRUG>; together the three cover the standard antimicrobial PK/PD-index triad. As with AUCMIC_<DRUG>, the ratio-carrying form is used when the source paper’s denominator cannot be sourced as a single fixed strain MIC; use the CMAX_<DRUG> family (CMAX_APTM, CMAX_M1) plus a model mic parameter whenever it can, because that form lets a user re-target the model to an isolate of different susceptibility.
• Units: unitless (a concentration divided by a concentration). Document per-model via covariateData[[CMAXMIC_CEFQ]]$units if a different susceptibility convention is reported.
• Type: continuous
• Scope: specific
• Reference category: n/a – set to 0 for the untreated blank-control record so the sigmoid term vanishes and the predicted change reduces to the zero-exposure value e0.
• Source aliases:
  • C max /MIC (combine) – printed name in Wei 2026 Table 5 (both the plasma and lung blocks) and Table 6. Wei 2026 reports plasma values of 77.74-687.74 and lung values of 230.95-1721.18 across the six combination arms, with EC50 estimates of 57.921 (plasma) and 555.674 (lung).
• Example models: Wei_2026_cefquinome_enrofloxacin_plasma_cmaxmic.R, Wei_2026_cefquinome_enrofloxacin_lung_cmaxmic.R (Wei 2026 inhibitory sigmoid Emax PK/PD integration for cefquinome plus a fixed 20 mg/kg enrofloxacin background against Klebsiella pneumoniae CLS2 in a chick pneumonia model).
• Notes: Specific scope – cefquinome-specific. The denominator is MIC(combine) and the ratio-carrying rationale is identical to AUCMIC_CEFQ; see that entry’s Notes for the arm-by-arm reconstruction showing that lung MIC(combine) takes two different values (0.0168 ug/mL single-dose, 0.0085 ug/mL split-dose) within the same fitted model. Wei 2026 ranked Cmax/MIC last of the three indices in plasma (R^2 = 0.803 against 0.867 for AUC72h/MIC and 0.953 for %T>MIC) and reported no 3 log10 CFU/mL target for it, concluding that cefquinome behaves as a time-dependent rather than a concentration-dependent agent in this combination; the fits are packaged because Table 6 reports them in full alongside the others.

FTMIC_CEFQ (canonical for the cefquinome percentage-of-interval-above-MIC PK/PD index)

• Description: Percentage of a 24 h dosing interval during which the cefquinome concentration exceeds the MIC of the challenge isolate (%T > MIC), supplied to an antibacterial inhibitory sigmoid Emax PK/PD-index model. Unlike AUCMIC_<DRUG> and CMAXMIC_<DRUG>, this index has no ratio to split: it is a duration statistic already expressed relative to the MIC, so it is always carried whole and there is no <X>_<DRUG> plus mic alternative form. Follows the FTMIC_<DRUG> naming already in use for FTMIC_TZP in Wenker_2024_piperacillin_tazobactam.R, and is likewise on the PERCENT scale (0-100), not the 0-1 fraction scale – the fitted EC50 is on the same scale, so a fraction-scaled column would silently mis-scale the sigmoid.
• Units: percent (0-100). Document per-model via covariateData[[FTMIC_CEFQ]]$units if a source reports the fraction rather than the percentage.
• Type: continuous
• Scope: specific
• Reference category: n/a – set to 0 for the untreated blank-control record so the sigmoid term vanishes and the predicted change reduces to the zero-exposure value e0.
• Source aliases:
  • %T > MIC (combine) – printed name in Wei 2026 Table 5 (both the plasma and lung blocks), Table 6 and Materials and methods, ‘Analysis of antibacterial effects and fitting of PK/PD’, which defines it as “the proportion of time within 24 h that the drug concentration exceeded MIC (combine)”, computed in WinNonlin. Wei 2026 reports plasma values of 20.13-63.89% and lung values of 29.00-99.55% across the six combination arms, with EC50 estimates of 35.748% (plasma) and 50.059% (lung).
• Example models: Wei_2026_cefquinome_enrofloxacin_plasma_tmic.R, Wei_2026_cefquinome_enrofloxacin_lung_tmic.R (Wei 2026 inhibitory sigmoid Emax PK/PD integration for cefquinome plus a fixed 20 mg/kg enrofloxacin background against Klebsiella pneumoniae CLS2 in a chick pneumonia model).
• Notes: Specific scope – cefquinome-specific. The threshold is MIC(combine), not the cefquinome monotherapy MIC; Wei 2026 compared the two directly and found that the monotherapy-MIC version of %T > MIC placed the bactericidal ENR20 + CEQ10-split arm below the conventional 40% cephalosporin threshold, which is why all its fitting uses MIC(combine). See the AUCMIC_CEFQ Notes for the arm-by-arm MIC(combine) reconstruction. This is the index Wei 2026 selected as its primary result: the plasma fit has the highest R^2 (0.953) and lowest AIC (14.695) of the six in Table 6, and its 3 log10 CFU/mL target of 37.062% is the paper’s headline PK/PD breakpoint.

CP_MILTEFOSINE_UGML (canonical for instantaneous miltefosine plasma concentration as a time-varying PD driver)

• Description: Instantaneous plasma concentration of miltefosine (hexadecylphosphocholine; oral alkylphosphocholine antileishmanial) supplied directly as a time-varying covariate column rather than computed from a coupled PK model. Used in PD-only antiparasitic kill-rate models that consume an upstream miltefosine population PK trajectory as input because the source analysis was a sequential PK-then-PD fit with the PK model fixed from a previously published popPK analysis.
• Units: ug/mL (= mg/L). NOTE the unit is NOT the one Table 2 prints. Verrest 2024 Table 2 gives lambda_MF in ug^-1 * L * h^-1, which would put this column in ug/L; that reading is falsified by the paper’s own Table 4 (see Notes), and the self-consistent unit is ug/mL, matching the UGML token already used by CP_RITUXIMAB_UGML and CP_RACSOTALOL_UGML.
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as a linear term in the drug-induced parasite-elimination rate, kdrug = kdrug + slope_miltefosine * CP_MILTEFOSINE_UGML, where slope_miltefosine (the paper’s lambda_MF, 0.0010) has units of mL/(ug*h). Set to 0 for study arms that do not receive miltefosine and outside the drug-exposure window. Reference values observed: miltefosine accumulates slowly over a 28-day 2.5 mg/kg/day course and, because of its long half-life, maintains above-target exposure for roughly 50 days from start of treatment (Verrest 2024 Fig 4 and Results 3.2.1).
• Source aliases:
  • individual predicted miltefosine concentration (Verrest 2024 Methods 2.4.2 and Fig 2 C drug; derived from the Palic 2020 non-linear paediatric-VL population PK model, J Antimicrob Chemother 75:3260-3268, and supplied to NONMEM as a data column).
• Example models: Verrest_2024_leishmania.R (linear parasite-kill term lambda_MF * CP_MILTEFOSINE_UGML in the Leishmania blood-parasite-dynamics PK-PD model; the driver covers the AmB+MF10D, MFC28D, and MFA28D arms).
• Notes: Specific scope; miltefosine-specific. The drug-specific naming follows the established CP_<drug>_<units> precedent (CP_OXY_NGML, CP_FBX_NGML, CP_MORPH_NGML, CP_GLASDEGIB_NGML, CP_RACSOTALOL_UGML) and the standing instruction on CP_MGL to register a drug-specific canonical rather than overload the generic name. Unlike most members of the family, an nlmixr2lib miltefosine popPK model DOES exist – modellib('Dorlo_2017_miltefosine'), fit to the same LEAP0208 trial (NCT01067443) that supplies three of Verrest 2024’s five arms – so users can generate this trajectory from the registry instead of supplying observed data; note that Verrest 2024 itself used the Palic 2020 model, which is a different (non-linear, paediatric) miltefosine popPK analysis. Unit provenance. Table 2 prints lambda_MF in ug^-1 * L * h^-1, implying a ug/L driver. Applying that literally makes kdrug about 5.2 /h on the 2.5 mg/kg/day miltefosine regimen (mean plasma concentration over Days 0-10 is roughly 5 ug/mL = 5180 ug/L), which drives the parasite load onto its 1 parasite/mL floor within hours. Verrest 2024 Table 4 instead reports the MFC28D arm declining only from 4583 parasites/mL on Day 1 to 2027 on Day 10, an implied kdrug of 0.0075 /h; the ug/mL reading gives 0.0010 * 5.18 = 0.0052 /h and reproduces it. The printed unit is therefore off by a factor of 1000 and this register entry records the corrected one. Distinct from CP_FEXINIDAZOLE_M1M2_UGML, the sibling driver registered alongside it.

CP_FEXINIDAZOLE_M1M2_UGML (canonical for instantaneous summed fexinidazole sulfoxide + sulfone whole-blood concentration as a time-varying PD driver)

• Description: Instantaneous whole-blood concentration of the SUM of the two active fexinidazole metabolites – fexinidazole sulfoxide (M1) and fexinidazole sulfone (M2) – supplied directly as a time-varying covariate column rather than computed from a coupled PK model. The parent nitroimidazole fexinidazole is NOT part of the driver: Verrest 2024 Fig 2 defines the drug concentration for the fexinidazole arm as “the sum of M1 and M2”, the two metabolites that carry the antiparasitic activity. Used in PD-only antiparasitic kill-rate models that consume an upstream fexinidazole/M1/M2 population PK trajectory as input.
• Units: ug/mL (= mg/L). As for CP_MILTEFOSINE_UGML, this is NOT the unit Table 2 prints for lambda_fexi; see Notes.
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as a linear term in the drug-induced parasite-elimination rate, kdrug = kdrug + slope_fexinidazole * CP_FEXINIDAZOLE_M1M2_UGML, where slope_fexinidazole (the paper’s lambda_fexi, 0.0011) has units of mL/(ug*h). Set to 0 for study arms that do not receive fexinidazole. Reference values observed: the Fexi10D regimen (1800 mg/day for 4 days then 1200 mg/day for 6 days) produced persistent drug exposure of about 11 days from start of treatment (Verrest 2024 Fig 4A and Results 3.2.1).
• Source aliases:
  • sum of M1 and M2 (Verrest 2024 Fig 2 legend and Methods 2.4.2; measured in dried blood spots by the assay of Tarral 2014 Clin Pharmacokinet 53:565-580 and predicted per subject by a fexinidazole/M1/M2 population PK model held in an unpublished internal report).
• Example models: Verrest_2024_leishmania.R (linear parasite-kill term lambda_fexi * CP_FEXINIDAZOLE_M1M2_UGML in the Leishmania blood-parasite-dynamics PK-PD model; the driver covers the Fexi10D arm).
• Notes: Specific scope; bound both to fexinidazole and to the summed-active-metabolite definition of the driver. The metabolite-sum semantics are load-bearing – a future analysis that drives a PD model from parent fexinidazole, or from M1 and M2 as two separate slots, must register a parallel canonical rather than overload this name. The upstream PK model is an unpublished internal report with no parameter values reported anywhere in Verrest 2024 or its S1 File, so no fexinidazole popPK model exists in the nlmixr2lib registry and the user must supply the trajectory. Whole blood rather than plasma is the source matrix (dried blood spots), an exception to the CP_ family’s usual plasma semantics that is recorded here rather than by a distinct prefix, since the covariate’s role – an externally supplied instantaneous systemic concentration driving a PD model – is identical. Unit provenance. The same 1000-fold discrepancy as CP_MILTEFOSINE_UGML. Inverting the Fexi10D arm’s published decline (Table 4: 3732 parasites/mL on Day 1 to 105 on Day 10, with kGR = 0.0037 /h) gives an implied kdrug of 0.020 /h and hence a driver of 0.020 / 0.0011 = 18 in this column’s units. Whole-blood M1 + M2 after 1800 mg/day fexinidazole is in the tens of ug/mL, not the tens of ug/L, so the column is in ug/mL. Distinct from CP_MILTEFOSINE_UGML, the sibling driver registered alongside it.

CP_FRUQUINTINIB_NGML (canonical for instantaneous parent fruquintinib plasma concentration as a time-varying PD driver)

• Description: Instantaneous total plasma concentration of parent fruquintinib – the selective inhibitor of vascular endothelial growth factor receptors 1, 2 and 3 – at the time of each PD observation, supplied directly as a time-varying covariate column rather than computed from a coupled PK model. Used in concentration-QTc (C-QTc) exposure-response models where the ECG endpoint is paired with a plasma sample drawn at the same nominal time point.
• Units: ng/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as the linear term (slope + etaslope) * CP_FRUQUINTINIB_NGML in a linear mixed-effects DeltaQTc model whose slope is reported directly in ms per ng/mL, so no in-model unit rescaling is required. Set to 0 for placebo records and outside the drug-exposure window; the concentration-slope term then collapses to 0. Reference values observed: geometric mean steady-state Cmax approximately 290 ng/mL total (unbound approximately 13.6 ng/mL at 95.3% protein binding) after fruquintinib 5 mg once daily on days 1-21 of a 28-day cycle, and 580 ng/mL at twice the geometric mean steady-state Cmax (Zhou 2025 Discussion).
• Source aliases:
  • CONC1 – the term name in the Zhou 2025 supplementary model-development logs (Supplementary Tables 1 and 2, glossed “CONC1, fruquintinib concentration”).
  • Fruquintinib Conc. – the parameter-table row label in Zhou 2025 Supplementary Table 5.
• Example models: Zhou_2025_fruquintinib_QTcP_parent.R (the paper’s additional C-QTc analysis driven by parent concentration; slope 0.00778 ms per ng/mL, 95% CI -0.00247 to 0.018, P = 0.1377 – a genuine negative result, so the covariate is present and load-bearing structurally even though its effect is not statistically significant).
• Notes: Specific scope; plasma matrix. Distinct from CP_FRUQUINTINIB_M11_NGML, the sibling metabolite driver registered alongside it – the two are separate analytes measured from the same sample, and Zhou 2025 fits them in three separate model families (parent only, metabolite only, and both together), so a model must never substitute one for the other. Zhou 2025 reports that only 16 of 936 (1.7%) parent concentrations in the active arm were below the limit of quantification, versus 326 of 936 (34.8%) for M11; BLQ values were set to zero and retained. No fruquintinib population PK model exists in the nlmixr2lib registry, so a downstream user must supply the trajectory (observed samples, or an upstream popPK source).

CP_FRUQUINTINIB_M11_NGML (canonical for instantaneous fruquintinib metabolite M11 plasma concentration as a time-varying PD driver)

• Description: Instantaneous total plasma concentration of M11, the major metabolite of fruquintinib, at the time of each PD observation, supplied directly as a time-varying covariate column rather than computed from a coupled PK model. This is the driver of the FINAL model of the Zhou 2025 fruquintinib cardiac-safety analysis: of the three C-QTc model families the authors fitted (parent only, metabolite only, both), the metabolite-only family had the lowest AIC.
• Units: ng/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as the linear term (slope + etaslope) * CP_FRUQUINTINIB_M11_NGML in a linear mixed-effects DeltaQTc model whose slope is reported directly in ms per ng/mL, so no in-model unit rescaling is required. Set to 0 for placebo records and outside the drug-exposure window. Reference values observed: geometric mean steady-state Cmax 77 ng/mL after fruquintinib 5 mg once daily, and 154 ng/mL at twice that value – the two scenarios tabulated in Zhou 2025 Tables 3 and 4. The QTcP model predicts the upper bound of the 90% CI of mean DeltaDeltaQTcP to exceed 10 ms at 262 ng/mL (3.4-fold the geometric mean steady-state Cmax); the QTcF model predicts it not to exceed 10 ms up to 177 ng/mL (2.3-fold).
• Source aliases:
  • CONC2 – the term name in the Zhou 2025 supplementary model-development logs (Supplementary Tables 2 and 3, glossed “CONC2, M11 concentration”).
  • M11 conc. / M11 Conc. – the parameter-table row labels in Zhou 2025 Table 2 and Supplementary Table 4.
• Example models: Zhou_2025_fruquintinib_QTcP_M11.R (the paper’s final model, on the population-based-corrected endpoint; slope 0.0339 ms per ng/mL, P = 0.0212), Zhou_2025_fruquintinib_QTcF_M11.R (the supportive analysis on the Fridericia-corrected endpoint; slope 0.0477 ms per ng/mL, P = 0.0035).
• Notes: Specific scope; plasma matrix. Parent-qualified by design. The CP_<analyte>_<units> family already contains both a bare-M-code form (CP_M2_NGML, CP_M18_NGML, CP_M20_NGML – all abemaciclib metabolites, with the parent named only in the entry Description) and a parent-qualified form (CP_FEXINIDAZOLE_M1M2_UGML), so precedent alone did not settle the naming. The parent-qualified form was chosen by the maintainers (PMC12555101, 2026-08-28) because metabolite M-codes are drug-local identifiers: a bare CP_M11_NGML would be a latent collision the first time another drug’s M11 is extracted, whereas qualifying by parent drug makes the name collision-proof. Prefer this form for new metabolite drivers; the three bare abemaciclib M-codes remain canonical for the models that already use them. Distinct from CP_FRUQUINTINIB_NGML, the sibling parent driver registered alongside it – Zhou 2025 fits parent and metabolite in separate model families, so the two must never be substituted for one another. 326 of 936 (34.8%) M11 concentrations in the active arm were below the limit of quantification and were set to zero and retained, which is why the authors also report the parent-driven cross-check. No fruquintinib or M11 population PK model exists in the nlmixr2lib registry, so a downstream user must supply the trajectory.

CP_IVOSIDENIB_NGML (canonical for instantaneous ivosidenib plasma concentration as a time-varying PD driver)

• Description: Instantaneous plasma concentration of ivosidenib (AG-120, an inhibitor of mutant isocitrate dehydrogenase 1) at the time of each PD observation, supplied directly as a time-varying covariate column rather than computed from a coupled PK model. Used in concentration-QTc models where the ECG endpoint is paired with a time-matched plasma sample.
• Units: ng/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as the linear term slope * CP_IVOSIDENIB_NGML of a direct-effect DeltaQTcF model whose slope is reported in ms per ng/mL, so no in-model unit rescaling is required. Set to 0 before the first dose. Reference values observed: geometric-mean Cmax 6551 ng/mL at 500 mg once daily (Jiang 2021 Results ‘QTc analysis’).
• Source aliases:
  • Ivosidenib concentration – the x-axis label of Jiang 2021 Figure 5 (time-matched plasma concentration, ng/mL).
• Example models: Jiang_2021_ivosidenib_QTcF.R (typical-value linear C-QTc model, slope 0.00258 ms per ng/mL, pooled across AG120-C-001, AG120-C-002 and AG120-C-004; the companion popPK model Jiang_2021_ivosidenib.R can generate the concentration trajectory).
• Notes: Specific scope; plasma matrix, total (bound plus unbound) concentration. Follows the drug-specific CP_<drug>_<units> pattern set out in the CP_MGL entry.

CP_AMILORIDE_NGML (canonical for instantaneous amiloride plasma concentration as a time-varying PD driver)

• Description: Instantaneous plasma concentration of amiloride (a potassium-sparing diuretic) supplied directly as a time-varying covariate column rather than computed from a coupled PK model. Used in cardiovascular-systems (CVS) hemodynamic models that take diuretic exposure as an external input to an Emax inhibition of the stroke-volume production rate.
• Units: ng/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as the Hill term of an Emax inhibition expression EFF = -emax * CP_AMILORIDE_NGML / (ec50 + CP_AMILORIDE_NGML) inside the SV production term (Snelder 2014 Equations 6 and 7). Set to 0 outside the exposure window and for every animal not receiving amiloride. Reference values observed: the fitted EC50 in conscious spontaneously hypertensive rats is 245 ng/mL (Snelder 2014 Table 5) after a single 10 mg/kg p.o. dose.
• Source aliases:
  • C (Snelder 2014 Equations 6 and 7, “the drug concentration in plasma”; values in ng/mL because Table 5 reports EC50 in ng/mL).
• Example models: Snelder_2014_cardiovascular_rat.R (Emax inhibition of the SV production rate; Emax fixed at 1 and EC50 = 245 ng/mL).
• Notes: Specific scope; amiloride-specific. Follows the established CP_<drug>_<units> precedent, whose CP_MGL entry explicitly directs new drivers to register a drug-specific member rather than overload the generic name. Snelder 2014 Table 2 took the amiloride concentration-time profile from the two-compartment-with-liver-compartment model of Segre et al. 1998 (Wistar rats) and publishes NO PK parameter values for it; no amiloride PK model exists in the nlmixr2lib registry either, so downstream users must supply this column externally. Snelder 2014 Table 3 Assumption 3 states that the PK was assumed not to differ between rat strains. Amiloride carries the least precise EC50 in the paper (RSE 25.1%, 95% CI 125-365 ng/mL), which is unsurprising given it was studied at a single dose level in three animals.

CP_ATROPINE_NGML (canonical for instantaneous atropine plasma concentration as a time-varying PD driver)

• Description: Instantaneous plasma concentration of atropine (a non-selective muscarinic acetylcholine-receptor antagonist) supplied directly as a time-varying covariate column rather than computed from a coupled PK model. Used in cardiovascular-systems (CVS) hemodynamic models that take vagolytic exposure as an external input to a linear STIMULATION of the heart-rate production rate.
• Units: ng/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as the linear term EFF = SL * CP_ATROPINE_NGML inside the HR production term (Snelder 2014 Equations 6 and 7). Set to 0 outside the exposure window. Reference values observed: the fitted slope in conscious spontaneously hypertensive rats is 0.00149 per ng/mL (Snelder 2014 Table 5) after a single 10 mg/kg p.o. dose.
• Source aliases:
  • C (Snelder 2014 Equations 6 and 7; values in ng/mL).
• Example models: Snelder_2014_cardiovascular_rat.R (linear stimulation of the HR production rate; SL = 0.00149 per ng/mL).
• Notes: Specific scope; atropine-specific. Follows the established CP_<drug>_<units> precedent. The only STIMULATING driver in the Snelder CVS family – every other CP_<drug>_NGML column registered for these models inhibits a production rate, so the sign of the effect must be read from the model file rather than assumed. Snelder 2014 Table 2 based the profile on the two-compartment model of Perlstein et al. 2002 (Sabra rats), which was dosed i.v. whereas Snelder 2014 dosed p.o.; only the absorption rate constant was re-estimated (Ka = 1.17 1/h, RSE 59.9%, 95% CI -0.204 to 2.54 – the single imprecise estimate in Table 5, which the authors accepted after showing the system parameters were insensitive to it). The disposition parameters are not published in Snelder 2014, so Ka alone does not reconstruct the profile and this column must be supplied externally.

CP_AMLODIPINE_NGML (canonical for instantaneous amlodipine plasma concentration as a time-varying PD driver)

• Description: Instantaneous plasma concentration of amlodipine (a dihydropyridine calcium-channel blocker) supplied directly as a time-varying covariate column rather than computed from a coupled PK model. Used in cardiovascular-systems (CVS) hemodynamic models that take antihypertensive exposure as an external input to an Emax inhibition of the total-peripheral-resistance production rate.
• Units: ng/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as the Hill term of an Emax inhibition expression EFF = -emax * CP_AMLODIPINE_NGML / (ec50 + CP_AMLODIPINE_NGML) inside the TPR production term (Snelder 2013 Equations 4 and 6; Snelder 2014 Equations 6 and 7). Set to 0 outside the exposure window. Reference values observed: the fitted potency in conscious spontaneously hypertensive rats is IC50 = 185 ng/mL in the Snelder 2013 basic CVS model and EC50 = 82.8 ng/mL in the Snelder 2014 extended CVS model, over doses of 0.3-10 mg/kg p.o.
• Source aliases:
  • C (Snelder 2013 Equation 6 and Snelder 2014 Equations 6 and 7; values in ng/mL).
• Example models: Snelder_2013_cardiovascular_rat.R (Emax inhibition of the TPR production rate; Emax fixed at 1 and IC50 = 185 ng/mL), Snelder_2014_cardiovascular_rat.R (Emax inhibition of the TPR production rate; Emax fixed at 1 and EC50 = 82.8 ng/mL).
• Notes: Specific scope; amlodipine-specific. Follows the established CP_<drug>_<units> precedent. Both Snelder papers built the amlodipine profile from a one-compartment model whose absorption rate constant was back-derived in Berkeley Madonna from the half-life, Vd, F and Tmax reported by Stopher et al. 1988 (Sprague-Dawley rats); neither publishes PK parameter values and no amlodipine PK model exists in the nlmixr2lib registry, so downstream users must supply this column externally. Amlodipine is load-bearing in both papers: in Snelder 2013 it is the one compound whose omission materially changes a system parameter (FB1 moves from 0.00379 to 0.00454 1/mmHg; Figure 5), and in Snelder 2014 it is the paradigm compound for the HR-and-MAP-only site-of-action analysis, where re-fitting to HR and MAP alone with the system parameters fixed recovered EC50 = 84.9 ng/mL (95% CI 75.4-94.4), indistinguishable from the full-data value.

CP_ENALAPRIL_NGML (canonical for instantaneous enalapril plasma concentration as a time-varying PD driver)

• Description: Instantaneous plasma concentration of enalapril (an angiotensin-converting-enzyme inhibitor) supplied directly as a time-varying covariate column rather than computed from a coupled PK model. Used in cardiovascular-systems (CVS) hemodynamic models that take antihypertensive exposure as an external input to an Emax inhibition of a hemodynamic production rate.
• Units: ng/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – in Snelder_2013_cardiovascular_rat.R it enters an Emax inhibition of the TPR production term directly (EFF = emax * CP_ENALAPRIL_NGML / (ic50 + CP_ENALAPRIL_NGML), Snelder 2013 Equations 4 and 6). In Snelder_2014_cardiovascular_rat.R it instead drives an effect compartment (dCe/dt = ke0 * (CP_ENALAPRIL_NGML - Ce), ke0 = 0.163 1/h, Snelder 2014 Equation 8), and the Emax inhibition of BOTH the TPR and the SV production rates is a function of that effect-site concentration rather than of the column itself. Set to 0 outside the exposure window and for every animal not receiving enalapril. Reference values observed: IC50 = 2410 ng/mL (Snelder 2013 Table 6) and EC50 = 1200 ng/mL (Snelder 2014 Table 5).
• Source aliases:
  • C (Snelder 2013 Equation 6 and Snelder 2014 Equations 6-8; values in ng/mL because both parameter tables report potency in ng/mL).
• Example models: Snelder_2013_cardiovascular_rat.R (Emax inhibition of the TPR production rate in the drug-independent CVS model; Emax fixed at 1 and IC50 = 2410 ng/mL), Snelder_2014_cardiovascular_rat.R (effect-compartment-delayed Emax inhibition of BOTH TPR and SV production with one shared EC50 = 1200 ng/mL).
• Notes: Specific scope; enalapril-specific. Follows the established CP_<drug>_<units> precedent. Both Snelder papers obtained the enalapril concentration-time profile from a two-compartment model with Michaelis-Menten elimination refitted in NONMEM to data read off Lin et al. 1988 (and, in 2014, also Li et al. 2007) in Sprague-Dawley rats, and publish NO CL / V / Vmax / Km value for it; no enalapril PK model exists in the nlmixr2lib registry either, so downstream users must supply this column externally. Both papers state that the PK was assumed not to differ between rat strains. Note the modelling difference between the two example models: Snelder 2013 measured no cardiac output and so resolved only the primary TPR effect, whereas Snelder 2014 measured CO and resolved a second, SV-mediated effect with the same EC50 (the separately estimated values had overlapping confidence intervals) plus a 4.3 h effect-compartment delay attributed to the renin-angiotensin-aldosterone cascade. A dataset column feeding the 2014 model is therefore consumed through one more state than the same column feeding the 2013 model.

CP_FASUDIL_NGML (canonical for instantaneous fasudil plasma concentration as a time-varying PD driver)

• Description: Instantaneous plasma concentration of fasudil (a Rho-kinase inhibitor) supplied directly as a time-varying covariate column rather than computed from a coupled PK model. Used in cardiovascular-systems (CVS) hemodynamic models that take vasodilator exposure as an external input to an Emax inhibition of the total-peripheral-resistance production rate.
• Units: ng/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as the Hill term of an Emax inhibition expression EFF = -emax * CP_FASUDIL_NGML / (ec50 + CP_FASUDIL_NGML) inside the TPR production term (Snelder 2013 Equations 4 and 6; Snelder 2014 Equations 6 and 7). Set to 0 outside the exposure window. Reference values observed: the two example models disagree by roughly 1900-fold – IC50 = 321 ng/mL in Snelder 2013 Table 6 versus EC50 = 0.172 ng/mL in Snelder 2014 Table 5, over doses of 3-30 mg/kg p.o. in both.
• Source aliases:
  • C (Snelder 2013 Equation 6 and Snelder 2014 Equations 6 and 7; values in ng/mL).
• Example models: Snelder_2013_cardiovascular_rat.R (Emax inhibition of the TPR production rate; Emax fixed at 1 and IC50 = 321 ng/mL), Snelder_2014_cardiovascular_rat.R (Emax inhibition of the TPR production rate; Emax fixed at 1 and EC50 = 0.172 ng/mL).
• Notes: Specific scope; fasudil-specific. Follows the established CP_<drug>_<units> precedent. Both Snelder papers built the fasudil profile from a one-compartment model whose absorption rate constant and lag time were back-derived in Berkeley Madonna from the half-life, AUC and Cmax reported by Ikegaki et al. 2001 (Wistar-Kyoto rats); neither publishes PK parameter values and no fasudil PK model exists in the nlmixr2lib registry, so downstream users must supply this column externally. Scale warning: the ~1900-fold potency discrepancy between the two example models is a genuine between-publication difference, not a transcription artefact – each value is internally consistent with its own reported RSE and 95% confidence interval (321 with CV 40.5% in 2013; 0.172 with RSE 18.4% and CI 0.110-0.234 in 2014) – and neither paper reconciles it. Because the driver is a user-supplied column rather than a coupled PK model, a fasudil concentration-time profile constructed to be consistent with one paper’s potency scale must not be reused with the other’s.

CP_HCTZ_NGML (canonical for instantaneous hydrochlorothiazide plasma concentration as a time-varying PD driver)

• Description: Instantaneous plasma concentration of hydrochlorothiazide (HCTZ, a thiazide diuretic) supplied directly as a time-varying covariate column rather than computed from a coupled PK model. Used in cardiovascular-systems (CVS) hemodynamic models that take diuretic exposure as an external input to an Emax inhibition of a cardiac-output or stroke-volume production rate.
• Units: ng/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as the Hill term of an Emax inhibition expression EFF = -emax * CP_HCTZ_NGML / (ec50 + CP_HCTZ_NGML) inside the CO production term in Snelder 2013 (Equations 4 and 6) and inside the SV production term in Snelder 2014 (Equations 6 and 7), which is the same site of action expressed at the finer resolution the 2014 model provides. Set to 0 outside the exposure window. Reference values observed: IC50 = 12 300 ng/mL (Snelder 2013 Table 6) and EC50 = 28 900 ng/mL (Snelder 2014 Table 5).
• Source aliases:
  • C (Snelder 2013 Equation 6 and Snelder 2014 Equations 6 and 7; values in ng/mL).
• Example models: Snelder_2013_cardiovascular_rat.R (Emax inhibition of the CO production rate; Emax fixed at 1 and IC50 = 12 300 ng/mL), Snelder_2014_cardiovascular_rat.R (Emax inhibition of the SV production rate; Emax fixed at 1 and EC50 = 28 900 ng/mL).
• Notes: Specific scope; hydrochlorothiazide-specific. The abbreviated stem HCTZ is the source papers’ own abbreviation and is the universally recognised short form. Follows the established CP_<drug>_<units> precedent. Both Snelder papers built the HCTZ profile from a one-compartment model taking Ke, Ka and Vd from Asdaq & Inamdar 2009 (Wistar-Kyoto rats) with the bioavailability term back-calculated from the reported AUC; neither publishes PK parameter values and no hydrochlorothiazide PK model exists in the nlmixr2lib registry, so downstream users must supply this column externally. Snelder 2013 reported that the 1 mg/kg HCTZ response was under-predicted and attributed it to the Emax = 1 constraint (the 0.1-3 mg/kg range had not spanned no-effect to maximal effect) rather than to the system model; Snelder 2014 addressed this directly by adding a second occasion at 10 and 30 mg/kg, which is why the two EC50 estimates differ by rather more than the other shared compounds.

CP_PRAZOSIN_NGML (canonical for instantaneous prazosin plasma concentration as a time-varying PD driver)

• Description: Instantaneous plasma concentration of prazosin (a selective alpha-1 adrenoceptor antagonist) supplied directly as a time-varying covariate column rather than computed from a coupled PK model. Used in cardiovascular-systems (CVS) hemodynamic models that take antihypertensive exposure as an external input to an inhibition of the total-peripheral-resistance production rate.
• Units: ng/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – in Snelder_2013_cardiovascular_rat.R it enters an Emax inhibition of the TPR production term (EFF = emax * CP_PRAZOSIN_NGML / (ic50 + CP_PRAZOSIN_NGML), Equations 4 and 6). In Snelder_2014_cardiovascular_rat.R it instead enters a power model (EFF = -SL * CP_PRAZOSIN_NGML^POW, Equation 7 with SL = 0.328 and POW = 0.0910), the only non-Emax concentration-effect form in that paper. Set to 0 outside the exposure window. Reference values observed: IC50 = 0.133 ng/mL (Snelder 2013 Table 6), over doses of 0.04-5 mg/kg p.o. in both papers.
• Source aliases:
  • C (Snelder 2013 Equation 6 and Snelder 2014 Equations 6 and 7; values in ng/mL).
• Example models: Snelder_2013_cardiovascular_rat.R (Emax inhibition of the TPR production rate; Emax and IC50 both estimated, 0.213 and 0.133 ng/mL), Snelder_2014_cardiovascular_rat.R (power-model inhibition of the TPR production rate; SL = 0.328, POW = 0.0910).
• Notes: Specific scope; prazosin-specific. Follows the established CP_<drug>_<units> precedent. Both Snelder papers built the prazosin profile from a one-compartment model whose CL and Vd were allometrically scaled to the rat from the New Zealand white rabbit model of Hamilton et al. 1985 – the only cross-species PK extrapolation in either analysis (Snelder 2014 Table 3 Assumption 3). Snelder 2013 estimated the absorption rate constant simultaneously with the PD; Snelder 2014 found it very high and imprecise and FIXED it to 99 1/h, a near-instantaneous-absorption idealisation adopted to remove correlations between drug-specific parameters and shorten run times. Neither publishes PK parameter values and no prazosin PK model exists in the nlmixr2lib registry, so downstream users must supply this column externally. Prazosin is the poorly-identified potency in both papers, for different reasons: in Snelder 2013 the IC50 has CV 110% (95% CI -0.15 to 0.4 ng/mL) because Emax and IC50 were estimated jointly, and in Snelder 2014 the power exponent of 0.0910 means the maximum effect was never reached, so the concentration-effect curve is nearly flat and the absolute scale of this column matters less than for any other driver in the family. Note also that the Snelder 2014 power form is unbounded above, unlike an Emax form, and must not be extrapolated far beyond the studied dose range.

CP_SAPANISERTIB_NGML (canonical for instantaneous sapanisertib plasma concentration as a time-varying PD driver)

• Description: Instantaneous total plasma concentration of sapanisertib (MLN0128 / INK128 / TAK-228; an investigational orally administered inhibitor of both the mTORC1 and mTORC2 complexes), supplied directly as a time-varying covariate column rather than computed from a coupled PK model. Used in the PD-only linear mixed-effects concentration-effect models of the Patel 2020 dedicated-QTc (DQT) study, which take sapanisertib exposure as an external input to a concentration-DeltaQTcI, concentration-DeltaQTcF or concentration-DeltaRR equation.
• Units: ng/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as the linear term slope * CP_SAPANISERTIB_NGML in the concentration-effect expression Delta<endpoint> = e0 + slope * CP_SAPANISERTIB_NGML (Patel 2020 Figures 4A, 4C and S2; the slopes are reported in msec per ng/mL, so no in-model unit rescaling is needed). Set to 0 outside the drug-exposure window (the concentration-slope term then collapses to the intercept e0). Reference values observed: after the study’s single 40 mg oral dose, geometric mean Cmax 297 ng/mL (%CV 52.9), median Tmax 1.52 h (range 0.50-24.00), geometric mean AUCinf 2480 ng*h/mL (%CV 80.4), geometric mean CL/F 16.1 L/h (%CV 46.3) and mean t1/2 9.5 h (SD 2.9) (Patel 2020 Results ‘Pharmacokinetics’ and Table S2). Two lower-dose projection anchors carried over from prior single-agent studies are geometric mean Cmax 36.9 ng/mL (4 mg once daily) and 235 ng/mL (30 mg once weekly) (Patel 2020 Table 4 footnotes a and b).
• Source aliases:
  • sapanisertib plasma concentration / Concentration (ng mL-1) (Patel 2020 Results ‘Relationship of Sapanisertib Concentration and QTc’ and the x-axis label of Figures 4A, 4C and S2; measured by a validated LC-MS/MS method with deuterated sapanisertib as internal standard, dynamic range 1-1000 ng/mL, reported in ng/mL throughout, no value transformation) – used in Patel_2020_sapanisertib_QTcI.R, Patel_2020_sapanisertib_QTcF.R and Patel_2020_sapanisertib_RR.R.
• Example models: Patel_2020_sapanisertib_QTcI.R, Patel_2020_sapanisertib_QTcF.R, Patel_2020_sapanisertib_RR.R (PD-only linear mixed-effects concentration-effect models for DeltaQTcI, DeltaQTcF and DeltaRR in patients with advanced solid tumors).
• Notes: Specific scope; plasma matrix. The drug-specific naming follows the established CP_<drug>_<units> precedent (CP_GLASDEGIB_NGML, CP_TUVUSERTIB_NGML, CP_RACSOTALOL_UGML, CP_FRUQUINTINIB_NGML), which CP_MGL’s own Notes explicitly direct new drugs to (“register a drug-specific canonical … rather than overloading this name”). Patel 2020 fit the concentration-effect models directly to observed, time-matched plasma concentrations sampled alongside triplicate Holter ECGs; the paper does not develop a population PK model of its own – the PK is characterised by non-compartmental analysis only – and no sapanisertib popPK model exists in the nlmixr2lib registry, so a downstream user must supply the concentration trajectory. All three Patel 2020 models share this single column, so the same simulated trajectory drives the QTcI, QTcF and RR endpoints consistently. Note the sign asymmetry the column produces across those three models: the two QTc slopes are negative (interval shortening with rising concentration, the paper’s central negative finding) while the RR slope is positive (heart-rate slowing); a user who assumes a common sign convention across the set will misread one of them. Ratified canonically alongside the Patel 2020 sapanisertib DQT concentration-effect extraction.

AUCR (canonical for the drug-drug-interaction AUC ratio of the modelled drug)

• Description: Dimensionless fold change in the modelled (victim) drug’s own area under the plasma concentration-time curve caused by a co-administered interacting drug (perpetrator), relative to the same drug given alone: AUCR = AUC(victim, with precipitant) / AUC(victim, without precipitant). Carries the magnitude of a pharmacokinetic drug-drug interaction as a continuous covariate rather than as a binary co-medication flag, so a model can be simulated across a graded range of interaction strengths (and can be inverted to answer “how strong an interaction is still acceptable at this dose?”). Typically enters apparent clearance as a divisor when the dose is fixed, since at constant dose and constant bioavailability AUC is inversely proportional to CL/F: cl = exp(lcl) * <other covariate terms> / AUCR. Values are usually taken from a drug-interaction compilation (e.g. the University of Washington Drug Interaction Database) rather than measured in the modelling dataset.
• Units: (unitless ratio)
• Type: continuous
• Scope: general
• Reference category: n/a – reference value is 1.0, meaning no interacting drug is co-administered. Values > 1 denote inhibition of the victim’s clearance; values < 1 would denote induction.
• Source aliases:
  • AUCR – printed name in Okada 2024 Table 1, Table 3 and Methods (“area under the blood concentration-time curve increase ratio”).
• Example models: Okada_2024_triazolam.R (oral triazolam PK/PD; Table 3 clearance model CL = theta3 * (age/30)^theta5 * 1/AUCR, driven across the 15 literature AUCR values of Table 1 from ranitidine 1.31 to ritonavir 40.7).
• Notes: General scope: the quantity is defined relative to whichever drug the model describes, so any popPK/PK-PD model that parameterises interaction magnitude continuously can reuse the column without renaming it. Deliberately distinct from three neighbouring families. (1) The AUC_<DRUG> family (AUC_RTV, AUC_IBRU, AUC_VERUB, AUC_CARBO, AUC_GEM, AUC_GCV, AUC_LCM, AUC_CBZ, AUC_PAZO, AUC_AMPH, AUC_ADU/_DON/_GAN/_LEC, AUC_BAST_FW) carries an absolute interval-integrated exposure of a named drug in concentration-time units; AUCR is a dimensionless ratio of the victim’s exposure to itself. (2) The binary CONMED_* DDI indicators (CONMED_CYP3A4_INH, CONMED_CYP3A4_INH_STRONG/_MOD/_HI/_LO, CONMED_CYP3A4_IND, CONMED_ITRACONAZOLE, CONMED_AZOLE, CONMED_RTV, CONMED_COBICISTAT,…) record only whether an interacting drug is present, not how strong the interaction is; use AUCR when the model’s conclusions depend on the graded magnitude and a CONMED_* flag when the paper fits a categorical shift. (3) The CP_<drug>_<units> family (CP_RIF_UM, CP_GDC_UM, CP_PRB_MGL) carries an instantaneous perpetrator plasma concentration feeding a mechanistic (time-varying) inhibition term; AUCR is a static, already-integrated, victim-side summary of the whole interaction. Note also that AUCR is unrelated to the similarly-spelled UACR (urine albumin-to-creatinine ratio). A model that needs the interaction split by perpetrator identity should carry AUCR alongside a CONMED_<INN> flag rather than encoding the perpetrator into this column’s name. Name and general scope ratified by the maintainers (PMC11370030) alongside the Okada 2024 triazolam extraction.

AUC_PAROX (canonical for cumulative paroxetine AUC over the first week of treatment)

• Description: Area under the plasma paroxetine concentration-time curve accumulated from treatment initiation to the end of the first week of treatment (AUC 0-1 week). Per-subject, time-fixed: a single early-exposure summary carried as a baseline covariate into a downstream PK/PD model, not a per-dosing-interval AUC. Not measured directly – predicted for each patient from a companion population PK model applied to that patient’s week-1 dosing history. Shigetome 2025 screened eight paroxetine exposure indices (measured trough; population-PK-predicted trough, Cmax and single-dose AUC; and cumulative AUC to weeks 1, 2, 4 and 6) and retained only the first-week cumulative AUC, which is what makes the cumulative-to-a-fixed-early-timepoint convention the defining property of this column rather than an incidental one.
• Units: ng*h/mL (equivalently ug*h/L). Document per-model via covariateData[[AUC_PAROX]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as a ratio-power term normalised to a cohort reference value. Reference value observed: 2610.43 ngh/mL (Shigetome 2025 Results 3.3 printed equation and NONMEM Text S3). Observed distribution in the 50-patient PK/PD cohort: 2764.9 +/- 703.1 ngh/mL in the remission group and 2472.2 +/- 385.6 in the non-remission group, overall range 1642.4-5813.6 (Table S5).
• Source aliases:
  • AUC_W1 – NONMEM $INPUT column name in Shigetome 2025 Text S3; identical orientation, no value transformation.
  • AUC 0-1week – the paper’s prose and figure-legend form.
• Example models: Shigetome_2025_paroxetine_madrs.R (enters the maximum enhancement rate in depression severity as the subtracted power term emax - (AUC_PAROX / 2610.43)^-8.56, so that higher first-week exposure raises the attainable MADRS improvement).
• Notes: Specific scope because the column meaning is tied to paroxetine and to the cumulative-first-week convention. Member of the AUC_<DRUG> family (AUC_CARBO, AUC_GEM, AUC_GCV, AUC_PAZO, AUC_RTV, AUC_VERUB, AUC_ADU, AUC_DON, AUC_GAN, AUC_LEC, AUC_IBRU, AUC_LCM, AUC_CBZ, AUC_AMPH, AUC_LEN), registered as the drug-specific sibling that the AUC_BAST_FW entry directs future first-week-AUC models to create rather than overloading that teaching-dataset name. Distinct from the other AUC_<DRUG> members in that all of those are per-interval or per-cycle exposures that are updated as treatment proceeds, whereas AUC_PAROX is cumulative from time zero and then frozen. A future paroxetine model using a per-interval AUC, or a cumulative AUC to a different horizon (weeks 2, 4 or 6, all of which Shigetome 2025 computed but did not retain), should register a parallel canonical rather than overload this name. The -8.56 exponent makes the term extremely steep: emax reaches zero near AUC_PAROX = 1541 ng*h/mL and the term is essentially saturated above the reference value, so the covariate must not be extrapolated outside the observed range.

AUC_BTP (canonical for bitopertin steady-state AUC over the 24 h dosing interval)

• Description: Individual bitopertin (RG1678, a glycine transporter 1 inhibitor) area under the plasma concentration-time curve over the once-daily 24 h dosing interval at steady state, used as the exposure driver of the Emax inhibition of hemoglobin synthesis in semi-mechanistic erythropoiesis PKPD models. Time-varying step-wise: constant while the subject is on a given dose level and 0 otherwise. Rognas 2025 derives it inside the control stream as DRUG = ADOS_BTP / ICL_BTP * TREAT_BTP – the actual last dose amount (mg) divided by the individual empirical-Bayes apparent clearance (L/h) from an upstream bitopertin population PK model that the paper does not report (“data not shown”) – multiplied by a treatment indicator. TREAT_BTP is 1 from the time of a dose until two days after the last dose and 0 thereafter, so the drug effect carries over for two days past treatment end; downstream users encode that carry-over in the AUC_BTP time course rather than in the model body.
• Units: mg*h/L (equivalently ug*h/mL). Must be in the same units as the model’s AUC50 so the Emax term is dimensionless. Document per-model via covariateData[[AUC_BTP]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – enters via the Emax form Imax * AUC_BTP^gamma / (AUC50^gamma + AUC_BTP^gamma) with gamma fixed to 1. Set to 0 for placebo subjects and for off-treatment periods, whereupon the inhibition term vanishes exactly and the erythropoiesis system rests at its homeostatic baseline. Reference value observed: AUC50 = 16.50 mgh/L (Rognas 2025 Table 2). The paper’s own simulations specify exposure as multiples of AUC50 rather than as doses: 0.5 x AUC50 = 8.25 mgh/L gives 20% inhibition and 2 x AUC50 = 33 mg*h/L gives 40% inhibition (Methods “Simulations”; reproduced exactly by the packaged model).
• Source aliases:
  • DRUG – Rognas 2025 Supplementary Appendix section 4 NONMEM control stream $PK (the computed exposure metric actually used in $DES).
  • ICL_BTP – the $INPUT column carrying the upstream empirical-Bayes apparent clearance from which DRUG is derived; this is NOT the canonical (it is a clearance, and belongs to the CL_INDIV / CLI family).
  • AUCss – the name used in the paper’s Methods narrative (“Individual AUCss values were used as the exposure metric driving the bitopertin drug effect”).
• Example models: Rognas_2025_bitopertin.R (founding example – drives the Emax inhibition of the corpuscular-hemoglobin input rate kinMch, and, when the mechanism switches are overridden, of the precursor recruitment rate and precursor differentiation rate for the paper’s hypothetical Mechanisms B and C).
• Notes: Specific scope; bitopertin-specific and tied to the once-daily steady-state 24 h AUC convention. Member of the AUC_<DRUG> family (AUC_CARBO, AUC_GEM, AUC_GCV, AUC_PAZO, AUC_RTV, AUC_VERUB, AUC_ADU, AUC_DON, AUC_GAN, AUC_LEC, AUC_IBRU, AUC_LCM, AUC_CBZ, AUC_AMPH, AUC_LEN); AUC_VERUB is the closest structural analogue, being likewise a steady-state dosing-interval AUC standing in for a PK ODE and driving an Emax inhibition of a biosynthetic input rate. Because the upstream bitopertin population PK model is unpublished, a downstream user must either supply per-subject AUC values directly or express the exposure as a multiple of AUC50 as the paper’s own simulations do.

DOSE_LIDOCAINE_MG (canonical for total administered lidocaine dose over the treatment episode)

• Description: Total mass of lidocaine administered to a subject over the whole treatment episode, in milligrams – the loading infusion plus the entire continuous infusion, summed. Time-fixed per subject. Distinct from a per-administration or per-record dose level: this is a cumulative episode total, so it is only meaningful once the treatment episode is fully specified.
• Units: mg
• Type: continuous
• Scope: specific
• Reference category: n/a – used as a power term (DOSE_LIDOCAINE_MG / 312)^exponent, where 312 mg is the He 2025 study-typical episode total derived from Table 1 (mean loading dose 86.07 mg plus median continuous rate 57.97 mg/h times mean infusion duration 3.90 h).
• Source aliases:
  • DOSE – the column name printed in the He 2025 Supplementary Table S2 abbreviation list (“DOSE, Total doses of lidocaine”).
• Example models: He_2025_lidocaine.R (power effects on three parameters simultaneously: the lidocaine-to-MEGX fraction metabolised, exponent 0.669; the peripheral volume, exponent 1.27; and GX clearance, exponent 1).
• Notes: A drug-specific member of the DOSE_<drug>_<units> family, required here rather than the bare DOSE canonical because rxode2’s event-table translator (etTrans) consumes a column literally named DOSE and never exposes it to model; a model that reads it from an event table fails at solve time with The following parameter(s) are required for solving: DOSE. Confirmed for He_2025_lidocaine.R on the rxUi solve path. Because the covariate is an episode total rather than a dose level, it extrapolates aggressively: it spans roughly 200-500 mg across the He 2025 trial but reaches 6480 mg in that paper’s own 72 h postoperative-infusion simulation, where the exponent of 1.27 on the peripheral volume inflates V2 more than fortyfold. Treat predictions at episode totals outside the fitted range as illustrative.

AUC_ISTRA (canonical for istradefylline steady-state daily AUC)

• Description: Istradefylline area under the plasma concentration-time curve at steady state over the once-daily 24 h dosing interval, used as the systemic-exposure driver of the Knebel 2012 istradefylline exposure-response models for percentage OFF time and for the dyskinesia, dizziness and nausea adverse-event endpoints. Time-fixed per subject in the source analysis. Not an observed AUC: Knebel 2012 generated it as the individual empirical-Bayes prediction from a companion population PK model (a two-compartment model with first-order absorption; Knebel 2012 reference 6, Knebel W, Rao N, Uchimura T, et al., J Clin Pharmacol, published online ahead of print 3 March 2010), which is not currently part of nlmixr2lib, so downstream users must supply the column themselves.
• Units: ng*hr/mL (equivalently ug*h/L). Must be in the same units as the model’s EC50 so the Emax term is dimensionless. Document per-model via covariateData[[AUC_ISTRA]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – enters an ordinary Emax form EmaxI * AUC / (EC50 + AUC) in the efficacy model, sigmoid Emax forms Emax * AUC^gamma / (EC50^gamma + AUC^gamma) in the dyskinesia and dizziness models, and an unbounded power form SLOP * AUC^power in the nausea model. Set to 0 for placebo subjects; all four forms then vanish exactly, leaving the disease-progression/placebo-response trajectory (efficacy) or the baseline logit (adverse events). Reference (half-maximal) values observed: 1690 nghr/mL for percentage OFF time in the base model and 1860 nghr/mL in the final full covariate model (Knebel 2012 Table II); 2380 nghr/mL for dyskinesia and 2770 nghr/mL for dizziness (Knebel 2012 Table IV).
• Source aliases:
  • AUC / AUCss – printed names in Knebel 2012 (the individual-parameter equations write AUCi, and the Results and Table II row labels write AUCss).
• Example models: Knebel_2012_istradefylline_offtime.R (Emax effect on percentage OFF time, additive to the disease-progression/placebo-response component, EC50 1860 nghr/mL), Knebel_2012_istradefylline_dyskinesia.R (sigmoid Emax on the dyskinesia log-odds, EC50 2380 nghr/mL, Hill 2.94), Knebel_2012_istradefylline_dizziness.R (sigmoid Emax on the dizziness log-odds, EC50 2770 ng*hr/mL, Hill 10), Knebel_2012_istradefylline_nausea.R (unbounded power effect on the nausea log-odds, slope 0.00218, exponent 0.635).
• Notes: Specific scope because the column meaning is tied to istradefylline and to the once-daily steady-state 24 h AUC convention. Member of the AUC_<DRUG> family (AUC_CARBO, AUC_GEM, AUC_GCV, AUC_PAZO, AUC_RTV, AUC_VERUB, AUC_ADU, AUC_DON, AUC_GAN, AUC_LEC, AUC_IBRU, AUC_LEN, AUC_LCM, AUC_CBZ, AUC_AMPH, AUC_PAROX); AUC_IBRU and AUC_LEN are the closest structural analogues, each likewise a daily steady-state exposure handed to Emax-shaped PD models in place of a PK ODE, and each likewise requiring the user to compute the column from an upstream popPK model that is not in the library. Knebel 2012 reports only two numeric dose-to-exposure anchors, both approximate: a median exposure of 1690 nghr/mL corresponds to 5 mg/d (Discussion) and approximately 2500 nghr/mL is the approximate median at 10 mg/d (Results, Figure 2); the full per-dose distribution across the 5-60 mg/d range appears only graphically, as a box-and-whisker panel in Figure 4. Effects on exposure itself – smoking lowers AUCss by 38 percent, CYP3A4 inhibitors raise it by 35 percent (Knebel 2012 Discussion) – belong to the upstream PK model, not to any model using this column. A future istradefylline model using a different exposure metric (Cmax, Ctrough, or a cumulative rather than per-interval AUC) should register a parallel canonical rather than overload AUC_ISTRA.

CONC_ADA_NGML (canonical for calibrated total anti-drug-antibody mass concentration)

• Description: Calibrated total (free plus drug-bound) anti-drug-antibody concentration on a mass-concentration scale, measured by a drug-tolerant immunoassay – typically a precipitation / acid-dissociation format, which dissociates ADA from circulating drug so the readout is not confounded by drug interference. Measured in every subject regardless of positivity status, so the column is strictly positive and carries no special encoding for ADA-negative subjects: ADA-negatives hold a real, non-zero measured concentration. Where a source paper also reports positivity, that positivity is a derived dichotomy obtained by applying an assay-specific cutoff to this column, not a property of the column itself. Time-varying, matched in time to the PK sample.
• Units: ng/mL (baked into the canonical name per the CONC_<ANALYTE>_<UNIT> family convention; document per-model via covariateData[[CONC_ADA_NGML]]$units).
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (CONC_ADA_NGML / ref)^exponent. Reference values observed: 10 ng/mL (Song 2025, cohort mean, which coincides with that assay’s positivity cutoff).
• Source aliases:
  • ADA – printed name in Song 2025 Table 2 (theta ADA-CL, theta ADA-F) and in the p. 380 final-model equations, where it enters as (ADA/10)^theta. A bare ADA column name is ambiguous across the ADA family; check the source’s Methods for whether the modelled quantity is a concentration, a titer, or a positivity flag before mapping it here.
• Example models: Song_2025_infliximab.R (founding example; ng/mL, reference 10; power effects on BOTH clearance, exponent 0.022, and subcutaneous bioavailability, exponent -0.213, via e_conc_ada_cl / e_conc_ada_f).
• Notes: Member of the analyte-concentration CONC_<ANALYTE>_<UNIT> family (siblings CONC_VORI_NGML, CONC_RIF_MGL, CONC_TOB_MGL, CONC_BAI_UM, CONC_AGONIST_M,…), which bakes the unit token into the canonical name in every member. The unit token is load-bearing here beyond the family convention: ADA_TITER is already registered as a dimensionless dilution factor, and the whole risk this canonical exists to manage is a user conflating the two ADA quantities – an unqualified ADA_CONC would invite exactly that, CONC_ADA_NGML forecloses it. The analyte token is ADA rather than a specific antibody because the assay is defined by its target (the therapeutic protein) rather than by a single molecular species; the drug the ADA is directed against is identified per-model, not in the column name. Distinct from [[ADA_TITER]], whose defining, load-bearing semantic is a 0 (American linear-titer convention) or 1 (British reciprocal-dilution convention) encoding for ADA-negative subjects – a semantic that is actively false for this column, and whose 0 would make the (ADA/ref)^theta power form undefined; reusing ADA_TITER for a ng/mL quantity would also be a unit lie. Distinct from [[ADA_POS]], the binary positivity indicator, which is what this column reduces to once a cutoff is applied and is the right choice when a paper estimates an on/off coefficient rather than a continuous exposure-response. Distinct from [[ADA_MISSING]] and [[ADA_POSOLD]], the missing-result and older-assay-generation companions of ADA_POS. Prefer this canonical over dichotomising when the source paper estimates a continuous ADA effect: Song 2025’s Discussion argues the continuous encoding is the point (“while binary classification is intuitive, it can lead to information loss”), and collapsing it to ADA_POS would discard an estimated parameter with nothing to replace it.

CONC_ADA_AUML (canonical for anti-drug-antibody level on an arbitrary-unit concentration scale)

• Description: Anti-drug-antibody level reported on an assay-specific ARBITRARY-UNIT concentration scale (AU/mL), calibrated against an in-house standard rather than against a mass standard – the readout of an antigen-binding radioimmunoassay. Measured in every subject regardless of positivity status, so the column is strictly positive and carries no special encoding for ADA-negative subjects. Where a source paper also reports positivity, that positivity is a derived dichotomy obtained by applying an assay-specific cutoff to this column (30 AU/mL in Pan 2026), not a property of the column itself. Time-varying, matched in time to the PK sample.
• Units: AU/mL (baked into the canonical name per the CONC_<ANALYTE>_<UNIT> family convention; document per-model via covariateData[[CONC_ADA_AUML]]$units).
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (CONC_ADA_AUML / ref)^exponent. Reference values observed: 76.03 AU/mL (Pan 2026, the normalising constant of the Data S1 Code S1 CLARIA block).
• Source aliases:
  • ARIA – Pan 2026 NONMEM $INPUT (Data S1 Code S1), the paper’s contraction of “ADA by RIA”. Note that Pan 2026’s dataset carries a separate CENSORARIA_IDV censoring flag and a distinct ANTIADL column; ARIA is the modelled quantitative level.
• Example models: Pan_2026_adalimumab.R (founding example; (CONC_ADA_AUML/76.03)^0.368 on apparent clearance of subcutaneous adalimumab in psoriasis).
• Notes: Member of the analyte-concentration CONC_<ANALYTE>_<UNIT> family (siblings CONC_ADA_NGML, CONC_VORI_NGML, CONC_RIF_MGL, CONC_BAI_UM, CONC_AGONIST_M, …), which bakes the unit token into the canonical name in every member. Distinct from [[CONC_ADA_NGML]] only in the unit scale – the semantics are identical – but an arbitrary-unit readout is not convertible to ng/mL without the originating assay’s own standard, so values under the two canonicals must never be pooled or compared numerically. Distinct from [[ADA_TITER]], whose defining, load-bearing semantic is a 0 or 1 encoding for ADA-negative subjects: that semantic is actively false here, and a 0 would make the (ADA/ref)^theta power form undefined. ADA_TITER’s Units field does admit “assay units (log2 or arbitrary)”, which is why the distinction has to be drawn on the zero-encoding rather than on the units alone. Distinct from [[ADA_POS]], the binary positivity indicator this column reduces to once a cutoff is applied; prefer this canonical over dichotomising when the source paper estimates a continuous ADA effect, since collapsing it would discard an estimated parameter. Ratified 2026-09-02 alongside the Pan 2026 adalimumab extraction.

AUC_PEMBRO (canonical for pembrolizumab serum AUC at steady state over a 6-week interval)

• Description: Per-subject pembrolizumab area under the serum concentration-time curve at steady state over a 6-week window (AUCss-6weeks), used as the exposure driver in exposure-response analyses of pembrolizumab. Six weeks is the canonical window because it contains an integer number of dosing intervals for both the Q2W (three doses) and Q3W (two doses) pembrolizumab regimens, which is exactly why the founding paper chose it. A steady-state, time-independent metric rather than a cumulative one: Chatterjee 2016 Methods argues that a time-dependent exposure metric would be confounded by early dropout, because patients who progress early accumulate less total exposure and would manufacture a spurious positive exposure-response relationship.
• Units: mg*day/L (equivalently ug*day/mL; the two are numerically identical and both spellings appear in the founding paper – Methods prints “mg/l x day” while the Figure 2 axis prints “ug.day/ml”). Document per-model via covariateData[[AUC_PEMBRO]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as a normalised power (log-linear) term (AUC_PEMBRO / ref)^exponent. Reference value observed: 7079 mg*day/L (Chatterjee 2016, the population-typical AUCtypical,ss-6weeks stated in Methods).
• Source aliases:
  • AUCss-6weeks / AUCss-6wk – Chatterjee 2016 main-article and supplementary notation.
• Example models: Chatterjee_2016_pembrolizumab.R (power effect (AUC_PEMBRO / 7079)^0.196 on the tumor kill rate kdeath of a bi-exponential tumor-size model in advanced NSCLC; the exponent’s 95% CI overlaps zero, and the paper’s conclusion is that response is flat over 2-10 mg/kg), Chatterjee_2017_pembrolizumab_consolidated.R (the same power form and the same 7079 mgday/L normalisation on kdeath in advanced melanoma, but with SEPARATE exponents per prior-ipilimumab stratum – 0.131 IPI-naive, 0.100 IPI-experienced – because dose assignment was confounded with prior-ipilimumab status in the pooled dataset; neither is significant, P = 0.20 and P = 0.25), Chatterjee_2017_pembrolizumab_mixture.R (the same paper’s earlier four-class mixture model, where exposure was NOT retained on the kill rate and instead enters the second-stage multinomial regression on the class logits as an additive term centered at 6000 mgday/L, coefficient 5.61e-5 per mg*day/L).
• Notes: Member of the AUC_<drug> exposure-metric family (AUC_ADU, AUC_DON, AUC_GAN, AUC_LEC, AUC_CARBO,…). Not a model output: the founding analysis has no PK sub-model and obtains the value as dose / CL using post-hoc individual clearances from a separate pembrolizumab population-PK analysis, published as Ahamadi 2017 and packaged in this library as Ahamadi_2017_pembrolizumab.R. To reproduce the metric, compute AUC_PEMBRO = (total mg administered over 6 weeks) / CL with CL in L/day. Because the window is fixed at six weeks rather than at one dosing interval, a Q2W regimen at the same mg/kg dose yields 1.5x the AUC of the Q3W regimen; do not confuse this column with a per-dosing-interval AUC.

AUC_BV_ADC (canonical for time-averaged brentuximab vedotin antibody-drug conjugate AUC up to an event)

• Description: Per-subject time-averaged AUC of the brentuximab vedotin antibody-drug conjugate (ADC), computed from post hoc popPK predictions over actual doses received and averaged over the time to the first occurrence of the worst-grade event (safety endpoints), to progression or censoring (efficacy), or over the treatment duration when no event occurred. A metric of exposure INTENSITY that absorbs dose reductions and delays, expressed per 14-day dosing interval (Suri 2019 Figure 4 axis values of about 20-110 ug*day/mL match the per-interval AUC of a 1.2 mg/kg every-2-weeks regimen).
• Units: ug*day/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – enters LINEARLY and uncentred on the logit (b0 + b1 * AUC_BV_ADC).
• Source aliases:
  • ADC AUC/time, ADC AUC (ug-day/mL) until <event> – Suri 2019 text and Figure 4 axis labels.
• Example models: Suri_2019_brentuximab_pn.R (grade 2 or worse peripheral neuropathy), Suri_2019_brentuximab_fn_adc.R (febrile neutropenia, with CONMED_GCSF), Suri_2019_brentuximab.R.
• Notes: Member of the AUC_<drug> exposure-metric family. Named for the drug AND the analyte because the same source derives a parallel payload metric (AUC_BV_MMAE) and fits separate ADC-driven and MMAE-driven models for the same endpoint (febrile neutropenia). Not interchangeable with a steady-state single-interval AUC: the event-truncated averaging window differs per endpoint and per subject. Suri 2019 Figure 4a gives the ADC AUC/time quartile boundaries of the efficacy analysis set as 50.4, 58.2 and 65.8 ug*day/mL.

AUC_BV_MMAE (canonical for time-averaged monomethyl auristatin E AUC after brentuximab vedotin, up to an event)

• Description: Per-subject time-averaged AUC of unconjugated monomethyl auristatin E (MMAE) released from brentuximab vedotin, computed from post hoc popPK predictions over actual doses received and averaged over the time to the first occurrence of the worst-grade event, or over the treatment duration when no event occurred, expressed per 14-day dosing interval (Suri 2019 Figure 4 axis values of about 2-41 ng*day/mL).
• Units: ng*day/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – enters LINEARLY and uncentred on the logit (b0 + b1 * AUC_BV_MMAE).
• Source aliases:
  • MMAE AUC/time, MMAE AUC (ng-day/mL) until <event> – Suri 2019 text and Figure 4 axis labels.
• Example models: Suri_2019_brentuximab_fn_mmae.R (febrile neutropenia), Suri_2019_brentuximab_neutropenia.R (grade 4 or worse neutropenia), Suri_2019_brentuximab_teae.R (any grade 3 or worse TEAE), Suri_2019_brentuximab.R.
• Notes: Member of the AUC_<drug> exposure-metric family; the ADC counterpart is AUC_BV_ADC. Scoped to brentuximab vedotin rather than to MMAE generally because MMAE exposure per unit of conjugate differs between vedotin ADCs (different drug-antibody ratios, targets and formation kinetics); another vedotin should register its own AUC_<drug>_MMAE column. MMAE exposure falls by about half from the first to later dosing intervals (Suri 2019), so the time-averaged value depends on how long the averaging window runs.

AUC_BRIG_SCAN (canonical for brigatinib daily AUC averaged over a disease-assessment scan interval)

• Description: Per-subject brigatinib daily AUC averaged over a disease-assessment (CT / MRI) scan interval, used as the exposure driver of exposure-efficacy models. Gupta 2022 (Methods S1) defines the daily AUC on day i as the increment of the cumulative AUC, AUCD(i) = CAUC(i + 1) - CAUC(i) with CAUC(i) = cumulative dose to day i divided by the individual Bayesian CL/F of the Gupta 2021 population PK model, over the actually administered doses; this column is the mean of AUCD over one scan interval. In the STATIC models it is the interval between the last two scans preceding the event (progression), the best confirmed response, or censoring – one value per subject. In the time-varying Cox model Gupta_2022_brigatinib_pfs_scan_tv.R it is supplied as a piecewise-constant column holding each successive interval’s mean.
• Units: ug*h/mL/day
• Type: continuous
• Scope: specific
• Reference category: n/a – enters LINEARLY and uncentred (logit_ref + slope * AUC_BRIG_SCAN for the logistic models; log(HR) * AUC_BRIG_SCAN for the Cox relative hazards). Calibration: Gupta 2022 Figure 2a gives the PFS analysis-set median (range) as 15.2 (0, 77.0) ug*h/mL/day and the quartile medians as 9.7, 13.0, 17.1 and 28.3. For an uninterrupted 180 mg once-daily course the value equals 180 mg / CL/F, whose ALTA-1L geometric mean (5th, 95th percentile) is 21.3 (10.1, 44.6) (Gupta 2022 Results).
• Source aliases:
  • Time-averaged AUC between last two scans (ug.h/mL/day) – Gupta 2022 Figure 2b / 2c x-axis label.
  • Daily time-averaged AUC between successive disease assessments – Gupta 2022 Results, time-varying PFS analysis.
• Example models: Gupta_2022_brigatinib_orr.R, Gupta_2022_brigatinib_iorr.R, Gupta_2022_brigatinib_pfs.R (static), Gupta_2022_brigatinib_pfs_scan_tv.R (time-varying).
• Notes: Member of the AUC_<DRUG> exposure-metric family with a metric qualifier, following AUC_MBV_SS / AUC_MBV_DAY: Gupta 2022 uses four brigatinib exposure definitions in one paper (AUC_BRIG_SCAN, AUC_BRIG_EVT, AUC_BRIG_D8_14, AUC_BRIG_DAY) and they are not interchangeable. A scan-interval mean absorbs dose reductions and interruptions within that interval, which is why the static PFS model shows a counter-intuitive positive hazard ratio (patients with long PFS accumulate dose reductions and so carry a lower last-interval exposure).

AUC_BRIG_EVT (canonical for brigatinib daily AUC time-averaged from the first dose to an adverse event)

• Description: Per-subject brigatinib daily AUC time-averaged from the first dose to the first occurrence of the adverse event of interest, or to the end of treatment for a subject without the event: cumulative dose over that window divided by the individual Bayesian CL/F of the Gupta 2021 population PK model and by the number of days (Gupta 2022 Methods, Exposure-safety analyses, and Methods S1). Used as the exposure driver of landmark exposure-safety logistic regressions.
• Units: ug*h/mL/day
• Type: continuous
• Scope: specific
• Reference category: n/a – enters LINEARLY and uncentred on the logit (logit_ref + slope * AUC_BRIG_EVT). Gupta 2022 Supporting Figures S6 and S7 plot the observed values over roughly 6 to 70 ug*h/mL/day.
• Source aliases:
  • Time-Averaged Exposure (ug.h/mL/day) – Gupta 2022 Supporting Figures S6 and S7 x-axis label.
  • time-averaged AUC to the first occurrence of an event (or to the end of treatment if no event) – Gupta 2022 Methods.
• Example models: Gupta_2022_brigatinib_cpk_grade3.R, Gupta_2022_brigatinib_ast_grade3.R, Gupta_2022_brigatinib_alt_grade3.R, Gupta_2022_brigatinib_amylase_grade3.R, Gupta_2022_brigatinib_lipase_grade3.R, Gupta_2022_brigatinib_aesi_grade3.R, Gupta_2022_brigatinib_hyperglycemia_grade2.R, Gupta_2022_brigatinib_hypertension_grade2.R, Gupta_2022_brigatinib_bradycardia_grade2.R, Gupta_2022_brigatinib_rash_grade2.R, Gupta_2022_brigatinib_ast_grade2.R, Gupta_2022_brigatinib_alt_grade2.R, Gupta_2022_brigatinib_amylase_grade2.R, Gupta_2022_brigatinib_pulmonary_grade2.R.
• Notes: Member of the AUC_<DRUG> family; see AUC_BRIG_SCAN for the four sibling brigatinib metrics. The averaging window is endpoint- AND subject-specific, so the same patient carries a different value in each adverse-event model, and because the window includes the 7-day 90 mg lead-in an early event is attached to a lower exposure than the 180 mg maintenance AUC. The closest analogues are AUC_ALIS and AUC_BV_ADC (treatment-course averages up to an event). Do not substitute a steady-state AUC(0-24).

AUC_BRIG_D8_14 (canonical for brigatinib daily AUC time-averaged over days 8 to 14 of cycle 1)

• Description: Per-subject brigatinib daily AUC time-averaged over days 8 to 14 of cycle 1 – the first week after the dose increase from the 7-day 90 mg once-daily lead-in to 180 mg once daily – computed over actual doses from the individual Bayesian CL/F of the Gupta 2021 population PK model. Gupta 2022 introduced it to represent exposure early after the step-up, when pancreatic-enzyme elevations cluster.
• Units: ug*h/mL/day
• Type: continuous
• Scope: specific
• Reference category: n/a – enters LINEARLY and uncentred on the logit. Gupta 2022 Figure 3 plots the observed values over roughly 6 to 54 ug*h/mL/day.
• Source aliases:
  • Averaged Daily AUC Day 8-14 (ug.h/mL/day) – Gupta 2022 Figure 3 x-axis label.
• Example models: Gupta_2022_brigatinib_lipase_grade3_d8_14.R (odds ratio 1.05 per ug*h/mL/day, p = 0.039), Gupta_2022_brigatinib_amylase_grade2_d8_14.R (odds ratio 1.06, p = 0.016).
• Notes: Member of the AUC_<DRUG> family; see AUC_BRIG_SCAN for the sibling metrics. Unlike AUC_BRIG_EVT the window is fixed by protocol day, so one patient has one value shared by both models that use it.

AUC_BRIG_DAY (canonical for brigatinib daily AUC on each treatment day, time-varying)

• Description: Brigatinib AUC over each treatment day, AUCD(i) = CAUC(i + 1) - CAUC(i) with CAUC(i) = cumulative dose to day i divided by the individual Bayesian CL/F of the Gupta 2021 population PK model (Gupta 2022 Methods S1). A TIME-VARYING covariate, one value per day, that changes at every dose change or interruption.
• Units: ug*h/mL/day
• Type: continuous
• Scope: specific
• Reference category: n/a – enters LINEARLY and uncentred as the time-dependent covariate of a Cox model (log(HR) * AUC_BRIG_DAY). For an uninterrupted 180 mg day it equals 180 mg / CL/F.
• Source aliases:
  • daily time-varying AUC – Gupta 2022 Results, Exposure-efficacy analysis.
• Example models: Gupta_2022_brigatinib_pfs_daily_tv.R (hazard ratio 1.01 per ug*h/mL/day, p = 0.69).
• Notes: Member of the AUC_<DRUG> family; see AUC_BRIG_SCAN for the sibling metrics. Distinct from AUC_MBV_DAY, which is a single per-subject value on the calendar day of an event rather than a daily time series.

DOSE_VGB_MGD (canonical for total daily vigabatrin dose)

• Description: Patient’s total daily vigabatrin dosage (mg/day), summed across the twice-daily administrations and NOT normalised by body weight. Per-record (time-varying) covariate: dosage was titrated in every contributing study, so the column is updated as the patient escalates. Set to 0 mg/day during a baseline run-in and for placebo subjects.
• Units: mg/day
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as the driver of the “normalized dosage” exposure surrogate dosenorm = DOSE_VGB_MGD * (WT/60)^e_wt_dosenorm, which in turn drives a quadratic drug effect exp(aquad * (dosenorm/3000)^2 - blin * (dosenorm/3000)). A value of 0 makes dosenorm 0 and the drug-effect term exactly 1.
• Source aliases:
  • DOSE – used in Nielsen_2015_vigabatrin.R (Nielsen 2015 Equation 10e, “a patient’s actual dosage (in mg)”).
• Example models: Nielsen_2015_vigabatrin.R (daily seizure-count dose-response; adult flat dosages of 1, 3 and 6 g/day and pediatric weight-banded or mg/kg dosing, pooled across five trials).
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family (siblings: DOSE_VPA_MGD for valproic acid, DOSE_TPM_MGD for topiramate, DOSE_EMPA_MGD, DOSE_LOR_MGD). Flat mg/day rather than per-kg because the pooled analysis carries both adult flat dosing and pediatric mg/kg dosing on a single mass scale; a future extraction that needs the per-kilogram scale should register a sibling DOSE_VGB_MGKGD rather than reuse this name. Study 118 dosing-weight cap: in the Nielsen 2015 pool, participants in Study 118 weighing more than 60 kg received their mg/kg dosage as if they weighed 60 kg. The authors corrected the analysis dataset for this before the final model run, so a simulation reproducing that study must cap the dosing weight at 60 kg when converting mg/kg to mg/day – the cap belongs in the dataset that fills this column, not in model(). Distinct from the rxode2 / nlmixr2 event column amt, which carries the amount of an individual administration; this is a per-record covariate carrying the current daily-dose LEVEL. Never name a dose covariate column bare DOSE: rxode2::etTrans() consumes a column of that name (any casing) before model() sees it.

CSS_LEC (canonical for lecanemab average steady-state serum concentration from an upstream PK fit)

• Description: Subject-specific (and, in use, time-varying) average steady-state serum concentration of the anti-amyloid monoclonal antibody lecanemab, generated externally by the lecanemab population PK model (Hayato et al., cited as references 22 and 24 of Cao 2026) and fed into a downstream disease model as the drug-exposure driver. In Cao_2026_lecanemab.R it drives two additive clearance-enhancement terms on the amyloid aggregation cascade: Eff_LECAF = 1 + DELECAF * CSS_LEC on A-beta protofibril clearance and Eff_LECAP = 1 + DELECAP * CSS_LEC on A-beta plaque clearance. Treated as time-varying so treatment, washout and open-label-extension windows are expressed by switching the column between 0 and the on-treatment Css,av – algebraically identical to the source paper’s if (treatment window)... else Eff = 1 construction and generalising to any number of windows.
• Units: ug/mL (micrograms per millilitre, equivalently mg/L). The paired drug-effect parameters carry the reciprocal units mL/ug. Document per-model via covariateData[[CSS_LEC]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – enters directly into the linear drug-effect expressions. Set to 0 to disable the drug effect (placebo arms, pre-treatment natural-history periods, post-discontinuation washout). Reference value for the approved 10 mg/kg IV Q2W regimen: the Cao 2026 Vpop301 Css,av distribution (Supplementary Figure E21) is log-normal with a mode near 115-130 ug/mL and a median near 135 ug/mL, spanning roughly 30-370 ug/mL.
• Source aliases:
  • CSSAV01 / CSSAV02 (Cao 2026 Methods pseudocode; one column per treatment window, selected by the TXSTR / TXEND window bounds) – collapse the per-window columns into a single time-varying CSS_LEC column for nlmixr2lib.
  • Css,av (paper narrative symbol).
• Example models: Cao_2026_lecanemab.R (ug/mL; time-varying input on the linear protofibril and plaque clearance-enhancement terms of an 11-ODE Alzheimer’s-disease QSP model).
• Notes: Member of the CSS_<drug> canonical family (siblings: CSS_RBV ribavirin, CSS_ALBIFN albinterferon alfa-2b, CSS_DFO deferoxamine) for a steady-state average concentration supplied by an upstream PK model rather than simulated in the same model. The LEC drug abbreviation matches the existing AUC_LEC entry. Distinct from AUC_LEC, which is the lecanemab AUC over a 4-week dosing interval: the two are not interconvertible without the dosing interval, and a model must use whichever quantity its own drug-effect parameter was estimated against. Cao 2026 additionally re-uses the same exposure channel for donanemab, aducanumab and gantenerumab via drug-specific exposure multipliers scaled to lecanemab’s Css,av, but does not tabulate those multipliers; a future extraction that needs them should register sibling CSS_<drug> columns rather than overloading this one. Founding example: Cao_2026_lecanemab.R.

CP_EFALIZUMAB_UGML (canonical for instantaneous efalizumab concentration as a time-varying target-engagement driver)

• Description: Instantaneous concentration of efalizumab (hu1124; humanised anti-CD11a monoclonal antibody, withdrawn 2009), supplied directly as a time-varying covariate column rather than computed from a coupled PK model. Used in QSP target-engagement models that consume an external efalizumab exposure profile as the driver of CD11a occupancy and CD11a down-modulation. Converted to umol/L inside model() as CP_EFALIZUMAB_UGML / mw_efa * 1000 (equivalently, divided by 148.841) because the CD11a pool and the Michaelis-Menten constant are reported in umol/L.
• Units: ug/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – after conversion to umol/L it enters the quasi-steady-state occupancy fraction C / (Km + C) that partitions the total CD11a pool between free and bound receptor. Set to 0 for untreated / pre-dose periods and for placebo arms, which returns both the total and the free CD11a pool to the baseline abundance Rmax. Reference values observed (Chetty 2014 Figure 2C, predicted concentration-time profile for the Gottlieb 2002 escalating regimen): weekly peaks rising from about 5 ug/mL in week 1 to about 30 ug/mL in weeks 5 to 7, with pre-dose troughs rising from about 0.5 ug/mL to about 7 ug/mL. Single-dose profiles for 1, 3 and 10 mg/kg are in Figure 2A.
• Source aliases:
  • systemic concentration – the Chetty 2014 Figure 2 y-axis label; the paper has no named data column because the profile is produced internally by its Simcyp Mechanistic FcRn PBPK model.
• Example models: Chetty_2014_efalizumab_cd11a_qsp.R (drives the CD11a target-turnover and quasi-steady-state occupancy limbs; Km = 0.000573 umol/L, Rmax = 0.01 umol/L).
• Notes: Specific scope; efalizumab-specific. Follows the auto-approved CP_<drug>_<units> family (CP_RITUXIMAB_UGML, CP_RIB_NGML, CP_ABE_NGML, CP_MILTEFOSINE_UGML,…), in ug/mL rather than ng/mL to match the monoclonal-antibody sibling CP_RITUXIMAB_UGML and the paper’s own Figure 2 axis. Matrix is deliberately unqualified, and that is a limitation of the source, not of this entry. Chetty 2014 never states whether target binding in its Simcyp implementation was driven from plasma or from a tissue interstitial concentration; its Introduction notes only that “input to the PD model can be done from a tissue interstitial compartment and not just from plasma, which is important when modeling membrane bound receptors”. Driving Chetty_2014_efalizumab_cd11a_qsp.R with the plasma concentrations of the paper’s own Figure 2C predicts free CD11a at roughly 0.1 to 3 percent of baseline, against the 9 to 24 percent of its Figure 3; inverting the model on the Figure 3 values implies a driver of roughly 0.05 to 0.16 ug/mL, one to two orders of magnitude below plasma. Users must therefore state which matrix their trajectory represents. Under the quasi-steady-state approximation the quantity that enters the binding term is the FREE (target-unbound) concentration; free and total differ by at most Rmax = 0.01 umol/L (1.49 ug/mL), immaterial above roughly 15 ug/mL and material below 1 ug/mL. The Chetty 2014 PBPK layer is a Simcyp platform port that is not extracted, so a downstream user must supply the trajectory; Stein_2018_efalizumab.R is an in-registry two-compartment QSS-TMDD efalizumab PK model (Stein and Peletier 2018, fitted to the Bauer 1999 data) that can serve that role, though it originates from a different publication and its own target layer is not comparable – Stein and Peletier report ke(R) = ke(CR) = 4400 /day as practically unidentifiable, giving a baseline target pool roughly 5000-fold smaller than the Chetty 2014 Rmax. Use its drug concentration, not its target states.

CP_SORAFENIB_NGML (canonical for instantaneous sorafenib concentration as a concentration-response driver)

• Description: Instantaneous total concentration of sorafenib – the multitargeted tyrosine kinase inhibitor – supplied directly as a covariate column rather than computed from a coupled PK model. Registered for in-vitro transporter concentration-response models, where the column carries the sorafenib concentration in the cell-assay buffer (a Transwell donor-compartment concentration, or an incubation-medium concentration) rather than a plasma concentration; the quantity and units are identical to the plasma-matrix members of the family, so a single column serves both roles.
• Units: ng/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as the argument of a concentration-response expression: the hyperbolic saturation term papp_ba = emax * (1 - C / (km + C)) in the BCRP model, and the Hill term vbl_accum = e0 + (emax - e0) * C^hill / (ic50^hill + C^hill) in the P-gp model. At C = 0 the BCRP model returns its tracer-concentration maximum emax and the P-gp model returns its baseline e0. Reference values observed (Agarwal 2011): BCRP assay donor concentrations 2 ng/mL to 30 ug/mL with apparent affinity km = 3.6 ng/mL; P-gp inhibition assay concentrations 20 ng/mL to 60 ug/mL with ic50 = 15900 ng/mL (15.9 ug/mL). For clinical scale, sorafenib plasma concentrations on the accepted 100-400 mg b.i.d. regimen are reported at 1-15 uM (Agarwal 2011 Discussion, citing Richly 2006 and Duran 2007).
• Source aliases:
  • C (Agarwal 2011 Methods Equations 2 and 3, glossed “the concentration of sorafenib in the donor compartment” for the BCRP model and “the concentration of sorafenib” for the P-gp model; reported in ng/mL on the Figure 5 axis and in ug/mL on the Figure 6A axis, so the ug/mL form is converted losslessly by x1000).
• Example models: Agarwal_2011_sorafenib_bcrp.R (saturation of BCRP-mediated basolateral-to-apical permeability in MDCKII-Bcrp1 monolayers; column carries the donor-side concentration), Agarwal_2011_sorafenib_pgp.R (sigmoidal inhibition of P-gp measured as fold increase in vinblastine accumulation in MDCKII-MDR1 cells; column carries the incubation-medium concentration).
• Notes: Specific scope; sorafenib-specific. Follows the established CP_<drug>_<units> precedent (CP_OXY_NGML, CP_FBX_NGML, CP_GLASDEGIB_NGML, CP_TUVUSERTIB_NGML), and reuses a single column across two in-vitro assays on different unit scales, as CP_TUVUSERTIB_NGML already does for a patch-clamp bath concentration. Molecular-weight caveat, load-bearing for any mass/molar conversion. Agarwal 2011 reports both potencies in dual units – 3.6 ng/mL = 5.5 nM (implying about 655 g/mol) and 15.9 ug/mL = 25 uM (implying 636 g/mol) – neither of which is the sorafenib free base at 464.8 g/mol. Both are consistent with sorafenib tosylate, 637.0 g/mol, the salt form the authors purchased. A user who converts these ng/mL values to molar units with the free-base weight will disagree with every printed molar figure in the paper by about 37%. The distinct existing model Jain_2011_sorafenib.R is an oral popPK model in patients and does not use this column (it computes its own plasma concentration), so there is no collision. No mouse or human sorafenib PK model in the registry drives these curves; a downstream user must supply the concentration trajectory. Ratified canonically alongside the Agarwal 2011 sorafenib transporter extraction, as a well-formed member of the auto-approved CP_<drug>_<units> family.

AUC_BTZ043TOT (canonical for BTZ-043total (BTZ-043 plus metabolite M2) AUC over the 24 h dosing interval)

• Description: Area under the plasma concentration-time curve over the once-daily 24 h dosing interval for BTZ-043total – the sum of the benzothiazinone antituberculosis drug BTZ-043 and its metabolite M2, reported as BTZ-043 mass equivalents. The analyte is defined by the assay: M2 cannot be quantified directly, so the bioanalytical method back-converts all M2 to BTZ-043 under acid conditions and measures the sum, and M2 is then obtained by difference against a second assay (Koele 2025 Methods ‘PK analysis’). Per-subject and time-fixed within the exposure-response analysis: Koele 2025 derives it by post-hoc estimation from the companion population PK model on the intensive-sampling day (Day 12 for Stage 1, Day 14 for Stage 2) and carries it into the PD run as a data column.
• Units: ng*h/mL. Document per model via covariateData[[AUC_BTZ043TOT]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – a continuous exposure regressor entering an Emax function, Emax * AUC_BTZ043TOT / (EC50 + AUC_BTZ043TOT), so the coefficient scale is set by the EC50 of 16 900 ng/mL*h. Reference values observed: Koele 2025 tabulates no AUC distribution, but states that the estimated EC50 is “similar to the exposure obtained after a 500 mg BTZ-043 dose following a standard breakfast”, which is the anchor to use when choosing a simulation range.
• Source aliases:
  • EXMET – Koele 2025 supplementary pharmacodynamic control-stream $INPUT list and data dictionary (;EXMET = Exposure metric).
• Example models: Koele_2025_btz043_bacterialload.R (drives the first-phase slope of the bilinear decline in sputum bacterial load for both the log10 cfu and the log10 TTP endpoints through a single shared EC50; Koele 2025 Table 3 – founding example).
• Notes: Specific scope: the column names one drug’s composite analyte and the once-daily 24 h steady-state convention. Member of the AUC_<DRUG> family (AUC_RTV, AUC_ANV, AUC_VAREN, AUC_LEN, …). The TOT token is load-bearing and the units are mass-based even though the companion PK model works in molar units. Koele 2025 also fitted the same exposure-response with M2 AUC alone (dOFV -13.1 versus -12.1 for BTZ-043total, no significant difference, attributed to the high correlation between the two exposures) and with M1 AUC (no relationship identified), and selected BTZ-043total because M2 cannot be measured directly; a future model driven by the parent alone or by M2 alone should register a parallel canonical rather than overload this one. To build the column from modellib('Koele_2025_btz043'), integrate Cc + Cc_m2 (nmol/L) over the 24 h interval and multiply by 431.39/1000 = 0.43139 ng per nmol.

AUC_MBV_SS (canonical for maribavir steady-state AUC over the dosing interval on the last day of exposure)

• Description: Area under the maribavir plasma concentration-time curve at steady state, computed over the dosing interval on the last day of exposure, used as the drug-exposure covariate in exposure-efficacy logistic regressions. Derived per subject from the empirical-Bayes individual parameters of a maribavir population PK model rather than measured directly, so this column is supplied as data and the exposure-response model carries no PK layer.
• Units: ug*h/mL (= mg*h/L). Sun 2025 estimates the coefficient per increment of 10 ug*h/mL, so a model using this column divides by 10 at the call site; document the per-model convention in covariateData[[AUC_MBV_SS]]$notes.
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as an uncentred linear regressor on the logit, slope * (AUC_MBV_SS / 10). Calibration: the AURORA maribavir arm median steady-state AUC is ~170 ug*h/mL at the 400 mg twice-daily dose (Sun 2025 Table 3), and Figure 3 draws the fitted curve across roughly 50-600 ug*h/mL.
• Source aliases:
  • AUCss of maribavir – printed row label in Sun 2025 Table S3 and Figure 3, defined in the figure legend as “area under the plasma concentration-time curve on the last day of exposure”.
• Example models: Sun_2025_maribavir_cmv_clearance_wk8.R (slope +0.00920 per 10 ug*h/mL, p = 0.503), Sun_2025_maribavir_cmv_clearance_wk16.R (slope -0.0131 per 10 ug*h/mL, p = 0.247).
• Notes: Distinct from AUC_MBV_DAY, its sibling in the same paper, and the distinction is load-bearing: AUC_MBV_SS is a steady-state interval AUC used for the two efficacy endpoints, whereas AUC_MBV_DAY is the AUC on the calendar day of a specific adverse event and is used for the fourteen safety endpoints. Substituting one for the other silently changes both the exposure metric and the analysis population time-point. Follows the AUC_<DRUG> sibling family (AUC_GCV, AUC_RTV, AUC_LCM, AUC_CBZ, …), with a metric qualifier appended because this drug needs two distinct AUC definitions in one paper. Downstream users should compute the column from a maribavir popPK model in the library – Sun_2025_maribavir.R is the model Sun 2025 itself used.

AUC_MBV_DAY (canonical for maribavir AUC on the calendar day of an adverse event)

• Description: Area under the maribavir plasma concentration-time curve over the 24 h of the calendar day on which a treatment-emergent adverse event occurred, used as the drug-exposure covariate in exposure-safety logistic regressions. For subjects who never experienced the event, the source analysis uses the exposure over the corresponding on-treatment day; where a subject had more than one event of the same type, the exposure on the day of the most severe event was retained (Sun 2025 Methods 2.4). Derived per subject from the empirical-Bayes individual parameters of a maribavir population PK model, so this column is supplied as data.
• Units: ug*h/mL (= mg*h/L). Sun 2025 estimates every coefficient per increment of 10 ug*h/mL, so a model using this column divides by 10 at the call site.
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as an uncentred linear regressor on the logit, slope * (AUC_MBV_DAY / 10). Calibration: the Figure S3 quartile boundaries across the fourteen panels run from roughly [0, 116), [116, 201), [201, 313) to [313, 951] ug*h/mL, so the observed range spans about an order of magnitude.
• Source aliases:
  • AUCday of maribavir – printed row label in every Sun 2025 Figure S3 parameter table, defined in the figure footnote as “area under the curve on the day of the adverse event”.
• Example models: Sun_2025_maribavir_dysgeusia.R, Sun_2025_maribavir_nausea.R, Sun_2025_maribavir_vomiting.R, Sun_2025_maribavir_diarrhea.R, Sun_2025_maribavir_neutropenia.R, Sun_2025_maribavir_immunosuppressant_increase.R, Sun_2025_maribavir_infection.R, Sun_2025_maribavir_gvhd.R, Sun_2025_maribavir_renal_disorder.R, Sun_2025_maribavir_anemia.R, Sun_2025_maribavir_pyrexia.R, Sun_2025_maribavir_headache.R, Sun_2025_maribavir_thrombocytopenia.R, Sun_2025_maribavir_sae.R (all fourteen carry a POSITIVE slope, between +0.0374 and +0.0769 per 10 ug*h/mL, and all fourteen are statistically significant).
• Notes: Distinct from AUC_MBV_SS – see that entry. Note the apparent tension with the paper’s own abstract, which reports that steady-state exposures were “not significantly associated with any adverse events other than nausea and vomiting”: that statement is about AUC_MBV_SS (main-paper Figure 4), whereas all fourteen AUC_MBV_DAY relationships in Figure S3 are significant. The two metrics answer different questions – a day-of-event AUC is partly a consequence of when in the treatment course the event happened, so it is the weaker causal claim of the two. Follows the AUC_<DRUG> sibling family with a metric qualifier.

DOSE_NEMOLIZUMAB_MG (canonical for per-arm or per-subject assigned nemolizumab dose)

• Description: Assigned nemolizumab (anti-interleukin-31 receptor A monoclonal antibody) dose in mg per subcutaneous administration; 0 when the arm/subject did not receive nemolizumab (placebo or a comparator arm).
• Units: mg (per administration)
• Type: continuous
• Scope: specific
• Reference category: n/a – 0 means “this arm/subject did not receive nemolizumab”. Consumed as (DOSE_NEMOLIZUMAB_MG > 0) in both founding models, because Takechi 2025 found no dose-response over the studied range and modelled the maximum effect as a constant.
• Source aliases:
  • Treatment arm / dosing regimen – Takechi 2025 Table 3 (Yokozeki et al., Staender et al., Kwatra et al. rows) and Supplementary Table S1.
• Example models: Takechi_2025_nemolizumab_ppnrs.R (per-subject; 0 = placebo, 30 = 30 mg Q4W after a 60 mg loading dose, 60 = 60 mg Q4W), Takechi_2025_nemolizumab_mbma_iga.R (per study arm; the pooled “30 mg, 60 mg Q4W” arms of OLYMPIA 1 and OLYMPIA 2 and the weight-based 0.5 mg/kg Q4W arm of SPR.115828 are all recorded as presence only), Takechi_2025_nemolizumab_mbma_ppnrs.R.
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family. Presence indicator in practice: neither founding model reads the magnitude, so a value of 30 and a value of 60 give identical predictions; the column is kept numeric rather than binary so the randomised arm is recorded faithfully and a future dose-response extension has somewhere to read the dose from. Two of the Takechi 2025 MBMA arms are not clean flat milligram doses – OLYMPIA 1 and OLYMPIA 2 pooled their 30 mg and 60 mg arms into a single reported arm, and SPR.115828 dosed 0.5 mg/kg – so a dose-response extension must go back to the individual trial reports rather than trusting a single number in this column. Sibling of DOSE_DUPILUMAB_MG, the comparator in the same meta-analysis.

DOSE_DUPILUMAB_MG (canonical for per-arm assigned dupilumab dose)

• Description: Assigned dupilumab (anti-interleukin-4 receptor alpha monoclonal antibody) dose in mg per subcutaneous administration; 0 when the arm/subject did not receive dupilumab.
• Units: mg (per administration)
• Type: continuous
• Scope: specific
• Reference category: n/a – 0 means “this arm did not receive dupilumab”. Consumed as (DOSE_DUPILUMAB_MG > 0) in the founding model.
• Source aliases:
  • Dupilumab dose – Takechi 2025 Table 3 (Yosipovitch et al. PRIME and PRIME2 rows), “300 mg, Q2W”.
• Example models: Takechi_2025_nemolizumab_mbma_iga.R (per study arm; the only administered value in the meta-analysis is 300 mg Q2W, so no dose-response is identifiable and the column acts as a presence indicator selecting the dupilumab-specific Edrug and Kd), Takechi_2025_nemolizumab_mbma_ppnrs.R.
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family. Every included dupilumab arm used the same 300 mg Q2W regimen, so this column carries only two distinct values (0 and 300) across the founding meta-analysis and the drug effect is estimated as a constant. Note the DIFFERENT dosing interval from its sibling DOSE_NEMOLIZUMAB_MG (Q2W vs Q4W): the two columns are not a common per-administration dose metric and must not be compared to each other numerically. Dupilumab appears elsewhere in this package as an atopic-dermatitis comparator; a model that needs an atopic-dermatitis dupilumab regimen must confirm the interval before reusing this column.

CONC_LUMEFANTRINE_168H (canonical for the individual capillary whole-blood lumefantrine concentration at the 168-hour (day 7) landmark from an upstream population PK fit)

• Description: Individual capillary whole-blood concentration of the antimalarial lumefantrine at the fixed 168 h (day 7) landmark after the first of the six twice-daily artemether-lumefantrine doses, carried as a per-episode exposure metric that drives a downstream exposure-response model. Member of the CONC_ family in the drug-and-landmark-specific form CONC_<ANALYTE>_<LANDMARK> documented under [[CONC_SOTRO_168H]] – the analyte and the landmark hour are both carried in the name rather than out-of-band in per-model notes. Not an observation in the downstream model: it is either the measured day 7 sample or the value read off the subject’s own predicted profile from the companion population PK fit, depending on which the source analysis used.
• Units: ng/mL
• Type: continuous
• Scope: specific
• Reference category: n/a as a continuous term. In the founding model it enters only through a dichotomy at 200 ng/mL whose reference stratum is at or above the cut; the threshold itself is carried as a separate fixed() model parameter rather than being baked into this column, following the Rayner_2013_oseltamivir_shedding.R precedent. Reference values observed (Tchaparian 2016): measured median 216 ng/mL (IQR 136-345, n = 216); model-predicted median 202.4 ng/mL (IQR 143.1-321.6, 95% prediction interval 55.0-589.5).
• Source aliases:
  • day 7 capillary whole-blood lumefantrine concentration – Tchaparian 2016 Methods and Table 3 row label, written as Day 7 lumefantrine concentration <200 ng/mL once dichotomised.
  • Day 7 concentration (ng/mL) – Tchaparian 2016 Table 2 secondary-parameter row, the model-predicted form.
• Example models: Tchaparian_2016_lumefantrine_recurrence.R (founding example; the exposure driver of the adjusted Cox model for 28-day recurrent parasitaemia, entering as a 200 ng/mL dichotomy with two stratum-specific log hazard ratios gated on trimethoprim-sulfamethoxazole prophylaxis – log(2.97) without prophylaxis and log(0.50) with it – and additionally on a continuous natural-log scale in the comparative hr_c7 output, log(0.80) per log unit).
• Notes: The 168 h landmark is nominal, counted from the first dose. Sources state it indirectly: Tchaparian 2016 Methods give “day 7 (108 hours after the last dose)”, and since the last of the six twice-daily doses falls at 60 h the nominal landmark is 168 h. That paper’s Supplement reports the ACTUAL median recorded collection time as 6.89 days (IQR 6.82-6.94), about 165 h, so a model that needs the measured rather than the nominal time should say so in its own notes. The matrix is load-bearing and is NOT interchangeable with plasma. These are capillary whole-blood concentrations from dried spots on tartaric-acid-pretreated filter paper; Tchaparian 2016 devote a Discussion paragraph to the point that venous plasma concentrations for the same drug run higher, so a plasma-based day 7 lumefantrine concentration must not be substituted into this column without rescaling. Day 7 lumefantrine is the single most widely reported antimalarial exposure metric – the WWARN individual-patient meta-analysis established the same 200 ng/mL threshold – so specific scope here reflects only that one paper has so far used it as a model input column, and promotion to general scope is appropriate once a second does. Sibling landmark on the same analyte: [[CONC_LUMEFANTRINE_72H]]. Do not substitute one landmark for the other: the two coefficients differ by design and the paper’s headline comparison is precisely between them. Ratified 2026-09-21 alongside the Tchaparian 2016 lumefantrine extraction (full INN spelling, consistent with the pre-existing STIM_LUMEFANTRINE_NM for this analyte).

CONC_LUMEFANTRINE_72H (canonical for the individual capillary whole-blood lumefantrine concentration at the 72-hour (day 3) landmark from an upstream population PK fit)

• Description: Individual capillary whole-blood concentration of lumefantrine at the fixed 72 h (day 3) landmark after the first of the six twice-daily artemether-lumefantrine doses, carried as a per-episode exposure metric. Sibling of [[CONC_LUMEFANTRINE_168H]] in the CONC_<ANALYTE>_<LANDMARK> family form documented under [[CONC_SOTRO_168H]]. This landmark sits near the peak of the lumefantrine profile rather than on its terminal decline, which is why it is roughly an order of magnitude larger than the day 7 value and why the two are not interchangeable.
• Units: ng/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – enters on a continuous natural-log scale, referenced in the founding model to the observed median of 2777 ng/mL carried as a fixed() parameter. Reference values observed (Tchaparian 2016): measured median 2777 ng/mL (IQR 1672-4760, n = 187).
• Source aliases:
  • day 3 capillary whole-blood lumefantrine concentration – Tchaparian 2016 Methods and Results.
• Example models: Tchaparian_2016_lumefantrine_recurrence.R (founding example; the exposure driver of the comparative hr_c3 output, e_conc_lumefantrine_72h_haz = log(0.51) per one-unit rise in natural log concentration among children not receiving trimethoprim-sulfamethoxazole prophylaxis).
• Notes: Nominal landmark counted from the first dose; Tchaparian 2016 Methods give it as “day 3 (12 hours after the last dose)” and the last dose is at 60 h. The Supplement reports the ACTUAL median recorded collection time as 2.86 days (IQR 2.80-2.93), about 69 h. Same capillary-whole-blood matrix caveat as the day 7 sibling – see [[CONC_LUMEFANTRINE_168H]] Notes; do not substitute a plasma value. Unlike the day 7 sibling this landmark is NOT dichotomised in the founding source: the paper reports it only per one-unit change in natural log concentration, and no threshold analysis for it is published. The reason the column exists at all is the founding paper’s headline comparison, that a one-log-unit rise in day 3 concentration cuts the 28-day recurrence hazard by 49% against only 20% for day 7, which the authors advance as a case for sampling at the earlier and logistically easier day. Ratified 2026-09-21 alongside the Tchaparian 2016 lumefantrine extraction.

DOSE_WUZHI_MG48H (canonical for cumulative Wuzhi capsule dose over the preceding 48 hours)

• Description: Total Wuzhi capsule dose, in mg, taken during the 48 h before the observation (e.g. 11.25 mg twice daily -> 45). Per-record, time-varying: recomputed at each observation from the Wuzhi dosing history. 0 = no Wuzhi capsule in the window. Wuzhi capsule is the Schisandra sphenanthera extract described under [[CONMED_WUZHI]]; the mg figure is the capsule strength as reported by the source paper (11.25 mg per capsule in the founding example).
• Units: mg per 48 h
• Type: continuous
• Scope: specific
• Reference category: 0 mg (no Wuzhi capsule). In the founding model the dose is binned rather than used continuously: CL/F multiplier 1 at 0 mg, 0.566 below 45 mg, 0.783 at exactly 45 mg, 0.598 above 45 mg.
• Source aliases:
  • WZ / ‘48h cumulative dosage of Wuzhi Capsule’ – used in Chen_2021b_tacrolimus.R (Chen 2021 Table 2 rows WZ<45mg, WZ=45mg, WZ>45mg, WZ=0mg).
• Example models: Chen_2021b_tacrolimus.R (four-bin multiplier on tacrolimus CL/F in Chinese adult kidney transplant recipients).
• Notes: Member of the auto-approved DOSE_<drug>_<units> family; the drug token is the product name, following [[CONMED_WUZHI]], because the herbal preparation has no single INN. The MG48H unit token records the 48-h accumulation window, which is part of the covariate’s definition: the same regimen gives a different value under a 24-h window. Use [[CONMED_WUZHI]] instead when a paper models only presence / absence. The founding model’s bins are non-monotone (the lowest dose gives the largest CL/F reduction), which the paper attributes to an inhibition-then-induction interplay; a simulation should use only the regimens that populate the bins in the source data (22.5, 45, 67.5, 90 and 135 mg per 48 h).

Count / Markov-feedback PD covariates

These columns are specific to count / Markov / time-to-event PD models that fit per-record event counts (e.g., daily or monthly seizure counts) with optional dependence on the previous-period count. Register names retain their source-paper conventions where those names are unambiguous and readable.

TRT_PHASE (canonical for double-blind active-treatment-phase indicator)

• Description: 1 = the record falls within the active double-blind treatment phase (drug + placebo effects are switched on); 0 = baseline / run-in / off-treatment (drug + placebo effects are zeroed).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (baseline / off-treatment).
• Source aliases:
  • Q2 – used in the Schoemaker 2018 LEV / BRV pediatric extrapolation (DDMODEL00000239) as the treatment-phase gating multiplier on the combined placebo + drug-effect log-rate term (LE = LS0 + Q2 * LTRTE).
• Example models: Schoemaker_2018_levetiracetam.R (DDMODEL00000239).
• Notes: New canonical because the source name Q2 collides with the canonical PK parameter q2 (inter-compartmental clearance to peripheral2) – q2 could not be used as a covariate column without confusing source-trace lookups. Useful for any phase-gated PD model where placebo and drug effects are constrained to a specific study period; analogous to a “treatment-on” indicator.

PDV (canonical for previous-period observed count (Markov-feedback state))

• Description: Previous-period observed count, supplied per-record as a covariate input to capture Markov dependence on the previous count. For daily-count datasets PDV is the observation on the immediately preceding day; for monthly (or other aggregate) records PDV is the observed count from the immediately preceding interval. The biological identity of the count is paper-specific (seizures in Ahn 2010 / Schoemaker 2018, contrast-enhancing MS lesions in Velez de Mendizabal 2013); the canonical concept is “the immediately preceding observed count, used as a Markov-state covariate”.
• Units: (count, non-negative integer)
• Type: count
• Scope: specific
• Reference category: n/a – typically used as <param> * PDV / (ES50 + PDV) (Markov-Hill) or as a linear coefficient <param> * PDV in additive-mean Markov models (Velez de Mendizabal 2013 equation 5).
• Source aliases: PDV – used in the Schoemaker 2018 LEV/BRV pediatric extrapolation (DDMODEL00000239), in the Ahn 2010 pregabalin Markov seizure-count model that the Schoemaker 2018 publication cites as the precursor structure, and in the Velez de Mendizabal 2013 MS CEL-count NB nested MAK2 model.
• Example models: Schoemaker_2018_levetiracetam.R (DDMODEL00000239), VelezdeMendizabal_2013_multipleSclerosis.R (additive-mean Markov form, equation 5), Chan_2020_fenebrutinib_acr.R (a BINARY 0/1 count: the patient’s responder status at the previous visit for the same ACR threshold as the current record, entering the logit linearly as 0.934 * PDV; 0 at the first post-baseline visit).
• Notes: Specific scope because the column’s interpretation (a Markov feedback state) is intrinsic to count-likelihood Markov-feedback PD models. nlmixr2 / rxode2 cannot natively express a Markov dependence of an observation on the immediately preceding observation as a model state – supplying PDV as a per-record data column is the way the maintainers approved to preserve the published structure. For records where the Markov term should not contribute (e.g., monthly counts in a mixed daily / monthly cohort, or the first monthly observation per subject) two conventions exist: (a) the sentinel value -99, used in Schoemaker_2018_levetiracetam.R where the model gates the Markov contribution on CHILD = 1 so the sentinel is multiplied by zero and is harmless; (b) the natural zero value PDV = 0, used in VelezdeMendizabal_2013_multipleSclerosis.R for the first per-subject observation, which makes the Markov contribution exactly zero without any indicator gating. Pick whichever convention matches the source paper’s encoding for the model at hand. The companion second-order Markov-state covariate is [[PPDV]] (the observed count two periods prior); see the PPDV entry for the second-order form.

PPDV (canonical for second-prior-period observed count (second-order Markov-feedback state))

• Description: Observed count from two periods prior, supplied per-record as a covariate input to capture a second-order Markov dependence on the count two intervals before the current one. For monthly-count datasets PPDV is the observed count from two monthly MRIs ago; for daily-count datasets it would be the observation from two days prior if a second-order term were used. Companion to the first-order [[PDV]] state.
• Units: (count, non-negative integer)
• Type: count
• Scope: specific
• Reference category: n/a – typically used as a linear coefficient <param> * PPDV in additive-mean Markov-count models (Velez de Mendizabal 2013 equation 5).
• Source aliases: PPDV – used in the Velez de Mendizabal 2013 MS CEL-count NB nested MAK2 model (paper notation “PPDV” = “previous-previous DV”, paralleling PDV = “previous DV”).
• Example models: VelezdeMendizabal_2013_multipleSclerosis.R (equation 5 second-order Markov term; theta_PPDV = 0.150, materially smaller than theta_PDV = 0.447 per the source paper’s decreasing-pattern observation across Markov orders).
• Notes: Specific scope because the column’s interpretation (a second-order Markov feedback state) is intrinsic to count-likelihood Markov-feedback PD models. Set PPDV = 0 at the first two per-subject observations (no second-prior interval) so the second-order Markov contribution is exactly zero; this matches the convention adopted in VelezdeMendizabal_2013_multipleSclerosis.R. The Velez de Mendizabal 2013 paper also explored a third-order Markov term PPPDV – “previous-previous-previous DV”, coefficient theta_PPPDV – but the third-order fit improvement was not statistically significant, and the third-order column is not registered until / unless a future model retains it.

NDAYS (canonical for number of days in the count-record interval)

• Description: Number of days in the interval over which the observed seizure count was tabulated. Multiplies the per-day rate to give the expected count for the record (e.g., LAMB = exp(LE) * NDAYS).
• Units: days
• Type: count
• Scope: general
• Reference category: n/a – appears as a multiplier on a per-day rate.
• Source aliases: NDAYS – used in the Schoemaker 2018 LEV/BRV pediatric extrapolation (DDMODEL00000239).
• Example models: Schoemaker_2018_levetiracetam.R (DDMODEL00000239).
• Notes: General scope because the count-interval-length concept is shared across any count or rate-based PD model that mixes record granularity (e.g., daily and monthly counts in the same dataset). For pure-daily or pure-monthly cohorts the column is constant; including it as a covariate keeps the model usable in mixed-granularity simulations.

MOMENT (canonical for endotracheal suctioning procedural state)

• Description: Procedural state at the time of the pain / distress assessment in invasive-ventilation studies, used to select between separate pre-procedure / intra-procedure / post-procedure baseline parameters. Coded 1 = before suctioning, 2 = during suctioning, 3 = after suctioning.
• Units: (categorical, 3 levels)
• Type: categorical
• Scope: specific
• Reference category: 1 (before suctioning).
• Source aliases: MOMENT – used in the Valitalo 2017 IRT morphine PD model (DDMODEL00000247).
• Example models: Valitalo_2017_morphine.R (DDMODEL00000247; selects between the three baseline pain typical values presuct / suct / aftsuct and the matching 3x3 correlated etas).
• Notes: Specific scope because the column’s meaning is tied to a particular study procedure (endotracheal suctioning during mechanical ventilation in neonates). The Simons 2003 cohort that Valitalo 2017 re-analysed scheduled pain assessments around suctioning events, so MOMENT changes within-subject at each scheduled assessment.

ITEM (canonical for pain-assessment item identifier in IRT graded-response models)

• Description: Identifier of the specific pain-assessment item being scored at each observation row, used to dispatch between the IRT discrimination / difficulty parameter sets in a graded-response model. Valitalo 2017 coding: 1 = COMFORT-B alertness; 2 = COMFORT-B calmness/agitation; 3 = COMFORT-B respiratory response; 5 = COMFORT-B body movement; 7 = COMFORT-B facial tension; 12 = VAS (cm, range 0-10); 25 = PIPP brow bulge; 26 = PIPP eye squeeze; 27 = PIPP nasolabial furrow; 28 = NIPS total.
• Units: (categorical, 9-10 levels depending on cohort)
• Type: categorical
• Scope: specific
• Reference category: n/a – selects per-item parameter sets rather than acting as a reference contrast.
• Source aliases: ITEM – used in the Valitalo 2017 IRT morphine PD model (DDMODEL00000247).
• Example models: Valitalo_2017_morphine.R (DDMODEL00000247; switches the IRT graded-response discrimination / difficulty parameters per row).
• Notes: Specific scope because the integer-to-item mapping is tied to the Valitalo 2017 NM-TRAN dataset’s coding. Other IRT graded-response models in the library (when they are added) may use different integer codings and should register their own canonical (e.g., ITEM_<study>) when the codings collide. The COMFORT-B “muscle tension” item is omitted from the Valitalo 2017 coding because it could not be assessed from video recordings.

OBSTYPE (canonical for VAS observer type)

• Description: Observer type for visual-analogue-scale pain assessments: 1 = investigator (video-based), 2 = bedside nurse. Used to select between observer-specific VAS difficulty / discrimination THETAs and between observer-specific residual-error SDs.
• Units: (categorical, 2 levels)
• Type: categorical
• Scope: specific
• Reference category: 1 (video investigator).
• Source aliases: OBSTYPE – used in the Valitalo 2017 IRT morphine PD model (DDMODEL00000247).
• Example models: Valitalo_2017_morphine.R (DDMODEL00000247; selects between diff_vas_video vs diff_vas_bedside, discr_vas_video vs discr_vas_bedside, and addSd_vas_video vs addSd_vas_bedside).
• Notes: Specific scope because the binary observer-coding is tied to the Valitalo 2017 NM-TRAN dataset. Future IRT models with a different observer split (e.g., parent / nurse / physician) should register their own canonical rather than overloading this 2-level coding.

Race / ethnicity

Canonical pattern: RACE_<GROUP>. Use one indicator per race/ethnicity group the source models. Reference category is the implicit 0 = all other groups; document explicitly which groups are in the reference. When the source uses composite groups (e.g., “Black or Other”), name them accordingly (RACE_BLACK_OTHER) and list the components in notes. The base RACE_<GROUP> indicators are scope: general; composite groupings are scope: specific because the grouping is tied to the study’s analysis plan.

RACE_BLACK (canonical for Black / African American race indicator)

• Description: 1 = Black / African American, 0 = other.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (document the actual reference groups used).
• Source aliases:
  • BLACK – used in Hu_2026_clesrovimab.R, Robbie_2012_palivizumab.R.
• Example models: Zhu_2017_lebrikizumab.R (canonical form), Robbie_2012_palivizumab.R, Zhou_2021_remimazolam.R (multiplicative ratios 0.87 on CL and 0.839 on all three volumes versus White and Asian pooled; source RACE.EQ.1), Wada_2023_sparsentan.R (log-additive effect on the apparent central volume: exp(0.309 * RACE_BLACK), i.e. 36% higher Vc/F than the White reference; paired with RACE_ASIAN so both = 0 selects White).

RACE_WHITE (canonical for White race indicator)

• Description: 1 = White, 0 = non-White.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (non-White; complement composition depends on the source paper, typically pooling Black/African American, Asian, American Indian/Alaska Native, Native Hawaiian/Pacific Islander, Other, Not reported, Unknown). Some papers (e.g., Hu 2014) instead use the Caucasian (RACE_WHITE = 1) subgroup as the typical-value reference; the column encoding is unchanged but the model implements the effect on (1 - RACE_WHITE).
• Source aliases:
  • RACE (with values 1 = White / 0 = non-White) – used in Lin_2024_casirivimab.R. Source column name RACE is generic; the canonical name is intentionally explicit because some other models use RACE for a different dichotomy.
  • RACE (Caucasian-vs-non-Caucasian dichotomy as named in Hu 2014 Table 2) – used in Hu_2014_bapineuzumab.R. Same canonical column name and 1 = White / 0 = non-White encoding; the typical-value reference is the Caucasian subgroup, so the model implements the 15% non-Caucasian effect on (1 - RACE_WHITE).
• Example models: Lin_2024_casirivimab.R (multiplicative fractional change on CL relative to non-White reference), Hu_2014_bapineuzumab.R (multiplicative 15% increase in CL for non-Caucasian relative to Caucasian reference).
• Notes: Used by papers that dichotomize race as White vs. non-White rather than decomposing into separate group indicators. Sign and reference-category interpretation are inverted relative to RACE_BLACK / RACE_ASIAN / etc.; do NOT combine RACE_WHITE with the decomposed indicators in the same model. The model’s typical-value reference category (which subgroup gets the unmodified lcl / lvc) varies between papers – Lin 2024 uses non-White as the reference, Hu 2014 uses Caucasian (White) as the reference; both share the same canonical column encoding.

RACE_BLACK_OTH (canonical for composite Black/Other group)

• Description: 1 = Black/African American or Other race, 0 = other groups.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 = White or Native Hawaiian/Pacific Islander (Clegg 2024 grouping).
• Source aliases: BLACK_OTH – used in Clegg_2024_nirsevimab.R.
• Example models: Clegg_2024_nirsevimab.R.
• Notes: Kept distinct from RACE_BLACK because the composite is not interchangeable.

RACE_BLACK_HISPANIC (canonical for composite Black / Hispanic American group)

• Description: 1 = subject self-identifies as Black / African American or as Hispanic American; 0 = White or Asian. Paper-specific composite developed by Thoueille 2023, which first fitted a separate tenofovir clearance per ethnic group (rich model) and then pooled the levels into this two-way split, retained because the reduced grouping cost only dOFV = +2 relative to the full ethnicity model (P > 0.05).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 = White or Asian (Thoueille 2023 grouping; the paper’s model-building cohort contained only White 71%, Black 23%, Hispanic American 4% and Asian 2%, with no “Other” level, so the indicator and its reference exactly partition the cohort).
• Source aliases:
  • Paper’s ethnicity levels “Black” + “Hispanic American” – decompose into RACE_BLACK_HISPANIC = as.integer(ETHNICITY %in% c("Black", "Hispanic American")). Thoueille 2023 Table 2 and Table S3 label the coefficient row “theta_Black or Hispano American” (the papers’s spelling of “Hispanic”).
• Example models: Thoueille_2023_tenofovir_full.R, Thoueille_2023_tenofovir_alafenamide.R (multiplicative fractional effect on apparent tenofovir clearance: (1 + e_race_black_hispanic_cl * RACE_BLACK_HISPANIC) with e_race_black_hispanic_cl = 0.119 and 0.122 respectively, i.e. ~12% higher CL/F than the White / Asian reference).
• Notes: Mutually exclusive with the decomposed constituents [[RACE_BLACK]] and [[RACE_HISPANIC]] – a model uses either this composite or the two separate indicators, never both, because the composite carries a single pooled coefficient that the constituents would double-count. Distinct from [[RACE_BLACK_OTH]] (Black or Other, reference White / Native Hawaiian-Pacific Islander) and, importantly, from [[RACE_NONBLACK_NONWHITE]], which places Black in the reference group and therefore carries the opposite sign convention. Specific scope: the grouping is paper-specific and reflects the empirical clustering of Thoueille 2023’s ethnicity covariate step rather than any biological homogeneity across the two constituent populations; future papers that split Black from Hispanic, or that add an “Other” level, should register their own composite rather than overload this one.

RACE_ASIAN (canonical for Asian race indicator)

• Description: 1 = Asian, 0 = other.
• Units: (binary)
• Type: binary
• Scope: general
• Source aliases: ASIAN – used in Hu_2026_clesrovimab.R, Robbie_2012_palivizumab.R, Fau_2020_isatuximab.R. RAAS (race-Asian-vs-other indicator as named in Bajaj 2017 Table 1) – used in Bajaj_2017_nivolumab.R. RACEN (race-numeric indicator as named in Lu 2019 / Shi 2020 NONMEM control stream; ASIAN = 1 if RACEN == 1) – used in Lu_2019_polatuzumab.R.
• Example models: Zhu_2017_lebrikizumab.R (canonical form), Robbie_2012_palivizumab.R, Bajaj_2017_nivolumab.R, Fau_2020_isatuximab.R, Lu_2019_polatuzumab.R (multiplicative factor e_asian_vc = 0.929 on acMMAE Vc, i.e., 7.1% lower V1 in Asian patients; verbatim Shi 2020 (PMID 32770353) ethnicity-sensitivity re-quote of the Lu 2019 popPK Asian-race covariate), Wada_2023_sparsentan.R (log-additive effect on the apparent central volume: exp(0.265 * RACE_ASIAN), i.e. 30% higher Vc/F than the White reference; paired with RACE_BLACK so both = 0 selects White).

RACE_ASIAN_AMIND_OTH (canonical for composite Asian / American Indian / Other group)

• Description: 1 = Asian, American Indian / Alaska Native, or Other race; 0 = White or Black. Composite indicator that pools the smaller-N race groups in a population dominated by White and Black subjects, with White + Black serving as the reference category. Distinct from RACE_ASIAN_AMIND_MULTI (Clegg 2024 grouping that includes Multiracial; pooled against a different reference), from RACE_ASIAN_OTH (within-Asian-population sub-indicator), and from RACE_BLACK_OTH (different composite).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 = White or Black (the larger-N pooled group used as the reference in the source paper).
• Source aliases: none formally; Frey 2013’s NONMEM control stream uses the inline race classification rather than a separate named column.
• Example models: Frey_2013_tocilizumab.R (multiplicative fractional effect on the DAS28 first-order loss rate Kout: Kout * (1 - 0.25 * RACE_ASIAN_AMIND_OTH) – Kout is 25% lower in the Asian/AmInd/Other composite group relative to the White+Black reference).
• Notes: Specific scope because the composite grouping is defined by the source paper’s analysis plan rather than by a uniform external standard. Do not combine with the decomposed RACE_ASIAN, RACE_OTHER, etc. indicators in the same model; the composite indicator is mutually exclusive with the decomposition. Frey 2013 uses TWO distinct race covariates: this RACE_ASIAN_AMIND_OTH indicator on Kout (DAS28-PD-side; pools Asian + AmInd + Other vs White+Black) AND the within-Asian RACE_ASIAN_OTH indicator on CL (PK-side; isolates the “Other Asian” subgroup within the Asian-only cohort).

RACE_ASIAN_AMIND_MULTI (canonical for composite group)

• Description: 1 = Asian, American Indian / Alaskan Native, or Multiple races, 0 = other.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 = White, Black / African American, Native Hawaiian / Pacific Islander, or Other (Clegg 2024 grouping).
• Source aliases: ASIAN_AMIND_MULTI – used in Clegg_2024_nirsevimab.R.
• Example models: Clegg_2024_nirsevimab.R.
• Notes: Clegg 2024 applies this covariate to both CL and V2 with different coefficients.

RACE_ASIAN_OTH (canonical for Asian-other composite race indicator)

• Description: 1 = subject self-identifies as Asian-other (Asian heritage outside the locally-dominant Asian subgroup, e.g. non-Chinese in a Chinese-dominant cohort, or “Other Asian” as a study-defined catch-all category). 0 = otherwise. Reference category is the cohort’s dominant race grouping (typically Chinese or White, depending on the study).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0. Document the dominant subgroup explicitly in covariateData[[RACE_ASIAN_OTH]]$notes for every model that uses this covariate.
• Source aliases: none.
• Example models: Brown_2017_osimertinib.R (paper’s “Asian (not Japanese or Chinese)” composite indicator with linear effect on apparent clearance of the AZ5104 metabolite; reference category Caucasian), Johnson_2025_osimertinib.R (the same osimertinib program’s AURA3 + FLAURA update; paper’s “Asian (non-Chinese, non-Japanese)” arm of a five-level grouped race covariate, with an exponential effect exp(0.182 * RACE_ASIAN_OTH) on apparent clearance of the AZ5104 metabolite and reference category White. Retained as Scope: specific rather than promoted to general: Brown 2017 and Johnson 2025 are the predecessor and successor analyses of one clinical program, so they are not two independent papers evidencing a general grouping), Inoue_2025_valemetostat.R (multiplicative effect on unbound valemetostat clearance, 1.25 – the largest non-AAG covariate effect in that model), Inoue_2025_valemetostat_orr_bicr.R, Inoue_2025_valemetostat_anemia.R, Inoue_2025_valemetostat_neutropenia.R, Inoue_2025_valemetostat_thrombocytopenia.R, Inoue_2025_valemetostat_teae_grade3.R, Inoue_2025_valemetostat_dose_reduction.R, Inoue_2025_valemetostat_dose_interruption.R, Yang_2025_osimertinib.R (the FLAURA2 six-study update of the same osimertinib program; paper’s “Asian (excluding Chinese and Japanese)” arm of a five-level race covariate, ETHL = 1 in the source NONMEM control stream, with linear effect (1 + 0.161274 * RACE_ASIAN_OTH) on apparent AZ5104 clearance and reference category White), Li_2019_ceftazidime.R (Li 2019’s RCE = 3 level, ‘non-Chinese, non-Japanese Asian’, against a pooled non-Asian reference; carries a -0.161 proportional clearance shift of its own AND a share of the -0.27 pooled Asian VOLUME shift that Li 2019 applies to all three of its Asian levels), Li_2019_avibactam.R (the same RCE = 3 level with a -0.0865 clearance shift; the avibactam half of the combination retains ONLY this Asian level, fixing the Black, Chinese, Japanese and Other coefficients to zero, so the two companion models of one fixed-ratio product disagree about which race levels survive covariate selection), Cirincione_2018_apixaban_asian_subgroups.R (the paper’s “other Asian” arm of an ad hoc three-way stratification of a previously single Asian-race effect; linear effect (1 + e_race_asian_oth_cl * RACE_ASIAN_OTH) on total apparent apixaban clearance with coefficient -0.0427, and reference category non-Asian. This is the first ratification whose dominant reference grouping is non-Asian rather than an Asian subgroup, and the first outside the osimertinib / valemetostat oncology programmes; the companion indicators are RACE_JAPANESE and RACE_KOREAN rather than a Chinese-dominant comparator, so “other Asian” here means Asian heritage that is neither Japanese nor Korean, 4.15 percent of the 4,385-subject dataset). Franzese_2022_ceftazidime.R (the pediatric + adult refit of Li_2019_ceftazidime.R: proportional shift of -0.136 on ceftazidime CL, supplementary Table S2 theta18, plus the pooled Asian Vc shift of -0.135, theta23; reference the non-Asian grouping; the avibactam companion holds its race thetas at 0, so Franzese_2022_avibactam.R lists it under covariatesDataExcluded).
• Notes: Distinct from RACE_ASIAN_AMIND_MULTI (a 4-way composite of Asian + American Indian + Multiple Races), RACE_ASIAN_AMIND_OTH (a 3-way Asian + AmInd + Other composite against a White+Black reference, used in Frey 2013), and RACE_BLACK_OTH (different composite). RACE_ASIAN_OTH is a within-Asian-population sub-indicator, not a multi-race composite. Kept separate from RACE_ASIAN by the maintainers’ decision because the paper’s “Other Asian” category is its own grouping, not an alias of “Asian (any)”. Brown 2017 uses Caucasian (not Chinese) as the dominant reference.

RACE_ASIAN_NORTHEAST (canonical for North East Asian composite race indicator)

• Description: 1 = North East Asian heritage (worldwide Chinese, Japanese, or Korean), 0 = non-North East Asian. Composite indicator analogous to RACE_ASIAN but specifically restricted to the East Asian subgroup most-relevant to ICH E5 ethnic-sensitivity / Asian-region bridging analyses.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (any non-North East Asian race, including South / Southeast Asian, White, Black, etc.).
• Source aliases:
  • RACE_NEAS – prior canonical name (pre-2026-06-19 readability standardization).
  • RAC4 – used in Zhou_2021_belimumab.R (Zhou 2021 Table 2 footnote d).
• Example models: Zhou_2021_belimumab.R (multiplicative factor 1.07 on V1), Xiong_2022_tepotinib.R (fractional effect +1.40 on MSC2571109A inter-compartmental clearance; Xiong 2022 groups Japanese with Other East Asian participants, against a non-East Asian reference).
• Notes: Distinct from the broader RACE_ASIAN (which can include South / Southeast Asian populations) because Zhou 2021 specifically tested whether Chinese/Japanese/Korean patients had different PK from the rest of the dataset; the analysis explicitly compared RAC4 (North East Asian) against alternative race definitions and chose RAC4 by AIC.

RACE_ASIAN_EAST_SE (canonical for the East Asian / Japanese / South East Asian composite race indicator)

• Description: 1 = East Asian, Japanese, or South East Asian heritage; 0 = any other race. Composite indicator for a paper-defined grouping that pools the three East/South-East Asian heritage categories of a multi-level race column while deliberately EXCLUDING Central / South Asian heritage, which the same analysis assigns to a separate smaller-N composite.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0, with White/Caucasian as the model’s implicit reference when paired on the same parameter with RACE_BLACK and RACE_ASIAN_CENTRAL_ARABIC_AMIND_OTH (all three indicators = 0 selects White/Caucasian).
• Source aliases:
  • RACE1 = 2 – used in Siederer_2016_fluticasoneFuroate.R (Siederer 2016 Sect. 2.3.1 grouping “RACE1 = 2-East Asian, Japanese, and South East Asian”) and in Mehta_2018_fluticasoneFuroate.R (Mehta 2018 Supplementary Table 1 $PK block IF(RACE1.EQ.2) CLRACE1 = ( 1 + THETA(7)), the same grouping carried forward unchanged).
• Example models: Siederer_2016_fluticasoneFuroate.R (log-additive effect on apparent inhaled clearance: exp(-0.211 * RACE_ASIAN_EAST_SE), i.e. CL/F 19% lower than the White/Caucasian reference, giving a 23-30% higher steady-state AUC0-24 for inhaled fluticasone furoate), Mehta_2018_fluticasoneFuroate.R (same grouping and the same log-additive placement on CL/F, but the coefficient is held at a structural zero: Mehta 2018 refit the Siederer structure on FULFIL pooled with the historical data, states that “a covariate analysis was not planned” and that “the same covariate relationship was assumed”, and reports no RACE1 row in its Table 1 – the deposited control stream’s race $THETA entries are initial estimates seeded from the historical fit).
• Notes: Specific scope because the composite grouping is defined by the source paper’s analysis plan (driven by the small N available in the individual race categories) rather than by a uniform external standard. Distinct from RACE_ASIAN (which also includes Central / South Asian), from RACE_ASIAN_NORTHEAST (Chinese / Japanese / Korean only – excludes South East Asian, includes Korean), and from RACE_JAPANESE (Japanese heritage alone). Do not combine with the decomposed RACE_JAPANESE / RACE_ASIAN indicators in the same model; the composite is mutually exclusive with its decomposition.

RACE_ASIAN_CENTRAL_ARABIC_AMIND_OTH (canonical for the Central Asian / White-Arabic / American Indian / Other composite race indicator)

• Description: 1 = Central or South Asian heritage, White-Arabic / North African heritage, American Indian / Alaska Native, or race category “other”; 0 = any other race. Composite indicator that pools the residual smaller-N race categories of a multi-level race column in an analysis that already carries separate indicators for its two largest non-reference groups.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0, with White/Caucasian as the model’s implicit reference when paired on the same parameter with RACE_BLACK and RACE_ASIAN_EAST_SE (all three indicators = 0 selects White/Caucasian).
• Source aliases:
  • RACE1 = 4 – used in Siederer_2016_fluticasoneFuroate.R (Siederer 2016 Sect. 2.3.1 grouping “RACE1 = 4-Asian Central, White Arabic, American Indian/Native Alaskan, and ‘other’”) and in Mehta_2018_fluticasoneFuroate.R (Mehta 2018 Supplementary Table 1 $PK block IF(RACE1.EQ.4) CLRACE1 = ( 1 + THETA(9)), the same grouping carried forward unchanged).
• Example models: Siederer_2016_fluticasoneFuroate.R (log-additive effect on apparent inhaled clearance: exp(-0.265 * RACE_ASIAN_CENTRAL_ARABIC_AMIND_OTH), i.e. CL/F 23% lower than the White/Caucasian reference; the coefficient is imprecise – %RSE 50.6, 95% CI -0.528 to -0.002 – because the group is under 2% of the dataset), Mehta_2018_fluticasoneFuroate.R (same grouping and the same log-additive placement on CL/F, with the coefficient held at a structural zero because Mehta 2018 re-used the Siederer covariate structure without re-estimating or reporting it – see the RACE_ASIAN_EAST_SE entry for the full reasoning).
• Notes: Specific scope because the composite grouping is paper-defined. Distinct from RACE_NONBLACK_NONWHITE (which pools ALL non-White non-Black races including East Asian, and takes White + Black as its reference) and from RACE_ASIAN_AMIND_OTH (Frey 2013 grouping, which pools all Asian heritage and takes White + Black as its reference). The distinguishing features here are that East / South East Asian heritage is carved out into RACE_ASIAN_EAST_SE instead, that White-Arabic / North African heritage is pooled in rather than counted as White, and that the reference category is White/Caucasian alone rather than White + Black.

RACE_MULTI (canonical for multiracial indicator)

• Description: 1 = multiracial, 0 = other.
• Units: (binary)
• Type: binary
• Scope: general
• Source aliases: MULTIRACIAL – used in Hu_2026_clesrovimab.R.
• Example models: Hu_2026_clesrovimab.R.

RACE_OTHER (canonical for race-category ‘Other’ indicator)

• Description: 1 = race category “Other,” 0 = not.
• Units: (binary)
• Type: binary
• Scope: general
• Source aliases:
  • OTHER – used in Robbie_2012_palivizumab.R.
• Example models: Zhu_2017_lebrikizumab.R, Robbie_2012_palivizumab.R.

RACE_NONBLACK_NONWHITE (canonical for the non-Black non-White composite race indicator)

• Description: 1 = race is anything other than White or Black (composite that pools Asian + Hispanic + American Indian + Other or analogous smaller-N groups from a multi-level race classification), 0 = White or Black. Composite indicator used by papers that collapse a multi-level race column into a three-level effective structure with White as the reference, Black as a separate indicator, and “non-Black ethnic origin” as the composite.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 = White or Black (the larger-N pooled reference group in the source paper). Pair with RACE_BLACK on the same model so the two indicators are mutually exclusive and the implicit White reference category is reproduced cleanly.
• Source aliases: none formally; Fisher 2008’s NONMEM control stream uses the 6-level inline race classification rather than a separate composite column.
• Example models: Fisher_2008_fosamprenavir.R (multiplicative CL/F factor 1.06 vs the White reference; the composite pools Race = 3 (Asian), 4 (Hispanic), 5 (American Indian), 6 (Other) into a single indicator alongside the separate RACE_BLACK indicator).
• Notes: Specific scope because the composite grouping is paper-defined and is mutually exclusive with the decomposed RACE_ASIAN, RACE_HISPANIC, RACE_AMIND, RACE_OTHER indicators in the same model (do not combine the composite indicator with its constituent decompositions). Distinct from RACE_ASIAN_AMIND_OTH (Frey 2013 grouping; explicitly excludes Hispanic) and from RACE_BLACK_OTH (Clegg 2024 grouping; includes Black in the composite).

RACE_HISPANIC (canonical for Hispanic / Latino ethnicity indicator)

• Description: 1 = Hispanic / Latino, 0 = non-Hispanic. Used by papers that report Hispanic as a separate category alongside Black, Asian, and Other rather than as a distinct ethnicity dimension.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (non-Hispanic; document the paper-specific reference race composition per-model).
• Source aliases:
  • HISPANIC – used in Robbie_2012_palivizumab.R.
  • Hispanic ethnicity indicator – used in Overgaard_2016_liraglutide.R (Overgaard 2016 Fig. 2 / Table S1 “Cov. Hispanic”; the paper reports race and ethnicity as separate dimensions but the column encoding, 1 = Hispanic vs 0 = non-Hispanic, is identical to the canonical form – per-model notes field records the ethnicity-vs-race interpretation).
• Example models: Robbie_2012_palivizumab.R (fractional effect on CL; additional effect on Vc), Overgaard_2016_liraglutide.R (log-scale multiplicative effect on CL/F, coefficient +0.08 per Table S1).
• Notes: In the US Office-of-Management-and-Budget (OMB) classification Hispanic is an ethnicity rather than a race, but clinical PK analyses frequently treat it as one of the race indicators. When a paper treats Hispanic as a race, use this column; otherwise encode ethnicity separately. Register-wise, this follows the RACE_<GROUP> indicator-decomposition pattern. Overgaard 2016 explicitly reports race and ethnicity as separate covariate dimensions but the numerical Hispanic-vs-non-Hispanic encoding matches this canonical – the per-model covariateData[[RACE_HISPANIC]]$notes records the “reported as ethnicity” interpretation.

RACE_HISPANIC_MISSING (canonical for Hispanic / Latino ethnicity-not-reported indicator)

• Description: 1 = the subject’s Hispanic / Latino ethnicity was recorded as Unknown / not reported; 0 = ethnicity was recorded (Hispanic or not Hispanic). The third level of a three-level ethnicity factor whose reference is not-Hispanic-or-Latino, carried as its own indicator rather than collapsed onto that reference.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (ethnicity recorded). Mutually exclusive with [[RACE_HISPANIC]]; both 0 selects the not-Hispanic-or-Latino reference level.
• Source aliases:
  • ethnic: Unknown – printed row label in Chan 2025 Table S10G.
• Example models: Chan_2025_atezolizumab_isr.R (log-odds shift of +2.39 on the probability of an injection site reaction, odds ratio 10.9, 95 percent CI 0.0684-258, p = 0.2559).
• Notes: The ethnicity analogue of [[ADA_MISSING]], and registered for the same reason: when a source model gives the unknown / not-reported level its own coefficient, dropping that coefficient silently redefines the reference group of the sibling indicator from “recorded as not Hispanic” to “not recorded as Hispanic, including unknown”, which biases the sibling coefficient a downstream user reproduces. Retain the column whenever the source model retained the level, even – as in the founding example – when the estimate is wholly uninformative (a confidence interval spanning three orders of magnitude). Interpret any non-zero estimate with caution: a missing-ethnicity indicator is a proxy for a data-collection pattern, not a biological covariate, and the missingness mechanism is rarely stated. Follows the <CONCEPT>_MISSING family convention shared with ADA_MISSING, PDL1_TUM_MISSING, TUM_EGFR_MUT_MISSING, AAG_MISSING and CYP2C9_MISSING. Note that the register carries the ethnicity dimension under the RACE_ prefix by convention – see the [[RACE_HISPANIC]] Notes for why – so this name preserves the sibling relationship even though the underlying variable is an ethnicity rather than a race.

RACE_JAPANESE (canonical for Japanese-heritage race indicator)

• Description: 1 = Japanese heritage, 0 = non-Japanese. Used when Japanese subjects form a distinct subgroup in the study design (e.g., ICH E5 bridging analyses or studies with a dedicated Japanese healthy-volunteer cohort).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (non-Japanese).
• Source aliases:
  • JAPANESE_HV – used in Wang_2017_benralizumab.R (Japanese healthy-volunteer cohort indicator; the healthy-volunteer vs. asthma-patient distinction is captured separately, not in this covariate).
  • RaceJapanese – used in Rini_2013_axitinib.R (Rini 2013 Results “PK model” typical-value clearance equation; Results “Effects of covariates” tested Asian and Japanese separately and in combination and retained only the Japanese indicator).
• Example models: Wade_2015_certolizumab.R (multiplicative fractional effect on V/F; Wade 2015 breaks Japanese [RACE.EQ.8] out separately from RACE_ASIAN), Wang_2017_benralizumab.R (multiplicative factor 1.34 on Vc), Rini_2013_axitinib.R (linear-proportional effect on CL: (1 - 0.249 * RACE_JAPANESE), i.e. 24.9% lower clearance; the Rini 2013 Discussion cautions that the effect may be confounded by the lower body weight and higher age of the Japanese subjects in the pooled dataset, and that a dedicated Caucasian-vs-Japanese healthy-volunteer comparison found no axitinib PK difference).
• Notes: Distinct from RACE_NEAS (North East Asian composite, includes Chinese, Japanese, and Korean) and from RACE_ASIAN. Use RACE_JAPANESE only when the source paper breaks out Japanese heritage as its own indicator; do not aggregate with other Asian groups when the paper keeps them separate.

RACE_CHINESE (canonical for Chinese-heritage race indicator)

• Description: 1 = Chinese heritage, 0 = non-Chinese. Used when Chinese subjects form a distinct subgroup alongside Japanese, Asian-other, and other race categories (e.g., multiregional oncology trials enrolling Chinese, Japanese, and other Asian cohorts as separate strata).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (non-Chinese; document the paper-specific reference race composition per-model).
• Source aliases: none formally; source NONMEM control streams typically use ad-hoc names (e.g., CHINESE, RACE.EQ.X).
• Example models: Brown_2017_osimertinib.R (linear additive effect (1 + 0.17 * RACE_CHINESE) on apparent clearance of the AZ5104 metabolite; reference category Caucasian).
• Notes: Distinct from RACE_NEAS (North East Asian composite, includes Chinese, Japanese, and Korean) and from RACE_ASIAN. Use RACE_CHINESE only when the source paper breaks out Chinese heritage as its own indicator alongside RACE_JAPANESE and RACE_ASIAN_OTH; do not aggregate with other Asian groups when the paper keeps them separate. Parallels the established RACE_JAPANESE entry.

RACE_PAPUAN (canonical for Papuan / indigenous Melanesian Indonesian heritage race indicator)

• Description: 1 = Papuan (indigenous Melanesian heritage native to the Indonesian provinces of Papua / West Papua, the western half of the island of New Guinea), 0 = non-Papuan (typically Indonesian residents of mainland-Indonesian Austronesian origin, e.g., Javanese / Sumatran transmigrants enrolled at Papua-based clinical sites; possibly also any non-Indonesian subjects in mixed cohorts). Used as a binary race / ethnicity indicator in clinical pharmacology studies conducted at sites in Indonesian Papua where the local population is a mixture of indigenous Melanesian Papuans and mainland-Indonesian transmigrants, and where PK differences across the two groups are clinically relevant.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (non-Papuan).
• Source aliases:
  • Paper’s binary Papuan indicator paired with the f multiplicative covariate term in Brussee 2016 Table 2 – used in Brussee_2016_arginine.R (values 1 = Papuan, 0 = non-Papuan; same orientation as the canonical).
• Example models: Brussee_2016_arginine.R (multiplicative factor f = 1.9 on L-arginine CL: CL_i = CL * (WT/60)^k1 * f^RACE_PAPUAN – Papuan patients have ~1.9x higher arginine clearance than the non-Papuan reference).
• Notes: Distinct from all existing RACE_<GROUP> indicators. Papuans are an indigenous Melanesian / Oceanic population genetically and culturally distinct from RACE_ASIAN (mainland-Indonesian transmigrants are typically Austronesian Asian by international population-genetics classification). The Brussee 2016 cohort was enrolled exclusively at the Mitra Masyarakat Hospital in Timika, Papua, so a paired REGION_INDONESIA indicator is not needed for that model – future models that mix Papuan and non-Papuan Indonesian sites may need such a region indicator separately.

RACE_IND_CHI_TWN (canonical for Indian/Chinese/Taiwanese composite race indicator (Schmid 2017 nintedanib))

• Description: 1 = subject self-identifies as Indian (South Asian), Chinese, or Taiwanese; 0 = otherwise (Caucasian, Black, Korean, or other Asian). Paper-specific composite developed by Schmid 2017 in pooling South Asian Indian and East Asian Chinese/Taiwanese subgroups together as a single nintedanib-bioavailability (F1) effect category against a Caucasian/Black/other-Asian reference. The grouping reflects the empirical clustering of the F1 covariate analysis (Schmid 2017 Table 3 ethnic-origin row); it does not imply biological homogeneity across the three constituent populations.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 = Caucasian, Black, Korean, or other Asian (paired with RACE_KOREAN to capture the Korean subgroup separately).
• Source aliases:
  • Paper’s ethnic-origin categorical levels “Indian / Chinese / Taiwanese origin” – decompose into RACE_IND_CHI_TWN = as.integer(ETHNIC %in% c("Indian", "Chinese", "Taiwanese")).
• Example models: Schmid_2017_nintedanib.R (multiplicative effect 1.33 on nintedanib relative bioavailability F1; reference category Caucasian / Black / other Asian).
• Notes: Specific scope: the composite grouping is paper-specific to Schmid 2017 and is unlikely to recur unchanged in other popPK extractions; future papers that distinguish South Asian from East Asian (or that lump Korean with Chinese) should register their own composite rather than overload this one. The constituent subgroups Chinese / Indian / Taiwanese remain available as individual indicators (RACE_CHINESE is canonical; RACE_INDIAN was added alongside this entry; RACE_TAIWANESE is not yet canonical but can be added when a future paper requires it).

RACE_KOREAN (canonical for Korean-heritage race indicator)

• Description: 1 = Korean heritage (subject self-identifies as Korean or is enrolled at a Korean-population stratum of a multiregional trial), 0 = non-Korean. Used when Korean subjects form a distinct subgroup whose PK differs from the Caucasian / Black / non-Korean Asian reference (e.g., Schmid 2017 nintedanib showed Korean patients have ~22% lower F1 than the Caucasian reference).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (non-Korean; document the paper-specific reference race composition per-model in the covariate’s notes).
• Source aliases: none formally; source NONMEM control streams typically use ad-hoc names (e.g., KOREAN, RACE.EQ.K).
• Example models: Schmid_2017_nintedanib.R (multiplicative effect 0.781 on nintedanib relative bioavailability F1; reference category Caucasian / Black / other Asian), vandenBerg_2021_finerenone.R (multiplicative effect 1.29^RACE_KOREAN on Vc/F; 2.4% of the FIDELIO-DKD cohort were Korean, and the paper Discussion notes the effect “may be a spurious finding based on limited data”), Jayanti_2026_pyrazinamide.R (reference category is specifically Indonesian, not a generic non-Korean mixture: the cohort is exactly 160 Korean and 160 Indonesian patients matched 1:1 on weight and age, so the single indicator partitions it completely. Used as a structural stratifier rather than a covariate effect – it selects which of two separately estimated typical clearances, which of two separately estimated clearance variances, and which of two diabetes covariate forms applies)., Usman_2022_valproic_acid.R (the effect enters on (1 - RACE_KOREAN), with Korean as the reference: the pooled cohort is 99 South Korean and 92 Pakistani patients, and the paper’s centre indicator CENT (1 = Pakistani) equals 1 - RACE_KOREAN. The term is the multiplicative linear factor (1 + 0.386 * (1 - RACE_KOREAN)) on intravenous valproic acid clearance. The indicator partitions the cohort completely, so for a patient who is neither Korean nor Pakistani RACE_KOREAN = 0 extrapolates to the Pakistani clearance).
• Notes: Parallels the established RACE_CHINESE / RACE_JAPANESE entries – individual East Asian nationality indicators when the source paper distinguishes them as distinct subgroups rather than pooling under RACE_ASIAN_NORTHEAST. Korean is the third NE-Asian individual nationality canonical alongside Chinese and Japanese; together they exhaust the RACE_ASIAN_NORTHEAST composite (Chinese + Japanese + Korean) so a model that uses all three individual indicators can omit RACE_ASIAN_NORTHEAST without information loss. Distinct from RACE_ASIAN_NORTHEAST (composite) and from RACE_ASIAN_OTH (within-Asian-population sub-indicator).

RACE_INDIAN (canonical for Indian (South Asian) race indicator)

• Description: 1 = South Asian Indian heritage (subject self-identifies as Indian; refers to nationals/descendants of India and not to American Indian / Native American groups), 0 = non-Indian. Used when Indian subjects form a distinct subgroup whose PK differs from the Caucasian / Black / non-Indian Asian reference (e.g., Schmid 2017 nintedanib showed Indian patients have ~90% higher BIBF 1202 F2 than the Caucasian / Black reference).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (non-Indian; document the paper-specific reference race composition per-model in the covariate’s notes).
• Source aliases: none formally; source NONMEM control streams typically use ad-hoc names (e.g., INDIAN, RACE.EQ.I).
• Example models: Schmid_2017_nintedanib.R (multiplicative effect 1.90 on BIBF 1202 relative bioavailability F2; paired with a non-Indian-Asian indicator at 1.20; reference category Caucasian / Black).
• Notes: Parallels RACE_CHINESE / RACE_JAPANESE / RACE_KOREAN – individual South / East Asian nationality indicators when the source paper distinguishes them as distinct subgroups. DISTINCT from RACE_ASIAN_AMIND_OTH and RACE_ASIAN_AMIND_MULTI, both of which use “AmInd” / “AMIND” to mean American Indian / Alaska Native (a North American indigenous population), not South Asian Indian. This is a recurring source of confusion in race-canonical lookups; reviewers should verify the meaning whenever a paper mentions “Indian” patients in a popPK cohort.

RACE_COLOURED (canonical for Coloured (South African population group) race indicator)

• Description: 1 = subject self-identifies as Coloured, the South African census population group of mixed ancestry that is the majority group in the Western Cape; 0 = any other self-identified race. Registered because South African clinical-trial cohorts routinely report Black / Coloured / White as their three self-identified race strata, and the Coloured stratum has no existing canonical: it is neither RACE_BLACK nor RACE_WHITE nor RACE_MULTI (which denotes a subject reporting more than one race on a US-style form, not a distinct census group).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (non-Coloured; document the paper-specific reference race composition per model in the covariate’s notes – Black and White in the founding cohort).
• Source aliases:
  • RACE – Koele 2025 supplementary control-stream data dictionary, ;RACE = Race [1=Cape-colored, 0=other]; same orientation as the canonical, so no value transformation is needed.
  • “Cape-coloured” – the regional descriptor Koele 2025 Table 1 and Results use for the same group.
• Example models: Koele_2025_btz043.R (multiplicative factor 0.762 on apparent BTZ-043 clearance, i.e. clearance 24% slower than in the Black / White reference; Koele 2025 Table 2 ‘Cape-coloured race on CL/F (%) 76.2 (65.4-88.5)’. The Discussion hypothesises a genetic polymorphism in a BTZ-043-metabolizing enzyme and states the effect should be confirmed in a larger population; the founding cohort was 65% Black, 34% Coloured and 1% White).
• Notes: The canonical name uses the bare census group name COLOURED rather than the regional form CAPE_COLOURED because “Coloured” is the term used on South African census and clinical-trial case-report forms nationally, while “Cape Coloured” narrows it to one province and would not fit a Gauteng or KwaZulu-Natal cohort reporting the same stratum. Spelled with the South African (British) orthography, matching the source. Sibling of RACE_BLACK / RACE_WHITE, with which it is mutually exclusive and jointly exhaustive in the typical South African three-stratum design; a model needs at most two of the three indicators, the third being the reference. Distinct from RACE_MULTI, which encodes a subject who selected more than one race category on a US-style multi-select form: that is an answer-format artefact, whereas Coloured is a single self-identified group with its own demographic history, so pooling the two would put subjects from different populations under one coefficient.

Geographic / enrollment-country indicators

Geographical study-site region indicators. Distinct from race / ethnicity (RACE_*), which describe subject ancestry; these describe the geographical location of the clinical trial site that enrolled the subject. Used in multi-regional studies (typically those including bridging analyses for Japan or East Asia) to capture region-specific clinical-practice or unmeasured-environment effects on PK that remain after accounting for body weight, race, and laboratory covariates. Encoded as a set of mutually exclusive binary indicators with US as the implicit reference category (all indicators = 0). When a paper groups some non-US regions with US (e.g., Hong 2025 groups US and Japan as the DXd CL reference), the model code uses only the indicators that distinguish the non-reference groups; the data column for the grouped region (e.g., REGION_JAPAN) is still recorded so the same dataset can serve other parameters that do separate that group.

REGION_JAPAN (canonical for Japan study-site / enrollment-country indicator)

• Description: 1 = study site in Japan (or patient enrolled in Japan, depending on source paper’s reporting), 0 = study site / enrollment country outside Japan. Geographical Japan indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-Japan sites / enrollment; specific reference set varies per model – e.g., Hong 2025 Dato-DXd CL uses “any non-Japan region”, whereas Hong 2025 DXd CL groups Japan with US into the reference).
• Source aliases:
  • COUNTRY_JPN – retired canonical; used in Yin_2021_trastuzumabDeruxtecan.R as an enrollment-country (not study-site-region) indicator. Some papers report country-of-enrollment rather than site region; both map to REGION_JAPAN when the binary contrast is Japan vs. non-Japan.
• Example models: Hong_2025_datopotamab.R (multiplicative effect 1 + (-0.219) = 0.781 on Dato-DXd linear clearance), Yin_2021_trastuzumabDeruxtecan.R (multiplicative effect 0.903 on CL_intact and 0.738 on V2_intact when REGION_JAPAN = 1; Yin 2021 retained Japan enrollment-country over Japanese race because the two were highly confounded, correlation -0.81), Morcos_2023_copanlisib.R (fractional effect -0.204 on copanlisib clearance, i.e. 20.4% lower CL for the 61 of 712 patients enrolled at Japanese sites; Morcos 2023 attributes the shift to “general differences in body weight” even though body weight itself did not survive the covariate search).
• Notes: Distinct from RACE_JAPANESE (subject ancestry). A subject of Japanese ancestry enrolled at a US site has RACE_JAPANESE = 1 but REGION_JAPAN = 0. Some papers (e.g., Yin 2021) report enrollment country rather than study-site region; both are encoded as REGION_JAPAN when the binary contrast is Japan vs. non-Japan. Paired with REGION_EUROPE and REGION_ROW in multi-regional studies (e.g., Hong 2025); REGION_JAPAN = 0 for a US-only cohort.

REGION_EUROPE (canonical for Europe study-site indicator)

• Description: 1 = study site in Europe, 0 = study site outside Europe. Geographical study-site region.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-Europe study sites; specific reference set varies per model).
• Source aliases: none yet; canonical name preferred.
• Example models: Hong_2025_datopotamab.R (multiplicative effect 1 + 0.240 = 1.240 on DXd clearance versus US/Japan reference), Naik_2016_vortioxetine.R (additive intercept-shift form: TVCL_EU = 39 L/hr is the typical CL/F when REGION_EUROPE = 1, versus USA reference TVCL = 51 L/hr), Mandema_2005_gemcabene_mbma.R (TRIAL-level rather than subject-level: an additive -4 percentage-point shift on the placebo LDL-C intercept in European trials versus the North American reference, e0_arm = e0 + e_region_europe_e0 * REGION_EUROPE), Sun_2025_maribavir_cmv_clearance_wk8.R, Sun_2025_maribavir_cmv_clearance_wk16.R, Sun_2025_maribavir_dysgeusia.R, Sun_2025_maribavir_anemia.R (all four are maribavir exposure-response logistic regressions in which REGION_EUROPE is paired with REGION_ASIAPACIFIC against a NORTH AMERICA reference rather than the family’s usual US reference; the coefficient changes sign between endpoints, -1.20 on the primary efficacy logit but +1.53 on the anemia logit).
• Notes: Pair with REGION_JAPAN and REGION_ROW to encode multi-regional study membership; subjects with all three indicators = 0 are in the “US” reference group. In a model-based meta-analysis the indicator is a property of the TRIAL, not of a subject, and it applies to every arm of that trial (placebo and active alike) – so a trial-level REGION_EUROPE effect that sits on the intercept cancels out of any within-trial active-minus-control contrast (see Mandema_2005_gemcabene_mbma.R).

REGION_ROW (canonical for Rest-of-World study-site indicator)

• Description: 1 = study site in Rest of World (i.e., not US, Japan, or Europe), 0 = study site in US / Japan / Europe. Residual-region indicator for multi-regional studies.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (US / Japan / Europe study sites; specific reference set varies per model).
• Source aliases:
  • REGION_RoW – mixed-case variant in some source publications (e.g., Naik 2016).
• Example models: Hong_2025_datopotamab.R (multiplicative effect 1 + 0.196 = 1.196 on DXd clearance versus US/Japan reference), Naik_2016_vortioxetine.R (additive intercept-shift form: TVCL_RoW = 38 L/hr is the typical CL/F when REGION_ROW = 1, versus USA reference TVCL = 51 L/hr; the RoW group for Naik 2016 spans study sites in Canada, Australia, and Asia), Gidal_2018_eslicarbazepine_dizziness.R (additive logit shift +0.668 versus a EUROPE reference, paired with REGION_NORTHAMERICA and REGION_LATINAMERICA; note that the same paper’s efficacy models invert the roles and make Rest of World the reference group instead).
• Notes: “Rest of the World” composition is paper-specific (e.g., Hong 2025 = study sites outside US, Japan, and Europe; Naik 2016 = Canada, Australia, and Asia). Document the subject set in covariateData[[REGION_ROW]]$notes per model.

REGION_FRANCE (canonical for France study-site / enrollment-country indicator)

• Description: 1 = study site in France (or French-population enrollment), 0 = otherwise. Country-level study-site indicator used when a multi-regional trial reports a France-vs-other contrast in popPK / PD parameters (cultural / behavioural-response differences, regional dosing-practice differences, etc.).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-France study sites; in Holford 1992 tacrine the reference cohort is the US protocol 970-01).
• Source aliases:
  • PROT == 4 – protocol-number alias used in Holford 1992 (the France cohort is protocol 970-04, encoded as PROT = 4; the US cohort is protocol 970-01, encoded as PROT = 1). Derive REGION_FRANCE = as.integer(PROT == 4).
• Example models: Holford_1992_tacrine.R (multiplicative scale factors on baseline ADAS-cog (1 + e_region_france_s0 * REGION_FRANCE) with e_region_france_s0 = 0.08; on placebo potency (1 + e_region_france_betap * REGION_FRANCE) with e_region_france_betap = 0.76; on placebo elimination half-time (1 + e_region_france_t12elp * REGION_FRANCE) with e_region_france_t12elp = 1.78 – so the France cohort has 8 percent higher baseline ADAS-cog, 76 percent larger placebo response, and 2.78x longer placebo wash-out half-time relative to the US cohort).
• Notes: Distinct from RACE_FRENCH (subject ancestry, no canonical at present). Holford 1992 introduces this as a behavioural-response covariate rather than a PK exposure covariate; the underlying mechanism is hypothesised cultural / clinical-trial-conduct differences, not pharmacology. Pair with REGION_JAPAN, REGION_EUROPE, REGION_ROW etc. when the same model needs to encode multiple geographic contrasts. Specific scope until a second model ratifies the name; at that point promote to general.

REGION_MOZAMBIQUE (canonical for Mozambique study-site / enrollment-country indicator)

• Description: 1 = study site in Mozambique, 0 = otherwise. Country-level study-site indicator used in multi-country sub-Saharan African trials of intermittent preventive treatment of malaria in pregnancy (IPTp).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-Mozambique sites; specific reference set varies per model – in de Kock 2017 the reference is Mali).
• Source aliases: none yet; canonical name preferred.
• Example models: deKock_2017_sulfadoxinePyrimethamine.R (multiplicative -20.2% effect on apparent pyrimethamine clearance; +57.6% scaling on observed pyrimethamine concentrations; +21.2% scaling on observed sulfadoxine concentrations; the on-observation scalings capture residual site-specific differences in apparent bioavailability or dried-blood-spot sample handling).
• Notes: Specific scope because the canonical models a single Mozambique vs non-Mozambique contrast within a specific multi-site trial; the reference set is paper-specific (e.g., Mali in de Kock 2017). Pairs with REGION_SUDAN and REGION_ZAMBIA for a 4-country sub-Saharan African IPTp trial design with Mali as the implicit reference.

REGION_SUDAN (canonical for Sudan study-site / enrollment-country indicator)

• Description: 1 = study site in Sudan, 0 = otherwise. Country-level study-site indicator used in multi-country sub-Saharan African trials.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-Sudan sites; specific reference set varies per model – in de Kock 2017 the reference is Mali).
• Source aliases: none yet; canonical name preferred.
• Example models: deKock_2017_sulfadoxinePyrimethamine.R (+15.5% scaling on observed sulfadoxine concentrations; +33.2% scaling on observed pyrimethamine concentrations).
• Notes: Specific scope; pairs with REGION_MOZAMBIQUE and REGION_ZAMBIA.

REGION_ZAMBIA (canonical for Zambia study-site / enrollment-country indicator)

• Description: 1 = study site in Zambia, 0 = otherwise. Country-level study-site indicator used in multi-country sub-Saharan African trials.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-Zambia sites; specific reference set varies per model – in de Kock 2017 the reference is Mali).
• Source aliases: none yet; canonical name preferred.
• Example models: deKock_2017_sulfadoxinePyrimethamine.R (-24.8% scaling on observed sulfadoxine concentrations; -5.4% scaling on observed pyrimethamine concentrations).
• Notes: Specific scope; pairs with REGION_MOZAMBIQUE and REGION_SUDAN.

REGION_TANZANIA (canonical for Tanzania study-site / enrollment-country indicator)

• Description: 1 = study site in Tanzania, 0 = otherwise. Country-level study-site indicator used in multi-country sub-Saharan African trials. In Pillay-Fuentes Lorente 2024 the contrast is Tanzania (Pemba Island) versus Cote d’Ivoire, and the paper refers to it as the “study population” covariate rather than by country name.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-Tanzania sites; the specific reference set varies per model – in Pillay-Fuentes Lorente 2024 the reference is Cote d’Ivoire).
• Source aliases: Country, study population.
• Example models: PillayFuentesLorente_2024_albendazole.R (exponential categorical effect on both apparent metabolite clearances: exp(0.56) = 1.75, i.e. 75% higher albendazole sulfoxide CL/F, and exp(0.38) = 1.46, i.e. 46% higher albendazole sulfone CL/F, in the Tanzanian study population).
• Notes: Specific scope, following the same pattern as REGION_MOZAMBIQUE / REGION_SUDAN / REGION_ZAMBIA: the canonical models a single Tanzania vs non-Tanzania contrast within one multi-site trial, and the comparator set is paper-specific. Note that a country indicator in this family is a proxy for whatever differs between the study populations (in Pillay-Fuentes Lorente 2024 the authors attribute it to drug-metabolising-enzyme polymorphisms, host-parasite interaction, nutritional status and gut microbiome differences), not to geography as such.

REGION_POLAND (canonical for Poland EPPICC enrollment-country indicator)

• Description: 1 = subject enrolled at the Polish EPPICC cohort site (Medical University Warsaw / Regional Hospital of Infectious Disease), 0 = otherwise. Country-level enrollment-country indicator used in the multi-country European Pregnancy and Paediatric HIV Cohort Collaboration (EPPICC) HIV/HCV coinfection study.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-Poland EPPICC sites; specific reference set varies per model – in Majekodunmi 2017 the reference is Ukraine).
• Source aliases: none yet; canonical name preferred.
• Example models: Majekodunmi_2017_HIV_HCV_CD4_recovery.R (additive +0.44 shift on pre-ART CD4 z-score intercept; Ukraine reference).
• Notes: Specific scope; pairs with REGION_RUSSIA, REGION_SWITZERLAND, REGION_UK, REGION_SPAIN, REGION_GERMANY, REGION_ITALY for the EPPICC 8-country pediatric HIV cohort with Ukraine as the implicit reference. Distinct from a Polish race/ethnicity indicator.

REGION_RUSSIA (canonical for Russia EPPICC enrollment-country indicator)

• Description: 1 = subject enrolled at the Russian EPPICC cohort site (Republican Hospital of Infectious Diseases, St Petersburg), 0 = otherwise. Country-level enrollment-country indicator used in the EPPICC HIV/HCV coinfection study.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-Russia EPPICC sites; specific reference set varies per model – in Majekodunmi 2017 the reference is Ukraine).
• Source aliases: none yet; canonical name preferred.
• Example models: Majekodunmi_2017_HIV_HCV_CD4_recovery.R (additive +0.69 shift on pre-ART CD4 z-score intercept; Ukraine reference).
• Notes: Specific scope; pairs with REGION_POLAND, REGION_SWITZERLAND, REGION_UK, REGION_SPAIN, REGION_GERMANY, REGION_ITALY for the EPPICC 8-country pediatric HIV cohort with Ukraine as the implicit reference.

REGION_SWITZERLAND (canonical for Switzerland EPPICC enrollment-country indicator)

• Description: 1 = subject enrolled at the Swiss Mother and Child HIV Cohort Study (MoCHiV), 0 = otherwise. Country-level enrollment-country indicator used in the EPPICC HIV/HCV coinfection study.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-Switzerland EPPICC sites; specific reference set varies per model – in Majekodunmi 2017 the reference is Ukraine).
• Source aliases: none yet; canonical name preferred.
• Example models: Majekodunmi_2017_HIV_HCV_CD4_recovery.R (additive +0.02 shift on pre-ART CD4 z-score intercept; Ukraine reference).
• Notes: Specific scope; pairs with the other EPPICC REGION indicators.

REGION_UK (canonical for United Kingdom EPPICC enrollment-country indicator)

• Description: 1 = subject enrolled in the UK Collaborative HIV Paediatric Study (CHIPS), 0 = otherwise. Country-level enrollment-country indicator used in the EPPICC HIV/HCV coinfection study.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-UK EPPICC sites; specific reference set varies per model – in Majekodunmi 2017 the reference is Ukraine).
• Source aliases: none yet; canonical name preferred.
• Example models: Majekodunmi_2017_HIV_HCV_CD4_recovery.R (additive -17.5 shift on pre-ART CD4 z-score intercept with Ukraine reference; magnitude implausibly large for a z-score effect and anchored on a UK cohort of only 2 subjects – reproduced verbatim per the published table and flagged in the model file and vignette as a small-sample artifact).
• Notes: Specific scope; pairs with the other EPPICC REGION indicators. Distinct from any Britain/England/Scotland-specific indicator.

REGION_SPAIN (canonical for Spain EPPICC enrollment-country indicator)

• Description: 1 = subject enrolled at the Spanish Paediatric HIV Network (CoRISpe; Madrid and Barcelona), 0 = otherwise. Country-level enrollment-country indicator used in the EPPICC HIV/HCV coinfection study.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-Spain EPPICC sites; specific reference set varies per model – in Majekodunmi 2017 the reference is Ukraine).
• Source aliases: none yet; canonical name preferred.
• Example models: Majekodunmi_2017_HIV_HCV_CD4_recovery.R (additive +2.89 shift on pre-ART CD4 z-score intercept; Ukraine reference).
• Notes: Specific scope; pairs with the other EPPICC REGION indicators.

REGION_GERMANY (canonical for Germany EPPICC enrollment-country indicator)

• Description: 1 = subject enrolled in the German Competence Network on HIV-infected Children, 0 = otherwise. Country-level enrollment-country indicator used in the EPPICC HIV/HCV coinfection study.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-Germany EPPICC sites; specific reference set varies per model – in Majekodunmi 2017 the reference is Ukraine).
• Source aliases: none yet; canonical name preferred.
• Example models: Majekodunmi_2017_HIV_HCV_CD4_recovery.R (additive +0.34 shift on pre-ART CD4 z-score intercept; Ukraine reference).
• Notes: Specific scope; pairs with the other EPPICC REGION indicators.

REGION_ITALY (canonical for Italy EPPICC enrollment-country indicator)

• Description: 1 = subject enrolled in the Italian Register for HIV-infection in Children, 0 = otherwise. Country-level enrollment-country indicator used in the EPPICC HIV/HCV coinfection study.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-Italy EPPICC sites; specific reference set varies per model – in Majekodunmi 2017 the reference is Ukraine).
• Source aliases: none yet; canonical name preferred.
• Example models: Majekodunmi_2017_HIV_HCV_CD4_recovery.R (additive -3.63 shift on pre-ART CD4 z-score intercept; Ukraine reference; small-sample subgroup with n = 2).
• Notes: Specific scope; pairs with the other EPPICC REGION indicators.

REGION_MOLDOVA (canonical for Moldova study-site / enrollment-country indicator)

• Description: 1 = study site in Moldova, 0 = otherwise. Country-level study-site indicator used when a multi-regional trial reports a Moldova-vs-other contrast in popPK / PD parameters.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-Moldova sites; specific reference set varies per model – in Cantillon 2018 the reference is the pooled USA, India, Philippines and Malaysia sites).
• Source aliases:
  • GEOG 5, GEOG5 – Cantillon 2018 notation, the fifth level of a five-level “geographic area of clinical site” factor (USA, India, Philippines, Malaysia, Moldova).
• Example models: Cantillon_2018_brilaroxazine.R (multiplicative effect on the PD Emax only, Emaxj = Emax * (1 + c1 * REGION_MOLDOVA) with c1 = -1.93, so Emax moves from -31.6 to +29.4 PANSS units and predicted PANSS RISES with cumulative exposure at those sites).
• Notes: Specific scope. A worked example of a region indicator that the source authors interpret as a MEASUREMENT artefact rather than as pharmacology: Cantillon 2018 found the site effect on the PANSS endpoint only, with no geographic-site effect on the pharmacokinetics, and reads it as an isolated rating discrepancy at one multi-centre site (“such data would be considered as an outlier”). Encode the effect as published, but a downstream user reproducing the paper’s base dose-response relationship should set the indicator to 0. A sign flip of this magnitude on a PD parameter is the signature of a single-site artefact rather than a regional pharmacological difference – contrast the REGION_JAPAN entries, where the effects are modest multiplicative shifts on clearance.

REGION_EASTASIA (canonical for East Asian region-of-origin study-site indicator)

• Description: 1 = subject enrolled at a study site in the East Asian region, 0 = enrolled elsewhere. Multi-country regional grouping (broader than a single-country REGION_<COUNTRY> indicator, narrower than a Rest-of-World REGION_ROW bucket). The exact country membership is protocol-specific and must be documented in per-model covariateData[[REGION_EASTASIA]]$notes. Distinct from the RACE_ASIAN* family: REGION_EASTASIA records where a subject was enrolled, not self-reported race; source papers routinely test the two separately and can find one significant and the other not.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-East-Asian region).
• Source aliases:
  • EASIAFL – East Asian flag; used in Cammarata_2024_sulbactam_durlobactam.R (Table 1 rows CLEASIAFL1 and V1EASIAFL1).
  • EAST_ASIA, REGION = East Asia – variant spellings of the same regional grouping.
  • RGN / RGNF – Zhou 2018 names the covariate RGN and its coefficient RGNF (“effect of RGN on F”); the two levels are “East (i.e. Asia/Japan)” and “West”.
• Example models: Zhou_2018_alisertib.R (Japan, Singapore, Taiwan, Hong Kong or South Korea; 59 of 422 analysis-set patients, 14.0%; a SINGLE coefficient of -0.341 multiplies CL/F, V1/F, Q/F and V2/F together, which is algebraically how a change in relative bioavailability propagates through an oral model – the reciprocal 1/(1-0.341) = 1.517 is the “51.7% (95% CI 36.5%, 70.1%) higher F” the paper reports, and it is the largest region effect in this family. Zhou 2018 is also the clearest worked case of region and race being separated empirically rather than by assumption: the 8 patients of Asian race enrolled in the West had dose-normalized AUC 475 nM.h/mg against 482 for Western non-Asians and 797 for Asians in East Asia, so the effect tracks enrolment region and not ancestry – though with n = 8 the paper declines to call the attribution conclusive), Cammarata_2024_sulbactam_durlobactam.R (China, Taiwan, or South Korea; proportional shifts of -0.199 on durlobactam total CL and -0.263 on durlobactam Vc, with no sulbactam region effect; 45 of 373 pooled subjects, 12.1%), Mandema_2011_biologicDMARDs_mbma.R (Japan or Taiwan; a STUDY-ARM-level trial-location indicator rather than a subject-level one, scaling the biologic Emax by 1 + 0.33 in the 4 of 50 trials carried out in East Asia), Larson_2026_sulbactam_durlobactam_pediatric_allometry.R (the pediatric ‘Adult with Allometry’ adaptation of that model, which retains the region term unchanged and sets it to 0 – poster P-444 states ‘East Asian flag was set to “not East Asian”’ for the simulated pediatric cohort)., Wang_2020_delamanid.R (Northeast Asia = China, Japan or Korea, 99 of 744 MDR-TB patients, 13.3%; multiplicative 1.53 on relative bioavailability, paired with the sibling REGION_SOUTHEASTASIA), Fediuk_2021b_ertugliflozin_esea.R (Hong Kong, Republic of Korea or Taiwan, plus the Japanese subjects of a phase 1 PK/PD study; acts only through the paper’s E/SE Asian indicator RACE_ASIAN * (REGION_EASTASIA + REGION_SOUTHEASTASIA) – Asian subjects at US or European sites are non-E/SE Asian – with multiplicative 1.17 on CL/F and 2.48 on Vc/F)
• Notes: Scope specific, matching every other member of the REGION_* family. Sibling to the single-country REGION_JAPAN / REGION_FRANCE /… entries and to REGION_EUROPE / REGION_ROW. When a source paper reports BOTH a region effect and a race effect, register the race arm separately under the RACE_* family and keep the two columns distinct – Cammarata 2024 explicitly found race and country of origin non-significant while East Asian region was significant.

REGION_SOUTHEASTASIA (canonical for Southeast Asian region-of-enrollment study-site indicator)

• Description: 1 = subject enrolled at a study site in the Southeast Asian region, 0 = enrolled elsewhere. Multi-country regional grouping, the Southeast Asian sibling of REGION_EASTASIA. The exact country membership is protocol-specific and must be documented in per-model covariateData[[REGION_SOUTHEASTASIA]]$notes. Records where a subject was enrolled, not self-reported race.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-Southeast-Asian region; state per model whether REGION_EASTASIA is carried separately so the reference set is explicit).
• Source aliases:
  • SE Asian, Southeast Asia (Philippines) – Wang 2020 Table 2 row label and Table 4 row F1,SEAsian.
• Example models: Wang_2020_delamanid.R (Southeast Asia = the Philippines, 200 of 744 MDR-TB patients, 26.9%; multiplicative 1.40 on the relative bioavailability of delamanid, paired with REGION_EASTASIA for Northeast Asia; the reference is non-Asian regions, including the 2 Asian-race patients enrolled in Peru), Fediuk_2021b_ertugliflozin_esea.R (Malaysia, the Philippines or Thailand; summed with REGION_EASTASIA and multiplied by RACE_ASIAN to form the paper’s E/SE Asian indicator, multiplicative 1.17 on CL/F and 2.48 on Vc/F).
• Notes: Member of the REGION_* family, scope specific like its siblings. Distinct from REGION_EASTASIA (Northeast Asia) and narrower than REGION_ASIAPACIFIC. Distinct from the RACE_ASIAN* race indicators because the founding paper groups Asian-race patients by where they were enrolled (Asian-race patients enrolled in Peru fall in the non-Asian reference).

REGION_EASTEUROPE (canonical for Eastern European region-of-enrollment study-site indicator)

• Description: 1 = subject enrolled at a study site in the Eastern European region, 0 = enrolled elsewhere. Multi-country regional grouping (broader than a single-country REGION_<COUNTRY> indicator, narrower than a Rest-of-World REGION_ROW bucket). The exact country membership is protocol-specific and must be documented in per-model covariateData[[REGION_EASTEUROPE]]$notes. Distinct from REGION_EUROPE, not a synonym of it: multi-regional trials routinely define Eastern Europe and (western) Europe as two separate levels of one region variable, and folding one into the other silently merges two arms the source fitted apart.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (any non-Eastern-European region).
• Source aliases:
  • EE – used in Jin_2025_benralizumab_aaer.R (Jin 2025 Resource 17 row beta_Base, EE).
  • REGION01 = 2, REGION = Eastern Europe – the parent five-level region variable in Jin 2025 Resource 3, whose levels are ‘1: Asia, 2: Eastern Europe, 3: Europe, 4: North America, 5: Rest of the world’.
• Example models: Jin_2025_benralizumab_aaer.R (log-scale effect of -0.329 on the baseline asthma exacerbation rate, i.e. a 28% lower baseline rate at Eastern European sites; the only region contrast retained in the final longitudinal exacerbation-rate model), Chan_2020_fenebrutinib_acr.R (the reference region of the fenebrutinib ACR exposure-response model, 62% of patients; paired with REGION_USA, with Latin America derived as 1 - REGION_USA - REGION_EASTEUROPE).
• Notes: Scope specific, matching every other member of the REGION_* family. Same sub-continental grain as REGION_EASTASIA, which is its closest structural sibling. As with REGION_EASTASIA, keep this column distinct from the RACE_* family: it records where a subject was enrolled, not self-reported race, and a source may find one significant and the other not (Jin 2025 retains both REGION_EASTEUROPE on the exacerbation baseline and RACE_CHINESE on the exacerbation Emax).

REGION_CHINA (canonical for China study-site / enrollment-country indicator)

• Description: 1 = subject enrolled at a study site in China, 0 = enrolled elsewhere. Single-country member of the REGION_* family, recording where a subject was enrolled rather than self-reported race or ethnicity. Distinct from REGION_EASTASIA, which is a broader multi-country regional grouping: encoding a China-specific effect as East Asian silently widens the claim beyond what the source fitted. Distinct also from the RACE_ASIAN* family, which records self-reported race.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-China / rest of world).
• Source aliases:
  • POP – Zhang 2025 Data S4 control streams, coded 2 = Region X (China) and 1 = non-Region X, so REGION_CHINA = POP - 1.
  • CN, cn, CHINA, REGION = China – variant spellings of the same enrollment-country indicator.
• Example models: Zhang_2025_drugA.R (multiplicative effect on PASI IC50, ic50 * R_regionX^REGION_CHINA, with R_regionX = 2.6 as the paper’s borderline clinically relevant inter-regional difference), Fediuk_2021b_ertugliflozin_china.R (mainland China, studies I-10 and III-5; acts only together with RACE_ASIAN: RACE_ASIAN * REGION_CHINA is the paper’s ‘Asian from mainland China’ indicator, multiplicative 1.04 on CL/F and 1.44 on Vc/F, and RACE_ASIAN * (1 - REGION_CHINA) its ‘Asian from the rest of the world’ indicator, 1.08 and 2.15; non-Asian subjects are the reference).
• Notes: Scope specific, matching every other member of the REGION_* family. Sibling to the single-country REGION_JAPAN / REGION_FRANCE / REGION_ITALY /… entries. Provenance caveat for the founding example: the Zhang 2025 article anonymises the region as “Region X” throughout and never names China; the identification comes from the authors’ own deposited Data S4 R code comments (n_cn “sample size for cn (Region X)” and “add label for Chinese and ROW populations (POP=2 is Region X, POP=1 is non-Region X)”).

REGION_USA (canonical for United States study-site / enrollment-country indicator)

• Description: 1 = subject enrolled and treated at a United States study site, 0 = enrolled outside the United States. Time-fixed per subject. Records where a subject was enrolled, not self-reported race or ancestry – source papers routinely test region and race separately and can find one relevant and the other not (Fukae 2024 carries REGION_USA, RACE_WHITE and RACE_BLACK as three simultaneous covariates in the same model).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (enrolled outside the United States). In a strictly two-region cohort this is the complement of REGION_JAPAN; document the cohort’s region membership in per-model covariateData[[REGION_USA]]$notes.
• Source aliases:
  • Geographic region: United States – Fukae 2024 Table 3 row label; used in the Fukae_2024_valemetostat_* exposure-safety family. No value transformation.
  • US, USA, COUNTRY = United States, REGION = US – variant spellings of the same enrollment-country indicator.
• Example models: Fukae_2024_valemetostat_anemia.R, Fukae_2024_valemetostat_anc_decrease.R, Fukae_2024_valemetostat_plt_decrease.R, Fukae_2024_valemetostat_teae_grade3.R, Fukae_2024_valemetostat_dose_interruption.R, Fukae_2024_valemetostat_dose_reduction.R (all six Fukae 2024 exposure-safety logistic models; log-odds effects on both the logit intercept and the unbound-AUCss exposure slope, relative to the Japan-enrolled reference; 31 of 102 pooled R/R NHL patients, 30.4%), Chan_2020_fenebrutinib_acr.R (multiplier 2.13 on the fenebrutinib Emax of a longitudinal ACR20/50/70 logistic model, relative to the Eastern European reference; the third region, Latin America, is derived as 1 - REGION_USA - REGION_EASTEUROPE).
• Notes: Scope specific, matching every other member of the REGION_* family (REGION_JAPAN, REGION_FRANCE, REGION_UK, REGION_SPAIN,…). Distinct from the RACE_* family; see REGION_EASTASIA for the multi-country regional-grouping form.

REGION_CANADA (canonical for Canada study-site / enrollment-country indicator)

• Description: 1 = subject enrolled and treated at a Canadian study site, 0 = enrolled outside Canada. Time-fixed per subject. Country-level study-site indicator used in multi-country trials that estimate a per-country shift in bioavailability, clearance or an observation-scaling factor.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-Canada sites; the specific reference set varies per model – in Gafar 2026 the reference is Indonesia, paired with REGION_VIETNAM for the third country).
• Source aliases:
  • FRM (value 1) – Gafar 2026 encodes the country contrast through the formulation column of its NONMEM dataset because country and rifampicin product are completely confounded in the 2R2 trial (every Canadian participant received re-compounded Rofact capsules). No value transformation beyond selecting the level.
  • CTR, SITE, COUNTRY = Canada – variant spellings of the same enrollment-country indicator.
• Example models: Gafar_2026_rifampicin.R (multiplicative factor 0.782 on prehepatic bioavailability, i.e. -21.8% versus the Indonesian reference; 87 of 440 participants, 19.8%).
• Notes: Scope specific, matching every other member of the REGION_* family. Use this rather than a FORM_* name when the source paper’s retained covariate is the country contrast even where the mechanism the authors propose is the formulation: Gafar 2026 fitted the two Vietnamese formulations separately and found no improvement (dOFV 1 point, 3 df, P >.05), so the identified effect is country-level and the formulation attribution is an interpretation. Record the country / formulation confounding in per-model covariateData[[REGION_CANADA]]$notes. Distinct from the RACE_* family.

REGION_VIETNAM (canonical for Vietnam study-site / enrollment-country indicator)

• Description: 1 = subject enrolled and treated at a Vietnamese study site, 0 = enrolled outside Vietnam. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-Vietnam sites; the specific reference set varies per model – in Gafar 2026 the reference is Indonesia, paired with REGION_CANADA for the third country).
• Source aliases:
  • FRM (values 2 and 3 pooled) – Gafar 2026 encodes the country contrast through the formulation column; values 2 (Svizera) and 3 (Mekophar) are the two Vietnamese products and are pooled into a single indicator in the final model.
  • CTR, SITE, COUNTRY = Vietnam – variant spellings of the same enrollment-country indicator.
• Example models: Gafar_2026_rifampicin.R (multiplicative factor 0.877 on prehepatic bioavailability, i.e. -12.3% versus the Indonesian reference; 88 of 440 participants, 20.0%).
• Notes: Scope specific. See REGION_CANADA for the country-versus-formulation naming rationale. Indonesia, the reference country in Gafar 2026, carries no indicator of its own: a participant with REGION_CANADA = 0 and REGION_VIETNAM = 0 is in the reference group.

REGION_ASIAPACIFIC (canonical for Asia Pacific study-site indicator)

• Description: 1 = study site in the Asia Pacific region, 0 = study site outside it. Geographical study-site region indicator for multi-regional trials that report a three-way North America / Europe / Asia Pacific split.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-Asia-Pacific study sites; the specific reference set varies per model – in Sun 2025 the reference is North America, with Europe carried by its own REGION_EUROPE indicator).
• Source aliases:
  • Enrolling region: Asia Pacific – printed row label in Sun 2025 Table S2; Asia Pacific [N = 41] in the Table S3 and Figure S3 parameter tables; Asia Pacific vs. North America in Table S4.
• Example models: Sun_2025_maribavir_cmv_clearance_wk8.R (coefficient -1.45 on the primary efficacy logit), Sun_2025_maribavir_cmv_clearance_wk16.R (-1.49), Sun_2025_maribavir_dysgeusia.R (-0.455), Sun_2025_maribavir_anemia.R (+1.14, OR 3.12).
• Notes: Broader than REGION_JAPAN and REGION_EASTASIA, which it therefore does NOT alias: Asia Pacific as a trial-operations region conventionally spans East Asia, South-East Asia and Oceania, so a subject enrolled in Australia has REGION_ASIAPACIFIC = 1 and REGION_EASTASIA = 0. Distinct from RACE_ASIAN (subject ancestry): Sun 2025 Table S2 records 41 subjects enrolled in Asia Pacific but 31 of Asian race, so the two columns disagree for at least ten subjects in the founding cohort alone – a concrete reminder that region and ancestry are not interchangeable. Where a paper’s three-way split has North America rather than the US as the reference, the reference arm is “both REGION_EUROPE and REGION_ASIAPACIFIC equal 0” and no separate North America indicator is registered. Scope: specific until a second model ratifies the name; at that point consider promoting to general alongside the rest of the REGION_* family.

REGION_MPUMALANGA (canonical for the Mpumalanga (South Africa) study-site indicator)

• Description: 1 = study site is Mpumalanga, South Africa, 0 = otherwise. Sub-country study-site indicator for pooled sub-Saharan African antimalarial analyses that estimate an effect for an individual site rather than for a whole country.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-Mpumalanga sites; the reference set is paper-specific – in de Kock 2018 the sulfadoxine reference group is Magude, Bancoumana, Bela Vista, Catuane and Chileka).
• Source aliases: none yet; canonical name preferred.
• Example models: deKock_2018_sulfadoxinePyrimethamine.R (one of three sites sharing a single estimated -39.7% multiplicative scaling of the predicted sulfadoxine whole-blood concentration; carries no pyrimethamine effect).
• Notes: First member of the REGION_* family to name a study SITE below country level. The extension is deliberate: the family is documented as a “study-site / enrollment-country indicator”, and a pooled analysis with several sites in one country (de Kock 2018 has five Mozambican sites) cannot express its site effects with country indicators. Per-site indicators are preferred over one composite column per estimated group because the grouping is the source paper’s modelling choice, not a property of the data – model() sums the member indicators into a group-membership flag. Compare REGION_MOZAMBIQUE / REGION_SUDAN / REGION_ZAMBIA, which serve the same role in the sibling deKock_2017_sulfadoxinePyrimethamine.R where each country contributed exactly one site.

REGION_BOANE (canonical for the Boane (Mozambique) study-site indicator)

• Description: 1 = study site is Boane, Mozambique, 0 = otherwise. Sub-country study-site indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-Boane sites; in de Kock 2018 the sulfadoxine reference group is Magude, Bancoumana, Bela Vista, Catuane and Chileka).
• Source aliases: none yet; canonical name preferred.
• Example models: deKock_2018_sulfadoxinePyrimethamine.R (shares the -39.7% sulfadoxine observation-scaling effect with REGION_MPUMALANGA and REGION_NAMAACHA; carries no pyrimethamine effect).
• Notes: One of the four Allen et al. sites in the de Kock 2018 pool (Boane, Catuane, Magude, Namaacha). Nested inside REGION_MOZAMBIQUE; a model that needs the country contrast rather than the site contrast should use that column instead of summing the site indicators.

REGION_NAMAACHA (canonical for the Namaacha (Mozambique) study-site indicator)

• Description: 1 = study site is Namaacha, Mozambique, 0 = otherwise. Sub-country study-site indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-Namaacha sites; the reference set differs per drug in de Kock 2018 – see Notes).
• Source aliases: none yet; canonical name preferred.
• Example models: deKock_2018_sulfadoxinePyrimethamine.R (the only site carrying a scaling effect for BOTH drugs: it shares the -39.7% sulfadoxine observation scaling with REGION_MPUMALANGA and REGION_BOANE, and it is alone in the -22.0% pyrimethamine observation-scaling group).
• Notes: Namaacha contributed patients to two of the four pooled trials in de Kock 2018 – Barnes et al. (n = 91) and Allen et al. (n = 72) – and appears twice in that paper’s Table 1 for that reason; the covariate model treats the site as one group regardless of which trial enrolled the patient, so this indicator is set for both. A worked caution on why per-site indicators beat composite group columns: Namaacha belongs to a different group for each drug, so a composite-per-group encoding would need the same subject flagged in two mutually inconsistent columns.

REGION_BANCOUMANA (canonical for the Bancoumana (Mali) study-site indicator)

• Description: 1 = study site is Bancoumana, Mali, 0 = otherwise. Sub-country study-site indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-Bancoumana sites; in de Kock 2018 the pyrimethamine reference group is Magude, Mpumalanga, Boane and Chileka).
• Source aliases: none yet; canonical name preferred.
• Example models: deKock_2018_sulfadoxinePyrimethamine.R (one of three sites sharing a single estimated +20.2% multiplicative scaling of the predicted pyrimethamine whole-blood concentration; carries no sulfadoxine effect).
• Notes: The sole site of the Tekete et al. trial (n = 114) in the de Kock 2018 pool, and the only Malian site; a model needing the country contrast can use this column directly.

REGION_BELAVISTA (canonical for the Bela Vista (Mozambique) study-site indicator)

• Description: 1 = study site is Bela Vista, Mozambique, 0 = otherwise. Sub-country study-site indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-Bela-Vista sites; in de Kock 2018 the pyrimethamine reference group is Magude, Mpumalanga, Boane and Chileka).
• Source aliases: none yet; canonical name preferred.
• Example models: deKock_2018_sulfadoxinePyrimethamine.R (shares the +20.2% pyrimethamine observation-scaling effect with REGION_BANCOUMANA and REGION_CATUANE; carries no sulfadoxine effect).
• Notes: The two-word place name is run together without a separator so the canonical stays a single token, matching REGION_EASTASIA and REGION_ASIAPACIFIC. Nested inside REGION_MOZAMBIQUE.

REGION_CATUANE (canonical for the Catuane (Mozambique) study-site indicator)

• Description: 1 = study site is Catuane, Mozambique, 0 = otherwise. Sub-country study-site indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-Catuane sites; in de Kock 2018 the pyrimethamine reference group is Magude, Mpumalanga, Boane and Chileka).
• Source aliases:
  • Cutuane – the spelling used in the de Kock 2018 Table 1 column header. The same site is spelled Catuane in that paper’s Table 2 footnotes and Results text, which is where the covariate effect itself is reported, so Catuane is the canonical spelling.
• Example models: deKock_2018_sulfadoxinePyrimethamine.R (shares the +20.2% pyrimethamine observation-scaling effect with REGION_BANCOUMANA and REGION_BELAVISTA; carries no sulfadoxine effect).
• Notes: The smallest site in the de Kock 2018 pool (n = 33). Nested inside REGION_MOZAMBIQUE.

Pediatric comorbidities

REGION_NORTHAMERICA (canonical for North America study-site indicator)

• Description: 1 = study site in North America, 0 = study site elsewhere. Geographical study-site region, coarser than the country-level REGION_USA / REGION_CANADA members and used when a multiregional trial stratifies at the continent level.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0. The specific reference set varies per model and MUST be documented per model – in the Gidal 2018 family alone it is Europe for the adverse-event models and Rest of World for the efficacy models.
• Source aliases:
  • NA_i – Gidal 2018 Appendix S1 Equations E-3, E-8, E-11 and E-12. Note that this source spelling collides with R’s NA; rename to the canonical when assembling input data.
• Example models: Gidal_2018_eslicarbazepine_dizziness.R (additive logit shift +1.04 versus a EUROPE reference, paired with REGION_LATINAMERICA and REGION_ROW), Gidal_2018_eslicarbazepine_ssf.R, Gidal_2018_eslicarbazepine_weekly_seizures.R (additive baseline shifts of +0.460 on the log scale and +1.12 seizures per week respectively, this time versus a REST-OF-WORLD reference and paired with REGION_WESTERNEUROPE and REGION_LATINAMERICA).
• Notes: Pair with the other REGION_* indicators to encode multiregional study membership; the group with every indicator at 0 is the reference. The founding paper is a caution worth carrying: it uses TWO DIFFERENT region splits in the same publication – a Europe-referenced three-indicator split for the adverse-event models and a Rest-of-World-referenced three-indicator split for the efficacy models – so a region encoding is not transferable even between models from one paper, let alone between papers. Distinct from REGION_USA, which is country-level; use that when a paper separates the United States from Canada.

REGION_LATINAMERICA (canonical for Latin America study-site indicator)

• Description: 1 = study site in Latin America, 0 = study site elsewhere. Geographical study-site region covering Mexico, Central America, South America and the Spanish- and Portuguese-speaking Caribbean.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0; the specific reference set varies per model and must be documented per model. See the caution under REGION_NORTHAMERICA.
• Source aliases:
  • LA_i – Gidal 2018 Appendix S1 Equations E-3, E-5, E-8, E-11 and E-12.
• Example models: Gidal_2018_eslicarbazepine_dizziness.R (additive logit shift +1.13 versus a Europe reference, the largest of that model’s three regional shifts), Gidal_2018_eslicarbazepine_somnolence.R (the ONLY region indicator retained for that endpoint, +0.661 versus a reference of the POOLED remaining three regions – a different reference again from its sibling models), Gidal_2018_eslicarbazepine_ssf.R, Gidal_2018_eslicarbazepine_weekly_seizures.R (baseline shifts of +0.310 on the log scale and +0.958 seizures per week versus a Rest-of-World reference).
• Notes: Pair with the other REGION_* indicators; the group with every indicator at 0 is the reference. A Latin American site is conventionally NOT folded into REGION_ROW when both appear in the same model, since the point of carrying the indicator is to separate it from the residual region.

REGION_WESTERNEUROPE (canonical for Western Europe study-site indicator)

• Description: 1 = study site in Western Europe, 0 = study site elsewhere. A geographical sub-region of REGION_EUROPE, and the western counterpart of the already-registered REGION_EASTEUROPE.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0; the specific reference set varies per model and must be documented per model. See the caution under REGION_NORTHAMERICA.
• Source aliases:
  • WEU_i – Gidal 2018 Appendix S1 Equations E-8, E-9, E-10, E-11 and E-12. The paper abbreviates the group “WE” in its main text and figure legends.
• Example models: Gidal_2018_eslicarbazepine_ssf.R (the only covariate in that model acting on BOTH the baseline and the Emax: +0.228 on the log-scale baseline seizure frequency and +0.242 on a negative Emax, so a Western European patient is predicted to have both a higher baseline and a smaller treatment benefit), Gidal_2018_eslicarbazepine_response.R (-0.668 on the responder logit versus a reference of the pooled non-Western-European remainder), Gidal_2018_eslicarbazepine_weekly_seizures.R (+0.751 seizures per week on the baseline count).
• Notes: Deliberately a SEPARATE canonical from REGION_EUROPE rather than an alias of it, on the same footing as the already-registered REGION_EASTEUROPE: a paper that splits Europe into east and west assigns its eastern sites to a different group (usually the residual REGION_ROW), so silently mapping a Western Europe column onto the whole-continent indicator would put those patients in the wrong stratum. Use REGION_EUROPE when the paper treats the continent as one block and this member only when it names Western Europe specifically. Pair with the other REGION_* indicators; the group with every indicator at 0 is the reference.

DIS_CLD_PREM (canonical for chronic lung disease of prematurity)

• Description: 1 = chronic lung disease of prematurity (bronchopulmonary dysplasia, BPD), 0 = no CLD. Time-fixed per subject (diagnosis at study entry).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no CLD of prematurity).
• Source aliases:
  • CLD_PREM – prior canonical name used in Robbie_2012_palivizumab.R prior to the 2026-06-19 DIS_ prefix standardization.
  • CLD – used in Robbie_2012_palivizumab.R.
  • BPD – bronchopulmonary-dysplasia shorthand.
• Example models: Robbie_2012_palivizumab.R (fractional +20% effect on CL).
• Notes: Standard pediatric / neonatology comorbidity flag; ties to palivizumab’s label population (high-risk preterm infants) and may re-appear in future pediatric mAb PK analyses (RSV, parenteral nutrition, etc.).

Comorbidities

DIS_DIAB (canonical for diabetes-mellitus comorbidity indicator)

• Description: 1 = patient has diabetes mellitus comorbidity, 0 = no diabetes comorbidity. Time-fixed at study entry per subject. Type 1 vs Type 2 are pooled at the column level; record the per-model Type-2-specific detail (if any) in the per-model covariateData[[DIS_DIAB]]$source_name / notes fields rather than maintaining a parallel canonical.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no diabetes comorbidity).
• Source aliases:
  • DIAB – prior canonical name (pre-2026-06-19 DIS_ prefix standardization) used in Chen_2022_guselkumab.R, Yao_2018_guselkumab.R, Overgaard_2019_semaglutide.R.
  • T2DM – prior parallel canonical (merged into DIS_DIAB on 2026-06-19) used in NA_NA_paracetamol.R (DDMODEL00000228), Lu_2014_sglt_qsp.R, Sharma_2018_naltrexone_bupropion.R.
  • Diabetes – used in Sherer_2012_AAA.R (Sherer 2012 Methods page 2 symbol “Diabetes”).
  • T2D – used in Guiastrennec_2016_gastric_emptying.R (matched-cohort flag, 1 = T2D patient vs 0 = matched nondiabetic control).
  • ZDF – used in Han_2018_methionineMetabolismCycle.R (Zucker Diabetic Fatty rat as T2DM animal model; ZDF/Gmi fa/fa coded as DIS_DIAB = 1 vs ZDF/Gmi fa/? non-diabetic littermate control coded as DIS_DIAB = 0).
• Example models:
  • Chen_2022_guselkumab.R (multiplicative effect on CL/F: 1.15^DIS_DIAB, +15% in patients with diabetes).
  • Yao_2018_guselkumab.R (multiplicative effect on CL/F: 1.12^DIS_DIAB, +12% in patients with diabetes).
  • Overgaard_2019_semaglutide.R (multiplicative effects 1.12 on CL and 0.544 on ka per Overgaard 2019 Table 4; T2D-vs-normoglycaemic stratification).
  • Sherer_2012_AAA.R (additive shift on the first derivative of AAA growth rate with size beta2: e_diab_b2 = -0.32/year for diabetics; cohort prevalence 14%).
  • Sharma_2018_naltrexone_bupropion.R (Type-2 diabetes mellitus stratifier on Emax, kout, kpro, baseline BW; from Sharma 2018 study 6, NB-304).
  • Lu_2014_sglt_qsp.R (multiplicative +17.6% shift on the typical-value Vmax2 of SGLT2 in the renal-glucose-reabsorption QSP model: Vmax2_DIS_DIAB = 110 mmol/h vs Vmax2_healthy = 93.5 mmol/h per Lu 2014 Table 2 calibration).
  • NA_NA_paracetamol.R (DDMODEL00000228).
  • Guiastrennec_2016_gastric_emptying.R (multiplicative -81.1% depression of POTcarbC, the carbohydrate potency on CCK release; all other parameters are common across cohorts).
  • Han_2018_methionineMetabolismCycle.R (preclinical rat MMC model; multiplicative-on-log-scale T2DM effects on five rate constants K_SH (+16 %), K_HM (-92 %), K_HC (-95 %), K_HP (-86 %), K_PH (-99 %) derived from Han 2018 Table 1 ZDF/control ratios).
  • Xu_2020_clopidogrel_pbpk.R (PBPK-PD of clopidogrel in CAD patients with DM: multiplies the CYP1A2 / 2B6 / 2C9 / 2C19 / 3A4 Vmax by 1.23 / 0.55 / 1.26 / 0.54 / 0.62 and the CES1 intrinsic clearances by 1.27, and replaces the six stomach-to-colon transit rate constants; Xu 2020 Methods ‘PBPK-PD Model in CAD Patients With DM’. The paper’s DM cohort is CAD + DM, so it is used together with DIS_IHD = 1).
• Notes: Captures pre-existing diabetes mellitus as a comorbidity in non-diabetes-primary indications (e.g., psoriatic arthritis, psoriasis, vascular disease) AND as a primary stratifier in diabetes-population studies (semaglutide, SGLT-QSP, naltrexone-bupropion T2DM substudies). Distinct from a primary disease-state indicator like DIS_UC. Type 1 vs Type 2 mellitus are pooled onto the single DIS_DIAB column; if a paper distinguishes them, record “T2DM” or “Type 2” in the per-model covariateData[[DIS_DIAB]]$source_name notes, and if a future paper genuinely needs a binary Type-2-vs-Type-1 split (e.g., a popPK/PD analysis stratifying by HbA1c level), register DIS_DIAB_TYPE2 at that point. Diabetic patients tend to have higher inflammation and altered IgG turnover, which can manifest as modest changes in monoclonal-antibody clearance. In vascular populations (Sherer 2012) diabetes is associated with slower AAA growth, possibly via aberrant monocyte-matrix interactions (Golledge 2008 mechanism cited in Sherer 2012 Discussion). Overgaard 2016 (weight-management liraglutide) reports T2DM as one level of a three-level “baseline glycaemic status” covariate (normoglycaemic reference, prediabetic, T2DM); the T2DM level maps to DIS_DIAB = 1 and the prediabetic level uses the sibling canonical [[DIS_PREDIAB]].

DIS_RENAL (canonical for history-of-renal-disease comorbidity indicator)

• Description: 1 = patient has a history of (or current) renal disease of any type, 0 = no renal disease. A pooled medical-history flag covering acute kidney disease, chronic kidney disease and other renal diagnoses together; time-fixed at study entry per subject. Use it when a source paper collapses a multi-level renal-diagnosis field into one binary covariate and estimates a single coefficient for it.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no history of renal disease).
• Source aliases:
  • REN – used in Yoon_2023_vancomycin.R (Yoon 2023 Results section 3.2: “REN = 1 for renal disease and 0 for no renal disease”).
  • CKD / kidney status – used in Baklouti_2021_iohexol_dog.R (Baklouti 2021 Table 1: “Healthy CKD - (Code 0)”, “Diseased CKD + (Code 1)”; dogs classed by IRIS guidelines).
• Example models: Yoon_2023_vancomycin.R (proportional shift of -0.237 on vancomycin clearance, i.e. 23.7% lower CL, applied multiplicatively alongside the continuous creatinine-driven renal-function factor; Yoon 2023 Table 1 resolves the source field into none / acute kidney disease / chronic kidney disease / others and the final model pools the latter three), Baklouti_2021_iohexol_dog.R (veterinary: exponential shift of -0.379 on log iohexol clearance, i.e. 31.5% lower GFR, in dogs with IRIS-diagnosed chronic kidney disease, alongside a centred linear-exponential serum-creatinine effect; chronic kidney disease is the only renal diagnosis in that cohort, so the pooled flag carries it exactly).
• Notes: Sits in the comorbidity / medical-history family with DIS_DIAB and DIS_HYPERT, and is deliberately distinct from the two other renal groups in this register. It is not a member of the RENALIMP_MILD / RENALIMP_MOD / RENALIMP_SEV / RENALIMP_ESRD family, which stratifies measured renal function against the regulatory creatinine-clearance thresholds; DIS_RENAL records a diagnosis and carries no threshold. It is also distinct from DIS_ARF, which is acute renal failure specifically, and from RRT_*, which record renal-replacement treatment. A model may carry DIS_RENAL together with a continuous renal-function covariate (CRCL, CREAT) and the two are multiplicative, not alternatives – the founding model does exactly that, so do not assume a diagnosis flag replaces the continuous relationship the way an RENALIMP_ESRD indicator often does. If a future paper reports acute and chronic renal disease as separate coefficients, register DIS_AKD / DIS_CKD at that point rather than overloading this column.

DIS_PNEUMONIA (canonical for pneumonia comorbidity / co-infection indicator)

• Description: 1 = patient has pneumonia, 0 = no pneumonia. A general pneumonia flag with no acquisition-setting qualifier, time-fixed at study entry per subject.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no pneumonia).
• Source aliases:
  • PNE – used in Yoon_2023_vancomycin.R (Yoon 2023 Results section 3.3: “PNE = 1 indicates pneumonia and 0 indicates no pneumonia”).
• Example models: Yoon_2023_vancomycin.R (additive shift of +0.0058/h on the transit rate constant of the C-reactive-protein cascade, shortening the mean transit time from 9.62 to 6.65 days; 32.0% of the 128-patient PD cohort, Yoon 2023 Table 2).
  • Kawaguchi_2021_cefiderocol.R (the enrolment indication rather than a comorbidity: a single multiplicative factor x0.981 on cefiderocol CL for patients enrolled with pneumonia in the phase 3 APEKS-NP and CREDIBLE-CR studies, plus the shared x1.39 any-infection factor on V1. The enrolled pneumonia was hospital-acquired, ventilator-associated or health care-associated, but the model carries one pneumonia level with no acquisition-setting split, and health care-associated pneumonia has no DIS_<infection> column, so the general flag is used rather than DIS_HABP + DIS_VABP).
• Notes: Distinct from DIS_HABP and DIS_VABP, which are members of the mutually exclusive DIS_<infection> index-infection-type set used to classify the infection an antibacterial trial enrolled against. DIS_PNEUMONIA is the general comorbidity / co-infection flag for papers that record pneumonia without classifying it as hospital-acquired or ventilator-associated, and it belongs to the medical-history family alongside DIS_DIAB and DIS_HYPERT. The founding model uses it in exactly that sense – pneumonia as a secondary infection prolonging vancomycin therapy, not as the enrolment indication. Do not substitute DIS_HABP / DIS_VABP when the source does not state the acquisition setting, and do not use DIS_PNEUMONIA as a catch-all parent of those two: a paper that classifies its pneumonia arms should use the specific columns so the strata stay separable.

DIS_PREDIAB (canonical for prediabetes / impaired glucose regulation indicator)

• Description: 1 = subject has prediabetes / impaired glucose regulation at baseline (e.g., impaired fasting glucose 5.6-6.9 mmol/L, impaired glucose tolerance, or HbA1c 5.7-6.4 % per ADA criteria), 0 = subject is not prediabetic (typically normoglycaemic; T2DM subjects are usually stratified onto the sibling canonical DIS_DIAB in the same model). Time-fixed at study entry per subject.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (not prediabetic; typically normoglycaemic, but paper-defined).
• Source aliases:
  • Prediabetes – used in Overgaard_2016_liraglutide.R (Overgaard 2016 Fig. 2 / Table S1 “Cov. prediabetes”; one of the three levels of a “baseline glycaemic status” covariate alongside normoglycaemic reference and T2DM).
• Example models: Overgaard_2016_liraglutide.R (multiplicative log-scale effect on CL/F, coefficient +0.00 per Table S1; effect is negligible per the paper’s own analysis).
• Notes: Sibling to DIS_DIAB in the disease-state family. Used when a source paper stratifies a diabetes / obesity cohort into normoglycaemic vs prediabetic vs T2DM strata and reports covariate effects on each stratum separately. Only ratified when both DIS_PREDIAB and DIS_DIAB are needed to reproduce the paper’s three-level glycaemic decomposition – when a paper only reports a single “any dysglycaemia” indicator, use DIS_DIAB alone (or the appropriate primary-disease-state canonical). ADA prediabetes criteria: FPG 100-125 mg/dL (5.6-6.9 mmol/L), 2-h PG 140-199 mg/dL (7.8-11.0 mmol/L) on 75 g OGTT, or HbA1c 5.7-6.4 % (39-46 mmol/mol).

DIS_HYPERT (canonical for hypertension comorbidity / medical-history indicator)

• Description: 1 = patient has a history of (or current) hypertension as a comorbidity; 0 = no hypertension. Time-fixed at study entry per subject (medical-history flag rather than time-varying blood-pressure measurement).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no hypertension comorbidity).
• Source aliases:
  • HYPERT – prior canonical name used in Girard_2012_pimasertib.R prior to the 2026-06-19 DIS_ prefix standardization.
  • MHHY (medical history of hypertension) – used in Girard_2012_pimasertib.R.
• Example models: Girard_2012_pimasertib.R (additive shift on the cumulative-logit AE-score model: theta_mhhy * DIS_HYPERT; +0.539 logit units in patients with prior hypertension).
• Notes: Companion to DIS_DIAB (diabetes-mellitus comorbidity); both are baseline binary medical-history flags collected from clinical-history forms. Captures any prior or current hypertension diagnosis, regardless of treatment status; if a future model needs to separate treated vs untreated hypertension, register a refinement (DIS_HYPERT_TREATED).

DIS_HYPERLIP (canonical for hyperlipidemia / hypercholesterolemia diagnosis indicator)

• Description: 1 = participant carries a diagnosis of hyperlipidemia (equivalently hypercholesterolemia / dyslipidemia) at study entry; 0 = no hyperlipidemia diagnosis (typically a healthy-volunteer or normolipidemic comparator arm pooled into the same analysis). Time-fixed per subject. Captures the generic acquired / polygenic lipid disorder that defines the target population of lipid-modifying therapies, as distinct from the monogenic familial forms carried by DIS_HEFH and DIS_HOFH.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no hyperlipidemia diagnosis).
• Source aliases:
  • Hyperlipidemia – Jadhav 2023 Table 2 covariate row name.
  • HLD, HCHOL, DYSLIP – common clinical-dataset abbreviations for the same diagnosis flag.
• Example models: Jadhav_2023_bempedoicAcid.R, Jadhav_2023_bempedoicAcid_ldlc.R (proportional shift CL/F * (1 + (-0.0945) * DIS_HYPERLIP), i.e. 9.5% lower apparent bempedoic acid clearance in participants with hyperlipidemia relative to the pooled healthy / non-hyperlipidemic reference; Jadhav 2023 Table 2).
• Notes: Not derivable from DIS_HEALTHY by complement, and not mutually exclusive with DIS_DIAB: the Jadhav 2023 popPK dataset contained 49 participants with diabetes alone plus 310 with hyperlipidemia and diabetes (Table 1 footnote c), so all three flags can be jointly set. Distinct from DIS_HEFH / DIS_HOFH, which identify the heterozygous and homozygous familial-hypercholesterolemia subsets; a HeFH patient normally also carries DIS_HYPERLIP = 1, so record the per-model convention in covariateData[[DIS_HYPERLIP]]$notes whenever both appear. Also distinct from the continuous lipid-panel canonicals (LDLC, HDLC), which carry the measured concentration rather than the diagnosis.

DIS_FRDA (canonical for Friedreich’s ataxia patient indicator)

• Description: 1 = participant has Friedreich’s ataxia (FRDA), the autosomal-recessive FXN trinucleotide-repeat disorder causing progressive sensory and cerebellar ataxia with hypertrophic cardiomyopathy; 0 = participant does not. Time-fixed per subject (a primary disease-state stratifier, not a comorbidity flag). Use it in pooled analyses that mix an FRDA patient cohort with healthy volunteers and/or a different patient cohort, where the FRDA stratum carries its own disposition or bioavailability effect.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (not Friedreich’s ataxia; typically healthy volunteers, and any other-disease stratum whose exposures were shown not to differ from healthy volunteers).
• Source aliases:
  • FA – used in Hu_2025_vatiquinone.R (Hu 2025 Section 2.3 covariate equations, “FAi is set to 1 for participants with FA and is set to 0 for healthy volunteers”; same orientation as the canonical).
• Example models:
  • Hu_2025_vatiquinone.R (two linear-deviation effects from Hu 2025 Table 2: Frel *= (1 + (-0.501) * DIS_FRDA) on relative bioavailability, theta16, and CL *= (1 + (-0.406) * DIS_FRDA) on apparent clearance, theta15. The two act in opposite exposure directions – lower bioavailability but also slower clearance – and partly cancel, netting a 19% lower steady-state AUC and 24% lower Cmax,ss in FRDA patients per Hu 2025 Table 3).
• Notes: The canonical name uses the standard genetic-disease abbreviation FRDA rather than the source papers’ FA, because a bare FA column is ambiguous against “fatty acid” and “folic acid” in a pharmacometric register. When a pooled dataset contains a third disease stratum, check how the source paper coded it before assigning the reference level: Hu_2025_vatiquinone.R codes its 54 patients with other mitochondrial diseases as DIS_FRDA = 0 because Hu 2025’s Discussion reports that stratum’s exposures were not distinguishable from healthy volunteers, so the reference level pools two clinically distinct groups. If a future paper estimates a separate effect for a non-FRDA mitochondrial-disease stratum, register a sibling indicator for it rather than overloading this column. Distinct from the ataxia-severity rating scales (mFARS / FARS), which are continuous progression endpoints rather than a diagnosis flag; register those separately if a PD model needs them.

DIS_IHD (canonical for ischemic heart disease comorbidity / medical-history indicator)

• Description: 1 = patient carries a diagnosis of ischemic (coronary) heart disease – stable angina, prior myocardial infarction, prior revascularisation, or a documented obstructive-coronary-disease diagnosis – recorded at or before study entry; 0 = no ischemic heart disease. Time-fixed per subject (medical-history flag, not a measured biomarker).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no ischemic heart disease).
• Source aliases:
  • IHD – Elhefnawy 2023 Table 1 / Table 3 covariate row name.
  • CAD, CHD, CORONARY – common clinical-dataset abbreviations for the same medical-history flag.
• Example models: Elhefnawy_2023_recurrent_ischemic_stroke_dm.R (multiplicative log-linear effect on the Gompertz hazard of recurrent ischemic stroke among patients with diabetes: exp(0.876 * DIS_IHD), adjusted hazard ratio 2.40, 95% CI 1.79-3.20; Elhefnawy 2023 Cureus 15(12):e50794 Table 3), Elhefnawy_2023_recurrent_ischemic_stroke_nondm.R, Elhefnawy_2023_recurrent_ischemic_stroke.R (log-linear effect on the recurrent-ischemic-stroke hazard: exp(0.745 * DIS_IHD), i.e. an adjusted hazard ratio of 2.10; Elhefnawy 2023 Table 3 theta6), Xu_2020_clopidogrel_pbpk.R (coronary artery disease in a PBPK-PD model: scales every blood flow by the CAD/healthy cardiac-output ratio 0.90 and the CLOP-AM platelet-inactivation constant kirre by 0.7; Xu 2020 Eq 18 and Methods ‘PBPK-PD Model in CAD Without DM Patients’).
• Notes: Companion to DIS_HYPERT, DIS_DIAB and DIS_HYPERLIP in the baseline-comorbidity family; all four are commonly collected together from a cardiovascular-registry medical-history form and are not mutually exclusive. Distinct from DIS_ACS, which identifies a cohort enrolled during an acute coronary syndrome rather than a chronic medical-history flag, and from the study-arm-level prevalence covariate DIS_CHD_PERCENT used in meta-analytic models. Ascertainment varies by source (physician diagnosis, discharge coding, medication history), so record the per-model definition in covariateData[[DIS_IHD]]$notes. If a future model needs to separate prior myocardial infarction from angina without infarction, register a refinement (DIS_MI) rather than overloading this flag.

Surgical history / disease state

SPINAL_BLOCK (canonical for spinal (neuraxial) block height)

• Description: The most cephalad dermatome anaesthetised by a neuraxial (spinal / epidural) block, recorded as a thoracic segment number (5 = T5, 6 = T6,…). Determines how much of the sympathetic chain is blocked and therefore the expected severity of the resulting vasodilatory hypotension.
• Units: thoracic segment number
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (SPINAL_BLOCK / ref)^exponent or as a linear per-segment regression term. Reference values observed: 5 (T5; Dings 2026 population median; cohort means 5.79 for cafedrine/theodrenaline and 5.02 for ephedrine).
• Source aliases:
  • Spinal block height – Dings 2026 Table 2 and Tables A2 / A3 row label; same orientation, no value transformation.
  • SPINAL – Dings 2026 Table A1 covariate-effect label.
• Example models: Dings_2026_cafedrine_theodrenaline_ephedrine.R (power effect on the additive anaesthesia blood-pressure effect: eff_anaes_bp = -9.78 * (SPINAL_BLOCK/5)^-2.05), Dings_2026_neonatal_acidosis.R (linear per-segment terms of +0.00534 on umbilical-arterial pH and +0.3 mmol/L on base excess).
• Notes: The orientation is inverted relative to the plain-English phrase “block height”, and this must be documented in every model that uses it. Because the value is a segment number, a lower number means an anatomically higher (more extensive) block: T4 blocks more sympathetic outflow than T6. So a lower SPINAL_BLOCK predicts a larger fall in blood pressure and worse neonatal acid-base status. That inverse orientation is what reconciles the negative power exponent in the founding paper’s Table A1 with the positive per-segment coefficients in its Tables A2 / A3: all three say the same thing, that a higher block is worse. Do not “correct” the sign of a coefficient on this column without first confirming which orientation the source used. General scope – neuraxial block height applies to any obstetric, orthopaedic or abdominal regional-anaesthesia model.

T_ANAESTHESIA (canonical for time from anaesthesia induction to the modelled treatment)

• Description: Elapsed time between induction of anaesthesia and the start of the modelled treatment (time zero of the model). A positive number: anaesthesia precedes treatment. Lets a model account for how far an anaesthesia-driven physiological effect has already progressed when the treatment begins. Member of the auto-approved T_<event> canonical family.
• Units: min
• Type: continuous
• Scope: general
• Reference category: n/a – enters an exponential decay term. Reference values observed: 7 min (Dings 2026 typical parturient).
• Source aliases:
  • TIMEAN – Dings 2026 Eq. A11 symbol; same orientation and units, no transformation.
• Example models: Dings_2026_cafedrine_theodrenaline_ephedrine.R (sets the initial condition of the effect_anaesthesia state, effect_anaesthesia(0) = exp(-kdel * T_ANAESTHESIA), so that the anaesthesia blood-pressure effect already realised before treatment is absorbed into the observed baseline and only the remaining fall is predicted forward).
• Notes: General scope.

T_INCISION (canonical for time from the modelled treatment to surgical incision)

• Description: Time of surgical incision relative to model time zero (the modelled treatment). Used to derive a time-varying 0/1 indicator inside model() as t >= T_INCISION, which avoids carrying a time-varying covariate column. Member of the auto-approved T_<event> family.
• Units: min
• Type: continuous
• Scope: general
• Reference category: n/a. Reference values observed: 5 min after treatment (Dings 2026 typical parturient).
• Source aliases:
  • I – Dings 2026 Eqs A12-A14 use the derived indicator directly (0 before incision, 1 after).
• Example models: Dings_2026_cafedrine_theodrenaline_ephedrine.R (iu = (t >= T_INCISION) + (t >= T_UTEROTOMY), which adds eff_event_hr = 2.75 beats/min and eff_event_pp = 2.39 mmHg of pulse pressure per event).
• Notes: General scope. Supplying the event time as a time-fixed covariate and deriving the indicator in model() is strongly preferred over supplying the indicator as a time-varying column: rxode2 silently drops covariate columns assigned onto an rxEt object, so a time-varying indicator is a recurring source of vignettes that render cleanly while modelling nothing.

T_UTEROTOMY (canonical for time from the modelled treatment to uterotomy)

• Description: Time of uterotomy (incision of the uterus during caesarean section) relative to model time zero. Used to derive a 0/1 indicator inside model() as t >= T_UTEROTOMY. Member of the auto-approved T_<event> family.
• Units: min
• Type: continuous
• Scope: specific
• Reference category: n/a. Reference values observed: 9 min after treatment (Dings 2026 typical parturient).
• Source aliases:
  • U – Dings 2026 Eqs A12-A14 use the derived indicator directly.
• Example models: Dings_2026_cafedrine_theodrenaline_ephedrine.R (second of the two intraoperative stimulation events; see T_INCISION).
• Notes: Specific scope – uterotomy is meaningful only for caesarean-section cohorts.

T_CORDCLAMP_SAMPLING (canonical for time from umbilical-cord clamping to blood sampling)

• Description: Delay between clamping the umbilical cord and drawing the umbilical-arterial blood-gas sample. A pre-analytical covariate rather than a physiological one: continued anaerobic metabolism in the sampled blood raises measured lactate with elapsed time, so this delay must be adjusted for when comparing neonatal acid-base results. Member of the auto-approved T_<event> family.
• Units: min
• Type: continuous
• Scope: specific
• Reference category: n/a – linear regression term. Reference values observed: cohort means 7.81 min (cafedrine/theodrenaline) and 4.65 min (ephedrine) in Dings 2026 Table 2; the between-arm difference is significant (p < 0.001).
• Source aliases:
  • Cord clamping to blood sampling – Dings 2026 Table A4 row label.
• Example models: Dings_2026_neonatal_acidosis.R (+0.708 mg/dL lactate per minute of delay – the largest single term in the lactate regression, and a confounder for the between-treatment lactate comparison precisely because the delay itself differed by arm).
• Notes: Specific scope – an obstetric / neonatal blood-gas pre-analytical covariate. Worth carrying even when not significant, because it is a measurement artefact that can otherwise be mistaken for a treatment effect.

T_INFUSION_LAG (canonical for a data-supplied infusion-delivery lag time)

• Description: Delay between starting an infusion device (typically a syringe pump driving a low-volume infusion through a line whose dead space is comparable to the infused volume) and the drug actually entering the circulation. Supplied per subject (or per dose record) as a known value, NOT estimated, and applied as an absorption lag on the infusion into the central compartment (alag(central) <- T_INFUSION_LAG, NONMEM ALAG1 = TLAG). The lag shifts the whole infusion later; it does not change the infusion duration. Member of the auto-approved T_<event> family.
• Units: h
• Type: continuous
• Scope: general
• Reference category: n/a – enters the model as the lag itself. Reference values observed: 5-40 min (0.083-0.667 h) by body-weight dosing band in Neroutsos 2022 (40 min below 16 kg, 5 or 10 min above 34 kg).
• Source aliases:
  • TLAG – Neroutsos 2022 supplement NONMEM $INPUT column; same meaning, carried in hours on the NONMEM time axis.
• Example models: Neroutsos_2022_busulfan.R (paediatric intravenous busulfan by syringe pump; the per-band values were determined in an in-vitro simulation of the infusion set-up and entered in the dataset from Neroutsos 2022 Table 2).
• Notes: General scope: any low-rate paediatric or neonatal infusion model that corrects for line dead space can reuse it. Distinct from an ESTIMATED absorption lag (ltlag / tlag parameters in ini()), which is a model parameter rather than a data column, and from MIX_LAGGED_ABS, a mixture-class indicator for oral absorption. Carry it in the model’s time units (hours) even when the source tabulates minutes.

POD (canonical for post-operative day)

• Description: Days elapsed since the qualifying surgical event (e.g., solid-organ transplantation, major resection). Time-varying within subject; integer- or fractional-day valued; rises monotonically from 0 at the day of surgery. Captures time-since-surgery effects on PK that are not explained by other covariates – e.g., post-transplant clearance of immunosuppressants typically declines toward a steady value over the first weeks-to-months as graft function, fluid status, hematocrit, and corticosteroid taper stabilise.
• Units: days
• Type: continuous
• Scope: general
• Reference category: n/a – typically used in centred-deviation form (1 + e_pod_param * (POD - ref_pod)), sometimes with an upper cap (e.g., values > 180 fixed to 180 days when the residual time-varying effect plateaus).
• Source aliases:
  • POD – used in Bergmann_2014_tacrolimus.R (capped at 180 days; centred at 22.7 days, the dataset mean).
  • PTD – “posttransplantation day”; used in Suzuki_2024_mycophenolic_acid.R. Same orientation and units, no value transformation.
  • PTD (two-level category) – used in Liu_2022_tacrolimus.R and Liu_2022_tacrolimus_cyp3a5.R. Liu 2022 records PTD = 1 before and PTD = 2 from post-transplant day 28 (HSCT); the model takes the continuous POD column and derives the indicator (POD >= 28) internally, with an exponential effect exp(-0.32 * (POD >= 28)) on tacrolimus CL.
  • DFLT – “days from liver transplantation”; used in Shoji_2021_vancomycin.R. Same orientation and units, but Shoji 2021 defines it as the days from transplant to the FIRST day of the vancomycin treatment episode, so it is a per-episode constant rather than a per-observation day count; supply the episode-start value on every row.
• Example models: Bergmann_2014_tacrolimus.R (linear deviation from POD = 22.7 days on tacrolimus CL/F; coefficient -0.0021 per day implies a 0.21% per-day decrease in apparent oral clearance with a 180-day plateau), Suzuki_2024_mycophenolic_acid.R (exponential effect on relative bioavailability, exp(0.956 * POD / 84) with the divisor being the maximum POD in the analysis dataset rather than a centring value, so the factor is 1 at POD = 0 and rises to 2.8-fold by day 90; renal-transplant recipients on mycophenolate mofetil), Zhou_2025_tacrolimus.R (median-normalised power form (POD / 49)^0.14 on tacrolimus CL/F – note the positive exponent, i.e. clearance rises with time post-transplant, the opposite direction to Bergmann 2014; see Notes), Shoji_2021_vancomycin.R (median-normalised power form (POD / 17)^-0.09 on vancomycin CL in pediatric liver-transplant recipients, POD held constant at its value on the first day of vancomycin; the negative exponent makes the term infinite at POD = 0, so the model needs POD > 0), Chen_2020_ciclosporin.R (median-normalised power (POD / 51.5)^0.749 on ciclosporin CL/F in paediatric bone marrow transplant recipients – the qualifying event is the transplant infusion rather than surgery; the paper’s column is also POD, ‘days post-transplant’), Li_2019_cyclosporine.R (days since allogeneic HSCT; linear declines (1 - (POD - 9) * 0.00703) on intravenous cyclosporine CL and (1 - (POD - 9) * 0.0197) on V – the V factor reaches zero at POD 59.8, inside the 0-67 day observed range, so the published form is only usable below about POD 50).
• Notes: Time-varying within subject; the per-row value is the integer day count from the date of surgery to the observation date. Negative values are legitimate when a cohort contributes pre-surgical observations: the Suzuki 2024 renal-transplant dataset begins 14 days pre-transplantation, and its control stream gates the covariate effect off for those rows (IF (PTD.GE.0) THEN FPTD = EXP(...) ELSE FPTD = 1). A model consuming POD over a pre-surgical window must reproduce whatever gate the source applies rather than feeding a negative day count into the covariate equation. For solid-organ-transplant cohorts, POD is the conventional NONMEM $INPUT column name. When the source paper reports a different name (DPT for “days post-transplant”, TX_DAY, T_POSTOP), record the alias here. Distinct from TIME (rxode2 time clock) and from OCC (integer-valued occasion / period indicator for IOV). When a paper uses POD jointly with an IOV occasion column, both can coexist in the dataset: POD enters the typical-value covariate equation, OCC multiplexes the IOV etas. The 180-day cap in Bergmann 2014 is data-driven (most observations are within the first 90 days post-transplant, so the linear effect is identifiable only over that window) – document any per-model cap in covariateData[[POD]]$notes.

TTD (canonical for time to death)

• Description: Days remaining from the observation time to the patient’s recorded time of death (TTD >= 0; falls to 0 on the day of death). Time-varying per subject. Available only retrospectively in observational palliative-care / end-of-life cohorts; not usable as a prospective predictor.
• Units: days
• Type: continuous
• Scope: specific
• Reference category: n/a – used in a first-order exponential decay form on a structural PK parameter. The Franken 2015 morphine model parameterises CL(TTD) = CL_pop - theta_D * exp(-theta_rate * TTD), so the decay term vanishes as TTD -> infinity (asymptotic CL far from death) and reaches its peak drop theta_D as TTD -> 0 (day of death).
• Source aliases:
  • TTD – used in Franken_2015_morphine.R (Franken 2015 NONMEM column for time-to-death in days; paper Eq. 3 and Table 2).
• Example models: Franken_2015_morphine.R (Franken 2015 Clin Pharmacokinet; first-order exponential decay term on morphine clearance with theta_D = 17.6 L/h and theta_rate = 0.13 /day; CL drops from 47.5 to 29.9 L/h as TTD goes from infinity to 0).
• Notes: Specific scope because the covariate’s semantics depend on a palliative-care / observational-cohort study design where time of death is known by retrospective abstraction. The Franken 2015 model interprets the TTD-dependent CL change as a composite of physiological end-of-life processes (e.g., reduced hepatic blood flow, cachexia) that are not captured by standard blood-chemistry covariates. For prospective simulation of a virtual cohort without a known time of death, set TTD to a large value (e.g., > 50 days) to recover the asymptotic CL_pop. Distinct from the canonical POD (post-operative day, monotonically increasing from a surgery date) – TTD counts down to death rather than up from a procedure.

POSTTX_DAY1 (canonical for first-24-hours-post-transplant indicator)

• Description: Binary indicator for the first 24 hours post-transplant. 1 = the observation falls within the first 24 hours (day 1) post-transplant; 0 = otherwise. Time-varying per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (any observation outside the first 24 hours post-transplant).
• Source aliases:
  • first day post-transplant – used in Storset_2014_tacrolimus.R.
• Example models: Storset_2014_tacrolimus.R (multiplicative ~2.68-fold increase in oral bioavailability on day 1: fdepot *= 2.68^POSTTX_DAY1; Storset 2014 Table 2 final theory-based model retains the day-1 factor with subject-level IIV of 57% CV on the day-1 multiplier).
• Notes: Distinct from the continuous POD (post-operative day) canonical above – POSTTX_DAY1 is a derived binary indicator (operationally POSTTX_DAY1 = as.integer(POD < 1) when both columns are present in the dataset). The two coexist in the same model when the source paper uses POD-based continuous effects on some parameters and a separate binary day-1 effect on others (Storset 2014 retains the binary day-1 factor explicitly because the day-1 oral bioavailability is ~2.68-fold higher than the rest of the post-transplant period and is not well captured by a continuous POD effect). Storset 2014 Discussion attributes the day-1 oral-bioavailability spike to candidate mechanisms including methylprednisolone-bolus inhibition of intestinal CYP3A / P-glycoprotein, surgery-related inflammation, anaesthesia / opioid effects on gut motility, and reduced food intake – but no single mechanism was identifiable in the data. The 2.68-fold factor was retained because it produced a 209-point OFV decrease and was crucial for predicting concentrations measured on the first post-transplant day. In Storset 2014 the day-1 effect carries its own subject-level eta (BSV 57% CV on the day-1 factor); only subjects with day-1 observations contribute to that eta.

POSTTX_STABLE (canonical for stable-versus-early post-transplant stage indicator)

• Description: Binary post-transplant stage indicator for cohorts sampled at two separated stages after transplantation. 1 = stable post-transplant state (sampled years after transplantation); 0 = early post-transplant stage (sampled within the first week or so after transplantation). Constant per subject in a cohort that samples each patient at one stage.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (early post-transplant stage).
• Source aliases:
  • Stage – used in Wang_2022_mycophenolic_acid.R (Methods: “posttransplant stages (the early stage as 0 and the stable stage as 1)”; Table 3 coefficients dCLdStage, dVdStage). Same orientation, no value transformation.
• Example models: Wang_2022_mycophenolic_acid.R (Phoenix NLME exponential categorical effects on MPA CL/F, 23.36 * exp(-0.82 * POSTTX_STABLE), and V/F, 78.07 * exp(-2.54 * POSTTX_STABLE), in adult renal-transplant recipients sampled on days 4-8 or 5.5-10 years after transplantation).
• Notes: Use this canonical when the source paper itself codes the stage as a binary column and the cohort contains only the two separated stages, so no day-count threshold between them is stated or identifiable. Do NOT derive it from POD with a maintainer-chosen cutoff: in Wang 2022 every patient was sampled either within the first week (4.88 +/- 1.01 days) or at least 5 years (2499.94 +/- 467.26 days) after transplantation, and any threshold between 8 days and 5 years would be invented. When a source states its own threshold on a day count (e.g. Barau_2012_mycophenolic_acid.R, which dichotomises POD <= 180 inside model()), keep POD as the input column and derive the indicator in model() instead. Distinct from POSTTX_DAY1 (first 24 hours only, time-varying) and POD (continuous days since transplantation). The stage is a composite of everything that changes between the first week and the stable state – corticosteroid dose, proton-pump-inhibitor co-medication, renal function, haemoglobin and albumin in Wang 2022 – so the coefficient is not attributable to any single mechanism.

PFA (canonical for perioperative intra-operative fluid administration volume)

• Description: Total volume of intra-operative fluid administered during a surgical procedure – saline / albumin / fresh-frozen-plasma / blood / platelet replacement summed across the entire procedure. Time-fixed per subject (a single per-subject scalar capturing the full intra-operative perfusion load). Used in solid-organ-transplant and other major-surgery popPK models where the volume expansion from peri-operative fluid resuscitation is large enough to perturb post-operative body composition and apparent volume of distribution beyond what pre-operative body weight alone explains.
• Units: mL
• Type: continuous
• Scope: specific
• Reference category: n/a – used algebraically. The Oualha 2018 paediatric-LT model uses PFA additively with pre-operative body weight to define a transient post-operative body-weight curve: BW(t) = (BWPREOP + PFA/1000) * (1 - (1 - fbw) * t^hill_bw / (tbw50^hill_bw + t^hill_bw)), with the /1000 conversion applying the 1 mL ~ 1 g fluid-density-1 convention so PFA in mL maps onto a kg increment of body weight. Future surgical / critical-care popPK papers may use PFA differently (additive on Vc, multiplicative on CL); the per-paper parameterisation should be recorded in covariateData[[PFA]]$notes.
• Source aliases:
  • PFA – used in Oualha_2018_enoxaparin.R (Oualha 2018 BJCP; total intra-operative fluid volume in mL summed over the entire liver-transplant procedure; cohort median 2634 mL, range 1008-6520 mL).
• Example models: Oualha_2018_enoxaparin.R (Oualha 2018 paediatric liver-transplant cohort; PFA in mL adds to BWPREOP after conversion to L, defining a transient post-operative body-weight curve BW(t) that drives the time-varying allometric scaling of V; the Hill / fBW / tBW50 parameters of the BW(t) curve are jointly estimated with the enoxaparin PK).
• Notes: Specific scope until a second surgical / critical-care popPK model ratifies the name; at that point promote to general. The 1 mL ~ 1 g (density-1) convention is conventional and adequate for resuscitation fluids (crystalloids and blood products); a more precise mass-balance accounting could account for the slightly higher density of packed RBCs (~1.08 g/mL) or 5% albumin (~1.02 g/mL), but the difference is well within the IIV captured by lfbw / ltbw50. Distinct from URINE_FLOW (instantaneous urine flow rate, time-varying within an observation interval) and from WT (the pre- or post-operative body weight scalar). Founding example: Oualha 2018 enoxaparin (paediatric liver transplantation, intra-operative fluid resuscitation typical of major paediatric surgery). When a future model needs a separate accounting of crystalloid vs colloid vs blood-product volumes, register sibling canonicals (PFA_CRYST, PFA_COLLOID, PFA_BLOOD_PROD) rather than overloading this entry.

PFA_NET_RATE (canonical for net intra-operative fluid balance expressed as an hourly rate)

• Description: Net volume of fluid retained per hour of surgery – the volume of fluid infused plus the volume of blood infused minus the volume of urine output, divided by the operation time. A RATE (mL/h), and a NET quantity: unlike PFA it debits the measured urine output and normalizes by operation duration, so a long operation with a large crystalloid load and brisk diuresis can have a smaller PFA_NET_RATE than a short operation with a modest load and oliguria. Time-fixed per subject (one scalar per operation). Used in perioperative popPK models as a composite surrogate for the intra-operative expansion of circulating volume and hepatic blood flow, which the operative literature holds responsible for the altered disposition of high-extraction drugs during general anaesthesia.
• Units: mL/h
• Type: continuous
• Scope: specific
• Reference category: n/a – used algebraically. Nie 2023 enters it as a power function on intercompartmental clearance normalized to the cohort MEAN, Q = 245 * (PFA_NET_RATE / 617.96)^-0.58. Note that the normalizing constant is the mean (617.96 mL/h), not the cohort median (563.56 mL/h) that the paper’s generic covariate equation (Eq. 2) nominates; the printed final-model equation governs. Future perioperative popPK papers may center on a median or use a linear or exponential form; record the per-paper parameterisation in covariateData[[PFA_NET_RATE]]$notes.
• Source aliases:
  • HNF (hourly net fluid volume infused) – used in Nie_2023_nalbuphine.R (Nie 2023 Front Pharmacol; defined in the Table 1 footnote as HNF = (FVI + BVI - UVO) / OT for fluid volume infused, blood volume infused, urine volume output and operation time; model-building cohort 617.96 +/- 247.61 mL/h, median 563.56, range 234.26-1202.25).
• Example models: Nie_2023_nalbuphine.R (Nie 2023 nalbuphine popPK in adults under general anaesthesia; the ONLY covariate retained by forward inclusion / backward elimination, entering intercompartmental clearance Q as (PFA_NET_RATE / 617.96)^-0.58 and reducing the between-subject variability of Q from 35.5% to 17.9%. The Discussion attributes the effect to intra-operative changes in hepatic blood flow, nalbuphine having a hepatic extraction ratio of 0.5-0.7, but the simulated impact on plasma concentration across the 10th-90th percentile of the covariate is small enough that the authors recommend against dose adjustment).
• Notes: Specific scope until a second perioperative popPK model ratifies the name; at that point promote to general. Distinct from PFA (perioperative intra-operative fluid administration volume): PFA is a gross cumulative VOLUME in mL over the whole procedure with no urine debit, whereas PFA_NET_RATE is a net BALANCE per hour. The two are not interconvertible without the urine output and the operation duration, so do not substitute one for the other or derive one from the other silently. Distinct from URINE_FLOW (the urine-output term alone, and time-varying within an observation interval) and from Q_CVVH / CRRT_EFFLUENT (extracorporeal circuit flows, which are machine settings rather than a patient fluid balance). Named in the PFA family that the PFA entry anticipates, following its instruction to register siblings rather than overload the base entry. Ratified canonically on 2026-09-04 alongside the Nie 2023 nalbuphine extraction.

CUM_FLUID_BAL_PCT (canonical for cumulative fluid balance expressed as a percentage of admission body weight)

• Description: Running sum, over the days of an intensive-care admission, of the daily net fluid balance (all recorded intakes minus all recorded outputs), divided by the ADMISSION body weight and expressed as a percentage. Time-varying within subject and monotone only when the patient stays net-positive: a day of diuresis subtracts, and the value can be negative outright. Known clinically as “percent fluid overload” or “cumulative fluid balance (%)”; it is the standard critical-care fluid-accumulation metric, preferred over an absolute volume because it is already normalized for patient size and therefore comparable between a 4 kg infant and a 90 kg adult. Used as a covariate on distribution volume in critical-care popPK models of small hydrophilic drugs (beta-lactams, aminoglycosides), whose extracellular distribution expands with fluid accumulation.
• Units: %
• Type: continuous
• Scope: general
• Reference category: n/a – used algebraically, and NOT normalized to a reference value: zero cumulative fluid balance is the natural, dimensionally meaningful origin, so the covariate enters uncentred. Morales Junior 2025 applies it as the exponential multiplier vc = vc_pop * (WT/70) * exp(0.026 * CUM_FLUID_BAL_PCT), i.e. +2.6% central volume per percentage point of cumulative fluid balance. A future paper may use a linear (1 + slope * CUM_FLUID_BAL_PCT) or power form; record the per-paper parameterisation in covariateData[[CUM_FLUID_BAL_PCT]]$notes.
• Source aliases:
  • Cum%FB – used in MoralesJunior_2025_cefepime.R (Morales Junior 2025 Clin Pharmacokinet; defined in Sect. 2.5 as “the sum of all the previous days’ and present day’s percentage of fluid balance (sum of each day’s net fluid balance / PICU admission body weight x 100)”). Same quantity, no value transformation.
• Example models: MoralesJunior_2025_cefepime.R (exponential effect on the central volume of distribution, coefficient 0.026 per percentage point, Table 2; cohort day-1 to day-7 medians 3.3 / 4.5 / 5.5 / 6.2 / 4.9 / 5.7 / 4.0% with IQRs spanning -0.1 to 15.1%; the Monte Carlo simulations set the covariate to 0% for every simulated patient and the Discussion sanctions that as the fallback when fluid-balance data are unavailable).
• Notes: Distinct from PFA and PFA_NET_RATE. All three describe fluid handling but none is convertible into another: PFA is a gross intra-operative administered VOLUME in mL with no output debit, PFA_NET_RATE is a net intra-operative balance per hour in mL/h, and this entry is a net cumulative balance over days expressed as a PERCENTAGE of admission body weight. The differing normalisations (none / operation duration / admission weight) and the differing accumulation windows (one operation / one operation / a whole ICU admission) mean a value from one can never substitute for another. Deliberately named outside the perioperative PFA_ prefix for that reason: this is an intensive-care admission-scale quantity, not a perioperative one, and reusing the PFA prefix would assert a surgical setting the covariate does not have. The denominator is the ADMISSION body weight, a fixed per-subject scalar – a model that also carries a time-varying WT must not re-derive this covariate from the current weight, because the two denominators diverge exactly when fluid accumulates, which is the situation the covariate exists to describe. A paper reporting the same accumulation in absolute volume (mL or L) rather than as a percentage should register a sibling CUM_FLUID_BAL_ML instead of reusing this column, following the PFA / PFA_NET_RATE precedent. General scope on the first paper: the definition is paper-independent standard critical-care nomenclature, and the same group reports having used the identical covariate on central volume in a companion PICU meropenem model (Morales Junior 2025 Discussion, reference 28), which is not yet in this registry.

TX_LIVER (canonical for liver (hepatic) transplant indicator)

• Description: Binary indicator for liver / hepatic-graft transplant recipients in a pooled solid-organ-transplant cohort. 1 = liver transplant; 0 = non-liver solid-organ transplant (kidney, heart, or lung). Time-fixed per subject (assigned at transplantation date).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-liver solid-organ graft – typically kidney as the most common non-liver comparator in pooled solid-organ-transplant analyses).
• Source aliases:
  • Hepatic trans_CL – used in Nanga_2019_tacrolimus_metaanalysis.R (Table 3 covariate-effect label; encoded as a multiplicative power coefficient theta^TX_LIVER on CL/F).
• Example models: Nanga_2019_tacrolimus_metaanalysis.R (multiplicative effect on apparent oral clearance: cl_typ *= 0.38^TX_LIVER, so liver-graft recipients have CL/F 62% lower than non-liver recipients at the same body weight and post-transplant day; Nanga 2019 Table 3 ‘Hepatic trans_CL’ = 0.38), Perrottet_2009_ganciclovir.R (Perrottet 2009 pools lung and liver under a single CL slope; the model applies a shared multiplicative effect via (TX_LUNG + TX_LIVER): pooled lung-or-liver theta_lung/liver = 1.17 vs theta_kidney = 1.68, so lung-or-liver recipients have a CL slope 30% lower than kidney recipients at the same GFR and sex; Perrottet 2009 Table 3).
• Notes: Specific scope because the reference complement (non-liver solid-organ graft) is paper-defined – in Nanga 2019 the complement is exclusively kidney-transplant patients (Table 5: 201 liver, 80 kidney). Future pooled solid-organ-transplant popPK models that distinguish liver from non-liver patients should reuse this canonical; if a future model decomposes the non-liver class into separate kidney / heart / lung indicators, register sibling canonicals (TX_KIDNEY, TX_HEART, TX_LUNG) rather than overloading this entry. If instead a model contrasts transplant recipients against NON-transplant patients, that is TX_ANY, not this entry – TX_LIVER = 0 does not distinguish a non-liver graft recipient from a subject who was never transplanted. Distinct from POSTTX_DAY1 (first-24-hour-post-transplant indicator, time-varying) and from POD (continuous post-transplant day); all three can coexist when a model parameterises both organ type and time-after-transplantation effects.

DONOR_DECEASED (canonical for deceased-vs-living donor indicator (solid-organ transplant))

• Description: Binary indicator for the source of the transplanted solid-organ graft. 1 = recipient received the graft from a deceased (post-mortal / cadaveric) donor; 0 = recipient received the graft from a living donor. Time-fixed per subject (assigned at transplantation date). Donor source is a graft-quality / preservation-status proxy: deceased-donor grafts typically experience longer cold-ischaemia times, more variable preservation conditions, and a higher rate of slow / delayed graft function than living-donor grafts; in solid-organ-transplant popPK studies these differences can translate into a measurable shift in apparent oral clearance of drugs that undergo hepatic / intestinal CYP3A metabolism (the mechanism is not fully understood and is hypothesised to involve recipient catabolic state, free-fraction shifts, or other unmeasured covariates).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (living-donor graft – the more common subgroup in most paediatric and adult kidney-transplant cohorts).
• Source aliases:
  • donor / Donor – common NONMEM $INPUT form (used in Andrews_2017_tacrolimus.R; Andrews 2017 codes the column as a binary living-vs-deceased indicator).
• Example models: Andrews_2017_tacrolimus.R (multiplicative effect on apparent oral clearance: living-donor recipients have CL/F 26% lower than deceased-donor recipients of the same body weight, CYP3A5 genotype, eGFR, and hematocrit; Andrews 2017 Table 2 final-model coefficient: theta_donor_living = 0.74 for the living-donor cohort relative to the deceased-donor reference, equivalent to deceased-donor recipients having ~35% higher CL/F than living-donor recipients as reported in Section 3.4).
• Notes: General scope because donor source is a universally applicable solid-organ-transplant cohort attribute, captured in transplant-registry data for kidney, liver, heart, lung, and pancreas recipients. The reference category is 0 = living donor (not 0 = deceased donor) to match the “1 = the perturbed / index condition” convention used elsewhere in the register (e.g., RRT_HEMODIAL_STATUS, POSTTX_DAY1, TX_LIVER); the index condition is “received a deceased-donor graft” because it carries the more variable graft quality and is the condition that perturbs clearance upward in published kidney-tacrolimus models. Source papers that encode the column as “1 = living donor” (e.g., Andrews 2017 itself parameterises the equation with a multiplier on living-donor recipients) should still record the column under the canonical orientation: set DONOR_DECEASED = 1 - source_living_indicator and document the value inversion in covariateData[[DONOR_DECEASED]]$notes. Distinct from the donor-genotype canonical CYP3A5_EXPR_DONOR: donor-source (deceased vs living) is a logistic / graft-procurement covariate, while donor-genotype is a pharmacogenetic covariate; both can coexist in the same dataset when the source paper genotypes both recipients and donors.

PRIOR_GAST (canonical for prior gastrectomy)

• Description: Prior (partial or total) gastrectomy indicator, 1 = prior gastrectomy, 0 = no prior gastrectomy. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no prior gastrectomy).
• Source aliases:
  • GAST – used in Yamada_2025_zolbetuximab.R.
• Example models: Yamada_2025_zolbetuximab.R (fractional effects on CLss, CLT, V1).
• Notes: Applicable to any PK model where gastrointestinal anatomy affects absorption, first-pass, or protein turnover; not inherently oncology-specific. No distinction between partial vs total gastrectomy unless the source paper separates them.

PRIOR_SPLEN (canonical for prior splenectomy)

• Description: Prior splenectomy indicator, 1 = prior splenectomy (spleen surgically removed before or concurrent with the modelled treatment period), 0 = spleen intact. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (spleen intact).
• Source aliases:
  • splenectomy – used in PerezRuixo_2015_oxaliplatin_platelet_dynamics.R (Perez-Ruixo 2015 Table I: subject-level binary covariate on the platelet random-destruction rate constant ks).
• Example models: PerezRuixo_2015_oxaliplatin_platelet_dynamics.R (multiplicative effect on ks: ks_i = ks * Phi^PRIOR_SPLEN with Phi = 0.475, i.e., splenectomized patients have a 47.5% reduction in the random platelet destruction rate constant, prolonging platelet lifespan from 3.23 days in patients with an intact spleen to 7.78 days in splenectomized patients; Perez-Ruixo 2015 The AAPS Journal Table I and Discussion).
• Notes: Follows the PRIOR_GAST / PRIOR_TNF / PRICORT naming pattern for prior surgical-history and prior-treatment indicators (short surgical stem, all uppercase, binary 0/1). General scope because the mechanism (loss of splenic platelet sequestration / phagocytic clearance) is applicable across therapeutic areas – Petrov 2024 romiplostim also identified splenectomy as a subgroup with distinct platelet PK/PD though encoded it as a subpopulation split rather than a data-driven column. Distinct from INTRAOP (active-surgery indicator, time-varying) and from POD (post-operative day, monotonically increasing continuous).

TX_HEART (canonical for heart (cardiac) transplant indicator)

• Description: Binary indicator for heart / cardiac-graft transplant recipients in a pooled solid-organ-transplant cohort. 1 = heart transplant; 0 = non-heart solid-organ transplant (kidney, lung, or liver). Time-fixed per subject (assigned at transplantation date).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-heart solid-organ graft – typically kidney as the most common non-heart comparator in pooled solid-organ-transplant analyses).
• Source aliases:
  • graft_type – pooled categorical column with levels K / H / Lu / Li; decompose into binary TX_HEART = as.integer(graft_type == "H") and document the decomposition in covariateData[[TX_HEART]]$notes. Used in Perrottet_2009_ganciclovir.R.
• Example models: Perrottet_2009_ganciclovir.R (multiplicative effect on the CL-vs-GFR slope: theta_heart = 0.86 vs theta_kidney = 1.68, so heart-transplant recipients have a CL slope 49% lower than kidney recipients at the same GFR and sex; Perrottet 2009 Table 3), Fromage_2025_mycophenolic_acid.R (pooled with TX_LUNG and TX_HCT into the source’s single group-(ii) level carrying one estimated effect on the second gamma absorption rate constant: -0.59 on log(MAT2), shortening the second-peak mean absorption time from 4.13 h to 2.29 h versus the renal / hepatic reference; Fromage 2025 Table 2).
• Notes: Specific scope because the reference complement is paper-defined; sibling to TX_LIVER, TX_LUNG, and the implicit kidney-reference cohort. Distinct from TX_ANY, whose reference category is a non-transplant patient rather than a different solid-organ graft.

TX_LUNG (canonical for lung (pulmonary) transplant indicator)

• Description: Binary indicator for lung / pulmonary-graft transplant recipients in a pooled solid-organ-transplant cohort. 1 = lung transplant; 0 = non-lung solid-organ transplant (kidney, heart, or liver). Time-fixed per subject (assigned at transplantation date).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-lung solid-organ graft – typically kidney as the most common non-lung comparator in pooled solid-organ-transplant analyses).
• Source aliases:
  • graft_type – pooled categorical column with levels K / H / Lu / Li; decompose into binary TX_LUNG = as.integer(graft_type == "Lu") and document the decomposition in covariateData[[TX_LUNG]]$notes. Used in Perrottet_2009_ganciclovir.R.
• Example models: Perrottet_2009_ganciclovir.R (Perrottet 2009 pools lung and liver under a single CL slope; the model applies a shared multiplicative effect via (TX_LUNG + TX_LIVER): pooled lung-or-liver theta_lung/liver = 1.17 vs theta_kidney = 1.68, so lung-or-liver recipients have a CL slope 30% lower than kidney recipients at the same GFR and sex; Perrottet 2009 Table 3), Fromage_2025_mycophenolic_acid.R (pooled with TX_HEART and TX_HCT into the source’s single group-(ii) level carrying one estimated effect on the second gamma absorption rate constant: -0.59 on log(MAT2), shortening the second-peak mean absorption time from 4.13 h to 2.29 h versus the renal / hepatic reference; Fromage 2025 Table 2).
• Notes: Specific scope because the reference complement is paper-defined; sibling to TX_LIVER, TX_HEART, and the implicit kidney-reference cohort. Distinct from TX_ANY, whose reference category is a non-transplant patient rather than a different solid-organ graft. Future models that estimate distinct lung-vs-liver effects (rather than pooling) should keep this canonical for lung and continue to use TX_LIVER for liver.

TX_LUNG_BILAT (canonical for bilateral (vs single) lung graft laterality indicator)

• Description: Binary indicator of lung-graft laterality within a cohort in which every subject is a lung transplant recipient. 1 = bilateral (double) lung transplant; 0 = single (unilateral) lung transplant. Time-fixed per subject (assigned at transplantation). Used as a proxy for the magnitude of the surgical insult: bilateral transplantation requires a longer duration of cardiopulmonary bypass, a median sternotomy / clamshell rather than a thoracotomy, and carries a greater ischemia-reperfusion burden, so it acts on early postoperative drug disposition through the associated critical-illness physiology rather than through any lung-specific elimination pathway.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (single / unilateral lung transplant).
• Source aliases:
  • transplant type – used in Miano_2024_tacrolimus.R (Miano 2024 Methods “Clinical covariates” lists “transplant type (single vs bilateral)”; Table 2 footnote a reads “Single vs bilateral”, with single-lung as the reference level carrying the factor 1.0).
• Example models: Miano_2024_tacrolimus.R (multiplicative effect on CL/F with a time-varying coefficient: theta^TX_LUNG_BILAT with theta = 0.48 on postoperative days 1-3 and theta = 0.82 on postoperative days 4-14, i.e. bilateral recipients clear tacrolimus 52% more slowly than single-lung recipients in the first three postoperative days and 18% more slowly thereafter; Miano 2024 Table 2 and the displayed CL/F equation in Results. The time-varying specification was required: a single time-invariant coefficient gave Delta-OFV = -2.87 and did not reach significance, whereas the early/late split gave Delta-OFV = -576.71 – Miano 2024 Table 3).
• Notes: Distinct in kind from TX_LUNG, whose contrast is a different organ (“1 = lung transplant; 0 = non-lung solid-organ transplant”). TX_LUNG_BILAT partitions within a lung-only cohort on graft laterality, so in such a cohort TX_LUNG = 1 for every subject and carries no information. The two can coexist in a pooled multi-organ dataset, where TX_LUNG_BILAT is defined only on the TX_LUNG = 1 subset; models that do so must state the coding for non-lung recipients (typically 0, folded into the single-lung reference) in covariateData[[TX_LUNG_BILAT]]$notes. Also distinct from SURG_SEV_MAJOR (a generic major-vs-minor procedure-severity flag that is not transplant-specific) and from INTRAOP / POD (which locate an observation in time relative to surgery rather than describing the procedure). Papers that report the reversed orientation (a single-lung indicator) should still record their column under TX_LUNG_BILAT with the values flipped and the flip documented per-model, so the bilateral-equals-1 orientation is preserved across the register. A future cohort that estimates a heart-lung or lobar-transplant contrast should register a sibling canonical rather than overloading this one.

TX_ANY (canonical for any-solid-organ-transplant indicator against a non-transplant reference)

• Description: Binary indicator that the subject is the recipient of any solid-organ graft, contrasted against a non-transplant patient. 1 = recipient of any solid-organ transplant (kidney, liver, heart, lung, pancreas, or a combination); 0 = non-transplant patient. Time-fixed per subject (assigned at transplantation date). Used when a population PK analysis pools transplant recipients with patients who have never been transplanted and tests the transplant-vs-non-transplant contrast as a PK covariate – typically because transplant recipients differ from the rest of the cohort in renal function, immunosuppressive co-medication, and general physiological reserve in ways the measured covariates do not fully capture.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-transplant patient; the complement group is defined per-model – in Yuen 1995 it is the CMV-retinitis and CMV-urine-shedding patients who make up 48 of the 53 subjects).
• Source aliases:
  • T – single-letter column name used in Yuen_1995_ganciclovir.R (Yang 2023 Table 3 footnote: “T: T = 0 for non-transplant patients and 0.76 for transplant patients”; the source folds the coefficient into the column’s coded value, so store the canonical column as a plain 0 / 1 indicator and carry 0.76 as a named ini parameter).
  • transplant (yes/no) – the covariate label used in Yang 2023 Table 4 for the same column.
• Example models: Yuen_1995_ganciclovir.R (fractional reduction of the renal-elimination term of clearance: cl = exp(lcl) + e_wt_crcl_cl * WT * CRCL/100 * (1 - 0.76 * TX_ANY) * (1 - 0.41 * DIS_CMV_RETINITIS), i.e. a transplant recipient’s weight-and-creatinine-clearance-driven clearance term is 76% lower than a non-transplant patient’s at the same weight and creatinine clearance; Yang 2023 Table 3 and its footnote).
• Notes: Distinct in kind from the organ-specific canonicals TX_LIVER / TX_HEART / TX_LUNG, which partition within a transplant cohort (their reference category is a different solid-organ graft, not the absence of a transplant). TX_ANY and the organ-specific siblings can coexist in one dataset – TX_ANY = 1 holds for every subject with any TX_* indicator set – but TX_ANY must NOT be inferred from the organ-specific indicators when the cohort contains non-transplant patients, because a subject with TX_LIVER = 0 may be either a non-liver graft recipient or a non-transplant patient. Also distinct from DIS_CMV (transplant-recipient-with-CMV vs non-CMV subject: that canonical conflates transplant status with CMV infection and its reference group is a healthy volunteer or phase I participant, not a non-transplant patient) and from DONOR_DECEASED (graft source among transplant recipients). Scope: specific, because the reference complement is paper-defined.

TX_HCT (canonical for haematopoietic cell (bone marrow / stem cell) transplant indicator)

• Description: Binary indicator for haematopoietic cell transplant recipients – bone marrow, peripheral blood stem cell, or cord blood graft – in a cohort that also contains other indications. 1 = haematopoietic cell transplant recipient; 0 = not a haematopoietic cell transplant recipient (the complement is paper-defined; in Fromage_2025_mycophenolic_acid.R it is renal or hepatic transplantation as the reference group, with cardiac and pulmonary transplantation and autoimmune disease as the other levels, while in Prohn_2021_letermovir_hsct.R it is healthy phase I volunteers pooled into the same fit). Time-fixed per subject (assigned at transplantation date).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (non-haematopoietic-cell-transplant patient; the complement group is paper-defined).
• Source aliases:
  • bone marrow transplantation – the wording used in the covariate definition of Fromage_2025_mycophenolic_acid.R (Fromage 2025 Section 2.2.2).
  • haematopoietic stem cell (HSC) recipients, HSCT, HSC, BMT – equivalent designations for the same graft class; all map to TX_HCT.
  • Indication – pooled categorical column whose levels include renal / hepatic / cardiac / pulmonary / bone marrow transplantation and autoimmune disease; decompose into binary TX_HCT = as.integer(Indication == "bone marrow transplantation") and document the decomposition in covariateData[[TX_HCT]]$notes.
• Example models: Prohn_2021_letermovir_hsct.R (the indicator that separates the 363 HSCT recipients from the 36 healthy phase I participants pooled into the same fit; it switches BOTH oral bioavailability – 0.346 without cyclosporine in HSCT recipients against 1.00 fixed in healthy participants – and the absorption rate, 0.150 against 1.26 1/h; Prohn 2021 Table 3), Fromage_2025_mycophenolic_acid.R (one of three transplant indicators – TX_HEART / TX_LUNG / TX_HCT – that the source pools into a single group-(ii) level carrying one estimated effect on the second gamma absorption rate constant: beta_b2 = +0.59 on log(b2), equivalently -0.59 on log(MAT2), halving the mean absorption time of the second peak from 4.13 h to 2.29 h relative to the renal / hepatic reference; Fromage 2025 Table 2 and Discussion).
• Notes: Promoted from Scope: specific to Scope: general on the strength of a second, independent source using the column with identical value semantics (Prohn 2021 letermovir, a different drug, indication and modelling group from the founding Fromage 2025 mycophenolic-acid entry). What varies between the two is only the paper-defined complement group, which the Description already records per model, so the column generalises without a semantic change. Sibling to the solid-organ indicators TX_LIVER / TX_HEART / TX_LUNG; distinct in kind from them in that a haematopoietic cell graft is not a solid organ, so a model pooling both (as Fromage 2025 does) is pooling across graft classes and should say so in its covariateData notes. Fromage_2025_mycophenolic_acid.R does NOT use TX_ANY, so no statement about whether that model’s TX_ANY includes HCT is required; a future model combining the two must decide explicitly and record it in the TX_ANY entry. Where a source pools several transplant types under one estimated effect, keep the per-type columns canonical and form the pooled indicator inside model(), so a later model that separates the effects can reuse the same columns.

DIS_ALOPECIA_AT_AU (canonical for alopecia totalis / universalis indicator within an alopecia areata cohort)

• Description: Binary indicator that a participant’s alopecia areata has progressed to alopecia totalis (AT; complete scalp hair loss) or alopecia universalis (AU; complete scalp and body hair loss), the severe end of the alopecia areata spectrum, as opposed to patchy alopecia areata not meeting AT or AU criteria. 1 = AT or AU; 0 = non-AT/AU alopecia areata. Time-fixed per subject (classification at study entry). Used when an analysis contrasts two alopecia areata presentations inside an alopecia areata cohort, rather than contrasting alopecia areata subjects against a non-alopecia reference.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-AT/AU alopecia areata, i.e. patchy disease).
• Source aliases:
  • AT/AU status, disease severity – the labels used in Huh_2024_ritlecitinib.R (Huh 2024 Section 2.3 lists it among the tested covariates as “disease severity (alopecia totalis [AT]/alopecia universalis [AU] status)”; Section 3.2 and Table 1 use the “non-AT/AU” and “AT/AU” group names).
• Example models: Huh_2024_ritlecitinib.R (the only covariate retained in the final longitudinal exposure-response model of the Severity of Alopecia Tool score, entering on three parameters: separate baseline typical values on the transformed scale, 11.6 for AT/AU vs 1.92 for non-AT/AU; separate maximum placebo effects, 2.75 for AT/AU vs 0 fix for non-AT/AU; and a -60.1% fractional shift on the drug-chain turnover half-life, 3.11 vs 7.80 weeks. Huh 2024 Table 2).
• Notes: This is a within-alopecia-areata severity contrast and is therefore ORTHOGONAL to DIS_ALOPECIA_AREATA, which encodes alopecia areata versus a non-alopecia reference cohort (healthy volunteers and other inflammatory-disease cohorts in Wojciechowski_2023_ritlecitinib_base.R / _updated.R). Both can appear in one dataset, and every subject in a DIS_ALOPECIA_AT_AU cohort – AT/AU and non-AT/AU alike – carries DIS_ALOPECIA_AREATA = 1; reusing DIS_ALOPECIA_AREATA for the severity split would overload it with incompatible semantics. Same construction as DIS_CMV_RETINITIS (a symptomatic-presentation split nested inside DIS_CMV). AT and AU are deliberately pooled into one level because the source pools them; register a sibling canonical only if a future paper estimates AT and AU effects separately. Note that AT/AU participants have a baseline Severity of Alopecia Tool score of 100 by definition of the classification, so this indicator is collinear with a baseline-SALT-of-100 indicator in any alopecia areata dataset – an analysis that carries both must say which one it means. Scope: specific, because the reference complement is a disease-specific presentation.

DIS_AUTOIMMUNE (canonical for pooled autoimmune-disease indication indicator)

• Description: Binary indicator that the treatment indication is an autoimmune disease, pooled across the specific autoimmune diagnoses present in the cohort, contrasted against a non-autoimmune indication. 1 = autoimmune-disease indication; 0 = non-autoimmune indication (the complement is paper-defined; in Fromage_2025_mycophenolic_acid.R it is transplantation). Time-fixed per subject. Use this only when the source pools two or more distinct autoimmune diagnoses into a single estimated level and reports no per-diagnosis effect; when the source names a single autoimmune disease, use that disease’s own canonical instead.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-autoimmune indication; the complement group is paper-defined – renal or hepatic transplantation in Fromage 2025).
• Source aliases:
  • autoimmune diseases – the level name used in Fromage_2025_mycophenolic_acid.R (Fromage 2025 Section 2.2.2), covering systemic lupus erythematosus, lupus nephritis and nephrotic syndrome.
  • AID – the abbreviation used in the Fromage 2025 Abstract.
  • Indication – pooled categorical column; decompose into binary DIS_AUTOIMMUNE = as.integer(Indication == "autoimmune disease") and document the decomposition in covariateData[[DIS_AUTOIMMUNE]]$notes.
• Example models: Fromage_2025_mycophenolic_acid.R (effect on the second gamma absorption rate constant: beta_b2 = -0.17 on log(b2), equivalently +0.17 on log(MAT2), lengthening the mean absorption time of the second peak from 4.13 h to 4.90 h relative to the renal / hepatic reference; Fromage 2025 Table 2 reports this effect with 188% RSE and Section 3.2.2 states it “can be disregarded”, but the authors retained it for their pcVPC simulations and the packaged model retains it accordingly).
• Notes: Distinct from the single-disease canonicals DIS_SLE, DIS_RA, DIS_UC, DIS_PBC, DIS_ALOPECIA_AREATA, DIS_VITILIGO and the like: those identify one named diagnosis, whereas DIS_AUTOIMMUNE deliberately pools several. Prefer the single-disease canonical whenever the source resolves the diagnosis; DIS_AUTOIMMUNE exists because some sources (Fromage 2025 among them, whose autoimmune group holds only 13 patients) have too few subjects per diagnosis to estimate separate effects and therefore pool lupus, lupus nephritis and nephrotic syndrome into one level. A dataset may legitimately carry both DIS_AUTOIMMUNE = 1 and DIS_SLE = 1 for the same subject; DIS_AUTOIMMUNE must NOT be inferred from the absence of the single-disease indicators, because an autoimmune diagnosis outside the registered set would then be missed. Scope: specific, because the reference complement is paper-defined.

ANESTH_GA (canonical for general-anaesthesia cohort indicator over the PK observation window)

• Description: 1 = the subject received general anaesthesia (any induction / maintenance regimen) at some point within the pharmacokinetic observation window; 0 = the subject was conscious and unanaesthetised throughout. A cohort / per-administration indicator, not a time-varying window: it stays 1 for the whole of an anaesthetised subject’s profile, including samples drawn after extubation. Captures the same composite of confounded perturbations that INTRAOP does – anaesthetic-agent vasodilation, intra-operative fluid loading, altered cardiac output and regional perfusion, surgical stress – but attributes it to having been anaesthetised rather than to being mid-procedure, which is the only contrast a dataset can support when the anaesthetised subjects are sampled post-operatively or sparsely.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (conscious, unanaesthetised).
• Source aliases:
  • ANEST – used in Pelligand_2016_robenacoxib_cat.R (Pelligand 2016 Methods “Inclusion of covariate relationships” lists ANEST among the four screened covariates as “perioperative status”; Table 2 carries it as two separate V1 rows labelled “ANEST = 0” and “ANEST = 1”).
• Example models:
  • Pelligand_2016_robenacoxib_cat.R (additive effect on the log central volume of distribution, vc = exp(lvc + e_anesth_ga_vc * ANESTH_GA + etalvc) with e_anesth_ga_vc = log(0.33 / 0.166) = 0.687, so an anaesthetised cat carries 1.99 times the conscious central volume – 0.166 -> 0.330 L/kg. The only covariate retained by the paper’s BIC backward elimination out of anaesthesia, age, body weight and sex screened on CL, V1 and Ftot. The anaesthetised cohort is 36 clinical cats given robenacoxib s.c. around elective ovariohysterectomy under buprenorphine / acepromazine premedication, propofol induction and isoflurane maintenance; the reference cohort is 47 conscious laboratory cats).
• Notes: Distinct from INTRAOP, and the distinction is load-bearing rather than cosmetic: INTRAOP is time-varying within a subject and its reference category is explicitly “pre- or post-operative”, so coding an anaesthetised subject’s post-extubation samples as INTRAOP = 1 would assert something the source contradicts. Use INTRAOP when the source resolves the active-surgery window per record (Stricker 2013); use ANESTH_GA when the source contrasts an anaesthetised cohort against an unanaesthetised one (Pelligand 2016). The two can coexist in a dataset that pools both designs. Distinct from CONMED_SEVO / ETSEVO (and the anticipated CONMED_ISO / CONMED_DES), which name a specific volatile agent: ANESTH_GA is agent-agnostic and must be used when the source attributes the effect to general anaesthesia as a whole rather than to one drug – Pelligand 2016 Discussion names acepromazine, propofol and isoflurane as candidate mechanisms and identifies none, so narrowing the column to any single agent would misattribute the effect. Also distinct from SURG_SEV_MAJOR (per-procedure severity), SURG_CARD_VASOACTIVE (post-cardiotomy low-output state), T_ANAESTHESIA (elapsed time from induction, continuous), PFA / PFA_NET_RATE (measured fluid volumes) and POD / POSTTX_DAY1 (post-surgical recovery time course). Specific scope until a second paper ratifies the name; at that point promote to general. In the founding paper anaesthesia is perfectly confounded with cohort and with sparse sampling – all anaesthetised cats came from the perioperative study and all were female – which Pelligand 2016 flags as a design limitation; record any such confounding in covariateData[[ANESTH_GA]]$notes per model.

INTRAOP (canonical for intra-operative period indicator (active surgery))

• Description: 1 = the observation time falls within the intra-operative period (active surgery, between the start of the surgical procedure and skin closure / end of surgery); 0 = pre- or post-operative. Time-varying within subject. The exact boundaries of the intra-operative window are paper-specific (e.g., “from end of induction to skin closure”, “from the post-loading-dose PK sample to end of surgery”, “from incision to last suture”) – document the per-model boundaries in covariateData[[INTRAOP]]$notes. Used as a categorical covariate to capture composite physiological changes during active surgery (hemodynamic effects of general anaesthesia, blood loss, fluid resuscitation, surgical stress, temperature management) that are jointly unidentifiable individually in typical perioperative PK datasets.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (pre- or post-operative; i.e., not during active surgery).
• Source aliases:
  • (paper-prose) – used in Stricker_2013_aminocaproicAcid.R (Stricker 2013 Methods “Full covariate model” and Table 4: the intra-operative period is defined as “the time immediately after the post-loading dose PK sample through the end of the surgery”; CL and V1 are estimated separately for pre/postoperative vs intraoperative periods, with multiplicative shifts of 0.89 on CL and 0.80 on V1 during the intra-operative window).
• Example models: Stricker_2013_aminocaproicAcid.R (multiplicative effect on CL and V1: cl = cl_postop * 0.89^INTRAOP, vc = vc_postop * 0.80^INTRAOP; reference category is pre/postoperative).
• Notes: Distinct from POD (post-operative day – continuous days since surgery, rising monotonically from 0) and POSTTX_DAY1 (binary first-24-hours-post-transplant indicator). INTRAOP captures the active-surgery window itself, whereas POD and POSTTX_DAY1 capture the post-surgical recovery time course. The three canonicals can coexist in a single dataset that pools perioperative and post-discharge PK samples. The intra-operative effect is typically interpreted as a composite of multiple confounded physiological perturbations and is not directly attributable to any single mechanism (e.g., Stricker 2013 Discussion: “the model results may be simply reflecting the net effect of a possible increased CL due to blood loss and decreased CL due to other confounding intra-operative factors”). Pairs with a future INTRAOP_BL (per-subject calculated intra-operative blood loss) canonical when a source paper reports a continuous blood-loss column alongside or instead of the binary indicator – that is a separate canonical to be proposed when needed.

SURG_SEV_MAJOR (canonical for major (vs minor) surgical procedure severity indicator)

• Description: 1 = the surgical procedure is classified as major or high-risk according to a paper-specific severity taxonomy (e.g., major orthopedic replacement, major abdominal / thoracic / vascular surgery); 0 = the procedure is classified as minor (e.g., central-line placement, small ENT / ophthalmic procedures, dental extraction, biopsy). Time-fixed per surgical procedure (a subject undergoing multiple surgeries in the study window may have different SURG_SEV_MAJOR values on each occasion). Distinct from INTRAOP (a time-varying within-surgery window indicator) – SURG_SEV_MAJOR is a per-procedure severity classification that persists across the perioperative period. Captures composite physiological differences between major and minor surgery (larger blood loss / fluid resuscitation, longer duration, higher acute-phase response, more consumption of coagulation factors) that can shift factor concentrate PK in bleeding disorders.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (minor surgical procedure).
• Source aliases:
  • severity of surgical procedure (paper-prose 0/1 indicator, 1 = major) – used in Hazendonk_2016_factor_viii.R (Hazendonk 2016 Table 5: 0.93 ^ severity on CL, so major-surgery cases show 7% lower typical FVIII clearance than minor-surgery cases; classification per Koshy et al. 1995 with major and high-risk collapsed together).
• Example models: Hazendonk_2016_factor_viii.R (multiplicative effect on CL: (1 + (-0.07) * SURG_SEV_MAJOR) – equivalent to 0.93 ^ SURG_SEV_MAJOR when binary; reference category 0 = minor surgery).
• Notes: General scope because major-vs-minor surgery severity classification recurs across perioperative popPK models for hemostatic factor concentrates and other drugs where surgical stress affects clearance. The exact severity taxonomy is paper-specific – Hazendonk 2016 follows Koshy et al. 1995 (a hemophilia-specific major / minor / high-risk classification, with major and high-risk collapsed into SURG_SEV_MAJOR = 1), while other surgical popPK models may use ASA physical-status class, RACHS-1 (cardiac only), or ad-hoc paper-specific criteria. Record the source-paper severity taxonomy in covariateData[[SURG_SEV_MAJOR]]$notes. Distinct from INTRAOP (time-varying within-surgery window indicator, applicable to any procedure regardless of severity), from RACHS1 (ordinal cardiac-surgery-specific risk score, cardiac only), and from MAJORSURG_HISTORY (would be a lifetime prior-major-surgery indicator, not yet ratified). Sign of the effect on CL is paper-specific and physiologically ambiguous: major surgery can plausibly increase FVIII clearance through blood loss and consumption OR decrease it through VWF-mediated stabilization from the surgical stress response (Hazendonk 2016 Discussion attributes the 7% decrease to confounding by age – older patients underwent more major procedures).

SURG_CARD_VASOACTIVE (canonical for cardiothoracic surgery with concomitant vasoactive support)

• Description: 1 = the subject underwent cardiothoracic surgery AND is receiving concomitant vasoactive / inotropic agents (milrinone, dopamine, dobutamine, epinephrine, norepinephrine, vasopressin, or equivalent) over the drug-therapy window; 0 = neither, or cardiothoracic surgery without concomitant vasoactive agents. Both halves of the conjunction are required for a 1 – the surgery alone does not qualify – because the column is a marker of the post-cardiotomy LOW-CARDIAC-OUTPUT state rather than of the operation itself. Captures the composite haemodynamic perturbation of that state (reduced renal perfusion, cardiopulmonary-bypass-associated systemic inflammatory response, elevated acute-kidney-injury risk) as it depresses renal drug clearance. Typically time-fixed per subject over a single course of therapy.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no cardiothoracic surgery, or cardiothoracic surgery without concomitant vasoactive agents).
• Source aliases:
  • CTS – used in Chen_2023_vancomycin.R (Chen 2023 Methods, ‘Patients and setting’: “We defined CTS patients as those who underwent CTS and received concomitant vasoactive agents (including milrinone, dopamine, epinephrine or norepinephrine) during vancomycin therapy”); same orientation as the canonical, no value transformation.
• Example models: Chen_2023_vancomycin.R (multiplicative power-of-binary effect on CL: cl *= e_cts_cl ^ SURG_CARD_VASOACTIVE with e_cts_cl = 0.761, i.e. 23.9% lower vancomycin clearance in post-cardiothoracic-surgery children on vasoactive support relative to the non-CTS reference; Chen 2023 Table 3 theta_1 and the ‘The final model’ equation block).
• Notes: Specific scope pending corroboration by a second model that ratifies the same conjunction semantics. Deliberately distinct from three neighbours that each capture only one half of it, and none of which should be substituted: CONMED_INOTROPE is vasoactive coadministration in ANY patient regardless of surgical history; SURG_SEV_MAJOR is a major-vs-minor procedure severity classification with no haemodynamic-support requirement and no cardiac specificity; and T_CPB / CPB_ON / CPB_REWARM describe the bypass run itself (its duration, or an intra-operative phase) rather than the post-operative state. A future paper that retains a PLAIN cardiothoracic- or cardiac-surgery indicator with no vasoactive-agent requirement should register a sibling (e.g. SURG_CARD) rather than reuse this column, because the two select materially different subsets – Chen 2023’s conjunction picked out 68 of 673 subjects (10.1%), a strict subset of everyone who had the operation. Record the per-paper vasoactive-agent list in covariateData[[SURG_CARD_VASOACTIVE]]$notes, as the pooled drug set is paper-specific.

T_CPB (canonical for total cardiopulmonary bypass duration during the most recent cardiac surgery)

• Description: Total time the patient spent on cardiopulmonary bypass (CPB) during their most recent cardiac-surgery operation. Time-fixed per subject (the duration of the bypass run is a single scalar set at the close of surgery). Captures composite physiological / pharmacological effects of CPB on post-operative drug disposition (hemodilution from priming volume, systemic inflammatory response, transient hepatic / renal hypoperfusion).
• Units: minutes
• Type: continuous
• Scope: general
• Reference category: n/a – typically enters as a power-centred effect (T_CPB / ref)^exponent. Reference value is paper-specific (cohort median total CPB time): Su 2016 uses 60 minutes (overall cohort median, Table 3).
• Source aliases:
  • TBYP – used in Su_2016_dexmedetomidine.R (Su 2016 NONMEM column for total bypass time in minutes; Methods ‘Full Covariate Model’ equation and Table 4 footnote).
• Example models: Su_2016_dexmedetomidine.R (power effect on CL: (T_CPB / 60)^(-0.31); negative exponent so longer bypass time reduces post-operative CL).
• Notes: General scope because CPB time is a routine clinical covariate in any popPK study of patients undergoing open heart surgery. Distinct from T_POST_ECMO (time AFTER decannulation from ECMO – T_CPB captures the DURATION of the bypass procedure during surgery, not time since the end of it; T_POST_ECMO is time-varying within subject whereas T_CPB is time-fixed). Distinct from POD (post-operative day) and INTRAOP (binary intra-operative indicator). Per the T_ canonical family, the T_ prefix denotes a procedure-related time covariate; for T_CPB the “event” is the bypass run itself and the column carries the duration of that run. Future paediatric / adult cardiac-surgery popPK extractions that retain CPB time should reuse this canonical.

T_SURG (canonical for duration of the surgical operation (operative time))

• Description: Total duration of the patient’s most recent surgical operation, from start to end of surgery (the “operative time”). Time-fixed per subject: a single scalar set at the close of surgery. Member of the T_<event> family in its duration sense, like T_CPB (bypass duration) and T_PUMP (pumping-session duration). Not the time elapsed SINCE surgery – use POD (post-operative day) or a T_POST_<event> column for that.
• Units: h
• Type: continuous
• Scope: general
• Reference category: n/a – typically a power-centred effect (T_SURG / ref)^exponent. Reference values observed: 5 h (Wang 2020 caspofungin, printed in the final-model equations).
• Source aliases:
  • OPT – Wang 2020 final-model equations and Table 4 (‘Theta OPT on CL’, ‘Theta OPT on VC’); used in Wang_2020_caspofungin.R.
• Example models: Wang_2020_caspofungin.R (power effects on clearance and central volume: (T_SURG / 5)^1.30 and (T_SURG / 5)^0.93; longer lung-transplant operations raised both).
• Notes: Distinct from T_CPB, which is the duration of the cardiopulmonary-bypass run within a cardiac operation rather than of the operation as a whole, and from T_ANAESTHESIA / T_INCISION, which are times relative to the modelled treatment. Hours, not minutes: a paper reporting operative time in minutes must divide by 60 (and scale any centring value) before using this column.

CPB_ON (canonical for cardiopulmonary bypass phase indicator, before rewarming begins)

• Description: 1 = the observation or dosing record falls within the cardiopulmonary bypass (CPB) phase proper, i.e. from commencement of bypass until rewarming begins; 0 = pre-CPB, during rewarming, or post-CPB. Time-varying within subject. Captures the acute intra-CPB physiological perturbation – hemodilution by the circuit priming volume, non-pulsatile flow, hypothermia, altered regional perfusion, and drug sequestration onto circuit components – as it applies during bypass but before the rewarming manoeuvre. Pairs with CPB_REWARM (below): the two are mutually exclusive, and the patient is on the bypass circuit whenever CPB_ON + CPB_REWARM == 1. A source paper that records the intra-operative CPB phase as a single multi-level column (commonly coded 0 = pre-CPB, 1 = CPB, 2 = warming, 3 = post-CPB) decomposes into these two binary indicators per the standing policy of decomposing categorical covariates rather than testing levels inside model; the post-CPB level carries no indicator of its own and therefore collapses onto the pre-CPB reference unless a paper retains a distinct post-CPB effect (in which case register a sibling CPB_POST).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (pre-CPB or post-CPB; i.e. off the bypass circuit).
• Source aliases:
  • CPB phase (0/1/2/3, level 1) – used in Khaowroongrueng_2024_sufentanil.R (Khaowroongrueng 2024 Methods ‘CPB-adjusted model’: “CPB phases were recorded in the dataset as 0, 1, 2, and 3, denoting the pre-CPB, CPB, warming, and post-CPB phases, respectively”; the paper’s Equations 1-4 give a plain multiplicative fold-change on the typical parameter value within each phase).
• Example models: Khaowroongrueng_2024_sufentanil.R (multiplicative effect on the central volume during the CPB phase only: vc = exp(lvc) * e_cpb_vc^CPB_ON with e_cpb_vc = 2.74, a 2.74-fold hemodilution-driven increase in V1; the clearance effect spans both bypass windows and so uses e_cpb_cl^(CPB_ON + CPB_REWARM) with e_cpb_cl = 2.80; reference category is pre-/post-CPB).
• Notes: General scope because the intra-operative CPB phase is a routine time-varying covariate in any popPK study of open-heart surgery, and the CPB-versus-rewarming split recurs whenever a paper resolves the bypass run into sub-phases. Distinct in kind from T_CPB, which is the time-fixed total duration of the bypass run (continuous, minutes) rather than a within-surgery window indicator – the two can coexist in one dataset. Distinct from INTRAOP, whose window is the whole active-surgery period: in a CPB cohort every record is intra-operative, so INTRAOP would be constant 1 and carry no information, which is precisely why this finer canonical is needed. Distinct from the ECMO family (T_ECMO, T_POST_ECMO, ECMO_PUMP_SPEED) – a different extracorporeal circuit, and those time covariates are continuous rather than phase indicators. Distinct from the subject-level extracorporeal-therapy status flags RRT_CRRT_STATUS / RRT_HEMODIAL_STATUS. Because the name reads as though it should cover the whole time on the circuit, record explicitly in covariateData[[CPB_ON]]$notes that rewarming is excluded and is carried by CPB_REWARM.

CPB_REWARM (canonical for cardiopulmonary bypass rewarming phase indicator)

• Description: 1 = the observation or dosing record falls within the rewarming phase of cardiopulmonary bypass, i.e. from the start of active rewarming until separation from the bypass circuit; 0 = pre-CPB, during the CPB phase proper, or post-CPB. Time-varying within subject. Mutually exclusive with CPB_ON. Separates the rewarming manoeuvre from the preceding hypothermic bypass phase, which matters when a paper finds a covariate effect whose window is one but not the other (e.g. a clearance effect that persists through rewarming while a hemodilution-driven volume effect does not). The rewarming target is paper-specific (e.g. a rectal temperature of at least 36 C before separation from CPB) – document the per-model boundary in covariateData[[CPB_REWARM]]$notes.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (pre-CPB, CPB proper, or post-CPB).
• Source aliases:
  • CPB phase (0/1/2/3, level 2) – used in Khaowroongrueng_2024_sufentanil.R (Khaowroongrueng 2024 Methods ‘CPB-adjusted model’; level 2 of the paper’s single 0/1/2/3 phase column is the “warming” phase, and Methods ‘CPB system’ records that “all patients were rewarmed to a rectal temperature of >= 36 C before they were separated from CPB”).
• Example models: Khaowroongrueng_2024_sufentanil.R (enters only through the summed clearance window cl = exp(lcl + etalcl) * e_cpb_cl^(CPB_ON + CPB_REWARM) with e_cpb_cl = 2.80; Khaowroongrueng 2024 Results found the CPB effect on clearance during the CPB and rewarming phases “exhibited similarity; therefore, the CPB effect in these phases was estimated using the same typical value”, while the central-volume effect was retained for the CPB phase only, so CPB_REWARM does not appear in the vc equation).
• Notes: General scope for the same reason as CPB_ON. The pair exists so that two different covariate windows over one bypass run are separately addressable: a shared effect across both is written e_<cov>_<param>^(CPB_ON + CPB_REWARM) (the sum is 0 or 1 because the indicators are mutually exclusive), while a CPB-only or rewarming-only effect uses the single indicator. Do NOT encode a shared effect by widening CPB_ON to include rewarming – that would make the two canonicals overlap and silently break any model that needs the narrow window. Distinct from T_CPB (time-fixed total bypass duration) and INTRAOP (whole active-surgery window). If a future paper retains a distinct post-CPB effect, register a sibling CPB_POST rather than reusing either of these; in the absence of such an indicator the post-CPB phase collapses onto the pre-CPB reference.

CPB_POST (canonical for post-cardiopulmonary-bypass (postoperative) period indicator)

• Description: 1 = the observation or dosing record falls after separation from the cardiopulmonary bypass circuit, i.e. the postoperative period; 0 = pre-CPB, during the CPB phase proper, or during rewarming. Time-varying within subject. Mutually exclusive with CPB_ON and CPB_REWARM, so the three indicators partition the perioperative record and CPB_ON + CPB_REWARM + CPB_POST == 0 identifies the pre-CPB reference window. Captures the disposition changes that persist after the circuit is removed and that are therefore not attributable to the circuit itself – most commonly renal impairment from CPB-associated acute kidney injury and ischemia-reperfusion, altered hemodynamics after cardiac stunning, and residual hemodilution from intra-operative fluid support. Register this indicator only when a paper retains a distinct post-CPB effect; in its absence the post-CPB phase collapses onto the pre-CPB reference (see CPB_ON).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (pre-CPB; the phase before commencement of bypass, when CPB_ON and CPB_REWARM are also 0).
• Source aliases:
  • OCC == 4 – used in Chu_2024_allopurinol.R (Chu 2024 ESM S3 NONMEM control stream: FLAG3 = 0 then IF(POC.EQ.4) FLAG3=1 ; post-cpb, with the postoperative typical value selected by TVCLA = (CLA_PNA**FLAG1) * (CLA_CPB**FLAG2) * (CLA_POST**FLAG3)).
• Example models: Chu_2024_allopurinol.R (allopurinol and oxypurinol in neonates with critical congenital heart disease; four separate postoperative fractional changes – e_cpb_post_cl = 2.33, e_cpb_post_vp = 1.42, e_cpb_post_cl_oxy = 0.23, e_cpb_post_vc_oxy = 1.48 – each multiplying the at-birth postnatal-preoperative baseline rather than the CPB-phase value. Note the sign asymmetry that makes a distinct post-CPB indicator necessary rather than an extension of CPB_ON: allopurinol clearance recovers above its at-birth value after bypass while oxypurinol clearance falls further, to 0.05 L/h against 0.12 L/h on bypass and 0.21 L/h at birth, which the authors attribute to post-CPB acute kidney injury – the two analytes move in opposite directions across the same phase boundary and no single on-bypass effect can express that).
• Notes: General scope for the same reason as CPB_ON and CPB_REWARM: the postoperative window is a routine time-varying stratum in any popPK study spanning open-heart surgery. The three phase indicators are designed to compose – an effect confined to the bypass run uses CPB_ON + CPB_REWARM, an effect confined to the postoperative window uses CPB_POST, and a paper that resolves neither sub-phase simply sets CPB_ON = 1 for the whole run with CPB_REWARM = 0, which leaves the summed window unchanged. Distinct from T_CPB (time-fixed total bypass duration in minutes, a per-subject scalar) and from INTRAOP (the active-surgery window, which by construction excludes the postoperative period this indicator names). Distinct also from ICU_ADM_POSTSURG, a time-fixed admission-etiology flag rather than a within-subject phase window. Where a paper’s postoperative effect is mechanistically a renal-function effect, prefer this phase indicator over a substituted renal covariate unless the paper actually fitted a measured renal marker – Chu_2024_allopurinol.R reports no creatinine or eGFR covariate, and encoding the drop in oxypurinol clearance as a renal-function term would attribute to measured data an effect the model estimated from the phase alone.

ICSHUNT_R2L (canonical for right-to-left intracardiac shunt indicator)

• Description: Binary indicator that the patient’s cardiac anatomy produces a right-to-left intracardiac shunt with pulmonary blood flow to systemic blood flow ratio Qp:Qs < 1. 1 = right-to-left shunt present (e.g., single-ventricle physiology after stage 2 palliation – Glenn or hemi-Fontan); 0 = no right-to-left shunt (Qp:Qs >= 1). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no right-to-left shunt; Qp:Qs >= 1).
• Source aliases:
  • intracardiac shunt – used in Su_2016_dexmedetomidine.R (Su 2016 Methods ‘Full Covariate Model’ definition; encoded in the source dataset as a binary indicator and entered in the CL equation as a multiplicative factor when shunt = 1).
• Example models: Su_2016_dexmedetomidine.R (multiplicative effect on CL: CL is multiplied by 1.24 when ICSHUNT_R2L = 1; Su 2016 Methods hypothesise the mechanism is shunt-induced reduction in first-pass lung extraction combined with increased hepatic perfusion for a drug with relatively high hepatic extraction ratio).
• Notes: General scope because intracardiac shunting is a defining anatomical feature in any congenital-heart-disease popPK cohort and is likely to recur in future paediatric / adult cardiac-surgery popPK studies. The Qp:Qs < 1 threshold is the standard clinical definition for right-to-left shunting. Distinct from TX_HEART (heart-transplant indicator), RACHS1 (paediatric cardiac surgery risk category), and INTRAOP (binary intra-operative indicator). Future extractions that report a continuous Qp:Qs ratio could either (a) reuse ICSHUNT_R2L as a binary derived from the continuous ratio (Qp:Qs < 1) or (b) register a sibling continuous canonical QP_QS.

Disease state (cross-population indicators)

DIS_CMV (canonical for transplant-recipient-with-cytomegalovirus-infection disease-state indicator)

• Description: 1 = hematopoietic cell transplant (HCT) or solid organ transplant (SOT) recipient with cytomegalovirus (CMV) infection/disease; 0 = non-CMV subject (healthy volunteer, or a phase I participant such as a renal- or hepatic-impairment cohort member, pooled into the same analysis as the reference group). Time-fixed per subject. Used when a population PK model pools healthy volunteers / phase I participants with transplant recipients who have CMV infection and tests the patient-vs-healthy contrast as a PK covariate.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-CMV subject; the complement group is defined per-model – typically healthy volunteers plus the phase I special-population cohorts).
• Source aliases:
  • HSCMV – NONMEM control-stream column name used in Sun_2023_maribavir.R.
• Example models: Sun_2023_maribavir.R (log-scale additive shift on CL/F: exp(e_dis_cmv_cl * DIS_CMV) with e_dis_cmv_cl = -0.280346, i.e. CL/F is 0.756x lower in transplant recipients with CMV than in the healthy 70-kg reference subject), Sun_2025_maribavir.R (the same log-scale additive shift on CL/F in the updated pooled analysis, e_dis_cmv_cl = log(0.678), i.e. CL/F is 32% lower in transplant recipients with CMV; source column HSCMV, derived as HS == 2 with the control-stream comment ;HV reference).
• Notes: Distinct from the organ-specific transplant indicators TX_LIVER / TX_HEART / TX_LUNG, which record which organ was transplanted rather than the presence of CMV infection, and from DIS_HEALTHY, which is the complementary healthy-cohort indicator. DIS_CMV is preferred over encoding this contrast as DIS_HEALTHY = 1 - DIS_CMV because the source’s structural reference subject is explicitly the non-CMV individual (Sun 2023 Table S2: “The reference population is a 70-kg individual without CMV administered a 800 mg maribavir dose”), so keeping the patient group as the indicator preserves the published meaning of the reference THETA. Also distinct from AUC_GCV, which supplies ganciclovir exposure to CMV viral-load PD models rather than flagging CMV disease state. Also distinct from DIS_CMV_RETINITIS, which is a within-CMV-positive disease-presentation split (retinitis vs no retinitis) and is orthogonal to this entry: both indicators can appear in one dataset, and a DIS_CMV_RETINITIS cohort carries DIS_CMV = 1 throughout. Scope: specific; promote to general if a second paper pools transplant recipients with CMV against a non-CMV reference with the same semantics. Sun_2025_maribavir.R is a second attestation with exactly those semantics, but it is deliberately NOT treated as discharging the promotion condition: it is the same authors’ updated pooled analysis of the same drug programme (it reuses the same HSCMV column and the same healthy-volunteer reference), so it is a second paper rather than a second independent line of evidence that the concept generalises. Promote when a paper from a different drug programme pools a CMV-infected transplant cohort against a non-CMV reference. ### DIS_CMV_RETINITIS (canonical for CMV-retinitis indicator within a CMV-positive cohort)
• Description: Binary indicator that a CMV-positive subject has CMV retinitis, i.e. symptomatic end-organ CMV disease of the retina, as opposed to CMV positivity without retinitis. 1 = CMV retinitis; 0 = CMV-positive without retinitis (in the founding study, asymptomatic CMV urine shedding). Time-fixed per subject. Used when a population PK analysis contrasts two CMV presentations inside a CMV-infected cohort, rather than contrasting CMV-infected subjects against CMV-negative ones.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (CMV-positive without retinitis; the complement group is defined per-model – in Yuen 1995 it is the 17 patients shedding CMV in urine).
• Source aliases:
  • CMV – column name used in Yuen_1995_ganciclovir.R (Yang 2023 Table 3 footnote: “CMV: CMV = 0 for CMV-shedding patients and 0.41 for patients with CMV retinitis”; the source folds the coefficient into the column’s coded value, so store the canonical column as a plain 0 / 1 indicator and carry 0.41 as a named ini parameter).
  • CMV-shedding or CMV retinitis – the covariate label used in Yang 2023 Table 4 for the same column.
• Example models: Yuen_1995_ganciclovir.R (fractional reduction of the renal-elimination term of clearance: cl = exp(lcl) + e_wt_crcl_cl * WT * CRCL/100 * (1 - 0.76 * TX_ANY) * (1 - 0.41 * DIS_CMV_RETINITIS), i.e. a retinitis patient’s weight-and-creatinine-clearance-driven clearance term is 41% lower than an asymptomatic urine-shedding patient’s at the same weight and creatinine clearance; Yang 2023 Table 3 and its footnote).
• Notes: This is a within-CMV-positive contrast and is therefore ORTHOGONAL to DIS_CMV, which encodes transplant-recipient-with-CMV-infection versus a non-CMV reference (healthy volunteer or phase I participant). Both can appear in one dataset, and every subject in a DIS_CMV_RETINITIS cohort – retinitis and non-retinitis alike – would carry DIS_CMV = 1; reusing DIS_CMV for the retinitis split would overload it with incompatible semantics and would change the meaning of the published reference THETA in Sun_2023_maribavir.R. If a future paper contrasts a different symptomatic CMV site (colitis, pneumonitis, oesophagitis) against asymptomatic CMV positivity, register a sibling canonical on the same pattern (e.g. DIS_CMV_COLITIS) rather than widening this entry; if a paper pools several sites into a single “end-organ disease” indicator, that is the point at which a DIS_CMV_ORGAN parent becomes worth registering. Scope: specific, because the reference complement is paper-defined.

DIS_COVID19 (canonical for COVID-19 patient disease-state indicator)

• Description: 1 = participant with coronavirus disease 2019 (COVID-19; laboratory-confirmed SARS-CoV-2 infection) enrolled as a patient in the source analysis; 0 = non-COVID-19 subject (healthy volunteer, or a phase I special-population participant such as a renal- or hepatic-impairment cohort member, pooled into the same analysis as the reference group). Time-fixed per subject. Used when a population PK / PD model pools phase I participants with COVID-19 patients and tests the patient-vs-non-patient contrast as a covariate.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (non-COVID-19 subject; the complement group is defined per-model – healthy volunteers plus phase I special-population cohorts in Chan 2023, a historical CML/GIST oncology cohort in Said 2025).
• Source aliases:
  • PTST (patient status) – NONMEM control-stream column name used in Chan_2023_nirmatrelvir.R (Chan 2023 Data S1 $PK: CLPTST = 1; IF(PTST.EQ.1) CLPTST = 1+THETA(17)).
  • COVID – NONMEM control-stream column name used in Said_2025_imatinib.R (Said 2025 Data S1 $PK: V = EXP(THETA(2) + THETA(8)*COVID + ETA(2))).
• Example models: Chan_2023_nirmatrelvir.R (fractional multiplier on nirmatrelvir CL: 1 + e_dis_covid19_cl * DIS_COVID19 with e_dis_covid19_cl = -0.341, i.e. CL is 34.1% lower in adults with COVID-19 than in the phase I reference participant at the same weight and renal function), Said_2025_imatinib.R (log-additive shift on the apparent central volume of a joint imatinib / N-desmethyl-imatinib model: vc = exp(lvc + e_dis_covid19_vc * DIS_COVID19 + etalvc) with e_dis_covid19_vc = log(1.2), i.e. V1/F1 is ~20% larger in pooled COVID-19 ARDS patients than in the CML/GIST oncology reference cohort, attributed to elevated capillary permeability).
• Notes: Distinct from DIS_HEALTHY, and DIS_COVID19 = 1 - DIS_HEALTHY is NOT a valid re-expression here: the reference group in Chan 2023 contains renal- and hepatic-impairment cohorts who are neither healthy nor COVID-19 patients, so the two indicators are not complements. Same structural pattern as DIS_CMV (transplant-recipient-with-CMV vs pooled phase I reference), which is the closest precedent in the DIS_ family. Also distinct from SARS_VLOAD (baseline SARS-CoV-2 viral load) and SARS_SEROPOS (baseline serostatus), which quantify infection within a COVID-19-only cohort rather than flagging cohort membership. Chan 2023 Discussion cautions that the COVID-19-on-CL effect and the 150-mg-tablet formulation effect on F1 are likely partially confounded because the 150-mg tablet was evaluated only in a single single-dose healthy-participant study. Scope promoted from specific to general on 2026-08-21 alongside the Said 2025 imatinib extraction, which is the second paper to pool COVID-19 patients against a non-COVID-19 reference under the same semantics and so satisfies the promotion condition this entry originally set. Note that the non-COVID-19 reference cohort is paper-defined and differs between the two example models (phase I participants in Chan 2023, CML/GIST oncology outpatients in Said 2025), so the effect size is not transferable between papers – record the actual complement in covariateData[[DIS_COVID19]]$notes. Also distinct from DIS_ARDS: Said 2025 describes its COVID-19 arm as C-ARDS throughout, but codes 1 for both the ARDS-by-Berlin-criteria InventCOVID patients and the supplemental-oxygen-only CounterCOVID patients, so the flag is cohort membership and not an ARDS diagnosis. Use DIS_ARDS only when the paper genuinely contrasts ARDS against non-ARDS.

DIS_CHB (canonical for chronic hepatitis B virus infection disease-state indicator)

• Description: 1 = participant with chronic hepatitis B virus (HBV) infection enrolled as a patient in the source analysis; 0 = participant without chronic HBV infection (typically a healthy volunteer from a pooled phase 1 cohort). Time-fixed per subject. Used when a population PK / PD analysis pools healthy participants with chronic-HBV patients and tests the patient-vs-healthy contrast as a covariate.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (participant without chronic HBV infection; the complement group is defined per-model – in Youssef 2024 it is the 21 healthy volunteers of phase 1 study 213725, 4.2 percent of the pooled analysis population).
• Source aliases:
  • Disease status (chronic HBV infection vs healthy) – the covariate label used in Youssef 2024 Supplementary Table 3.
• Example models: Youssef_2024_bepirovirsen.R (fractional multipliers on two disposition parameters of an apparent three-compartment antisense-oligonucleotide model: tlag = exp(ltlag + etaltlag) * (1 + e_dis_chb_tlag * DIS_CHB) with e_dis_chb_tlag = -0.466, and vp = exp(lvp + etalvp) * (1 + e_dis_chb_vp * DIS_CHB) with e_dis_chb_vp = -0.337; i.e. chronic-HBV patients absorb bepirovirsen with a 46.6 percent shorter lag time and have a 33.7 percent smaller shallow peripheral volume than the healthy reference participant).
• Notes: Encodes which disease a subject has, which is what the DIS_ family is for – contrast HBSAG_BL_LOG10 and HBEAG_POS, which quantify the serologic state within an HBV-infected cohort and are orthogonal to this entry (every subject carrying HBEAG_POS would also carry DIS_CHB = 1). Named DIS_CHB rather than DIS_HBV because the abbreviation must distinguish the chronic disease state from acute HBV infection, which the library already models separately in BoivinChampeaux_2026_hepatitisB_qsp.R; CHB is the standard hepatology abbreviation and matches the brevity of DIS_UC / DIS_CMV / DIS_OUD. Distinct from DIS_HEALTHY, and preferred over re-expressing this contrast as DIS_HEALTHY = 1 - DIS_CHB, for the same reason DIS_CMV is: Youssef 2024’s structural reference subject is explicitly the healthy participant (Table 2 rows are labelled “ALAG1 (h), healthy” and “V3/F (L), healthy”, with the chronic-HBV group carrying the proportional shift), so keeping the patient group as the indicator preserves the published meaning of the reference THETA and lets the model file quote Table 2’s printed values unchanged. Also distinct from HEPIMP* and HEPFUNC_REL, which grade hepatic impairment severity rather than recording an infection diagnosis – Youssef 2024 excluded participants with cirrhosis or liver failure, so its chronic-HBV cohort is not a hepatic-impairment cohort. Scope: specific, because the reference complement is paper-defined; promote to general if a second paper pools chronic-HBV patients against a non-HBV reference with the same semantics.

DIS_MALARIA_ACUTE (canonical for acute-phase (vs convalescent / healthy) Plasmodium malaria episode indicator)

• Description: 1 = the record falls in the acute symptomatic phase of a Plasmodium malaria episode; 0 = convalescent or healthy. Time-varying WITHIN subject, which is what separates this canonical from every other DIS_ sibling: it marks a phase of an episode, not membership of a disease cohort. In the founding design each patient contributes acute records on treatment days 1-2 and convalescent records on day 7 of the same admission, and then a further set of healthy records months later, so one subject carries both values of the indicator at both visits. Record per model where the acute/convalescent boundary falls, because it is set by the treatment schedule and the parasite-clearance time rather than by a laboratory threshold.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (convalescent or healthy; the complement is defined per-model – in Kloprogge 2015 it pools day-7 convalescent records from the malaria admission with every record from the post-partum healthy-volunteer visit).
• Source aliases:
  • “disease status” / “disease effect” – the wording Kloprogge 2015 uses in Table 2 and Results for this binary; the source NONMEM column is a plain DISEASE flag.
• Example models: Kloprogge_2015_artesunate.R (fractional multiplicative effect on the absolute oral bioavailability of artesunate: fdepot *= (1 + e_dis_malaria_acute_fdepot * DIS_MALARIA_ACUTE) with e_dis_malaria_acute_fdepot = 0.866, i.e. F is 86.6% higher during acute malaria than in the convalescent / healthy reference, which the authors attribute to a reduced first-pass effect during acute infection; dOFV = -66.4. It acts on bioavailability and on nothing else – neither artesunate nor dihydroartemisinin disposition is affected – which is the paper’s central finding).
• Notes: Registered in the DIS_<condition> family alongside the acute-state siblings DIS_ARF, DIS_CRITILL, DIS_ARDS and DIS_SEPSIS. The _ACUTE suffix is load-bearing and must not be dropped: a bare DIS_MALARIA would read as “has malaria”, which is false for the day-7 convalescent records that carry 0 here even though those patients are mid-treatment for a confirmed infection. Encoding the phase rather than the diagnosis is what lets a within-subject design dissect the disease effect from a second, subject-level effect (PREG in the founding model); a cohort-membership encoding would confound the two, because every acute record in Kloprogge 2015 also happens to be a pregnant record. Distinct from PARA and LNPC (asexual parasite count per microlitre, raw and log-transformed), which quantify parasite burden continuously within an infected cohort and would be the right columns for a paper that grades severity rather than dichotomising phase; and from BODYTEMP, the continuous vital sign that several Mahidol-Oxford malaria models (Kloprogge_2013_lumefantrine.R, Kloprogge_2014_quinine.R) use as their febrile-state covariate instead. Kloprogge 2015 tested two richer disease descriptors – a mildly/moderately-unwell dichotomy from raised creatinine, blood urea nitrogen or liver function tests (or fever > 37.5 degC with tachycardia > 100 beats/min), and a count of fulfilled severity criteria – and retained neither, so the plain binary is the published form. Scope: specific because the acute/convalescent boundary is design-defined; promote to general if a second paper carries an acute-versus-convalescent malaria phase indicator with the same semantics.

DIS_UC (canonical for ulcerative colitis disease-state indicator)

• Description: 1 = ulcerative colitis patient, 0 = non-UC (e.g., healthy volunteer or non-IBD indication). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-UC subject; the complement group is defined per-model – typically healthy volunteers and/or patients with another indication such as asthma).
• Source aliases:
  • UC – used in Hua_2015_anrukinzumab.R.
• Example models: Hua_2015_anrukinzumab.R (multiplicative fractional increase in CL, +72.8%, on top of weight and albumin effects), Wojciechowski_2023_ritlecitinib_updated.R (two linear fractional effects from Wojciechowski 2023 Table 2: CL/F *= (1 - 0.560 * DIS_UC) and F *= (1 - 0.224 * DIS_UC); the CL/F reference category is the healthy-participant cohort while the F reference category is healthy participants plus the RA, alopecia areata and vitiligo cohorts, and the UC cohort also contributes to the pooled inflammatory-disease group that scales the IIV and proportional-residual-error magnitudes), Maleki_2024_brepocitinib.R (additive shift of -1.16 1/h on the oral absorption rate constant Ka, the only structural disease-state effect in a nine-study brepocitinib popPK model pooling healthy participants with six immuno-inflammatory indications; the reference complement is therefore healthy participants plus the alopecia-areata / hidradenitis-suppurativa / psoriatic-arthritis / plaque-psoriasis / vitiligo cohorts. Additive rather than multiplicative because Maleki 2024 Table 2 classifies Ka as a normally distributed parameter; the paper attributes the ~46% slower absorption to UC pathophysiology – gastrointestinal transit time, GI fluid composition and permeability, and altered abundance of metabolising enzymes and transporters), Ternant_2021_infliximab.R (log-additive shift on the central baseline TNF-alpha level of a double central-peripheral QSS-TMDD model, R0_C = 3.3 * exp(0.57 * DIS_UC) nM; reference is Crohn’s disease. Together with DIS_CD it also rebuilds the paper’s IBD switch, IBD = DIS_CD + DIS_UC, which turns the target-mediated terms on; subjects with both indicators 0 are the ankylosing-spondylitis linear-kinetics reference).
• Notes: Used when a population PK model pools UC patients with a non-UC reference population (e.g., Hua 2015: healthy volunteers + asthma patients + UC patients) and UC disease status is tested as a PK covariate. Distinct from DISEXT_EP / DISEXT_OTHER, which operate within a UC-only cohort (disease extension). Start as scope: specific; promote to general if a second paper pools UC with a non-UC reference.

DIS_UVEITIS (canonical for active-uveitis comorbidity indicator)

• Description: 1 = subject has active uveitis (intraocular inflammation) concurrent with the modelled treatment; 0 = no active uveitis. Used when a population PK / PD analysis carries uveitis as a comorbidity within a cohort defined by some other primary disease – typically juvenile idiopathic arthritis, whose commonest extra-articular manifestation is uveitis (prevalence 10-20%) – rather than as a cohort-membership flag against a non-uveitis reference population. Time-varying in principle (uveitis flares and remits under treatment); record per-model whether the source column is subject-level “ever active during treatment” or a genuinely time-varying flag.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no active uveitis during the modelled treatment; the complement is the rest of the primary-disease cohort, so the reference group is paper-defined).
• Source aliases:
  • UVEITIS – used in NassarSheikhRashid_2024_adalimumab.R (screened as “uveitis during treatment with adalimumab (yes/no)”).
• Example models: NassarSheikhRashid_2024_adalimumab.R (categorical power form on apparent clearance per the paper’s Equation 1, CL/F *= 1.44^DIS_UVEITIS: JIA patients with concomitant active uveitis clear adalimumab 44% faster than JIA patients without, which the authors offer as a pharmacokinetic rationale for the escalated – up to weekly – adalimumab dosing that uveitis patients are observed to need in clinical practice).
• Notes: This is a within-primary-disease comorbidity contrast, structurally the same shape as DIS_CMV_RETINITIS (symptomatic ocular end-organ disease inside a CMV-positive cohort) rather than DIS_UC / DIS_COVID19 (cohort membership against a pooled non-disease reference). Do not reuse it to flag membership of a uveitis-only trial cohort against healthy volunteers – that is a different contrast and would overload the reference category; register a sibling canonical if such a paper appears. Distinct from PGA_PT and other disease-activity scores, which are graded rather than binary, and from the arthritis-activity flag some JIA papers carry alongside it (Nassar-Sheikh Rashid 2024 reports active arthritis in 68% of patients but did not test it as a covariate). Founding paper’s ascertainment caveat is worth carrying forward: in a retrospective TDM cohort, concentrations are measured preferentially when treatment looks clinically ineffective, which can preferentially sample low-concentration uveitis patients and inflate the apparent clearance effect. Scope: specific because the reference complement is the paper’s own primary-disease cohort; promote to general if a second paper tests active uveitis as a comorbidity covariate against the same kind of reference.

DIS_DUOD_ULCER (canonical for duodenal-ulcer disease-state indicator)

• Description: 1 = duodenal-ulcer patient (peptic ulcer disease of the duodenum, endoscopically confirmed), 0 = non-duodenal-ulcer reference (e.g., healthy volunteer, or other non-peptic-ulcer indication pooled in the source analysis). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-duodenal-ulcer subject; the complement group is paper-defined – for Yu 2024 the reference is the pooled healthy-subject cohort across 4 phase I trials).
• Source aliases:
  • Disease status – Yu 2024 paper narrative and Table 3 footnote (1 = duodenal ulcer, 0 = healthy).
• Example models: Yu_2024_ilaprazole.R (exponential effects exp(0.290 * DIS_DUOD_ULCER) on CL and exp(0.356 * DIS_DUOD_ULCER) on Vp of the 2-compartment IV ilaprazole model; reference category is the pooled healthy-subject cohort from CTR20132848 / CTR20140147 / CTR20150686 / CTR20150685, and the phase IIa duodenal-ulcer cohort CTR20132846 required an ulcer diameter <= 15 mm with no combined ulcer bleeding; Yu 2024 Table 3 and Eqs. 7-9).
• Notes: Distinct from ENDO_ULCER (IBD mucosal-ulcer activity indicator scored within an IBD cohort at baseline ileocolonoscopy) because DIS_DUOD_ULCER is a peptic-duodenal-ulcer-patient vs non-patient cohort indicator, following the DIS_UC / DIS_PSORIASIS / DIS_HAE pattern of specific-disease-vs-non-disease pooled-cohort covariates. Distinct from DIS_UC and any future IBD indicators because peptic ulcer disease is a non-IBD upper-GI condition. Also distinct from DIS_HEALTHY: that canonical carries a healthy-vs-pooled-patient contrast with 0 = patient, whereas a paper enrolling a single named disease cohort alongside healthy participants is encoded with the disease-specific indicator so the source coefficients and typical values transfer without a sign flip or re-baselining. The shorter forms DIS_DU (collides with other DU acronyms), DIS_PUD, and DIS_PEPTIC_ULCER (both broader than the duodenal-only cohort Yu 2024 enrolled) were considered and rejected. Scope: specific because the complement reference category is paper-defined; promote to general if a second paper pools duodenal-ulcer patients with a non-peptic-ulcer reference.

DIS_OUD (canonical for opioid use disorder disease-state indicator)

• Description: 1 = participant with opioid use disorder (OUD; DSM-defined opioid dependence, the population enrolled in opioid-agonist-treatment trials of buprenorphine, methadone, or extended-release depot formulations), 0 = non-OUD reference (typically a healthy participant enrolled in a phase 1 cohort of the same pooled analysis). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-OUD participant; the complement group is paper-defined – for Bjornsson 2023 the reference is the pooled healthy-participant cohort from the two phase 1 trials, which is also the most common category at 62 percent and therefore the reference for the source’s fractional-difference categorical covariate model).
• Source aliases:
  • Population – Bjornsson 2023 Table 3 covariate rows (“Population covariate on Vc”, “Population covariate on Fq1w1”) with footnote a “Patients with OUD versus healthy volunteers”; the source NMTRAN column name is not separately reported.
• Example models: Bjornsson_2023_buprenorphine.R (two effects in the final buprenorphine / CAM2038 population PK model: a fractional difference of +2.69 on the central volume, vc * (1 + e_dis_oud_vc * DIS_OUD), giving 64.3 L in healthy participants versus 237 L in participants with OUD; and an additive -0.671 on the logit of the CAM2038 weekly-depot fast-pathway dose fraction Fq1w1, giving 45.5 percent versus 29.9 percent. Bjornsson 2023 Table 3 and Results 3.1.3).
• Notes: Follows the DIS_<disease> family (DIS_UC, DIS_HAE, DIS_HOFH, DIS_DUOD_ULCER,…) rather than the complementary DIS_HEALTHY canonical, per the rule recorded in the DIS_DUOD_ULCER notes: a paper enrolling a single named disease cohort alongside healthy participants is encoded with the disease-specific indicator so the source coefficients and typical values transfer without a sign flip or re-baselining. Bjornsson 2023 is exactly that shape – its reference (most common) category is the healthy cohort and both published coefficients are expressed as OUD-versus-healthy – so DIS_OUD keeps Table 3’s typical values (Vc = 64.3 L, Fq1w1 = 45.5 percent) as the model file’s structural means. Distinct from OPIOID_PATIENT_TYPE (the Mann 2022 opioid-naive versus chronic-opioid-user indicator, which stratifies respiratory-depression PD sensitivity within a non-OUD analgesia setting rather than flagging a substance-use-disorder diagnosis) and from CAR_OPIOID (time-varying fraction of mu-opioid receptors bound by an agonist, a mechanistic input rather than a cohort label). Scope: specific because the complement reference category is paper-defined; promote to general if a second paper pools an OUD cohort with a non-OUD reference.

DIS_SASTHMA (canonical for moderate-to-severe asthma disease-state indicator)

• Description: 1 = moderate-to-severe asthma patient, 0 = not (e.g., healthy volunteer, mild-to-moderate asthma, or other indication). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-moderate-to-severe-asthma subject; the complement group is defined per-model).
• Source aliases:
  • sAsthma – used in Hua_2015_anrukinzumab.R.
• Example models: Hua_2015_anrukinzumab.R (multiplicative fractional change in SC bioavailability, -30.9%).
• Notes: The moderate-to-severe vs. mild-to-moderate asthma cutoff is protocol-defined; Hua 2015 uses FEV1 55-80% and ACQ-5 >= 2 for “moderate to severe” (study 4) versus FEV1 > 70% and ACQ-5 <= 1 for “mild to moderate” (study 1). Scope: specific because the severity threshold is tied to a particular analysis plan; future asthma-severity indicators with different thresholds should register as separate canonicals.

DIS_EOE (canonical for eosinophilic esophagitis disease-state indicator)

• Description: 1 = patient with eosinophilic esophagitis (EoE), 0 = non-EoE reference (in the founding analysis, a healthy adult volunteer). Time-fixed per subject. EoE is a chronic, type-2-inflammation-driven immune-mediated disease of the oesophagus diagnosed by peak intraepithelial eosinophil count on biopsy; it is a separately labelled indication for several type-2 biologics, so cohort membership is a stable clinical fact rather than a study-design construct.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-EoE subject; the complement group is paper-defined – for Nguyen 2026 the reference is the 202 healthy adults contributing dense single-dose IV and SC PK from the six phase I studies of Table S1).
• Source aliases:
  • EOE – the NONMEM $INPUT-derived indicator in the Nguyen 2026 supplementary control stream, entering as CLEOE = THETA(18), VMAXEOE = THETA(20) and VSSEOE = THETA(25); the Table 2 row labels write it out as Patient with EoE ... (REF: healthy volunteer). Same orientation, no value transformation.
• Example models: Nguyen_2026_dupilumab.R (three simultaneous exponential shifts in a dupilumab popPK model whose structural parameters were fixed to a healthy-volunteer-only fit: linear clearance ratio 0.944, maximum nonlinear elimination rate ratio 0.782, and a ratio of 1.26 shared by the central and peripheral volumes – i.e. patients with EoE clear dupilumab slightly more slowly by the linear route, saturate the nonlinear route sooner, and distribute into a 26% larger steady-state volume than healthy volunteers. Because every structural parameter was fixed from the healthy-volunteer model, setting DIS_EOE = 0 recovers that published healthy-volunteer model exactly – founding example).
• Notes: Registered in the DIS_<condition> family and follows the DIS_GERD / DIS_UC / DIS_DUOD_ULCER pattern of a named-disease cohort carried against a pooled non-disease reference. Distinct from DIS_GERD: both are upper-gastrointestinal conditions presenting with oesophageal symptoms and the two are frequently considered in the same differential, but EoE is an eosinophil-mediated inflammatory disease defined histologically while GERD is acid-reflux-mediated, and a subject can carry either independently. Also distinct from DIS_HEALTHY, which carries the complementary healthy-vs-pooled-patient contrast with 0 = patient; a paper enrolling a single named disease cohort alongside healthy participants is encoded with the disease-specific indicator so the source coefficients and typical values transfer without a sign flip or re-baselining (the same reasoning recorded under DIS_GERD and DIS_DUOD_ULCER). Do not conflate the cohort-membership indicator with the disease-severity marker some EoE papers carry alongside it: Nguyen 2026 screened baseline peak esophageal intraepithelial eosinophil count (EEOS, reported in eosinophils per high-power field at 400x) as a separate continuous covariate on CL, Vss and Vmax and did not retain it, and that microscopy field count is in any case not the same quantity as the volumetric blood count the canonical EOS column carries. Scope: specific because the complement reference category is paper-defined; promote to general if a second paper pools EoE patients with a non-EoE reference.

DIS_SBS (canonical for short bowel syndrome patient indicator)

• Description: 1 = patient with short bowel syndrome (SBS) dependent on parenteral support, 0 = non-SBS subject (the complement is paper-defined; in the founding analysis it pools healthy volunteers and subjects with renal or moderate hepatic impairment from phase 1 studies). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 1 (SBS patient) in the founding model, whose typical values describe SBS patients and whose effect is applied to the non-SBS complement as e_nonsbs_<param>^(1 - DIS_SBS); the column itself keeps the family orientation 1 = disease.
• Source aliases:
  • POP1 – Marier 2021 supplement control stream, IF(POP1.EQ.0) CLPOP1 = (1 + THETA(15)), IF(POP1.EQ.1) CLPOP1 = 1; Table S8 row label ‘x 0.668 if non-SBS’. Same orientation (POP1 = 1 is SBS), no value transformation.
• Example models: Marier_2021_teduglutide.R (CL/F x 0.668 in non-SBS subjects, Marier 2021 Table S8; Results: ‘The typical CL/F in healthy subjects was ~33% lower than in patients with SBS’).
• Notes: Registered in the DIS_<disease> family and follows the DIS_EOE / DIS_UC pattern of a named-disease cohort carried against a pooled non-disease reference. Not DIS_HEALTHY: the founding complement includes subjects with renal and hepatic impairment, who are not healthy, so 1 - DIS_HEALTHY would misclassify them. Scope: specific because the complement group is paper-defined; promote to general if a second paper pools SBS patients with a non-SBS reference.

DIS_SSC_ILD (canonical for systemic-sclerosis-associated interstitial lung disease indicator)

• Description: 1 = the subject’s interstitial lung disease arises from systemic sclerosis (scleroderma), 0 = fibrosing ILD of any other aetiology (idiopathic pulmonary fibrosis, other chronic fibrosing ILD with a progressive phenotype, hypersensitivity pneumonitis, surfactant protein deficiency, autoimmune ILD not due to systemic sclerosis, drug- or radiation-induced pneumonitis, post-HSCT fibrosis). Time-fixed per subject. SSc-ILD is a separately labelled nintedanib indication and is recorded in the source cohorts as its own yes/no field alongside, not instead of, the ILD-aetiology category, so a subject can legitimately carry both an ILD-aetiology group and DIS_SSC_ILD = 1.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (non-SSc-ILD). This is the majority group in every cohort seen so far: 82% of the Hartmann 2026 pediatric PK set and 83% of its FVC set.
• Source aliases:
  • SSCSUB4 – the NONMEM $INPUT column in the Hartmann 2026 Data S1 pediatric popPK control stream, selected by IF(SSCSUB4.EQ.0) F1DIAG6 = 1 / IF(SSCSUB4.EQ.1) F1DIAG6 = (1 + THETA(7)). Same orientation, no value transformation. The companion SSCSUB / SSCSUB5 columns in the same $INPUT are alternative sub-groupings that the final model does not use.
  • SSc-ILD – the Hartmann 2026 Table 1 and Table S4 row label, and the Table S5 parameter-row label SSc-ILD on Frel.
• Example models: Hartmann_2026_nintedanib.R (fractional change of -0.139 on nintedanib relative bioavailability: f_sscild <- 1 + e_sscild_fdepot * DIS_SSC_ILD, a pre-specified mechanistic covariate carried from the adult nintedanib meta-model through the NONMEM NWPRI prior – founding example).
• Notes: General scope: systemic-sclerosis-associated ILD is a standard clinical diagnosis with a stable, paper-independent meaning (it is the indication studied in the SENSCIS trial and an approved nintedanib indication), not a study-design construct. It is a disease-aetiology indicator and is therefore distinct from the ILD-severity and lung-function covariates (fev1pp-family endpoints, FVC percent predicted) and from the pooled DIS_SASTHMA / DIS_CF pulmonary disease-state siblings, none of which it overlaps. Registered in the DIS_<condition> family. The two Hartmann 2026 exposure-response models list it in covariatesDataExcluded rather than covariateData: the adult FVC models carried an SSc-ILD study effect on baseline and on slope, but it was not retained in the pediatric fits because SSc-ILD was not sufficiently prevalent in that population, so no pediatric coefficient exists to encode.

DIS_PJIA (canonical for polyarticular juvenile idiopathic arthritis disease-state indicator)

• Description: 1 = polyarticular juvenile idiopathic arthritis (pJIA) patient, 0 = non-pJIA (e.g., adult rheumatoid arthritis or other indication). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-pJIA subject; the complement group is defined per-model – typically adult RA in pooled abatacept analyses).
• Source aliases:
  • JIA – used in Gandhi_2021_abatacept.R and Zhong_2026_abatacept.R.
• Example models: Gandhi_2021_abatacept.R (additive coefficient on logit-F: pJIA patients have markedly higher SC bioavailability than RA reference), Zhong_2026_abatacept.R (additive coefficient +3.08 on logit-F transferred verbatim from a previous internal JIA PPK model that matches Gandhi 2021’s published value).
• Notes: Used when a population PK model pools pJIA patients with a non-pJIA reference population (e.g., Gandhi 2021: pooled adult RA + pediatric pJIA; Zhong 2026: pooled adult RA + pediatric pJIA + adult/pediatric HM) and pJIA disease/age status is tested as a PK covariate (typically on bioavailability rather than CL). Distinct from CHILD and ADOLESCENT, which are pure age-band indicators independent of indication. Scope: specific; promote to general if a third paper pools pJIA with a non-pJIA reference and the reference category remains adult RA.

DIS_CANCER (canonical for advanced-solid-tumor / oncology cohort indicator)

• Description: 1 = patient with an advanced or metastatic solid tumor (the oncology cohort in a pooled multi-indication PK/PD analysis), 0 = non-oncology subject (healthy volunteer or non-oncology disease cohort pooled in the source analysis). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-oncology subject; the complement group is paper-defined – typically the union of healthy volunteers and a non-oncology disease cohort such as cGVHD pooled in the source analysis).
• Source aliases: none; source NONMEM / Monolix control streams typically derive the indicator from a POP or STUDY categorical alongside DIS_HEALTHY.
• Example models: Yang_2024_axatilimab.R (multiplicative effect on baseline NCMC: BL_NCMC x exp(1.22 x DIS_CANCER + 0.618 x DIS_HEALTHY); reference category cGVHD when both indicators are 0), Bonate_2004_apomine.R (log-additive shifts on baseline CL/F and Vc/F for the advanced-solid-tumor cohort vs the healthy adult-male reference: e_cancer_cl = log(10.2 / 40.7) = -1.384 and e_cancer_vc = log(7.11 / 12.3) = -0.548; reference category 0 = healthy adult male volunteer), Silva_2024_apx3330.R (log-additive shift on apparent oral clearance for the advanced-solid-tumor cohort vs the healthy Japanese male volunteer reference: e_cancer_cl = 0.409, Silva 2024 Table 2 ‘Combined data’ row beta SubjectSource,CL, applied as CL/F x exp(0.409 x DIS_CANCER) per the paper’s Eq. 2; reference category 0 = healthy Japanese male volunteer. The source paper names the covariate ‘subject source’ and chose it over serum albumin deliberately – it stands in for the whole bundle of between-cohort differences (disease state, age, ethnicity and albumin), which is exactly the paper-defined-complement semantic that makes this entry specific), Kubota_2018_naldemedine.R (multiplicative power-form effect on apparent central volume: 1.27^DIS_CANCER, i.e. Vc/F 27% above the healthy reference in cancer patients with opioid-induced constipation, from phase 2b study 1108V9222 and phase 3 study 1331V9236; paired with DIS_HEALTHY to carry a THREE-level health-status covariate – see that entry for the full coding scheme. Note that the source cohort is cancer patients with OIC generally rather than the advanced/metastatic solid-tumour population named in this entry Description; that is within the paper-defined-complement semantics).
• Notes: Used together with DIS_HEALTHY to decompose a three-level “participant population” categorical (cGVHD reference, advanced solid tumor, healthy volunteer) into two orthogonal binary indicators (Yang 2024 form), or as a single binary stratifier between an oncology cohort and a healthy-volunteer reference (Bonate 2004 and Silva 2024 form). Scope: specific because the disease-pooling reference category is paper-defined (Yang 2024 reference is patients with cGVHD; Bonate 2004 reference is healthy adult males; Silva 2024 reference is healthy Japanese male volunteers). A “study source” / “data source” categorical that separates an oncology cohort from a healthy-volunteer cohort pooled in the same analysis maps to DIS_CANCER rather than to a new STUDY_* indicator, because the modelled contrast is the disease cohort rather than a specific trial protocol.

DIS_CANCER_PED (canonical for pediatric oncology cohort indicator)

• Description: 1 = pediatric patient receiving cancer-directed therapy (any malignancy, including hematologic cancers such as leukemia and lymphoma as well as solid tumors / blastomas), 0 = pediatric patient admitted for a non-oncology indication (e.g. infection, surgery, transplant). Time-fixed per subject. Distinct from DIS_CANCER, which is restricted to advanced/metastatic solid tumors in adults; DIS_CANCER_PED is the pediatric variant in the DIS_CANCER* family and explicitly covers leukemia-dominant pediatric cohorts.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-oncology pediatric patient; the complement group is paper-defined – in Llanos-Paez 2020 the complement is pediatric patients admitted for various non-oncology indications, with appendicitis and kidney disease / urinary tract infection the most common).
• Source aliases:
  • ONCOLOGY – Llanos-Paez 2020 NONMEM column with the same orientation (1 = oncology, 0 = nononcology); maps directly to DIS_CANCER_PED.
• Example models: LlanosPaez_2020_gentamicin.R (multiplicative cohort shifts on V1 (-0.154) and Q (-0.321) relative to the nononcology baseline; CL has no oncology effect), deAlwis_1998_ondansetron.R (cohort indicator used together with DIS_HEALTHY and AGE thresholds to switch the per-subject proportional-residual-error magnitude across five paper-defined sub-populations; DIS_CANCER_PED = 1 routes to the paediatric-chemotherapy stratum (Table 1 group 4, propSd 0.178), DIS_CANCER_PED = 0 paired with DIS_HEALTHY = 0 routes to the paediatric-general-anaesthesia stratum (Table 1 group 5, propSd 0.145); reference complement is the paediatric general-anaesthesia cohort (study 4) plus all non-paediatric subjects), Damle_2021_dalteparin.R (pediatric venous-thromboembolism cohort with and without cancer; the paper’s CANCERST = 1 flags patients WITHOUT cancer, so the multiplicative CL/F factor 0.885 is applied as e_nocancer_cl^(1 - DIS_CANCER_PED) and the reference patient has cancer).
• Notes: Use DIS_CANCER_PED rather than DIS_CANCER whenever the source paper’s “oncology” cohort includes hematologic malignancies (leukemia / lymphoma) or pediatric blastomas, because DIS_CANCER is canonically restricted to advanced/metastatic solid tumors. Reference-category complement is paper-defined (Llanos-Paez 2020 complement is the pooled pediatric non-oncology admissions cohort; de Alwis 1998 complement is the paediatric general-anaesthesia sub-cohort plus all non-paediatric subjects). Scope: specific because the complement is paper-defined. Covariate-effect parameters drop the DIS_ prefix per the existing DIS_CANCER -> e_cancer_* convention (Yang 2024); use e_cancer_ped_<param>.

DIS_HEALTHY (canonical for healthy-participant cohort indicator)

• Description: 1 = healthy participant (no diagnosis), 0 = patient (any diagnosis represented in the pooled cohort). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (patient subject; the complement group is the union of disease cohorts pooled in the source analysis).
• Source aliases: none known; healthy-participant indicators in source NONMEM control streams typically use ad-hoc names (e.g., HV, HEALTHY, DIS_HV).
• Example models: Liu_2024_saf189s.R (multiplicative power-form effect on apparent central volume: 0.784^DIS_HEALTHY, i.e. V/F 21.6% lower in the 24 healthy Chinese volunteers of study STL31147 than in the ALK-positive ALK-inhibitor-naive patient reference; also gates the TUM_ALK_MUT term so that healthy participants, who are not ALK-positive, do not fall into the ROS1-positive-or-unknown patient cell), Nikanjam_2019_siltuximab.R (multiplicative effects: 0.77 on CL, 0.83 on Vss; reference category is the pooled non-healthy oncology cohort), Okada_2025_rocatinlimab.R (multiplicative shift 1 - 0.532 on Vmax when 1; reference complement is the pooled atopic-dermatitis + ulcerative-colitis + plaque-psoriasis patient cohort), Yang_2024_axatilimab.R (multiplicative effect on baseline NCMC: BL_NCMC x exp(1.22 x DIS_CANCER + 0.618 x DIS_HEALTHY); reference category cGVHD), Goel_2016_Sonidegib.R (multiplicative power-form effect on CL/F: 2.96^DIS_HEALTHY; reference category is the pooled cancer-patient cohort across X2101 / X1101 / A2201), Brown_2017_osimertinib.R (linear factor (1 + 0.44 x DIS_HEALTHY) on apparent osimertinib clearance and (1 + 1.25 x DIS_HEALTHY) on apparent AZ5104 clearance; reference category is the pooled NSCLC cohort across AURA / AURA2), Lu_2015_vismodegib.R (additive-on-log-scale shift on ka via exp(0.671 * DIS_HEALTHY) and on F via exp(0.881 * DIS_HEALTHY) gated on the Phase I formulation indicator; reference category is the pooled cancer-patient cohort across SHH3925g / SHH4610g / SHH4476g), Gupta_2016_lenvatinib.R (multiplicative power-form effect on CL/F: 1.15^DIS_HEALTHY; reference category is the pooled solid-tumor / thyroid-cancer patient cohort across 15 phase 1-3 studies; healthy subjects show +15 percent CL/F vs cancer patients), Bienczak_2025_ligelizumab.R (log-additive exp(-0.087 * DIS_HEALTHY) on apparent ligelizumab CL/F; reference category is the pooled chronic-spontaneous-urticaria patient cohort across C2201 / C2202 / C2302 / C2303), Lu_2015_tacrolimus.R (multiplicative factor 1 / 0.562 = 1.78 on CL/F at DIS_HEALTHY = 1; reference category is the adult Chinese orthotopic liver-transplant recipient cohort; healthy-volunteer CL/F is 1.78x patient CL/F at ALT = 0. Paper Eq. 10 uses the reverse-coded SubPop indicator – the model file re-expresses it as DIS_HEALTHY = 1 - SubPop), Li_2017_CC292.R (binary stratifier on the proportional residual error magnitude: HNP cohort uses propSd = sqrt(0.234) and patient cohort uses propSd = sqrt(0.659); reference category 0 is the pooled relapsed/refractory B-cell-malignancy patient cohort from AVL-292-003; source column HNP), Yoneyama_2017_emicizumab.R (exponential effects exp(-0.232 * DIS_HEALTHY) on CL/F and exp(-0.175 * DIS_HEALTHY) on Vd/F; reference category is the pooled Japanese male adult/adolescent severe-hemophilia-A patient cohort across the Japanese MAD phase I and phase I/II extension; paper Eq. 2 uses the reverse-coded PATIENT indicator and the model file re-expresses it as DIS_HEALTHY = 1 - PATIENT, shifting the structural typicals to the patient state), Kleideiter_2017_cebranopadol.R (multiplicative effect on bioavailability: f_disease *= 0.837 for healthy volunteers relative to the LBP/OA chronic-pain reference; paired with DIS_DPN and DIS_BUNIONECTOMY to form the four-level disease-status stratification), Kleideiter_2018_cebranopadol.R (multiplicative power-form effect on bioavailability F: factor 0.837, i.e. about -16% F for healthy adults vs the pooled nociceptive-pain (LBP and OA) reference cohort; sibling indicators DIS_DPN and DIS_BUN carry the diabetic-polyneuropathy and bunionectomy effects in the same four-level disease-status encoding; Kleideiter 2018 Table 13), Taubert_2018_finafloxacin.R (log-additive effects on the canonical lcl_renal + lcl_nonren decomposition: exp(+0.985 * DIS_HEALTHY) on CL_renal and exp(+0.068 * DIS_HEALTHY) on CL_nonren; reference category is the cUTI patient cohort (Trial III), with healthy effects rederived from paper Table 3 CL_total = 20.9 L/h * patient effect (1 - 0.29) and FER1 = 0.40 / FER2 = 0.21; source column PATIENT in the paper, re-expressed as DIS_HEALTHY = 1 - PATIENT), Goggin_2004_emfilermin.R (log-additive exp(+0.4325 * DIS_HEALTHY) on apparent CL/F, where +0.4325 = -log(0.649); the patient reference category is the IVF-ET premenopausal cohort with recurrent implantation failure (Study 3, n = 39) and the model file shifts lcl to the IVF-ET state lcl = log(57 * 0.649) = log(37.0) so the +0.4325 shift restores the healthy-postmenopausal-women-typical 57 L/h at DIS_HEALTHY = 1; source column TYPE in the paper, re-expressed as DIS_HEALTHY = 1 - TYPE), Klunder_2017_upadacitinib.R (paired healthy/RA structural means lcl_h / lcl_ra and lvc_h / lvc_ra gated by DIS_HEALTHY, with cohort-specific log-normal IIV on CL/F and Vc/F; reference category 0 is the adult RA cohort. The CL/F means encode the paper’s verbatim Table 3 contrast lcl_ra = log(39.7 * 0.76); Vc/F means are identical because Klunder 2017 reports no disease-state effect on typical Vc/F, only on its ISV), Bulitta_2010_ceftazidime.R (log-additive effects on CL and on V1 / V2 / V3 of the 3-compartment ceftazidime IV model: exp(log(1 / 1.17) * DIS_HEALTHY) on CL and exp(log(1 / 1.01) * DIS_HEALTHY) shared across V1, V2, and V3; reference category 0 is the cystic-fibrosis patient cohort, and Bulitta 2010 Table 3 reports FCYFCL = 1.17 and FCYFVSS = 1.01 with the healthy-volunteer cohort as the paper’s structural reference; the model file re-expresses the paper’s CF-vs-HV scale factors onto the canonical DIS_HEALTHY orientation so the typical-value parameters equal the Table 3 CF column), Desai_2016_isavuconazole.R (multiplicative fractional effect on peripheral volume V_p: 1 + e_dis_healthy_vp * DIS_HEALTHY with e_dis_healthy_vp = -0.3765, so healthy V_p baseline is about 38% lower than the patient reference at BMI 24.80 kg/m^2; reference category is the pooled SECURE-trial invasive-aspergillosis / other-filamentous-fungi patient cohort; source column SP (1 = patient, 0 = healthy) re-expressed as DIS_HEALTHY = 1 - SP; assignment of Table 5 theta_4 = 417 L to patients and theta_11 = 260 L to healthy subjects was inferred from the Discussion typical-value report V_p ~390 L (patients) / ~292 L (healthy)), deAlwis_1998_ondansetron.R (cohort indicator used together with DIS_CANCER_PED and AGE thresholds to switch the per-subject proportional-residual-error magnitude across five paper-defined sub-populations: DIS_HEALTHY = 1 routes the subject into one of three volunteer strata selected by AGE (young < 45 y -> propSd 0.125, elderly 45-75 y -> propSd 0.133, aged >= 75 y -> propSd 0.169), DIS_HEALTHY = 0 routes the subject into one of two paediatric strata selected by DIS_CANCER_PED (chemotherapy -> propSd 0.178, anaesthesia -> propSd 0.145); reference complement under DIS_HEALTHY = 0 is the pooled paediatric-patient cohort from studies 3 and 4), Iida_2008_nicorandil.R (multiplicative log-additive effects on apparent CL, V1, Q, V2 of the 2-compartment IV nicorandil model: exp(log(1 / 1.94) * DIS_HEALTHY) on CL, exp(log(1 / 1.39) * DIS_HEALTHY) on V1, exp(log(1 / 0.519) * DIS_HEALTHY) on Q, exp(log(1 / 4.06) * DIS_HEALTHY) on V2; reference category 0 is the acute heart failure (AHF) patient cohort, and Iida 2008 Table 2 reports POP_CL, POP_V1, POP_Q, POP_V2 with multiplicative FCL = 1.94, FV1 = 1.39, FQ = 0.519, FV2 = 4.06 carrying the AHF-vs-healthy contrast with the healthy-volunteer cohort as the paper’s structural reference; the model file re-expresses the paper’s HV-vs-AHF scale factors onto the canonical DIS_HEALTHY orientation so the typical-value parameters equal the AHF cohort), Shoji_2011_pregabalin.R (binary stratifier on the combined proportional + additive residual error magnitudes: healthy cohort uses propSd = 0.220 and addSd = 0.0239 ug/mL; patient cohort uses propSd = 0.285 and addSd = 0.236 ug/mL; reference complement under DIS_HEALTHY = 0 is the pooled post-herpetic neuralgia (PT01-PT04) and diabetic peripheral neuropathy (PT05) patient cohort; Shoji 2011 Table 3 final model rows for residual variability), Farrell_2013_conestatAlfa.R (cohort indicator gating the endogenous-baseline structural mean and IIV between healthy volunteers and patients with hereditary angioedema: DIS_HEALTHY = 1 selects lrbase_hv = log(0.901) with etalrbase_hv (12.7% CV) and DIS_HEALTHY = 0 selects lrbase_hae = log(0.176) with etalrbase_hae (54.4% CV); reference complement is the pooled adolescent / adult HAE-patient cohort from Studies C1 1101-01 / 1202-01 / 1203-01 / 1205-01 / 1304-01), PerezRuixo_2006_tipifarnib.R (ratio-form multiplicative effects on five structural parameters: CL ratio 1.21, V2 ratio 0.55, Q4 ratio 8.83, V4 ratio 2.66, and Ka ratio 2.31 applied as ratio^DIS_HEALTHY; reference category is the pooled adult-cancer cohort across 14 phase 1, 2, and 3 studies; healthy subjects show +21 percent CL and 2.31-fold higher Ka than cancer patients, while V2 is 0.55-fold smaller), Yin_2020_pexidartinib.R (multiplicative effect on CL/F: 1.26x when DIS_HEALTHY = 1 corresponding to a 21 percent lower steady-state AUC0-24 in healthy subjects vs patients per Yin 2020 Results; the residual-error proportional SD also switches by DIS_HEALTHY: 0.297 for patients vs 0.196 for healthy subjects; reference category 0 is the pooled TGCT / other solid-tumour patient cohort from Studies PLX108-01 and ENLIVEN), Melhem_2018_g_csf.R (log-additive shifts on the granulopoiesis parameters KINT and STM2: KINT_HV / KINT_PT = 0.197 / 0.113 = 1.743 log-multiplier and STM2_HV / STM2_PT = 5.21 / 3.89 = 1.339 log-multiplier applied on the healthy-volunteer arm; reference category 0 is the adult cancer-patient-on-chemotherapy cohort in the primary Dataset A reference model; healthy-volunteer subjects show approximately 74 percent higher receptor internalisation rate and 34 percent higher maximum stimulation of bone-marrow transit compared with cancer patients on chemotherapy), Chen_2023_nemonoxacin.R (multiplicative power-form effect on the peripheral volume of distribution Vp: e_dis_healthy_vp^DIS_HEALTHY with e_dis_healthy_vp = 1 / 1.23 = 0.813; reference category 0 is the pooled Chinese community-acquired-pneumonia cohort from the phase II (TG-873870-C-3) and phase III (TG-873870-C-4) nemonoxacin trials, and the healthy stratum is the phase I (TG-873870-C-1) cohort; Chen 2023 Eq. 8 prints V3 = 28 x (BW/70) x 1.23^DisStat against the reverse-coded DisStat indicator (0 = healthy, 1 = CAP patient), so the model file re-expresses it as DIS_HEALTHY = 1 - DisStat and shifts lvp to the patient state log(28 x 1.23) = log(34.4) so DIS_HEALTHY = 1 restores the printed healthy typical 28 L), Galluppi_2021_ulotaront.R (log-additive shift exp(-log(0.809) * DIS_HEALTHY) = exp(+0.2119 * DIS_HEALTHY) on apparent CL/F; reference category 0 is the pooled adult-schizophrenia-patient cohort across seven phase I studies plus the phase II acute + 6-month open-label extension studies, with the structural typical lcl = log(32.5 * 0.809) = log(26.29) shifted from the paper’s HV-typical 32.5 L/h to the patient state so exp(+0.2119) restores 32.5 L/h at DIS_HEALTHY = 1; source column PATIENT in the paper, re-expressed as DIS_HEALTHY = 1 - PATIENT), Mitra_2026_ziftomenib.R (multiplicative effects on parent ziftomenib CL/F (2.59x when DIS_HEALTHY = 1, log-scale THETA(12) = 0.950; reference category is R/R AML patients), on FM (0.348x, logit-scale THETA(17) = -1.62), on KO-739 Vc (0.197x, log-scale THETA(21) = -1.62), and on KO-516 Vc (0.154x, log-scale THETA(22) = -1.87); source paper column ‘HV’), Marathe_2023_belzutifan.R (binary stratifier on the proportional residual error magnitude only, with no effect on any structural or random-effect PK parameter: healthy participants use propSd = 0.26 (RESHV) and patients use propSd = 0.29 (RES PAT) per Marathe 2023 Table 2; reference complement under DIS_HEALTHY = 0 is the pooled advanced-RCC / other-advanced-solid-tumor / VHL-RCC patient cohort from Studies 1 and 4. The source control stream switches on STUDY (PROP = THETA(8) IF (STUDY.EQ.1.OR.STUDY.EQ.4)), and because Studies 1 and 4 are exactly the two patient studies the switch is re-expressed on the canonical DIS_HEALTHY orientation. Note that disease status was separately screened as a structural covariate on CL/F and V2/F and was NOT retained, so this model carries a residual-error-only disease effect), Lee_2023_tripegfilgrastim.R (the paper’s “study population” covariate on the tripegfilgrastim PDMDD parameters VD/F and KD: Lee 2023 Table 2 reports both typical values directly rather than the exponential coefficients, so the model file encodes lvd = log(12.7) / e_hv_vd = log(4.7 / 12.7) and lkd = log(16.2) / e_hv_kd = log(42.2 / 16.2); reference category 0 is the Korean pediatric solid-tumor cohort receiving chemotherapy, matching the orientation of the upstream Melhem_2018_g_csf.R framework this model inherits. In this analysis the indicator is fully confounded with age stratum and chemotherapy exposure – every healthy participant is an adult aged 20-38 and every patient is pediatric aged 6-17 and on chemotherapy – which the paper acknowledges as its principal limitation), Ganesan_2023_tebipenem.R (cohort indicator gating (a) three multiplicative structural effects from Ganesan 2023 Eq. 4-6 – Vc/F * (1 - 0.290 * DIS_HEALTHY), Vp/F * (1 - 0.245 * DIS_HEALTHY), Ka * (1 + 0.368 * DIS_HEALTHY) – and (b) which of two cohort-specific CL/F IIV variances applies (0.0614, 24.8 %CV for the 99 healthy phase 1 subjects vs 0.328, 57.2 %CV for the 647 infected ADAPT-PO patients), hosted on paired lcl_healthy / lcl_patient anchors fixed at 0 because the source reports no infection-status effect on the typical CL/F; reference category 0 is the pooled cUTI / acute-pyelonephritis patient cohort, and the paper’s (1 - Infected) term is re-expressed as DIS_HEALTHY), Zhang_2023_brazikumab.R (linear fractional effect on CL: 1 - 0.362 * DIS_HEALTHY, i.e. healthy participants clear brazikumab 36.2 percent more slowly than patients at the same baseline albumin; reference category 0 is the pooled mild-to-severe / moderate-to-severe Crohn’s disease cohort across the phase Ib NCT01258205 and phase IIa NCT01714726 studies, and Zhang 2023 attributes the faster patient clearance to faecal protein loss through the inflamed gut wall; source column GRP coded 1 = CD patient / 2 = healthy, re-expressed as DIS_HEALTHY = GRP - 1), Zhang_2023_brazikumab_il22.R, Zhang_2023_brazikumab_crp.R (same CL effect carried forward as a fixed() parameter into the two biomarker-driven CDAI indirect-response PK/PD variants; every subject in their phase IIa efficacy cohort has DIS_HEALTHY = 0, so the term is inert there and is retained only to keep the PK layer identical to the parent popPK model), Cleary_2023_risdiplam.R (linear fractional effect on apparent clearance: CL/F x (1 + 0.524 * DIS_HEALTHY), i.e. healthy adults clear risdiplam 1.524-fold faster than patients at the same weight and age; reference category 0 is the pooled spinal-muscular-atrophy type 1 / 2 / 3 cohort across NCT03032172 / NCT02908685 / NCT02913482, and all 61 healthy participants are adults from NCT02633709 and NCT03988907. Cleary 2023 Table 2 prints the effect only as “Healthy subjects on CL/F | Factor | 0.524” and never states the functional form, so the linear encoding is an inferred assumption rather than a sourced one: the literal multiplicative reading would place healthy adults 48% below patients, contradicting ESM Fig. S4 (post-hoc CL/F 5.60 vs 3.52 L/h), the Discussion’s “approximately 30% lower” and the Mech-PPK Table 3 adult intrinsic-clearance ratios of 1.374 (CYP3A) and 1.375 (FMO3)), AitOudhia_2024_sotatercept.R (binary stratifier on the log-scale (lnorm) residual error magnitude only, with no effect on any structural or random-effect PK parameter: healthy participants use expSd = sqrt(0.0570) = 0.239 and patients use expSd = sqrt(0.0357) = 0.189 per Ait-Oudhia 2024 Table 2 rows “Residual variability in HV / in PAH (log units)”; reference complement under DIS_HEALTHY = 0 is the pooled pulmonary-arterial-hypertension cohort from PULSAR, SPECTRA, and STELLAR, and the healthy stratum is the two phase 1 post-menopausal-women studies (SAD and MAD). As in Marathe_2023_belzutifan.R, disease status was separately screened as a structural covariate, entered during forward selection, and was NOT retained during backward elimination, so this model carries a residual-error-only disease effect), Zhao_2025_paracetamol.R (multiplicative factor 1 / 1.58 = 0.633 on the apparent paracetamol volume of distribution V_pcm/F, which propagates to all three metabolite volumes through the fixed 0.18 ratio; reference category 0 is the spinal-muscular-atrophy patient cohort and the structural typical is shifted to lvc = log(63.5 * 1.58) so DIS_HEALTHY = 1 restores the paper’s printed healthy-control typical of 63.5 L/70 kg; source column disease coded 1 = SMA, re-expressed as DIS_HEALTHY = 1 - disease. Same SMA-vs-healthy contrast as Cleary_2023_risdiplam.R, on the opposite structural parameter), Maleki_2024_brepocitinib.R (no effect on any typical structural parameter; the indicator instead gates two variability layers at once – (a) the magnitude of the Box-Cox transformed interindividual variability on CL/F and Vc/F, where the tabulated omegas 0.42 and 0.19 are the healthy-participant values and patients carry multiplicative uplifts of 1 + 0.44 and 1 + 0.29 applied to the realised eta, and (b) which pair of combined residual-error SDs applies, healthy participants using addSd = 0.89 ng/mL with propSd = 0.10 and patients using addSd = 0.45 ng/mL with propSd = 0.05; reference complement under DIS_HEALTHY = 0 is the pooled alopecia-areata / hidradenitis-suppurativa / psoriatic-arthritis / plaque-psoriasis / ulcerative-colitis / vitiligo cohort. Maleki 2024 Table 2 codes the source covariate as “Health status (healthy vs patient)” with the effects reported on the patient state, so both uplifts are applied on (1 - DIS_HEALTHY)), Riccobene_2016_ceftaroline.R (multiplicative power-form effects on the active metabolite of the ceftaroline fosamil prodrug: 3.32^DIS_HEALTHY on ceftaroline CL and 3.67^DIS_HEALTHY on ceftaroline Vc; reference category 0 is the pooled infected-patient cohort of the upstream adult analysis of 10 phase 1 / 1 phase 2 / 4 phase 3 studies. Riccobene 2016 Supplemental Table 1 prints theta16^PAT = 3.32 and theta15^PAT = 3.67 against the reverse-coded PAT indicator (0 = healthy, 1 = patient) and the model file re-expresses them as DIS_HEALTHY = 1 - PAT, leaving lcl_ceftaroline = log(3.76) and lvc_ceftaroline = log(3.18) on the patient state so DIS_HEALTHY = 1 restores the healthy-adult typicals 12.5 L/h and 11.7 L. Unusually, the supplement’s own footnote asserts the opposite orientation (“theta15 and theta16 fixed to one for healthy subjects”); it is contradicted by the paper’s Table 2 noncompartmental results, which the model reproduces in BOTH dose arms only when both multipliers are active for the all-healthy ELF cohort, and which it misses by 3.5x on AUC0-tau and 2.4x on Cmax under the footnote’s orientation. Applying only one of the two multipliers fails as well, so the pair is jointly identified by the back-solve), Roepcke_2023_rezafungin.R (linear fractional effects on the 3-compartment rezafungin IV model: CL x (1 - 0.276 x DIS_HEALTHY) and V1 x (1 - 0.222 x DIS_HEALTHY); reference category 0 is the composite “disease state” cohort that Roepcke 2023 built during covariate analysis by pooling patients with candidemia and/or invasive candidiasis (STRIVE / ReSTORE) with the hepatically impaired subjects of Study CD101.IV.1.15, while that study’s matched normal-hepatic-function controls were grouped with the healthy subjects. Table 2 footnote a defines the indicator on the canonical orientation already (I_healthy, 0 = no / 1 = yes), so no re-expression was needed. The composite grouping replaced separate study on CL and infection status on V1 effects; because low albumin marks both hepatic impairment and infection, this indicator is partly collinear with the ALB effect on V23 in the same model), Yin_2024_soticlestat.R (-22.8% effect on linear elimination clearance carried by PATIENTS, i.e. on the complement (1 - DIS_HEALTHY), because the published typical CL of 4.2 L/h is the healthy-volunteer anchor and the reference complement is the pooled DEE / Dravet-syndrome / Lennox-Gastaut-syndrome / 15q-duplication-syndrome / CDKL5-deficiency-disorder cohort), Gu_2025_rivaroxaban.R (log-additive effect on apparent clearance: exp(log(6.48 / 8.35) * DIS_HEALTHY) = exp(-0.2535 * DIS_HEALTHY); reference category 0 is the Chinese NVAF cohort treated by radiofrequency ablation (Study 2, n = 105), and Gu 2025 reports the two typical clearances directly (6.48 L/h healthy, 8.35 L/h patient) rather than a coefficient, so the model file anchors lcl = log(8.35) on the patient state; the paper calls this covariate the “morbid state”), Huang_2025_dexmedetomidine.R (linear fractional effect on the absorption rate constant of intranasal dexmedetomidine: KA x (1 + 1.05 x DIS_HEALTHY), i.e. healthy-volunteer KA is 2.05-fold the patient KA, equivalently patient KA is 49% of healthy KA; reference category 0 is the phase III NCT04383418 cohort of adults undergoing elective abdominal surgery, and the healthy stratum is the phase I CTR20191868 / CTR20171118 cohort. Huang 2025 applies the Methods Eq. 7 piecewise categorical form Pi = PTV if COV = type1, PTV x (1 + theta) if COV = type2 with Table 4 state on KA = 1.05; the orientation and the additive-(1 + theta) reading are each confirmed twice over by the Abstract’s “approximately 49%” ratio, the Discussion’s “KA in healthy volunteers was approximately 1 h-1”, and the published Figure 4 healthy-to-patient Cmax ratio of 1.48-1.51 at 40/60/80/100 kg (a multiplicative x 1.05 reading would give ~1.02). In this analysis the indicator is fully confounded with study, sampling density (rich phase I vs sparse phase III) and surgical/anaesthetic context, which the authors flag as a likely partial explanation of the effect size), Ozdin_2025_dexamethasone.R (dual role: (a) multiplicative power-form effect on CL, e_dis_healthy_cl^DIS_HEALTHY with e_dis_healthy_cl = 1 / 0.899 = 1.112, i.e. healthy adults clear dexamethasone about 11% faster than patients at the same body weight, and (b) binary stratifier selecting which of two combined proportional + additive residual-error magnitudes applies (propSd 0.181 / addSd 0.0013 ng/mL for the healthy cohort versus propSd 0.295 / addSd 24.3 ng/mL for the patient cohort); reference category 0 is the ataxia-telangiectasia cohort of the phase 3 ATTeST study (NCT02770807) and the healthy stratum is the phase 1 single-dose study (NCT01925859). The source uses two reverse-coded columns that partition the pooled dataset identically – PTNT (0 = healthy, 1 = patient) for the CL effect and PHAS (1 = phase 1, 3 = phase 3) for the $ERROR branch – because every phase 1 subject is a healthy adult and every phase 3 subject is an AT patient; both are re-expressed as DIS_HEALTHY = 1 - PTNT and the structural typical is shifted to the patient state lcl = log(20.8 * 0.899) = log(18.70)), Jian_2025_peginterferon_alfa_2b.R (linear factor (1 + (-0.405) x (1 - DIS_HEALTHY)) on the SC absorption rate constant ka of peginterferon alfa-2b, i.e. chronic hepatitis B patients absorb 40.5% more slowly; note the INVERTED reference relative to most entries here – Jian 2025 uses HEALTHY as the reference category and reports the effect on the CHB arm, so the model evaluates the complement (1 - DIS_HEALTHY); reference complement under DIS_HEALTHY = 0 is the phase 2 CHB cohort NCT01143662 and the healthy stratum is the phase 1 single-ascending-dose trial), Xu_2025_aficamten.R (linear fractional effects on two structural parameters: CL/F * (1 + 0.356 * DIS_HEALTHY) and Vp/F * (1 + 0.266 * DIS_HEALTHY) per Xu 2025 Table 1 and Figure S3; reference category 0 is the pooled symptomatic obstructive-hypertrophic-cardiomyopathy cohort from REDWOOD-HCM (phase 2) and SEQUOIA-HCM (phase 3), and the healthy stratum is the 264 participants across seven phase 1 studies. Because oHCM is the reference, the paper’s typical values describe a male participant with oHCM weighing 80 kg; healthy participants show 23%-24% lower steady-state AUCtau, Cmax, and Ctau (Xu 2025 Table 2). Source column PTYPE, where PTYPE == 1 identifies healthy participants), Riccobene_2017_ceftaroline.R (multiplicative power-form effects on the active metabolite of the ceftaroline fosamil prodrug, applied to the PATIENT state as 3.23^(1 - DIS_HEALTHY) on ceftaroline CL and 4.33^(1 - DIS_HEALTHY) on ceftaroline Vc; reference category 1 is the healthy-adult stratum, so lcl_ceftaroline = log(3.28) and lvc_ceftaroline = log(3.6) are the healthy typicals as printed in Supplemental Table S2 and the multipliers move them to the patient state 10.6 L/h and 15.6 L. Supplemental Equation S1 prints both terms under the heading “For patients” and Supplemental Table S2 writes them as theta16^PAT / theta15^PAT against PAT (0 = healthy, 1 = patient), so no re-expression of the printed typicals was needed. The orientation is confirmed by back-solving Dose/CL against the paper’s own steady-state AUC0-24 simulations: all seven pediatric age bands of Tables 3-5 plus the adult reference arm – every one of which is a patient cohort – land within about 13 percent under this orientation and roughly 4-fold off under the opposite one. Note this is the OPPOSITE orientation to the sibling Riccobene_2016_ceftaroline.R, whose all-healthy ELF cohort required the multipliers to be active; the two analyses share a structural model but were fitted to different pooled datasets, and each orientation is the one pinned by its own paper’s published exposures), Mao_2012_vernakalant.R (an unusually load-bearing use of this indicator, with THREE simultaneous roles: (a) a log-additive effect exp(0.676 * DIS_HEALTHY) on the intercompartmental clearance Q, taking it from 1.34 L/h/kg in patients to 2.63 L/h/kg in healthy volunteers; (b) an INTERACTION gate selecting which CYP2D6 extensive-metabolizer coefficient applies to CL – theta5 = 0.472 for healthy EM subjects versus theta6 = 0.156 for patient EM/UM subjects – so the CYP2D6 effect on clearance is not a main effect at all; and (c) a binary stratifier on the proportional residual error magnitude, healthy volunteers using propSd = sqrt(0.056) = 0.237 and patients propSd = sqrt(0.079) = 0.281. Mao 2012 codes the source covariate as (Subject = Volunteer), which is already the canonical orientation, so no re-expression was needed. Reference complement under DIS_HEALTHY = 0 is the pooled atrial-fibrillation / atrial-flutter cohort from Scene 2 and ACT I-IV; the healthy stratum is just 8 male volunteers from a single phase 1 study, and the paper itself cautions that “this comparison is limited by the small number of volunteers included in the study cohort”. The stratum is additionally confounded with age (volunteer mean 31.2 vs patient mean 63.4 years), sex (all volunteers male) and dosing regimen (a flat 240 mg rather than mg/kg)), Pohl_2022_linzagolix_e2.R (dual role in a PK-oestradiol exposure-response model, with no PK parameter involved at all: (a) it selects which of two estimated typical baseline serum oestradiol values applies – lbl_e2_patient = log(59.1 pg/mL) for endometriosis patients versus lbl_e2_healthy = log(26.6 pg/mL) for healthy premenopausal volunteers, a roughly 2-fold contrast that Pohl 2022 Table 5 reports as two separate structural parameters rather than as a coefficient, so the switch is applied on the log scale to keep the model mu-referenced in etalbl_e2; and (b) it is a binary stratifier on the log-transform-both-sides residual error magnitude, patients using expSd = sqrt(0.610) = 0.781 and healthy volunteers expSd = sqrt(0.241) = 0.491. Reference category 0 is the pooled endometriosis cohort of the EDELWEISS phase 2b study and the Japanese KLH1204 study; the healthy stratum is the phase 1 studies 16-OBE2109-011 and 17-OBE2109-008. The Supporting Information section 3.2 writes the switch as theta_BASE,i = theta_1 if patient and theta_10 if healthy volunteer, which is already the canonical orientation, so no re-expression was needed. Note that the disease contrast here is on an endogenous-biomarker baseline rather than on drug disposition, which is expected: linzagolix PK carries no disease-state effect in the companion Pohl_2022_linzagolix.R popPK model), Gibbs_2017_evolocumab.R (the first entry here in which the indicator acts on a TARGET-TURNOVER set point rather than on a drug disposition parameter or a residual-error magnitude: a multiplicative power-form effect 0.637^DIS_HEALTHY on the baseline PCSK9 concentration BASE_PCSK9 of the quasi-steady-state TMDD model, per Gibbs 2017 Table 2 theta 1, “fold change in baseline PCSK9 for healthy subjects vs statin-treated patients”. Reference category 0 is the phase 1b hypercholesterolemic cohort on stable low- to moderate-intensity statin therapy (Study 20080398) and the healthy stratum is the phase 1a single-ascending-dose cohort (Study 20080397). The orientation is over-determined by the paper’s own Results, which print both typical values – 5.27 nM (379 ng/mL) in statin-treated patients and 5.27 x 0.637 = 3.36 nM (242 ng/mL) in healthy subjects – so no re-expression was needed. The effect is unusually load-bearing for a single covariate because BASE_PCSK9 propagates three ways: it sets the PCSK9 synthesis rate ksyn = kdeg x BASE_PCSK9, it sets the target-mediated component of evolocumab clearance (statin-treated patients therefore show about 20 percent lower unbound evolocumab Cmax and AUClast, Gibbs 2017 Table 1), and through the LDL-C baseline balance it sets the maximum achievable LDL-C reduction BASE_PCSK9/(IC50 + BASE_PCSK9). The indicator is confounded with statin exposure by construction, since statins raise PCSK9 and more than 97 percent of the phase 1b cohort was statin-treated), Bhatnagar_2024_upadacitinib.R (the axSpA analysis is entirely at DIS_HEALTHY = 0 – every subject is a patient and the control streams hard-code it – but the fixed structural CL/F is inherited from an upstream pooled healthy-volunteer + patient model and is expressed on the HEALTHY scale, so the indicator must be supplied as 0 to reproduce any published axSpA exposure), Ly_2023_cabozantinib.R (binary stratifier on the proportional residual error magnitude only, with no effect on any structural or random-effect PK parameter: healthy volunteers use propSd = 0.266 and pooled cancer patients propSd = 0.363 per Ly 2023 Table 2 rows “Residual variability / Healthy subjects” and “/ Patients”, the latter footnoted “Pooled subjects with various cancer types”. Reference complement under DIS_HEALTHY = 0 is the pooled radioiodine-refractory-DTC / RCC / CRPC / HCC cohort across COSMIC-311 and five other phase 3 studies (1682 of 1745 subjects); the healthy stratum is the 63 subjects of the phase 1 single-dose tablet study XL184-020, which was also the only study with serial rather than sparse PK sampling – so the indicator is confounded with sampling density, as in Marathe_2023_belzutifan.R and AitOudhia_2024_sotatercept.R. Ly 2023 arrived at the split structurally rather than by covariate selection: Results 3.1.1 states the residual error “was best described by a RE term for healthy volunteers and a separate RE term for patients with various cancer types”, and the separate patient population indicator WAS screened against the structural parameters and not retained – Results 3.1.2 reports “The CL/F and Vc/F range was overlapping in DTC patients and other cancer types” from post hoc estimates of the final model), Hughes_2022_brepocitinib.R (residual-error stratification only: selects between expSdHv = 0.527 and expSdPat = 0.875, the log-scale residual SDs Hughes 2022 Table 3 tabulates as “Proportional RUV (phase 1;CV)” and “(phase 2;CV)”. The mapping from the paper’s phase 1 / phase 2 study split onto this health-status column is licensed by the Abstract, which restates the same two numbers as applying to “healthy volunteers” and “patients”: all three phase 1 studies enrolled healthy participants and both phase 2 studies enrolled patients. Note that patient status was ALSO screened as a structural covariate on CL/F and on Frel in the same paper’s full covariate model and retained in neither – the Discussion reports “no statistically or clinically significant difference between healthy participants and patient populations with plaque psoriasis or alopecia areata” – so in this model DIS_HEALTHY moves only the residual error, never a typical value. Contrast the companion Maleki_2024_brepocitinib.R entry, where the same column scales the IIV magnitudes as well as the residual error), Choi_2025_denosumab.R (exponential effects on the absorption rate constant ka, the inter-compartmental clearance Q/F and the baseline RANKL concentration R0; reference category is the postmenopausal-osteoporosis patient cohort of the Phase III SB16-3001 study, and the 1 level is the healthy male volunteers of the Phase I SB16-1001 study), Qin_2025_ropeginterferon.R, Qin_2025_ropeginterferon_hct.R, Qin_2025_ropeginterferon_plt.R, Qin_2025_ropeginterferon_wbc.R (an unusual use worth noting: DIS_HEALTHY gates no structural PK parameter at all, but instead switches OFF the chronic decline of the target binding capacity from Rtot0 to Rtot,SS. Qin 2025 Methods 2.4.1 states that ‘In healthy subjects who received only a single dose, as well as in PV patients prior to the start time of target mediation (TSTART), Rtot was simply Rtot0’, so the 48 healthy phase I volunteers keep a constant binding capacity while the 78 polycythaemia vera patients lose theirs with a 27-day half-life. Reference category is the PV patient cohort. In the three pharmacodynamic siblings the column must be set to 0 for every subject, since only patients contributed hematologic observations), Zhou_2025_fruquintinib_poppk.R (fractional-change form on fruquintinib apparent volume of distribution: V/F * (1 - 0.0908 * DIS_HEALTHY), i.e. V/F 9.08% lower in the 42 healthy volunteers of NCT04645940 and NCT04557397 than in the 515 patients with metastatic colorectal cancer or other advanced solid tumors who form the reference; the paper’s own forest plot shows the resulting change in steady-state exposure is <= 2%, i.e. not clinically meaningful), Du_2025_repotrectinib.R (three roles at once: a LINEAR-deviation effect on central volume vc * (1 + (-0.854) * DIS_HEALTHY), so healthy volunteers carry a central volume 14.6% of the patient value; a stratifier selecting between two clearance IIV terms, because healthy volunteers received at most 3 doses and autoinduction of CL is far better characterised in patients; and a stratifier selecting between two combined proportional-plus-additive residual error magnitudes. Source column HV, same orientation, no value transformation. Patients are the reference in all three roles), Schlachter_2026_atogepant.R (selects between the two apparent-clearance typical values printed side by side in Schlachter 2026 Table 2: 22.9 L/h in healthy participants and 17.4 L/h in patients with migraine, encoded as the fractional effect 1 + 0.3161 * DIS_HEALTHY on the patient reference. Patients are the reference because the source’s Section 2.3.1 parameterises the typical value for the level constituting the largest proportion of the population, and patients supply 1005 of the 1356 subjects; the source states the same contrast the other way round, as clearance being 23.7% lower in patients. Episodic and chronic migraine were found not to differ and both take 0), Inoue_2025_valemetostat.R, Tamai_2017_lenvatinib.R (multiplicative power-form effect on CL/F: 1.19^DIS_HEALTHY, i.e. apparent clearance 19% higher in the 232 healthy phase 1 clinical-pharmacology subjects than in the pooled cancer-patient reference; the successor fit to Gupta_2016_lenvatinib.R, which estimated 1.15 on the same contrast before study 202 was added), Desai_2016_isavuconazole_secure.R (two simultaneous roles. (a) A multiplicative fractional effect on peripheral volume V_p: 1 + e_dis_healthy_vp * DIS_HEALTHY with e_dis_healthy_vp = -0.3765, so healthy V_p baseline is about 38% lower than the patient reference at BMI 24.80 kg/m^2. (b) A binary stratifier selecting which of two published clearance IIV terms applies, healthy subjects using omega^2 = 0.09346 (31.30% CV) and patients omega^2 = 0.32921 (62.44% CV), encoded as two mutually-exclusive gated etas so exactly one is active per subject – the same dual-variance pattern as Li_2017_CC292.R and Mao_2012_vernakalant.R, but on an IIV term rather than on the residual error. Reference category is the pooled SECURE-trial invasive-aspergillosis / other-filamentous-fungi patient cohort; source column SP (1 = patient, 0 = healthy), the reverse of the canonical orientation, so the model file re-expresses it as DIS_HEALTHY = 1 - SP. Note this is the phase 1 + phase 3 SECURE pooled analysis (doi:10.1128/AAC.02819-15), a DIFFERENT paper from the same author’s same-year hepatic-impairment analysis carried in Desai_2016_isavuconazole.R (doi:10.1128/AAC.02942-15), which does not use this covariate re-expressed as DIS_HEALTHY = 1 - SP; assignment of Table 5 theta_4 = 417 L to patients and theta_11 = 260 L to healthy subjects was inferred from the Discussion typical-value report V_p ~390 L (patients) / ~292 L (healthy)), Bhagunde_2019_imipenem.R (linear proportional effect on the central volume of imipenem: (1 + -0.29 * DIS_HEALTHY), i.e. a healthy volunteer’s V1 is 0.71 times a patient’s at the same body weight; reference category is the pooled complicated-intra-abdominal-infection / complicated-urinary-tract-infection / hospital-acquired-and-ventilator-associated-pneumonia patient cohort, which is also the more common group at 73.0% of 855 participants. The orientation is settled by the deposited final-model control stream rather than by the table, which prints only a row labelled Healthy: the stream codes IF(HLTH.EQ.0) V1_IPHLTH = 1 ; Most common against IF(HLTH.EQ.1) V1_IPHLTH = ( 1 + THETA(12)), so the tabulated typical V1 of 15.83 L is the PATIENT value. Bhagunde 2019 attributes the direction to intravenous fluid loading in hospitalized and intensive-care patients. Because the covariate enters V1 and not CL it leaves steady-state AUC unchanged, which is why the paper reports no health-status effect on exposure. The companion Bhagunde_2019_relebactam.R model, fitted in the same NONMEM run, carries the same covariate only in covariatesDataExcluded: health status on relebactam CL was the least significant pair in the analysis and was dropped at backward deletion at dMVOF -9.11 against a 10.83 threshold), Kubota_2018_naldemedine.R (carries the healthy level of a THREE-level health-status covariate alongside DIS_CANCER. Kubota 2018 codes health status with two source indicators over three levels – healthy subjects are the cell where BOTH vanish – so this model recovers the source chronic-non-cancer-pain indicator inside model() as (1 - DIS_HEALTHY) * (1 - DIS_CANCER) and applies the published factor 1.20 to it. DIS_HEALTHY therefore appears in no term of its own here: it is load-bearing purely as one of the two columns that separate the three levels, and a subject mis-coded healthy picks up the wrong Vc/F factor), Thakkar_2018_tafenoquine.R (multiplicative power-form effects on BOTH apparent volumes, in OPPOSITE directions: 1.35^DIS_HEALTHY on V2/F and 0.347^DIS_HEALTHY on V3/F, reference category the acute P. vivax malaria patient cohort of DETECTIVE parts 1 and 2. The split sign is the point of the covariate rather than an artefact – the authors read the higher central volume in healthy volunteers as the absence of background chloroquine plus dehydration in acute malaria, and the higher peripheral volume in patients as infection-related vascular leak), Nicolas_2019_alirocumab.R (two-level switch on the LDL-C turnover rate Kout rather than a multiplicative shift: Kout = 0.0948 * DIS_HEALTHY + 0.2393 * (1 - DIS_HEALTHY) per day, i.e. patients with hypercholesterolemia turn LDL-C over 2.52-fold faster than healthy volunteers. The source column DISST runs the OTHER way, 1 = patient, so DIS_HEALTHY = 1 - DISST; only 150/2799 (5.36%) of the pooled cohort were healthy volunteers and none of them received a statin, making disease state and statin use strongly collinear), Kim_2019_voriconazole.R (selects which of the two published additive residual-error magnitudes applies to an observation – 0.208 mg/L for the 93 healthy volunteers of studies 1-4 versus 0.799 mg/L for the 100 TDM patients of study 5, fitted independently per Kim 2019 Methods and Table 2 – rather than acting on a structural parameter; the first registered use of this column on the residual-error model), Alghamdi_2019_cycloserine.R (exponential shift on apparent clearance, exp(-0.660 * (1 - DIS_HEALTHY)), i.e. CL/F 48.3% lower in the 235 pooled tuberculosis / nontuberculous-mycobacteria patients than in the 12 intensively sampled healthy volunteers, so the fitted intercept CL/F = 2.00 L/h is the HEALTHY value and 1.03 L/h the patient value. Two features worth carrying forward. First, the reference orientation is inverted relative to the canonical: the paper’s indicator is the PATIENT level, so the model file rebuilds 1 - DIS_HEALTHY in order to carry the published -0.660 and the published healthy-subject intercept verbatim rather than re-anchoring the intercept on the patient stratum. Second, the authors explicitly decline to interpret the effect as disease biology – the healthy arm was fasted, medication-free, renally normal and intensively sampled over 48 h while the patient arms were none of those (Discussion), so this column is doing duty as a pooled study-conduct contrast and should not be reused as a disease-severity term), Salinger_2019_pretomanid.R (multiplicative power-form effect on apparent oral clearance, e_dis_healthy_cl^DIS_HEALTHY with e_dis_healthy_cl = 1.16, i.e. CL/F 16% higher in the 211 healthy North American subjects than in the tuberculosis patients who make up the rest of the 14-study pooled analysis. The source control stream derives the indicator as HV = 1 - DS - MDR, so it is the complement of the two disease-state indicators rather than an independently recorded column), Kuchimanchi_2022_diroximelFumarate.R (used through its complement: the source column PTST = 1 for a patient with relapsing-remitting multiple sclerosis, so the model computes patient <- 1 - DIS_HEALTHY and applies fractional changes (1 - 0.284 * patient) to monomethyl fumarate clearance and (1 - 0.122 * patient) to 2-hydroxyethyl succinimide clearance; the healthy volunteer is the reference, the opposite orientation to the register’s stated default, and phase I participants with renal impairment take DIS_HEALTHY = 1), Shimizu_2021_lusutrombopag.R (QSP thrombopoiesis model; selects between two population parameter SETS rather than scaling one parameter – DIS_HEALTHY = 1 takes the healthy-subject baselines TPO0 1.4 pM and PLT0 20 x 10^4/uL with splenic sequestration fixed at 1/3, DIS_HEALTHY = 0 takes the chronic-liver-disease-with-thrombocytopenia baselines TPO0 0.78 pM and PLT0 4 x 10^4/uL with 2500 platelets per megakaryocyte; the complement cohort is Japanese CLD patients with platelets below 5 x 10^4/uL), Chan_2020_fenebrutinib.R (log-scale effect exp(log(1.52) * DIS_HEALTHY) on CL/F – 52% higher apparent clearance in the 78 phase 1 healthy volunteers than in the RA-patient reference; also switches the proportional residual error to a time-after-dose-decaying healthy-volunteer form), Zuo_2020_zanamivir.R (entered inverted as the source flag FLU = 1 - DIS_HEALTHY; hospitalized suspected-or-confirmed influenza patients have CL multiplied by 0.756 and the CL random effect scaled by 3.10 relative to the healthy-volunteer structural reference, i.e. the reference category is 1 in this model; the patient group includes influenza-negative/unknown hospitalized subjects, which is why the infection-agnostic healthy indicator is used), Yu_2020_inotersen.R (proportional (1 + theta * DIS_HEALTHY) effects in BOTH the PK and the PD layer of a sequential PK/PD model: +0.111 on CL/F and -0.284 on Vc/F (Yu 2020 Table 2) and +0.269 on baseline serum transthyretin (Table 3); reference category 0 is the hATTR-polyneuropathy patient cohort of the phase 2/3 NEURO-TTR study and its open-label extension, and the healthy stratum is the 51 phase 1 volunteers. Only the PD term’s orientation is stated by the source – the Table 3 footnote calls it the proportional shift in baseline TTR ‘of healthy volunteers compared to hATTR patients’ – and the PK terms are read on the same orientation, which reproduces the paper’s hATTR-patient steady-state AUC (Table 4, 89.9 ug*h/mL vs Dose / CL = 88.2) and its Discussion remark that patient post hoc CL/F (3.18 L/h) is close to the 3.40 L/h typical value), Vaddady_2020_doravirine.R (linear factor (1 - 0.205 x DIS_HEALTHY) on apparent doravirine V/F, i.e. V/F 20.5 percent lower in the 341 phase 1 healthy volunteers than in people living with HIV-1; reference category is the pooled treatment-naive plus virologically suppressed switch HIV-1 cohort. Source flag is 1 = healthy, 0 = HIV-infected per the predecessor Yee 2019 Table 2 footnote), Ou_2021_zanubrutinib.R (separate typical log CL/F per health-status group in the source Eq. 5, 5.13 x Patient + 4.77 x HV, encoded as the patient CL/F of 170 L/h plus a log-ratio shift e_dis_healthy_cl = log(118/170) for the 90 healthy volunteers pooled with 542 patients with B-cell malignancies; reference category is the pooled B-cell-malignancy patient cohort), Edlund_2022_acalabrutinib.R (categorical factors (1 + 0.467)^DIS_HEALTHY on CL/F and (1 - 0.556)^DIS_HEALTHY on Vp/F for the 138 healthy subjects pooled with 575 patients with B-cell malignancies; reference category is the patient cohort), Xiong_2022_tepotinib.R (fractional effects on tepotinib apparent peripheral volume, -0.810, and on MSC2571109A peripheral volume, +2.31; also selects healthy-participant IIV variances on tepotinib CL/F and F and on MSC2571109A CL), Gupta_2022_mobocertinib.R (linear factors (1 + 0.787 x DIS_HEALTHY) on apparent mobocertinib central volume and (1 + 0.900 / 0.738 / 0.909 x DIS_HEALTHY) on the apparent mobocertinib / AP32960 / AP32914 clearances; reference category is the pooled metastatic NSCLC cohort).
• Notes: Used when a population PK model pools healthy participants with patients across heterogeneous indications and the healthy-vs-patient contrast is retained as a covariate. Scope: specific because the complement reference category is paper-defined (Nikanjam 2019 reference is “all non-healthy, non-Castleman, non-SMM tumor types”; Okada 2025 reference is the pooled AD+UC+psoriasis patient cohort; Yang 2024 reference is patients with cGVHD; Goel 2016 reference is the pooled cancer-patient cohort with advanced solid tumors or BCC; Brown 2017 reference is the pooled advanced NSCLC cohort; Lu 2015 vismodegib reference is the pooled cancer-patient cohort with advanced solid tumors / metastatic or locally-advanced BCC; Lu 2015 tacrolimus reference is the adult Chinese liver-transplant recipient cohort). The retired canonical name DIS_HV (healthy-volunteer) was renamed on 2026-05-11 because “volunteer” terminology is discouraged for clinical-trial participants.

DIS_MIGRAINE_EPISODIC (canonical for episodic-migraine patient indicator)

• Description: 1 = patient with episodic migraine (fewer than 15 headache days per month; includes frequent episodic migraine), 0 = otherwise. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0. Used together with the mutually exclusive sibling DIS_MIGRAINE_CHRONIC; a subject with both indicators 0 is in the reference group (healthy participant in Baker 2020).
• Source aliases: DS = EM / “theta if EM” – disease-state category in the FDA Clinical Pharmacology Review of BLA 761119 (eptinezumab), Table 12; used in Baker_2020_eptinezumab.R.
• Example models: Baker_2020_eptinezumab.R (exponentiated factors exp(-0.231) on CL and exp(-0.311) on Vc relative to healthy participants).
• Notes: DIS_HEALTHY cannot carry this contrast because it is a single healthy-vs-any-patient binary; the episodic / chronic split needs two indicators. The DIS_MIGRAINE_<subtype> form was chosen over DIS_EM / DIS_CM, which collide with other disease abbreviations (CMV, CML, cardiomyopathy).

DIS_MIGRAINE_CHRONIC (canonical for chronic-migraine patient indicator)

• Description: 1 = patient with chronic migraine (15 or more headache days per month, at least 8 of them migraine days), 0 = otherwise. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0. Mutually exclusive with DIS_MIGRAINE_EPISODIC; a subject with both indicators 0 is in the reference group (healthy participant in Baker 2020).
• Source aliases: DS = CM / “theta if CM” – disease-state category in the FDA Clinical Pharmacology Review of BLA 761119 (eptinezumab), Table 12; used in Baker_2020_eptinezumab.R.
• Example models: Baker_2020_eptinezumab.R (exponentiated factors exp(-0.272) on CL and exp(-0.422) on Vc relative to healthy participants).
• Notes: See DIS_MIGRAINE_EPISODIC. Migraine-population models that fit episodic and chronic migraine as separate models (e.g. FiedlerKelly_2020_fremanezumab_em.R / _cm.R) do not need these indicators.

DIS_DPN (canonical for diabetic polyneuropathy disease-state indicator)

• Description: 1 = patient with painful diabetic peripheral polyneuropathy (DPN; type 1 or 2 diabetes mellitus with clinically documented painful peripheral neuropathy), 0 = non-DPN subject (e.g., chronic low back pain, osteoarthritis, post-bunionectomy acute pain, or healthy volunteer). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-DPN subject; the complement reference is paper-defined – for the Kleideiter cebranopadol models the typical-value reference is the pooled nociceptive-pain (low back pain + osteoarthritis) cohort, with the sibling canonicals DIS_HEALTHY and DIS_BUNIONECTOMY carrying the healthy-volunteer and bunionectomy effects respectively).
• Source aliases:
  • DIS – Kleideiter 2017 (paper Table 13 categorical disease status decomposed to binary DIS_DPN).
• Example models: Kleideiter_2017_cebranopadol.R (multiplicative effect on bioavailability: f_disease *= 1.132 for DPN patients relative to the LBP/OA nociceptive-pain reference; the value 1.132 reflects the 2018 erratum correction in which Table 13 rows 27-28 for bunionectomy and DPN were swapped), Kleideiter_2018_cebranopadol.R (DPN-vs-LBP/OA bioavailability ratio 1.132 applied as ratio^DIS_DPN; one level of the four-level disease-status stratification {LBP/OA reference, healthy, DPN, bunionectomy}, Kleideiter 2018 Table 13 erratum-corrected), James_2025_fepixnebart.R (additive log-scale effect on the central volume of fepixnebart: VCSTUDY = -0.102 for DPNP relative to a chronic-low-back-pain-only reference, i.e. a fold change of exp(-0.102) = 0.903; the osteoarthritis level of the same three-level term is carried by the sibling DIS_OA. The reference group differs from the Kleideiter models, where osteoarthritis is pooled INTO the reference and therefore needs no indicator – which is precisely why this canonical’s scope is specific and each model must document its own complement).
• Notes: Used when a population PK model pools DPN patients with a non-DPN reference cohort (typically the chronic-pain LBP/OA reference plus other strata) and DPN disease status is retained as a covariate. Pairs with DIS_HEALTHY and DIS_BUNIONECTOMY in the Kleideiter four-level disease-status stratification (LBP/OA reference, healthy, DPN, bunionectomy). Distinct from the existing DIS_DIAB canonical (binary type-1-or-type-2 diabetes-mellitus comorbidity indicator; pre-2026-06-19 names DIAB and T2DM were merged into DIS_DIAB) because DIS_DPN flags the specific painful-polyneuropathy complication of diabetes used as a chronic-pain clinical-trial enrollment category (cebranopadol phase IIa trials 10 and 12; phase II trial 14). Distinct from the DIS_HEALTHY canonical because both indicators may be 0 simultaneously in a pooled chronic-pain analysis where DIS_HEALTHY = 0 means ‘patient’ but DIS_DPN = 0 may still mean ‘non-DPN patient’ (e.g., LBP / OA or bunionectomy patients). Scope: specific because the disease-pooling reference category is paper-defined.

DIS_BUNIONECTOMY (canonical for post-bunionectomy acute-pain cohort indicator)

• Description: 1 = patient with moderate-to-severe acute pain following primary unilateral first-metatarsal bunionectomy surgery, 0 = non-bunionectomy subject (e.g., chronic low back pain, osteoarthritis, painful diabetic polyneuropathy, or healthy volunteer). Time-fixed per subject within the postoperative analgesic-dosing window.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-bunionectomy subject; the complement reference is paper-defined – for the Kleideiter cebranopadol models the typical-value reference is the pooled nociceptive-pain (low back pain + osteoarthritis) cohort, with the sibling canonicals DIS_HEALTHY and DIS_DPN carrying the healthy-volunteer and DPN effects respectively).
• Source aliases:
  • DIS – Kleideiter 2017 (paper Table 13 categorical disease status decomposed to binary DIS_BUNIONECTOMY).
• Example models: Kleideiter_2017_cebranopadol.R (multiplicative effect on bioavailability: f_disease *= 1.801 for bunionectomy patients relative to the LBP/OA nociceptive-pain reference; the value 1.801 reflects the 2018 erratum correction in which Table 13 rows 27-28 for bunionectomy and DPN were swapped; in the corrected Table 14 bunionectomy patients had 80% higher exposure than the reference, the largest disease-status effect in the analysis), Kleideiter_2018_cebranopadol.R (bunionectomy-vs-LBP/OA bioavailability ratio 1.801 applied as ratio^DIS_BUNIONECTOMY; one level of the four-level disease-status stratification {LBP/OA reference, healthy, DPN, bunionectomy}, Kleideiter 2018 Table 13 erratum-corrected).
• Notes: Bunionectomy patients are an acute-pain clinical-trial enrollment category distinct from the chronic-pain LBP / OA / DPN categories in the same analysis cohort. Use this canonical when a pooled analgesic / opioid population PK analysis retains a post-bunionectomy cohort as a covariate group; future bunionectomy-anchored analgesic models should extend the example list and document the complement reference. Pairs with DIS_HEALTHY and DIS_DPN in the Kleideiter four-level disease-status stratification. Distinct from POD (continuous post-operative day) and POSTTX_DAY1 (first-24h-post-transplant indicator) – those describe surgical-recovery time windows rather than the bunionectomy-cohort enrollment label itself. The shorter form DIS_BUN is deliberately avoided to prevent confusion with the common BUN (blood urea nitrogen) laboratory abbreviation. Scope: specific because the disease-pooling reference category is paper-defined.

DIS_OA (canonical for osteoarthritis-pain cohort indicator)

• Description: 1 = patient enrolled in the osteoarthritis (typically knee) chronic-pain cohort, 0 = non-osteoarthritis subject (e.g. chronic low back pain, painful diabetic peripheral neuropathy, post-bunionectomy acute pain, or healthy volunteer). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-osteoarthritis subject; the complement reference is paper-defined – in James_2025_fepixnebart.R the reference indication is chronic low back pain, with the sibling canonical DIS_DPN carrying the diabetic-peripheral-neuropathic-pain level of the same three-level term).
• Source aliases:
  • OA – James 2025 Table 3 footnote d (VCSTUDY = 0, -0.102, 0.00735 for patients with CLBP, DPNP, and OA patients, respectively; the three-level categorical disease/study term decomposed to the binary indicators DIS_DPN and DIS_OA).
  • Disease, Study – James 2025 Table 2 covariate-run labels, where the same variable is called both (each phase 2 study enrolled exactly one indication, so study and indication are the same column).
• Example models: James_2025_fepixnebart.R (additive log-scale effect on the central volume: VCSTUDY = 0.00735 for osteoarthritis relative to the chronic-low-back-pain reference, i.e. a fold change of exp(0.00735) = 1.01. The effect is essentially null and its RSE is 408.2%; it is carried because the paper retained the whole three-level term after backward deletion, not because the OA level is itself informative).
• Notes: Sibling of DIS_DPN and DIS_BUNIONECTOMY in the analgesic / chronic-pain enrollment-category family. The distinction from those two is which cohort the paper made its reference: in the Kleideiter cebranopadol models osteoarthritis is pooled INTO the reference (the “LBP/OA nociceptive-pain” group) and therefore needs no indicator, whereas James 2025 separates all three indications, so OA needs its own column. Do not treat this canonical as interchangeable with the reference-group membership of the Kleideiter models – the same trial population can be split differently by different analyses, which is exactly why the scope is specific and the complement reference must be documented per model. Distinct from FLARE, which is an osteoarthritis-pain TRIAL-DESIGN descriptor (whether the protocol required a washout-induced pain flare at randomization) rather than the enrollment-indication label. Distinct also from any osteoarthritis SEVERITY grade (Kellgren-Lawrence and similar); this column carries cohort membership only.

DIS_CASTLEMAN (canonical for Castleman’s disease indicator)

• Description: 1 = Castleman’s disease (multicentric or unicentric), 0 = not Castleman’s disease. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-Castleman subject; the complement group is the union of healthy volunteers and other indications pooled in the source analysis).
• Source aliases: none known; source NONMEM control streams typically use ad-hoc names (e.g., CD, CASTLEMAN).
• Example models: Nikanjam_2019_siltuximab.R (multiplicative +24% effect on CL; no Vss effect).
• Notes: Castleman’s disease is a lymphoproliferative disorder strongly associated with elevated IL-6 levels; it is the only FDA-approved indication for siltuximab. Scope: specific because the disease-pooling reference category is paper-defined.

DIS_HOFH (canonical for homozygous familial hypercholesterolemia patient indicator)

• Description: 1 = patient with homozygous familial hypercholesterolemia (HoFH), 0 = non-HoFH subject (typically healthy volunteer or another reference cohort pooled in the source analysis). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-HoFH subject; the complement group is paper-defined – for Pu 2021 the reference is the pooled healthy-volunteer cohort).
• Source aliases:
  • DISTYPN – used in Pu_2021_evinacumab.R (Pu 2021 NM-TRAN $INPUT column for HoFH-vs-HV disease type, 1 = HoFH).
• Example models: Pu_2021_evinacumab.R (multiplicative exp(theta * DIS_HOFH) factor on Vmax with theta = -0.289, i.e. HoFH patients show ~25% lower target-mediated Vmax than the HV reference; biologically consistent with the LDLR-pathway disruption in HoFH altering ANGPTL3 catabolic kinetics), Bihorel_2025_evinacumab.R (the refinement of that model; same multiplicative form on Vmax, reported as the proportional multiplier 0.750 and encoded as exp(log(0.749573) * DIS_HOFH). Note the reference group here is the pooled phase 1 participant cohort – adults with elevated triglycerides and/or LDL-C who were otherwise healthy – which the paper states explicitly rather than a healthy-volunteer cohort per se).
• Notes: Used when a population PK model pools HoFH patients with healthy volunteers (or another non-HoFH cohort) and HoFH disease status is retained as a covariate. Distinct from a heterozygous-FH (HeFH) indicator because HoFH patients have markedly higher baseline LDL-C (untreated levels often > 500 mg/dL) and a more pronounced response to LDLR-independent therapies. Scope: specific because the reference category is paper-defined.

DIS_HAE (canonical for hereditary angioedema patient indicator)

• Description: 1 = patient with hereditary angioedema (HAE-C1INH-Type1, HAE-C1INH-Type2, or HAE-nC1INH), 0 = healthy volunteer (or other non-HAE reference cohort pooled in the source analysis). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-HAE subject; the complement group is paper-defined – for Diep 2026 the reference is the pooled healthy-volunteer cohort from NCT03263507 and ISIS 721744-CS9).
• Source aliases: paper narrative “patient with HAE” / “healthy volunteer” subgroup labels driving the Diep 2026 disease-status covariate effects on Vc/F, Q/F, baseline PKK, and IC50.
• Example models: Diep_2026_donidalorsen.R (linear (1 + theta * DIS_HAE) multiplicative effects on apparent central volume Vc/F (theta = +0.426, +42.6%), apparent intercompartmental clearance Q/F (theta = -0.261, -26.1%), baseline plasma prekallikrein BL (theta = -0.132, -13.2%), and donidalorsen IC50 on PKK production (theta = +0.770, +77.0%) for patients with HAE vs healthy volunteers), Garcia_2025_garadacimab.R, Garcia_2025_garadacimab_hae_attack.R (log-domain multiplicative effect on garadacimab clearance, exp(log(1.05) * DIS_HAE), i.e. 1.05-fold higher CL in patients with HAE than in healthy volunteers; source column PAT, which codes 2 = patient with HAE).
• Notes: Used when a population PK/PD model pools HAE patients with healthy volunteers and HAE disease status is retained as a covariate. The three molecular HAE subtypes (HAE-C1INH-Type1, HAE-C1INH-Type2, HAE-nC1INH) are pooled in this indicator following the Diep 2026 analysis; if a future paper resolves subtype-specific covariate effects, separate canonical indicators (e.g., DIS_HAE_C1INH_T1) can be added without conflicting with this pooled indicator. Scope: specific because the complement reference category is paper-defined.

DIS_HAE_ACUTE (canonical for acute hereditary-angioedema attack at dosing indicator)

• Description: 1 = the dose was given during an acute hereditary angioedema (HAE) attack; 0 = no attack in progress at dosing. It marks a phase of an episode, not membership of a disease cohort, which is what separates it from DIS_HAE. Record per model how the source defines the attack window; in the founding model it is an attack with onset within 12 hours before icatibant administration (Wang 2021 Table S2 footnote a). Constant within subject for single-dose designs; a multi-dose or multi-attack design should carry it per dosing occasion.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no attack at dosing; the complement is paper-defined – in Wang 2021 it pools all healthy adults with the pediatric patients with HAE who were dosed outside an attack).
• Source aliases:
  • “HAE attack” / “HAE attack status” / “occurrence of an HAE attack within 12 hours of dosing” – Wang 2021 Table 2, Table S2 and Methods.
• Example models: Wang_2021_icatibant.R (log-domain multiplicative effect on icatibant apparent clearance, exp(log(0.911) * DIS_HAE_ACUTE), i.e. CL/F 9% lower when dosed during an attack; RSE 50.2%).
• Notes: Registered in the DIS_<condition> family on the pattern of DIS_MALARIA_ACUTE; the _ACUTE suffix is load-bearing for the same reason. DIS_HAE_ACUTE = 1 implies DIS_HAE = 1, but the converse does not hold: in Wang 2021, 10 of the 31 pediatric patients with HAE were dosed without an attack and carry 0. A model that pools healthy volunteers with HAE patients and estimates both a disease-state and an attack effect needs both columns. Every adult patient with HAE in Wang 2021 (study JE049-2101) also carries 1, so the paper’s Results wording that the effect applies to “pediatric patients who experienced an acute attack” describes where the data came from rather than restricting the covariate to children.

DIS_PBC (canonical for primary biliary cirrhosis disease-state indicator)

• Description: 1 = patient with primary biliary cirrhosis (PBC), an autoimmune destruction of intrahepatic bile ducts causing chronic cholestasis; 0 = non-PBC subject (healthy reference cohort or other non-PBC reference). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-PBC subject; complement reference is paper-defined – for Zuo 2016 the reference is the pooled healthy-adult cohort from Xiang 2011 / Dilger 2012 / Hess 2004).
• Source aliases: paper narrative “patient with PBC” / “healthy” subgroup labels in Zuo 2016.
• Example models: Ooi_2026_elafibranor.R, Ooi_2026_elafibranor_gft1007.R (fractional difference in the PBC population: apparent peripheral volume 28.4% lower for the parent, apparent inter-compartmental clearance 21.4% higher for the metabolite; Ooi 2026 Table S3), Zuo_2016_UDCA.R (multiplicative scaling on liver-to-biliary rate constants when DIS_PBC = 1: K_LB,0 scaled by 0.10 – 90% reduction; K_LB,1 scaled by 0.30 – 70% reduction; K_LB,2 scaled by 0.10 – 90% reduction; reproduces the Zuo 2016 Figure 3 PBC simulation).
• Notes: Used when a systems / popPK model adapts a healthy-state structural model to a PBC population via fixed disease-state scaling on hepatic-excretion rate constants. Scope: specific because the structural adaptation form (which K parameters are scaled, by how much) is paper-defined; future PBC extractions that re-estimate or alter the scaling pattern can extend the example-models list. Distinct from DIS_HEPATIMP (hepatic-impairment severity categorical), DBIL (direct bilirubin biomarker), and ALP (cholestasis biomarker), which describe pathophysiology rather than the disease label itself.

DIS_CIRRHOSIS (canonical for cirrhosis disease-state indicator)

• Description: 1 = subject has cirrhosis (end-stage hepatic fibrosis, from any aetiology); 0 = subject does not. Time-fixed per subject. Records the cirrhosis diagnosis as a disease state, which is what the DIS_ family is for.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no cirrhosis).
• Source aliases:
  • CIRRHOSIS – Osawa 2018 source-column convention, defined in the Table 2 footnote as “CIRRHOSIS, cirrhosis (yes vs no)”.
• Example models: Osawa_2018_asunaprevir.R (exponential effect -0.428 on asunaprevir CL/F, i.e. apparent clearance 0.65-fold that of a non-cirrhotic subject; compensated cirrhosis only, 8.3% of the cohort).
• Notes: Record the compensation status per model. Trials enrolling hepatically-impaired patients almost always restrict to compensated cirrhosis (Child-Pugh A), and the PK consequences of decompensated disease are far larger – Osawa 2018 cites a dedicated hepatic-impairment study finding 5- to 10-fold and 20- to 30-fold asunaprevir exposure increases in Child-Pugh B and C respectively, against a 1.5-fold effect for the compensated-cirrhosis indicator fitted here. A downstream user who reads this indicator as spanning all cirrhosis will under-predict decompensated exposure by an order of magnitude, so put the enrolled severity in covariateData[[DIS_CIRRHOSIS]]$notes. Distinct from the HEPIMP* family and HEPFUNC_REL, which grade hepatic impairment severity on a classification scheme (NCI ODWG, Child-Pugh) rather than recording a diagnosis; use DIS_CIRRHOSIS when the paper carries a plain yes/no cirrhosis flag. Distinct from DIS_PBC (primary biliary cirrhosis), which names one specific cholestatic aetiology rather than the fibrotic end state, and from NASF (a NASH activity-plus-fibrosis score, in which cirrhosis is the top fibrosis stage). Frequently collinear with transaminase covariates in the same model – Osawa 2018 notes that its cirrhotic subjects also had higher baseline and time-varying AST, so the cirrhosis and AST terms are partly redundant markers of the same loss of metabolic capacity. Specific scope because the enrolled severity band is paper-defined; promote to general once a second paper ratifies a plain cirrhosis yes/no flag with comparable semantics.

DIS_DMD (canonical for Duchenne muscular dystrophy patient indicator)

• Description: 1 = patient with Duchenne muscular dystrophy (DMD), 0 = non-DMD subject (healthy volunteer or other reference cohort). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-DMD subject; the complement group is the reference cohort the source analysis pools alongside the DMD population – typically healthy adult volunteers).
• Source aliases: SPOP (Wojciechowski 2022 study-population indicator with the same orientation: 1 = DMD pediatric patient, 0 = healthy adult volunteer).
• Example models: Wojciechowski_2022_domagrozumab.R (additive 1 + theta shift on baseline myostatin and on the joint kdeg/kint axis; theta_BASE = -0.641, theta_kdegkint = -0.900), Hajjar_2018_DMD_6MWT.R (gates the latent-disease coefficients ALPHA and BETA – both zero for DIS_DMD = 0 – and scales the 6MWT production rate KIN by the multiplier KCOV = 0.63 for DMD subjects; reference cohort is the Henricson 2012 healthy-boy controls).
• Notes: Used when a population PK/PD model pools DMD patients with a non-DMD reference population and DMD disease status is retained as a covariate. Scope: specific because the reference category is paper-defined.

DIS_SMM (canonical for smoldering multiple myeloma indicator)

• Description: 1 = smoldering (asymptomatic) multiple myeloma, 0 = not smoldering MM. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-SMM subject; the complement group is the union of healthy volunteers and other indications pooled in the source analysis).
• Source aliases: none known; source NONMEM control streams typically use ad-hoc names (e.g., SMM, SMOLDMM).
• Example models: Nikanjam_2019_siltuximab.R (multiplicative -23% effect on Vss; no CL effect).
• Notes: Smoldering multiple myeloma is an asymptomatic plasma-cell disorder distinct from active multiple myeloma; pooled with the Nikanjam 2019 cohort that also included MGUS, multiple myeloma, RCC, ovarian, and other tumor types. Scope: specific because the disease-pooling reference category is paper-defined.

DIS_MM (canonical for active multiple myeloma disease indicator)

• Description: 1 = active (non-smoldering) multiple myeloma, 0 = other hematologic malignancy or reference group. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-MM subject; the complement group is defined per-model – in Ogasawara 2020 the reference is the union of MDS, AML, and non-Hodgkin lymphoma cohorts).
• Source aliases:
  • MM – prior canonical name (pre-2026-06-19 DIS_ prefix standardization) and source-paper column in Ogasawara_2020_durvalumab.R. The short MM token also collides with the SI unit millimolar (used in STIM_*_MM in-vitro stimulus concentrations and RIC50 in mM) and with the melanoma abbreviation in some oncology datasets (e.g., TUMTP == "MM" in Aoyama 2012 where MM means malignant melanoma); the DIS_MM form disambiguates.
• Example models: Ogasawara_2020_durvalumab.R (multiplicative factor 0.820 on Vc per Ogasawara 2020 Table 3 footnote c; active (non-smoldering) multiple myeloma cohort from studies MEDI4736-MM-002 and -MM-005), Li_2015_pomalidomide.R (the widest use of this indicator so far: ONE column drives five separate model elements – multiplicative ratios of 0.913 on CL/F, 1.20 on V2/F, 8.46 on V3/F and 3.71 on Q/F, a switch between two absolute absorption lag times, and a switch between two log-scale residual-error magnitudes. Two features are worth carrying forward. First, the reference category here is a HEALTHY normal participant rather than another malignancy, because the pooled dataset is two healthy-volunteer studies plus four relapsed/refractory multiple myeloma studies – so the same canonical spans both the “MM versus other haematologic malignancy” contrast of Ogasawara 2020 and a “MM versus healthy” contrast, and the per-model reference_category field is what distinguishes them. Second, the effect is a DISTRIBUTION effect, not an elimination one: Li 2015 argues the 8-fold larger apparent peripheral volume reflects preferential tumour uptake, and supports it by noting that a dedicated CYP3A4/CYP1A2 inhibition study moved the alpha phase rather than the beta phase).
• Notes: Distinct from DIS_SMM (smoldering multiple myeloma) and MM_NIGG (the immunoglobulin-subtype stratifier within multiple myeloma cohorts).

DIS_EMD (canonical for extramedullary disease indicator)

• Description: 1 = extramedullary disease (plasmacytoma or plasma-cell infiltration outside the bone marrow) present at baseline / screening, 0 = disease confined to the bone marrow. Time-fixed per subject. An aggressive-phenotype and high-tumour-burden marker in multiple myeloma and related plasma-cell malignancies; prevalence is roughly 15-20% at initial diagnosis and higher in relapsed/refractory populations.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (medullary / bone-marrow-confined disease).
• Source aliases:
  • MEDFL – used in Collins_2023_belantamab_mprotein.R; the same column name appears in the related GSK belantamab mafodotin exposure-response analyses.
  • EMD, EMD_BL, EXTRAMED – plausible alternative NONMEM $INPUT forms.
• Example models: Collins_2023_belantamab_mprotein.R (multiplicative factor 0.108 on the effect-compartment rate constant KEO of the serum M-protein tumour-growth-inhibition model when DIS_EMD = 1, per Collins 2023 Table 1; 95% CI 0.0617-0.187).
• Notes: Distinct from DIS_MM (active vs smoldering multiple-myeloma disease status) and DIS_SMM (smoldering multiple myeloma), and complementary to both: a patient with active multiple myeloma may or may not have extramedullary disease. Also distinct from the ulcerative-colitis disease-extent canonicals DISEXT_EP / DISEXT_OTHER, which describe the anatomical extent of inflammatory bowel disease rather than a plasma-cell malignancy outside the marrow. Covariate-effect parameters drop the DIS_ prefix per the DIS_CANCER -> e_cancer_* convention; use e_emd_<param>. Scope: specific because the concept is mechanistically bound to plasma-cell malignancies.

DIS_PNH (canonical for paroxysmal nocturnal hemoglobinuria indicator)

• Description: 1 = paroxysmal nocturnal hemoglobinuria (PNH) patient, 0 = non-PNH subject (healthy volunteer or another indication pooled in the source analysis). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-PNH subject; the complement group is paper-defined – for Lin 2024 it pools healthy adult volunteers and CHAPLE disease patients, while for Lee 2024 it is the healthy phase I subjects only).
• Source aliases: none known; source NONMEM control streams typically use ad-hoc names (e.g., PNH, DPNH); Lee 2024 calls the column “subject group (healthy subjects versus PNH patients)”.
• Example models: Lin_2024_pozelimab.R (additive-fractional +34.07% effect on Vc; no CL or Vp effect; reference category pools healthy volunteers and CHAPLE patients), Lee_2024_eculizumab.R (separately estimated typical values per subject group on Vc, on the terminal-complement-activity baseline E0 and on Imax, encoded as log-scale shifts e_pnh_vc, e_pnh_rbase_tca, e_pnh_imax_tca; reference category is the healthy phase I cohort), Crass_2024_pegcetacoplan.R (fractional linear effect on clearance, cl * (1 + 0.257 * DIS_PNH), i.e. patients with PNH clear pegcetacoplan about 26% faster than the pooled non-PNH reference of healthy and renal-impairment participants; Crass 2024 ESM Table 3 theta 9. The indicator additionally selects which of three log-scale residual-error strata applies, jointly with STUDY_PEGCET_PHASE3), Crass_2024_pegcetacoplan_hemoglobin.R, Crass_2024_pegcetacoplan_ldh.R (the same clearance effect carried forward as a fixed() parameter into the two biomarker PK/PD variants; every subject in their PK/PD analysis set has DIS_PNH = 1, so the term is inert there and is retained only to keep the PK layer identical to the parent popPK model).
• Notes: Paroxysmal nocturnal hemoglobinuria is a rare hematological disease characterized by uncontrolled complement activation on red blood cells; treated with C5-targeted complement inhibitors (eculizumab, ravulizumab, pozelimab). Scope: specific because the disease-pooling reference category is paper-defined.

DIS_PH1 (canonical for primary hyperoxaluria type 1 disease-state indicator)

• Description: 1 = patient with primary hyperoxaluria type 1 (PH1, autosomal-recessive AGXT mutation causing alanine-glyoxylate aminotransferase deficiency and hepatic oxalate overproduction), 0 = subject without PH1 pooled in the source analysis (primary hyperoxaluria type 2 or type 3 patient, or healthy volunteer). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-PH1 subject; the complement group is paper-defined – for Zhang 2025 it pools PH2 patients and healthy volunteers).
• Source aliases:
  • PH – Zhang 2025 Table 2 covariate-formula footnote, defined there as “PH = 1 for PH1 subjects and 0 for PH2 and healthy volunteers”.
  • PH1, PHTYPE, DPH1 – plausible alternative NONMEM $INPUT forms.
• Example models: Zhang_2025_nedosiran.R, Zhang_2025_nedosiran_uoxcr.R (power-form multiplicative factor 1.32^DIS_PH1 on the slow subcutaneous absorption rate constant ka1, per Zhang 2025 Table 2 ka1.PH = 1.32, 95% CI 1.07-1.57; no effect on CL/F, Vc/F, Q/F, Vp/F or Vmax – the paper reports PH subtype had no effect on nedosiran exposure), Zhang_2024_nedosiran.R.
• Notes: Primary hyperoxaluria is a group of three genetically distinct disorders of hepatic glyoxylate metabolism; PH1 accounts for 70-80% of cases and is the most severe. Register a sibling DIS_PH2 / DIS_PH3 only if a future paper distinguishes those subtypes as their own indicators rather than pooling them into the reference category. Covariate-effect parameters drop the DIS_ prefix per the DIS_CANCER -> e_cancer_* convention; use e_ph1_<param>. Scope: specific because the reference category is paper-defined (whether healthy volunteers, PH2 patients, or both are pooled into it varies by analysis). Distinct from DIS_HEALTHY, which is a healthy-volunteer cohort indicator: a PH2 patient is neither DIS_PH1 = 1 nor DIS_HEALTHY = 1.

DIS_MDS_AML (canonical for MDS or AML disease-type indicator)

• Description: 1 = patient with myelodysplastic syndrome (MDS) or acute myeloid leukemia (AML), 0 = other hematologic malignancy or reference group. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-MDS/AML subjects; the complement group is defined per-model – typically multiple myeloma and non-Hodgkin lymphoma in Ogasawara 2020).
• Source aliases:
  • MDSAML – prior canonical name (pre-2026-06-19 DIS_ prefix standardization) and combined indicator used directly in source analyses; MDSAML is the column name in Ogasawara_2020_durvalumab.R.
• Example models: Ogasawara_2020_durvalumab.R (multiplicative factor 1.26 on CL; reference group is the union of MM and NHL subjects), vanIersel_2018_posaconazole.R (multiplicative effect on relative bioavailability F1: fdepot *= (1 + (-0.165) * DIS_MDS_AML) – 16.5% lower F1 in AML/MDS patients relative to the non-AML/MDS reference (healthy volunteers and HSCT recipients pooled); van Iersel 2018 Table 2 final-model ‘AML/MDS on F1’ = -0.165).
• Notes: Use DIS_MDS_AML as a combined MDS+AML indicator when the source paper collapses the two diagnoses into one covariate. If a future paper separates MDS and AML as distinct indicators, register DIS_MDS and DIS_AML separately. Scope: specific because the reference category is paper-defined.

DIS_AML (canonical for acute myeloid leukemia disease-state indicator)

• Description: 1 = patient with acute myeloid leukemia (AML), 0 = non-AML subject (the complement group in a pooled multi-indication PK analysis). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-AML subject; the complement group is paper-defined – for Xu 2023 the reference is patients with advanced solid tumors, alongside the parallel DIS_MDS and DIS_CMML indicators that decompose the hematologic-malignancy cohort).
• Source aliases:
  • DISEASE_abb == "AML" – used in Xu_2023_MBG453.R (the Monolix supplement Appendix S2 encodes disease as the categorical column DISEASE_abb with categories {AML, CMML, MDS, Solid_Tumor} and reference Solid_Tumor; the canonical column carries the binary as.integer(DISEASE_abb == "AML")).
  • AML3 – used in Vaddady_2024_quizartinib.R. NOTE the source column is the complement: the NONMEM stream codes AML3 = 1 for NON-AML subjects (healthy volunteers and subjects with hepatic impairment), so the canonical column carries DIS_AML = 1 - AML3 and every reported effect is applied via (1 - DIS_AML).
• Example models: Xu_2023_MBG453.R (exponential effect on CL: exp(-0.0146 * DIS_AML); not statistically significant in the full covariate model but retained because Xu 2023 used the full-covariate-model approach), Vaddady_2024_quizartinib.R (reference category INVERTED – AML patients are the reference and the non-AML effect is estimated: Frel multiplier 1.73, fractional change -0.188 on quizartinib ka, +0.843 on AC886 CL; the column also selects which stratum-specific IIV variance applies to quizartinib Frel and to AC886 CL), Kang_2020_quizartinib.R (reference category 0 = healthy volunteers, so the column is the paper’s flagPATIENT directly: ka is the level 1.68 1/h in AML patients vs 0.874 1/h, relative F1 multiplier 0.599, and fractional change +0.082 on quizartinib CL; the column also selects the patient-specific AC886 CL IIV variance and the patient-specific residual magnitudes of both analytes).
• Notes: Use DIS_AML (rather than the combined DIS_MDS_AML) when the source paper separates AML from MDS as distinct indicators. Scope: specific because the disease-pooling reference category is paper-defined – and it genuinely varies: Xu 2023 makes non-AML (advanced solid tumors) the reference, whereas Vaddady 2024 makes AML patients the reference. Always read the paper’s stated reference before assigning a sign to any effect on this column, and record the direction in covariateData[["DIS_AML"]]$reference_category.

DIS_MDS (canonical for myelodysplastic syndrome disease-state indicator)

• Description: 1 = patient with myelodysplastic syndrome (MDS), 0 = non-MDS subject (the complement group in a pooled multi-indication PK analysis). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-MDS subject; the complement group is paper-defined – for Xu 2023 the reference is patients with advanced solid tumors, alongside the parallel DIS_AML and DIS_CMML indicators).
• Source aliases:
  • DISEASE_abb == "MDS" – used in Xu_2023_MBG453.R (Monolix supplement Appendix S2 categorical column; reference category Solid_Tumor).
• Example models: Xu_2023_MBG453.R (exponential effect on CL: exp(-0.149 * DIS_MDS); statistically significant, p = 0.021 – patients with MDS have ~14% lower CL than the solid-tumor reference).
• Notes: Use DIS_MDS (rather than the combined DIS_MDS_AML) when the source paper separates MDS from AML as distinct indicators. Scope: specific because the disease-pooling reference category is paper-defined.

DIS_CMML (canonical for chronic myelomonocytic leukemia disease-state indicator)

• Description: 1 = patient with chronic myelomonocytic leukemia (CMML), 0 = non-CMML subject (the complement group in a pooled multi-indication PK analysis). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-CMML subject; the complement group is paper-defined – for Xu 2023 the reference is patients with advanced solid tumors, alongside the parallel DIS_AML and DIS_MDS indicators that decompose the hematologic-malignancy cohort).
• Source aliases:
  • DISEASE_abb == "CMML" – used in Xu_2023_MBG453.R (Monolix supplement Appendix S2 categorical column; reference category Solid_Tumor).
• Example models: Xu_2023_MBG453.R (exponential effect on CL: exp(-0.0411 * DIS_CMML); not statistically significant in the full covariate model but retained because Xu 2023 used the full-covariate-model approach).
• Notes: CMML is a clonal myeloid malignancy with overlapping features of MDS and myeloproliferative neoplasms. Scope: specific because the disease-pooling reference category is paper-defined.

DIS_CML (canonical for chronic myeloid leukemia disease-state indicator)

• Description: 1 = patient with chronic myeloid leukemia (CML), 0 = non-CML subject (the complement group in a pooled analysis; a healthy volunteer in the founding example). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-CML subject; the complement group is paper-defined – in Kang 2025 it is the 23 Korean healthy male volunteers of NCT06461078 pooled with the 24 Chinese CML-CP patients of NCT03722420).
• Source aliases:
  • CML – used in Kang_2025_radotinib.R (Kang 2025 equation 1 and Table 2 footnote a: “disease status: 0 = healthy volunteers, 1 = patients with chronic myeloid leukemia”).
• Example models: Kang_2025_radotinib.R (fractional effect on apparent oral clearance applied via the COMPLEMENT: cl <- exp(lcl + etalcl) * (1 + 0.646 * (1 - DIS_CML)) * circ_cl, so the published typical CL/F of 23.0 L/h is the CML-CP value and a healthy volunteer takes 23.0 * 1.646 = 37.9 L/h, i.e. clearance is 39.2% slower in patients).
• Notes: CML is a myeloproliferative neoplasm driven by the BCR-ABL1 fusion gene arising from the Philadelphia chromosome; it is the indication for the BCR-ABL1 tyrosine kinase inhibitors (imatinib, nilotinib, radotinib, …). Distinct from the neighbouring DIS_CMML (chronic myelomonocytic leukemia, a different disease with mixed MDS/MPN features) – the two abbreviations differ by one letter and must not be conflated. Also distinct from TUMTP_LEUKEMIA, which pools AML/ALL/CLL/CML against other tumour types rather than isolating CML. Like the DIS_AML alias in Vaddady_2024_quizartinib.R, the founding example’s published coefficient multiplies (1 - DIS_CML) rather than DIS_CML, because the paper’s typical value is stated for the patient group; check the direction against the source equation before reusing an effect size. Kang 2025 Discussion warns that its disease effect is confounded with ethnicity and sex (all healthy volunteers Korean male, all patients Chinese of both sexes), so the coefficient is not cleanly a disease effect – record the actual complement cohort in covariateData[["DIS_CML"]]$notes. Scope: specific because the reference complement is paper-defined.

DIS_BCPALL (canonical for B-cell precursor acute lymphoblastic leukemia disease-state indicator)

• Description: 1 = B-cell precursor acute lymphoblastic leukemia (BCP-ALL), 0 = B-cell non-Hodgkin’s lymphoma (NHL) or other non-BCP-ALL indication pooled in the source analysis. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-BCP-ALL; in the Wu 2024 cohort this is the adult B-cell NHL stratum).
• Source aliases:
  • ALL – used in Wu_2024_inotuzumab.R (Wu 2024 calls it the “ALL effect” and notes it bundles disease type with the corresponding bioanalytical assay difference).
  • PTST – used in Garrett_2019_inotuzumab.R, where the source column is coded 1 = NHL / 2 = ALL rather than 0/1 (Garrett 2019 Online Resource 3 $INPUT); the main-text equations of that paper call the same indicator PTSTALL. Recoding to the canonical 0/1 leaves the published coefficients verbatim because NHL is the reference category in both codings.
• Example models: Wu_2024_inotuzumab.R (additive fractional-change effects on CL1 (-0.767) and CL2 (-0.362), and gates the BLSTABL and AGE effects on kdes; for kdes itself a -0.924 fractional change for BCP-ALL), Garrett_2019_inotuzumab.R (the predecessor adult-only analysis: additive fractional-change effects on cl_exp_inf (-0.745) and cl_exp_kdes (-0.860), and gates the BLSTPB effect on cl_exp_kdes).
• Notes: Used when a population PK model pools BCP-ALL patients with a non-BCP-ALL reference (e.g., Wu 2024: pooled adult B-cell NHL + adult BCP-ALL + pediatric BCP-ALL; Garrett 2019: pooled adult B-cell NHL + adult B-cell ALL). Scope: specific because the complement reference category is paper-defined (in both inotuzumab analyses the reference is the adult B-cell NHL stratum). The “ALL effect” theta conflates two physiologically distinct sources of variation – B-cell tumor type (NHL vs ALL surface CD22 burden) and bioanalytical method (ELISA for adult NHL vs HPLC-MS for ALL) – and cannot be split with the available data, because in both papers each assay was used exclusively in one tumour type; document this confounding when comparing across populations.

DIS_ALL (canonical for acute lymphoblastic leukemia disease-state indicator)

• Description: 1 = patient with acute lymphoblastic leukemia (ALL) of any lineage, 0 = any other indication pooled in the source analysis. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-ALL; the complement group is paper-defined – for Schreib 2024 the reference is every other hematopoietic-stem-cell-transplantation indication in the cohort, i.e. AML, chronic granulomatous disease, HLH/XLP, hemoglobinopathies, primary immunodeficiencies, metabolic diseases, neuroblastoma, and others).
• Source aliases:
  • ALL – used in Schreib_2024_busulfan.R (Table 3 covariate theta_k5, effect of ALL on ln(k); the source dataset carries a nine-level diagnosis group factor and the canonical column is as.integer(DiaGroup == "ALL")).
• Example models: Schreib_2024_busulfan.R (exponential effect on the elimination rate constant: exp(-0.210 * DIS_ALL), about a 19% lower k and CL; 13 of 124 patients, 10%).
• Notes: Lineage-agnostic ALL indicator. Distinct from DIS_BCPALL, which is specifically B-cell precursor ALL against a paper-defined reference of B-cell non-Hodgkin lymphoma or other non-BCP-ALL indication; DIS_BCPALL additionally conflates disease type with a bioanalytical-assay difference in its founding model and is not interchangeable with DIS_ALL. Use DIS_ALL when the source pools all ALL lineages against a general non-ALL reference. Scope: specific because the complement reference category is paper-defined.

DIS_TALL (canonical for T-cell acute lymphoblastic leukemia disease-subtype indicator)

• Description: 1 = patient whose disease subtype is T-cell acute lymphoblastic leukemia (T-ALL), 0 = any other ALL / lymphoblastic-lymphoma subtype pooled in the source analysis. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-T-ALL; the complement group is paper-defined – for Lin 2023 the reference pools B-cell ALL 74.1%, T-cell lymphoblastic lymphoma 9.6% and B-cell lymphoblastic lymphoma 0.6%).
• Source aliases:
  • Disease subtype (T-ALL) – used in Lin_2023_asparaginaseErwiniaRecombinant.R (Lin 2023 Table 2 row label; the printed final-model equation multiplies clearance by 0.771 [if T-ALL disease subtype]).
• Example models: Lin_2023_asparaginaseErwiniaRecombinant.R (multiplicative fractional-change effect on serum-asparaginase-activity clearance: 0.771^DIS_TALL, i.e. 22.9% lower CL in T-ALL; 26 of 166 patients, 15.7%).
• Notes: The indicator is T-ALL specifically and NOT T-lineage generally: in the founding model the 9.6% of patients with T-cell lymphoblastic lymphoma sit in the reference group, because Lin 2023 screened primary disease (ALL vs LBL) and disease subtype (B-cell vs T-cell) as separate covariates and retained only the T-ALL cell of that cross-classification. Do not widen this name to “T-lineage” without registering a separate canonical. Distinct from DIS_ALL (lineage-agnostic ALL against a general non-ALL reference) and from DIS_BCPALL (B-cell precursor ALL against a B-cell non-Hodgkin-lymphoma reference); the three are complementary rather than interchangeable, and a source that resolves B- versus T-lineage within an all-ALL cohort wants DIS_TALL. Scope: specific until a second ALL popPK model ratifies the reference-category semantics.

DIS_HLHXLP (canonical for hemophagocytic lymphohistiocytosis / X-linked lymphoproliferative disease indicator)

• Description: 1 = patient with hemophagocytic lymphohistiocytosis (HLH) or X-linked lymphoproliferative disease (XLP), 0 = any other indication pooled in the source analysis. Time-fixed per subject. The two diagnoses are pooled into one indicator because the source analyses that use it treat them as a single disease group.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (neither HLH nor XLP; the complement group is paper-defined – for Schreib 2024 the reference is every other hematopoietic-stem-cell-transplantation indication in the cohort).
• Source aliases:
  • HLH/XLP – used in Schreib_2024_busulfan.R (Table 3 covariate theta_dk2; the source dataset carries a nine-level diagnosis group factor and the canonical column is as.integer(DiaGroup == "HLH/XLP")).
• Example models: Schreib_2024_busulfan.R (additive effect of -0.145 on kel_exp_famp, the fractional amplitude of the decline in the elimination rate constant over a course of therapy, taking it from -0.167 to -0.312, i.e. from a 17% to a 31% fall in k and CL; 14 of 124 patients, 11%).
• Notes: Splitting into separate DIS_HLH and DIS_XLP columns was considered and declined: the founding source pools the two into one 14-patient diagnosis group and estimates a single coefficient, so a split cannot be sourced from it. Register DIS_HLH / DIS_XLP separately only if a future source estimates them apart. Scope: specific because the complement reference category is paper-defined.

DIS_MAS (canonical for macrophage activation syndrome (secondary HLH) disease-state indicator)

• Description: 1 = patient whose haemophagocytic syndrome is macrophage activation syndrome (MAS), i.e. secondary HLH arising in a rheumatic disease – most commonly Still’s disease (systemic juvenile idiopathic arthritis / adult-onset Still’s disease); 0 = primary (genetic) HLH. Time-fixed per subject. Distinguishes the two aetiologies in a pooled HLH population PK analysis.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (primary HLH; the complement group is paper-defined – for Brossard 2025 the reference is the 44 primary-HLH patients of NCT01818492 pooled with the compassionate-use cohort).
• Source aliases:
  • MAS – used in Brossard_2025_emapalumab.R (supplementary Table S2 covariate CLNL_MAS).
• Example models: Brossard_2025_emapalumab.R (coefficient fixed at -1 on the non-linear, IFN-gamma-dependent clearance component, i.e. clnl = ... * (1 + e_mas_clnl * DIS_MAS) sets CLNL to exactly zero in MAS patients; this is the paper’s central PK finding – MAS patients do not reach the total-IFN-gamma levels that drive target-mediated disposition, so emapalumab clearance is linear in MAS).
• Notes: Distinct from DIS_HLHXLP, which pools HLH with X-linked lymphoproliferative disease into a single transplant-indication group against a non-HLH reference: DIS_MAS instead partitions within a haemophagocytic population, separating secondary (MAS) from primary HLH. A model that needs all three strata should carry DIS_MAS alongside a separate primary-HLH indicator rather than overloading this column. Scope: specific because the complement reference category is paper-defined. Pairs naturally with IFNG, since the mechanism the indicator encodes is the absence of IFN-gamma-driven target-mediated clearance.

DIS_BCELLNHL (canonical for B-cell non-Hodgkin lymphoma disease-state indicator)

• Description: 1 = patient with B-cell non-Hodgkin lymphoma (B-cell NHL), 0 = non-B-cell-NHL subject (the complement group in a pooled multi-indication PK analysis of hematologic malignancies). Time-fixed per subject. Used when a population PK model treats B-cell NHL as its own indicator alongside a sibling CLL/SLL reference and other lymphoma / leukemia strata.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-B-cell-NHL; the complement group is paper-defined – for Kemal 2026 the reference is chronic lymphocytic leukemia / small lymphocytic lymphoma (CLL/SLL), alongside the parallel DIS_WM and DIS_OTHER_HEME indicators that decompose the hematologic-malignancy cohort).
• Source aliases:
  • INDIC == 1 – used in Kemal_2026_nemtabrutinib.R (Kemal 2026 NONMEM control stream in the supplement encodes disease as the categorical column INDIC with categories {B-cell NHL, CLL/SLL, WM, Other, MZL, FL, MCL, RT} and reference CLL/SLL; the canonical column carries the binary as.integer(INDIC == 1)).
• Example models: Kemal_2026_nemtabrutinib.R (fractional-change effects on CL/F (-0.166) and Vc/F (0.00224); reference category CLL/SLL), Inoue_2025_valemetostat.R.
• Notes: Distinct from DIS_BCPALL (B-cell precursor acute lymphoblastic leukemia). Use DIS_BCELLNHL when the source paper treats B-cell NHL as one indication category in a pooled analysis of chronic B-cell malignancies (CLL/SLL, WM, DLBCL, follicular, mantle-cell, marginal-zone) and the model estimates a disease-specific effect on PK. Scope: specific because the complement reference category is paper-defined.

DIS_WM (canonical for Waldenstrom’s macroglobulinemia disease-state indicator)

• Description: 1 = patient with Waldenstrom’s macroglobulinemia (WM), 0 = non-WM subject. Time-fixed per subject. WM is a rare indolent B-cell lymphoma / IgM-secreting lymphoplasmacytic disorder that is sometimes pooled with other B-cell malignancies in oncology popPK analyses.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-WM; the complement group is paper-defined – for Kemal 2026 the reference is CLL/SLL, alongside the parallel DIS_BCELLNHL and DIS_OTHER_HEME indicators).
• Source aliases:
  • INDIC == 3 – used in Kemal_2026_nemtabrutinib.R (Kemal 2026 NONMEM control stream in the supplement; reference category CLL/SLL).
• Example models: Kemal_2026_nemtabrutinib.R (fractional-change effects on CL/F (0.0718) and Vc/F (0.152); reference category CLL/SLL).
• Notes: Scope: specific because the complement reference category is paper-defined.

DIS_OTHER_HEME (canonical for pooled ‘other hematologic malignancy’ indicator)

• Description: 1 = patient with a hematologic malignancy other than the explicitly-modeled sibling indications (typically pools some subset of MZL, FL, MCL, Richter’s transformation, or unclassified ‘other’), 0 = one of the explicitly-modeled indications or the reference disease. Time-fixed per subject. The exact composition of the pool is paper-specific and must be documented in per-model covariateData[[DIS_OTHER_HEME]]$notes.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (subject falls into one of the sibling indication indicators or the reference; the reference disease is paper-defined – for Kemal 2026 the reference is CLL/SLL).
• Source aliases:
  • INDIC IN (4, 5, 6, 7, 8) – used in Kemal_2026_nemtabrutinib.R (Kemal 2026 NONMEM control stream pools ‘Other’, marginal zone lymphoma (MZL), follicular lymphoma (FL), mantle cell lymphoma (MCL), and Richter’s transformation (RT) into a single indicator).
• Example models: Kemal_2026_nemtabrutinib.R (fractional-change effects on CL/F (-0.0244) and Vc/F (-0.0200); reference category CLL/SLL).
• Notes: Different-composition pools do not share coefficients; every model that uses DIS_OTHER_HEME must document the exact set of pooled diseases in covariateData[[DIS_OTHER_HEME]]$notes so downstream users know what the coefficient represents. When a paper analyzes MZL, FL, MCL, RT, or a specific ‘other’ indication as its own indicator, register a sibling per-disease canonical (e.g. DIS_MZL, DIS_FL, DIS_MCL, DIS_RT) rather than reusing DIS_OTHER_HEME. Scope: specific because the pool composition is paper-defined.

DIS_PV (canonical for active polycythemia vera disease-state indicator)

• Description: 1 = patient with active polycythemia vera (PV), 0 = the non-PV complement of the myeloproliferative-neoplasm cohort. Time-fixed per subject. PV, essential thrombocythaemia (ET) and primary myelofibrosis (MF) are the three classical BCR-ABL1-negative myeloproliferative neoplasms; PV is distinguished by erythrocytosis with an expanded red-cell mass. The indicator marks the active PV phenotype, not a PV history: a patient whose PV has progressed to post-PV myelofibrosis takes the value 0, because the post-PV MF state is clinically and pharmacokinetically indistinguishable from primary MF.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-PV subject; the complement group is paper-defined – in Ogasawara 2019 it pools primary myelofibrosis, post-PV myelofibrosis, post-ET myelofibrosis and essential thrombocythaemia, i.e. 407 of 452 subjects).
• Source aliases:
  • PV – Ogasawara 2019 Table 2 rows “PV on CL/F” and “PV on V2/F” and the footnote b / c equations, where the factor is written 1.54(if PV) / 1.87(if PV). Same orientation, no value transformation.
• Example models: Ogasawara_2019_fedratinib.R (multiplicative factors applied as e_dis_pv_cl^DIS_PV and e_dis_pv_vc^DIS_PV: 1.54 on apparent clearance and 1.87 on apparent central volume, so a typical PV patient has CL/F 20.0 L/h against 13.0 L/h in the MF/ET reference group. Unlike the DIS_CML and DIS_AML precedents, the published coefficient multiplies the indicator directly rather than its complement, because the paper’s typical value is stated for the MF/ET reference group).
• Notes: Distinct from DIS_ET and a future DIS_MF should a paper need to isolate those two arms separately – Ogasawara 2019 found no difference between MF and ET and therefore pooled them into the reference, so this single indicator is not evidence that the three-way contrast collapses in general. Distinct from TUM_JAK2_V617F_VAF, which measures the size of the driver clone rather than the diagnosis: over 95 percent of PV patients carry JAK2 V617F, but so do roughly half of ET and MF patients, so the VAF column cannot substitute for the diagnosis. Also distinct from the neighbouring DIS_CML and DIS_CMML, which are BCR-ABL1-positive and MDS/MPN-overlap entities respectively. Scope: specific because the reference complement is paper-defined; promote to general if a second paper ratifies the same PV-versus-other-MPN contrast.

DIS_SAD (canonical for secondary antibody deficiency indicator)

• Description: 1 = secondary antibody deficiency (SAD) patient (hypogammaglobulinaemia from external causes such as B-cell-depleting therapy, haematological malignancy, or other immunosuppression), 0 = primary immunodeficiency (PID) patient. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (PID patient; the complement category is the genetic / inborn-error-of-immunity primary immunodeficiency cohort pooled with SAD in the source analysis).
• Source aliases: SID / “secondary immunodeficiency” (Tortorici 2019); none otherwise known; source NONMEM control streams typically use ad-hoc names (e.g., SAD, IMD, DIS).
• Example models: Cheng_2026_immunoglobulin.R (multiplicative theta^DIS_SAD factors on CL (0.542) and on baseline IgG (CBAS, 0.541); reference category PID), Tortorici_2019_immunoglobulin.R (multiplicative ratio 2.841 on Vc, reference category PID; the source term there is “secondary immunodeficiency (SID)”).
• Notes: Used when a population PK model pools PID and SAD pediatric or adult patients receiving immunoglobulin replacement therapy (IgRT) and tests SAD-vs-PID as a covariate. Distinct from the disease-state indicators that pool oncology / autoimmune indications: DIS_SAD specifically partitions hypogammaglobulinaemia by its underlying mechanism (genetic vs. acquired). Scope: specific because the SAD cohort composition is paper-defined (in Cheng 2026, 75% post-rituximab and 25% post-CAR-T cell therapy; in Tortorici 2019 the secondary cohort is predominantly haematological-malignancy-associated). The broader label “secondary immunodeficiency (SID)” maps here rather than to a near-synonym column: both name the same acquired-versus-inborn antibody-deficiency contrast with PID as the reference category.

DIS_AD (canonical for Alzheimer’s disease patient indicator)

• Description: 1 = participant with Alzheimer’s disease (clinical AD diagnosis), 0 = non-AD subject (typically healthy volunteer pooled in the source analysis). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-AD subject; the complement group is paper-defined – for Perez-Ruixo 2025 the reference is the pooled healthy-volunteer cohort).
• Source aliases: none known; source NONMEM control streams typically use ad-hoc names (e.g., AD, STATUS, DISGRP).
• Example models: PerezRuixo_2025_posdinemab.R (acts on baseline free p217+tau in CSF, R0; healthy R0 = 0.793 pmol/L vs AD R0 = 5.995 pmol/L, a 656% relative increase, no PK-parameter effects).
• Notes: Used when a population PK/PD model pools healthy volunteers with Alzheimer’s disease patients and the AD-vs-HV contrast is retained as a covariate on a target-related parameter (e.g., baseline p-tau, baseline p217+tau). Scope: specific because the complement reference category is paper-defined.

DIS_AD_LMCI (canonical for baseline late-mild-cognitive-impairment stage indicator on the Alzheimer’s disease continuum)

• Description: 1 = baseline clinical diagnosis of late mild cognitive impairment (LMCI); 0 = otherwise. Time-fixed per subject (a baseline stratum, not a time-varying disease state). One of four indicators – DIS_AD_LMCI, DIS_AD_MCI, DIS_AD_MILD, DIS_AD_DEMENTIA – that jointly encode the baseline clinical stage along the Alzheimer’s disease continuum as binary contrasts against the earliest stage, early mild cognitive impairment (EMCI), which is the reference level (all four indicators 0). The four-indicator form (rather than an ordinal score) is required because source papers estimate a separate, non-monotonic coefficient per stage.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0; the reference level of the stage covariate is EMCI (early mild cognitive impairment), represented by setting all four DIS_AD_* stage indicators to 0. The four indicators are mutually exclusive.
• Source aliases:
  • bDIAG / DIAG (Monolix categorical baseline-diagnosis covariate) – used in Cao_2026_lecanemab.R; derive DIS_AD_LMCI = as.integer(bDIAG == "LMCI").
  • DX_BL (the ADNI baseline-diagnosis column, with levels CN / SMC / EMCI / LMCI / AD).
• Example models: Cao_2026_lecanemab.R (multiplicative log-scale effect on the ADAS-Cog scaling factor alpha_adas: exp(0.13 * DIS_AD_LMCI)).
• Notes: Distinct from DIS_AD, which is a binary Alzheimer’s-patient-vs-healthy-volunteer cohort flag; this family stratifies within an all-AD cohort by clinical stage. Pooled AD datasets routinely carry more than one diagnostic vocabulary – Cao 2026 merges ADNI (EMCI / LMCI / mild AD dementia) with the lecanemab Studies 201 and 301 (MCI / mild dementia due to AD) – so DIS_AD_MCI and DIS_AD_LMCI coexist, as do DIS_AD_MILD and DIS_AD_DEMENTIA, and the mapping from source category to indicator must be documented per-model in covariateData[[...]]$notes. Founding example: Cao_2026_lecanemab.R.

DIS_AD_MCI (canonical for baseline mild-cognitive-impairment stage indicator on the Alzheimer’s disease continuum)

• Description: 1 = baseline clinical diagnosis of mild cognitive impairment (MCI) recorded without an early / late qualifier; 0 = otherwise. Time-fixed per subject. See DIS_AD_LMCI for the family encoding.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0; family reference level EMCI.
• Source aliases:
  • bDIAG / DIAG (Monolix categorical baseline-diagnosis covariate): derive DIS_AD_MCI = as.integer(bDIAG == "MCI").
• Example models: Cao_2026_lecanemab.R (log-scale effects exp(0.12 * DIS_AD_MCI) on the A-beta monomer aggregation rate kAggAO and exp(0.14 * DIS_AD_MCI) on the ADAS-Cog scaling factor alpha_adas; 61.9-64.1% of the lecanemab Study 201 / 301 cohorts).
• Notes: See DIS_AD_LMCI.

DIS_AD_MILD (canonical for baseline mild-dementia-due-to-AD stage indicator)

• Description: 1 = baseline clinical diagnosis of mild dementia due to Alzheimer’s disease; 0 = otherwise. Time-fixed per subject. See DIS_AD_LMCI for the family encoding.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0; family reference level EMCI.
• Source aliases:
  • bDIAG / DIAG (Monolix categorical baseline-diagnosis covariate): derive DIS_AD_MILD = as.integer(bDIAG == "Mild_AD").
• Example models: Cao_2026_lecanemab.R (log-scale effects exp(0.032 * DIS_AD_MILD) on kAggAO, exp(0.18 * DIS_AD_MILD) on alpha_cdr and exp(0.16 * DIS_AD_MILD) on alpha_adas; 35.9-38.1% of the lecanemab Study 201 / 301 cohorts).
• Notes: See DIS_AD_LMCI. Kept distinct from DIS_AD_DEMENTIA because pooled analyses that merge trial and observational cohorts estimate separate coefficients for the two vocabularies’ dementia categories (Cao 2026 Supplementary Table S14: 0.18 for Mild_AD vs 0.39 for AD on alpha_cdr).

DIS_AD_DEMENTIA (canonical for baseline Alzheimer’s-dementia stage indicator)

• Description: 1 = baseline clinical diagnosis of Alzheimer’s dementia recorded under a vocabulary that does not carry a separate “mild dementia” level (in Cao 2026, the ADNI category “Mild AD dementia”, printed as the level AD in Supplementary Table S14); 0 = otherwise. Time-fixed per subject. See DIS_AD_LMCI for the family encoding.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0; family reference level EMCI.
• Source aliases:
  • bDIAG (Monolix categorical baseline-diagnosis covariate): derive DIS_AD_DEMENTIA = as.integer(bDIAG == "AD").
  • DX_BL (ADNI baseline-diagnosis column): derive from the AD level.
• Example models: Cao_2026_lecanemab.R (log-scale effects exp(0.39 * DIS_AD_DEMENTIA) on both alpha_cdr and alpha_adas; 27.6% of the ADNI cohort).
• Notes: See DIS_AD_LMCI and DIS_AD_MILD.

DIS_COPD (canonical for chronic obstructive pulmonary disease patient indicator)

• Description: 1 = patient with chronic obstructive pulmonary disease (clinical COPD diagnosis, typically moderate-to-severe per GOLD criteria), 0 = non-COPD subject (typically healthy volunteer pooled in the source analysis). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-COPD subject; the complement group is paper-defined – for Lahu 2010 the reference is the pooled phase I healthy-volunteer cohort).
• Source aliases:
  • COPD – used in Lahu_2010_roflumilast.R (paper text covariate symbol in equation 6 and 7).
• Example models: Lahu_2010_roflumilast.R (linear additive effects on roflumilast parent CL (-39.4%) and V1 (+184%) and on roflumilast N-oxide CL (-7.9%) and Vd (-21.4%); reference category 0 = pooled phase I healthy volunteers, 1 = pooled phase II/III moderate-to-severe COPD patient), Facius_2018_roflumilast.R (linear additive phase II-III patient effects on KA (-73.3%), parent CL (-55.2%), N-oxide CL (-24.4%), and N-oxide central V3 (-20.7%) on the joint Lahu 2010 base model re-estimated on the OPTIMIZE + REACT phase III COPD dataset; reference category 0 = the implicit phase I healthy-volunteer cohort that backed the upstream Lahu 2010 fixed structural parameters), Yang_2013_losmapimod.R (mutually-exclusive proportional residual variance switch: sigma^2_prop = 0.061 for non-COPD subjects vs sigma^2_prop,COPD = 0.268 for COPD subjects, reproducing the NONMEM IF (COPD.EQ.1) EPS(2) ELSE EPS(1) $ERROR pattern; reference category 0 = pooled healthy volunteers plus rheumatoid arthritis patients), Siederer_2016_vilanterol.R.
• Notes: Used when a population PK/PD model pools healthy volunteers (with or without additional non-COPD patient cohorts such as rheumatoid arthritis in Yang 2013) with COPD patients and the COPD-vs-non-COPD contrast is retained either as a covariate on structural PK parameters or as a switch on residual-error magnitude. Scope: specific because the complement reference category and the COPD-severity inclusion criteria are paper-defined.

DIS_OBESE (canonical for obese cohort / health-status stratum indicator)

• Description: 1 = subject belongs to the source analysis’s obese stratum, 0 = otherwise. The unqualified obesity indicator: use it when a paper classifies subjects as obese without stating a threshold, or states an ordinary-obesity threshold (BMI >= 30 kg/m^2), as opposed to the morbid-obesity threshold that [[DIS_OBESE_MORBID]] carries. Time-fixed per subject. Frequently appears as one non-reference level of a multi-level health-status field alongside [[DIS_DIAB]], in which case the reference level is the “normal / healthy” stratum, recovered inside model() as 1 - DIS_OBESE - DIS_DIAB.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (non-obese). When the column is one level of a multi-level health-status field, 0 means “not in the obese stratum” and the reference stratum is defined jointly with the other level indicators (e.g. normal = DIS_OBESE == 0 & DIS_DIAB == 0).
• Source aliases:
  • HSTATUS (level “obese”) – used in Larizza_2018_cpeptideKinetics_M0.R and Larizza_2018_cpeptideKinetics_M4.R, where the source HSTATUS column carries three mutually exclusive levels (normal / obese / diabetic) and the model estimates a separate typical short half-life and amplitude fraction in each.
• Example models: Larizza_2018_cpeptideKinetics_M0.R, Larizza_2018_cpeptideKinetics_M4.R (selects the obese-stratum typical values of the C-peptide short half-life, 4.554 min in M0 vs 5.000 min for normal, and of the amplitude fraction, 0.782 vs 0.764; the Van Cauter 1992 / Magni 2000 cohort of 207 subjects stratified as normal, obese or diabetic).
• Notes: Registered as the unqualified sibling of [[DIS_OBESE_MORBID]], following the severity-suffix pattern the register already applies to DIS_CHF_NYHA2 / _NYHA3 / _NYHA4, DIS_INTELLDIS_MODSEV and DIS_TB_XDR_STRICT, and following the explicit instruction in the DIS_INTELLDIS_MODSEV Notes to register a sibling rather than widen an existing severity-qualified entry. The two are not interchangeable: a morbidly obese cohort (BMI > 40, typically bariatric-surgery patients) has a materially different prevalence and effect size from an ordinary obese stratum, which is exactly why the qualifier is in the name. Do not substitute BMI (the continuous body-mass-index covariate used for parameter scaling) – DIS_OBESE is a cohort-membership flag and does not encode a threshold for general use; record the paper’s own definition, or its absence, in the per-model covariateData[[DIS_OBESE]]$notes. General scope because the “obese vs not obese” contrast is a population descriptor that recurs across therapeutic areas, unlike the pooled-cohort-specific complement DIS_OBESE_MORBID carries.

DIS_OBESE_MORBID (canonical for morbidly obese cohort indicator)

• Description: 1 = morbidly obese patient (BMI > 40 kg/m^2 in the canonical definition; typical pooled-analysis enrollment criterion is bariatric-surgery patients), 0 = non-obese subject (typically healthy volunteer pooled in the source analysis). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-obese subject; the complement group is paper-defined – for de Hoogd 2017 the reference is the pooled healthy-volunteer cohort from Sarton 2000 and Romberg 2004).
• Source aliases: none known; source NONMEM control streams typically use ad-hoc names (e.g., OBESE, MO, COHORT).
• Example models: deHoogd_2017_morphine.R (selects per-cohort proportional residual error magnitudes for each of three observed species – morphine, M3G, M6G – after a pooled-cohort fit of 20 morbidly obese surgical patients and 20 healthy volunteers), Smit_2020_vancomycin.R (selects the cohort-specific between-subject variance on vancomycin clearance – 5.28% FIX in 8 nonobese healthy volunteers vs 24.7% in 20 morbidly obese bariatric-surgery patients, Smit 2020 Table 2 – with no effect on any typical value).
• Notes: Used when a population PK or PK/PD model pools morbidly obese patients with a non-obese reference population (typically healthy volunteers) and the cohort indicator selects per-cohort parameter values (residual error magnitudes, study-specific bioavailability, or similar). Distinct from BMI (which is the continuous body-mass-index covariate used for parameter scaling) – DIS_OBESE_MORBID is the binary cohort-membership flag and does not encode a specific BMI threshold for general use; the threshold is paper-defined. Scope: specific because the complement reference category is paper-defined.

DIS_INTELLDIS_MODSEV (canonical for complication with moderate or severe intellectual disability)

• Description: 1 = patient has a comorbid moderate or severe intellectual disability, 0 = no intellectual disability (reference). The canonical threshold is the one used by the founding paper: ICD-10 codes F70-F79 restricted to the moderate-or-severe range, stated by the source as consistent with IQ < 50. Time-fixed per subject. In epilepsy cohorts the covariate is typically a marker of a more severe underlying encephalopathy and correlates strongly with polytherapy, so a model that carries it may be unable to also carry the concomitant-AED indicators without collinearity (see Notes).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no intellectual disability, or an intellectual disability milder than the moderate-or-severe threshold).
• Source aliases:
  • Intellectual disability – used in Ogusu_2014_valproic_acid.R (Ogusu 2014 Eq. 9 superscript Intellectual disability, defined in the text below the equation as “intellectual disability = 1 if an intellectual disability was present, and was otherwise 0”; Table 3 row “Intellectual disability on BASE”; Table 1 row “Intellectual disability”, 131/237 = 55.3% of the PK cohort and 97/169 = 57.4% of the PK-PD cohort).
• Example models: Ogusu_2014_valproic_acid.R (additive shift on the intercept of a logistic regression for the probability of gamma-glutamyltransferase elevation: logit_ggt_elevation <- logit_ref + e_intelldis_logit * DIS_INTELLDIS_MODSEV + ... with e_intelldis_logit = 3.62 log-odds; the largest single effect in that model, Ogusu 2014 Table 3, RSE 28.4%, bootstrap 95% CI 1.87-8.94).
• Notes: The severity qualifier is load-bearing and belongs in the name, not only in this Description: an unrestricted “any intellectual disability” definition would have a materially different prevalence and a materially different effect size, so it is a different covariate, not a looser reading of this one. Register a sibling DIS_INTELLDIS_ANY if a future paper uses the unrestricted definition, rather than widening this entry. The severity-suffix pattern follows the established family precedents DIS_CHF_NYHA2 / _NYHA3 / _NYHA4, [[DIS_OBESE_MORBID]], DIS_TB_XDR_STRICT and [[DIS_INFECT_CSSSI_SEV]]. The name deliberately avoids the shorter DIS_ID, because ID is the universal subject-identifier column in every nlmixr2 / NONMEM dataset and DIS_ID reads as “disease identifier” at the use site. Collinearity warning. Ogusu 2014 reports that intellectual disability was significantly associated with an increased number of co-administered CBZ, PB and PHT (P < 0.0001), and for that reason dropped [[CONMED_CBZ]], [[CONMED_PB]] and [[CONMED_PHT]] from the final PK-PD model even though each was individually significant – so a model carrying DIS_INTELLDIS_MODSEV on a PD endpoint may legitimately omit the enzyme-inducing-AED indicators it carries on the PK side, and the intellectual-disability coefficient should be read as partly a proxy for that polytherapy. Ratified by the maintainers 2026-09-02 (PMC9833552; a register search confirmed DIS_ID, DIS_DD, “mental retardation”, “developmental delay”, “learning disability” and “NEURODEV” all absent; [[DIS_FXS]] is fragile X syndrome, a cause of intellectual disability rather than the condition itself).

HSCT_URD_7OF8 (canonical for hematopoietic stem cell transplant from a 7-of-8 HLA-matched unrelated donor)

• Description: 1 = patient received an allogeneic hematopoietic stem cell transplant (HSCT) from an unrelated donor (URD) HLA-matched at 7 of 8 alleles (single-allele mismatch), 0 = otherwise (the union of patients not in this transplant cohort, including non-HSCT patients and HSCT recipients matched at all 8 alleles). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (not in the 7-of-8-matched HSCT cohort; the complement group is paper-defined – for Zhong 2026 it pools the RA + pJIA studies and the 8-of-8-matched HSCT cohort, with the latter encoded by the parallel HSCT_URD_8OF8 indicator).
• Source aliases:
  • COHORT7 – used in Zhong_2026_abatacept.R (Zhong 2026 NM-TRAN indicator for ABA2 study Cohort 7/8).
• Example models: Zhong_2026_abatacept.R (exponential coefficient -0.326 on CL; the single-allele-mismatch HSCT cohort exhibits ~28% lower abatacept clearance than the reference complement).
• Notes: Used together with HSCT_URD_8OF8 to decompose a three-level “transplant cohort” categorical (non-HSCT-cohort / 7-of-8 / 8-of-8) into two orthogonal binary indicators. The 7-of-8 cohort represents a higher GvHD-risk population because of the single-allele HLA mismatch. Scope: specific because the reference complement (the union of non-transplant disease cohorts pooled in the source analysis) is paper-defined.

HSCT_URD_8OF8 (canonical for hematopoietic stem cell transplant from an 8-of-8 HLA-matched unrelated donor)

• Description: 1 = patient received an allogeneic hematopoietic stem cell transplant (HSCT) from an unrelated donor (URD) HLA-matched at all 8 alleles (full match), 0 = otherwise (the union of patients not in this transplant cohort, including non-HSCT patients and HSCT recipients matched at 7 of 8 alleles). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (not in the 8-of-8-matched HSCT cohort; the complement group is paper-defined – for Zhong 2026 it pools the RA + pJIA studies and the 7-of-8-matched HSCT cohort, with the latter encoded by the parallel HSCT_URD_7OF8 indicator).
• Source aliases:
  • COHORT8 – used in Zhong_2026_abatacept.R (Zhong 2026 NM-TRAN indicator for ABA2 study Cohort 8/8).
• Example models: Zhong_2026_abatacept.R (exponential coefficient -0.0934 on CL and +0.257 on VC; the fully-HLA-matched HSCT cohort exhibits a small CL decrease and a larger VC increase relative to the reference complement).
• Notes: Used together with HSCT_URD_7OF8 to decompose a three-level “transplant cohort” categorical (non-HSCT-cohort / 7-of-8 / 8-of-8) into two orthogonal binary indicators. The 8-of-8 cohort is the lower-risk HLA-matching configuration. Scope: specific because the reference complement (the union of non-transplant disease cohorts pooled in the source analysis) is paper-defined.

DIS_PSORIASIS (canonical for plaque psoriasis disease-state indicator)

• Description: 1 = plaque psoriasis patient, 0 = non-psoriasis subject (e.g., atopic dermatitis, ulcerative colitis, or healthy volunteer). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-psoriasis subject; the complement group is paper-defined – the union of other disease cohorts pooled in the source analysis).
• Source aliases: none known; source NONMEM control streams typically use a categorical DIS indicator (e.g., Okada 2025: DIS=1 for psoriasis, DIS=0 for healthy, DIS=2 for UC, DIS=3 for AD), decomposed into a binary DIS_PSORIASIS indicator at ingestion.
• Example models: Okada_2025_rocatinlimab.R (multiplicative shift 1 - 0.372 on linear CL when 1; reference complement is the pooled atopic dermatitis + ulcerative colitis + healthy-volunteer cohort), Warren_2025_apremilast.R (multiplicative factor 1.09 on CL/F when 0; reference complement is “other disease or missing”, i.e. the FDA Otezla popPK model treats unknown disease status the same as a non-psoriasis indication, and Warren 2025 simulates atopic dermatitis patients at DIS_PSORIASIS = 0), Wojciechowski_2022_abrocitinib.R (shares one additive +0.489 shift on logit absolute bioavailability with [[DIS_ATOPIC_DERMATITIS]]; reference complement is healthy volunteers, with mild and moderate hepatic impairment carried by HEPIMP_MILD / HEPIMP_MOD), Okubo_2021_apremilast.R (multiplicative factor 0.834 on CL/F when 1; reference complement is healthy subjects together with the psoriatic-arthritis / rheumatoid-arthritis ‘other’ cohort of one phase 1 study, which the final model did not separate from healthy).
• Notes: Used when a population PK model pools plaque-psoriasis patients with a non-psoriasis reference population and psoriasis disease status is retained as a covariate. Scope: specific because the disease-pooling reference category is paper-defined – the two registered models differ in it (Okada 2025 pools three named non-psoriasis cohorts; the FDA Otezla model behind Warren 2025 pools every non-psoriasis indication together with subjects whose disease status is missing), so the 1-side is transferable across papers but the 0-side is not.

DIS_ATOPIC_DERMATITIS (canonical for atopic dermatitis disease-state indicator)

• Description: 1 = patient with atopic dermatitis (atopic eczema), 0 = not (healthy volunteer or another disease cohort). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the paper-defined non-atopic-dermatitis complement; healthy volunteers in the founding model).
• Source aliases:
  • PTST = 2 – Wojciechowski 2022 Online Resource 2 patient-type column (0 = healthy, 1 = psoriasis, 2 = atopic dermatitis, 3 = mild and 4 = moderate hepatic impairment), decomposed into binary indicators at ingestion.
• Example models: Wojciechowski_2022_abrocitinib.R (founding example; shares one additive shift of +0.489 on logit absolute bioavailability with [[DIS_PSORIASIS]], e_dis_f * (DIS_PSORIASIS + DIS_ATOPIC_DERMATITIS), because the paper could not distinguish the two inflammatory skin diseases; reference is healthy volunteers).
• Notes: Spelled out because DIS_AD is already the Alzheimer’s-disease indicator. Scope specific because the 0 side is the union of whatever other cohorts a source pools.

DIS_RA (canonical for rheumatoid arthritis disease-state indicator)

• Description: 1 = adult rheumatoid arthritis patient, 0 = non-RA subject (e.g., healthy volunteer, Crohn’s disease, systemic lupus erythematosus, or other indication). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-RA subject; the complement group is paper-defined – the union of other cohorts pooled in the source analysis).
• Source aliases: none known; source NONMEM control streams typically derive the indicator from a POPULATION / STUDY categorical alongside other DIS_* indicators.
• Example models: Li_2018_PF04236921.R (log-shift e_ra_cl = log(0.00588 / 0.00546) = 0.0741 on linear CL, log-shifts on baseline CRP / IC50 / logit-Imax; one of three orthogonal indicators (DIS_RA / DIS_CD / DIS_SLE) decomposing the four-level HV / RA / CD / SLE cohort with HV as the reference category), Wojciechowski_2023_ritlecitinib_base.R, Wojciechowski_2023_ritlecitinib_updated.R (linear fractional effect on apparent clearance CL/F *= (1 + e_ra_cl * DIS_RA) with e_ra_cl = -0.439 in the base model and -0.496 in the updated model; reference category is the healthy-participant cohort, and the RA cohort also contributes to the pooled inflammatory-disease group that scales the IIV and proportional-residual-error magnitudes; one of four orthogonal indicators (DIS_RA / DIS_UC / DIS_ALOPECIA_AREATA / DIS_VITILIGO) decomposing the source’s multi-level PTST patient-type column), Klunder_2019_upadacitinib.R (multiplicative effect on apparent clearance, CL/F *= 0.754 for RA patients relative to healthy volunteers, encoded on the log scale as e_ra_cl = log(0.754); the source control stream carries the identical polarity as pref1 = 1; IF(RA.EQ.1) pref1 = THETA(12), so the source column maps onto the canonical orientation with no value transformation).
• Notes: Use when a population PK model pools adult RA patients with a non-RA reference population and RA disease status is retained as a covariate distinct from CD / SLE / other indications. Distinct from DIS_PJIA (polyarticular juvenile idiopathic arthritis, a pediatric-cohort sibling indicator). Scope: specific because the disease-pooling reference category is paper-defined. Note that the other two upadacitinib models in the library carry an RA-vs-healthy contrast under the COMPLEMENTARY DIS_HEALTHY column rather than this one, and should not be added to the Example models list above: the Klunder 2017 predecessor gates paired healthy/RA structural means and random effects rather than applying a single multiplier, and the Bhatnagar 2024 successor studies axial spondyloarthritis, so its patient level is not an RA cohort at all.

DIS_CIA (canonical for collagen-induced arthritis (preclinical) disease-state indicator)

• Description: 1 = animal with collagen-induced arthritis (CIA), the standard immunisation-induced rodent model of rheumatoid arthritis, studied at peak disease; 0 = age-, sex- and strain-matched healthy control animal. Time-fixed per animal within a study conducted at a single disease stage. Deliberately distinct from DIS_RA, whose Description is scoped to the adult human rheumatoid-arthritis patient: an induced rodent model and a human disease cohort differ in kind, and a covariate query that pooled them would return preclinical arms to a user asking for human RA models.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (healthy control animal from the same study; the control cohort is paper-defined).
• Source aliases:
  • CIA – the near-universal abbreviation in the preclinical inflammation literature; source datasets typically carry it as a group label (CIA vs healthy, or CF / HF / CM / HM when crossed with sex) rather than as a 1/0 column.
  • AA (adjuvant-induced arthritis) is not an alias: it is a different induction protocol with a different time course, and a paper using it should register a sibling indicator rather than reuse this one.
• Example models: Li_2017_naproxen_rat.R (Lewis rats at peak CIA – day 16 post induction in females, day 21 in males – versus healthy controls; selects the unbound plasma clearance, the unbound distribution clearance and the peripheral volume from the Arthritic / Healthy pairs of Table 4, selects the interstitial-to-plasma albumin ratio f_alb_isf (E/P = 0.9 arthritic vs 0.5 healthy), and jointly with SEXF selects the four albumin association constants kaff_* / kaff2_* of Table 1).
• Notes: Ratified 2026-09-02 alongside the Li 2017 naproxen extraction (PMC5399645). The species itself is recorded in population$species and in the model file’s _rat stem, so this indicator carries only the disease contrast and does not re-encode species. Follows the register’s own granularity precedent that arthritis cohorts differing in kind get separate indicators – DIS_RA is kept explicitly “Distinct from DIS_PJIA” for the same reason. A study that pools CIA animals with a different disease model rather than with healthy controls should say so in the per-model covariateData[[DIS_CIA]]$notes, since the reference category is paper-defined.

DIS_TCT_COLITIS (canonical for T-cell-transfer colitis (preclinical IBD model) disease-state indicator)

• Description: 1 = animal with colitis induced by adoptive transfer of naive CD45RB-high T cells into an immunodeficient (SCID or RAG-deficient) recipient, the standard T-cell-driven rodent model of inflammatory bowel disease; 0 = control animal from the same study that did not receive the T-cell transfer. Time-fixed per animal within a study conducted at a single disease stage. Deliberately distinct from DIS_UC, DIS_CD and IBD_CD, whose Descriptions are scoped to human patients: an induced rodent model and a human disease cohort differ in kind (same reasoning as DIS_CIA versus DIS_RA).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-transferred control animal from the same study; the control cohort is paper-defined).
• Source aliases:
  • IBD / non-IBD group labels (Zheng 2020 Table 1, “IBD mice” versus “Non-IBD mice”).
  • Chemically induced colitis models (dextran sulfate sodium, DSS; trinitrobenzene sulfonic acid, TNBS) are not aliases: they are different induction mechanisms with different time courses and should register sibling indicators rather than reuse this one.
• Example models: Zheng_2020_CNTO5048_mouse_mpbpk.R (female SCID mice 21 days after CD45RB-high T-cell transfer versus non-transferred SCID controls; selects the serum Michaelis-Menten capacity Vmax, the colon reflection coefficient, the colon clearance, and the fixed colon ISF volume and lymph flow from the IBD / non-IBD pairs of Table 2, and sets the serum and colon TNF baselines to zero in controls, where TNF was below the assay limit).
• Notes: Ratified 2026-09-29 alongside the Zheng 2020 CNTO 5048 extraction (PMC7531524), as a family-conforming DIS_<disease> preclinical sibling of DIS_CIA. The species is recorded in population$species and in the model file’s _mouse stem, so this indicator carries only the disease contrast.

DIS_CD (canonical for Crohn’s disease state indicator (multi-indication pooled analyses))

• Description: 1 = Crohn’s disease patient, 0 = non-CD subject (e.g., healthy volunteer, rheumatoid arthritis, systemic lupus erythematosus, or other indication). Time-fixed per subject. Distinct from IBD_CD, which is a pooled-UC+CD discriminator with UC as the reference category; DIS_CD is used when the complement group is a heterogeneous non-IBD cohort rather than UC specifically.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-CD subject; the complement group is paper-defined – the union of other cohorts pooled in the source analysis).
• Source aliases: none known; source NONMEM control streams typically derive the indicator from a POPULATION / STUDY categorical alongside other DIS_* indicators.
• Example models: Li_2018_PF04236921.R (log-shift e_cd_cl = log(0.00946 / 0.00546) = 0.5499 on linear CL – a 73 percent higher typical CL in CD vs the HV reference, consistent with the paper’s reported 60 percent higher CL in CD when other covariates are held at reference; also log-shifts on baseline CRP, IC50, and logit-Imax), Ternant_2021_infliximab.R (no coefficient of its own: Crohn’s disease is the reference category of the DIS_UC effect on central baseline TNF-alpha. DIS_CD enters only through the IBD switch IBD = DIS_CD + DIS_UC, which activates the central and peripheral QSS-TMDD terms; the complement with both indicators 0 is the ankylosing-spondylitis cohort used as the linear-kinetics reference).
• Notes: Use when a population PK model pools CD patients with a non-IBD reference population and Crohn’s disease status is retained as a covariate. When the analysis pools CD with UC only (and tests CD-vs-UC as a discriminator), use IBD_CD instead. Scope: specific because the disease-pooling reference category is paper-defined.

DIS_SLE (canonical for systemic lupus erythematosus disease-state indicator)

• Description: 1 = systemic lupus erythematosus patient, 0 = non-SLE subject (e.g., healthy volunteer, rheumatoid arthritis, Crohn’s disease, or other indication). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-SLE subject; the complement group is paper-defined – the union of other cohorts pooled in the source analysis).
• Source aliases: none known; source NONMEM control streams typically derive the indicator from a POPULATION / STUDY categorical alongside other DIS_* indicators.
• Example models: Li_2018_PF04236921.R (log-shift e_sle_cl = log(0.00643 / 0.00546) = 0.1632 on linear CL, log-shifts on baseline CRP / IC50 / logit-Imax, and a Hill coefficient effect e_sle_gamma = log(1.55) = 0.4383 shifting gamma from 1 in the HV/RA/CD reference to 1.55 in SLE), Pitsiu_2023_atacicept.R (residual-error stratification only: selects between two proportional SDs, 0.188 for healthy volunteers and 0.251 for patients with SLE; the paper tested SLE as a structural and bioavailability covariate and found no PK difference, so the flag has no effect on the structural model).
• Notes: Use when a population PK model pools SLE patients with a non-SLE reference population and SLE disease status is retained as a covariate. Future SLE-anchored extractions (e.g., anifrolumab, belimumab) that include SLE as a covariate against a non-SLE comparator should extend the example list. Distinct from BGENE21 / BGENE21_HIGH (continuous / binary IFN-21-gene scores within an SLE cohort) – those operate within an SLE-only population whereas DIS_SLE is the across-cohort disease-state flag. Scope: specific because the disease-pooling reference category is paper-defined.

DIS_INFECT_CSSSI_SEV (canonical for complicated skin and skin-structure infection severity indicator)

• Description: 1 = severe complicated skin and skin-structure infection (cSSSI), 0 = not severe cSSSI. Within-cohort severity indicator: stratifies severity inside an already-defined cSSSI / acute bacterial skin and skin-structure infection (ABSSSI) population, not a disease-vs-non-disease cohort indicator. Time-fixed per subject in the source analysis. The exact clinical criteria that classify a patient as severe vs not severe are protocol-defined within the trial dataset and may differ across antimicrobial development programs; per-model covariateData[[DIS_INFECT_CSSSI_SEV]]$notes should document the underlying definition when the source paper provides one.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (not severe cSSSI).
• Source aliases:
  • SOI (severity-of-infection indicator; same orientation, 1 = severe / 0 = not severe) – used in Lodise_2018_iclaprim.R (ASSIST-1 / ASSIST-2 phase 3 cSSSI trials).
• Example models: Lodise_2018_iclaprim.R (additive-linear shift on inter-compartmental clearance Q: q_typ = exp(lq) + e_infect_csssi_sev_q * DIS_INFECT_CSSSI_SEV with e_infect_csssi_sev_q = +13.5 L/h, so severe-cSSSI patients have Q rise from 1.85 L/h to 15.35 L/h relative to non-severe patients).
• Notes: Specific scope because the “severe cSSSI” definition is protocol-defined; future antimicrobial popPK papers that test a severity-of-infection contrast in a different infection class (pneumonia, HABP/VABP, bloodstream infection, bone and joint infection) should register sibling canonicals (e.g., DIS_INFECT_PNEUM_SEV, DIS_HABP_SEV) rather than overloading this entry. Distinct from DIS_SASTHMA and other disease-state indicators (which contrast a disease cohort with a non-disease reference) – DIS_INFECT_CSSSI_SEV operates within an already-cSSSI cohort. The covariate-effect parameter naming drops the DIS_ prefix per the existing DIS_CANCER -> e_cancer_* / DIS_CANCER_PED -> e_cancer_ped_* convention; here that gives e_infect_csssi_sev_<param>.

DIS_HABP (canonical for hospital-acquired bacterial pneumonia infection-type indicator)

• Description: 1 = the subject’s index infection is hospital-acquired bacterial pneumonia (HABP), 0 = otherwise. Infection-TYPE indicator (which infection the subject has), not a within-class severity indicator. Member of the mutually exclusive DIS_<infection> cohort-indicator set used by antimicrobial popPK analyses that pool healthy volunteers with several infected cohorts; the shared reference category (all indicators 0) is normally the uninfected healthy-volunteer stratum, but the reference must be documented per model because some analyses use one infection type as reference instead.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no HABP; in Cammarata_2024_sulbactam_durlobactam.R and Xie_2025_aztreonam_avibactam.R alike the shared all-zero reference is the uninfected Phase 1 subject).
• Source aliases:
  • INFTYPN = 1 – numeric infection-type flag level; used in Cammarata_2024_sulbactam_durlobactam.R (Table 1 rows CLINFTYPN1, V3INFTYPN1, and the merged V1INFTYPN1&2).
• Example models: Cammarata_2024_sulbactam_durlobactam.R (proportional shifts of -0.424 on sulbactam total CL and +0.836 on sulbactam Vc; durlobactam Vc uses a merged HABP-or-VABP coefficient of +1.52 applied to DIS_HABP + DIS_VABP), VanWart_2025_telavancin.R (a single coefficient shared across DIS_BACTEREMIA + DIS_HABP + DIS_VABP: +0.418 on total CL, +0.578 on Vc, +0.329 on Vp; second ratifying model), Patel_2022_imipenem.R / Patel_2022_relebactam.R (one ‘Pneumonia’ coefficient on DIS_HABP + DIS_VABP: -0.38 / -0.43 on CL and -0.39 / -0.29 on V1, against a pooled healthy + cIAI + cUTI reference).
  • Cammarata_2024_sulbactam_durlobactam.R (proportional shifts of -0.424 on sulbactam total CL and +0.836 on sulbactam Vc; durlobactam Vc uses a merged HABP-or-VABP coefficient of +1.52 applied to DIS_HABP + DIS_VABP).
  • Xie_2025_aztreonam_avibactam.R (shares the merged +0.931 cIAI-or-HAP-or-VAP coefficient applied to DIS_CIAI + DIS_HABP + DIS_VABP on the central volume of BOTH aztreonam and avibactam; neither drug carries a pneumonia-specific clearance term).
• Notes: Plain DIS_<disease> cohort indicator in the same shape as DIS_HEALTHY / DIS_CANCER / DIS_MDS / DIS_AML. Sibling to DIS_CIAI, DIS_VABP, DIS_CUTI, DIS_BACTEREMIA, and DIS_AP. The DIS_INFECT_<TYPE> prefix suggested by the DIS_INFECT_CSSSI_SEV Notes was considered and not chosen, because DIS_INFECT_CSSSI_SEV is a severity-WITHIN-cohort indicator while these are type-of-infection COHORT indicators (maintainers’ decision recorded with the Cammarata 2024 extraction, 2026-07-27). Covariate-effect naming drops the DIS_ prefix: e_habp_<param>, or e_habp_vabp_<param> / e_ciai_habp_vabp_<param> when a model shares one coefficient across several infection types.

DIS_VABP (canonical for ventilator-associated bacterial pneumonia infection-type indicator)

• Description: 1 = the subject’s index infection is ventilator-associated bacterial pneumonia (VABP), 0 = otherwise. Member of the mutually exclusive DIS_<infection> cohort-indicator set. See DIS_HABP for the shared-reference-category discipline.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no VABP; shared all-zero reference documented per model).
• Source aliases:
  • INFTYPN = 2 – numeric infection-type flag level; used in Cammarata_2024_sulbactam_durlobactam.R (Table 1 rows CLINFTYPN2, V3INFTYPN2, and the merged V1INFTYPN1&2).
• Example models: Cammarata_2024_sulbactam_durlobactam.R (proportional shifts of -0.298 on sulbactam total CL and +1.43 on sulbactam Vc; shares the merged +1.52 HABP-or-VABP coefficient on durlobactam Vc because the separate HABP and VABP durlobactam terms ‘did not yield significantly different effects on PK from one another’), VanWart_2025_telavancin.R (shares one coefficient with DIS_HABP and DIS_BACTEREMIA on CL, Vc and Vp; Van Wart 2025 Table 1 pools HABP and VABP into a single demographic stratum and never reports them separately; second ratifying model).
  • Cammarata_2024_sulbactam_durlobactam.R (proportional shifts of -0.298 on sulbactam total CL and +1.43 on sulbactam Vc; shares the merged +1.52 HABP-or-VABP coefficient on durlobactam Vc because the separate HABP and VABP durlobactam terms ‘did not yield significantly different effects on PK from one another’).
  • Xie_2025_aztreonam_avibactam.R (shares the merged +0.931 cIAI-or-HAP-or-VAP coefficient on the central volume of both drugs; Xie 2025 reports that steady-state exposures were similar between VAP and cIAI patients while HAP patients had the highest, yet the final model carries no VAP-specific coefficient distinct from HAP).
• Notes: Sibling to DIS_HABP and DIS_CIAI. When a source paper merges the pneumonia arms into one coefficient, keep the covariate COLUMNS distinct and apply the shared coefficient to their sum inside model() – do not collapse to a single DIS_HABP_VABP column, because sibling analyses (and the same paper’s other analyte) may separate them again. Both ratifying models so far merge them, in different groupings (Cammarata 2024 across HABP + VABP, Xie 2025 across cIAI + HAP + VAP), which is exactly why the columns stay separate.

DIS_CUTI (canonical for complicated urinary tract infection infection-type indicator)

• Description: 1 = the subject’s index infection is a complicated urinary tract infection (cUTI), 0 = otherwise. Member of the mutually exclusive DIS_<infection> cohort-indicator set. Papers that enroll a combined “cUTI including acute pyelonephritis” cohort but report separate model coefficients for the two should pair this column with DIS_AP; papers that report a single pooled coefficient should use this column alone and say so in the per-model notes.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no cUTI; shared all-zero reference documented per model).
• Source aliases:
  • INFTYPN = 3 – numeric infection-type flag level; used in Cammarata_2024_sulbactam_durlobactam.R (Table 1 rows V1INFTYPN3, CLINFTYPN3, V3INFTYPN3).
• Example models: Cammarata_2024_sulbactam_durlobactam.R (proportional shifts of +0.343 on durlobactam Vc, -0.157 on sulbactam total CL, and +0.17 on sulbactam Vc; cUTI subjects came from the Phase 2 study CS2514-2017-0003, in which acute pyelonephritis was reported as a separate infection-type level).
  • Cammarata_2024_sulbactam_durlobactam.R (proportional shifts of +0.343 on durlobactam Vc, -0.157 on sulbactam total CL, and +0.17 on sulbactam Vc; cUTI subjects came from the Phase 2 study CS2514-2017-0003, in which acute pyelonephritis was reported as a separate infection-type level).
  • Xie_2025_aztreonam_avibactam.R (proportional shifts of +0.222 on avibactam CL and +1.5 on avibactam Vc; both terms are avibactam-only because only 3 of the 431 subjects in the aztreonam data set had cUTI against 707 of 2,635 in the avibactam data set, so the aztreonam coefficients are absent and must not be borrowed).
  • Chandorkar_2015_ceftolozane.R (multiplicative exp(beta) factors x1.21 on CL and x1.21 on Vc; ceftolozane-only, with no companion coefficient in Chandorkar_2015_tazobactam.R).
  • Kawaguchi_2021_cefiderocol.R (a pooled cUTI-or-AUP flag whose effect on cefiderocol CL is study-specific – x1.27 in the phase 2 APEKS-cUTI study and x0.872 in the phase 3 CREDIBLE-CR study – selected by STUDY_CEFIDEROCOL_PHASE2; also carries the shared x1.39 any-infection factor on V1).
• Notes: Sibling to DIS_CIAI / DIS_HABP / DIS_VABP / DIS_BACTEREMIA / DIS_AP. All three ratifying models are fixed-ratio beta-lactam / beta-lactamase-inhibitor combinations in which the cUTI effect is carried by one analyte only, and in each case the reason is the composition of the cUTI sub-cohort rather than an incomplete table. Check the per-analyte subject counts before assuming a missing coefficient is an omission – and check the DOSING as well as the counts: in Chandorkar 2015 the asymmetry is absolute rather than merely lopsided, because the Phase 2 cUTI study (Umeh 2010, NCT00921024) administered ceftolozane WITHOUT tazobactam, so the tazobactam data set contains no cUTI subject at all and the paper says so explicitly (“note there were no tazobactam data from cUTI patients”).

DIS_BACTEREMIA (canonical for bacteremia / bloodstream-infection infection-type indicator)

• Description: 1 = the subject’s index infection is bacteremia (bloodstream infection), 0 = otherwise. Member of the mutually exclusive DIS_<infection> cohort-indicator set. Whether bacteremia secondary to another focus counts as bacteremia or as the primary focus is protocol-specific and must be documented per model.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no bacteremia; shared all-zero reference documented per model).
• Source aliases:
  • INFTYPN = 4 – numeric infection-type flag level; used in Cammarata_2024_sulbactam_durlobactam.R (Table 1 rows V1INFTYPN4, CLINFTYPN4, V3INFTYPN4).
• Example models: Cammarata_2024_sulbactam_durlobactam.R (proportional shifts of +3.32 on durlobactam Vc – the largest single covariate effect in the model – plus -0.444 on sulbactam total CL and +1.85 on sulbactam Vc; bacteremia due to Acinetobacter baumannii-calcoaceticus complex was one of the Phase 3 enrollment categories), VanWart_2025_telavancin.R (uncomplicated Staphylococcus aureus bacteremia from a single Phase 2 study, 18 of 1,205 subjects; shares one coefficient with DIS_HABP and DIS_VABP on CL, Vc and Vp; second ratifying model), Larson_2026_sulbactam_durlobactam_pediatric_allometry.R (retains the same three bacteremia shifts unchanged; poster P-444 set infection type to bacteremia for the whole simulated pediatric cohort, so this is the level the published pediatric exposure and target-attainment results were generated at), Trang_2019_plazomicin.R (plazomicin popPK; BSI carries its own -18.9% shift on total CL and is grouped with DIS_HABP / DIS_VABP for the Vc and CLd2 shifts. The CLd1:BSI and Vp1:BSI relationships were estimated but dropped after the nonparametric bootstrap for unacceptably poor precision).
  • Kawaguchi_2021_cefiderocol.R (the paper’s ‘bloodstream infection/sepsis’ category from the phase 3 CREDIBLE-CR study, n = 20: x1.08 on cefiderocol CL plus the shared x1.39 any-infection factor on V1; sepsis is pooled into this level, with no separate DIS_SEPSIS coefficient).
• Notes: Scope specific until a second model ratifies it; sibling to DIS_HABP / DIS_VABP / DIS_CUTI / DIS_AP / DIS_CSSSI.

DIS_BACTEREMIA_COMP (canonical for complicated Staphylococcus aureus bacteraemia adjudicated-diagnosis indicator)

• Description: 1 = the subject’s adjudicated final diagnosis is COMPLICATED bacteraemia – bloodstream infection with a deep-seated or metastatic focus, persistent bacteraemia, or an infected implanted device – 0 = otherwise. Severity-resolved member of the DIS_<infection> set, used when a source paper’s endpoint-adjudication committee splits bacteraemia into complicated and uncomplicated strata and estimates a separate covariate effect for each. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0. In a paper that decomposes an adjudicated-diagnosis categorical into indicators, the reference is the ALL-ZERO state across every member of the set, which is usually “no adjudicated diagnosis on record” rather than a clinical category – document which per model.
• Source aliases:
  • IEAC [Complicated Bacteraemia] / IEAC = 4 – Xu 2017 supplementary Table S2 row label for theta 18; IEAC is the source’s independent external adjudication committee final-diagnosis covariate, whose five levels are 1 = left-sided infective endocarditis, 2 = complicated right-sided IE, 3 = uncomplicated right-sided IE, 4 = complicated bacteraemia, 5 = uncomplicated bacteraemia.
• Example models: Xu_2017_daptomycin.R (multiplicative factor e_bacteremia_comp_cl = 1.10 on clearance in both the dialysis and the not-on-dialysis equation; Xu 2017 Table S2, 4% RSE; 58 of 459 subjects, the most frequent adjudicated diagnosis in that cohort).
• Notes: Distinct from DIS_BACTEREMIA, which is the unresolved bacteraemia-vs-other-infection-type indicator: use DIS_BACTEREMIA when the paper treats bacteraemia as one level of an infection-TYPE covariate, and this pair when it additionally splits by severity and estimates two coefficients. In a cohort where every bacteraemic subject falls into exactly one of the two strata, DIS_BACTEREMIA = DIS_BACTEREMIA_COMP + DIS_BACTEREMIA_UNCOMP; do not supply both the lumped and the split columns to one model. The complicated / uncomplicated boundary is protocol-defined and must be recorded per model. Paired with DIS_BACTEREMIA_UNCOMP and, in the founding source, with DIS_IEAC_LEFT / DIS_IEAC_RIGHT_COMP / DIS_IEAC_RIGHT_UNCOMP. Names ratified by the maintainers’ ruling 2026-09-21 (PMC5306496, with the spellings the maintainers specified): the endocarditis levels carry the DIS_IEAC_ prefix after the paper’s adjudication-committee label, and the severity qualifiers are _COMP / _UNCOMP rather than _COMPLICATED / _UNCOMPLICATED.

DIS_BACTEREMIA_UNCOMP (canonical for uncomplicated Staphylococcus aureus bacteraemia adjudicated-diagnosis indicator)

• Description: 1 = the subject’s adjudicated final diagnosis is UNCOMPLICATED bacteraemia (bloodstream infection clearing on therapy with no metastatic focus, no endocarditis and no retained infected device); 0 = otherwise. The lower-severity partner of DIS_BACTEREMIA_COMP; see that entry for the full rationale.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (all-zero across the adjudicated-diagnosis indicator set = no adjudicated diagnosis on record).
• Source aliases:
  • IEAC [Uncomplicated Bacteraemia] / IEAC = 5 – Xu 2017 supplementary Table S2 row label for theta 19.
• Example models: Xu_2017_daptomycin.R (multiplicative factor e_bacteremia_uncomp_cl = 1.13 on clearance; Xu 2017 Table S2, 5% RSE; 37 of 459 subjects).
• Notes: Counter-intuitively the founding source estimates a slightly LARGER clearance increase for uncomplicated (1.13) than for complicated (1.10) bacteraemia; the two 95% intervals overlap heavily and the difference is not interpreted by the authors. Transcribe the printed values rather than ordering them by severity. Names ratified by the maintainers’ ruling 2026-09-21 (PMC5306496, with the spellings the maintainers specified): the endocarditis levels carry the DIS_IEAC_ prefix after the paper’s adjudication-committee label, and the severity qualifiers are _COMP / _UNCOMP rather than _COMPLICATED / _UNCOMPLICATED.

DIS_IEAC_LEFT (canonical for left-sided infective endocarditis adjudicated-diagnosis indicator)

• Description: 1 = the subject’s adjudicated final diagnosis is LEFT-sided infective endocarditis (aortic or mitral valve involvement); 0 = otherwise. Left-sided IE differs prognostically and haemodynamically from right-sided disease – it is associated with systemic embolisation and a substantially worse outcome – which is why adjudication committees in Gram-positive-infection antimicrobial trials separate the two rather than recording a single “endocarditis” level. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (all-zero across the adjudicated-diagnosis indicator set = no adjudicated diagnosis on record; 337 of 459 subjects, 73%, in the founding source).
• Source aliases:
  • IEAC [LIE] / IEAC = 1 – Xu 2017 supplementary Table S2 row label for theta 15.
• Example models: Xu_2017_daptomycin.R (multiplicative factor e_ieac_left_cl = 1.16 on clearance; Xu 2017 Table S2, 12% RSE; 9 of 459 subjects).
• Notes: Head of the three-member infective-endocarditis indicator set DIS_IEAC_LEFT / DIS_IEAC_RIGHT_COMP / DIS_IEAC_RIGHT_UNCOMP, which together with DIS_BACTEREMIA_COMP / DIS_BACTEREMIA_UNCOMP decomposes a five-level adjudicated-diagnosis categorical. Register a separate DIS_IEAC_LEFT_COMP / DIS_IEAC_LEFT_UNCOMP pair only if a future source actually splits the left-sided stratum by severity; the founding source does not, and an all-zero column would fail the structural-cleanliness check. Distinct from DIS_ENDOCARDITIS-style lumped indicators (not currently registered) and from DIS_BACTEREMIA. Names ratified by the maintainers’ ruling 2026-09-21 (PMC5306496, with the spellings the maintainers specified): the endocarditis levels carry the DIS_IEAC_ prefix after the paper’s adjudication-committee label, and the severity qualifiers are _COMP / _UNCOMP rather than _COMPLICATED / _UNCOMPLICATED.

DIS_IEAC_RIGHT_COMP (canonical for complicated right-sided infective endocarditis adjudicated-diagnosis indicator)

• Description: 1 = the subject’s adjudicated final diagnosis is COMPLICATED right-sided infective endocarditis (tricuspid or pulmonic valve involvement with septic pulmonary emboli, persistent bacteraemia, large vegetation or extracardiac complication); 0 = otherwise. Member of the adjudicated-diagnosis indicator set headed by DIS_IEAC_LEFT.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (all-zero across the set).
• Source aliases:
  • IEAC [Complicated RIE] / IEAC = 2 – Xu 2017 supplementary Table S2 row label for theta 16.
• Example models: Xu_2017_daptomycin.R (multiplicative factor e_ieac_right_comp_cl = 1.3 on clearance; Xu 2017 Table S2, 9% RSE; 13 of 459 subjects).
• Notes: Right-sided IE is the labelled indication for daptomycin in Staphylococcus aureus bacteraemia, so it is a heavily represented stratum in that drug’s registrational database and likely to recur. See DIS_IEAC_LEFT for the set-level reference-category rule. Names ratified by the maintainers’ ruling 2026-09-21 (PMC5306496, with the spellings the maintainers specified): the endocarditis levels carry the DIS_IEAC_ prefix after the paper’s adjudication-committee label, and the severity qualifiers are _COMP / _UNCOMP rather than _COMPLICATED / _UNCOMPLICATED.

DIS_IEAC_RIGHT_UNCOMP (canonical for uncomplicated right-sided infective endocarditis adjudicated-diagnosis indicator)

• Description: 1 = the subject’s adjudicated final diagnosis is UNCOMPLICATED right-sided infective endocarditis; 0 = otherwise. The lower-severity partner of DIS_IEAC_RIGHT_COMP.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (all-zero across the set).
• Source aliases:
  • IEAC [Uncomplicated RIE] / IEAC = 3 – Xu 2017 supplementary Table S2 row label for theta 17.
• Example models: Xu_2017_daptomycin.R (multiplicative factor e_ieac_right_uncomp_cl = 1.3 on clearance; Xu 2017 Table S2, 10% RSE; 5 of 459 subjects).
• Notes: The founding source’s point estimate coincides with the complicated-RIE factor at 1.3, but the two were estimated as separate thetas (16 and 17) on 13 and 5 subjects respectively; keep the columns separate rather than collapsing them onto one shared coefficient, or a re-fit cannot recover the published parameterisation. Names ratified by the maintainers’ ruling 2026-09-21 (PMC5306496, with the spellings the maintainers specified): the endocarditis levels carry the DIS_IEAC_ prefix after the paper’s adjudication-committee label, and the severity qualifiers are _COMP / _UNCOMP rather than _COMPLICATED / _UNCOMPLICATED.

DIS_AP (canonical for acute pyelonephritis infection-type indicator)

• Description: 1 = the subject’s index infection is acute pyelonephritis (AP), 0 = otherwise. Member of the mutually exclusive DIS_<infection> cohort-indicator set. Used when a source paper reports AP as a model level distinct from DIS_CUTI, which is common because AP trials enroll under a combined “cUTI including AP” protocol but the two present with materially different renal and distributional physiology.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no AP; shared all-zero reference documented per model).
• Source aliases:
  • INFTYPN = 5 – numeric infection-type flag level; used in Cammarata_2024_sulbactam_durlobactam.R (Table 1 rows CLINFTYPN5, V3INFTYPN5; the table footnote defines ‘AP, acute pyelonephritis’).
• Example models: Cammarata_2024_sulbactam_durlobactam.R (proportional shifts of -0.382 on sulbactam total CL and -0.704 on sulbactam Vc; durlobactam carries no AP term because no V1INFTYPN5 row appears in the durlobactam half of Table 1), Larson_2026_sulbactam_durlobactam_pediatric_allometry.R (retains the same two AP shifts unchanged, inert in the simulated pediatric cohort, which was assigned bacteremia), Trang_2019_plazomicin.R (plazomicin popPK; AP carries its own +13.0% shift on total CL, distinct from the cUTI+AP group shift it shares with DIS_CUTI on Vc, CLd1, Vp1 and CLd2), Li_2019_ceftazidime.R (proportional shift of -0.185 on ceftazidime Vc, Li 2019 Table 1 theta11, stacking MULTIPLICATIVELY on top of the DIS_CUTI factor rather than replacing it; the companion Li_2019_avibactam.R carries no AP term, another instance of the per-analyte asymmetry this entry’s Notes warn against synthesising away).
• Notes: Scope specific until a second model ratifies it; sibling to DIS_CUTI. A model may legitimately carry an AP coefficient for one analyte and not the other – do not synthesise a missing coefficient by borrowing the cUTI value.

DIS_CIAI (canonical for complicated intra-abdominal infection infection-type indicator)

• Description: 1 = the subject’s index infection is a complicated intra-abdominal infection (cIAI), 0 = otherwise. Member of the mutually exclusive DIS_<infection> cohort-indicator set used by antimicrobial popPK analyses that pool healthy volunteers with several infected cohorts; the shared reference category (all indicators 0) is normally the uninfected healthy-volunteer stratum, but the reference must be documented per model. cIAI is a broad protocol-defined class – perforated viscus, intra-abdominal abscess, complicated appendicitis, complicated cholecystitis, secondary peritonitis – and is the most commonly enrolled infection type across beta-lactam / beta-lactamase-inhibitor development programs.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no cIAI; in Xie_2025_aztreonam_avibactam.R the shared all-zero reference is the uninfected Phase 1 healthy subject, and in the Chandorkar_2015_* pair it is the pooled healthy-volunteer plus renal-impairment stratum).
• Source aliases:
  • cIAI – printed row label in Xie_2025_aztreonam_avibactam.R (Table S3 rows cIAI on CL_ATM, cIAI on CL_AVI, and the shared cIAI/NP on Vc) and in Chandorkar_2015_ceftolozane.R / Chandorkar_2015_tazobactam.R (Table 3 rows With cIAI).
• Example models: Xie_2025_aztreonam_avibactam.R (proportional shifts of +0.279 on aztreonam CL and +0.115 on avibactam CL, plus a share in the merged +0.931 cIAI-or-HAP-or-VAP coefficient applied to the central volume of both drugs; cIAI raises the clearance of both analytes, which is why Xie 2025 chose cIAI as the most conservative population for its dose-selection simulations), Chandorkar_2015_ceftolozane.R (multiplicative exp(beta) factors x1.22 on CL and x1.59 on Vc, and – uniquely among the registered models – the indicator additionally GATES a second covariate, switching the body-weight effect on Vc off entirely for cIAI subjects because the paper found no weight-volume correlation in that group), Chandorkar_2015_tazobactam.R (x1.47 on Vc only; the same paper’s tazobactam analyte carries no cIAI clearance term).
• Notes: Plain DIS_<disease> cohort indicator, sibling to DIS_HABP, DIS_VABP, DIS_CUTI, DIS_BACTEREMIA and DIS_AP, and registered in exactly the shape those five were ratified in on 2026-07-28. Distinct from DIS_PERIT: DIS_PERIT is a peritonitis indicator (inflammation of the peritoneum specifically, usable in any critically-ill cohort), whereas DIS_CIAI is the broader trial-protocol infection-TYPE class of which peritonitis is one presentation; a paper that enrols a cIAI cohort and separately flags which of those subjects had peritonitis would legitimately carry both columns. Covariate-effect naming drops the DIS_ prefix per the family convention: e_ciai_<param>, or e_ciai_habp_vabp_<param> when a model shares one coefficient across cIAI and the two pneumonia types. The indicator can gate another covariate rather than only shifting a parameter: in Chandorkar_2015_ceftolozane.R the body-weight exponent on Vc is multiplied by (1 - DIS_CIAI), because Chandorkar 2015 reports “in cIAI patients, there was no significant correlation between Vc and body weight given the large observed variability” – read a cIAI row that silently drops a covariate term its sibling rows carry as a deliberate interaction, not as a printing omission.

DIS_OTHER_INFECT (canonical for residual ‘other infection’ infection-type indicator)

• Description: 1 = the subject’s index infection falls in the source paper’s residual “other infection” level, i.e. an infected patient who is not in any of the explicitly-modeled sibling infection types (DIS_CUTI, DIS_CIAI, DIS_PNEUMONIA, DIS_HABP, DIS_VABP, …); 0 = one of the explicitly-modeled infection types or the uninfected reference. Member of the mutually exclusive DIS_<infection> cohort-indicator set. Time-fixed per subject. The composition of the residual level is paper-defined and must be documented in per-model covariateData[[DIS_OTHER_INFECT]]$notes.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (not in the residual infection level; in Zhang_2021_ceftolozane.R / Zhang_2021_tazobactam.R the shared all-zero reference is the healthy-participant stratum, including renal-impairment subjects without infection).
• Source aliases:
  • OINF – equation term in Zhang_2021_ceftolozane.R / Zhang_2021_tazobactam.R (Table 2 equations 2.14^OINF and 2.49^OINF; row label ‘Fold-change in Vc for other infections’).
• Example models: Zhang_2021_ceftolozane.R (x2.14 on ceftolozane Vc), Zhang_2021_tazobactam.R (x2.49 on tazobactam Vc). In both, the level is the 10 critically ill patients of the augmented-renal-clearance cohort of study MK-7625A-007 (group 2, CrCL >= 180 mL/min, single dose of ceftolozane/tazobactam).
• Notes: Named with the _OTHER_ infix used for other residual categories in the register (DIS_OTHER_HEME, RACE_OTHER, TUMTP_OTHER). Not an augmented-renal-clearance flag and not a critical-illness flag, even when the paper’s residual level happens to be an ARC or ICU cohort: in Zhang 2021, ventilated pneumonia patients with CrCL > 150 mL/min are coded DIS_PNEUMONIA = 1, DIS_OTHER_INFECT = 0, so substituting RENAL_ARC or DIS_CRITILL would stack the other-infection factor on top of the pneumonia factor for patients the model never assigned to it. Different-composition residual levels do not share coefficients. Covariate-effect naming drops the DIS_ prefix per the family convention: e_other_infect_<param>.

CARRAGEENAN (canonical for intraplantar-carrageenan inflammatory-challenge indicator)

• Description: Binary indicator for intraplantar injection of carrageenan suspension as an experimental inflammatory / hyperalgesic challenge. 1 = subject received an intraplantar carrageenan injection at the start of the experiment (the carrageenan-induced peripheral inflammation / thermal-hyperalgesia paradigm); 0 = subject received an intraplantar saline injection (sham control). Time-fixed per subject within an experiment.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (saline-injected sham animal; no induced inflammation).
• Source aliases:
  • CARRAGEENAN – used in VasquezBahena_2009_lumiracoxib_rat.R (1 = groups II-IX, 100 uL of 1% carrageenan suspension into the right hind paw; 0 = group I, 100 uL of 0.9% saline solution).
• Example models: VasquezBahena_2009_lumiracoxib_rat.R (switches the COX-2 synthesis-rate model: CARRAGEENAN = 0 selects the constant saline synthesis rate ks_cox2_saline and CARRAGEENAN = 1 selects the time-variant gamma function ks_cox2(t) = A * t^alpha * exp(-beta * t) driving the carrageenan-induced inflammation profile).
• Notes: The carrageenan-induced peripheral inflammation / hyperalgesia model (Winter 1962; Hargreaves 1988 thermal-hyperalgesia variant) is one of the most widely used preclinical assays for screening anti-inflammatory and analgesic drugs in rodents, so the canonical name is reusable for future preclinical extractions. Scope: specific until a second model ratifies the binary-switch semantics on a different PD framework. The Hargreaves-test thermal-hyperalgesia readout (LT paw withdrawal latency, seconds) is the typical observable when this indicator is in use, but other readouts (paw oedema, mechanical-allodynia von Frey threshold) are equally valid. Distinct from disease-state indicators (DIS_*) because the inflammatory state is experimentally induced at a defined time, not a chronic patient condition.

MENT_DISABLED (canonical for severe mental-disability indicator (paper-defined diagnosis))

• Description: 1 = subject diagnosed with severe mental disability by the pediatrics department per the source-paper clinical criteria; 0 = mentally intact (ASA class 1, normal mental ability). Time-fixed per subject. Subject-level indicator; the underlying definition is broad (severe developmental / cognitive impairment with impairment of daily-living activities irrespective of primary neurologic cause, per Shin 2014 Results) rather than a single diagnostic category.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (mentally intact; ASA class 1).
• Source aliases:
  • MEN – used in Shin_2014_sevoflurane.R (Shin 2014 Appendix 1 NONMEM $INPUT column; same orientation: MEN = 1 mentally disabled, MEN = 0 mentally intact).
• Example models: Shin_2014_sevoflurane.R (stratifies both C50 and the Hill coefficient of the sigmoid-Emax probability of return of consciousness during emergence from sevoflurane anesthesia: typical-value C50 0.37 vol % vs 0.19 vol % and gamma 16.4 vs 4.53 for intact vs disabled).
• Notes: Specific scope because the “severe mental disability” diagnostic criterion is paper-defined (Shin 2014 enrolled pediatric patients diagnosed with severe mental disability by the pediatrics department, without specifying a structured instrument); per-model covariateData[[MENT_DISABLED]]$notes should document the diagnostic basis when the source paper provides one. Distinct from DIS_* disease-state indicators (which name a specific disease) and from cognitive-score covariates (SCORE_ADAS_COG, SCORE_MMSE, SCORE_CDR_SOB, SCORE_FAQ) which are continuous measurements; MENT_DISABLED is a binary developmental / cognitive-impairment cohort indicator. Future emergence-from-anesthesia or perioperative PK/PD models that use a similar binary cognitive-impairment indicator may extend this entry; promote to general scope once a second model ratifies the name with the same diagnostic criterion family.

DIS_CF (canonical for cystic-fibrosis disease-state indicator)

• Description: 1 = cystic fibrosis patient, 0 = healthy participant (or non-CF reference cohort pooled in the source analysis). Time-fixed per subject. Distinct from the Harun_2019_cysticFibrosis.R model which is single-cohort (all CF) and therefore carries no DIS_CF column; DIS_CF is the canonical for popPK / PD analyses that pool a CF patient cohort with a non-CF reference (typically healthy volunteers) and retain the CF-vs-non-CF contrast as a covariate.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-CF subject; the complement is paper-defined – for Bulitta 2007 the reference is the pooled adult healthy-volunteer cohort).
• Source aliases: none known; source NONMEM control streams typically encode the cohort indicator with ad-hoc names (e.g., CF, CYS, PATIENT, or via a STUDY / GROUP categorical).
• Example models: Bulitta_2007_piperacillin.R (multiplicative power-form effects on volumes via fcyf_vss^DIS_CF with fcyf_vss = 0.926, i.e. about a 7.4% reduction in V1 and V2 for CF patients vs healthy adults of the same lean body mass; CL carries the boundary-estimated fcyf_cl^DIS_CF with fcyf_cl = 1.00, retained from the published model for traceability though it has no numerical effect at the published point estimate; reference category is the pooled adult healthy-volunteer cohort), Bulitta_2011_cefpirome.R (log-scale effects on the canonical lcl_renal + lcl_nonren decomposition and on the central / shallow-peripheral / deep-peripheral volumes: e_cf_cl_renal = log(1.07) = +0.06766 on CL_R, e_cf_cl_nonren = log(1.13) = +0.12222 on CL_NR, and e_cf_vc_vp_vp2 = log(0.98) = -0.02020 applied uniformly to V1, V2, V3; typical values anchored to the HV column of paper Table 3 with FCYF effects from paper Table 4 LBM-allometric row), Mascarenhas_2015_pentadecanoic_triheptadecanoic.R (the CF-vs-healthy contrast selects which relative-bioavailability branch applies in the malabsorption blood test: healthy subjects are the F = 1 reference, while CF subjects with pancreatic insufficiency take an enzyme-dependent F for each of the two co-administered fatty-acid analytes; here DIS_CF is used with WT rather than LBM because Mascarenhas 2015 scaled allometrically on total body weight at a 70 kg reference), Bulitta_2019_pefloxacin.R (same investigator group and the same fcyf_*^DIS_CF power-form idiom as the 2007 piperacillin entry, extended to three separate factors – 1.53 on renal clearance, 0.861 on non-renal clearance, 0.916 on both volumes – and additionally used to multiplex wholly separate typical values AND separate between-subject variances for bioavailability, absorption rate, reabsorption rate and the two metabolite formation fractions, which is the first use of this canonical as a group selector rather than only as a scale-factor exponent), Shah_2019_cefotiam.R (the ONLY registered model in which DIS_CF acts on no clearance or volume at all: every unbound disposition parameter is shared across cohorts, and the indicator serves purely – jointly with SEXF – to select which of four separately estimated plasma unbound fractions applies, fu = 0.500 fixed in HV males, 0.545 in HV females, 0.563 in CF males and 0.744 in CF females; the FCYF scale factors the same group used for cefpirome are the alternative encoding these authors explicitly removed when they estimated fu).
• Notes: Used when a population PK / PD analysis pools CF patients with a non-CF reference (typically healthy volunteers) and the CF-vs-non-CF contrast is retained as a covariate. Scope: specific until a second model ratifies the canonical; promote to general at that point. Pair with LBM rather than WT as the size descriptor when the source paper found LBM to reduce unexplained BSV more than total body weight, which is common in CF cohorts whose body composition differs substantially from healthy adults.

DIS_IPF (canonical for idiopathic-pulmonary-fibrosis disease-state indicator)

• Description: 1 = patient with idiopathic pulmonary fibrosis (IPF), 0 = non-IPF subject (typically a healthy volunteer pooled in the same analysis). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-IPF subject; the complement is paper-defined – for Taneja 2019 it is the pooled healthy-volunteer cohort of the first-in-human and drug-drug-interaction studies).
• Source aliases:
  • Health status (HV vs IPF) – Taneja 2019 Sect. 2.3.3 and Table 3 (‘VP2 in IPF patients’, ‘RUV in patients with IPF’). Free-text label, not a column header.
• Example models: Taneja_2019_glpg1690.R (fractional increase of 5.59 on apparent central volume, vc <- exp(lvc + etalvc) * (1 + 5.59 * DIS_IPF), plus log-scale multipliers of 0.209 and 0.500 on the GLPG1690 and LPA C18:2 residual VARIANCES; the authors attribute the volume effect to the sparse proof-of-concept sampling schedule, which took no samples near Tmax, rather than to a physiological difference).
• Notes: Member of the DIS_<DISEASE> family (siblings: [[DIS_CF]], [[DIS_DPN]]). Oriented with the patient = 1 because the source estimates the effect on the patient cohort; [[DIS_HEALTHY]] (healthy = 1) would invert the reading of every effect parameter. Distinct from the IPF-trial MBMA arm-level covariate FVC_PCTPRED, which describes lung function within an all-IPF population. Scope: specific until a second model ratifies the canonical.

DIS_SEPSIS (canonical for active-sepsis co-condition indicator)

• Description: 1 = active sepsis / septic syndrome at the start of (or during) the modeled PK interval, 0 = no sepsis. Time-fixed per subject for the typical popPK use-case (Han 2013), although time-varying use is permitted; document per-model in covariateData[[DIS_SEPSIS]]$notes whether the indicator is fixed at PK-sampling onset or updated dynamically over the observation period. Used as a binary co-condition indicator on PK parameters (clearance, volume) when a study population pools septic and non-septic subjects, typically in ICU / burn-ICU / critically ill cohorts.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no sepsis).
• Source aliases:
  • SEPS – used in Han_2013_fluconazole.R (Han 2013 Methods / Table 1 covariate with values 0 = no sepsis, 1 = sepsis; same orientation as the canonical).
• Example models: Han_2013_fluconazole.R (additive shift on the non-RRT CL arm: CL_norrt = theta1 + (CLCR/120)*theta4 + TBI*theta7 + SEPS*theta9 with theta9 = -0.369 L/h – septic patients have about 0.37 L/h lower CL in the non-RRT arm), Jeon_2014_piperacillin.R (additive shift on central volume V1: TVV1 = exp(lvc) + e_sepsis_vc * DIS_SEPSIS with e_sepsis_vc = 14.8 L – septic patients have V1 = 25.3 + 14.8 = 40.1 L versus 25.3 L in non-septic patients, consistent with capillary leakage and interstitial edema in sepsis).
• Notes: Distinct from chronic disease-state indicators (e.g., DIS_UC, DIS_HAE) – sepsis is an acute / transient co-condition rather than a chronic indication label. Distinct from CRP / IL6 / SAPS_II, which are biomarker / severity-score columns that may co-vary with clinical sepsis but are conceptually distinct from a binary clinical-sepsis diagnostic flag.

DIS_INFECT_ACTIVE (canonical for active clinical infection episode indicator)

• Description: 1 = the record falls within an active clinical infection episode; 0 = no active infection. Time-varying per record (an acute, transient episode), in contrast to chronic disease-state indicators. The clinical criterion is paper-specific and MUST be recorded in covariateData[[DIS_INFECT_ACTIVE]]$notes – e.g. Kloos 2021 used “fever (>38 degrees Celsius) and hospital admission or prescription of antibiotics”. Used as a multiplicative or additive shift on clearance (and occasionally volume) when acute infection / inflammation alters drug elimination, e.g. via activation of the mononuclear phagocyte system.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no active infection).
• Source aliases:
  • INFECTION – used in Kloos_2021_pegasparaginase.R (Kloos 2021 Table 2 / Table 3 covariate; same orientation as the canonical).
• Example models: Kloos_2021_pegasparaginase.R (multiplicative on CL: 1.38^DIS_INFECT_ACTIVE, i.e. 38 percent higher PEGasparaginase clearance during an infection, attributed by the authors to activation of the mononuclear phagocyte system that clears PEGylated asparaginase), AbdullahKoolmees_2024_voriconazole_pbpk.R (divides each hepatic CYP Vmax by a per-isoenzyme fold decrease – CYP2C19 1.79, CYP3A4 4.6, CYP2C9 1.5 – so severe infection lowers voriconazole intrinsic clearance, the opposite direction to Kloos 2021; the clinical criterion is the paper’s C-reactive-protein banding of its simulated patient types rather than a fever / antibiotic rule), Leeds_2013_tecovirimat_cyno.R (log-additive shift on Ka and on CL/F in cynomolgus monkeys, i.e. 48.1 percent faster tecovirimat absorption and 0.6 percent lower apparent oral clearance during an active monkeypox-virus infection; the criterion here is experimental rather than clinical – a record is infected if it follows intravenous inoculation with approximately 5e7 PFU of MPXV, and Leeds 2013 study 2 dosed each animal once 10 days before inoculation to obtain the uninfected occasion within animal).
• Notes: Distinct from [[DIS_SEPSIS]], which denotes a specific septic syndrome – DIS_INFECT_ACTIVE is the broader “any active clinical infection episode” flag and is typically time-varying where DIS_SEPSIS is often time-fixed at PK-sampling onset. Distinct from DIS_INFECT_CSSSI_SEV (an infection-severity grade in a specific-scope cSSSI indication). Distinct from its constituent measurements BODYTEMP, CRP, and the CONMED_<INN> antibiotic columns, which are biomarker / comedication columns that may co-vary with a clinical infection flag but are conceptually separate.

DIS_EDEMA (canonical for clinical edema presence indicator)

• Description: 1 = clinical edema present (typical clinical definition: puffy face and/or pitting peripheral edema), 0 = no clinical edema. Time-fixed per subject in the typical popPK use-case (Han 2013), although time-varying use is permitted; document per-model in covariateData[[DIS_EDEMA]]$notes whether the indicator is fixed at PK-sampling onset or updated dynamically. Used as a binary co-condition indicator on PK volume (most commonly) when third-space / extravascular fluid expansion is expected to alter drug distribution.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no clinical edema).
• Source aliases:
  • EDEM – used in Han_2013_fluconazole.R (Han 2013 Methods / Table 1 covariate with values 0 = no edema, 1 = clinical edema by the paper’s puffy-face and leg-pitting criterion; same orientation as the canonical).
• Example models: Han_2013_fluconazole.R (additive shift on V: V = (WT/65)*theta5 + EDEM*theta6 + TBI*theta8 with theta6 = 13.6 L – edematous patients have about 13.6 L larger V).
• Notes: Conceptually distinct from DIS_HAE (hereditary angioedema disease-state indicator) – DIS_EDEMA is a generic clinical-sign flag (e.g., burn-related capillary leak, congestive heart failure, hypoalbuminemia) rather than a disease classification. Distinct from a continuous fluid-balance / weight-gain column.

DIS_BURN_RECENT (canonical for recent-postburn-injury hypermetabolic-phase indicator)

• Description: 1 = within the paper-defined recent-postburn window (i.e., the acute hypermetabolic phase shortly after burn injury), 0 = beyond the window. Time-varying within a subject is permitted as time since injury crosses the cutoff; in Han 2013 the cutoff is 30 days postburn (the hypermetabolic response is maximized between days 7-17 and substantially resolved by day 30; the binary recoding captures patients whose physiology is dominated by the hypermetabolic period). The exact recency threshold is paper-specific – document the threshold in each model’s covariateData[[DIS_BURN_RECENT]]$notes. Used as a piecewise covariate on PK parameters (CL and / or V) that change during the acute hypermetabolic phase.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (beyond the recent-postburn window; i.e., resolved hypermetabolic phase).
• Source aliases:
  • TBI – used in Han_2013_fluconazole.R (Han 2013 Methods / Tables 2 and 3 covariate with values 1 = within 30 days of burn injury, 0 = at least 30 days postburn; same orientation as the canonical, with the recency threshold = 30 days documented per-model). NOTE: TBI in this register is a recoded binary covariate; it does NOT denote traumatic brain injury (the common medical abbreviation) and is NOT a continuous time-from-burn-injury value.
  • Burn-S – used in Lin_2021_vancomycin.R (Lin 2021 Methods covariate list; coded 1 = burn degree > 50%, acute convalescence, 2 = burn degree < 50%, 20 days after burn, with the non-burn majority of the ICU cohort also carrying code 2 per Table 1; DIS_BURN_RECENT = 2 - Burn-S, rebuilt in model()).
• Example models: Han_2013_fluconazole.R (recency threshold = 30 days; additive shifts on both CL and V: CL_norrt += TBI*theta7 with theta7 = 0.504 L/h and V += TBI*theta8 with theta8 = 9.61 L – recent-postburn patients have about 0.5 L/h higher CL and 9.6 L larger V relative to the resolved-phase reference), Lin_2021_vancomycin.R (severity + recency indicator: 1 = burn degree > 50% in acute convalescence, 0 = burn degree < 50% or more than 20 days after burn, and all non-burn ICU patients; additive-linear term in the CL equation, -0.43 L/h * Burn-S with Burn-S = 2 - DIS_BURN_RECENT, i.e. a severe acute burn raises CL by 0.43 L/h).
• Notes: Generalised binary recoding rather than a continuous time-from-burn-injury column. A future paper that uses a continuous days-from-burn-injury covariate (analogous to TPP for time-postpartum) would warrant its own canonical (suggested name: POSTBURN_DAYS); DIS_BURN_RECENT is reserved for binary phase indicators. The paper’s source-column name TBI is preserved as a documented alias even though it collides with the medical abbreviation for traumatic brain injury – this is the author’s chosen column name and the canonical avoids the collision.

DIS_ARDS (canonical for acute respiratory distress syndrome indicator)

• Description: 1 = acute respiratory distress syndrome (ARDS) present at the start of (or during) the modeled PK interval, 0 = no ARDS. Diagnostic standard is paper-specific (Berlin 2012 criteria, the older American-European Consensus Conference / AECC criteria, or a clinician-adjudicated diagnosis on the chart) – document per-model in covariateData[[DIS_ARDS]]$notes which definition was used. Used as a binary co-condition indicator on PK parameters (most often clearance and / or volume) when ARDS is hypothesized to alter drug disposition through pulmonary inflammation, increased pulmonary reactive oxygen species, ventilator-induced lung injury, or altered fluid balance and capillary leak.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no ARDS).
• Source aliases:
  • ARDS – used in Taubert_2016_linezolid.R (Taubert 2016 patient-group-1 baseline indicator; ARDS = 1 multiplies CL by 1.82, encoded inside model as a log-additive shift e_dis_ards_cl = log(1.82) per Taubert 2016 page 5256 covariate-model equation).
• Example models: Taubert_2016_linezolid.R (multiplicative factor on CL: typical-value CL is multiplied by 1.82 in ARDS patients, equivalent to a ~82% increase in linezolid clearance; the proposed mechanism is non-enzymatic oxidation of linezolid by elevated pulmonary reactive oxygen species in ARDS lungs per Taubert 2016 Discussion).
• Notes: Distinct from DIS_SEPSIS (acute septic syndrome co-condition) and from DIS_CKD / DIS_HEPATIC chronic organ-failure indicators – ARDS is an acute pulmonary syndrome that may coexist with sepsis but has its own pathophysiology and is captured separately when both are tested as covariates. General scope because ARDS has a stable internationally-standardised clinical definition (Berlin criteria) and is universally applicable as a critically-ill PK covariate. Time-varying within a subject is permitted as the syndrome resolves or worsens.

DIS_PERIT (canonical for peritonitis indicator)

• Description: 1 = peritonitis (infection / inflammation of the peritoneum, including primary spontaneous bacterial peritonitis, secondary peritonitis from a perforated viscus or anastomotic leak, and tertiary / persistent peritonitis) present at the start of (or during) the modeled PK interval, 0 = no peritonitis. Used as a binary co-condition indicator on PK parameters when peritoneal inflammation, intra-abdominal fluid sequestration, third-space loss, or altered hepatic synthetic function is hypothesized to change drug disposition.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no peritonitis).
• Source aliases:
  • Peritonitis – used in Taubert_2016_linezolid.R (Taubert 2016 patient-group-1 baseline indicator; peritonitis = 1 multiplies Vc by 1.53, encoded inside model as a log-additive shift e_dis_perit_vc = log(1.53) per Taubert 2016 page 5256 covariate-model equation).
• Example models: Taubert_2016_linezolid.R (multiplicative factor on Vc: typical-value central volume is multiplied by 1.53 in peritonitis patients, equivalent to a ~53% larger central volume of distribution, attributed in Taubert 2016 Discussion to third-space fluid accumulation in the peritoneal cavity).
• Notes: Distinct from PERIT_DIAL (chronic peritoneal-dialysis modality indicator) – DIS_PERIT is an acute infection / inflammation of the peritoneum, while PERIT_DIAL is a chronic therapeutic dialysis modality; the two have entirely different pathophysiologies and PK implications. Future papers that retain peritonitis as a covariate may extend Example models. General scope because peritonitis has a clear universal clinical definition and any future critically-ill PK model with the same encoding may reuse the canonical.

DIS_SCOL_IDIO (canonical for idiopathic-scoliosis aetiology indicator)

• Description: 1 = subject has idiopathic scoliosis (no underlying systemic / syndromic cause; classic adolescent idiopathic scoliosis or idiopathic kyphoscoliosis), 0 = otherwise (subject is in either the non-idiopathic-scoliosis cohort or the reference complement cohort). Time-fixed per subject. Used as a binary cohort indicator in popPK / popPD models that pool an adolescent scoliosis population with another paediatric surgical reference cohort and the scoliosis-aetiology effect on PK is retained as a covariate.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (not idiopathic-scoliosis; the complement reference is paper-defined – in Stricker 2015 the implicit reference cohort is the pooled infant craniofacial-reconstruction cohort from the earlier Stricker 2013 BJA paper).
• Source aliases:
  • paper narrative idiopathic scoliosis / idiopathic kyphoscoliosis diagnosis labels in Stricker 2015 Table 1, decomposed from a 3-level diagnosis categorical {craniofacial, idiopathic scoliosis, non-idiopathic scoliosis} into binary DIS_SCOL_IDIO with value 1 for the idiopathic-scoliosis cohort.
• Example models: Stricker_2015_aminocaproic_acid.R (multiplicative power-form effect on CL: cl *= 1.10^DIS_SCOL_IDIO; idiopathic-scoliosis adolescents have ~10% higher CL than the craniofacial-cohort reference at the same body weight and post-natal age, after allometric scaling; Stricker 2015 Table 5 ‘Impact of idiopathic spines’ = 1.10; effect retained for magnitude / precision reporting even though the authors note it is not clinically meaningful).
• Notes: Specific scope because the reference complement (infant craniofacial reconstruction cohort) is paper-defined. Pairs with DIS_SCOL_NONIDIO in the Stricker 2015 three-level surgical-cohort stratification {craniofacial reference, idiopathic scoliosis, non-idiopathic scoliosis}; both = 0 selects the craniofacial reference. Distinct from ADOLESCENT (pure age-cohort indicator independent of aetiology) and from CHILD (broader paediatric age-cohort indicator).

DIS_SCOL_NONIDIO (canonical for non-idiopathic / syndromic scoliosis aetiology indicator)

• Description: 1 = subject has non-idiopathic / syndromic scoliosis (cerebral palsy, Marfan syndrome, Ehlers-Danlos syndrome, neurofibromatosis type 1, spina bifida, congenital neuromuscular scoliosis, syringomyelia, 22q deletion, cortical dysgenesis, and similar systemic / syndromic aetiologies), 0 = otherwise. Time-fixed per subject. Used as a binary cohort indicator in popPK / popPD models that pool a syndromic-scoliosis paediatric population with another paediatric surgical reference cohort and the scoliosis-aetiology effect on PK is retained as a covariate.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (not non-idiopathic-scoliosis; the complement reference is paper-defined – in Stricker 2015 the implicit reference cohort is the pooled infant craniofacial-reconstruction cohort from the earlier Stricker 2013 BJA paper).
• Source aliases:
  • paper narrative diagnosis labels in Stricker 2015 Table 1 (cerebral palsy, Marfan, Ehlers-Danlos, neurofibromatosis, spina bifida, congenital neuromuscular scoliosis, syringomyelia, 22q deletion, cortical dysgenesis), decomposed from the 3-level diagnosis categorical {craniofacial, idiopathic scoliosis, non-idiopathic scoliosis} into binary DIS_SCOL_NONIDIO with value 1 for any of the syndromic-scoliosis aetiologies.
• Example models: Stricker_2015_aminocaproic_acid.R (multiplicative power-form effect on CL: cl *= 0.97^DIS_SCOL_NONIDIO; non-idiopathic-scoliosis adolescents have ~3% lower CL than the craniofacial-cohort reference at the same body weight and post-natal age, after allometric scaling; Stricker 2015 Table 5 ‘Impact of non-idiopathic spines’ = 0.97; the authors describe this effect as not clinically meaningful but retain it in the final model for precision estimation).
• Notes: Specific scope because the reference complement (infant craniofacial reconstruction cohort) is paper-defined and because the non-idiopathic class pools many aetiologically heterogeneous syndromes (cerebral palsy, connective-tissue disorders, neuromuscular disorders, neurocutaneous disorders, etc.) that future papers may want to resolve into separate sub-cohort indicators. Pairs with DIS_SCOL_IDIO in the Stricker 2015 three-level surgical-cohort stratification {craniofacial reference, idiopathic scoliosis, non-idiopathic scoliosis}; both = 0 selects the craniofacial reference.

DIS_CHF (canonical for compensated congestive heart failure indicator)

• Description: 1 = subject has compensated congestive heart failure (chronic CHF stabilised on diuretics, cardiac glycosides, or other background therapy; NYHA class typically not stratified at this level), 0 = no diagnosed CHF. Time-fixed per subject in popPK / popPK-PD analyses that include CHF as a baseline disease-state covariate.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no CHF).
• Source aliases:
  • chf – used in Thomson_1989_lisinopril.R (Thomson 1989 Table 2 cardiac-failure indicator; treated as a multiplicative factor 0.645 raised to chf on CL/F so the unaffected reference group keeps factor 1).
• Example models: Thomson_1989_lisinopril.R (multiplicative power-form effect on CL/F: e_chf_cl^DIS_CHF with e_chf_cl = 0.645; ~35% lower apparent clearance with compensated cardiac failure).
• Notes: Use this canonical when the source paper tests a binary cardiac-failure indicator (compensated, stable, on background therapy) as a structural covariate on PK parameters. Distinct from concomitant-medication indicators such as CONMED_SPIRON (which captures the diuretic / aldosterone-antagonist effect typically given alongside CHF therapy) and from haemodynamic / autonomic covariates such as HR and the blood-pressure outputs. For papers that decompose CHF severity into NYHA classes, prefer a per-class encoding (e.g., DIS_CHF_NYHA3, DIS_CHF_NYHA4) when the strata are retained in the final model. The covariate-effect parameter form is e_chf_<param> (drops the DIS_ prefix per the existing DIS_CANCER -> e_cancer_<param> precedent).

DIS_HFPEF (canonical for heart failure with preserved or mildly reduced ejection fraction indicator)

• Description: 1 = participant has symptomatic heart failure with a preserved or mildly reduced left ventricular ejection fraction (HFpEF or HFmrEF, conventionally LVEF >= 40%, usually together with elevated natriuretic peptides), 0 = no heart failure. Subject-level, time-fixed. Use for the “disease status” covariate of a pooled analysis whose only heart-failure cohort was enrolled on an explicit preserved- or mildly-reduced-ejection-fraction criterion.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no heart failure). In a pooled healthy-volunteer-plus-patient design the reference group is the healthy volunteers, but any non-heart-failure patient cohort enrolled for a different reason – a renal-impairment cohort, for instance – also takes 0, so the complement is “no heart failure” rather than “healthy”.
• Source aliases:
  • disease status (healthy volunteers or patients with HFpEF/HFmrEF) – used in Sunnaker_2026_mitiperstat.R (Sunnaker 2026 Table 2 row “Disease status”, Table 3 row “Effect of disease status (patients with HFpEF/HFmrEF vs. HVs) on CL/F”).
• Example models: Sunnaker_2026_mitiperstat.R (linear fractional effect on apparent clearance, CL/F * (1 + e_hfpef_cl * DIS_HFPEF) with e_hfpef_cl = -0.23, i.e. apparent clearance is 23% lower in the SATELLITE heart-failure cohort than in the reference no-heart-failure group; Sunnaker 2026 Table 3 final model – founding example).
• Notes: Member of the DIS_<CONDITION> family; covariate-effect parameter form is e_hfpef_<param>, dropping the DIS_ prefix per the DIS_CANCER -> e_cancer_<param> and DIS_CHF -> e_chf_<param> precedents. Registered separately from DIS_CHF rather than folded into it. DIS_CHF is an ejection-fraction-agnostic “compensated congestive heart failure” indicator whose founding example (Thomson_1989_lisinopril.R) predates the HFpEF/HFrEF split, and modern heart-failure trials enrol on ejection fraction as the primary phenotype criterion: HFpEF/HFmrEF and HFrEF differ in pathophysiology, comorbidity burden and background therapy, so a clearance shift estimated in one does not transfer to the other and the two coefficients must not share a column. A paper that enrols an ejection-fraction-unselected or explicitly reduced-ejection-fraction cohort should use DIS_CHF or a future DIS_HFREF sibling, not this entry. Distinct also from DIS_CHF_NYHA2 / DIS_CHF_NYHA3 / DIS_CHF_NYHA4, which stratify symptom severity within a heart-failure cohort rather than naming the ejection-fraction phenotype; a model may legitimately carry one of each. Scope specific until a second heart-failure model reuses it.

DIS_RETT (canonical for Rett syndrome disease-state indicator)

• Description: 1 = participant has Rett syndrome (RTT), a rare X-linked neurodevelopmental disorder caused by loss-of-function variants in MECP2; 0 = healthy volunteers and any other disease cohort. Subject-level, time-fixed. Rett syndrome occurs almost exclusively in females, so in a pooled analysis this indicator is effectively collinear with female sex within the patient cohort; document that collinearity per model rather than fitting both a sex and an RTT term on the same PK parameter unless the source paper did.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (healthy volunteers, per the usual pooled healthy-volunteer-plus-patient design).
• Source aliases:
  • Rett – used in Darwish_2025a_trofinetide.R and Darwish_2025b_trofinetide.R (Darwish 2025 Results, definition of the Rett_I indicator variable in the typical-value equations for CL and Vp).
  • RTT – the standard abbreviation used throughout the Rett syndrome literature and in the Darwish 2025 narrative text.
• Example models: Darwish_2025a_trofinetide.R (two proportional shifts from Darwish 2025 Table 2: cl * (1 + e_rett_cl * DIS_RETT) with e_rett_cl = -0.169 (16.9% lower clearance) and vp * (1 + e_rett_vp * DIS_RETT) with e_rett_vp = 0.616 (61.6% larger peripheral volume). The indicator additionally selects the disease-cohort residual-error magnitude, which Darwish 2025 pooled across Rett syndrome, fragile X syndrome, and traumatic brain injury – founding example), Darwish_2025b_trofinetide.R (two proportional shifts from Darwish 2025 Table 2: cl * (1 + e_rett_cl * DIS_RETT) with e_rett_cl = -0.169 (16.9% lower clearance) and vp * (1 + e_rett_vp * DIS_RETT) with e_rett_vp = 0.616 (61.6% larger peripheral volume). The indicator additionally selects the disease-cohort residual-error magnitude, which Darwish 2025 pooled across Rett syndrome, fragile X syndrome, and traumatic brain injury – founding example).
  • Darwish_2025b_trofinetide.R (the 14-study update of the same sponsor analysis: e_rett_cl = -0.146 (14.6% lower clearance) and e_rett_vp = 0.805 (80.5% larger peripheral volume), Darwish 2025b Table 2. These two shifts were the ONLY covariate effects to move by more than 10% when the DAFFODIL pediatric data were added – by +13.6% and +30.7% respectively – which makes them the sharpest available check that a transcription used the updated table rather than the stale equation block printed in that paper’s Table 2 footnote).
• Notes: Member of the DIS_<CONDITION> family of disease-state indicators. The covariate-effect parameter form is e_rett_<param> (drops the DIS_ prefix, per the DIS_CANCER -> e_cancer_<param> and DIS_CHF -> e_chf_<param> precedents). Distinct from DIS_FXS (fragile X syndrome) even though both are monogenic neurodevelopmental disorders studied with overlapping trofinetide programmes – the two cohorts have opposite sex distributions and Darwish 2025 estimated separate, differently-signed effects on different PK parameters, so they must not be collapsed into a single “neurodevelopmental disorder” indicator.

DIS_TBI (canonical for traumatic brain injury disease-state indicator)

• Description: 1 = participant has a traumatic brain injury (TBI); 0 = healthy volunteers and any other disease cohort. Subject-level, time-fixed. Used where a source paper pools a TBI cohort with healthy volunteers or other patient groups and retains a TBI shift on a PK parameter.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (healthy volunteers).
• Source aliases:
  • TBI – used in Darwish_2025a_trofinetide.R and Darwish_2025b_trofinetide.R (Darwish 2025 Results, definition of the TBI_I indicator variable in the typical-value equations for CL and Vp).
• Example models: Darwish_2025a_trofinetide.R (two proportional shifts from Darwish 2025 Table 2: cl * (1 + e_tbi_cl * DIS_TBI) with e_tbi_cl = 0.235 (23.5% higher clearance) and vp * (1 + e_tbi_vp * DIS_TBI) with e_tbi_vp = -0.752 (75.2% smaller peripheral volume). The indicator additionally selects the disease-cohort residual-error magnitude – founding example), Darwish_2025b_trofinetide.R (two proportional shifts from Darwish 2025 Table 2: cl * (1 + e_tbi_cl * DIS_TBI) with e_tbi_cl = 0.235 (23.5% higher clearance) and vp * (1 + e_tbi_vp * DIS_TBI) with e_tbi_vp = -0.752 (75.2% smaller peripheral volume). The indicator additionally selects the disease-cohort residual-error magnitude – founding example).
  • Darwish_2025b_trofinetide.R (the 14-study update: e_tbi_cl = 0.229 and e_tbi_vp = -0.753, Darwish 2025b Table 2, both within 3% of the earlier estimates).
• Notes: Member of the DIS_<CONDITION> family; covariate-effect parameter form e_tbi_<param>. Name-collision warning: DIS_BURN_RECENT documents a source alias also spelled TBI (used in Han_2013_fluconazole.R), where the author’s column name abbreviates a recoded burn-injury recency flag rather than traumatic brain injury. The two canonicals are unrelated. When a source paper uses a bare TBI column, read the paper’s own variable definition before mapping it – map to DIS_TBI only when the paper defines it as traumatic brain injury, and to DIS_BURN_RECENT when it defines it as time since burn injury.

DIS_TRAUMA (canonical for acute trauma or head injury comorbidity indicator)

• Description: 1 = patient has an acute trauma or head injury as a recorded comorbidity; 0 = otherwise. Subject-level, time-fixed. Used where an ICU popPK analysis carries trauma (including isolated head injury) as a comorbidity category associated with augmented renal clearance of hydrophilic drugs.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (any other comorbidity category; paper-defined).
• Source aliases:
  • COMO == 3 (trauma) or COMO == 4 (brain injury) – the FLAT2 indicator of the Werumeus Buning 2021 Appendix C NONMEM control stream, which pools the two comorbidity categories into one.
• Example models: WerumeusBuning_2021_ceftazidime.R (multiplicative factor 1.99^DIS_TRAUMA on the off-CVVH ceftazidime clearance in critically ill adults; 27/96 patients).
• Notes: Member of the DIS_<CONDITION> family; covariate-effect parameter form e_trauma_<param>. Werumeus Buning 2021 merged trauma with head injury because both act through “the hyperdynamic state with glomerular hyperfiltration”, so a head-injury-only patient is DIS_TRAUMA = 1. Distinct from DIS_TBI, which is a disease-COHORT indicator contrasted against healthy volunteers in a pooled trofinetide analysis, and from ICU_ADM_POLYTRAUMA, which is one of three mutually exclusive ICU admission etiologies (a trauma comorbidity can coexist with a medical or surgical admission category, as it does in Werumeus Buning 2021 Table 1). A paper that separates trauma from head injury should use DIS_TRAUMA for trauma and document the head-injury column separately.

DIS_HEME_MALIG (canonical for hematologic malignancy comorbidity indicator)

• Description: 1 = patient has a hematologic malignancy (leukemia, lymphoma, myeloma or similar) as a recorded comorbidity; 0 = otherwise. Subject-level, time-fixed. Used where a non-oncology popPK analysis (typically ICU anti-infectives) carries hematologic malignancy as a comorbidity category.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no hematologic malignancy; paper-defined, and may include solid-tumour malignancy).
• Source aliases:
  • COMO == 1 – the FLAT1 indicator of the Werumeus Buning 2021 Appendix C NONMEM control stream (“hematologic malignancy”).
• Example models: WerumeusBuning_2021_ceftazidime.R (multiplicative factor 1.57^DIS_HEME_MALIG on the off-CVVH ceftazidime clearance in critically ill adults; 14/96 patients; oncologic (solid) malignancy was a separate comorbidity category and sits in the reference group).
• Notes: Member of the DIS_<CONDITION> family; covariate-effect parameter form e_heme_malig_<param>. This is the whole-class comorbidity flag. Distinct from the per-disease indicators used in pooled hematologic-oncology analyses (DIS_AML, DIS_MM, DIS_OTHER_HEME and siblings), where every subject has a hematologic malignancy and the indicators contrast one disease against another, and from DIS_CANCER (advanced solid tumours).

DIS_FXS (canonical for fragile X syndrome disease-state indicator)

• Description: 1 = participant has fragile X syndrome (FXS), the X-linked trinucleotide-repeat expansion disorder of FMR1 and the most common inherited cause of intellectual disability; 0 = healthy volunteers and any other disease cohort. Subject-level, time-fixed. Clinically penetrant FXS predominates in males, so in a pooled analysis this indicator tends to be collinear with male sex within the patient cohort.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (healthy volunteers).
• Source aliases:
  • FXS – used in Darwish_2025a_trofinetide.R and Darwish_2025b_trofinetide.R (Darwish 2025 Results, definition of the FXS_I indicator variable in the typical-value equation for Vc).
• Example models: Darwish_2025a_trofinetide.R (proportional shift on central volume from Darwish 2025 Table 2: vc * (1 + e_fxs_vc * DIS_FXS) with e_fxs_vc = 1.15, i.e. a 115% larger central volume – more than a doubling, the largest single covariate effect in that model. The indicator additionally selects the disease-cohort residual-error magnitude – founding example), Darwish_2025b_trofinetide.R (proportional shift on central volume from Darwish 2025 Table 2: vc * (1 + e_fxs_vc * DIS_FXS) with e_fxs_vc = 1.15, i.e. a 115% larger central volume – more than a doubling, the largest single covariate effect in that model. The indicator additionally selects the disease-cohort residual-error magnitude – founding example).
  • Darwish_2025b_trofinetide.R (the 14-study update: e_fxs_vc = 1.16, Darwish 2025b Table 2, a 116% larger central volume and still the largest single covariate effect in the model).
• Notes: Member of the DIS_<CONDITION> family; covariate-effect parameter form e_fxs_<param>. See the DIS_RETT notes for why the two neurodevelopmental-disorder cohorts are registered separately rather than pooled.

DIS_CSSSI (canonical for complicated skin and skin-structure infection infection-type indicator)

• Description: 1 = the subject’s index infection is a complicated skin and skin-structure infection (cSSSI, also published as acute bacterial skin and skin-structure infection, ABSSSI), 0 = otherwise. Infection-TYPE indicator (which infection the subject has), not a within-cSSSI severity indicator. Member of the mutually exclusive DIS_<infection> cohort-indicator set used by antimicrobial popPK analyses that pool healthy volunteers with several infected cohorts; see DIS_HABP for the shared-reference-category discipline.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no cSSSI; in VanWart_2025_telavancin.R the shared all-zero reference across DIS_CSSSI / DIS_HABP / DIS_VABP / DIS_BACTEREMIA is the uninfected healthy subject).
• Source aliases: none known; source NONMEM control streams normally derive the indicator from an infection-type categorical alongside the sibling DIS_* columns.
• Example models: VanWart_2025_telavancin.R (proportional shifts of +0.228 on total CL, +0.313 on Vc and +0.118 on Vp; cSSSI is the largest infected stratum of the pooled analysis at 557 of 1,205 subjects, and carries its own coefficient distinct from the pooled bacteremia / HABP / VABP coefficient on the same three parameters), Bulitta_2024_contezolid.R (selects the separately estimated patient contezolid total clearance, 11.3 vs 10.2 L/h in healthy volunteers, AND its own between-subject variability, 0.538 vs 0.234, by multiplexing two etas; the 74 ABSSSI patients of phase 2 study MRX-I-03 are the only infected cohort, the reference is the healthy volunteer), Li_2021_tedizolid.R (linear fractional shifts of +0.220 on CL and +0.0987 on Vc in ABSSSI patients relative to healthy volunteers, in a pooled adult / adolescent / pediatric tedizolid analysis).
• Notes: Sibling to DIS_HABP, DIS_VABP, DIS_CUTI, DIS_BACTEREMIA and DIS_AP; adds the cSSSI member to that set using the DIS_<infection> pattern the maintainers ratified with the Cammarata 2024 extraction (2026-07-27), where the alternative DIS_INFECT_<TYPE> prefix was offered and not chosen. Distinct from DIS_INFECT_CSSSI_SEV, which is a severity-WITHIN-cohort indicator (severe vs non-severe cSSSI, Lodise_2018_iclaprim.R) rather than a type-of-infection cohort indicator: a model may legitimately carry both, DIS_CSSSI selecting the cSSSI cohort against a non-cSSSI reference and DIS_INFECT_CSSSI_SEV stratifying severity inside it. Covariate-effect naming drops the DIS_ prefix: e_csssi_<param>. Scope specific until a second model ratifies it.

DIS_GERD (canonical for gastroesophageal-reflux-disease patient indicator)

• Description: 1 = patient with gastroesophageal reflux disease (GERD), 0 = non-GERD reference (e.g., healthy volunteer, or another indication pooled in the source analysis). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-GERD subject; the complement group is paper-defined – for Yang 2024 the reference is the 68-subject phase 1 healthy-volunteer cohort screened free of any gastrointestinal disorder).
• Source aliases:
  • disease status / PT – Yang 2024 narrative and Table 2 footnote a (PT = 1 for patient, 0 for healthy volunteer).
• Example models: Yang_2024_zastaprazan.R (linear fractional effect on apparent clearance of the 2-compartment Erlang-absorption oral zastaprazan model: CL/F * (1 + e_dis_gerd_cl * DIS_GERD) with e_dis_gerd_cl = -0.414, i.e. CL/F 41.4% lower in patients with erosive GERD (17.2 L/h) than in healthy volunteers (29.4 L/h); reference category is the phase 1 healthy-volunteer cohort of Study 1, and the patient cohort is the 92-subject phase 2 erosive-GERD cohort of Study 2; Yang 2024 Table 2 and footnote a).
• Notes: Follows the DIS_UC / DIS_DUOD_ULCER / DIS_PSORIASIS / DIS_HAE pattern of specific-disease-vs-non-disease pooled-cohort indicators. Distinct from DIS_DUOD_ULCER, which is a peptic-duodenal-ulcer cohort indicator: both are upper-GI acid-related conditions treated with the same drug classes (PPIs and P-CABs), but they are different diagnoses and a subject can carry either independently. Also distinct from DIS_HEALTHY, which carries a healthy-vs-pooled-patient contrast with 0 = patient; a paper enrolling a single named disease cohort alongside healthy participants is encoded with the disease-specific indicator so the source coefficients and typical values transfer without a sign flip or re-baselining (the same reasoning recorded under DIS_DUOD_ULCER). If a future paper contrasts erosive against non-erosive GERD within a GERD-only cohort, register a separate within-cohort severity indicator on the DIS_CMV_RETINITIS pattern rather than widening or re-pointing this entry. Scope: specific because the complement reference category is paper-defined; promote to general if a second paper pools GERD patients with a non-GERD reference.

DIS_FANCONI (canonical for Fanconi anemia diagnosis indicator)

• Description: 1 = the subject’s underlying diagnosis is Fanconi anemia (FA), the inherited DNA-interstrand-crosslink-repair deficiency syndrome; 0 = the paper-defined non-Fanconi reference diagnosis. Time-fixed per subject. Fanconi anemia is a constitutional (germline) bone-marrow-failure syndrome, in contrast to the acquired immune-mediated marrow aplasia of idiopathic aplastic anemia, and the distinction matters pharmacologically because Fanconi cells are hypersensitive to alkylating and crosslinking agents, which is why Fanconi transplant protocols use markedly attenuated conditioning.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the paper-defined comparator diagnosis; in Mohanan_2017_fludarabine.R this is acquired aplastic anemia, pooling the AA / SAA / VSAA severity grades).
• Source aliases:
  • Diagnosis (coded 0 = AA/SAA/VSAA aplastic anemia, 1 = FA Fanconi anemia) – Mohanan 2017 Table 2 footnote a and the Results covariate-definition caption; same orientation as the canonical, no value transformation.
• Example models: Mohanan_2017_fludarabine.R (log-scale multiplicative effect on the BSA-normalised clearance of the fludarabine nucleoside F-ara-A, exp(e_fanconi_cl * DIS_FANCONI) with e_fanconi_cl = log(2.90 / 7.12) = -0.898, i.e. clearance in Fanconi anemia is 0.407 times that in aplastic anemia, equivalently the 2.46-fold higher aplastic-anemia clearance quoted in the Mohanan 2017 Abstract, P < 1e-6; reference category acquired aplastic anemia).
• Notes: Member of the DIS_<CONDITION> family; the covariate-effect parameter form drops the DIS_ prefix, giving e_fanconi_<param>, per the DIS_CANCER -> e_cancer_<param> and DIS_CHF -> e_chf_<param> precedents. Fanconi anemia is registered as the indicator and aplastic anemia as the reference, rather than the reverse, because the printed typical value that the source model quotes (7.12 L/h/m^2) belongs to the aplastic-anemia group; keeping the published orientation lets the source table’s values be carried into the model file unchanged without a sign flip or re-baselining, following the rule recorded in the DIS_OUD and DIS_CHB notes. Register a separate DIS_AA sibling if a future paper needs an acquired-aplastic-anemia indicator against a different reference – do not re-express this column as its complement, because the reference group is paper-defined and need not be aplastic anemia. Distinct from DIS_FRDA (Friedreich ataxia), whose abbreviation is deliberately spelled out for the same reason this entry is: a bare DIS_FA would be ambiguous against folic acid and fatty acid. Distinct also from DIS_PNH and DIS_MDS, the other non-malignant / pre-malignant marrow-failure states, which are separate diagnoses that a marrow-failure cohort may carry alongside this one. Scope: specific because the reference diagnosis is paper-defined; promote to general if a second paper contrasts Fanconi anemia against the same acquired-aplastic-anemia reference.

Treatment-emergent adverse-event indicators

Binary indicators flagging that a participant is currently experiencing a specific treatment-emergent adverse event, used where the event itself perturbs the drug’s pharmacokinetics (most often gastrointestinal events altering oral absorption or bioavailability). Naming follows AE_<EVENT>, where <EVENT> is the MedDRA-style preferred term in upper case.

These are deliberately not registered under the DIS_<CONDITION> family. DIS_* entries describe the participant’s underlying disease or comorbid state – the reason they are in the trial – and are almost always time-fixed. An AE_* entry describes a drug-caused, typically transient and time-varying event occurring during treatment. The distinction matters because both can appear in the same model on the same parameter: Darwish_2025a_trofinetide.R carries DIS_RETT (the disease under study) and AE_DIARRHEA (the drug’s most common adverse event) simultaneously, and conflating them would obscure which effect is disease-driven and which is treatment-emergent. Record per model in covariateData[[AE_<EVENT>]]$notes whether the source carried the flag per record (time-varying) or per subject (ever-experienced).

AE_DIARRHEA (canonical for concurrent treatment-emergent diarrhea indicator)

• Description: 1 = participant is experiencing treatment-emergent diarrhea at the time of the record; 0 = no diarrhea. Diarrhea shortens gastrointestinal transit time and reduces the window available for absorption, so it typically enters oral popPK models as a negative effect on bioavailability or a positive effect on the absorption rate constant.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no diarrhea).
• Source aliases:
  • Diar – used in Darwish_2025a_trofinetide.R and Darwish_2025b_trofinetide.R (Darwish 2025 Results, definition of the Diar_I indicator variable in the typical-value equation for F1; the paper explicitly describes it as a time-varying indicator).
  • DIAR – the all-caps form used in the Darwish 2025 Fig. 5 forest-plot panel labels.
• Example models: Darwish_2025a_trofinetide.R (proportional shift on oral bioavailability from Darwish 2025 Table 2: fdepot * (1 + e_diarrhea_f * AE_DIARRHEA) with e_diarrhea_f = -0.148, a 14.8% reduction in F1 while diarrhea is present. Diarrhea is trofinetide’s most common adverse event, affecting 52.4% of the Rett syndrome participants in the analysis dataset at some point during the studies, which is why the sponsor carried it as a structural covariate rather than screening it out – founding example), Darwish_2025b_trofinetide.R (proportional shift on oral bioavailability from Darwish 2025 Table 2: fdepot * (1 + e_diarrhea_f * AE_DIARRHEA) with e_diarrhea_f = -0.148, a 14.8% reduction in F1 while diarrhea is present. Diarrhea is trofinetide’s most common adverse event, affecting 52.4% of the Rett syndrome participants in the analysis dataset at some point during the studies, which is why the sponsor carried it as a structural covariate rather than screening it out – founding example).
  • Darwish_2025b_trofinetide.R (the 14-study update: e_diarrhea_f = -0.157, Darwish 2025b Table 2, a 15.7% reduction in F1 while diarrhea is present).
• Notes: Founding member of the AE_<EVENT> family. Carry the flag per dose record when the source models it as time-varying (the Darwish 2025 form) and per subject only when the source collapses it to an ever-experienced flag; state which in the per-model notes. Distinct from DIS_UC, DIS_DUOD_ULCER, and the other gastrointestinal DIS_* entries, which denote a pre-existing gastrointestinal disease rather than a treatment-emergent event.

Ophthalmology: geographic atrophy lesion characteristics

Baseline morphology and laterality descriptors of a geographic atrophy (GA) lesion secondary to age-related macular degeneration, as graded on fundus autofluorescence imaging of the study eye. All members are time-fixed per subject and are graded on the study eye; the fellow (non-randomised) eye carries no covariate in the models that use these columns. Distinct from the TUM* oncology lesion-burden family (TUM_SLD, TUMSZ, NTARGET_GE3), which describes solid-tumour burden under RECIST, and from BCVA_ETDRS (a functional visual-acuity score rather than an anatomic lesion descriptor).

DIS_GA_UNILATERAL (canonical for unilateral vs bilateral geographic atrophy indicator)

• Description: 1 = unilateral geographic atrophy at baseline (GA present in the randomised study eye only); 0 = bilateral GA (GA present in both eyes at baseline). Time-fixed per subject. Bilateral GA is an established indicator of accelerated atrophy (Desai 2022 Eye), so the unilateral stratum is the slower-progressing group.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (bilateral GA – the majority stratum). Reference values observed: 81% bilateral in the pooled FILLY + OAKS + DERBY pegcetacoplan population (Crass 2025 Table 1); 19% unilateral.
• Source aliases:
  • UNILATGA – Crass 2025 Data S1 equation notation for the indicator entering INIT_i,study (theta_11) and SLOPE_i,study (theta_12).
  • GA laterality – the Crass 2025 Table 1 demographic row name, reported as the complementary Unilateral / Bilateral pair.
• Example models: Crass_2025_pegcetacoplan_ga_progression.R, Crass_2025_pegcetacoplan_ga_doseresponse.R, Crass_2025_pegcetacoplan_ga_exposureresponse.R (prespecified structural covariate; proportional shifts (1 + e_unilat_rbase_study * DIS_GA_UNILATERAL) on the study-eye initial lesion area and (1 + e_unilat_slope_study * DIS_GA_UNILATERAL) on the study-eye progression rate; in the final PK/PD fit the slope coefficient -0.139 gives the paper’s 0.860-fold lower lesion growth for unilateral GA).
• Notes: General scope – laterality of GA is a routinely collected baseline characteristic in every GA natural-history and interventional trial, so future ophthalmology extractions should reuse this column. Because a subject with unilateral GA has no atrophic fellow eye, joint study-eye / fellow-eye models drop the fellow-eye records for that subject (Crass 2025 excluded fellow-eye observations for patients with unilateral GA at baseline who converted to bilateral disease during the study). The covariate-effect parameter form is e_unilat_<param> (drops the DIS_GA_ prefix per the existing DIS_CANCER -> e_cancer_<param> precedent).

DIS_GA_NONSUBFOVEAL (canonical for nonsubfoveal geographic atrophy lesion location indicator)

• Description: 1 = the study-eye GA lesion does not involve the fovea (nonsubfoveal / extrafoveal); 0 = subfoveal involvement. Time-fixed per subject; graded at baseline on fundus autofluorescence. Encoded in the “absence of foveal involvement” direction because that is the direction the source literature reports the effect in (nonsubfoveal lesions progress faster, so the coefficient is positive on the lesion growth rate).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (subfoveal involvement). Reference values observed: 63% subfoveal, 37% nonsubfoveal in the pooled Crass 2025 population (Table 1).
• Source aliases:
  • NOFOV – Crass 2025 Data S1 equation notation (theta_19).
  • GA lesion location – the Crass 2025 Table 1 demographic row name, reported as the complementary Subfoveal / Nonsubfoveal pair. A source paper reporting SUBFOVEAL = 1 for foveal involvement is the value-inverted alias: set DIS_GA_NONSUBFOVEAL = 1 - SUBFOVEAL and flip the sign convention of the effect accordingly.
• Example models: Crass_2025_pegcetacoplan_ga_progression.R, Crass_2025_pegcetacoplan_ga_doseresponse.R, Crass_2025_pegcetacoplan_ga_exposureresponse.R (proportional shift (1 + e_nonsubfov_slope_study * DIS_GA_NONSUBFOVEAL) on the study-eye progression rate only; in the final PK/PD fit the coefficient 0.118 gives the paper’s 1.12-fold higher lesion growth for nonsubfoveal lesions).
• Notes: General scope – foveal involvement is graded in every GA trial because it governs the visual-acuity consequences of atrophy. Registered in the nonsubfoveal-is-1 direction to match the source-paper convention; document the inversion in covariateData[[DIS_GA_NONSUBFOVEAL]]$notes when a future paper reports the complementary column. The covariate-effect parameter form is e_nonsubfov_<param>. Distinct from DIS_GA_UNIFOCAL (lesion multiplicity rather than lesion position).

DIS_GA_UNIFOCAL (canonical for unifocal vs multifocal geographic atrophy lesion indicator)

• Description: 1 = unifocal study-eye GA lesion (a single atrophic patch); 0 = multifocal lesion (two or more separate atrophic patches). Time-fixed per subject; graded at baseline on fundus autofluorescence. Multifocal lesions have a larger total perimeter for a given area and progress faster, so the unifocal stratum is the slower-progressing group.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (multifocal). Reference values observed: 70% multifocal, 30% unifocal in the pooled Crass 2025 population (Table 1; one FILLY patient had missing data).
• Source aliases:
  • UNIFOC – Crass 2025 Data S1 equation notation (theta_20).
  • GA focality – the Crass 2025 Table 1 demographic row name, reported as the complementary Unifocal / Multifocal pair.
• Example models: Crass_2025_pegcetacoplan_ga_progression.R, Crass_2025_pegcetacoplan_ga_doseresponse.R, Crass_2025_pegcetacoplan_ga_exposureresponse.R (proportional shift (1 + e_unifocal_slope_study * DIS_GA_UNIFOCAL) on the study-eye progression rate only; in the final PK/PD fit the coefficient -0.164 gives the paper’s 0.837-fold lower lesion growth for unifocal lesions).
• Notes: General scope. GA trial entry criteria commonly impose a minimum patch size for multifocal lesions (Crass 2025 required at least one lesion >= 1.25 mm^2 if GA was multifocal), so the multifocal stratum is not simply “any number of small patches”; record the per-paper entry criterion in covariateData[[DIS_GA_UNIFOCAL]]$notes. The covariate-effect parameter form is e_unifocal_<param>.

DRUSEN_GT20 (canonical for baseline drusen burden binarised at 20 intermediate or large drusen groups)

• Description: Binary indicator dichotomising baseline drusen burden in the study eye at 20 groups: 1 = more than 20 intermediate or large drusen groups (individual drusen diameter >= 63 micrometres, the AREDS simplified-severity-scale size threshold of Ferris 2005 AREDS report No. 18); 0 = 20 or fewer. Time-fixed per subject. Drusen are extracellular deposits between the retinal pigment epithelium and Bruch’s membrane and are the defining lesion of early / intermediate AMD; a higher baseline drusen burden is associated with slower subsequent GA lesion growth.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (20 or fewer intermediate or large drusen groups). Reference values observed: 45% of study eyes above the threshold and 55% at or below it in the pooled Crass 2025 population (Table 1; two OAKS patients had missing data).
• Source aliases:
  • MOREDR – Crass 2025 Data S1 equation notation (theta_21).
  • No. of intermediate or large drusen groups – the Crass 2025 Table 1 demographic row name, reported as the complementary <= 20 / > 20 pair.
• Example models: Crass_2025_pegcetacoplan_ga_progression.R, Crass_2025_pegcetacoplan_ga_doseresponse.R, Crass_2025_pegcetacoplan_ga_exposureresponse.R (proportional shift (1 + e_drusen_slope_study * DRUSEN_GT20) on the study-eye progression rate only; in the final PK/PD fit the coefficient -0.130 gives the paper’s 0.869-fold lower lesion growth for the high-drusen-burden stratum).
• Notes: Follows the count-covariate policy (“count covariate -> decomposed binary indicators, not a single integer count”) and the NTARGET_GE3 precedent of binarising a lesion count at a paper-defined threshold. No DIS_ prefix because drusen are an AMD imaging feature that is graded independently of the GA diagnosis, in the same way NTARGET_GE3 records RECIST lesion multiplicity without a disease-state prefix. Source papers supply the already-dichotomised indicator rather than a raw count; if a future paper reports the integer count, derive DRUSEN_GT20 = as.integer(count > 20) and record the raw column as a source alias. If a future paper uses a different split (e.g. > 10 or > 30) or a different size threshold, register a sibling canonical rather than overloading this name. The covariate-effect parameter form is e_drusen_<param>.

DIS_CHF_NYHA2 (canonical for NYHA class II heart-failure indicator)

• Description: 1 = subject has mild chronic heart failure classified as New York Heart Association (NYHA) functional class II, 0 = otherwise. Time-fixed per subject. Member of the mutually-exclusive DIS_CHF_NYHA2 / DIS_CHF_NYHA3 / DIS_CHF_NYHA4 trio prescribed by the DIS_CHF entry’s Notes for papers that retain NYHA strata in the final model; all three = 0 selects the non-heart-failure reference. NYHA class I is not given its own indicator because published HF pharmacokinetic cohorts almost never separate it from the healthy reference.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no heart failure, when DIS_CHF_NYHA2, DIS_CHF_NYHA3 and DIS_CHF_NYHA4 are all 0).
• Source aliases:
  • HF-II – used in Gu_2025_*_pbpk.R (Gu 2025 Table 1 and Table S3 heart-failure grade column).
• Example models: Gu_2025_digoxin_pbpk.R.
• Notes: Distinct from DIS_CHF, which is an unstratified compensated-CHF indicator; use DIS_CHF when the source treats heart failure as a single binary and this trio when it resolves NYHA classes. Distinct from the paediatric SCORE_ROSS severity instrument (NYHA is adult-only) and from DIS_HF_OR_LF_SEV (which pools severe heart failure with severe liver failure). The covariate-effect parameter form is e_nyha2_<param>, dropping the DIS_CHF_ prefix per the DIS_CANCER -> e_cancer_<param> precedent.

DIS_CHF_NYHA3 (canonical for NYHA class III heart-failure indicator)

• Description: 1 = subject has moderate chronic heart failure classified as New York Heart Association (NYHA) functional class III, 0 = otherwise. Time-fixed per subject. Member of the mutually-exclusive DIS_CHF_NYHA2 / DIS_CHF_NYHA3 / DIS_CHF_NYHA4 trio prescribed by the DIS_CHF entry’s Notes for papers that retain NYHA strata in the final model; all three = 0 selects the non-heart-failure reference. NYHA class I is not given its own indicator because published HF pharmacokinetic cohorts almost never separate it from the healthy reference.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no heart failure, when DIS_CHF_NYHA2, DIS_CHF_NYHA3 and DIS_CHF_NYHA4 are all 0).
• Source aliases:
  • HF-III – used in Gu_2025_*_pbpk.R (Gu 2025 Table 1 and Table S3 heart-failure grade column).
• Example models: Gu_2025_digoxin_pbpk.R.
• Notes: Distinct from DIS_CHF, which is an unstratified compensated-CHF indicator; use DIS_CHF when the source treats heart failure as a single binary and this trio when it resolves NYHA classes. Distinct from the paediatric SCORE_ROSS severity instrument (NYHA is adult-only) and from DIS_HF_OR_LF_SEV (which pools severe heart failure with severe liver failure). The covariate-effect parameter form is e_nyha3_<param>, dropping the DIS_CHF_ prefix per the DIS_CANCER -> e_cancer_<param> precedent.

DIS_CHF_NYHA4 (canonical for NYHA class IV heart-failure indicator)

• Description: 1 = subject has severe chronic heart failure classified as New York Heart Association (NYHA) functional class IV, 0 = otherwise. Time-fixed per subject. Member of the mutually-exclusive DIS_CHF_NYHA2 / DIS_CHF_NYHA3 / DIS_CHF_NYHA4 trio prescribed by the DIS_CHF entry’s Notes for papers that retain NYHA strata in the final model; all three = 0 selects the non-heart-failure reference. NYHA class I is not given its own indicator because published HF pharmacokinetic cohorts almost never separate it from the healthy reference.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no heart failure, when DIS_CHF_NYHA2, DIS_CHF_NYHA3 and DIS_CHF_NYHA4 are all 0).
• Source aliases:
  • HF-IV – used in Gu_2025_*_pbpk.R (Gu 2025 Table 1 and Table S3 heart-failure grade column).
• Example models: Gu_2025_digoxin_pbpk.R.
• Notes: Distinct from DIS_CHF, which is an unstratified compensated-CHF indicator; use DIS_CHF when the source treats heart failure as a single binary and this trio when it resolves NYHA classes. Distinct from the paediatric SCORE_ROSS severity instrument (NYHA is adult-only) and from DIS_HF_OR_LF_SEV (which pools severe heart failure with severe liver failure). The covariate-effect parameter form is e_nyha4_<param>, dropping the DIS_CHF_ prefix per the DIS_CANCER -> e_cancer_<param> precedent.

Epilepsy baseline seizure-severity indicators

Baseline seizure-severity indicators derived from a pre-trial seizure count (typically the number of seizures in the 3 months before study entry, decomposed by the source paper into severity bins). The count itself is decomposed into a set of mutually-exclusive binary indicators per the maintainers’ standing policy for count covariates (“count covariate -> decomposed binary indicators, not a single integer count”). Members are jointly restricted to at most one being = 1 for any subject; the reference category is 2-6 seizures in the previous 3 months (all indicators = 0 selects the reference).

NSP3M_LT2 (canonical for baseline seizure count fewer than 2 in the previous 3 months)

• Description: Binary indicator that the subject reported fewer than 2 seizures in the 3 months before trial start (baseline period). 1 = <2 baseline seizures (the low-severity bin), 0 = otherwise. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (2-6 baseline seizures per the Lindauer 2017 NSP3M decomposition; the reference bin). Subjects in the 7-50 or >50 bins have NSP3M_LT2 = 0 and are identified by NSP3M_7_50 = 1 or NSP3M_GT50 = 1 instead.
• Source aliases:
  • NSP3M (categorical, the <2 bin) – used in Lindauer_2017_lacosamide_seizure.R (Lindauer 2017 Section 2.5 splits NSP3M into four categories: <2, 2-6, 7-50, >50, with 2-6 as reference; the low-severity bin is registered here as NSP3M_LT2).
• Example models: Lindauer_2017_lacosamide_seizure.R (log-hazard shifts on both first-seizure and subsequent-seizure Weibull scale: e_nsp3m_lt2_1st = -1.12 (first event), e_nsp3m_lt2_2nd = -1.37 (subsequent events); the associated hazard ratios vs the 2-6 reference are 0.58 and 0.38 per Lindauer 2017 Table 4).
• Notes: Specific scope because the concept is tied to the pre-trial-seizure-count decomposition used in the SP0993 / N01061 monotherapy epilepsy trials (Lindauer 2017 references Abrantes et al. for the two-Weibull-sub-model approach with NSP3M-derived binary categories). Sibling canonicals NSP3M_7_50 and NSP3M_GT50 complete the 4-level decomposition. Data assemblers must enforce mutual exclusivity: at most one of {NSP3M_LT2, NSP3M_7_50, NSP3M_GT50} is 1 for each subject; if all three are 0 the subject belongs to the reference 2-6 bin. The covariate-effect parameter form is e_nsp3m_lt2_<param> (drops _LT2 structure into the coefficient name).

NSP3M_7_50 (canonical for baseline seizure count 7 to 50 in the previous 3 months)

• Description: Binary indicator that the subject reported 7 to 50 seizures in the 3 months before trial start (baseline period). 1 = 7-50 baseline seizures (the moderate-high-severity bin), 0 = otherwise. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (2-6 baseline seizures per the Lindauer 2017 NSP3M decomposition; the reference bin).
• Source aliases:
  • NSP3M (categorical, the 7-50 bin) – used in Lindauer_2017_lacosamide_seizure.R.
• Example models: Lindauer_2017_lacosamide_seizure.R (log-hazard shifts on both first-seizure and subsequent-seizure Weibull scale: e_nsp3m_7_50_1st = +1.94 (first event), e_nsp3m_7_50_2nd = +1.36 (subsequent events); the associated hazard ratios vs the 2-6 reference are 2.60 and 2.63 per Lindauer 2017 Table 4).
• Notes: Specific scope. Sibling of NSP3M_LT2 and NSP3M_GT50 in the 4-level NSP3M decomposition; see NSP3M_LT2 notes for the mutual-exclusivity rule and covariate-effect naming. The Lindauer 2017 Table 4 hazard ratio of 2.60 (90% CI 2.02-3.31) for the first-seizure hazard was highlighted in the paper Abstract as the primary NSP3M-severity finding.

NSP3M_GT50 (canonical for baseline seizure count greater than 50 in the previous 3 months)

• Description: Binary indicator that the subject reported more than 50 seizures in the 3 months before trial start (baseline period). 1 = >50 baseline seizures (the highest-severity bin), 0 = otherwise. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (2-6 baseline seizures per the Lindauer 2017 NSP3M decomposition; the reference bin).
• Source aliases:
  • NSP3M (categorical, the >50 bin) – used in Lindauer_2017_lacosamide_seizure.R.
• Example models: Lindauer_2017_lacosamide_seizure.R (log-hazard shifts on both first-seizure and subsequent-seizure Weibull scale: e_nsp3m_gt50_1st = +3.30 (first event), e_nsp3m_gt50_2nd = +2.53 (subsequent events); the associated hazard ratios vs the 2-6 reference are 5.09 and 6.09 per Lindauer 2017 Table 4).
• Notes: Specific scope. Sibling of NSP3M_LT2 and NSP3M_7_50 in the 4-level NSP3M decomposition; see NSP3M_LT2 notes for the mutual-exclusivity rule and covariate-effect naming.

Pulmonary / lung-disease biomarkers

LPAC182_BL (canonical for baseline plasma lysophosphatidic acid C18:2)

• Description: Per-subject baseline (pre-dose) plasma lysophosphatidic acid C18:2 (LPA C18:2, the most abundant plasma LPA species: an 18-carbon fatty-acid side chain with two unsaturated bonds). Time-fixed per subject. Used both as the baseline of a turnover-free inhibitory PD model and as a covariate on its potency and efficacy.
• Units: peak-area ratio (LPA C18:2 / LPA C17:0 internal standard); document per model via covariateData[[LPAC182_BL]]$units. The Taneja 2019 LC-MS/MS assay is non-quantitative because of endogenous LPA, so the value is a unitless ratio, not a molar concentration.
• Type: continuous
• Scope: specific
• Reference category: n/a – used with power scaling (LPAC182_BL / ref)^exponent and linear-deviation forms e * (LPAC182_BL - ref). Reference value observed: 0.36 (Taneja 2019 Sect. 2.3.9, the typical baseline of its typical-patient simulations).
• Source aliases:
  • LPA C18:2 BL, LPA C18:2,BL – Taneja 2019 Table 2, Table 3 and the Sect. 3.2 display equation.
• Example models: Taneja_2019_glpg1690.R (sets the PD baseline lpaC182 <- LPAC182_BL * (1 - imax * ceff / (ceff + ic50)), and enters as a power covariate on IC50, (LPAC182_BL / 0.36)^-0.473, and linearly on the probit of Imax, 0.5805 * (LPAC182_BL - 0.36); observed range 0.104-1.305 across studies, Table 2).
• Notes: Follows the <ANALYTE>_BL baseline shape ([[D25OH_BL]], [[GLN_BL]], [[PKK_BL]], [[FERRITIN_BL]]). Named for the specific lipid species because the assay measures only LPA C18:2; a model driven by total LPA or by another species (C16:0, C18:1, C20:4) needs its own canonical. Because the value is an assay-specific peak-area ratio, it is not transferable between laboratories without recalibration.

FEV1 (canonical for forced expiratory volume in 1 second)

• Description: Baseline forced expiratory volume in 1 second (FEV1) reported as an absolute volume in litres. Pulmonary-function spirometry endpoint; reflects large-airway airflow obstruction and is a standard covariate in chronic obstructive pulmonary disease, alpha-1 antitrypsin deficiency, asthma, and cystic-fibrosis disease-progression analyses. Distinct from FEV1 percent-predicted (the % predicted value standardises absolute FEV1 by reference equations from sex / age / height / ethnicity); use this canonical only for the absolute-litre value supplied as a covariate column.
• Units: L (absolute volume; document the exhalation-effort standard the source paper cites in covariateData[[FEV1]]$notes if non-default – ATS / ERS post-bronchodilator is the typical convention).
• Type: continuous
• Scope: general
• Reference category: n/a – used with linear-deviation forms (theta + e_fev1_param * (FEV1 - ref)) or power forms (FEV1 / ref)^exponent. Reference values observed: 1.6 L (Tortorici 2017, population median across the RAPID-RCT/RAPID-OLE A1-PI augmentation cohort); 2 L (Bartels 2021 Table 3, asthma).
• Source aliases: none yet; canonical name preferred. Source papers typically use the abbreviation FEV1 directly.
• Example models: Tortorici_2017_a1pi.R (linear-deviation effect on the lung-density decline rate: theta5 * (FEV1 - 1.6) with theta5 = +0.56 (g/L/year per L FEV1); lower-FEV1 patients have steeper natural decline rates independent of A1-PI exposure), Bartels_2021_mometasoneFuroate.R (power-form effects of baseline FEV1 on the apparent clearance and central volume of inhaled mometasone furoate in asthma, (FEV1 / 2)^-0.17 on CL/F and (FEV1 / 2)^-0.23 on Vc/F, normalised to the 2 L reference of Bartels 2021 Table 3).
• Notes: Distinct from FEV1 percent-predicted (which is a derived ratio with the reference-equation denominator built in; FEV1% is the outcome variable in Harun_2019_cysticFibrosis.R rather than a covariate column). Use FEV1 only when the source paper supplies the absolute-volume value; if the source supplies a percent-predicted value as a covariate, the canonical for that surface is FEV1_PCTPRED. A model that also OUTPUTS FEV1 must use [[FEV1_BL]] instead of this name – FEV1 is simultaneously a registered output-state canonical in inst/references/compartment-names.md, and a covariate column cannot share a name with the state it drives. Scope promoted to general by Bartels_2021_mometasoneFuroate.R, the second model to use the absolute-litre semantics (baseline FEV1 as a covariate on inhaled-drug PK in asthma, a different disease and a different parameter from the founding example).

FEV1_BL (canonical for the patient’s own baseline FEV1 supplied as a covariate to a model whose output is FEV1)

• Description: The patient’s own pre-treatment (baseline) absolute forced expiratory volume in 1 second, in litres, supplied as an input covariate column to a model whose output state is also FEV1. Semantically identical to the [[FEV1]] covariate – same measurement, same units, same time-fixed per-subject shape – and differing only in that the bare name is unavailable because FEV1 is simultaneously a registered output-state canonical (inst/references/compartment-names.md). Airways-disease longitudinal and landmark PD models routinely scale or offset their predicted FEV1 by the patient’s own baseline, so this collision is structural rather than incidental. Record which spirometric manoeuvre the baseline came from (pre- or post-bronchodilator, trough or peak) in covariateData[[FEV1_BL]]$notes; the two are not interconvertible without a reversibility assumption.
• Units: L (absolute volume)
• Type: continuous
• Scope: specific
• Reference category: n/a – entered normalised to a population central value in the founding model, CON * (FEV1_BL / 1.31) multiplying the zero-dose intercept. Reference values observed: 1.31 L (Ambery 2015, the Table 1 mean of the n = 347 PD analysis set, substituted for the unreported median that Eq. 2 actually normalises by). Linear-deviation and power forms are equally admissible; the functional form is per-paper and is not a property of this column.
• Source aliases:
  • Baseline FEV1 on Day 1 – used in Ambery_2015_batefenterol_fev1.R (Ambery 2015 Eq. 2 and Table 1; the day-1 pre-dose trough value, defined in Sect. 2.3 as a Vitalograph spirometry measurement).
  • FEVB / FEV1baseline – paper notation for the fitted baseline parameter of the same endpoint rather than an input column; when a source estimates the baseline it is a model parameter (lrbase), NOT this covariate. Jin_2025_benralizumab_fev1.R and Zhang_2025_dupilumab_fev1.R both take that route and carry their baseline FEV1 in covariatesDataExcluded.
• Example models: Ambery_2015_batefenterol_fev1.R (founding example; the sole retained covariate of a day-29 trough-FEV1 Emax dose-response in moderate-to-severe COPD, scaling the zero-dose intercept as e0 * CON * (FEV1_BL / 1.31) with CON = 19.0, so a patient one standard deviation above the cohort mean baseline is predicted about 35% higher at every dose).
• Notes: Use this name, not the bare [[FEV1]] covariate, whenever the model’s observation variable is the FEV1 output state; use bare [[FEV1]] when baseline FEV1 is a covariate on something else (Tortorici_2017_a1pi.R puts it on a lung-density decline rate). Distinct from [[FEV1_PCTPRED]], which is the same measurement divided by a reference-equation predicted value and is therefore dimensionless and not interconvertible without that equation, and from [[FEV1_PBD_ANCHOR]], which addresses a different problem (studies whose absolute FEV1 had to be reconstructed from a post-bronchodilator baseline plus a change from baseline). A baseline-suffix member of the established <MEASURE>_BL family ([[QTC_BL]], [[MAP_BL]], [[SBP_BL]], [[HR_BL]], [[DBP_BL]], [[HGB_BL]], [[OFFTIME_BL]], [[BMD_BL]], [[INS_BL]]), every one of which exists for the same reason: the baseline of a measurement that the model also predicts. Scope: specific until a second airways model ratifies the semantics; promote to general at that point.

FEV1_PCTPRED (canonical for baseline FEV1 as percent of the predicted value)

• Description: Baseline forced expiratory volume in 1 second expressed as a percent of the sex / age / height / ethnicity reference-equation predicted value (FEV1% predicted). The percent-predicted scaling normalises the absolute-litre FEV1 measurement against a healthy-reference standard so a single covariate value is comparable across patient ages and body sizes. Standard pulmonary-function covariate in cystic-fibrosis disease-severity and inhaled-therapy popPK analyses. Time-fixed at study entry / baseline.
• Units: % predicted (numeric percentage, e.g. 62.1 for the Ting 2014 population median; not a fraction 0.621).
• Type: continuous
• Scope: specific
• Reference category: n/a – used with power-form effects (FEV1_PCTPRED / ref)^exponent or linear-deviation forms (1 + e * (FEV1_PCTPRED - ref)). Reference values observed: 62.1 % (Ting 2014, population median across the combined three-study cystic-fibrosis cohort).
• Source aliases:
  • FEV1% predicted – used in Ting_2014_tobramycin_inhaled.R (paper Table 1 / Table 2). Free-text label, not a typical NMTRAN column header; downstream data sets should use the canonical column name FEV1_PCTPRED.
• Example models: Ting_2014_tobramycin_inhaled.R (power-form effect on apparent central volume of distribution: (FEV1_PCTPRED / 62.1)^-0.303; lower lung-function patients have larger apparent Vd/F, consistent with the paper’s hypothesis that worsening lung disease increases central-airway aerosol deposition).
• Notes: Distinct from absolute-litre FEV1 (canonical for the unstandardised volume) and from FEV1 as an outcome variable in disease-progression models (e.g. Harun_2019_cysticFibrosis.R, where percent-predicted FEV1 is the dependent variable rather than a covariate). The reference equation used to derive the percent-predicted value is paper-specific; document any non-default reference standard (Hankinson 1999 NHANES III, GLI 2012, Wang 1993, etc.) in covariateData[[FEV1_PCTPRED]]$notes per model. Scope: specific until a second model ratifies the percent-predicted semantics; promote to general at that point.

FVC_PCTPRED (canonical for baseline FVC as percent of the predicted value)

• Description: Baseline forced vital capacity expressed as a percent of the sex / age / height / ethnicity reference-equation predicted value (FVC% predicted). The percent-predicted scaling normalises the absolute-litre FVC measurement against a healthy-reference standard so a single covariate value is comparable across patient ages and body sizes. The standard measure of disease severity in idiopathic pulmonary fibrosis and other fibrosing interstitial lung diseases – the LOWER the value, the more severe the pulmonary-function impairment. Time-fixed at study entry / baseline. In a model-based meta-analysis the value is the study ARM’s mean or median baseline rather than an individual patient’s.
• Units: % predicted (numeric percentage, e.g. 76.2 for the Chan 2017 CAPACITY-2 arms; not a fraction 0.762).
• Type: continuous
• Scope: specific
• Reference category: n/a – used with power-form effects (FVC_PCTPRED / ref)^exponent or linear-deviation forms (1 + e * (FVC_PCTPRED - ref)). Reference values observed: 74 % (Chan 2017, the approximate median arm-level baseline across the 20-trial IPF meta-analysis dataset).
• Source aliases:
  • arm level baseline %predicted FVC – used in Chan_2017_ipf_fvc_mbma.R (Chan 2017 Table 2 note). Free-text label, not a column header; downstream data sets should use the canonical column name FVC_PCTPRED.
  • %predicted FVC, FVC% predicted, FVCpp (as a covariate).
• Example models: Chan_2017_ipf_fvc_mbma.R (power-form effect scaling the whole treatment effect rather than a PK parameter or a baseline: (FVC_PCTPRED / 74)^1.48, so an arm with higher baseline lung function is predicted to show a LARGER placebo-corrected treatment effect. The only covariate retained out of six prespecified ones; the paper imputed it by multivariate regression where unreported, and derived it from absolute FVC via the Crapo 1981 reference equation for the three nintedanib trials).
• Notes: Second member of the percent-predicted spirometry surface established by FEV1_PCTPRED, and added as a member of that surface rather than as a newly-minted concept: the construct, units, type and functional forms are identical and only the spirometric volume differs (vital capacity rather than one-second expiratory volume), exactly the relationship the output-side register already encodes between the registered fvcpp and fev1pp states. Should a third member appear, promote the pair to an explicit <MEASURE>_PCTPRED family heading. Distinct from the OUTPUT-side canonicals fvcpp (absolute FVC percent predicted as a modelled state) and fvcppcfb (its change from baseline) in inst/references/compartment-names.md – use FVC_PCTPRED only for the baseline value supplied as an input column, and note that a model may legitimately carry both, as Chan_2017_ipf_fvc_mbma.R does. Also distinct from an absolute-litre FVC, which has no canonical yet; register one following the FEV1 precedent if a model needs it. The reference equation used to derive the percent-predicted value is paper-specific; document any non-default standard (Crapo 1981, Hankinson 1999 NHANES III, GLI 2012) in covariateData[[FVC_PCTPRED]]$notes per model. Scope: specific until a second model ratifies the semantics; promote to general at that point.

A1PI (canonical for serum alpha-1 proteinase inhibitor concentration)

• Description: Baseline serum alpha-1 proteinase inhibitor (A1-PI; also known as alpha-1 antitrypsin, AAT) concentration. Used in alpha-1 antitrypsin deficiency (AATD) augmentation-therapy modelling as a per-subject pre-treatment exposure covariate (the subject’s endogenous A1-PI level at study entry, before any augmentation infusions). Time-fixed per subject. Distinct from a time-course of A1-PI used as a state variable – when the source paper carries A1-PI as the dynamic dependent variable (the augmentation model’s PD output), use Cc rather than A1PI.
• Units: umol/L (typical SI-convention reporting; also reported as mg/dL in US-convention papers – document the unit used in each model via covariateData[[A1PI]]$units). Conversion: 1 umol/L A1-PI ~= 5.2 mg/dL (using MW ~52 kDa). The 11 umol/L “putative protective threshold” used clinically corresponds to ~57 mg/dL.
• Type: continuous
• Scope: specific
• Reference category: n/a – used with power-form effects (A1PI / ref)^exponent on the post-treatment exposure intercept and slope. Reference values observed: 5.5 umol/L (Tortorici 2017, approximate median pre-treatment A1-PI among RAPID-RCT placebo-randomised patients, used as the normalisation denominator for power-form covariate effects).
• Source aliases: Cbase – used in Tortorici 2017’s published equation 6 to denote the baseline pre-treatment A1-PI value; the column name in the modelled dataset would be A1PI.
• Example models: Tortorici_2017_a1pi.R (two power-form effects: (A1PI/5.5)^theta5 with theta5 = +0.73 on the placebo-arm post-treatment exposure intercept, and (A1PI/5.5)^theta4 with theta4 = -0.12 on the dose-rate slope; together they encode the modest dose-exposure dependence on each subject’s endogenous A1-PI level).
• Notes: AATD enrolment criteria typically restrict A1PI to <= 11 umol/L (severe deficiency); reference / heterozygous PI*MZ phenotypes have higher levels. Document the source paper’s AATD-genotype enrolment criteria in covariateData[[A1PI]]$notes per model. Specific scope because the canonical is tied to AATD augmentation-therapy modelling; future PK / PD analyses of A1-PI (or AAT) in non-AATD contexts (acute-phase response, smoking-induced inflammation) should ratify general scope at that time.

Cystic fibrosis lung-disease indicators

AIR_TRAP_5Y (canonical for severe air trapping on chest HRCT scan at age 5 years)

• Description: 1 = subject had a non-zero “air trapping” component score on the validated Brody-II chest high-resolution computed tomography (HRCT) scan performed at age 5 years; 0 = air trapping component score of 0 (absent). Time-fixed per subject (the indicator captures the single end-of-study HRCT performed at age 5 in the Australasian Cystic Fibrosis Bronchoalveolar Lavage (ACFBAL) study). The Brody-II scoring system reports air trapping as a percentage of maximum possible HRCT score; the binary “present vs absent” dichotomy follows the same convention as Rosenow 2015 (Am J Respir Crit Care Med 191:1158-65).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no air trapping detected on the age-5 HRCT scan).
• Source aliases:
  • ATS5C – used in Harun_2019_cysticFibrosis.R (Harun 2019 NMTRAN $INPUT column, “presence of air trapping at age 5”; values 0 = absent, 1 = present).
• Example models: Harun_2019_cysticFibrosis.R (linear-deviation effect on baseline FEV1% predicted at age 5: (1 + e_at_baseline * AIR_TRAP_5Y) with coefficient -0.0417, i.e., subjects with severe air trapping at age 5 have a baseline FEV1% predicted approximately 4.17% lower than those without).
• Notes: Specific scope because the indicator is tied to the ACFBAL study’s standardised HRCT-at-age-5 protocol and the Brody-II scoring rubric; future paediatric CF lung-disease studies that score HRCT at a different age or use a different scoring system should register a separate canonical (AIR_TRAP_8Y, AIR_TRAP_PRAGMA, etc.) rather than overload this name.

HOSPRA (canonical for hospitalisation due to a pulmonary exacerbation)

• Description: Time-varying binary indicator of inpatient hospitalisation for management of a pulmonary exacerbation at the time of the FEV1% predicted measurement: 1 = subject is hospitalised because of a pulmonary exacerbation when the spirometry value is recorded, 0 = not hospitalised at that visit. Used as a per-visit covariate in disease-progression models of FEV1% decline in cystic fibrosis where pulmonary exacerbations are tracked from the Australian Cystic Fibrosis Data Registry (ACFDR) inpatient records.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (not hospitalised for a pulmonary exacerbation at the time of the FEV1% measurement).
• Source aliases:
  • HOSPRA – used in Harun_2019_cysticFibrosis.R (Harun 2019 NMTRAN $INPUT column, “hospitalisation at the time of FEV1% predicted measurement; 0=no, 1=yes”).
• Example models: Harun_2019_cysticFibrosis.R (linear-deviation effects on the disease-progression maximum drop and on the half-effect age: (1 + e_hpe_dmax * HOSPRA) with coefficient -0.22 on the maximum FEV1% drop and (1 + e_hpe_t50max * HOSPRA) with coefficient -0.235 on the age at which 50% of the maximum drop occurs; hospitalised visits accelerate both the magnitude and the onset of FEV1% decline).
• Notes: Specific scope because the canonical encoding pools all pulmonary exacerbation-driven hospitalisations into a single binary regardless of severity, duration, or treatment intensity; future CF / chronic-respiratory-disease studies that need to distinguish exacerbation severity (e.g., requiring intravenous antibiotics vs oral) should register a finer-grained canonical.

Infectious disease

LNPC (canonical for log-transformed admission Plasmodium parasitaemia)

• Description: Natural logarithm of the asexual Plasmodium parasite count (parasites per microlitre of blood) at study admission. Time-fixed per subject (one value per subject, captured at enrolment).
• Units: log(parasites/uL)
• Type: continuous
• Scope: specific
• Reference category: n/a – used with linear-deviation forms (1 + e * (LNPC - ref)). Reference values observed: 5.88 log(parasites/uL) (Birgersson 2019, population median in the pooled pregnant + non-pregnant Burkina Faso cohort).
• Source aliases: none formally; companion column PARA (raw asexual parasite count per microlitre) is provided alongside LNPC in the Birgersson 2019 NONMEM dataset but the model uses LNPC = log(PARA) as the active covariate.
• Example models: Birgersson_2019_artesunate.R (linear-deviation effect on relative bioavailability F1: F1LNPC = 1 + e_lnpc_f * (LNPC - 5.88); positive coefficient e_lnpc_f = +0.138 per unit increase in log-parasite-count, reflecting increased oral artesunate bioavailability with higher parasite burden).
• Notes: Disease-severity covariate specific to malaria PK models. Higher parasitaemia is a marker of more severe acute malaria infection and has been associated in the source publication with altered oral bioavailability of artesunate (presumably via gut-mucosal / first-pass effects of the febrile parasitised state). Scope: specific because the canonical reference value (5.88) is the Birgersson 2019 cohort median; future malaria-in-pregnancy or malaria-in-children PK models may legitimately reuse LNPC but should document their own cohort-specific reference value in covariateData[[LNPC]]$notes. Distinct from PARA (raw parasitaemia in parasites/uL), which is the companion canonical for models that apply the log transform inside model() rather than pre-computing it at dataset-assembly time, and that use a time-varying (last-observation-carried-forward) parasitaemia trajectory rather than the admission-only fixed value.

PARA (canonical for raw Plasmodium parasitaemia (parasites/uL))

• Description: Asexual Plasmodium parasite count per microlitre of blood. May be admission-only (time-fixed) or time-varying with serial counts across follow-up (last-observation-carried-forward is the typical NONMEM idiom in this lab’s malaria models). Document time-varying vs time-fixed mode in covariateData[[PARA]]$notes per model.
• Units: parasites/uL
• Type: continuous
• Scope: specific
• Reference category: n/a – used with log-transformed linear forms (1 + e * log10(PARA)) or (1 + e * (log10(PARA) - ref)) when the source paper applies the log transform inside the model. Reference values observed: implicit log10(PARA) = 0 (i.e. PARA = 1 parasite/uL) in Kloprogge 2014 (the linear form anchors F at the typical population estimate when PARA = 1).
• Source aliases:
  • PARA – used in Kloprogge_2014_quinine.R (raw parasitaemia in parasites/uL, time-varying via last-observation-carried-forward across the 7-day quinine treatment course), in Kloprogge_2018_lumefantrine.R (admission-only / time-fixed parasitaemia, centered at the model-building geometric-mean 15,800 parasites/uL = log10 4.2), in Tarning_2012_dihydroartemisinin.R (admission-only / time-fixed parasitaemia, centered at the typical-patient log10(PARA) = 3.98 reported in the Table 4 footnote of Tarning 2012 AAC), and in Hien_2017_cipargamin.R (admission-only / time-fixed parasitaemia in parasites/uL, used to seed the initial-condition split of the two-population parasite clearance PD model into sensitive and refractory subpools).
• Example models: Kloprogge_2014_quinine.R (linear effect on relative bioavailability with the log10 transform applied inside model(): F1PARA = 1 + e_para_f * log10(max(PARA, 1)) with e_para_f = +0.389 per log10 parasitaemia; gated by max(PARA, 1) so that PARA values below 1 parasite/uL (effectively zero / below detection) collapse to no covariate effect, matching the source paper’s “effect only during the acute phase when parasitaemia was above the limit of detection” wording), Tchaparian_2016_lumefantrine_recurrence.R (adjustment covariate of an adjusted Cox model on 28-day recurrent parasitaemia, entering the log-hazard as e_para_haz * (log(max(PARA, 1)) - log(17603)) with e_para_haz = log(1.04) per log unit, centred on the cohort geometric mean. Note the contrast with the Kloprogge entry above on the point this column most often goes wrong: that paper names log10 explicitly, whereas Tchaparian 2016 write only “parasite density (log transformed)” and the NATURAL log is inferred from the base the same paper names explicitly for its concentration analyses – so the transform base must be read off each source individually and never carried over from a sibling model).
  • Kloprogge_2014_quinine.R – linear effect on relative bioavailability with the log10 transform applied inside model(): F1PARA = 1 + e_para_f * log10(max(PARA, 1)) with e_para_f = +0.389 per log10 parasitaemia; gated by max(PARA, 1) so that PARA values below 1 parasite/uL (effectively zero / below detection) collapse to no covariate effect, matching the source paper’s “effect only during the acute phase when parasitaemia was above the limit of detection” wording.
  • Kloprogge_2018_lumefantrine.R – exponential effect on relative bioavailability with the log10 transform applied inside model(), centered on log10(15,800) = 4.2: F1PARA = exp(e_lnpc_f * (log10(max(PARA, 1)) - 4.2)) with e_lnpc_f = -0.643 per log10 parasitaemia (higher pre-treatment parasitaemia is associated with lower relative bioavailability, consistent with reduced visceral blood flow in more severe disease). Same max(PARA, 1) gating convention as Kloprogge 2014.
  • Tarning_2012_dihydroartemisinin.R – linear-deviation effect on relative bioavailability with the log10 transform applied inside model(), centered on log10(PARA) = 3.98 (the pooled-cohort median in the Thai-Myanmar-border dihydroartemisinin-piperaquine combination trial): fpara = 1 + e_para_f * (log10(max(PARA, 1)) - 3.98) with e_para_f = +0.278 per log10 unit (Results: ‘27.8% linear increase per unit logarithmic parasitemia’). Higher pre-treatment parasitaemia is associated with higher relative bioavailability of dihydroartemisinin (Tarning 2012 Discussion: ‘…consistent with a disease-related decrease in first-pass metabolism possibly compounded by reduced hepatic blood flow…’). Same max(PARA, 1) gating convention as the Kloprogge models; admission-only / time-fixed.
  • Ding_2024_piperaquine.R – linear-deviation effect on relative bioavailability with the log10 transform applied inside model(), centered on log10(PARA) = 2.83, the rounded log10 of the cohort median of 680 parasites/uL in the PREGACT African pregnant-women cohort (Table 1), giving an effective reference of 10^2.83 = 676 parasites/uL: f_para = 1 + e_para_f * (log10(max(PARA, 1)) - 2.83) with e_para_f = -0.119 per log10 unit (Table 3: ‘Baseline parasites count on F (%) = -11.9’; the Table 3 footnote prints the form [1 + theta x (log(parasitemia) - 2.83)]). Higher pre-treatment parasitaemia is associated with LOWER relative bioavailability of piperaquine – the opposite sign to the Tarning_2012_dihydroartemisinin.R entry above, and consistent with Hoglund 2017 (Ding 2024 Discussion: ‘relative bioavailability was correlated to disease severity’). The base of the logarithm is pinned to 10 by the paper itself: the Results simulations use ‘a baseline parasitemia of 676 parasites/uL’, which is 10^2.83 = 676.1. Same max(PARA, 1) gating convention as the Kloprogge and Tarning models (the Table 1 cohort minimum is 0 parasites/uL); admission-only / time-fixed.
  • Ding_2026_lumefantrine.R – linear-deviation effect on relative bioavailability with the log10 transform applied inside model(), centered on log10(PARA) = 4.56: f_para = 1 + e_para_f * (log10(max(PARA, 1)) - 4.56) with e_para_f = -0.116 per log10 unit (Table 4: ‘Baseline parasite density on F (%) = -11.6’; the Table 4 footnote prints the form [1 + (theta x (log(PBM)-4.56)]). Same negative direction and same Mahidol-Oxford idiom as Ding_2024_piperaquine.R by the same first author. The base of the logarithm is pinned to 10 by the centring constant itself: 10^4.56 = 36,300 parasites/uL sits inside the pooled cohort’s per-trial medians of 14,390-52,500 parasites/uL (Table 1), whereas exp(4.56) = 96 parasites/uL would be far below every reported median. Admission-only / time-fixed; same max(PARA, 1) gating convention as the Kloprogge and Tarning models.
  • Hien_2017_cipargamin.R – admission-only / time-fixed parasitaemia in parasites/uL (enrolment inclusion criterion 5,000-50,000 asexual parasites/uL per Hien 2017 Methods ‘Patients’). Not used as a bioavailability covariate; instead PARA seeds the parasite ODE compartments at simulation start: parasite_sensitive(0) = fsen * PARA and parasite_refractory(0) = (1 - fsen) * PARA. The ODE states carry the same parasites/uL units as PARA (no total-burden conversion applied inside the model), matching the Fig 3 individually-predicted-clearance-profile reporting scale.
• Notes: Companion to LNPC – both canonicals describe the same biological quantity (asexual Plasmodium parasite count) but differ in the transform location: LNPC is pre-transformed to natural log at dataset-assembly time and used as (LNPC - ref), whereas PARA is the raw count with the log transform applied inside model(). Use PARA when the source paper reports the covariate coefficient on the log10 scale (per log10 parasitaemia) and / or when parasitaemia is time-varying. Use LNPC when the source dataset pre-computes the natural log at admission. Scope: specific because the gating-at-1 convention and the log10-inside-model() idiom reflect the Mahidol-Oxford malaria popPK lab’s specific implementation; a future malaria PK model that uses a different gating threshold (e.g., LOQ-aware, or log base-e inside the model) or a different reference value should document its own convention in covariateData[[PARA]]$notes.

CT_TCRUZI_BASE (canonical for baseline Trypanosoma cruzi qPCR cycle threshold)

• Description: Pre-treatment (screening visit) cycle threshold from quantitative real-time PCR for T. cruzi satellite DNA in whole blood, reported as the mean of nine measurements (three venous blood samples, each assayed in triplicate). Time-fixed per subject. The scale is inverse to parasite burden: the cycle threshold counts the amplification cycles needed to cross the detection threshold, so a HIGHER Ct means less starting template, i.e. a LOWER circulating parasite density. Samples with no amplification within the assay’s cycle budget are reported at the ceiling value (40 in Assmus 2025).
• Units: (cycles)
• Type: continuous
• Scope: specific
• Reference category: n/a – used as a raw linear predictor on a logit scale. Reference values observed: 37.7 cycles (Assmus 2025, BENDITA mITT median; range 30.2-40.0), used to recover the exposure-response intercept from Assmus 2025 Fig 8.
• Source aliases:
  • Ct – used in Assmus_2025_benznidazole_qpcr.R (Assmus 2025 Table 1 “Baseline Ct”; Table 3 “Odds ratio, Ct value”).
• Example models: Assmus_2025_benznidazole_qpcr.R (additive shift on the qPCR-positivity logit: pqpcr = expit(b0_qpcr + e_auc_logit * auc_central + e_ct_tcruzi_base_logit * CT_TCRUZI_BASE) with e_ct_tcruzi_base_logit = log(0.9021) = -0.1030 per cycle; Assmus 2025 Table 3, mITT including placebo, AUCinf column).
• Notes: Because the Ct scale is inverted relative to a parasite count, the sign of any coefficient on this canonical reads opposite to the corresponding coefficient on a burden-scaled canonical: a NEGATIVE CT_TCRUZI_BASE coefficient on an event logit means that patients with a HIGHER parasite load (lower Ct) are more likely to experience the event. Always state the direction explicitly in covariateData[[CT_TCRUZI_BASE]]$notes; it is the single easiest sign error to make with this column. Do NOT convert a Ct into a parasite density to reuse [[PARA]] or [[LNPC]] – the conversion requires an assay-specific standard curve that clinical papers rarely publish, and both of those canonicals are Plasmodium-specific in any case. Ct values are only comparable within one assay and one laboratory, so a model reusing this canonical must record the assay in its notes (Assmus 2025 uses the validated T. cruzi satellite-DNA qPCR method of Duffy 2013, its reference 46). Specific scope: the only source is the BENDITA trial. Parallel in spirit to [[TTP_MGIT_BASE]], the other inverse-scaled baseline pathogen-burden canonical in this register (longer time-to-positivity likewise means a lower bacterial load), and registered as a distinct pathogen-quantification method following the convention set by [[SARS_VLOAD]] and [[HEV_VLOAD]]. Register a parallel canonical for any future paper that uses a qPCR cycle threshold for a different pathogen.

SARS_VLOAD (canonical for SARS-CoV-2 baseline viral load)

• Description: Baseline (pre-treatment) SARS-CoV-2 viral load measured from nasopharyngeal swab by RT-qPCR, reported as log10 RNA copies/mL.
• Units: log10 copies/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – used with power scaling (SARS_VLOAD / ref)^exponent. Reference values observed: 6.4 log10 copies/mL (Lin 2024, median in pooled COVID-19 cohort).
• Source aliases:
  • VIRAL – used in Lin_2024_casirivimab.R.
• Example models: Lin_2024_casirivimab.R (small negative exponent -0.0075 on CL).
• Notes: SARS-CoV-2-specific. For non-infected subjects, the value is encoded as 0 in the source dataset (below assay detection); the population-PK exponent is small enough that this 0 is absorbed by the reference shift. Register a parallel canonical for any future paper that uses a different infection (e.g., RSV, influenza).

HEV_VLOAD (canonical for hepatitis E virus baseline viral load)

• Description: Pre-treatment (per-subject anchor-time) hepatitis E virus (HEV) RNA concentration in plasma, reported by RT-qPCR. Used as a static per-subject covariate that supplies the initial condition for a virion-state ODE in target-cell-limited HEV viral-dynamics models (virus(0) <- HEV_VLOAD) and that feeds the steady-state derivation of the infection-rate and virion-production-rate constants (beta = kloss * rho / HEV_VLOAD, p = elim * HEV_VLOAD / rho) so that all three viral compartments (healthy hepatocytes, infected hepatocytes, virions) start at pre-treatment steady state.
• Units: IU/mL (international units per millilitre; the WHO-standardized HEV-RNA reporting unit since the 2011 first WHO international standard for HEV RNA, NIBSC code 6329/10).
• Type: continuous
• Scope: specific
• Reference category: n/a – subject-level baseline supplied as a covariate column. Reference values observed: 1.886e6 IU/mL (Mulder 2025 chronic-HEV SOT cohort median; range 527-1.68e8 IU/mL).
• Source aliases:
  • VLBASE – used in Mulder_2025_ribavirin.R.
• Example models: Mulder_2025_ribavirin.R (IU/mL; initial condition for the virus compartment; used in the baseline steady-state derivations for the infection-rate beta and virion-production rate p; cohort median 1,886,058 IU/mL).
• Notes: HEV-specific (parallel to SARS_VLOAD for SARS-CoV-2 and to viral-load canonicals for HCV / HBV / HIV that should be registered separately when first encountered). The IU/mL reporting convention is the WHO-standardized one for HEV; older HEV literature may report viral load in genomic copies/mL with a study-specific copies-to-IU conversion factor (roughly 1 IU ~ 0.3-2.5 copies depending on assay). The Mulder 2025 dataset is chronically infected subjects only (no zero baselines); papers that pool infected and non-infected subjects should encode the non-infected value as the assay LOD and document the convention in covariateData[[HEV_VLOAD]]$notes. Specific scope; promote to general only if a second paper ratifies the same HEV-RNA-baseline-as-initial-condition pattern.

HBSAG_BL_LOG10 (canonical for baseline hepatitis B surface antigen on the log10 IU/mL scale)

• Description: Baseline (pre-treatment) hepatitis B surface antigen (HBsAg) concentration in serum, expressed on the log10 IU/mL scale. HBsAg is a circulating viral protein shed by infected hepatocytes and is the quantity whose disappearance defines the “HBsAg loss” endpoint in chronic hepatitis B; a lower baseline is the single strongest published predictor of achieving that loss. In a model-based meta-analysis the column carries the aggregate study-strata-arm baseline summary rather than an individual measurement.
• Units: log10 IU/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – continuous, entered as a centred deviation (HBSAG_BL_LOG10 - ref) on the logit scale. Reference values observed: 3 log10 IU/mL (Hanan 2026, the across-arm population median, used as the reference in both the end-of-treatment and 24-week post-treatment models). Record the centring value in the model’s covariateData[["HBSAG_BL_LOG10"]]$notes rather than assuming a library-wide default.
• Source aliases:
  • Baseline HBsAg (log10 IU/mL) – used in Hanan_2026_peginterferon_alfa_eot_mbma.R and Hanan_2026_peginterferon_alfa_24wk_mbma.R (Hanan 2026 Table 2 covariate row “Baseline HBsAg (per log10 IU/mL, ref.=3)”; same quantity in the same units, no value transformation).
• Example models: Hanan_2026_peginterferon_alfa_eot_mbma.R (linear centred effect on the logit-scale probability of HBsAg loss at end of treatment: logit_p += -1.57 * (HBSAG_BL_LOG10 - 3), i.e. OR 0.21 per log10 IU/mL – each log10 unit of baseline HBsAg reduces the odds of HBsAg loss by 79%), Hanan_2026_peginterferon_alfa_24wk_mbma.R (same form on the 24-week post-treatment endpoint: logit_p += -1.58 * (HBSAG_BL_LOG10 - 3)).
• Notes: The log10 scale is deliberately explicit in the canonical name so that a raw IU/mL column can never be silently substituted: the founding paper itself quotes raw IU/mL elsewhere (2485 and 2081 IU/mL virtual-cohort baselines in Section 3.5, which are 3.395 and 3.318 on this scale), and a 3-vs-1000 substitution would be a three-order-of-magnitude error that still solves without complaint. Same discipline as BACT_PTT_LOG10CFU. A _VLOAD name would be wrong here: HBsAg is a circulating viral protein, not a viral load – HBV DNA is the viral load for this infection and needs its own separate HBV_VLOAD canonical when a paper first uses it (compare HEV_VLOAD, SARS_VLOAD, both of which are nucleic-acid quantities). Because the two are routinely reported side by side in hepatitis B trials, registering them under distinguishable names matters more here than for other infections. Derivation in an MBMA context is paper-specific and must be documented per model: Hanan 2026 Section 2.2 used the reported median log10 IU/mL when available, the mean when only a mean was reported, and the stratum midpoint when baseline HBsAg was reported as tightly defined strata (e.g. 0-500, 500-1000, 1000-1500 IU/mL), excluding studies that reported only broad non-continuous categories. Specific scope because the reference value and the aggregate-summary derivation are tied to the founding meta-analysis; promote to general when an individual-patient chronic-hepatitis-B model ratifies the same column.

HBEAG_POS (canonical for baseline hepatitis B e-antigen positive indicator)

• Description: 1 = baseline hepatitis B e-antigen (HBeAg) positive, 0 = HBeAg negative. HBeAg positivity marks the immune-tolerant / immune-active phase of chronic hepatitis B (high viral replication, generally higher HBsAg), whereas HBeAg-negative disease is the later, seroconverted phase; the two are the standard stratification of chronic HBV populations and predict different treatment responses. Time-fixed. In a model-based meta-analysis the column carries the study-strata-arm’s HBeAg stratum.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (HBeAg negative).
• Source aliases:
  • Baseline HBeAg status – used in Hanan_2026_peginterferon_alfa_eot_mbma.R (Hanan 2026 Table 2 covariate row “HBeAg status (ref = negative)”; same orientation as the canonical).
• Example models: Hanan_2026_peginterferon_alfa_eot_mbma.R (additive effect on the logit-scale probability of HBsAg loss at end of treatment: logit_p += 0.863 * HBEAG_POS, i.e. OR 2.37 (95% PI 1.34-4.17) – HBeAg-positive arms show more than twice the odds of HBsAg loss).
• Notes: Parallels the _POS suffix convention of HIV_POS, HCV_POS, TB_POS, ADA_POS and SARS_SEROPOS. Unlike those, it is not a coinfection or comorbidity flag: HCV_POS / HIV_POS / TB_POS mark a comorbid infection in a population whose primary indication is something else, whereas HBEAG_POS is a disease-phase / serologic marker within the primary indication (chronic HBV). It is likewise not a disease-state indicator of the DIS_ family, which is reserved for which disease a subject has rather than which serologic phase of it – the reason the DIS_HBEAG form was rejected at ratification. A companion HBEAG_POS_PCT (arm proportion HBeAg-positive) should be registered separately if a future meta-analysis carries mixed arms as a continuous proportion rather than as stratified rows. Mixed-cohort caution: models that enter this indicator on the logit scale are non-linear in it, so an arm containing both HBeAg-positive and HBeAg-negative participants has an aggregate response equal to the HBeAg-weighted average of the two subgroup probabilities, not the probability evaluated at the mean indicator – simulate the two strata separately and average.

SARS_SEROPOS (canonical for SARS-CoV-2 baseline serostatus positive)

• Description: 1 = SARS-CoV-2 spike or nucleocapsid antibody positive at baseline (prior infection or prior vaccination), 0 = seronegative or other / unknown.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (seronegative; “Other” / unknown serostatus is typically pooled into the reference per the source paper’s analysis plan).
• Source aliases:
  • SERPOS – used in Lin_2024_casirivimab.R.
• Example models: Lin_2024_casirivimab.R (multiplicative fractional change on CL).
• Notes: SARS-CoV-2-specific. The exact assay (anti-spike vs anti-nucleocapsid; vendor) varies by study; document per-model in covariateData[[SARS_SEROPOS]]$notes.

OXYSUP_LOW (canonical for low-flow supplemental oxygen indicator)

• Description: 1 = subject is receiving low-flow supplemental oxygen at baseline (e.g., nasal cannula, simple face mask), 0 = no supplemental oxygen at baseline.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no supplemental oxygen at baseline; the high-flow / mechanical-ventilation categories are encoded by the parallel OXYSUP_HIGH indicator).
• Source aliases:
  • OXYSTAT1 – used in Lin_2024_casirivimab.R.
• Example models: Lin_2024_casirivimab.R (multiplicative fractional change on CL; +10.6%).
• Notes: Decomposed indicator from a 4-level ordered categorical (no oxygen / low-flow / high-flow / mechanical ventilation). Use with the parallel OXYSUP_HIGH indicator. Register a separate OXYSUP_VENT canonical if a future analysis splits mechanical ventilation from high-flow oxygen.

OXYSUP_HIGH (canonical for high-flow supplemental oxygen indicator)

• Description: 1 = subject is receiving high-flow supplemental oxygen at baseline (high-flow nasal cannula, non-rebreather mask, non-invasive positive-pressure ventilation, OR mechanical ventilation pooled into the high-flow category), 0 = otherwise (no supplemental oxygen or low-flow).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no supplemental oxygen at baseline; document whether the source paper pooled mechanical ventilation into this indicator or treated it separately).
• Source aliases:
  • OXYSTAT2 – used in Lin_2024_casirivimab.R.
• Example models: Lin_2024_casirivimab.R (multiplicative fractional change on CL; +38.0%).
• Notes: Companion indicator to OXYSUP_LOW. In Lin 2024 the rare mechanical-ventilation cases were pooled into the high-flow indicator (n = 24 across the 7598-subject dataset).

HIV_POS (canonical for HIV-positive comorbidity indicator)

• Description: 1 = HIV-1 antibody positive at study entry, 0 = HIV-negative. Time-fixed per subject. Used as a binary comorbidity indicator on PK parameters (typically bioavailability or clearance) when a study population pools HIV-positive and HIV-negative subjects on a non-HIV primary indication (tuberculosis treatment, hepatitis treatment, etc.).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (HIV-negative).
• Source aliases:
  • HIV – used in Jonsson_2011_ethambutol.R (DDMODEL00000220 NMTRAN $INPUT column with values 0 = HIV negative, 1 = HIV positive; same orientation as the canonical).
• Example models: Jonsson_2011_ethambutol.R (multiplicative 1 + e_hiv_pos_f * HIV_POS shift on bioavailability; HIV-positive patients exhibit a 15.5% reduction in ethambutol bioavailability versus HIV-negative reference), Jonsson_2011_ethambutol_ddmore.R (multiplicative 1 + e_hiv_pos_f * HIV_POS shift on bioavailability; HIV-positive patients exhibit a 15.4% reduction in ethambutol bioavailability versus HIV-negative reference), Francis_2019_rifapentine.R (multiplicative 1 + e_hiv_pos_fdepot * HIV_POS shift on bioavailability; HIV-positive patients exhibit a 21.9 percent reduction in rifapentine bioavailability versus the HIV-negative reference, in 59 of 326 southern African adults with pulmonary tuberculosis. The same “HIV as a comorbidity flag in a TB-treatment cohort, acting on bioavailability” semantics as the two ethambutol entries above, and a larger effect; Francis 2019 notes the data were not sufficient to identify interactions with the individual concomitant antiretrovirals, so the column carries the whole HIV contrast including any unresolved drug-drug interaction).
• Notes: Parallels the _POS suffix convention used by ADA_POS, SARS_SEROPOS, and other serostatus / antibody-positivity indicators. Distinct from a primary disease-state indicator like DIS_HIV (not yet registered) – HIV_POS is a comorbidity flag in non-HIV-primary indications where HIV-vs-non-HIV is tested as a PK covariate.

TB_POS (canonical for active tuberculosis co-infection indicator)

• Description: 1 = active tuberculosis (typically pulmonary) at study entry, 0 = no active TB. Time-fixed per subject. Used as a binary comorbidity indicator on PK or PD parameters (typically albumin secretion, hepatic clearance, bioavailability, or disease-progression rates) when a study population pools TB-positive and TB-negative subjects on a non-TB primary indication (HIV ART, hepatitis treatment, malnutrition, etc.) or when a TB-cohort study pools TB subjects against a separately enrolled non-TB comparator group.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no active TB).
• Source aliases:
  • TB – used in Bisaso_2014_albumin.R (paper text and Table 1 stratification column with values 0 = HIV only, 1 = HIV + TB co-infection; same orientation as the canonical).
• Example models: Bisaso_2014_albumin.R (multiplicative additive shift on baseline albumin secretion rate Q0: Q0 = exp(lq0) * (1 + e_tb_pos_q0 * TB_POS) with e_tb_pos_q0 = -0.308; TB-positive subjects have ~30.8% lower Q0 than the HIV-only reference – equivalent to the paper text’s “44.2% lower” framing relative to the TB-HIV cohort).
• Notes: Parallels the _POS suffix convention used by HIV_POS, ADA_POS, SARS_SEROPOS, and other serostatus / disease-state indicators. Distinct from any TB-treatment-regimen indicator (e.g. CONMED_RIF_LPVR4 for concomitant rifampicin) – TB_POS is the active-disease flag; the medication exposure is a separate concept. In Bisaso 2014 all 158 TB-positive subjects were also on rifampicin-based anti-TB therapy, so the two are confounded in that single cohort; the canonical preserves the conceptual distinction for future studies that decouple them.

HCV_POS (canonical for HCV coinfection / hepatitis C virus positive indicator)

• Description: 1 = chronic hepatitis C virus (HCV) coinfection at study entry (HCV antibody positive and/or >= 2 positive HCV RNA detections on separate visits at least 3 months apart, with subjects of known spontaneous HCV clearance excluded), 0 = HCV uninfected / HCV cleared. Time-fixed per subject. Used as a binary comorbidity / coinfection indicator on PD parameters (e.g., immune-reconstitution recovery rates) or PK parameters when a study pools HCV-coinfected and HCV-uninfected subjects on a non-HCV primary indication (HIV ART, transplant medicine, IBD biologics, etc.).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (HCV uninfected).
• Source aliases:
  • HCV – used in Majekodunmi_2017_HIV_HCV_CD4_recovery.R (paper Methods ‘Definitions’ and Table 2 covariate ‘C:Coinf’ for HIV/HCV coinfected vs HIV monoinfected; same orientation as the canonical).
• Example models: Majekodunmi_2017_HIV_HCV_CD4_recovery.R (multiplicative fractional reduction on the CD4 z-score recovery-rate constant c: c = (c_pop + etac) * (1 + e_hcv_pos_c * HCV_POS) with e_hcv_pos_c = -0.77; HCV-coinfected children recover at 23% of the HIV-monoinfected rate – 0.357 /year versus 1.55 /year typical).
• Notes: Parallels the _POS suffix convention used by HIV_POS, TB_POS, ADA_POS, SARS_SEROPOS, and other serostatus / disease-state indicators. Distinct from any anti-HCV treatment-regimen indicator (e.g., pegylated interferon + ribavirin in Majekodunmi 2017’s coinfected subset of 10 children) and from any HCV-genotype indicator (1/2/3/4 distribution reported in Majekodunmi 2017 Table 1 but not used as a covariate). Distinct from a primary disease-state indicator like DIS_HCV (not yet registered) – HCV_POS is the coinfection / comorbidity flag in non-HCV-primary indications.

EARLY_ART (canonical for early-vs-delayed antiretroviral-treatment-initiation arm indicator)

• Description: Trial randomization-arm indicator: 1 = subject was randomized to initiate antiretroviral treatment (ART) early (within the first 14 days of admission, before nutritional recovery), 0 = subject was randomized to delayed ART initiation (after nutritional recovery, > 14 days from admission). Time-fixed per subject within the trial. The indicator captures the early-vs-delayed-ART contrast tested in the Archary 2019 / MATCH (Malnutrition and ART Timing in Children with HIV) trial in severely malnourished HIV-infected children; the early-ART arm exhibits ~31% higher abacavir bioavailability than the delayed-ART arm.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (delayed ART initiation; F = 1 typical-value reference).
• Source aliases:
  • EARLY – Archary 2019 source description (“randomized to early ART”); same orientation as the canonical (1 = early arm).
• Example models: Archary_2019_abacavir.R (multiplicative additive shift on bioavailability F: f(depot) <- (1 + e_earlyart_f * EARLY_ART) * exp(etalfdepot) with e_earlyart_f = 0.31; the early-arm IIV on F is 21.4%, the delayed-arm IIV on F is fixed at 0 in the source – see model-file vignette Errata).
• Notes: Specific scope because the early/delayed cutoff (14 days from admission) and the underlying nutritional-rehabilitation context are tied to the MATCH-trial design; future studies that test a similar early-vs-delayed contrast with a different time cutoff or non-nutritional context should register a new canonical (e.g., EARLY_ART_28D). Distinct from TRT_PHASE (which gates active-vs-baseline study-phase contributions on a per-record basis) and from DAY14 (which is a within-subject time-since-treatment-initiation landmark, not an enrolment arm).

DIS_TB_MDR (canonical for rifampicin-and-isoniazid-resistant (MDR or worse) tuberculosis indicator)

• Description: 1 = subject’s Mycobacterium tuberculosis isolate is resistant to at least isoniazid + rifampicin, i.e. multidrug-resistant (MDR) TB or any stratum beyond it – treatment-intolerant MDR, non-responsive MDR, pre-XDR and extensively-drug-resistant (XDR) TB; 0 = drug-susceptible TB, or a healthy subject in a pooled healthy-plus-patient analysis. Time-fixed per subject (one drug-susceptibility profile from baseline sputum culture). This is the inclusive resistance indicator: it is deliberately TRUE for XDR subjects, so a model that also needs an XDR-specific contrast pairs it with DIS_TB_XDR_STRICT and forms the MDR-but-not-XDR stratum as DIS_TB_MDR * (1 - DIS_TB_XDR_STRICT).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (drug-susceptible tuberculosis; in pooled analyses that also enrol healthy subjects, the healthy cohort is additionally flagged by DIS_HEALTHY, so the three levels are mutually distinguishable).
• Source aliases:
  • MDR combined with NIXPT – Salinger 2019 supplemental Table S3 carries a baseline MDR column plus a three-level NIXPT column (1 = treatment-intolerant MDR, 2 = non-responsive MDR, 3 = XDR) for the Nix-TB sub-study, and reconstructs the inclusive indicator in $PK as the expression (MDR - NIX + MDRIT + MDRNR + MDRXDR), which evaluates to 1 for every drug-resistant subject whether or not they were enrolled in Nix-TB. A data assembler derives DIS_TB_MDR = as.integer(MDR == 1 | NIXPT %in% 1:3).
• Example models: Kurosawa_2021_bedaquiline.R (bedaquiline CL/F 37.5% higher for healthy volunteers and drug-sensitive TB patients than for MDR-TB patients, (1 + e_nonmdr_cl)^(1 - DIS_TB_MDR) with e_nonmdr_cl = 0.375; McLeay 2014 Table 3), Salinger_2019_pretomanid.R (multiplicative power-form effect on apparent oral clearance, e_dis_tb_mdr_cl^DIS_TB_MDR with e_dis_tb_mdr_cl = 1.15, i.e. 15% higher CL/F in drug-resistant TB; paired with DIS_TB_XDR_STRICT on the apparent central volume, where the 1.44-fold MDR effect is applied to DIS_TB_MDR * (1 - DIS_TB_XDR_STRICT) and XDR subjects instead take a 1.75-fold effect).
• Notes: The inclusive sibling anticipated in the DIS_TB_XDR Notes, which directed future TB drug-resistance models to register siblings rather than overload that name. The three registered members now cover the usual contrasts: DIS_TB_MDR (any isoniazid + rifampicin resistance, XDR included), DIS_TB_XDR (pre-XDR + XDR pooled against MDR), and DIS_TB_XDR_STRICT (XDR alone, pre-XDR in the reference). Read the source’s own contrast wording before choosing – Salinger 2019 tabulates its clearance effect as “CL ~ MDR, TI/NR MDR, or XDR”, which is the inclusive reading, while its volume effects separate “V2 ~ MDR or TI/NR MDR” from “V2 ~ XDR”. Distinct from TB_POS (active tuberculosis as a co-infection in a non-TB primary indication) and from HIV_POS (a comorbidity flag, not a resistance stratum).

DIS_TB_XDR (canonical for (pre-)XDR vs MDR tuberculosis drug-resistance stratum indicator)

• Description: 1 = subject’s Mycobacterium tuberculosis isolate is classified as pre-extensively-drug-resistant (pre-XDR; resistant to isoniazid + rifampicin plus a second-line fluoroquinolone OR an injectable, but not both) or extensively-drug-resistant (XDR; resistant to isoniazid + rifampicin plus a second-line fluoroquinolone AND an injectable); 0 = multidrug-resistant (MDR; resistant to isoniazid + rifampicin only), drug-susceptible, or unclassified TB. Time-fixed per subject (one drug-susceptibility profile per subject from baseline sputum culture). The dichotomisation pools pre-XDR with XDR into the “1” stratum because Svensson 2017 found these two categories share the same slower-clearance bacterial-load phenotype (28.1% longer half-life of mycobacterial load decline relative to MDR).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (MDR / susceptible / missing-treated-as-MDR; the typical-value reference). Per Svensson 2017 Results paragraph 4, the 19% of subjects with missing TB-type information were assigned to the reference (MDR) group after testing that the missing group did not differ significantly from MDR.
• Source aliases: none formally; the Svensson 2017 source describes the variable in prose as the “(pre-)XDR” stratum, operationally encoded as 1 when TB type = pre-XDR or XDR and 0 otherwise (MDR + susceptible + missing).
• Example models: Svensson_2017_bedaquiline.R (multiplicative (1 + e_xdr_hl * DIS_TB_XDR) shift on the half-life of mycobacterial load decline; pre-XDR/XDR patients have a 28.1% longer MBL half-life than MDR/susceptible patients, translating to 2-4 weeks longer median time-to-sputum-culture-conversion per Svensson 2017 Table 3).
• Notes: Specific scope because the dichotomisation is tied to the Svensson 2017 multidrug-resistant tuberculosis cohort (TMC207-C208 Phase IIb registration trial) and the pre-XDR + XDR pooling convention may not transfer to studies that report a different TB-resistance categorisation (e.g., separate XDR vs pre-XDR stratification, or rifampicin-monoresistant TB as a distinct category). Future TB drug-resistance models should consider sibling canonicals (e.g., DIS_TB_XDR_STRICT for XDR-only, DIS_TB_RIFRES for rifampicin-monoresistant) rather than overloading this name; the DIS_TB_XDR_STRICT sibling anticipated here was subsequently ratified for Lin_2024_TB_multistate.R, which places pre-XDR in the reference group. Distinct from HIV_POS (a comorbidity indicator on TB-coinfected patients, not a drug-resistance stratum).

DIS_TB_XDR_STRICT (canonical for XDR-only tuberculosis drug-resistance stratum indicator)

• Description: 1 = subject’s Mycobacterium tuberculosis isolate is classified as extensively-drug-resistant (XDR; resistant to isoniazid + rifampicin plus a second-line fluoroquinolone AND a second-line injectable); 0 = every other resistance stratum, including pre-XDR (resistant to isoniazid + rifampicin plus a fluoroquinolone OR an injectable, but not both), MDR, drug-susceptible and unclassified TB. Time-fixed per subject (one drug-susceptibility profile per subject from baseline sputum culture). Pre-2021 WHO drug-resistance definitions.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-XDR: pre-XDR + MDR + drug-susceptible + missing-treated-as-MDR).
• Source aliases:
  • XDR (Lin 2024 NONMEM control-stream variable, derived as XDR = 0; IF(TBTYPE.EQ.4) XDR = 1 from a four-level TBTYPE column) – used in Lin_2024_TB_multistate.R.
• Example models: Lin_2024_TB_multistate.R (log-hazard effect on the surge amplitude of the sputum-culture-conversion hazard, SA12 = exp(lsa12 + ... + e_xdr_sa12 * DIS_TB_XDR_STRICT) with e_xdr_sa12 = -0.622792; exp(-0.622792) = 0.536 reproduces the paper’s reported “lambda12 decreased by 46% (95% CI, 24%-62%)” for XDR patients), Salinger_2019_pretomanid.R (multiplicative power-form effect on the apparent central volume, e_dis_tb_xdr_strict_vc^DIS_TB_XDR_STRICT with e_dis_tb_xdr_strict_vc = 1.75. The strict sibling is correct here because Salinger 2019 Table 1 tabulates V2 ~ XDR separately from V2 ~ MDR or TI/NR MDR, and the source control stream forms the two strata as mutually exclusive – the MDR volume effect is applied to DIS_TB_MDR * (1 - DIS_TB_XDR_STRICT), so an XDR subject takes the 1.75-fold effect INSTEAD OF, not on top of, the 1.44-fold MDR effect. On clearance the same model uses the inclusive DIS_TB_MDR, which is TRUE for XDR subjects, so the two canonicals appear in one model with deliberately different reference groups).
• Notes: The strict sibling of DIS_TB_XDR, and the two are NOT interchangeable: DIS_TB_XDR pools pre-XDR with XDR into the “1” stratum, whereas DIS_TB_XDR_STRICT places pre-XDR in the reference group. Choosing the wrong one misclassifies the pre-XDR patients, who are a substantial fraction of a typical drug-resistant TB cohort (95 of 402, 28%, in Lin 2024 Table 1). Read the source paper’s own contrast wording before picking: Lin 2024 states the effect for XDR-TB patients “compared with those with non-XDR-TB”, while Svensson 2017 reports a pooled “(pre-)XDR” effect. Specific scope because the dichotomisation depends on which resistance categories the source study enumerated. Distinct from HIV_POS (a comorbidity indicator on TB-coinfected patients, not a drug-resistance stratum). The sibling name was anticipated in the DIS_TB_XDR Notes at its 2026-05-21 ratification and is realised here for the Lin 2024 multistate extraction; patients with a missing resistance profile (~16% in Lin 2024) were assigned to the MDR reference group after the source tested that they did not differ significantly from MDR.

TBM_GRADE_II (canonical for baseline tuberculous-meningitis severity grade II indicator)

• Description: 1 = baseline tuberculous meningitis (TBM) severity grade II, 0 = otherwise (grade I or grade III). Time-fixed per subject (graded at enrolment). TBM grade is the modified British Medical Research Council (MRC) three-level severity scale: grade I = fully conscious (Glasgow coma score, GCS, 15) with no focal neurological signs; grade II = GCS 11-14, or GCS 15 with focal neurological signs; grade III = GCS <= 10. In young children the Blantyre coma score (BCS) replaces the GCS: grade I = BCS 4-5 without focal signs; grade II = BCS 2-3, or BCS 4-5 with focal signs; grade III = BCS <= 1 (Panjasawatwong 2020 Table 1 footnote b, which applies the BCS below 5 years of age and the GCS from 5 years). Paired with TBM_GRADE_III to reproduce the three-level grade with two mutually exclusive level indicators (grade I is the reference level, both indicators = 0).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0; family reference level is TBM grade I (TBM_GRADE_II = 0 and TBM_GRADE_III = 0).
• Source aliases:
  • TBM severity / TBM grade / MRC grade (ordinal column with levels I / II / III, or integer 1 / 2 / 3) – decompose: TBM_GRADE_II = as.integer(grade == 2).
• Example models: Panjasawatwong_2020_tbm_mortality.R (selects the grade II Weibull scale factor of the time-to-death hazard, lambda = 2.35e-6 1/h, Panjasawatwong 2020 Table S3).
• Notes: Level indicators, NOT cumulative thresholds – the same encoding as the ISS_II / ISS_III pair. TBM_GRADE_II and TBM_GRADE_III are mutually exclusive, so a grade III patient has TBM_GRADE_II = 0. Papers that fit TBM grade as a single ordinal covariate with one slope should still supply both indicators. The grade is a composite clinical score and is not an alias of a coma-score column; a model that uses the GCS or BCS directly needs its own column. Specific scope because the scale is defined only for tuberculous meningitis.

TBM_GRADE_III (canonical for baseline tuberculous-meningitis severity grade III indicator)

• Description: 1 = baseline tuberculous meningitis severity grade III (GCS <= 10, or BCS <= 1 in young children), 0 = otherwise (grade I or grade II). Time-fixed per subject. Paired with TBM_GRADE_II; see that entry for the full grading definition and the family encoding.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0; family reference level is TBM grade I (TBM_GRADE_II = 0 and TBM_GRADE_III = 0).
• Source aliases:
  • TBM severity / TBM grade / MRC grade (ordinal column with levels I / II / III, or integer 1 / 2 / 3) – decompose: TBM_GRADE_III = as.integer(grade == 3).
• Example models: Panjasawatwong_2020_tbm_mortality.R (selects the grade III Weibull scale factor of the time-to-death hazard, lambda = 7.94e-5 1/h, Panjasawatwong 2020 Table S3).
• Notes: Mutually exclusive with TBM_GRADE_II – see that entry. Grade III carries by far the highest mortality: 18 of the 100 children in Panjasawatwong 2020 were grade III at baseline, and the fitted hazard puts their 8-month survival near 50%, against about 98% for grade I.

TTP_MGIT_BASE (canonical for baseline mean time-to-positivity in mycobacterial growth indicator tube)

• Description: Subject-specific baseline (pre-treatment) mean time-to-positivity (TTP) in the mycobacterial growth indicator tube (MGIT) liquid-culture system, expressed in days. Computed in the source paper as the mean of three replicate spot-sputum-sample TTP values collected the day before the start of treatment. The MGIT system automatically records the time in hours from inoculation to a positive signal indicating presence of M. tuberculosis (cap at 42 days = negative); the per-subject baseline TTP is then used as a covariate driving the model’s starting mycobacterial load (a longer baseline TTP corresponds to fewer viable bacteria in the inoculum and a lower starting MBL). Time-fixed per subject in the Svensson 2017 source.
• Units: days
• Type: continuous
• Scope: specific
• Reference category: n/a – used in power form (TTP_MGIT_BASE / ref)^exponent. Reference value observed: 6.8 days (Svensson 2017 Table 1 cohort median, n = 191).
• Source aliases:
  • mTTP0 (Svensson 2017 paper symbol, also written mTTP_{0,i}) – used in Svensson_2017_bedaquiline.R.
• Example models: Svensson_2017_bedaquiline.R (power-form covariate on the starting mycobacterial load mbl0_i = exp(lmbl0) * (TTP_MGIT_BASE / 6.8)^e_ttp_mbl0 with e_ttp_mbl0 = -3.69; a four-fold longer median TSCC is predicted in patients with the lowest baseline TTP versus the highest, per Svensson 2017 Discussion paragraph 3), Lin_2024_TB_multistate.R (exponential-deviation covariate shared between the peak time and the surge width of the sputum-culture-conversion hazard, PT12 = exp(lpt12 - e_ttp_pt12sw12 * (TTP_MGIT_BASE - 9.06944)/7) and SW12 = exp(lsw12 + e_ttp_pt12sw12 * (TTP_MGIT_BASE - 9.06944)/7) with e_ttp_pt12sw12 = 0.442668; a longer baseline TTP moves the conversion peak earlier and widens the surge; cohort-specific reference 9.06944 days for the pooled C208 + C209 cohort, n = 402, and the divisor 7 converts the day deviation to weeks so the coefficient is per week of baseline TTP).
• Notes: Specific scope because the cohort-median reference value (6.8 days) is tied to the Svensson 2017 cohort (TMC207-C208 Phase IIb registration trial); future MGIT-based mycobacterial-load models should reuse this canonical but document their own cohort-specific reference in covariateData[[TTP_MGIT_BASE]]$notes (as Lin_2024_TB_multistate.R does for its 9.06944-day pooled-cohort median). Source columns are frequently recorded in HOURS by the MGIT instrument (Lin 2024’s MTTP column is centered at 217.6667 h = 9.06944 days); always convert to days when populating this canonical. Distinct from LNPC (log-transformed admission Plasmodium parasitaemia, a different pathogen-quantification method for malaria) and from SARS_VLOAD (SARS-CoV-2 RT-qPCR log10 copies/mL, a different infectious-disease baseline biomarker). MGIT samples without a positive signal within 42 days are classified as negative in the source paper; for the per-subject baseline mean, only the positive triplicate values are averaged (a subject with all-negative baseline samples has missing TTP_MGIT_BASE).

MBL_HL_WK2 (canonical for model-derived half-life of mycobacterial-load decline through week 2)

• Description: Per-subject half-life of the decline of the model-derived mycobacterial load (MBL) over the first 2 weeks of anti-tuberculosis treatment, expressed in weeks. A model-derived covariate: it is not measured directly but is a secondary metric read out of an upstream longitudinal mycobacterial-load PK-PD model, which integrates baseline bacterial burden, drug-resistance profile, prior anti-TB treatment and individual drug exposure over time into a single early-bactericidal-activity summary. A shorter half-life means faster bacterial clearance. Time-fixed per subject in the models that use it: the source computes the metric prospectively from data up to week 2 only (to avoid immortal-time bias) and carries the week-2 value forward thereafter, so a single scalar per subject reproduces the model exactly.
• Units: week
• Type: continuous
• Scope: specific
• Reference category: n/a – used as an exponential deviation exp(beta * (MBL_HL_WK2 - ref)). Reference value observed: 0.69443 weeks (Lin 2024 cohort median, hardcoded in the source NONMEM control stream and confirmed by Lin 2024 Figure 3, whose reference individual has “0.69 weeks of half-life of bacterial clearance at Week 2”). Lin 2024 Figure 6 reports the 5th and 95th percentiles as 0.33 weeks (high clearance) and 1.1 weeks (low clearance).
• Source aliases:
  • HL2 (Lin 2024 paper symbol and data column; the subscript denotes “up to week 2”) – used in Lin_2024_TB_multistate.R.
• Example models: Lin_2024_TB_multistate.R (enters BOTH surge parameters of the sputum-culture-conversion hazard through a single shared coefficient of opposite sign, SA12 = exp(lsa12 + e_hl2_sa12pt12 * (MBL_HL_WK2 - 0.69443) + ...) and PT12 = exp(lpt12 - ... - e_hl2_sa12pt12 * (MBL_HL_WK2 - 0.69443)) with e_hl2_sa12pt12 = -0.686145; because the coefficient is negative, a shorter half-life simultaneously raises the surge amplitude and moves the peak earlier).
• Notes: Must be strictly positive. The intended source is the half-life of the mycobacterial-load state of modellib('Svensson_2017_bedaquiline'), whose hl_i is already in weeks – no conversion is needed; supplying it from any other longitudinal bacterial-load model is acceptable provided the half-life is expressed in weeks and evaluated at week 2. Paired with MBL_END, the end-of-treatment load from the same upstream model; the two share the MBL_ prefix precisely because they are different readouts of one underlying mycobacterial-load trajectory, and the family extends naturally (MBL_BASE, MBL_HL_WK8,…). Distinct from TTP_MGIT_BASE, which is a directly measured baseline culture-growth time rather than a model-derived rate of on-treatment decline; the two are complementary and Lin 2024 retains both. Specific scope because the cohort-median reference value is tied to the pooled TMC207-C208 + TMC207-C209 cohort; models reusing this canonical should document their own reference in covariateData[[MBL_HL_WK2]]$notes.

MBL_END (canonical for model-derived mycobacterial load at the end of treatment)

• Description: Per-subject model-derived mycobacterial load (MBL) at the end of the investigational anti-tuberculosis treatment period, on the natural (untransformed) scale, in number of bacteria per sample inoculum. A model-derived covariate read out of an upstream longitudinal mycobacterial-load PK-PD model; the source describes it as “a composite of disease severity at baseline and on-treatment response throughout the whole treatment period”, because the upstream model integrates baseline burden, resistance profile and the full longitudinal individual drug exposure. Time-fixed per subject.
• Units: n bacteria per sample inoculum
• Type: continuous
• Scope: specific
• Reference category: n/a – used as a natural-log deviation exp(beta * (log(MBL_END) - log(ref))). Reference value observed: 5.5726e-05 bacteria per inoculum (Lin 2024 cohort median, hardcoded in the source NONMEM control stream as LOG(0.000055726)); log10(5.5726e-05) = -4.25, matching the “-4.3 log10(MMBLend)” reference individual of Lin 2024 Figure 3.
• Source aliases:
  • MMBLend (Lin 2024 paper symbol and data column; “model-derived Mycobacterial Load at end of treatment”) – used in Lin_2024_TB_multistate.R.
• Example models: Lin_2024_TB_multistate.R (log-linear effect on the recurrence hazard, gated to act only after week 26: lambda23 = exp(llambda23 + e_sexf_lambda23 * SEXF + e_mblend_lambda23 * (t > 26) * (log(MBL_END) - log(5.5726e-05))) with e_mblend_lambda23 = 0.0371081).
• Notes: Store on the NATURAL scale and let model take the log – the coefficient is per unit of natural log, so a per-log10-unit hazard ratio is exp(beta * ln(10)) (for Lin 2024, exp(0.0371081 * ln(10)) = 1.089). Confusing the natural-log and log10 conventions is the obvious failure mode here because the source paper’s figures are labelled in log10(MMBLend) while the control stream uses LOG (natural log). Must be strictly positive: log(0) is undefined and would propagate NaN through the entire ODE solve even at times before the gate opens, since 0 * -Inf is NaN. Units match the mycobacterial-load state of modellib('Svensson_2017_bedaquiline') (n bacteria per sample inoculum, typical starting MBL_0 = 2.14e3), so that model’s mbl state at the end of treatment can be supplied directly. Paired with MBL_HL_WK2 (see that entry for the shared MBL_ family rationale). Specific scope because both the cohort-median reference and the “end of treatment” timepoint (24 weeks in Lin 2024) are study-design-dependent; models reusing this canonical should document their own reference value and end-of-treatment definition in covariateData[[MBL_END]]$notes.

CD19_ABS (canonical for absolute CD19+ B-lymphocyte count)

• Description: Absolute peripheral-blood CD19+ B-lymphocyte count. Baseline or time-varying; document the time resolution per model via covariateData[[CD19_ABS]]$notes. Used as a continuous target-antigen-burden / B-cell-mediated-clearance biomarker in anti-CD20 (rituximab, ofatumumab, obinutuzumab) and other B-cell-targeted antibody popPK models, where the therapeutic mAb elimination increases with the size of the on-target B-cell pool via target-mediated drug disposition.
• Units: cells/mm^3 (equivalently cells/uL; the two labels are numerically equivalent).
• Type: continuous
• Scope: general – the absolute CD19+ B-cell count is the standard flow-cytometry biomarker of circulating B-cell burden and is expected to reuse across anti-CD20 / anti-CD79b / anti-CD22 popPK extractions where the source paper carries the count as an exogenous exposure covariate rather than as a modelled PD state.
• Reference category: n/a – enters as a centred power term (CD19_ABS / ref)^exponent against a population reference. Document the form and reference value in each model’s covariateData[[CD19_ABS]]$notes.
• Source aliases:
  • CD19 – bare-name source-paper column; common in clinical-PK papers (Lioger 2017 prints “CD19+ count” throughout Methods, Results, and Table 2).
  • CD19_COUNT – explicit-suffix variant.
• Example models: Lioger_2017_rituximab.R (time-varying pre-course CD19+ count as a covariate on the elimination rate constant k10; power exponent 0.035, reference 100 cells/mm^3; sign is positive because higher CD19+ counts correlate with faster rituximab elimination via target-mediated drug disposition).
• Notes: Distinct from BLBCELL (baseline-only CD19+ B-cell count used as a baseline covariate in ofatumumab / polatuzumab vedotin PK-PD models, where the count is measured prior to first dose only) – CD19_ABS is the general canonical for CD19+ counts that are carried as time-varying exogenous covariates measured at multiple visits. BLBCELL’s notes (“Distinct from a time-varying B cell count, which is the PD response variable rather than a covariate”) reflected the earlier register where the only registered anti-CD20 CD19+ use case was baseline; CD19_ABS fills the exogenous time-varying covariate slot (analogous to how CD4_ABS covers both baseline and time-varying CD4+ counts). Distinct from a modelled time-varying B-cell state, which is a PD response variable and would be encoded as an ODE state, not a covariate column.

CD4_ABS (canonical for absolute CD4+ T-lymphocyte count)

• Description: Absolute peripheral-blood CD4+ T-lymphocyte count. Baseline or time-varying; document the time resolution per model via covariateData[[CD4_ABS]]$notes. Used as a continuous disease-severity / immunosuppression-depth biomarker in HIV / AIDS popPK models and as a covariate on clearance for drugs whose disposition is altered by advanced HIV immunosuppression.
• Units: cells/mm^3 (equivalently cells/uL; the two labels are numerically equivalent).
• Type: continuous
• Scope: general – the absolute CD4 count is a textbook HIV-severity covariate that future popPK extractions in HIV / AIDS cohorts are expected to reuse.
• Reference category: n/a – enters either as a raw additive linear slope (McLachlan 1996, CL = exp(lcl) + e_crcl_cl * CRCL + e_cd4_abs_cl * CD4_ABS; coefficient in L/h per cell/mm^3) or as a centred linear / power term (CD4_ABS / ref)^exponent against a population reference (typical reference 50 or 200 cells/mm^3 in HIV cohorts). Document the form and reference value in each model’s covariateData[[CD4_ABS]]$notes.
• Source aliases:
  • CD4 – bare-name source-paper column; common in clinical-PK papers (McLachlan 1996 prints “CD4 cell count” in the abstract and “CD4+ T-lymphocyte count” in the Methods).
  • CD4_COUNT – explicit-suffix variant.
  • CD4_T_COUNT – verbatim “CD4+ T-lymphocyte count” form used in some clinical-PK papers.
• Example models: McLachlan_1996_fluconazole.R (raw additive linear slope on the CL covariate model: 0.00068 L/h per cell/mm^3; cohort mean 69 cells/mm^3 with 97 of 109 covariate-evaluable subjects below 200 cells/mm^3 per Table 3; entered as e_cd4_abs_cl * CD4_ABS in the additive intercept-plus-slopes CL regression alongside e_crcl_cl * CRCL).
• Notes: Distinct from the CD4 z-score family (e.g., the age-standardised z-score modelled by Majekodunmi_2017_HIV_HCV_CD4_recovery.R), which is a paediatric-recovery standardised metric on the natural-z-score scale and is not directly interchangeable with the absolute count. Distinct from CD4 percentage (CD4_PCT, the proportion of total lymphocytes that are CD4+; commonly reported in paediatric HIV studies and not yet registered) – register CD4_PCT separately if a future paper retains it. Distinct from HIV_POS (the binary serostatus indicator), which captures HIV-positive-vs-negative rather than the depth of immunosuppression within an HIV-positive cohort. The two can be paired in a single model when an HIV-positive subset is further stratified by CD4 count, but most HIV-popPK papers focus on either the comorbidity flag (mixed HIV+/HIV- cohort) or the absolute count (HIV+-only cohort), not both.

CD4CD8_RATIO (canonical for the CD4+/CD8+ T-lymphocyte ratio)

• Description: Ratio of CD4+ (helper) to CD8+ (cytotoxic) T lymphocytes, a dimensionless immunology readout. A low ratio marks a cytotoxic-skewed compartment (more CD8+ than CD4+ cells); a high ratio marks a helper-skewed compartment. Applicable to whole-blood T cells or, as in the founding example, to a specific engineered T-cell product (CD4+/CD8+ CAR-T cells). The population, the measurement matrix (whole blood vs a cell product), and the measurement timing (baseline vs a specific day post-treatment) must be documented per model in covariateData[[CD4CD8_RATIO]]$notes, because the ratio is a dynamic quantity.
• Units: (dimensionless)
• Type: continuous
• Reference category: n/a – enters as a continuous term. In the founding example it is a power term CD4CD8_RATIO^exponent referenced to a ratio value of 1 (e_cd4cd8_vmax1 = -0.385). Document the form and reference value in each model’s notes.
• Scope: general – the CD4+/CD8+ (helper:cytotoxic) T-cell ratio is a standard immunology index expected to recur across immunotherapy, HIV / AIDS, and transplant popPK-PD extractions.
• Source aliases:
  • CAR+ CD4/CD8 day7 / ratio of CD4+/CD8+ CAR-T cells on day seven (Mueller-Schoell 2021) – the day-7 ratio of CD4+ to CD8+ CAR-T cells; the timing (day 7) lives in the per-model notes, not the name.
  • CD4:CD8 ratio, CD4/CD8 – generic whole-blood forms.
• Example models: MuellerSchoell_2021_axicabtageneCiloleucel_qsp.R (day-7 CD4+/CD8+ CAR-T cell ratio as a power covariate on the reference-population baseline maximum expansion rate, Vmax1,base,ref * CD4CD8_RATIO^(-0.385), referenced to a ratio of 1; Equation 10 and Table 1).
• Notes: Distinct from the absolute-count family (CD4_ABS, CD8_*), which carry cells/uL rather than a ratio – a ratio of 1 is uninformative about whether both counts are high or both low. Distinct from FCD8TDP (Minucci_2024_CART_qsp.R), which is the CD8+ FRACTION of the infused drug PRODUCT (a manufacturing-composition covariate, values in [0,1]), not an in-vivo patient ratio; the two are not algebraic aliases (FCD8TDP = CD8 / (CD4 + CD8) whereas CD4CD8_RATIO = CD4 / CD8, and the founding papers measure them on different objects at different times). The IFNG_PBORATIO entry’s note that “no CD4_RATIO sibling is registered yet” is closed by this entry.

CD8_CD38DR_PCT (canonical for the percentage of circulating CD8+ T cells co-expressing CD38 and HLA-DR (systemic immune activation))

• Description: Percentage of peripheral-blood CD8+ T lymphocytes that co-express CD38 and HLA-DR, measured by multiparameter flow cytometry on peripheral blood mononuclear cells. This double-positive fraction is the standard quantitative index of systemic immune activation in HIV immunology: it is elevated in untreated HIV infection and in active tuberculosis, tracks disease activity independently of the depth of CD4 depletion, and is the mechanistic hypothesis by which inflammation suppresses hepatic drug-metabolising enzyme activity. Measured per subject per occasion (it changes with antiretroviral therapy and with anti-TB therapy), so it is a time-varying covariate at occasion resolution rather than a time-fixed baseline. Enters clearance as a continuous term normalised to a cohort reference level.
• Units: % of CD8+ T cells (a percentage of the parent CD8+ gate, not of total lymphocytes and not an absolute count; document the per-model reference/normalisation value in covariateData[[CD8_CD38DR_PCT]]$notes)
• Type: continuous
• Scope: general – the activated-CD8 double-positive fraction is the canonical systemic-immune-activation readout across HIV, tuberculosis, sepsis and post-transplant popPK cohorts, and inflammation-mediated suppression of drug metabolism is a recurring extraction class (the CRP and IL6 entries cover the soluble-marker side of the same mechanism).
• Reference category: n/a – enters as a median-normalised power term (CD8_CD38DR_PCT / ref)^exponent against the cohort median (36.9% in the founding example). Record the form and the normalising value in each model’s covariateData notes; the normalisation is what keeps the base typical value interpretable as the clearance at the cohort-median activation level.
• Source aliases:
  • %CD38+DR+CD8+ – the verbatim form printed by Vinnard 2017 (Table 1 and Figure 2B).
  • %CD38 + HLA-DR + |CD8 – the Table 2 parameter-row form of the same quantity.
  • CD38DR_CD8 / CD8_ACT – bare-name dataset column variants; same orientation as the canonical, no transformation, but confirm the source reports a percentage (0-100) rather than a fraction (0-1) before use.
• Example models: Vinnard_2017_isoniazid.R (median-normalised power term on apparent oral isoniazid clearance, immune_cl <- (CD8_CD38DR_PCT / 36.9)^e_cd8_cd38dr_pct_cl with exponent -0.31, fitted alongside a three-level NAT2 acetylator effect; higher immune activation lowers clearance and so raises isoniazid exposure).
• Notes: Report as a percentage (0 to 100) to match the flow-cytometry convention every source paper uses; a source supplying a 0-to-1 fraction must be multiplied by 100 on the way in. Because the canonical form is ratio-normalised, the percent-versus-fraction choice does not change the fitted covariate effect, but it does change what the normalising reference value means, so the two must be kept consistent within a model. Distinct from CD4_ABS (the absolute CD4+ count, which measures the depth of immunosuppression rather than the degree of activation) – the founding paper is the discriminating case, retaining CD8_CD38DR_PCT on clearance while explicitly rejecting CD4_ABS, and concluding that “immune activation rather than immune suppression drives the variability in NAT-2 activity”. Distinct from CRP and IL6, which are soluble downstream inflammatory mediators of the same axis and were also screened and rejected in the founding paper. The companion single-marker fractions that a source may tabulate alongside this one (%CD38+DR-CD8+, %CD38-DR+CD8+, and the three CD4+ analogues) are separate quantities; register CD4_CD38DR_PCT and the single-positive variants separately if a future paper retains one, following this entry’s naming pattern <parent gate>_<marker(s)>_PCT. The _PCT suffix follows the established percentage family (BODYFAT_PCT, HUMAN_SERUM_PCT) and the convention pre-named in the CD4_ABS notes for cell-population percentages.

DIS_ANTHRAX (canonical for active inhalational-anthrax infection indicator)

• Description: 1 = the subject has active inhalational anthrax (in the founding example, following aerosol challenge with a target 200 LD50 of Bacillus anthracis Ames-strain spores); 0 = healthy / unexposed. Time-fixed per subject in challenge-model designs, where an animal is allocated to either the challenged or the unexposed arm for the whole study. Distinguishes the infected cohort from the healthy cohort when a population PK model is fit simultaneously to both in order to estimate a disease effect on disposition.
• Units: (binary)
• Type: binary
• Scope: general – a disease-state indicator in the established DIS_<condition> family; anthrax medical countermeasures developed under the FDA Animal Rule (obiltoxaximab, raxibacumab, anthrax immune globulin) are a recurring extraction class.
• Reference category: 0 (healthy / unexposed).
• Source aliases:
  • INFECTED – used in Nagy_2017_obiltoxaximab.R (Nagy 2017 Results: “TMDD in infected NZW rabbits and cynomolgus macaques was approximated via parallel nonlinear elimination for infected animals only”).
• Example models: Nagy_2017_obiltoxaximab.R (switches on a parallel Michaelis-Menten elimination arm approximating protective-antigen target-mediated drug disposition: vmax <- DIS_ANTHRAX * vmaxRef * (WT / wtRefVmax)^e_wt_cl).
• Notes: In anthrax monoclonal-antibody models the biological content of this indicator is the presence of the drug’s binding target: protective antigen (PA) circulates only during active infection, so target-mediated disposition exists only when DIS_ANTHRAX = 1. That is why the founding paper’s human population PK model – fit to healthy volunteers only – has no fitted nonlinear arm, and why setting DIS_ANTHRAX = 1 for a human subject invokes a component transferred from the macaque rather than one estimated in humans. Distinct from DIS_INFECT_CSSSI_SEV (a severity grade for complicated skin and skin-structure infection) and from the pathogen-load covariate BACT_PTT_LOG10CFU, which quantifies how advanced the infection is rather than whether it is present; the two are complementary and a model may carry both.

BACT_PTT_LOG10CFU (canonical for prior-to-treatment quantitative bacteremia on the log10 CFU/mL scale)

• Description: Quantitative blood bacterial burden measured immediately prior to the start of treatment (“prior-to-treatment”, PTT), expressed as log10 colony-forming units per mL of blood. Per-subject scalar captured at the treatment-trigger time point. Serves as the disease-severity covariate in trigger-to-treat infection models, where treatment is initiated on a physiological trigger rather than at a fixed time and the bacterial burden at that moment is the dominant predictor of outcome.
• Units: log10 CFU/mL
• Type: continuous
• Scope: general – prior-to-treatment bacteremia is the standard severity covariate across bacterial-challenge medical-countermeasure programmes, not specific to anthrax.
• Reference category: n/a – enters on its natural log10 scale. In the founding example it appears twice in the same model: as an exponentiated penalty on the cure fraction, logit(psurv) = theta0 - (theta1 * BACT_PTT_LOG10CFU)^theta2 + Emax * dose / (ED50 + dose), and as a log-linear effect on the Weibull death rate, log(lambda) = lambda0 + lambda1 * BACT_PTT_LOG10CFU. A value of 0 denotes no detectable PTT bacteremia and makes the penalty term vanish.
• Source aliases:
  • PTT – used in Nagy_2017_obiltoxaximab_survival.R (Nagy 2017 Results, “Animal survival modeling”; the paper writes the covariate as “log10(PTT bacteremia)” and “PTT quantitative bacteremia”).
• Example models: Nagy_2017_obiltoxaximab_survival.R (Weibull cure-rate survival model; Supplementary Figure S2 stratifies the observed and predicted dose-response into quartiles of this covariate – [BLQ, 3.02], [3.03, 3.95], [3.96, 4.87], (4.87, 8.56] log10 CFU)………….
• Notes: Values below the assay’s limit of quantification are reported by the founding paper as “BLQ” and pooled into the lowest quartile; a model consuming this covariate must decide how to impute BLQ records and document the choice in covariateData[[BACT_PTT_LOG10CFU]]$notes (the founding vignette uses the quartile midpoint when reproducing the published figure). The _PTT token marks the measurement timing (at treatment trigger), which matters because bacteremia rises steeply during untreated infection – a post-treatment or peak bacteremia value is a different covariate and should be registered separately (e.g. BACT_PEAK_LOG10CFU) rather than folded into this name. The _LOG10CFU token marks the scale explicitly so that a raw-CFU column is never silently substituted; register BACT_PTT_CFU if a paper models the untransformed count. Distinct from LNPC (natural-log Plasmodium parasitaemia per microlitre), which is the malaria analogue on a different log base, a different organism and a different volume unit.

HCV_VLOAD (canonical for hepatitis C virus baseline viral load)

• Description: Baseline (pre-treatment) hepatitis C virus RNA concentration in serum or plasma, measured by a quantitative nucleic-acid assay. Carried as the CONTINUOUS concentration even when a model consumes it only in dichotomized form, so that one column can serve several thresholds and so that the threshold stays visible in model() rather than being baked into the dataset.
• Units: IU/mL (the WHO-standardized reporting unit for HCV RNA since the 1997 first WHO international standard, NIBSC code 96/790). Papers reporting copies/mL use an assay-specific conversion of roughly 1 IU ~ 2.7 copies for the older bDNA and RT-PCR assays; document the assay and any conversion per model in covariateData[[HCV_VLOAD]]$units and $notes.
• Type: continuous
• Scope: general
• Reference category: n/a – supplied as a subject-level baseline concentration. Threshold values observed: 800,000 IU/mL (Riggs 2012 popPK, the conventional “high viral load” cut-point in chronic-HCV interferon trials) and 400,000 IU/mL (Riggs 2012 SVR exposure-response, the lower cut-point of a three-level categorization).
• Source aliases:
  • Baseline HCV RNA – Riggs 2012 covariate-equation and Table I row label.
• Example models: Riggs_2012_albinterferon.R (dichotomized inside model() at 800,000 IU/mL, giving multiplicative effects 0.988 on CL/F, 0.989 on V/F, and 1.09 on ka per Table I), Riggs_2012_albinterferon_svr_gt1.R, Riggs_2012_albinterferon_svr_gt23.R (the same column categorized into three levels – below 400,000, the 400,000-to-800,000 reference band, and 800,000 or above – driving two logit-scale coefficients in each genotype stratum’s sustained-virologic-response logistic regression).
• Notes: Distinct from HCV_POS, which is a binary HCV-coinfection indicator used when HCV status is a comorbidity of the modelled population rather than the treated disease. Populate with the continuous concentration and binarize inside model() (hcvHigh <- (HCV_VLOAD >= 800000)), the same pattern the ALP entry prescribes for above-ULN indicators; a pre-binarized dataset column silently loses the threshold. For subjects below the assay limit of quantitation, record the LOQ value and document the convention per model, as SARS_VLOAD and HEV_VLOAD require. Note that the reporting SCALE differs across the <VIRUS>_VLOAD family – SARS_VLOAD is log10 copies/mL while HEV_VLOAD and HCV_VLOAD are linear IU/mL – so never carry a log10 value in this column.

HIV_VLOAD (canonical for human immunodeficiency virus type 1 baseline plasma viral load)

• Description: Baseline (pre-treatment / Day 1) plasma HIV-1 RNA concentration, measured by a quantitative nucleic-acid assay. Carried as the CONTINUOUS concentration on the LINEAR scale even when a model consumes it log-transformed or dichotomized, so that one column can serve several transforms and so that the transform stays visible in model() rather than being baked into the dataset. The HIV-1 member of the <VIRUS>_VLOAD family, pre-named in the HEV_VLOAD notes (“viral-load canonicals for HCV / HBV / HIV that should be registered separately when first encountered”).
• Units: copies/mL (c/mL). HIV-1 RNA has no IU-based WHO reporting convention in routine clinical use, unlike HCV and HEV; copies/mL is universal in the antiretroviral literature. Document the assay and its lower limit of quantitation per model in covariateData[[HIV_VLOAD]]$units and $notes.
• Type: continuous
• Scope: general – baseline plasma HIV-1 RNA is the standard disease-burden covariate across antiretroviral popPK and exposure-response analyses, and is the denominator of every virologic-response endpoint in the field.
• Reference category: n/a – a subject-level baseline concentration. Enters either as a median-normalized power term (HIV_VLOAD / ref)^exponent on the linear scale or as a median-centred log10 term beta * (log10(HIV_VLOAD) - ref10); a single paper can use BOTH forms for the same column across sibling models. Reference values observed: 44,940 c/mL and its log10 equivalent 4.65 (Parasrampuria 2025, the phase 3 BRIGHTE Day 8 exposure-efficacy cohort median).
• Source aliases:
  • BHIVRNA – Parasrampuria 2025 Table 3 and Table 4 covariate-equation notes; “baseline plasma HIV-1 RNA (c/mL)” in the abbreviations. Carries the LINEAR value in the Table 3 equation and the LOG10 value in the Table 4 equation, with the same symbol in both – read the centring constant to tell which scale is meant (a constant near 4-7 is log10; a constant in the tens of thousands is linear).
  • Baseline plasma HIV-1 RNA, baseline HIV RNA, VL, HIVRNA – common prose and dataset spellings.
• Example models: Parasrampuria_2025_temsavir_hivrna_day8.R (linear-scale median-normalized power term on the Emax drug-effect of a temsavir exposure-response model, (HIV_VLOAD / 44940)^0.150), Parasrampuria_2025_temsavir_resp05log_day8.R, Parasrampuria_2025_temsavir_resp1log_day8.R (the SAME column on the log10 scale, median-centred, as a logit-scale slope 0.836 * (log10(HIV_VLOAD) - 4.65) and 0.848 * (log10(HIV_VLOAD) - 4.65) in the two companion responder-rate logistic regressions – founding examples).
• Notes: Distinct from HIV_POS, the binary serostatus indicator used when HIV status is a comorbidity of the modelled population rather than the treated disease; the two are complementary and a model fit to a mixed healthy / HIV-infected cohort may carry both. Distinct from CD4_ABS and CD8_CD38DR_PCT, which measure the host immune consequence of infection rather than the viral burden. Populate with the linear concentration and transform inside model(); the founding paper is the discriminating case, consuming this one column on two different scales across three sibling models in a single publication. For subjects below the assay lower limit of quantitation (commonly 40, 50 or 200 c/mL depending on assay generation), record the LOQ value and document the convention per model, as SARS_VLOAD, HEV_VLOAD and HCV_VLOAD require. Note that the reporting SCALE differs across the <VIRUS>_VLOAD family – SARS_VLOAD is log10 copies/mL while HEV_VLOAD, HCV_VLOAD and this entry are linear – so never carry a log10 value in this column. Register HBV_VLOAD separately when first encountered, following this pattern.

IMMUNE_STATE (canonical for experimentally-induced immune status in a preclinical infection model)

• Description: Ordered categorical level of host immune competence in a preclinical infection model whose immune status is an experimental design factor rather than a patient characteristic: 0 = neutropenic, 1 = intermediate suppression, 2 = immunocompetent. Time-fixed per animal (the immunosuppression is induced by cytotoxic pretreatment completed before infection). Used when a paper runs the same infection model at two or more deliberately-set immune states and estimates state-specific parameters, so the column selects which parameter set applies rather than scaling a base value. Levels beyond 2 should not be invented; a study with a different number of arms maps its arms onto the same three semantic levels, and a genuinely finer gradation warrants extending this entry rather than a parallel column.
• Units: (categorical)
• Type: categorical
• Scope: general – the neutropenic / immunocompetent contrast is the standard axis of antibacterial PKPD animal work (neutropenic mouse thigh and lung infection models), not specific to one paper.
• Reference category: 0 (neutropenic). This is the reference for two independent reasons: it is the conventional preclinical infection-model condition against which the others are contrasted, and it is the level at which the immune contribution is typically not identifiable, so its immune-response parameter is fixed to zero and the other levels are increments on top of it.
• Source aliases:
  • immune status – the phrase used throughout Saporta_2026_meropenem.R (Saporta 2026 Materials and Methods, “Mouse lung infection model”; Table 1 carries the levels as the parameter subscripts neu / int / com).
• Example models: Saporta_2026_meropenem.R (three states induced by intraperitoneal cyclophosphamide at 4 and 1 days before infection – 200 + 150 mg/kg neutropenic, 75 + 50 mg/kg intermediate, none immunocompetent – selecting the apparent central volume 2.19 / 2.46 / 3.40 L/kg, the phagocytosis rate kphag 0 / 0.185 / 0.318 1/h, and the initial susceptible bacterial count 7.56 / 7.80 / 7.52 log10 CFU/lung).
• Notes: Kept as a single ordered column rather than a RACE_<GROUP>-style binary-indicator family because the levels are monotone and the founding paper’s immune-response parameterization is cumulative across them, so mutually-exclusive indicators would have to be kept consistent by the user for no gain. Named for the resulting immune STATE rather than for the immunosuppressive intervention (the rejected IMMUNOSUPP_LEVEL, which would also reverse the level ordering) because the state is what the model’s parameters are indexed by, and because a study could reach the same state by a non-pharmacological route (SCID / NOD / gamma-c knockout strains, which Saporta 2026 names as the natural extension of this design). Distinct from a measured immune biomarker. WBC, NEUT and the granulocyte counts quantify the realised immune status on a continuous scale; this column carries the assigned experimental arm. The two are not interchangeable and a paper may report both – Saporta 2026 measured blood granulocyte counts (0.2, 0.4 and 1.9 x 10^3 /mm3 across the three arms) and found that scaling the immune-response parameters by them fitted worse than the categorical arm. Distinct also from DIS_ disease-state indicators (which disease a subject has) and from the immunosuppressant co-medication flags of the CONMED_ family (the cyclophosphamide here is a design intervention that has washed out of the PK-relevant window, not a co-administered perpetrator drug).

DIS_PSEUDOMONAS_LUNG_ACUTE (canonical for acute Pseudomonas aeruginosa lung infection (preclinical) indicator)

• Description: 1 = animal with an acute Pseudomonas aeruginosa lung infection established by intratracheal inoculation of a planktonic (free, non-embedded) bacterial suspension and studied a few days later; 0 = otherwise. Time-fixed per animal. One arm of a mutually exclusive group set with DIS_PSEUDOMONAS_LUNG_CHRONIC and ALGINATE_BEAD_BLANK; a healthy, non-inoculated animal has all three at 0.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (healthy, non-inoculated control animal from the same study; the control cohort is paper-defined).
• Source aliases:
  • acute infection, acutely infected group – Dias 2022 Sections 2.3-2.4 and Table S2.
• Example models: Dias_2022_tobramycin_rat.R (male Wistar rats 7 days after intratracheal inoculation of 100 uL of 10^9 CFU/mL P. aeruginosa PA14; enters only through the inoculated-lung volume lv_lung_inoc (V3infected = 0.130 L vs 0.083 L healthy), clearance and central volume being those of healthy rats).
• Notes: Ratified 2026-10-02 alongside the Dias 2022 tobramycin extraction (PMC9228144), as a preclinical DIS_<disease> sibling of DIS_CIA and DIS_TCT_COLITIS. The species name is spelled out (PSEUDOMONAS, not PSAE) per the register’s preference for spelled-out new canonicals. Deliberately distinct from DIS_INFECT_ACTIVE, which is a time-varying clinical-episode flag and cannot separate an acute from a chronic model, and from IMMUNE_STATE, which carries the host’s immune arm rather than the infection. The species of host is recorded in population$species and the model file stem, not here. A planktonic infection with a different organism should register its own DIS_<ORGANISM>_LUNG_ACUTE sibling.

DIS_PSEUDOMONAS_LUNG_CHRONIC (canonical for chronic (alginate-bead) Pseudomonas aeruginosa lung infection (preclinical) indicator)

• Description: 1 = animal with a chronic Pseudomonas aeruginosa lung infection established by intratracheal inoculation of bacteria embedded in alginate (or agar) beads, the standard rodent model of the mucoid chronic infection of cystic-fibrosis airways, studied about two weeks later; 0 = otherwise. Time-fixed per animal. Mutually exclusive with DIS_PSEUDOMONAS_LUNG_ACUTE and ALGINATE_BEAD_BLANK.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (not chronically infected; in the founding example this pools healthy, acutely infected and blank-bead animals for the plasma parameters).
• Source aliases:
  • chronic infection, chronically infected group, subscript chronic on CLchronic / V1chronic – Dias 2022 Table 1 and Results.
• Example models: Dias_2022_tobramycin_rat.R (male Wistar rats 14 days after intratracheal inoculation of 50 uL of alginate beads carrying P. aeruginosa ATCC 27853; selects the chronic-infection clearance and central volume lcl_chronic / lvc_chronic (0.085 L/h, 0.323 L vs 0.047 L/h, 0.055 L) and the inoculated-lung volume lv_lung_inoc).
• Notes: Ratified 2026-10-02 with DIS_PSEUDOMONAS_LUNG_ACUTE. Distinct from the human cystic-fibrosis indicators (DIS_CF and the cystic-fibrosis lung-disease section), which describe a patient population rather than an induced animal infection: a covariate query pooling them would return preclinical arms to a user asking for CF patient models. The sterile-bead control of the same protocol is ALGINATE_BEAD_BLANK, not 0 on this column alone.

ALGINATE_BEAD_BLANK (canonical for sterile alginate-bead intratracheal control (preclinical) indicator)

• Description: 1 = animal inoculated intratracheally with sterile alginate beads carrying no bacteria, prepared exactly as the beads of a chronic bead-embedded lung-infection model; 0 = otherwise. Time-fixed per animal. The control that separates the effect of the alginate matrix itself from the effect of infection. Mutually exclusive with DIS_PSEUDOMONAS_LUNG_ACUTE and DIS_PSEUDOMONAS_LUNG_CHRONIC.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no blank-bead inoculation).
• Source aliases:
  • blank bead, blank-bead group – Dias 2022 Sections 2.3-2.4 and Table S2.
• Example models: Dias_2022_tobramycin_rat.R (male Wistar rats inoculated with 50 uL of sterile alginate beads; enters only through the inoculated-lung volume lv_lung_inoc, plasma PK being that of healthy rats).
• Notes: Ratified 2026-10-02 with the DIS_PSEUDOMONAS_LUNG_* pair. Deliberately not prefixed DIS_: a sterile-bead animal has no disease, and naming it as one would mislabel the control arm. Named for the material rather than for the route so that an agar-bead control (the other common bead matrix) registers an AGAR_BEAD_BLANK sibling instead of overloading this column.

TE_RESIST_MBV (canonical for treatment-emergent maribavir-resistant CMV mutation indicator)

• Description: 1 = a treatment-emergent cytomegalovirus mutation conferring resistance to maribavir was detected in the subject during the treatment course, 0 = no such mutation detected. Maribavir resistance arises principally through mutations in the viral UL97 kinase (the drug’s target) and, less commonly, UL27. The indicator is treatment-EMERGENT: every subject in the founding cohort was maribavir-susceptible at baseline, so the column records an on-treatment event rather than a baseline characteristic.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no treatment-emergent maribavir-resistance mutation).
• Source aliases:
  • TE CMV mutation conferring resistance to maribavir – printed row label in Sun 2025 Table S3; Treatment-emergent CMV mutation for maribavir resistance in Table S2.
• Example models: Sun_2025_maribavir_cmv_clearance_wk8.R (coefficient -3.05, OR 0.0475), Sun_2025_maribavir_cmv_clearance_wk16.R (coefficient -3.51, OR 0.0299).
• Notes: Because the mutation emerges DURING treatment, this covariate is a partial consequence of the treatment course rather than a pre-treatment stratifier, so a model using it is a descriptive association rather than a prognostic one usable at baseline. Sun 2025 Table S2 records 23 of 238 subjects (9.7%) positive, and CMV resistant at baseline: No 238 (100) confirms the baseline-susceptible starting point. The name carries the drug suffix _MBV because resistance is drug-specific: a CMV isolate resistant to maribavir is frequently still susceptible to ganciclovir or foscarnet and vice versa, so a generic TE_RESIST_CMV would conflate distinct phenotypes. A future paper reporting ganciclovir or letermovir resistance should register TE_RESIST_GCV / TE_RESIST_LMV as siblings. Scope: specific.

CMVDNA_HIGH (canonical for high baseline plasma CMV DNA load indicator)

• Description: 1 = baseline plasma cytomegalovirus DNA concentration in the HIGH stratum, 0 = the very-low or low stratum. Baseline viral load is the standard measure of infection burden at treatment start and a strong determinant of how long clearance takes.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (very low / low baseline CMV DNA, pooled – Sun 2025 Table S3 states the reference as “a very low/low level of CMV DNA at baseline”).
• Source aliases:
  • High CMV DNA level at baseline – printed row label in Sun 2025 Table S3; CMV DNA level at baseline: High in Table S2; High CMV load at baseline vs. low/very low in Table S4.
• Example models: Sun_2025_maribavir_cmv_clearance_wk8.R (coefficient -1.05, OR 0.350 – high baseline viral load roughly a third the odds of confirmed clearance at week 8), Sun_2025_maribavir_renal_disorder.R (coefficient -1.55, OR 0.212).
• Notes: The HIGH threshold is >= 9100 IU/mL, stated in the Sun 2025 Table S3 footnote a; a downstream user applying this column to another cohort must check that the source assay and threshold match, because CMV DNA thresholds are assay- and guideline-dependent. The source’s three-level stratification (very low / low / high; 71, 127 and 39 of 238 subjects respectively per Table S2) is collapsed to a binary in every fitted model, with very-low and low pooled into the reference – so this single indicator carries the whole published covariate and no CMVDNA_LOW sibling is needed for these models. A future paper fitting the three levels separately should register the decomposed set rather than overloading this name. Scope: specific.

CMV_PROPHY_PRIOR (canonical for prior CMV prophylaxis indicator)

• Description: 1 = the subject received antiviral prophylaxis against cytomegalovirus before the current treatment course, 0 = no prior CMV prophylaxis. Prophylaxis (e.g. letermovir, valganciclovir or high-dose valaciclovir after transplant) suppresses early CMV replication and so selects a subject whose infection emerged despite, or after, preventive therapy.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no prior CMV prophylaxis).
• Source aliases:
  • Prior use of CMV prophylaxis – printed row label in Sun 2025 Table S2 and Figure S3 panel b; Prior use of CMV prophylaxis vs. not in Table S4.
• Example models: Sun_2025_maribavir_nausea.R (coefficient -2.98, OR 0.0508, 95% CI 0.00313-0.826, p = 0.0363 – subjects with prior prophylaxis had markedly lower treatment-emergent nausea risk).
• Notes: Only 22 of 238 subjects (9.2%) were positive in the founding cohort, and the fitted odds ratio has a confidence interval spanning more than two orders of magnitude, so the effect size should be treated as poorly determined even though the p-value is nominally significant. The specific prophylactic agent is NOT recorded by the source, so this column cannot be decomposed by drug; a paper that does report the agent should use the CONMED_<INN> family instead of, or in addition to, this composite. Distinct from DIS_CMV (the disease-state indicator marking a transplant recipient with CMV infection) and from TX_HCT. Scope: specific.

Vaccine immunogenicity (placebo-corrected _RATIO family)

Study-arm-level immunogenicity readouts from preventive-vaccine trials, expressed as the ratio of the vaccine arm’s response to the concurrent placebo arm’s response on the original (untransformed) scale. The placebo correction is what makes these comparable across trials that used different assays, laboratories and units: the assay-specific scale cancels in the ratio, so a fold-rise from one sponsor’s platform can be pooled with another’s. A value of 1 means no vaccine-induced rise over placebo and is the natural null for every member of the family.

Members are named <ANALYTE>_RATIO, where <ANALYTE> is the measured immune marker. The family is deliberately distinct from the absolute-concentration canonicals of the same analytes (IFNG in pg/mL, ADA_TITER as a reciprocal dilution): a ratio is unit-free, is defined only relative to a concurrent control arm, and is a study-arm summary rather than a per-subject measurement. It is also distinct from anti-drug-antibody immunogenicity (ADA_POS, ADA_TITER, CONC_ADA_NGML), which measures an undesirable host response against a therapeutic; these columns measure the intended protective response against a pathogen.

Log base is per-model, not per-column. The register fixes the column as the ratio on the original scale; each model applies whatever transform its source specifies and documents it in covariateData[[...]]$notes. The founding paper uses two different bases in one model (base 2 for SNA_RSVA_PBORATIO, natural log for IFNG_PBORATIO), which is exactly why the transform is not baked into the column.

SNA_RSVA_PBORATIO (canonical for placebo-corrected serum neutralizing activity against RSV subtype A)

• Description: Study-arm-level ratio of the vaccine arm’s serum neutralizing activity (SNA) titer against respiratory syncytial virus subtype A to the concurrent placebo arm’s SNA titer, on the original (untransformed) scale, measured at the post-vaccination timepoint the source nominates (about 28 days after a single dose in the founding paper). Unit-free by construction. Not an individual-level measurement.
• Units: (fold; dimensionless ratio)
• Type: continuous
• Scope: specific
• Reference category: n/a – continuous. 1 is the null (equal titers in the vaccine and placebo arms); must be strictly positive, since models take its logarithm.
• Source aliases:
  • RSV-A SNA – Kashihara 2026 Table 2 column header, with footnote b “Immunogenicity values were presented as the ratio to placebo on the original scale.”
  • SNA (RSV-A) – Kashihara 2026 Table 1 “Immunogenicity outcomes” column.
• Example models: Kashihara_2026_rsvVaccine_sna_mbma.R, Kashihara_2026_rsvVaccine_snaCmi_mbma.R (enters as log2(SNA_RSVA_PBORATIO) per Kashihara 2026 Equation 2, driving a severity-specific – respectively common – slope on the negative-effect-size transform of vaccine efficacy).
• Notes: Subtype-qualified because RSV-A and RSV-B neutralizing titers are separate assays whose values are not interchangeable, even though the founding paper found them highly correlated (Spearman 0.89) and used only one at a time to avoid multicollinearity. A SNA_RSVB_RATIO sibling should be registered on the same terms when a model needs it; the founding paper’s RSV-B fit is a sensitivity analysis and was not extracted. Observed range across the seven founding trials is 1.6 to 13.9, so the calibrated span is roughly 1.5-14.

IFNG_PBORATIO (canonical for placebo-corrected vaccine-induced interferon-gamma response)

• Description: Study-arm-level ratio of the vaccine arm’s interferon-gamma (IFN-gamma) response to the concurrent placebo arm’s response, on the original (untransformed) scale, measured at the post-vaccination timepoint the source nominates. A cell-mediated-immunity (CMI) readout, typically an ELISpot or intracellular-cytokine-staining summary rather than a serum concentration. Unit-free by construction. Not an individual-level measurement.
• Units: (fold; dimensionless ratio)
• Type: continuous
• Scope: specific
• Reference category: n/a – continuous. 1 is the null (equal response in the vaccine and placebo arms, which zeroes any natural-log-transformed term); must be strictly positive.
• Source aliases:
  • IFN-g / IFN-gamma – Kashihara 2026 Table 2 column header, with footnote b “Immunogenicity values were presented as the ratio to placebo on the original scale.”
• Example models: Kashihara_2026_rsvVaccine_snaCmi_mbma.R (enters as log(IFNG_PBORATIO) – natural log, per Kashihara 2026 Equation 4 – driving the interferon-gamma slope on RSV-LRTD 3+ only).
• Notes: Not an alias of IFNG, which is an absolute total serum interferon-gamma concentration in pg/mL used for target-mediated-disposition scaling in anti-IFN-gamma antibody models. The two differ in every respect that matters to a user: unit-free ratio vs pg/mL, arm-level vs per-subject, control-referenced vs absolute, and functional T-cell readout vs circulating cytokine burden. A model must not silently substitute one for the other. In the founding paper only four of seven trials had an observed value (VANIR 3.5, D4420C00005 9.0, CYPRESS 12.8, RENOIR 20.6) and the rest were multiply imputed, so the calibrated span is roughly 3.5-20.6. The companion CD4+ T-cell CMI marker was collected in the same trials but excluded from the analysis for sparsity (two trials), so no CD4_RATIO sibling is registered yet.

Oncology

RCFB1MAX (canonical for week-1 maximum across-lesion relative change from baseline in SUVmax)

• Description: Per-subject scalar predictor entering the overall-survival hazard in the Schindler 2016 sunitinib joint SUVmax / SLD / OS-TTE / dropout model. Defined as the maximum (across the up-to-five tracked target lesions) of (SUVmax(t = 168 h) - SUVmax(0)) / SUVmax(0), i.e., the most negative (largest reduction) relative change in [18F]FDG-PET standardized uptake value at one week of sunitinib therapy. The OS Weibull hazard is lambh * alphh * t^(alphh - 1) * exp(theta_pred * RCFB1MAX) with alphh = 1 (degenerates to constant baseline hazard).
• Units: unitless (relative change; typically negative for responders).
• Type: continuous
• Scope: specific
• Reference category: n/a – entered directly into the OS hazard exponent. Sign convention: more negative RCFB1MAX (greater week-1 SUVmax suppression) reduces the OS hazard (Schindler 2016 reports a positive theta_pred = 5.36, so exp(5.36 * RCFB1MAX) is < 1 for negative RCFB1MAX).
• Source aliases:
  • RCFB1MAX (Schindler 2016 NONMEM $ERROR block intermediate; “max relative change in SUVmax from baseline at week 1 across lesions”). Computed inline in the source .mod from the on-the-fly SUVmax states at TIME = 168 h and reused in subsequent records.
• Example models: Schindler_2016_sunitinib.R (DDMODEL00000221).
• Notes: Specific scope because the metric is tied to the Schindler 2016 GIST-on-sunitinib joint biomarker / OS analysis. In the source NONMEM .mod RCFB1MAX is a record-loop state, not a true subject-level covariate – it is captured at FLAG = 1 / TIME = 168 h from the running SUVmax compartment values and reused on every subsequent record. nlmixr2 / rxode2 do not have an idiomatic equivalent of NONMEM’s record-loop persistent state, so the model file consumes RCFB1MAX as a per-subject input covariate; reproducing the source’s behavior requires a two-stage simulation (run the SUVmax + SLD ODEs first, compute RCFB1MAX per subject from the t = 168 h SUVmax values, then run the OS / dropout TTE arms with RCFB1MAX bound). The vignette virtual cohort follows this pattern.

RCFB6_SLD (canonical for the week-6 relative change from baseline in tumor sum-of-longest-diameters)

• Description: Per-subject relative (fractional) change from baseline in tumor size at week 6, i.e. at the end of treatment cycle 2, defined as (SLD(week 6) - SLD(0)) / SLD(0). Negative values are tumor shrinkage. The canonical early-tumor-shrinkage predictor of the drug-independent tumor-growth-inhibition-to-survival modeling framework (Claret 2009 and successors), in which a longitudinal TGI model supplies this single scalar per subject to a parametric survival or PFS model that contains no drug term of its own. Almost always a MODEL-DERIVED quantity taken from the individual predictions of a companion TGI model rather than a directly observed measurement.
• Units: unitless (fraction of baseline)
• Type: continuous
• Scope: general
• Reference category: n/a – entered linearly and uncentred into the log-scale location of an accelerated-failure-time model (or into a hazard exponent). Sign convention: a more negative value (greater shrinkage) lengthens survival, so the fitted coefficient on the AFT log-time scale is negative.
• Source aliases:
  • Fractional change in tumor size at week 6 – Bruno 2012 Table 1 / Table 2 covariate row label.
• Example models: Bruno_2012_capecitabine_docetaxel_os.R, Bruno_2012_capecitabine_docetaxel_pfs.R (founding examples; coefficients -0.801 on log median OS and -1.18 on log median PFS, both in months, with Bruno_2012_capecitabine_docetaxel_tgi.R supplying the value).
• Notes: Because the value is derived from a companion model rather than measured, using a model that consumes it requires a two-stage simulation: solve the TGI model, take the week-6 fractional change per subject, then bind it as a covariate. Distinct from RCFB1MAX (Schindler 2016), which is a week-1 across-lesion maximum relative change in [18F]FDG-PET SUVmax rather than a week-6 change in the RECIST sum of longest diameters, and is scope: specific to that joint model. Scope general because the same end-of-cycle-2 construct recurs across the whole Claret / Bruno TGI-to-survival family. Papers in that family that assess at the end of a longer cycle 2 (week 8) should register a sibling RCFB8_SLD rather than overload this entry, exactly as the MET_GE<n> family handles a different dichotomisation threshold.

MET_GE2 (canonical for baseline number of metastatic sites >= 2 indicator)

• Description: Binary indicator dichotomising the count of baseline metastatic sites at 2: 1 = patient has more than one (i.e. two or more) documented metastatic site at baseline, 0 = patient has one or fewer. Time-fixed per subject. Treated as a prognostic tumor-burden stratifier.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (one or fewer metastatic sites at baseline).
• Source aliases:
  • Number of metastases – Bruno 2012 Table 1 covariate row label for the “>1 versus <=1 metastatic site” split. When a source paper supplies the raw integer count column, derive MET_GE2 = as.integer(NMET >= 2).
• Example models: Bruno_2012_capecitabine_docetaxel_os.R (log-normal accelerated-failure-time effect on log median overall survival; Bruno 2012 enters the covariate on a 1 / 2 code rather than 0 / 1, so the model body writes e_metge2_tmed * (1 + MET_GE2) with e_metge2_tmed = -0.200, reproducing the Table 1 footnote’s “-0.2 x 2 = -0.4 for >1 metastases”).
• Notes: Sibling of MET_GE3 (Terranova 2022) and MET_GE4 (Bruno 2005), registered under the split explicitly anticipated by the MET_GE3 entry’s notes (“If a future paper uses a different split (e.g. >= 2 or >= 5), register a separate sibling (MET_GE2, MET_GE5)”). The >= 2 split is the “>1 versus <=1 metastatic site” dichotomisation Bruno 2012 used. A MET_GE2 = 1 patient may also have LMET = 1 or PERIT_CARC = 1; the columns are not mutually exclusive, and MET_GE2 >= MET_GE3 >= MET_GE4 holds by construction for the same patient.

PFS_EVENT (canonical for progression-free-survival event indicator observed during follow-up)

• Description: Binary indicator of a documented disease-progression event observed during the trial follow-up window (1 = the patient experienced a progression event during follow-up per the study’s response-assessment criteria; 0 = no progression event observed through end of follow-up). Time-fixed per subject in a post-hoc analysis (the indicator is assigned once follow-up is complete). Used by Rozman 2017 as a stratifying covariate on the target-mediated CL decay-rate parameter kdes of rituximab – patients who later experienced progression had 82.2% lower kdes (slower decay of the time-varying CL arm), interpreted mechanistically as sustained CD20 target burden in poor responders. Standard oncology-trial terminology derived from progression-free-survival endpoints (PFS event = first documented disease progression or death from any cause during follow-up).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no progression event observed during follow-up).
• Source aliases:
  • Rozman 2017 uses the prose descriptor “disease progression” / “with disease progression” / “no disease progression” throughout Methods / Results / Table 2; the paper does not disclose the NONMEM $INPUT column name. The canonical PFS_EVENT follows the widely used oncology-trial convention.
• Example models: Rozman_2017_rituximab.R (multiplicative effect on the time-varying CL decay-rate kdes: kdes = exp(lkdes + e_pfs_event_kdes * PFS_EVENT + etalkdes) with e_pfs_event_kdes = log(1 - 0.822) = log(0.178), reproducing the paper’s 82.2% reduction from 0.143/day in non-progressors to 0.0254/day in progressors per Rozman 2017 Table 2).
• Notes: This is an outcome-derived covariate: the indicator is assigned from longitudinal response-assessment data (typically CT / PET imaging plus clinical follow-up per the study’s response criteria) and then used post-hoc as a stratifying covariate on a PK parameter. That is unusual but is a legitimate published modeling choice; models using PFS_EVENT should document the post-hoc nature and the associated selection-bias caveat in the vignette Assumptions and deviations section. Time-fixed by construction. If a downstream paper treats progression as a time-varying Markov state (progression-free at time t vs. progressed by time t), register a distinct time-varying canonical rather than reusing PFS_EVENT. Distinct from LMET (baseline liver metastases – a baseline burden indicator, not a follow-up outcome) and from TUMSZ / TUM_SLD (baseline continuous tumor-size measures).

TUMSZ (canonical for baseline tumor size)

• Description: Baseline tumor size. For solid tumors, the sum of diameters of target lesions per RECIST; for classical Hodgkin lymphoma and lymphoma generally, the sum of products of perpendicular diameters (SPPD) or the sum of linear diameters of target lesions, depending on the source paper.
• Units: mm (for linear-diameter constructs); mm^2 (for SPPD constructs; record the per-model convention in covariateData[[TUMSZ]]$units and notes).
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (TUMSZ / ref)^exponent for continuous effects, or with a paper-specific threshold for categorical-stratum indicators (e.g., Gibiansky 2014 splits BSIZ at 1750 mm^2 into low- vs high-burden strata). Reference values observed: 41 mm (Zhou 2025); 63 mm (Budha 2023); 90 mm (Lu 2014, source reference 9 cm converted to mm); 1750 mm^2 threshold (Gibiansky 2014, SPPD).
• Source aliases:
  • LDIAM (Zhou 2025; pediatric lymphoma “linear diameter” of target lesions in mm).
  • TMBD (originally in cm; TUMSZ_mm = TMBD_cm * 10) – used in Lu_2014_trastuzumabemtansine.R.
  • BSIZ (Gibiansky 2014; baseline tumor size as the sum of products of perpendicular diameters of target lesions, mm^2; used as the categorical indicator (BSIZ <= 1750) in the obinutuzumab popPK model rather than as a continuous power covariate).
  • SIZE (He 2025; preoperative largest-lesion diameter of a liver tumor by imaging, originally in cm, converted to mm on ingestion). Not a RECIST sum of diameters – a single-lesion diameter used as an indirect surrogate for the remaining functional liver mass, so TUMSZ (the pooled tumor-burden register) applies rather than TUM_SLD.
• Example models: Budha_2023_tislelizumab.R (reference 63 mm), Gibiansky_2021_rituximab.R (CLL; source column BSIZ in mm^2, reference 7000 mm^2; power exponent 0.261 on the time-dependent target-mediated clearance component cl_exp_component; the paper’s missing-value codes -99 and 0 map to the reference value via if (TUMSZ <= 0)), He_2025_lidocaine.R (reference 50 mm, the Table 1 median 5 cm converted to mm; negative exponent -0.382 on lidocaine clearance, i.e. clearance falls as tumor size rises – the only extraction so far where tumor size acts as a liver-function proxy in a non-oncology-therapy popPK model), Lu_2014_trastuzumabemtansine.R (reference 90 mm; source column TMBD in cm, values converted to mm on ingestion), Zhou_2025_brentuximab.R (reference 41 mm; source column LDIAM is the sum of linear diameters of target lesions; effect on ADC clearance only), Gibiansky_2014_obinutuzumab.R (SPPD in mm^2; used as a categorical indicator (TUMSZ <= 1750) on the time-dependent clearance decay rate kdes, not as a continuous power covariate), Hansson_2013b_sunitinib.R (DDMODEL00000198; observed baseline tumor SLD used as the per-subject IC of the tumor-size ODE via the IPP-style proportional baseline-residual construction tumor(0) = TUMSZ * (1 + etaibase * propSd); the source .mod reads OBASE from DV at TIME=0/FLAG=4, but nlmixr2 / rxode2 cannot replicate the in-record assignment idiom and consumes the observed baseline as a covariate instead).
• Notes: Promoted to scope: general on 2026-04-20 as a conventional oncology baseline-tumor-size measure (RECIST for solid tumors, SPPD or sum-of-linear-diameters for lymphomas). The SPPD vs sum-of-diameters vs sum-of-linear-diameters convention is pooled onto a single column; document the per-model mixture where relevant. When the source paper reports tumor size in cm, convert to mm (the canonical unit) on data ingestion and scale the per-model reference accordingly so (TUMSZ / ref)^exp is numerically invariant. For SPPD constructs the natural unit is mm^2 (a product of two perpendicular diameters in mm); record that in the per-model covariateData[[TUMSZ]]$units field and do NOT cross-mix mm and mm^2 within a single ingest. When a source paper specifically reports the RECIST 1.1 “sum of longest diameters” of target lesions, use the more specific TUM_SLD canonical instead – TUMSZ remains the pooled-tumor-burden register.

TUM_SLD (canonical for sum of longest diameters of target lesions)

• Description: Baseline sum of longest diameters of target lesions per RECIST 1.1. More specific than the pooled TUMSZ canonical; use TUM_SLD when the source paper explicitly reports “sum of longest diameters” (or “sum of lesions”) as the tumor-burden metric, distinct from the pooled “sum of diameters / SPPD / sum of linear diameters” mixture covered by TUMSZ.
• Units: mm
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (TUM_SLD / ref)^exponent. Reference values observed: 70.0 mm (de Vries Schultink 2020 zenocutuzumab population median).
• Source aliases:
  • SoL / “sum of lesions” (de Vries Schultink 2020 zenocutuzumab) – same construct, mm.
• Example models: deVriesSchultink_2020_zenocutuzumab.R (reference 70.0 mm; power exponent 0.447 on Vmax of the parallel non-linear / Michaelis-Menten clearance).
• Notes: Distinct from TUMSZ (pooled tumor-size canonical covering RECIST sum-of-diameters / SPPD / sum-of-linear-diameters); TUM_SLD is the precise RECIST 1.1 sum-of-longest-diameters metric. When the source paper reports tumor burden in cm, convert to mm (the canonical unit) on data ingestion and scale the per-model reference accordingly so (TUM_SLD / ref)^exp is numerically invariant. Also used as the per-subject initial-condition input for tumour-growth / angiogenesis-inhibition (TGI) ODE models where the source paper sets the SLD state at time zero from the observed baseline SLD rather than estimating a typical-value baseline (e.g., Ouerdani_2015_pazopanib.R uses tumorSize(0) <- TUM_SLD).

TUM_VOL (canonical for volumetric tumor burden)

• Description: Tumor burden expressed as a volume, pooling every measurement modality that yields one. Two established sources: (a) baseline tumor volume measured by handheld caliper in subcutaneous-xenograft preclinical studies and similar small-animal models, computed from longest length and orthogonal width via volume = (length * width^2) / 2 (the standard ellipsoid-approximation formula used in xenograft pharmacology); and (b) clinical total tumor volume delineated on a tomographic image (CT, or the CT component of a SPECT/CT or PET/CT acquisition) and summed over all lesions. Used both as a covariate stratifier (size at randomisation, or tumor burden as a covariate on an uptake parameter) and as the per-subject initial-condition input for TGI ODE models where the source paper sets the tumor-volume state at time zero from the observed measurement rather than estimating a typical-value baseline. Usually a per-subject baseline held constant, but may be time-varying when the source re-measures it per treatment cycle.
• Units: mm^3
• Type: continuous
• Scope: general
• Reference category: n/a – typically used directly as the ODE initial state (tumorSize(0) <- TUM_VOL); not a covariate effect coefficient.
• Source aliases:
  • P0 (Ouerdani 2015 mouse model symbol for the observed initial tumour volume; mm^3 in the CAKI-2 xenograft cohort; range 100-250 mm^3 at randomisation per the paper’s preclinical Methods).
  • total tumor volume (Siebinga 2023 clinical CT-delineated volume summed over all PSMA-positive lesions; reported in mL, so multiply by 1000 to reach the canonical mm^3).
• Example models: Ouerdani_2015_pazopanib_mouse.R (Ouerdani 2015 preclinical TGI in CAKI-2 xenograft mice; per-subject TUM_VOL initialises the tumorSize state and is held constant per individual across the 24-day dosing window), Siebinga_2023_lu177psma617.R (clinical CT-delineated total tumor volume as a power covariate on the tumor uptake rate constant, (TUM_VOL / 1730)^0.705; re-measured before each treatment cycle).
• Notes: General scope because tumor-volume measurements have a shared meaning across papers regardless of drug, cell line, or measurement modality. Carried in mm^3 regardless of the unit the source reports (clinical papers usually report mL; 1 mL = 1000 mm^3), so convert on data ingestion and scale any per-model reference accordingly to keep (TUM_VOL / ref)^exp numerically invariant. Use TUM_SLD for clinical RECIST sum-of-longest-diameters (a length, not a volume) and TUMSZ for the pooled “baseline tumor burden as a covariate on PK” use case where the source does not resolve to a volume. The promotion follows the CRCL precedent, where creatinine-based and tracer-measured GFR are pooled onto one canonical because they play the same operational role. For clinical functional-imaging-segmented tumor volumes where a paper reports BOTH the whole-body burden and the segmented target-lesion subset separately, use the finer-grained siblings TUM_VOL_TOTAL / TUM_VOL_TARGET below instead, so the two volumes stay distinguishable; TUM_VOL remains correct when a paper reports a single volumetric tumor burden with no target-lesion subset.

TUM_VOL_TOTAL (canonical for total clinical imaging-segmented tumor volume over all lesions)

• Description: Total tumor volume summed over ALL identified lesions, delineated on clinical functional imaging (PET/CT, SPECT/CT) rather than by caliper. Represents whole-body tumor burden as a volume. The founding example uses semi-automatic threshold segmentation at 50% SUVmax on [68Ga]Ga-HA-DOTATATE PET/CT. Per-subject baseline measurement, held constant per individual.
• Units: mL
• Type: continuous
• Scope: general
• Reference category: n/a – the founding example uses it uncentred in an exponential tumor-sink term, kin_spleen = kin_spleen_pop * exp(0.4 * (-TUM_VOL_TOTAL / 1000)), with the volume converted to litres. Population median observed: 283 mL (Siebinga 2023 EJNMMI Phys Table 1, range 22.4-644).
• Source aliases:
  • V tumor total, i / “total tumor volume” (Siebinga 2023 EJNMMI Phys, Equation 4 and Table 1) – same construct, mL, no transformation.
  • Baseline metabolic tumour volume (Mueller-Schoell 2021 Supplementary Table S1) – CD19+ metabolic tumour volume by [18F]FDG PET-CT, mL, used as an ODE initial condition rather than in a covariate term.
• Example models: Siebinga_2023_ga68hadotatate.R, Siebinga_2023_lu177hadotatate.R (drives the tumor-sink effect on spleen uptake in the six-compartment semi-physiological HA-DOTATATE radioligand models), MuellerSchoell_2021_axicabtageneCiloleucel_qsp.R (the per-patient baseline CD19+ metabolic tumour volume is the initial condition of the CD19+ tumour state, tumor(0) <- TUM_VOL_TOTAL, and drives the CAR-T expansion Michaelis-Menten terms; cohort median 85.7 mL).
• Notes: Distinct from TUM_VOL (preclinical caliper volume in mm^3, whose own notes redirect clinical models elsewhere) and from TUMSZ / TUM_SLD (both LENGTHS in mm, not volumes). Distinct from its sibling TUM_VOL_TARGET, which covers only the segmented target-lesion subset; the distinction is load-bearing whenever a model uses both, because the whole-burden volume and the modelled-compartment volume drive different effects with different coefficients. Canonical unit is mL because clinical functional imaging reports volumes in mL; when a source reports cm^3 the values are numerically identical, and when a source reports mm^3 divide by 1000 on ingestion. Note that Siebinga_2023_lu177psma617.R pre-dates this entry and carries its clinical total tumor volume in TUM_VOL (converted to mm^3); new clinical extractions should use TUM_VOL_TOTAL.

TUM_VOL_TARGET (canonical for clinical imaging-segmented volume of the target lesions only)

• Description: Tumor volume summed over the segmented TARGET lesions only – the protocol-defined subset of lesions that a model carries as its tumor compartment – delineated on clinical functional imaging (PET/CT, SPECT/CT). In the founding example the target lesions are those with a diameter above 2 cm, to a maximum of five segmented lesions and two per organ system per patient, segmented at a 50% SUVmax threshold. Per-subject baseline measurement, held constant per individual.
• Units: mL
• Type: continuous
• Scope: general
• Reference category: n/a – the founding example uses it in a power term referenced to the cohort median, (TUM_VOL_TARGET / 80.0)^eff, and ALSO as the tumor compartment volume itself (v_tumor = TUM_VOL_TARGET / 1000 L). Population median observed: 80.0 mL (Siebinga 2023 EJNMMI Phys Table 1, range 7.81-212).
• Source aliases:
  • V tumor cmt, i / “tumor volume of target tumors (representing the tumor compartment)” (Siebinga 2023 EJNMMI Phys, Equation 5 and Table 1) – same construct, mL, no transformation.
• Example models: Siebinga_2023_ga68hadotatate.R, Siebinga_2023_lu177hadotatate.R (sets the tumor compartment volume and drives the power effect of tumor burden on tumor uptake in the six-compartment semi-physiological HA-DOTATATE radioligand models).
• Notes: A strict subset of TUM_VOL_TOTAL; register both when a paper reports both, because a model may legitimately use the target-lesion volume as the modelled compartment volume while using the whole-body burden to drive a sink effect on healthy tissue. Where the model uses this column as a compartment volume, a concentration-valued binding capacity automatically scales the compartment’s capacity with tumor size – record that dependency in the per-model covariateData notes so a downstream user does not double-count it. Distinct from TUM_SLD (RECIST sum of longest diameters, a length in mm) even when both describe target lesions: TUM_SLD is a length and TUM_VOL_TARGET is a volume, and they are not interconvertible without a shape assumption.

TUM_BURDEN_HIGH (canonical for high baseline tumour-burden indicator)

• Description: 1 = the subject had HIGH tumour burden at baseline by the source’s composite disease-specific criteria, 0 = low or intermediate tumour burden. A categorical counterpart to the continuous TUM_* burden measures, for analyses that stratify burden by clinical thresholds rather than carry a measured quantity.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (low or intermediate tumour burden, pooled – Hibma 2026 Figure 6 caption: ‘reference tumor burden (i.e., low/intermediate)’).
• Source aliases:
  • TumorBurden High – printed row label in Hibma 2026 ESM Table S2; High under the Tumor burden, n (%) block of Table 1.
• Example models: Hibma_2026_elranatamab_crs.R (coefficient 1.14, OR 3.11 – high baseline tumour burden roughly triples the odds of any-grade cytokine release syndrome after the first elranatamab step-up dose, in a landmark binomial logistic model alongside log(CTROUGH)).
• Notes: The threshold definition is disease-specific and must be recorded in each model’s covariateData[[TUM_BURDEN_HIGH]]$notes. Founding definition (Hibma 2026 Table 1 footnote, relapsed/refractory multiple myeloma): high = bone-marrow plasma-cell infiltrate >= 80% OR serum M-spike >= 5 g/dL OR serum free light chain >= 5000 mg/L; low = infiltrate < 50% AND M-spike < 3 g/dL AND free light chain < 3000 mg/L; subjects meeting neither are intermediate. General scope because high baseline tumour burden is a recurring cytokine-release / neurotoxicity covariate for T-cell-engaging therapies across tumour types; a downstream user applying this column to another cohort must check that the definition matches. The source’s three-level stratification is collapsed to a binary with low and intermediate pooled, so no TUM_BURDEN_LOW sibling is needed; a future paper fitting the levels separately should register the decomposed set rather than overloading this name. Distinct from TUMTP_MYELO (tumour TYPE) and MM_NIGG (myeloma immunoglobulin type), and from the continuous burden measures TUM_SLD, TUM_VOL, TUM_CELLS0 and CTDNA.

ORGVOL_KIDNEY (canonical for measured total kidney volume)

• Description: Measured total volume of the kidneys (both kidneys combined unless the model states otherwise), typically delineated on CT or the CT component of a SPECT/CT or PET/CT acquisition.
• Units: mL
• Type: continuous
• Scope: general
• Reference category: n/a – used directly as a physiologic input that sets a compartment’s sub-volumes, its perfusion and its permeability-surface-area product; not a covariate-effect coefficient.
• Source aliases:
  • Kidneys (measured volume) – Kletting 2016 J Nucl Med Table 1 column header; reported in mL, no value transformation.
• Example models: Budiansah_2025_dotatate_pbpk.R, Golzaryan_2025_lu177psmaIT_pbpk.R.
• Notes: Founding member of the ORGVOL_<ORGAN> family, whose members carry a per-patient measured organ volume that a physiologically-based model uses to build that organ’s sub-volumes rather than scaling it allometrically from body size. In the founding model the measured volume is split into vascular (5.5 %), interstitial (15 %) and intracellular (two thirds of the remainder) sub-volumes and multiplies the fitted sst2 binding-site density. Cohort range 125-233 mL. Distinct from KBF (renal blood flow, mL/min) – that is a flow, this is a volume.

ORGVOL_LIVER (canonical for measured total liver volume)

• Description: Measured total liver volume, typically delineated on CT or the CT component of a SPECT/CT or PET/CT acquisition.
• Units: mL
• Type: continuous
• Scope: general
• Reference category: n/a – used directly as a physiologic input that sets a compartment’s sub-volumes, its perfusion and its permeability-surface-area product; not a covariate-effect coefficient.
• Source aliases:
  • Liver (measured volume) – Kletting 2016 J Nucl Med Table 1 column header; reported in mL, no value transformation.
• Example models: Budiansah_2025_dotatate_pbpk.R.
• Notes: Member of the ORGVOL_<ORGAN> family; see ORGVOL_KIDNEY for the family rationale. In the founding model the measured volume is split into vascular (8.5 %) and interstitial (20 %) sub-volumes and multiplies the fitted hepatic sst2 binding-site density. Cohort range 1500-4876 mL, the upper end being a patient whose liver was largely replaced by neuroendocrine metastases. Distinct from LIVER_VOL_FRAC or any tumour-burden covariate: this is the whole organ including any intrahepatic disease.

ORGVOL_SPLEEN (canonical for measured total spleen volume)

• Description: Measured total spleen volume, typically delineated on CT or the CT component of a SPECT/CT or PET/CT acquisition.
• Units: mL
• Type: continuous
• Scope: general
• Reference category: n/a – used directly as a physiologic input that sets a compartment’s sub-volumes, its perfusion and its permeability-surface-area product; not a covariate-effect coefficient.
• Source aliases:
  • Spleen (measured volume) – Kletting 2016 J Nucl Med Table 1 column header; reported in mL, no value transformation.
• Example models: Budiansah_2025_dotatate_pbpk.R.
• Notes: Member of the ORGVOL_<ORGAN> family; see ORGVOL_KIDNEY for the family rationale. In the founding model the measured volume is split into vascular (12 %) and interstitial (20 %) sub-volumes and multiplies the fitted splenic sst2 binding-site density. The spleen is the highest-density normal-tissue somatostatin-receptor organ, so its volume is a material determinant of whole-body peptide distribution. Cohort range 110-320 mL. Splenectomised patients contribute no spleen observations; set the volume to a small positive placeholder rather than zero so that the sub-volume divisions in the ODEs stay finite, and exclude the spleen output from those subjects.

CTDNA (canonical for baseline circulating tumor DNA burden)

• Description: Baseline (pre-treatment, cycle 1 day 1) circulating tumor DNA burden in plasma, quantified by next-generation sequencing as the average number of mutant tumor molecules per millilitre of plasma (MMPM). ctDNA is shed into the circulation when tumor cells die by apoptosis or necrosis, so MMPM acts as a liquid-biopsy surrogate for total tumor burden that is independent of the RECIST target-lesion selection captured by TUM_SLD.
• Units: MMPM (mutant molecules per mL of plasma)
• Type: continuous
• Scope: general
• Reference category: n/a – used as the per-subject regressor / ODE initial-condition input rather than as a covariate effect coefficient. Models that fit ctDNA on the base-10 logarithmic scale (the common convention, because raw MMPM spans several orders of magnitude) derive the state initial condition inside model as rbase_ctdna <- log10(CTDNA); the register stores the untransformed MMPM value so the transformation is visible at the call site.
• Source aliases:
  • y0 (Ribba 2022 Eq. 1 symbol for the observed baseline used as a Monolix regressor).
  • ctDNA0 (Ribba 2022 Eq. 2 symbol for the same quantity in the joint ctDNA / SLD model).
• Example models: Ribba_2022_ctdna.R (Stein bi-exponential on log10 ctDNA; growth_ctdna(0) <- log10(CTDNA)), Ribba_2022_ctdna_sld_joint.R (joint ctDNA / SLD model; same initial-condition use alongside TUM_SLD).
• Notes: Deliberately NOT pooled with variant-allele-frequency (VAF) or ctDNA-tumor-fraction (cTF) readouts: MMPM is an absolute concentration of mutant molecules whereas VAF and cTF are dimensionless ratios of mutant to wild-type (or aneuploidy-derived) signal, so the two are not interconvertible without the wild-type denominator. A future VAF / cTF canonical should be registered separately (e.g. CTDNA_VAF) rather than aliased onto CTDNA. The assay platform matters for cross-study pooling – Ribba 2022 used the Roche AVENIO panel for the MMPM cohorts (Weber 2021 and OAK) and the FMI panel for the cTF cohort (IMspire170) – so record the panel in the per-model covariateData[[CTDNA]]$notes.

TUMTP_HODGKIN_CLASSICAL (canonical for classical Hodgkin lymphoma tumor-type indicator)

• Description: 1 = classical Hodgkin lymphoma (cHL) or Hodgkin lymphoma generally, 0 = other tumor types.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 = all other tumor types (e.g., NSCLC, EC, HCC, UC, GC, CRC, NPC, OC, “Other” solid tumors in the Budha 2023 cohort; systemic anaplastic large-cell lymphoma in the Zhou 2025 pediatric cohort).
• Source aliases:
  • TUMTP_CHL – prior canonical name (pre-2026-06-19 standardization audit).
  • TUMTP (categorical column with levels like cHL, GC,…) – decompose into TUMTP_HODGKIN_CLASSICAL = as.integer(TUMTP == "cHL").
  • DIS (Zhou 2025; integer code with DIS == 1 flagging HL) – decompose into TUMTP_HODGKIN_CLASSICAL = as.integer(DIS == 1). Zhou 2025 calls the complement “non-HL”; in the Zhou 2025 cohort the non-HL group is exclusively sALCL.
• Example models: Budha_2023_tislelizumab.R, Zhou_2025_brentuximab.R (effects on ADC Q2, MMAE central volume VM, and the ADC->MMAE conversion-decay rate ALFM; the Zhou 2025 paper anchors typical-value parameters to HL patients so the model uses (1 - TUMTP_HODGKIN_CLASSICAL) as the on-effect indicator with reference category 1 = HL), Zhao_2025_nivolumab.R (exponential coefficient -0.330 on baseline CL, i.e. CL about 28% lower than the second-line-NSCLC reference – the largest single covariate effect in that model; decomposed from the source’s categorical TTYPEN2 level 10, and the only tumor-type indicator retained alongside TUMTP_GASTRIC).
• Notes: Paired with TUMTP_GASTRIC in Budha 2023; a patient can have at most one of the indicators set to 1 (the remaining tumor types collapse into the reference 0 group). The reference category is the off-encoded value (0) by definition; when a source paper anchors typical-value parameters to the HL group rather than the non-HL group (as Zhou 2025 does), encode the effect as coef^(1 - TUMTP_HODGKIN_CLASSICAL) so the canonical column meaning (1 = cHL/HL) is preserved while the paper’s reference (HL) still receives multiplier 1.

HER2_ECD (canonical for HER2 shed extracellular domain concentration)

• Description: Baseline serum concentration of the shed extracellular domain of human epidermal growth factor receptor 2 (HER2). Serves as a soluble-antigen biomarker of HER2-mediated target-mediated drug disposition for HER2-directed mAbs / ADCs.
• Units: ng/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – used with power scaling (HER2_ECD / ref)^exponent. Reference 25 ng/mL used in Lu 2014; reference 8.23 ng/mL (population median) used in Bruno 2005, with a 200 ng/mL plateau cap.
• Source aliases:
  • ECD – used in Lu_2014_trastuzumabemtansine.R and Bruno_2005_trastuzumab.R.
• Example models: Lu_2014_trastuzumabemtansine.R (reference 25 ng/mL; exponent 0.035 on CL), Bruno_2005_trastuzumab.R (reference 8.23 ng/mL; power exponents 0.041 on CL and 0.105 on V; HER2_ECD capped at 200 ng/mL inside model() to reflect the Bruno 2005 plateau observation), deVriesSchultink_2018_trastuzumab_LVEF.R (inherits the Bruno 2005 trastuzumab popPK as a deterministic forcing function with the same 8.23 ng/mL reference and 200 ng/mL plateau cap; PK IIV is omitted per the source paper’s ‘fixed effect parameters’ clause, so HER2_ECD only enters via the typical-value PK covariate equations).
• Notes: Scoped specific because the covariate is meaningful only for HER2-targeted agents; if a non-HER2 paper uses a shed-antigen analog for a different target, register a target-specific canonical (e.g., EGFR_ECD) rather than reusing this one. Disambiguated from the covariate-columns register by the explicit HER2_ prefix.

TRAST_BL (canonical for baseline trastuzumab concentration from prior therapy)

• Description: Baseline serum concentration of residual unconjugated trastuzumab remaining from prior trastuzumab-containing therapy, measured at the start of a subsequent anti-HER2 treatment (e.g., trastuzumab emtansine). Encodes the magnitude of residual HER2-site competition from a previous trastuzumab exposure.
• Units: ug/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – used as a linear covariate on log(CL) via exp(coef * TRAST_BL). TRAST_BL = 0 corresponds to no detectable residual trastuzumab (reference condition used in Lu 2014).
• Source aliases:
  • TBL – used in Lu_2014_trastuzumabemtansine.R.
• Example models: Lu_2014_trastuzumabemtansine.R (linear-on-log coefficient -0.002 per ug/mL on CL).
• Notes: Scoped specific because the covariate is meaningful only for drugs that compete with trastuzumab at the HER2 binding site. Expect values clustered at 0 for trastuzumab-naive patients; Lu 2014 observed 0 at the 5th percentile and 54 ug/mL at the 95th percentile.

TROPONIN_T_MAX (canonical for peak post-anthracycline high-sensitive serum troponin T concentration)

• Description: Peak (maximum) high-sensitive serum troponin T concentration observed per subject during or after the anthracycline-containing portion of an adjuvant or neoadjuvant chemotherapy regimen, in ng/L. Used as a covariate on subsequent trastuzumab-induced LVEF decline (the peak captures the cumulative anthracycline-driven myocyte damage that sensitises the heart to subsequent HER2-directed therapy).
• Units: ng/L
• Type: continuous
• Scope: specific
• Reference category: n/a – median-centered power covariate in de Vries Schultink 2018: EC50_i = EC50_pop * (TROPONIN_T_MAX / 18)^(-1.16), where 18 ng/L is the cohort-median TROPONIN_T_MAX. A TROPONIN_T_MAX value at the cohort median leaves EC50 unchanged from its population value.
• Source aliases:
  • TRPmax – used in deVriesSchultink_2018_trastuzumab_LVEF.R.
• Example models: deVriesSchultink_2018_trastuzumab_LVEF.R (power covariate effect on EC50 with median-centered normalisation, exponent -1.16, explaining 15.1% of inter-individual variability in EC50; per de Vries Schultink 2018 Table 2 and the EC50 covariate equation in Results).
• Notes: Derived per-subject from the upstream K-PD anthracycline-troponin T model output (see deVriesSchultink_2018_anthracycline_troponinT.R). Treated as a time-fixed baseline covariate for the LVEF model: even though the troponin T peak occurs during anthracycline treatment, by the time the LVEF observations start (typically a few weeks after the last anthracycline dose), the peak is a determined historical scalar per subject. Specific scope until a second cardiotoxicity / cardiac-biomarker model legitimately reuses this indicator. Values are positive; sub-LLOQ peak values are imputed at LLOQ/2 (= 1.5 ng/L for the Roche Modular E hs-TnT assay used in de Vries Schultink 2018).

TUMTP_GASTRIC (canonical for gastric-cancer tumor-type indicator)

• Description: 1 = gastric cancer (GC) or adenocarcinoma of the gastroesophageal junction (GEJ), 0 = other tumor types.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 = all other tumor types (same reference group as TUMTP_HODGKIN_CLASSICAL).
• Source aliases:
  • TUMTP_GC – prior canonical name (pre-2026-06-19 standardization audit).
  • TUMTP (categorical column) – decompose into TUMTP_GASTRIC = as.integer(TUMTP == "GC").
  • TTYPE (Quartino 2019; categorical column with levels MBC, EBC, HV, AGC, Others) – decompose into TUMTP_GASTRIC = as.integer(TTYPE == "AGC").
  • TTYPE4 (Wang 2024; level 4 of a five-level tumor-type factor labelled “GCGEJ” in the source) – decompose into TUMTP_GASTRIC = as.integer(TTYPE4 == 1).
• Example models: Budha_2023_tislelizumab.R, Quartino_2019_trastuzumab.R (advanced gastric cancer; per-group typical-value switch on linear CL and Vc rather than an exponential multiplier), Wang_2024_sugemalimab.R (gastric + GEJ adenocarcinoma pooled; exponential coefficient log(1.13) on CL and log(1.14) on Vc), Zhao_2025_nivolumab.R (exponential coefficient 0.180 on baseline CL, i.e. CL about 20% higher than the second-line-NSCLC reference; decomposed from the source’s categorical TTYPEN2 level 7, in which every tumor type other than gastric cancer and classical Hodgkin lymphoma is collapsed to an “other” level that carries no retained effect; gastric cancer is 1.6% of the 3554-subject pooled cohort), Gao_2026_trastuzumabRezetecan.R (ADDITIVE shift of -0.129 1/day on the trastuzumab-rezetecan payload release-rate constant RAT, applied before the body-weight and tumor-size power terms; the reference group is NSCLC, not a pooled ‘other’ bucket; source column CT2, defined as gastric cancer or gastroesophageal junction adenocarcinoma).
• Notes: Follows the RACE_<GROUP> indicator-decomposition pattern. New oncology tumor types should be added as additional TUMTP_<GROUP> entries so the reference set stays explicit. “Advanced gastric cancer” (AGC), “gastric cancer” (GC), and “GC or adenocarcinoma of the gastroesophageal junction” (GCGEJ) are pooled onto a single TUMTP_GASTRIC indicator; document the per-paper stage-of-disease and GEJ-inclusion detail in covariateData[[TUMTP_GASTRIC]]$notes. ESCC (squamous histology) is captured by the separate TUMTP_ESCC indicator and is not pooled here.

TUMTP_OTHER (canonical for ‘other tumor types’ residual indicator)

• Description: 1 = heterogeneous “other” tumor-type pool (typically NSCLC plus miscellaneous solid tumors such as prostate, ovarian, and colorectal), 0 = one of the named tumor-type groups in the same analysis.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 = named tumor-type groups defined per-paper (e.g., MBC, EBC, HV, AGC in Quartino 2019). The complement of all TUMTP_<GROUP> indicators defined in the same model.
• Source aliases:
  • TUMTP_OTH – prior canonical name (pre-2026-06-19 standardization audit).
  • TTYPE (Quartino 2019) – decompose into TUMTP_OTHER = as.integer(TTYPE == "Others").
  • PAT2 (Sathe 2024, Sathe 2025) – integer-coded tumor type column with levels 1 (mTNBC), 2 (mUC or HR+/HER2- mBC), 4 (Other epithelial); the source NONMEM control stream collapses PAT2 = 1 and PAT2 = 2 into the reference (no effect) and applies the deviation only when PAT2 = 4. Decompose into TUMTP_OTHER = as.integer(PAT2 == 4).
  • tumtyp (Abegesah 2025) – integer-coded tumor type column whose level 1 the paper never labels (Table 4 names the row only “Tumor type 1 on CL” and the Fig. 1 tornado plot omits it, unlike the levels 2 and 3 bars labelled “TUM TYP Bladder” and “TUM TYP BTC”). Decompose into TUMTP_OTHER = as.integer(tumtyp == 1).
• Example models: Hu_2024_nivolumab.R (exponential effect on nivolumab baseline CL crossed with age band: adults exp(0.00699), adolescents 12-17 y exp(-0.223), young pediatric < 12 y exp(-0.580), relative to an adult-melanoma reference; adults additionally carry an Emax offset of +0.118), Hu_2024_ipilimumab.R (exponential effect on ipilimumab CL: adults exp(-0.698), pediatric subjects < 18 y exp(-0.462)), Quartino_2019_trastuzumab.R (per-group typical-value switch on linear CL; NSCLC plus a small residual group of prostate, ovarian, and other histologies), Sathe_2024_sacituzumab.R (multiplicative effect on tAB CL: -13.4% when TUMTP_OTHER = 1; “Other” pool = small-cell and non-small-cell lung cancer, colorectal cancer, esophageal cancer, pancreatic ductal adenocarcinoma, etc., n = 184; reference = pooled mTNBC + mUC + HR+/HER2- mBC, n = 345), Sathe_2025_sacituzumab.R (updated 3-study pooled analysis of the same drug adding TROPiCS-02 HR+/HER2- mBC data to Sathe 2024; multiplicative effect on tAB CL: -11.2% when TUMTP_OTHER = 1; “Other” pool composition and n = 184 unchanged from Sathe 2024; reference = pooled mTNBC + mUC + HR+/HER2- mBC, n = 605), Lacy_2018_cabozantinib.R (multiplicative fractional effect on CL/F = +0.178 and on Vc/F = -0.186 for “other malignancies” relative to the healthy-volunteer reference; n = 40 / 1534 = 2.6% of the cohort, all from Study XL184-001 mixed-malignancy first-in-human cohort), Wang_2024_sugemalimab.R (multiplicative effect exp(e_oth_cl * TUMTP_OTHER) on CL and exp(e_oth_vc * TUMTP_OTHER) on Vc, with the “Other” pool = the residual tumor-type bucket complementing TUMTP_LYMPH / TUMTP_GASTRIC / TUMTP_ESCC in the same model), Gastonguay_2005_efaproxiral.R (categorical power-model multiplier on SLPp50 = 0.854 for “other” cancer type vs lung-cancer reference; CATP level 4; cancer types not in {lung, breast, GBM} pooled into “other”, n = 72 / 451 = 16.0% of cohort), Yang_2026_encorafenib.R (linear categorical multiplier 1 + e_tumtp_other_cl * TUMTP_OTHER on the day-1 apparent clearance CL/F, coefficient -0.0938 relative to a melanoma reference; “other” pool = healthy participants n = 15 plus lung / NSCLC n = 98 plus other solid tumors n = 65, total n = 178 / 1310 = 13.6% of the cohort – note this is the rare case where the pool mixes healthy volunteers into a tumor-type indicator), Abegesah_2025_durvalumab.R (multiplicative effect on durvalumab CL of (1 - 0.0101); the one registered use where the pool composition is INFERRED rather than stated – see the Notes below), Wang_2026_serplulimab.R (selects the ‘other tumour types’ stratum of a seven-histology joint fit: baseline CL0 0.211 L/day and Vc 3.19 L, the highest clearance and therefore the lowest exposure of the seven; the pool complements Sq. NSCLC / non-sq. NSCLC / SCLC / ESCC / HCC / CRC and is populated mainly by the MSI-H/dMMR solid-tumour, cervical-cancer and head-and-neck cohorts; n = 116 / 2110 = 5.50% of the cohort), Gao_2026_trastuzumabRezetecan.R (ADDITIVE shift of +0.0464 1/day on the trastuzumab-rezetecan payload release-rate constant RAT; source column CT3. NOTE the INVERTED orientation relative to every other model in this list: here NSCLC is the model reference and ‘other tumor types’ carries its own retained effect, whereas elsewhere the ‘other’ pool is the residual bucket. RSE 78.0%, the least precisely estimated parameter in the paper, and its bootstrap 95% CI spans zero), Nguyen_2019_cabozantinib.R, Nguyen_2019_cabozantinib_liver_dysfunction.R (the updated refit of the Lacy 2018 integrated cabozantinib model; Nguyen 2019 Table 3 reports the effect as a direct MULTIPLICATIVE factor rather than the fractional change its predecessor used, so the ‘other malignancies’ pool carries 1.19^TUMTP_OTHER on CL/F and 0.762^TUMTP_OTHER on Vc/F in the final model, and 1.14 / 0.749 in the liver-dysfunction variant; pool composition unchanged from Lacy 2018, n = 40 subjects from the Study XL184-001 mixed-malignancy first-in-human cohort, now 2% of the enlarged 2023-subject cohort), Zhang_2019b_nivolumab.R (exponential effect on nivolumab baseline CL of exp(0.0718) = 1.074, against an NSCLC reference and alongside the named TUMTP_NPC indicator; not statistically significant, RSE 57.5% with a 95% CI of -0.0129 to 0.153 spanning zero; the pool is melanoma 116, renal cell carcinoma 35, colorectal cancer 33, prostate cancer 25 and hepatocellular carcinoma 2, i.e. 211 / 1200 = 17.6% of the cohort once the 959 NSCLC and 23 NPC subjects are removed), Cheng_2021_avadomide.R (‘other solid tumor’ in an avadomide cohort whose named strata are healthy (reference), DLBCL, PCNSL, NHL, HCC, GBM, MM and brain cancer; fractional effect CL/F x (1 - 0.364), the V2/F effect held at zero by the authors; n = 19 of 298), Nguyen_2022_cemiplimab.R (fractional change 1 + (-0.491) on T50 of the sigmoid time-dependent change in cemiplimab CL, against an NSCLC reference alongside TUMTP_CSCC and TUMTP_BCC; the pool is control-stream level TUMTYPN = 5 (‘OTHERS’), the miscellaneous advanced solid tumors of the first-in-human Study 1423, whose composition the paper does not enumerate), Gibiansky_2022_tisotumab.R (non-cervical tumors against a cervical-cancer reference, multiplicative 1.22^TUMTP_OTHER on the apparent MMAE clearance of a tisotumab vedotin ADC-MMAE model; the pool is every tissue-factor-expressing solid tumor other than cervical cancer in innovaTV 201, 202 and 207, n = 227 of 399).
• Notes: Scope: specific because the set of histologies collapsed into “Others” is defined by the analysis plan of the source paper; two papers’ TUMTP_OTHER columns are not interchangeable. Document the exact per-paper composition (e.g., “NSCLC + prostate + ovarian + other, n = 107 in Quartino 2019”; “small-cell + non-small-cell lung + CRC + esophageal + pancreatic ductal adenocarcinoma, n = 184 in Sathe 2024 and Sathe 2025 (same PAT2 = 4 pool carried across the two analyses)”; “advanced mixed malignancies enrolled in the FIH Study XL184-001, n = 40 in Lacy 2018”; “all non-{lung, breast, GBM} cancer types in the radiation-therapy cohort, n = 72 in Gastonguay 2005”) in covariateData[[TUMTP_OTHER]]$notes. A given subject can have at most one of the TUMTP_<GROUP> indicators (including TUMTP_OTHER) set to 1; all-zero means the reference group. One registered use has an inferred rather than a stated composition: Abegesah 2025 never labels its tumtyp = 1 stratum anywhere – Table 4 names the row only “Tumor type 1 on CL” and, unlike levels 2 and 3, it is absent from the Fig. 1 tornado plot – so its composition is inferred as the residual non-NSCLC / non-bladder / non-BTC pool (small-cell lung cancer plus the miscellaneous Study 1108 advanced solid tumours) and TUMTP_OTHER is used for that reason. That inference is safe to carry only because the coefficient there is negligible and not distinguishable from zero (RSE 185%, bootstrap 95% CI spanning zero); do NOT extend the same reasoning to an unlabelled stratum carrying a materially non-zero effect – refer it to the maintainers instead.

SPDL1 (canonical for soluble PD-L1 concentration)

• Description: Baseline (or time-varying) serum concentration of soluble programmed death-ligand 1 (sPD-L1). Serves as a circulating biomarker of target burden and immune activation for anti-PD-1/PD-L1 antibodies.
• Units: pg/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – used with power scaling (SPDL1 / ref)^exponent. Reference value observed: 173.8 pg/mL (study-population median in Ogasawara 2020).
• Source aliases: none; SPDL1 is the standard abbreviation used directly in source analyses.
• Example models: Ogasawara_2020_durvalumab.R (power effect on CL, exponent 0.0617, reference 173.8 pg/mL; time-varying; values below LOD imputed as LOD/2 = 33.55 pg/mL), Quartino_2019_trastuzumab.R (per-group typical-value switch on linear CL; NSCLC plus a small residual group of prostate, ovarian, and other histologies), Wang_2024_sugemalimab.R (heterogeneous solid-tumor residual group of n = 174; exponential coefficient log(0.885) on CL and log(0.926) on Vc; NSCLC is the reference group, not part of TUMTP_OTHER), deVries_2025_durvalumab.R (centred linear effect on the Michaelis-Menten Vmax rather than on linear CL: (1 + 0.00336 * (SPDL1 - 124.8)), reference 124.8 pg/mL; baseline, time-fixed; the only registered model that places sPD-L1 on the saturable elimination arm).
• Notes: Scope: specific because sPD-L1 is meaningful only for drugs targeting the PD-1/PD-L1 pathway. For other checkpoint biomarkers (e.g., soluble CTLA-4, soluble LAG-3) register new dedicated canonicals rather than reusing this one.
  • TTYPE3 (Wang 2024; level 3 of a five-level tumor-type factor labelled “Other” in the source) – decompose into TUMTP_OTHER = as.integer(TTYPE3 == 1).
• Example models: Quartino_2019_trastuzumab.R (per-group typical-value switch on linear CL; NSCLC plus a small residual group of prostate, ovarian, and other histologies), Wang_2024_sugemalimab.R (heterogeneous solid-tumor residual group of n = 174; exponential coefficient log(0.885) on CL and log(0.926) on Vc; NSCLC is the reference group, not part of TUMTP_OTHER), Ogasawara_2020_durvalumab.R (power effect on CL, exponent 0.0617, reference 173.8 pg/mL; time-varying; values below LOD imputed as LOD/2 = 33.55 pg/mL), deVries_2025_durvalumab.R (centred linear effect on the Michaelis-Menten Vmax rather than on linear CL: (1 + 0.00336 * (SPDL1 - 124.8)), reference 124.8 pg/mL; baseline, time-fixed; the only registered model that places sPD-L1 on the saturable elimination arm).
• Notes: Scope: specific because the set of histologies collapsed into “Others” is defined by the analysis plan of the source paper; two papers’ TUMTP_OTHER columns are not interchangeable. Document the exact per-paper composition (e.g., “NSCLC + prostate + ovarian + other, n = 107 in Quartino 2019”; “miscellaneous solid tumors excluding NSCLC, lymphoma, GCGEJ, and ESCC, n = 174 in Wang 2024”) in covariateData[[TUMTP_OTHER]]$notes. A given subject can have at most one of the TUMTP_<GROUP> indicators (including TUMTP_OTHER) set to 1; all-zero means the reference group.

MCPROT (canonical for serum monoclonal (M) protein concentration)

• Description: Serum monoclonal (M) protein concentration. Multiple-myeloma plasma-cell-burden marker secreted by the tumor clone; elevated MCPROT reflects higher tumor burden and (for IgG-secreting MM) competes with therapeutic IgG mAbs for FcRn-mediated salvage and target-mediated elimination. Typically time-varying – measured at multiple visits over the treatment course and supplied at every PK observation time via linear interpolation between measurements.
• Units: g/dL (US-convention; equivalent to 10 g/L SI). Document the unit used in each model via covariateData[[MCPROT]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – used as a continuous, log-linear effect on the Vmax of target-mediated elimination via exp(theta * MCPROT) (i.e., MCPROT enters un-log-transformed). Reference values observed: 0 g/dL (Ide 2020 Vmax,REF reference) and 2.0 g/dL (Ide 2020 figure-1 reference patient).
• Source aliases:
  • TMCPROT (time-varying serum M-protein concentration) – used in Ide_2020_elotuzumab.R. NONMEM column with imputation sentinel -99 for missing observations, replaced by population median 2.1 g/dL via IF(TMCPROT.EQ.-99) TMCPROT = 2.1.
• Example models: Ide_2022_elotuzumab.R (g/dL, time-varying, linearly interpolated between measurements; exp(0.27 * MCPROT) on Vmax, source column TMCPROT_Z with missing imputed to 2.05 g/dL), Ide_2020_elotuzumab.R (g/dL, time-varying; entered un-log-transformed as exp(0.277 * MCPROT) on Vmax of the Michaelis-Menten target-mediated elimination from the central compartment), Collins_2023_belantamab_mprotein.R (g/L, time-fixed baseline; seeds the initial condition of the modelled M-protein tumor state and is binarized as (MCPROT < 20) to carry a multiplicative factor 1.41 on the kill rate constant KD), Srimani_2022_ixazomib_mprotein.R (g/L, time-fixed observed baseline Y_BL; scales the relative M-protein states to g/L, sets the dropout-hazard M-protein threshold and enters the PFS hazard as a power covariate centred on 23 g/L).
• Notes: Specific scope because the column is mechanistically meaningful only for plasma-cell-targeting therapies in multiple myeloma (e.g., elotuzumab anti-SLAMF7, daratumumab anti-CD38, isatuximab anti-CD38, belantamab anti-BCMA, and BCMA-bispecifics / CAR-T). MCPROT decreases with treatment response; the time-varying form is the only correct way to capture the diminishing target-mediated-elimination component as the tumor regresses. In NONMEM datasets MCPROT is supplied at each event-row time, with linear interpolation between observations and last-observation-carried-forward beyond the last sample (Ide 2020 Methods). Distinct from MM_NIGG (which is the immunoglobulin subtype, an MM-disease stratifier that is time-fixed), SBCMA (soluble BCMA, a different MM tumor-burden biomarker for BCMA-targeting drugs), and B2M (beta-2-microglobulin, a renal-function-and-MM-disease-burden marker). The 1 g/dL = 10 g/L conversion lets future SI-convention papers register the same canonical with their own unit string.

T_MPROTEIN_NADIR (canonical for per-subject time of the serum M-protein nadir)

• Description: Per-subject time after the first dose at which the serum monoclonal (M) protein reached its nadir, supplied as data. In Srimani 2022 it was derived before the PK/PD fit by fitting combinations of exponential functions to each patient’s observed M-protein profile; “the minimum value of the model predicted concentration plus 0.5 g/L was defined as the nadir and the corresponding time was denoted t_nadir” (Methods). About 4.71% of patients had no reliably estimable nadir time; the resistant (regrowing) M-protein population grows from this time.
• Units: h (model time units; document the source scale in covariateData[[T_MPROTEIN_NADIR]]$notes).
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as a switch time, (t - T_MPROTEIN_NADIR) * (t > T_MPROTEIN_NADIR). A patient whose M-protein never regrows takes a value beyond the end of follow-up.
• Source aliases:
  • t_nadir – Srimani 2022 Methods and Equation 3.
• Example models: Srimani_2022_ixazomib_mprotein.R (resistant M-protein population exp(kL * (t - T_MPROTEIN_NADIR)) - 1 after the nadir; for simulation the vignette draws it from the model’s own relapse hazard).
• Notes: Member of the T_<event> family, but unlike most members (time-varying “time since X”) it is a STATIC per-subject event time, constant over the record. The analyte is named so a nadir time of another biomarker (e.g. neutrophil count) cannot collide. It is a data-derived quantity, not a model parameter, so it belongs here rather than in ini().

PDL1_TUM (canonical for tumor PD-L1 expression (tumor proportion score))

• Description: Baseline tumor PD-L1 expression measured by immunohistochemistry, reported as Tumor Proportion Score (TPS) – the percent of viable tumor cells with complete or partial membrane staining at any intensity. Used as a prognostic and predictive covariate for anti-PD-1 / anti-PD-L1 therapies. Distinct from SPDL1 (soluble serum PD-L1 in pg/mL); PDL1_TUM is a tissue biomarker, SPDL1 is a circulating biomarker.
• Units: percent (0-100)
• Type: continuous
• Scope: general (oncology)
• Reference category: n/a – enters typically as exp(theta * PDL1_TUM) (uncentred multiplicative) or 1 + theta * (PDL1_TUM - ref) (centred linear deviation). Reference values observed: 15 (Struemper 2025 population median across PD-1 inhibitor arms); 70 (Struemper 2025 INTR@PID imputed value when the 22C3 assay was unavailable).
• Source aliases: PD-L1, PDL1, TPS, PDL1_TPS – paper notation varies; map to canonical PDL1_TUM with no value transformation.
• Example models: Struemper_2025_tumorsize_OS_nsclc.R (exponential effect on the tumor shrinkage rate ks of a Stein bi-exponential TGI model; coefficient 0.00902 per percent; applied only to PD-1 inhibitor-containing treatment arms via a derived has_pd1 indicator that gates the effect when TRT is one of the eight pembrolizumab- or dostarlimab-containing categories).
• Notes: The PD-L1 IHC antibody clone (22C3, 73-10, SP142, SP263, etc.) determines cross-study comparability; document the assay per-model in covariateData[[PDL1_TUM]]$notes. Tumor Proportion Score (TPS) is the percent of TUMOR cells stained; Combined Positive Score (CPS) and Immune Cell (IC) score are distinct percent metrics that should be registered separately if a paper relies on them. PDL1_TUM is a tissue-level biomarker that may coexist in a model with the serum-level SPDL1. Scope: general because PDL1 IHC is a widely reused oncology covariate. Founding example: Struemper 2025 (exp(PDL1_TUM * 0.00902) on ks, paper Table 3 footnote b).

PDL1_TUM_MISSING (canonical for tumor-PD-L1-result-missing indicator)

• Description: Binary indicator for a subject whose baseline tumor PD-L1 tumor proportion score could not be assigned. 1 = no PD-L1 TPS available (no evaluable contemporaneous biopsy, or an assay result that could not be scored); 0 = a TPS value is available in PDL1_TUM. Companion to the continuous PDL1_TUM canonical, for source papers that retain “unknown PD-L1” as a distinct modelled category with its own estimated coefficient rather than imputing those subjects into a measured expression stratum.
• Units: (binary)
• Type: binary
• Scope: general (oncology)
• Reference category: 0 (PD-L1 TPS available). When PDL1_TUM_MISSING == 1, any indicator derived from PDL1_TUM (e.g. a TPS >= 50% or TPS < 1% stratum flag) must be gated off inside model so the unknown subject receives only its own coefficient; the value carried in PDL1_TUM for such a subject is then a placeholder with no effect.
• Source aliases:
  • PD-L1 "Unknown" – Chatterjee 2016 supplementary Table S3 category label and the PD-L1_3 coefficient row of supplementary Table S6.
• Example models: Chatterjee_2016_pembrolizumab.R (fractional-deviation effect on the tumor kill rate: kdeath *= (1 + 0.268 * PDL1_TUM_MISSING) relative to the TPS 1%-49% reference category; 60 of 505 subjects, supplementary Table S3).
• Notes: Follows the missing-data-indicator pattern established by ADA_MISSING (Suri 2018 brentuximab), CYP2C9_MISSING (Lane 2011 warfarin) and HEPIMP_MOD_OR_MISSING. Use only when the source paper explicitly fits a separate coefficient for the PD-L1-unknown subgroup; a paper that imputes unknown PD-L1 to a measured level (as Struemper 2025 does for the INTR@PID cohort, imputing TPS = 70) should set PDL1_TUM to the imputed value and NOT use this column. The estimated coefficient for a missing-PD-L1 stratum is generally poorly identified and biologically uninterpretable, because the stratum mixes PD-L1-positive and PD-L1-negative tumors – Chatterjee 2016’s supplementary text says exactly this and reports an RSE of 145% for it. Do not extrapolate a missingness coefficient to a measured subgroup. Scope: general (oncology) so that any anti-PD-1 / anti-PD-L1 exposure-response paper carrying an unknown-PD-L1 stratum can reuse the column.

PDL1_TUM_POS (canonical for tumor PD-L1 positivity indicator)

• Description: Binary indicator of baseline tumor PD-L1 positivity as called by the source study’s own immunohistochemistry assay and threshold. 1 = PD-L1-positive tumor, 0 = PD-L1-negative tumor. Companion to the continuous PDL1_TUM (tumor proportion score, percent) for the many exposure-response papers that report only a dichotomised positive / negative call, and to PDL1_TUM_MISSING for papers that additionally fit a distinct coefficient for the unassignable stratum.
• Units: (binary)
• Type: binary
• Scope: general (oncology)
• Reference category: study-specific – do NOT assume the negative level is the reference. The source’s “most frequent category” rule can make PD-L1-POSITIVE the reference, as it does in Chatterjee 2017 (739 of 1366 positive, 54.1%), which flips the sign convention of every PD-L1 coefficient in that model. Record the reference level in covariateData[[PDL1_TUM_POS]]$reference_category.
• Source aliases:
  • PDL1 coded 1 = positive / 0 = negative / 6 = unknown – Chatterjee 2017 NONMEM input column; the indicator is PDL1_TUM_POS = as.integer(PDL1 == 1) with PDL1_TUM_MISSING = as.integer(PDL1 == 6).
  • “PD-L1 status” – Chatterjee 2017 supplementary Table 5A row label (Negative / Positive / Unknown).
• Example models: Chatterjee_2017_pembrolizumab_consolidated.R (fractional-deviation effect on the tumor kill rate of a bi-exponential tumor-size model: kdeath * (1 - 0.614) for PD-L1-negative relative to the PD-L1-positive reference; the largest single dOFV in that covariate search, 37, P = 8.66e-09).
• Notes: Use this column, not a threshold derived from PDL1_TUM, whenever the source paper dichotomises PD-L1 without printing the percent cut-point or the assay clone – deriving a binary from the continuous column would require inventing a threshold the source never states. Chatterjee 2017 is exactly that case: it says only “PD-L1 expression level grouped into two categories (positive and negative)”. Conversely, when the source reports a TPS percentage, carry PDL1_TUM and derive the strata inside model() so the cut-point is auditable. The two columns may coexist in a dataset assembled for several models. Assay comparability is the usual confounder across studies (22C3, 73-10, SP142, SP263 clones, and melanoma-era prototype assays scoring tumor nests and stromal bands rather than a pure tumor proportion score), so record the assay per model. Gate this indicator off (set it to 0) whenever PDL1_TUM_MISSING == 1, so an unknown-status subject receives only its own coefficient.

LMET (canonical for baseline presence of liver metastases)

• Description: Binary indicator of radiologically documented liver metastases at baseline, 1 = liver metastases present, 0 = no liver metastases.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no liver metastases at baseline).
• Source aliases: none; LMET is the common NONMEM / clinical-PK abbreviation used directly by the source papers.
• Example models: Quartino_2019_trastuzumab.R (exponential effect on linear CL; +16.4% typical CL when LMET = 1, per Quartino 2019 Table 1 theta12 = 0.152).
• Notes: Liver metastases are associated with hepatic protein-synthesis impairment and altered IgG catabolism, making LMET a commonly tested covariate in oncology mAb population PK analyses. Scope: general so future oncology papers can reuse the canonical column. Time-fixed baseline indicator; if a source paper treats it as time-varying (progression during treatment), document in covariateData[[LMET]]$notes.

MET_GE4 (canonical for baseline number of metastatic sites >= 4 indicator)

• Description: Binary indicator dichotomising the count of baseline metastatic sites at 4, 1 = patient has four or more documented metastatic sites at baseline, 0 = patient has zero to three metastatic sites at baseline. Time-fixed per subject. Treated as a surrogate for tumor burden in oncology mAb popPK analyses.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (fewer than four metastatic sites at baseline).
• Source aliases:
  • MET – used in Bruno_2005_trastuzumab.R (Bruno 2005 Methods: “MET=1 if number of metastatic sites=4 or greater; otherwise MET=0”). When a source paper supplies the raw integer count column (NMET, N_METS, etc.) rather than the pre-binarised indicator, derive MET_GE4 = as.integer(NMET >= 4).
• Example models: Bruno_2005_trastuzumab.R (multiplicative effect on linear CL: typical CL multiplied by (1 + 0.221 * MET_GE4), i.e. +22.1% CL in MET_GE4 = 1 patients; reference MET_GE4 = 0 per Bruno 2005 Table 3).
• Notes: The >= 4 split is the dichotomisation used by Bruno 2005 to capture “patients with four or more metastatic sites” as a high-tumor-burden subgroup. Scope: general so future oncology popPK papers using the same dichotomisation can reuse this canonical column. If a future paper uses a different split (e.g. >= 2 or >= 5), register a separate MET_GEN canonical rather than overloading this entry. A MET_GE4 = 1 patient may also have LMET = 1; the two columns are not mutually exclusive (liver-only metastases would have LMET = 1, MET_GE4 = 0).

MET_GE3 (canonical for baseline number of metastatic sites >= 3 indicator)

• Description: Binary indicator dichotomising the count of baseline metastatic sites at 3, 1 = patient has three or more documented metastatic sites at baseline, 0 = patient has zero to two metastatic sites at baseline. Time-fixed per subject. Treated as a surrogate for tumor burden in oncology popPK / TGD analyses that centre the source-paper continuous integer count at 3.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (fewer than three metastatic sites at baseline).
• Source aliases:
  • NumMet (raw integer count 0-12; used in Terranova_2022_TGD_OS_gastric.R). When a source paper supplies the raw integer count column, derive MET_GE3 = as.integer(NumMet >= 3).
• Example models: Terranova_2022_TGD_OS_gastric.R (multiplicative-additive effect on Gompertzian baseline tumor size: BASE = exp(ltvBase + etalBase) * exp(0.143 * MET_GE3) – single-step binarisation of the paper’s continuous per-metastasis coefficient exp(0.143 * (NumMet - 3)); deviation from paper’s continuous form documented in vignette Errata per the count-covariate policy).
• Notes: Sibling of MET_GE4 (Bruno 2005) and NTARGET_GE3 (Struemper 2025). The >= 3 split matches Terranova 2022’s centering value of 3 (avelumab-arm median NumMet). Auto-approved under the count-covariate-decomposed-to-binary policy on 2026-07-24. If a future paper uses a different split (e.g. >= 2 or >= 5), register a separate sibling (MET_GE2, MET_GE5) rather than overloading this entry. A MET_GE3 = 1 patient may also have LMET = 1 (liver metastasis) or PERIT_CARC = 1 (peritoneal carcinomatosis); the three columns are not mutually exclusive.

PERIT_CARC (canonical for peritoneal carcinomatosis indicator)

• Description: Binary indicator of radiologically or surgically documented peritoneal carcinomatosis (diffuse peritoneal-surface metastatic spread) at baseline / re-baseline in a solid-tumor oncology cohort, 1 = peritoneal carcinomatosis present, 0 = absent. Time-fixed per subject at the assessed baseline. Distinct from PERIT_DIAL (peritoneal-dialysis treatment-status indicator; entirely different clinical concept) and from DIS_PERIT (peritonitis, an infectious inflammation of the peritoneum).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no peritoneal carcinomatosis at baseline).
• Source aliases:
  • Peritoneal carcinomatosis – used in Terranova_2022_TGD_OS_gastric.R (Terranova 2022 Table S2 categorical covariate on log-median OS, coded 1/0).
• Example models: Terranova_2022_TGD_OS_gastric.R (linear-additive effect on log-median OS: log_median_OS += -0.181 * PERIT_CARC; peritoneal carcinomatosis shortens median OS by ~16.6% relative to patients without it, one of the covariates flagged as meaningful in the paper’s +/-15% posterior-median threshold).
• Notes: Specific scope because peritoneal carcinomatosis is a gastrointestinal-cancer-associated metastasis pattern (colon, gastric, ovarian, appendiceal); future GI-oncology or ovarian-cancer popPK models retaining this covariate should extend the Example-models list rather than register a sibling. Sibling to LMET (liver metastasis) and MET_GE3 / MET_GE4 (aggregate metastatic-site count binarised at different thresholds); all four indicators can coexist for a patient with widespread advanced disease.

RESP_SD (canonical for RECIST re-baseline response = stable disease indicator)

• Description: Binary indicator, 1 = the patient’s RECIST 1.1 best response at re-baseline (typically the end of induction chemotherapy or the pre-maintenance assessment) is stable disease (SD); 0 = any other response category (CR, PR, PD, NED, NE, or non-CR/non-PD). Time-fixed at the re-baseline assessment. Used as one leaf of a decomposed RECIST-response categorical covariate under the count-covariate-decomposed-to-binary policy.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (RECIST response is anything other than SD – typically responder [CR/PR], non-CR/non-PD, PD, or NE).
• Source aliases:
  • Re-baseline stable disease vs other – used in Terranova_2022_TGD_OS_gastric.R (Terranova 2022 Table S2 categorical covariate; also enters TGD BASE equation).
  • RESSD – source NONMEM column name derived from the RECIST assessment.
• Example models: Terranova_2022_TGD_OS_gastric.R (multiplicative effect on Gompertzian baseline tumor size BASE = ... * (1 + 0.644 * RESP_SD) – SD patients have +64% larger baseline tumor size than the responder+PD+NE reference at re-baseline; and additive linear effect on log-median OS: log_median_OS += -0.0919 * RESP_SD).
• Notes: One of three parallel RECIST-response indicators (RESP_SD, RESP_NONPDCR, RESP_RESPONDER) that decompose the RECIST 1.1 response categorical into binary flags per the count-covariate-decomposed-to-binary policy. The three flags share the same source RECIST-response column and different subsets are used by different sub-models within the same paper (Terranova 2022 TGD uses RESP_SD + RESP_NONPDCR with implicit reference = responder/PD/NE; Terranova 2022 OS TTE uses RESP_SD + RESP_RESPONDER with implicit reference = non-responder-non-SD).

RESP_NONPDCR (canonical for RECIST re-baseline response = neither CR nor PR indicator)

• Description: Binary indicator, 1 = the patient’s RECIST 1.1 best response at re-baseline is NOT complete response (CR) and NOT partial response (PR) – i.e., the patient is a non-responder by RECIST criteria (any of stable disease [SD], non-CR/non-PD, no evidence of disease [NED], progressive disease [PD], or not evaluable [NE]); 0 = responder (CR or PR). Time-fixed at the re-baseline assessment. Complement of RESP_RESPONDER from a different reference-category framing; both are needed because different sub-models within the same analysis use different reference categories.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (responder: CR or PR at re-baseline).
• Source aliases:
  • RES_nonPR/nonCR – Terranova 2022 Supplementary Methods notation; equivalent to “response at re-baseline was not complete or partial response” per the paper’s definition. Coded 1/0.
• Example models: Terranova_2022_TGD_OS_gastric.R (multiplicative effect on Gompertzian baseline tumor size: BASE = ... * (1 - 0.0769 * RESP_NONPDCR) – non-responders have -7.69% smaller baseline tumor size than responders at re-baseline; used in the TGD sub-model only).
• Notes: Sibling of RESP_SD and RESP_RESPONDER. The name is deliberately explicit about “not CR / not PR” rather than the shorter RESP_NONRESP because the source paper’s coding is defined on the exact CR/PR complement rather than an arbitrary “responder / non-responder” binarisation of a broader ordinal.

RESP_RESPONDER (canonical for RECIST re-baseline response = responder (CR or PR) indicator)

• Description: Binary indicator, 1 = the patient’s RECIST 1.1 best response at re-baseline is either complete response (CR) or partial response (PR) – i.e., the patient is a responder by RECIST criteria; 0 = non-responder (any of SD, non-CR/non-PD, NED, PD, NE). Time-fixed at the re-baseline assessment. Complement of RESP_NONPDCR from a different reference-category framing; both are needed because different sub-models within the same analysis use different reference categories.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (non-responder at re-baseline; typically the pooled reference alongside a separate RESP_SD flag).
• Source aliases:
  • Re-baseline responder vs other – Terranova 2022 Table S2 OS TTE covariate label; coded 1/0.
• Example models: Terranova_2022_TGD_OS_gastric.R (additive linear effect on log-median OS: log_median_OS += 0.146 * RESP_RESPONDER; responder patients have ~15.7% longer median OS than non-responder-non-SD reference).
• Notes: Sibling of RESP_SD and RESP_NONPDCR. Explicitly RESPONDER rather than CRPR because responder-vs-other is the clinically meaningful contrast used by the source paper’s OS sub-model.

RESP_DCR (canonical for RECIST disease-control (CR, PR or SD vs PD) indicator)

• Description: Binary indicator of disease control by RECIST criteria at a protocol-defined early on-treatment assessment, 1 = the patient’s best response is complete response (CR), partial response (PR) or stable disease (SD); 0 = progressive disease (PD). This is the standard “disease control rate” (DCR) dichotomisation of the RECIST response categorical – the CR/PR/SD-versus-PD contrast – as distinct from the responder contrast (RESP_RESPONDER, CR/PR versus everything else) and the stable-disease flag (RESP_SD). Time-fixed at the assessment used by the source model; record the assessment week in covariateData[[RESP_DCR]]$notes because the DCR landmark differs between papers (week 8 in Claret 2014; end-of-induction re-baseline elsewhere).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (progressive disease, PD).
• Source aliases:
  • DCR – Claret 2014 (J Clin Pharmacol 54(3):253-257) Table 2 and Table 3 covariate row label; the paper defines it in Methods as “CR + PR + SD versus PD (dichotomic, disease control rate, DCR)”. Coded 1/0.
  • DC / DISCTRL – common NONMEM column-name variants for a disease-control flag.
• Example models: Claret_2014_DCR_OS_nsclc.R (additive effect on the log-normal AFT location parameter of overall survival: mu_os += e_dcr_mu_os * RESP_DCR with e_dcr_mu_os = 0.939, i.e. median OS is exp(0.939) = 2.56 times longer in patients with week-8 disease control), Claret_2014_DCR_PFS_nsclc.R (additive effect on the log-normal AFT location parameter of progression-free survival: mu_pfs += e_dcr_mu_pfs * RESP_DCR with e_dcr_mu_pfs = 1.378, i.e. median PFS is exp(1.378) = 3.97 times longer).
• Notes: Fourth member of the RESP_<contrast> family (RESP_SD, RESP_NONPDCR, RESP_RESPONDER, RESP_DCR) that decomposes the RECIST response categorical into binary flags under the count-covariate-decomposed-to-binary policy. All four share the same underlying source RECIST-response column and differ only in where the categorical is cut and which category is the reference; a paper’s published coefficient is defined on exactly one cut, so the flags are not interchangeable and cannot be derived from one another. RESP_DCR is the logical complement of a progressive-disease flag (RESP_DCR = 1 - as.integer(response == "PD")) and is NOT the complement of RESP_RESPONDER (patients with SD have RESP_DCR = 1 and RESP_RESPONDER = 0). Derive from a RECIST response column with RESP_DCR = as.integer(response %in% c("CR", "PR", "SD")).

NTARGET (canonical for baseline number of target lesions)

• Description: Count of target lesions identified at baseline per RECIST 1.1 (RECIST 1.1 permits at most five target lesions, at most two per organ). Time-fixed per subject. Captures lesion multiplicity / disease spread, which is distinct from the tumour-burden magnitude carried by TUM_SLD (the sum of those lesions’ longest diameters, in mm). Carry the raw count here; apply any dichotomisation the source model uses inside model(), or use the pre-dichotomised NTARGET_GE3 when the source itself only reports a binary split.
• Units: (count)
• Type: continuous
• Scope: general (oncology)
• Reference category: n/a – enters as a normalised power term (NTARGET / ref)^exponent, or linearly. Reference value observed: 3 (Chatterjee 2017, the denominator in the source control stream; 3 is also the reference centring in Struemper 2025).
• Source aliases:
  • NTARGET – NONMEM input column name in both Chatterjee 2017 and Struemper 2025.
  • “number of target lesions” – Chatterjee 2017 main-article Table 2 and supplementary Table 3 row labels.
• Example models: Chatterjee_2017_pembrolizumab_mixture.R (power effect on the estimated baseline size of the growing/shrinking tumour portion: BASEL1 * (NTARGET / 3)^0.654; the single strongest covariate relationship in that analysis, dOFV -102, P < 0.0001, supplementary Table 3).
• Notes: Register entry added because a power (or any other continuous) functional form on the lesion count cannot be recovered from the binary NTARGET_GE3; the two columns are complementary, not competing, and NTARGET_GE3 remains the right column for a source whose own model dichotomises. A source that maps a missing count to a covariate multiplier of 1 – as Chatterjee 2017 does via IF(NTARGET.EQ.-99) BAS1NTARGET = 1 – is reproduced without any missingness column by carrying the reference count itself (NTARGET = 3), which makes the normalised power term exactly 1; document that substitution in covariateData[[NTARGET]]$notes. Never carry the source sentinel (-99) into the column: a power term would return NaN or a wildly wrong multiplier and the failure is silent.

NNODAL (canonical for baseline number of affected lymph-node lesions)

• Description: Count of involved (affected) lymph-node lesions at baseline. Time-fixed per subject. Nodal involvement is a marker of regional and distant spread and is scored separately from the target-lesion count in RECIST-based datasets, because RECIST 1.1 measures pathological nodes by short axis and applies a different measurability threshold to them. Distinct from NTARGET (count of RECIST target lesions of any site) and from LYMPH_ABS (absolute circulating lymphocyte count, a haematology laboratory value).
• Units: (count)
• Type: continuous
• Scope: general (oncology)
• Reference category: n/a – enters linearly and uncentred in the founding model.
• Source aliases:
  • NNODAL – Chatterjee 2017 NONMEM input column name.
  • “number of lymph nodes affected” – Chatterjee 2017 main-article Methods and supplementary Table 3 row label.
• Example models: Chatterjee_2017_pembrolizumab_mixture.R (uncentred linear term on each non-reference class logit of the second-stage multinomial regression, coefficient 0.112 per node: more involved nodes raise the odds of every responder class relative to the progressor “escape” class; dOFV -15.3, P < 0.0001, supplementary Table 3).
• Notes: The founding model centres this covariate at 0 (THETA(5)*(NNODAL - 0.00) in the source control stream), so the coefficient is a per-node log-odds and the reference patient has no nodal involvement. A count is carried as a count here rather than decomposed into binary indicators because the source model’s functional form is linear in the count; decomposing would falsify the published equation. Where a source instead dichotomises nodal involvement, register a separate indicator (e.g. NNODAL_GE1) rather than overloading this column, following the NTARGET / NTARGET_GE3 precedent.

NTARGET_GE3 (canonical for baseline number of target lesions >= 3 indicator)

• Description: Binary indicator dichotomising the count of target lesions at baseline at 3, 1 = three or more target lesions at baseline (per RECIST 1.1), 0 = one or two target lesions at baseline. Time-fixed per subject. Distinct from TUM_SLD (continuous sum of longest diameters, mm), which captures tumor burden magnitude; NTARGET_GE3 captures lesion multiplicity / spread.
• Units: (binary)
• Type: binary
• Scope: general (oncology)
• Reference category: 0 (one or two target lesions at baseline).
• Source aliases:
  • NTARGET – raw integer count of target lesions per RECIST 1.1 (paper notation in Struemper 2025). When a source paper supplies the integer count column, derive NTARGET_GE3 = as.integer(NTARGET >= 3).
• Example models: Struemper_2025_tumorsize_OS_nsclc.R (multiplicative effect on the typical-value baseline tumor size TVTSb: TVTSb * (1 + 0.288 * NTARGET_GE3); +28.8% TVTSb for patients with three or more target lesions vs one or two).
• Notes: The >= 3 split aligns with the Struemper 2025 NSCLC analysis equation centring (NTARGET = 3 reference) even though the population median is 2 lesions (paper Table S3); the binarisation preserves the qualitative effect direction (more lesions => larger baseline tumor) while avoiding the linear extrapolation past the data range (paper data: NTARGET 1-8) that the source paper’s linear form implies. If a future oncology paper uses a different split (e.g. >= 2 or >= 5), register a separate NTARGET_GEN canonical rather than overloading this entry. Scope: general so future oncology popPK / popPD papers using the same dichotomisation can reuse this column. Founding example: Struemper 2025.

JOINT_SMALL (canonical for small-joint tenosynovial giant cell tumor (TGCT) involvement indicator)

• Description: 1 = the primary TGCT tumor is located in a small joint (e.g., hand / wrist / foot / ankle), 0 = the tumor is located in a large joint (e.g., knee / hip / shoulder / elbow; reference category). Time-fixed per subject; the anatomic joint-size classification is investigator-assigned per patient in TGCT trials. Small-joint TGCT is associated with lower baseline tumor size and greater response to CSF1R inhibition.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (large joint; the reference joint-size category in Yin 2021 Figure 2 caption reference patient).
• Source aliases:
  • JOINT_SIZE (with values small / large) – decompose into JOINT_SMALL = as.integer(JOINT_SIZE == "small") (Yin 2021 codes joint size as a binary categorical stratifier without publishing the raw column name).
• Example models: Yin_2021_pexidartinib.R (multiplicative effects exp(theta4)^JOINT_SMALL on baseline Y0 (0.581), exp(theta10)^JOINT_SMALL on kdrug (2.14), and exp(theta14)^JOINT_SMALL on konset (1.82) in the longitudinal RECIST tumor-size model; small-joint TGCT reduces baseline tumor size and increases drug-effect rate constant kdrug relative to large-joint reference).
• Notes: Specific scope because the small-vs-large joint-size stratifier is tied to musculoskeletal-oncology (TGCT) analyses; future TGCT popPK / popPD papers using the same dichotomisation should extend this entry’s example list rather than register a new canonical. The paper does not enumerate which anatomic joints are pooled as small vs large; this is an investigator-assigned per-patient classification. Distinct from SWOL_28JOINT (rheumatoid-arthritis 28-joint swollen-joint count from the DAS28 composite; integer 0-28) and from NTARGET_GE3 (RECIST target-lesion count binarised at 3).

TUMEXT_UPPER (canonical for upper-extremity tenosynovial giant cell tumor (TGCT) location indicator)

• Description: 1 = the primary TGCT tumor is located in an upper extremity (e.g., shoulder / elbow / wrist / hand), 0 = the tumor is located in a lower extremity (e.g., hip / knee / ankle / foot; reference category). Time-fixed per subject; the anatomic extremity classification is investigator-assigned per patient in TGCT trials. TGCT is overwhelmingly lower-extremity (~90% in the Yin 2021 cohort).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (lower extremity; the reference extremity category in Yin 2021 Figure 2 caption reference patient).
• Source aliases:
  • EXTREMITY (with values upper / lower) – decompose into TUMEXT_UPPER = as.integer(EXTREMITY == "upper") (Yin 2021 codes tumor extremity as a binary categorical stratifier without publishing the raw column name).
• Example models: Yin_2021_pexidartinib.R (multiplicative effects exp(theta3)^TUMEXT_UPPER on baseline Y0 (0.692), exp(theta9)^TUMEXT_UPPER on kdrug (0.527), and exp(theta13)^TUMEXT_UPPER on konset (7.82) in the longitudinal RECIST tumor-size model; upper-extremity TGCT has wide-CI non-significant effects on kdrug and konset in this cohort due to small upper-extremity cell counts, but the covariate is retained in the paper’s final model).
• Notes: Specific scope because the upper-vs-lower extremity stratifier is tied to musculoskeletal-oncology (TGCT) analyses; future TGCT popPK / popPD papers using the same dichotomisation should extend this entry’s example list rather than register a new canonical. Distinct from a hypothetical TUMEXT_LOWER canonical (would be the polar complement); the register uses the paper-published direction (upper as the non-reference indicator) to match Yin 2021 Table S2 sign convention. Complementary to JOINT_SMALL (small-vs-large joint-size stratifier); the two indicators describe orthogonal anatomic-location features of TGCT and may coexist in the same model.

ECOG_GE1 (canonical for Eastern Cooperative Oncology Group performance-status indicator, >= 1)

• Description: 1 if baseline Eastern Cooperative Oncology Group (ECOG) performance status is greater than or equal to 1, 0 if ECOG = 0. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (ECOG performance status = 0, i.e., fully active / asymptomatic).
• Source aliases:
  • PS / BPS – used in Bajaj_2017_nivolumab.R (BPS = “baseline performance status”; the one study using Karnofsky Performance Status was mapped to ECOG via Oken 1982 before binarization) and Zhang_2019_nivolumab.R (paper’s binary collapse PS=0 vs. PS>0).
  • ECOG_1 – alternative explicit form; equivalent to ECOG_GE1 when ECOG only takes values 0, 1, 2 in the analysis dataset (the typical oncology case).
  • ECOG_PS_GT0 – retired name used in earlier register drafts; semantically identical (>= 1 equals > 0 for integer ECOG scores).
  • ECOG101 (categorical 0/1/2 score with thresholding IF(ECOG101.GT.0.5)) – used in Ide_2020_elotuzumab.R. Decompose: ECOG_GE1 = as.integer(ECOG101 >= 1).
• Example models: Bajaj_2017_nivolumab.R (exponential effect on CL with coefficient 0.172), Zhang_2019_nivolumab.R (exponential effect exp(0.181) on baseline CL; additive effect -0.138 on the time-varying-CL Emax parameter), Ide_2020_elotuzumab.R (multiplicative effect on CL = 1.03; paired with ECOG_GE2 for separate ECOG=1 vs ECOG>=2 effects), Netterberg_2017_docetaxel.R (multiplicative effect on baseline ANC of the Friberg myelosuppression chain: BACOV *= (1 + theta * ECOG_GE1) with theta = 0.130; source column PERF with ordinal ECOG 0/1/2 values, binarized via ECOG_GE1 = as.integer(PERF >= 1) per Kloft 2006).
• Notes: Oncology papers conventionally report ECOG as an integer (0-5) but binarize at >= 1 because ECOG >= 2 is rare in trial cohorts. When a source paper provides the ordinal ECOG score separately, derive ECOG_GE1 = as.integer(ECOG >= 1). Zhang 2019 uses ECOG_GE1 on both baseline CL and the time-varying Emax parameter (unlike Bajaj 2017, which uses it on CL only); document the structural role in each model’s covariateData[[ECOG_GE1]]$notes. When a paper retains separate effects for ECOG = 1 vs ECOG >= 2 (Ide 2020), pair this column with ECOG_GE2 and supply both indicators in the event dataset.

ECOG_GE2 (canonical for Eastern Cooperative Oncology Group performance-status indicator, >= 2)

• Description: 1 if baseline Eastern Cooperative Oncology Group (ECOG) performance status is greater than or equal to 2, 0 if ECOG <= 1. Time-fixed per subject. Used in models that retain separate effects for ECOG = 1 vs ECOG >= 2 by pairing this column with ECOG_GE1.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (ECOG performance status <= 1; in models that pair ECOG_GE1 and ECOG_GE2, both indicators = 0 corresponds to ECOG = 0 and (ECOG_GE1 = 1, ECOG_GE2 = 0) corresponds to ECOG = 1).
• Source aliases:
  • ECOG101 (categorical 0/1/2 score with thresholding IF(ECOG101.GT.1.5)) – used in Ide_2020_elotuzumab.R. Decompose: ECOG_GE2 = as.integer(ECOG101 >= 2).
• Example models: Ide_2020_elotuzumab.R (multiplicative effect on CL = 1.15; paired with ECOG_GE1 to retain separate ECOG = 1 vs ECOG >= 2 effects).
• Notes: Parallels ECOG_GE1. Use only when the source paper reports a separate effect for ECOG >= 2 in addition to ECOG_GE1; otherwise ECOG_GE1 alone is sufficient. The paired (ECOG_GE1, ECOG_GE2) decomposition reproduces a three-level (ECOG = 0, ECOG = 1, ECOG >= 2) ordinal effect with two binaries.

TUMTP_BLADDER (canonical for bladder (urothelial) carcinoma tumor-type indicator)

• Description: 1 = bladder cancer (urothelial carcinoma of the bladder), 0 = other tumor types.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 = all other tumor types (in Abegesah 2025 the pooled reference stratum is non-small-cell lung cancer).
• Source aliases:
  • tumtyp (integer-coded tumor-type column; level 2 in Abegesah 2025) – decompose into TUMTP_BLADDER = as.integer(tumtyp == 2).
• Example models: Abegesah_2025_durvalumab.R (multiplicative effect on durvalumab CL of (1 + 0.0698), i.e. +7.0% relative to the pooled reference tumor type).
• Notes: Follows the TUMTP_NSCLC / TUMTP_SCLC / TUMTP_GASTRIC decomposition pattern. General scope because urothelial carcinoma is a high-frequency contrast in checkpoint-inhibitor popPK analyses – it was the first approved indication for both durvalumab and atezolizumab and recurs in pooled anti-PD-(L)1 cohorts (the sibling Hwang_2022_tremelimumab.R cohort is 17.1% urothelial bladder cancer). In Abegesah 2025 the tumor-type identity behind the Table 4 row named only “Tumor type 2 on CL” is recovered from the Fig. 1 tornado plot, which labels the +7% clearance bar “TUM TYP Bladder”; record that provenance in covariateData[[TUMTP_BLADDER]]$notes when a paper indexes tumor type numerically. Use this column for bladder / urothelial primaries specifically; upper-tract urothelial or mixed genitourinary pools that a paper does not resolve separately belong in TUMTP_OTHER.

TUMTP_BTC (canonical for biliary tract cancer tumor-type indicator)

• Description: 1 = biliary tract cancer (BTC; intrahepatic or extrahepatic cholangiocarcinoma and gallbladder carcinoma, as pooled by the source analysis), 0 = other tumor types.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 = all other tumor types (in Abegesah 2025 the pooled reference stratum is non-small-cell lung cancer).
• Source aliases:
  • tumtyp (integer-coded tumor-type column; level 3 in Abegesah 2025) – decompose into TUMTP_BTC = as.integer(tumtyp == 3).
• Example models: Abegesah_2025_durvalumab.R (multiplicative effect on durvalumab CL of (1 + 0.166), i.e. +16.6% relative to the pooled reference tumor type – the largest tumor-type effect in that model, and the only new covariate the analysis added to the inherited POSEIDON model; the stratum is the 314-patient TOPAZ-1 cohort).
• Notes: Follows the TUMTP_NSCLC / TUMTP_SCLC / TUMTP_HCC decomposition pattern. General scope because BTC has become a recurring checkpoint-inhibitor indication since the TOPAZ-1 (durvalumab) and KEYNOTE-966 (pembrolizumab) readouts, and the sibling Hwang_2022_tremelimumab.R cohort already carries a 6.2% biliary tract carcinoma stratum. Distinct from TUMTP_HCC (hepatocellular carcinoma): both are hepatobiliary but they are separately reported strata with different clearance effects, and papers that pool them say so. In Abegesah 2025 the tumor-type identity behind the Table 4 row named only “Tumor type 3 on CL” is recovered from the Fig. 1 tornado plot, which labels the +16.6% clearance bar “TUM TYP BTC”.

TUMTP_SCLC (canonical for small-cell-lung-cancer tumor-type indicator)

• Description: 1 = small cell lung cancer (SCLC), 0 = other tumor types.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 = all other tumor types (e.g., melanoma, NSCLC, RCC, HCC, CRC in the Sanghavi 2020 cohort; reference category is melanoma).
• Source aliases:
  • TUMTP (categorical column with levels including melanoma, NSCLC, SCLC, CRC, HCC, RCC) – decompose into TUMTP_SCLC = as.integer(TUMTP == "SCLC").
• Example models: Sanghavi_2020_ipilimumab.R (exponential coefficient -0.124 on CL), Wang_2026_serplulimab.R (selects the SCLC stratum of a seven-histology joint fit: baseline CL0 0.171 L/day and Vc 3.45 L, the lowest clearance and therefore the highest exposure of the seven; n = 390 / 2110 = 18.48% of the cohort, from the ASTRUM-005 Phase III trial).
• Notes: Follows the TUMTP_HODGKIN_CLASSICAL / TUMTP_GASTRIC decomposition pattern. SCLC is the only retained tumor-type indicator in the Sanghavi 2020 final model after backward elimination; the other tumor types collapse into the reference (melanoma) group.

TUMTP_NSCLC (canonical for non-small-cell-lung-cancer tumor-type indicator)

• Description: 1 = non-small cell lung cancer (NSCLC), 0 = other tumor types.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 = all other tumor types (in Ahamadi 2017 the implicit reference is melanoma, with the rare “other” cancer type pooled into the reference).
• Source aliases:
  • TUMTP (categorical column with levels including melanoma, NSCLC, other) – decompose into TUMTP_NSCLC = as.integer(TUMTP == "NSCLC").
• Example models: Ahamadi_2017_pembrolizumab.R (proportional change of +14.5% on CL for NSCLC patients relative to melanoma; the “other” cancer type cohort – 1.01% of the population – is pooled into the melanoma reference per the paper’s model description), Aoyama_2012_sepantronium.R.
• Notes: Follows the TUMTP_HODGKIN_CLASSICAL / TUMTP_GASTRIC / TUMTP_SCLC decomposition pattern. Scope: general because NSCLC is a high-frequency tumor-type contrast (with melanoma or “other” reference) across PD-1 / PD-L1 / chemotherapy popPK analyses, and is likely to recur in future extractions.

TUMTP_CSCC (canonical for cutaneous-squamous-cell-carcinoma tumor-type indicator)

• Description: 1 = cutaneous squamous cell carcinoma (CSCC; metastatic or locally advanced), 0 = other tumor types.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 = all other tumor types (in Nguyen 2022 the reference is NSCLC, with miscellaneous advanced solid tumors sharing the reference CL).
• Source aliases:
  • TUMTYPN (integer-coded tumor type; level 1 = CSCC in the Nguyen 2022 control stream) – decompose into TUMTP_CSCC = as.integer(TUMTYPN == 1).
• Example models: Nguyen_2022_cemiplimab.R (fractional change 1 + (-0.216) on cemiplimab baseline CL relative to NSCLC; n = 188 patients from Study 1540).
• Notes: Follows the TUMTP_<type> decomposition pattern. Distinct from TUMTP_NSCLC_SQUAM (squamous NSCLC histology) and TUMTP_ESCC (oesophageal squamous cell carcinoma): the CSCC primary is cutaneous.

TUMTP_BCC (canonical for basal-cell-carcinoma tumor-type indicator)

• Description: 1 = basal cell carcinoma (BCC; advanced, typically after hedgehog-inhibitor therapy), 0 = other tumor types.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 = all other tumor types (in Nguyen 2022 the reference is NSCLC).
• Source aliases:
  • TUMTYPN (integer-coded tumor type; level 2 = BCC in the Nguyen 2022 control stream) – decompose into TUMTP_BCC = as.integer(TUMTYPN == 2).
• Example models: Nguyen_2022_cemiplimab.R (fractional change 1 + (-0.211) on cemiplimab baseline CL and 1 + (-0.872) on the maximal time-dependent change in CL, so BCC patients show almost no decrease in CL over time; n = 132 patients from Study 1620).
• Notes: Follows the TUMTP_<type> decomposition pattern. Not to be confused with TUMTP_BCL (B-cell lymphoma).

TUMTP_NPC (canonical for nasopharyngeal-carcinoma tumor-type indicator)

• Description: 1 = nasopharyngeal carcinoma (NPC), 0 = other tumor types.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 = all other tumor types (in Zhang 2019 the explicit reference tumor type is NSCLC, with everything that is neither NSCLC nor NPC carried by the companion TUMTP_OTHER indicator).
• Source aliases:
  • NPC (Zhang 2019) – used directly as TUMTP_NPC = NPC.
  • TUMTP / TTYPE categorical columns carrying an NPC or nasopharyngeal level – decompose into TUMTP_NPC = as.integer(TUMTP == "NPC").
• Example models: Zhang_2019b_nivolumab.R (exponential coefficient +0.0889 on nivolumab baseline CL, i.e. exp(0.0889) = 1.093, a 9% higher clearance than the NSCLC reference; not statistically significant, RSE 90.9% with a 95% CI of -0.068 to 0.254 spanning zero; n = 23 / 1200 = 1.9% of the cohort, all of them Chinese, entering from the CheckMate 077 phase 1/2 study).
• Notes: Follows the TUMTP_HODGKIN_CLASSICAL / TUMTP_GASTRIC / TUMTP_SCLC / TUMTP_NSCLC decomposition pattern. Scope: general because NPC is an endemic-in-East-Asia histology that is routinely enrolled alongside NSCLC in Chinese and other Asian checkpoint-inhibitor trials, so the NPC-versus-NSCLC contrast is likely to recur with a stable meaning in future extractions. Distinct from TUMTP_OTHER: in the founding model both indicators coexist against the same NSCLC reference, NPC being a named histology of scientific interest and TUMTP_OTHER the residual bucket. When a paper reports an NPC effect, expect it to be imprecise – NPC cohorts in pooled solid-tumour popPK analyses are typically small (23 subjects in Zhang 2019), and the source authors there explicitly caution that “uncertainty is observed due to the small sample size (n = 23) of patients with NPC”.

TUMTP_NONLUNG (canonical for pooled non-lung-cancer tumor-type indicator)

• Description: 1 = any tumor type other than lung cancer; 0 = lung cancer (NSCLC or SCLC). This is the INVERSE-sense pooled indicator used by analyses whose cohort is predominantly lung cancer and which collapse every other histology into a single contrast level rather than fitting one indicator per histology. Use it only when the source paper itself pools that way; when the paper names individual non-lung histologies with separate coefficients, decompose into the per-histology members of the family (TUMTP_HCC, TUMTP_CRC, TUMTP_OTHER, …) instead.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 = lung cancer (per source paper; in Wang 2025 the reference group is the pooled NSCLC + SCLC cohort, 817 of 1144 subjects).
• Source aliases:
  • TUMTP (binary column coded 0 = lung cancer, 1 = all other tumor types) – use directly as TUMTP_NONLUNG = TUMTP.
  • TUMTP / CANCER_TYPE categorical columns with per-histology levels – collapse via TUMTP_NONLUNG = as.integer(!TUMTP %in% c("NSCLC", "SCLC", "lung")), but only if the source model fits a single pooled coefficient.
• Example models: Wang_2025_serplulimab.R (exponential coefficient -0.0887 on Vc; the pooled non-lung group is hepatocellular carcinoma 125, colorectal cancer 86 and “other” 116, i.e. 327 of 1144 subjects, and no tumor-type effect on CL was retained).
• Notes: Follows the TUMTP_HODGKIN_CLASSICAL / TUMTP_GASTRIC / TUMTP_SCLC / TUMTP_NSCLC decomposition pattern, but with the polarity reversed: the named group is the residual rather than the histology of interest. It is NOT interchangeable with TUMTP_OTHER, which is the leftover level alongside several named histology indicators; TUMTP_NONLUNG carries the entire non-reference cohort on its own. Scope: specific because the lung-vs-everything-else pooling is a property of a predominantly-lung cohort rather than a contrast that recurs with a stable meaning; a paper pooling differently should register or reuse the per-histology members instead.

TUMTP_NSCLC_NONADENO (canonical for non-adenocarcinoma-histology NSCLC sub-indicator)

• Description: 1 = NSCLC of a histology OTHER than adenocarcinoma (squamous-cell carcinoma, large-cell carcinoma, or other non-adenocarcinoma NSCLC subtype); 0 = adenocarcinoma NSCLC, NSCLC of unknown histology, or a non-NSCLC diagnosis. Within-NSCLC histology sub-indicator paired with TUMTP_NSCLC (which distinguishes NSCLC patients overall from other tumor types).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (adenocarcinoma NSCLC, unknown-histology NSCLC, or non-NSCLC subject; in particular IPF / non-cancer subjects in a mixed cohort are assigned 0 because they have no NSCLC histology).
• Source aliases:
  • Schmid 2017 categorical column with levels “NSCLC-adenocarcinoma” / “NSCLC-no adenocarcinoma” / “IPF or NSCLC of unknown histology” – decompose into TUMTP_NSCLC_NONADENO = as.integer(NSCLC_histology == "non-adenocarcinoma").
• Example models: Schmid_2017_nintedanib.R (multiplicative effect 1.36 on BIBF 1202 ka2 absorption rate; reference category adenocarcinoma NSCLC pooled with IPF and unknown-histology NSCLC).
• Notes: Follows the TUMTP_<type> decomposition pattern (auto-approved family). Within-NSCLC histology distinctions arise when a popPK / PD model in NSCLC differentiates adenocarcinoma from squamous-cell carcinoma; the paired adenocarcinoma reference can be made explicit by adding a TUMTP_NSCLC_ADENO sibling in a future extraction if needed. Distinct from TUMTP_NSCLC_SQUAM: non-adenocarcinoma NSCLC pools squamous with large-cell and other non-adenocarcinoma histologies, whereas TUMTP_NSCLC_SQUAM is the squamous-only indicator complementary to the pooled non-squamous group.

TUMTP_NSCLC_SQUAM (canonical for squamous-histology NSCLC sub-indicator)

• Description: 1 = squamous non-small cell lung cancer (Sq. NSCLC); 0 = any other tumor type, including non-squamous NSCLC. Within-NSCLC histology sub-indicator encoding the squamous / non-squamous dichotomy that oncology popPK analyses in lung cancer routinely stratify on, and that the NSCLC first-line treatment guidelines themselves split on.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (any non-squamous-NSCLC subject). When a model defines a full mutually exclusive TUMTP_<type> indicator set, the reference histology is the complement of all of them; in Wang_2026_serplulimab.R that complement is non-squamous NSCLC.
• Source aliases:
  • TUMTP (Wang 2026 serplulimab categorical column with levels HCC / CRC / Sq. NSCLC / Non-sq. NSCLC / SCLC / ESCC / Other) – decompose into TUMTP_NSCLC_SQUAM = as.integer(TUMTP == "Sq. NSCLC").
• Example models: Wang_2026_serplulimab.R (selects the squamous-NSCLC stratum’s baseline clearance 0.204 L/day and central volume 3.38 L from a seven-histology joint fit; n = 441 / 2110 = 20.90% of the cohort, from the ASTRUM-004 Phase III programme).
• Notes: Follows the TUMTP_<type> decomposition pattern (auto-approved family). Pairs with TUMTP_NSCLC (NSCLC overall vs other tumor types) and is distinct from TUMTP_NSCLC_NONADENO (which pools squamous with large-cell and other non-adenocarcinoma NSCLC). A paired explicit TUMTP_NSCLC_NONSQUAM is deliberately NOT registered here: in the founding model non-squamous NSCLC is the reference histology and is therefore the complement of the other six indicators, so an explicit column for it would be redundant and could double-count. Register it when a paper needs both indicators simultaneously against a non-lung reference.

TUMTP_HRPC (canonical for hormone-refractory prostate cancer tumor-type indicator)

• Description: 1 = hormone-refractory prostate cancer (HRPC), 0 = other tumor types. The historical term HRPC has since been displaced by CRPC (castration-resistant prostate cancer); the two terms refer to the same clinical entity. TUMTP_HRPC is the canonical column for both wordings (no separate TUMTP_CRPC is registered; if a future paper distinguishes hormone-naive metastatic prostate cancer from CRPC, register a more specific canonical at that time).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 = all other tumor types (per source paper; in Aoyama 2012 the implicit reference is NSCLC when paired with TUMTP_MEL = 0).
• Source aliases:
  • TUMTP (categorical column with levels including HRPC, NSCLC, MM) – decompose into TUMTP_HRPC = as.integer(TUMTP == "HRPC").
  • STUDY / CANCER_TYPE factor columns where one level is HRPC or CRPC – decompose identically.
• Example models: Aoyama_2012_sepantronium.R (proportional change of -4.5% on CL for HRPC patients relative to the NSCLC reference; ratio THETA_HRPC = 0.955 in the paper’s power form), Lacy_2018_cabozantinib.R (multiplicative fractional effect on CL/F = -0.009 and on Vc/F = -0.241 for CRPC patients relative to the healthy-volunteer reference; Lacy 2018 enrolled 823 CRPC patients across Studies XL184-203, XL184-306, and XL184-307, the largest cancer-type cohort in the integrated popPK analysis).
• Notes: Follows the TUMTP_HODGKIN_CLASSICAL / TUMTP_GASTRIC / TUMTP_SCLC / TUMTP_NSCLC decomposition pattern. Scope: general because prostate-cancer cohorts (under either HRPC or CRPC nomenclature) recur across small-molecule and targeted-therapy popPK analyses. The “hormone-refractory” wording reflects the 2010s convention; modern papers using CRPC map onto the same canonical.

TUMTP_MTC (canonical for medullary thyroid carcinoma tumor-type indicator)

• Description: 1 = medullary thyroid carcinoma (MTC), 0 = other tumor types. Time-fixed per subject. MTC is a distinct C-cell-derived thyroid malignancy (not papillary / follicular / anaplastic thyroid cancer); cabozantinib and vandetanib are the principal MTC-approved tyrosine-kinase inhibitors.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 = all other tumor types (per source paper; in Lacy 2018 the implicit reference is healthy volunteer when paired with all other TUMTP_* indicators = 0).
• Source aliases:
  • POP (Lacy 2018 NONMEM categorical population indicator with levels HV / RCC / CRPC / MTC / GB / Other) – decompose into TUMTP_MTC = as.integer(POP == "MTC").
• Example models: Lacy_2018_cabozantinib.R (multiplicative fractional effect on CL/F = +0.928 and on Vc/F = -0.07 for MTC patients relative to the healthy-volunteer reference; the +93% CL/F increase in MTC is the load-bearing finding of Lacy 2018 and explains why the MTC capsule label dose is 140 mg/day while the RCC tablet label dose is only 60 mg/day).
• Notes: Follows the TUMTP_HODGKIN_CLASSICAL / TUMTP_GASTRIC / TUMTP_SCLC / TUMTP_NSCLC decomposition pattern. Scope: general because MTC is the approved indication for Cometriq (cabozantinib capsule) and Caprelsa (vandetanib), and future TKI popPK papers in MTC populations are likely to recur.

TUMTP_RCC (canonical for renal cell carcinoma tumor-type indicator)

• Description: 1 = renal cell carcinoma (RCC), 0 = other tumor types. Time-fixed per subject. Includes both clear-cell and non-clear-cell histologies pooled at the cohort level; if a future paper distinguishes clear-cell from non-clear-cell, register a finer-grained canonical at that time.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 = all other tumor types (per source paper; in Lacy 2018 the implicit reference is healthy volunteer when paired with all other TUMTP_* indicators = 0).
• Source aliases:
  • POP (Lacy 2018 NONMEM categorical population indicator with levels HV / RCC / CRPC / MTC / GB / Other) – decompose into TUMTP_RCC = as.integer(POP == "RCC").
• Example models: Lacy_2018_cabozantinib.R (multiplicative fractional effect on CL/F = -0.129 and on Vc/F = -0.63 for RCC patients relative to the healthy-volunteer reference; Lacy 2018 enrolled 282 RCC patients in Study XL184-308 dosed at 60 mg/day cabozantinib tablet).
• Notes: Follows the TUMTP_HODGKIN_CLASSICAL / TUMTP_GASTRIC / TUMTP_SCLC / TUMTP_NSCLC decomposition pattern. Scope: general because RCC cohorts recur frequently across TKI, anti-PD-1, and anti-VEGF popPK analyses.

TUMTP_DTC (canonical for differentiated thyroid cancer tumor-type indicator)

• Description: 1 = differentiated thyroid cancer (DTC, i.e. papillary or follicular thyroid carcinoma including their variants; in most modelling cohorts specifically radioiodine-refractory DTC, RR-DTC), 0 = other tumor types or healthy subject. Time-fixed per subject. Distinct from TUMTP_MTC (medullary thyroid carcinoma), which arises from parafollicular C cells rather than the follicular epithelium and is a separate disease entity; do not pool the two.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 = all other tumor types (per source paper; in Majid 2024 the reference stratum is cancer patients with a solid tumor other than DTC, RCC or HCC, obtained by setting TUMTP_DTC, TUMTP_RCC, TUMTP_HCC and DIS_HEALTHY all to 0).
• Source aliases:
  • DTC (Majid 2024 Text S1 NONMEM flag derived from the study number: IF (STUD.EQ.303.OR.STUD.EQ.211.OR.STUD.EQ.201) DTC=1).
  • TUMTYP (Majid 2024 Text S1 categorical column with levels 1 = DTC, 2 = RCC, 3 = HCC, 4 = other solid tumor) – decompose into TUMTP_DTC = as.integer(TUMTYP == 1).
  • TUM / POP / CANCER_TYPE factor columns where one level is DTC, RR-DTC, papillary or follicular thyroid cancer – decompose identically.
• Example models: Majid_2024_lenvatinib.R (multiplicative power-form effect on CL/F: 0.951^TUMTP_DTC, i.e. -4.9 percent relative to the other-solid-tumor reference; 542 of 1921 pooled subjects were DTC).
• Notes: Follows the TUMTP_HODGKIN_CLASSICAL / TUMTP_GASTRIC / TUMTP_SCLC / TUMTP_NSCLC / TUMTP_RCC decomposition pattern. Scope: general because DTC cohorts recur across multikinase-inhibitor popPK analyses (lenvatinib, sorafenib, cabozantinib, vandetanib) and thyroid-cancer PK/PD work. Gupta_2016_lenvatinib.R records its DTC cohort only in population$disease_state because that model has no DTC covariate effect; that is not an omission of this column.

TUMTP_HCC (canonical for hepatocellular carcinoma tumor-type indicator)

• Description: 1 = hepatocellular carcinoma (HCC), 0 = other tumor types or healthy subject. Time-fixed per subject. Note that an HCC cohort usually also carries altered hepatic-function laboratory covariates (albumin, alkaline phosphatase, bilirubin, Child-Pugh class); when a model fits both this indicator and continuous liver-function markers, the indicator captures the residual tumor-type effect after those markers are accounted for.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 = all other tumor types (per source paper; in Majid 2024 the reference stratum is cancer patients with a solid tumor other than DTC, RCC or HCC).
• Source aliases:
  • HCC (Majid 2024 Text S1 NONMEM flag derived from the study number: IF (STUD.EQ.202.OR.STUD.EQ.304) HCC=1).
  • TUMTYP (Majid 2024 Text S1 categorical column with levels 1 = DTC, 2 = RCC, 3 = HCC, 4 = other solid tumor) – decompose into TUMTP_HCC = as.integer(TUMTYP == 3).
  • POP / CANCER_TYPE factor columns where one level is HCC or hepatocellular – decompose identically.
• Example models: Majid_2024_lenvatinib.R (multiplicative power-form effect on CL/F: 0.862^TUMTP_HCC, i.e. -13.8 percent relative to the other-solid-tumor reference; 534 of 1921 pooled subjects were HCC), Nguyen_2019_cabozantinib.R, Nguyen_2019_cabozantinib_liver_dysfunction.R (same multiplicative power form, 0.878^TUMTP_HCC on CL/F and 0.847^TUMTP_HCC on Vc/F in the final model, and 0.82 / 0.81 in the liver-dysfunction variant; 489 of 2023 pooled subjects were HCC, from studies XL184-203 and CELESTIAL XL184-309. The pair illustrates the interaction this entry’s Description warns about: the HCC effect on CL/F strengthens from 0.878 to 0.82 when NCI-ODWG hepatic-function indicators enter the same model, because 65-68% of the HCC patients carried mild impairment and the HEPIMP_MILD factor of 1.12 absorbs part of the tumor-type stratum. The reference here is the healthy volunteer, not another solid tumor), Cheng_2021_avadomide.R (fractional effect V2/F x (1 + 0.521) relative to the healthy-subject reference, the CL/F effect held at zero by the authors; n = 27 of 298).
• Notes: Follows the TUMTP_HODGKIN_CLASSICAL / TUMTP_GASTRIC / TUMTP_SCLC / TUMTP_NSCLC / TUMTP_RCC decomposition pattern. Scope: general because HCC cohorts recur across TKI, anti-PD-(L)1 and anti-VEGF popPK analyses. Distinct from the HEPIMP_* (NCI ODWG hepatic-impairment) family, which classifies hepatic function by bilirubin and AST regardless of tumor type; an HCC patient may have normal hepatic function and a non-HCC patient may be hepatically impaired, so the two families are not substitutes.

TUMTP_MEL (canonical for melanoma tumor-type indicator)

• Description: 1 = melanoma (any anatomic site; in advanced-solid-tumor cohorts typically unresectable stage III or IV cutaneous melanoma), 0 = other tumor types.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 = all other tumor types (per source paper; in Aoyama 2012 the implicit reference is NSCLC when paired with TUMTP_HRPC = 0).
• Source aliases:
  • TUMTP (categorical column with levels including MM, NSCLC, HRPC) – decompose into TUMTP_MEL = as.integer(TUMTP == "MM"). NOTE: in Aoyama 2012 the abbreviation MM denotes malignant (unresectable) melanoma, NOT multiple myeloma; do not confuse with the canonical MM register entry (which is specifically active multiple myeloma).
  • STUDY / CANCER_TYPE factor columns where one level is melanoma or MM (melanoma sense) – decompose identically.
• Example models: Aoyama_2012_sepantronium.R (proportional change of +24% on CL for melanoma patients relative to the NSCLC reference; ratio THETA_MM = 1.24 in the paper’s power form), Hu_2024_ipilimumab.R (exponential effect on ipilimumab CL for pediatric (< 18 y) melanoma: exp(-0.347) = 0.707, i.e. 29% lower than the adult-melanoma reference; adult melanoma is the reference level and carries no coefficient).
• Notes: Follows the TUMTP_HODGKIN_CLASSICAL / TUMTP_GASTRIC / TUMTP_SCLC / TUMTP_NSCLC decomposition pattern. The canonical name uses MEL (not MM) to disambiguate from the existing MM register entry for multiple myeloma. Scope: general because melanoma cohorts recur across PD-1 / PD-L1 / BRAF-inhibitor / small-molecule popPK analyses.

TUMTP_LYMPH (canonical for lymphoma (pooled) tumor-type indicator)

• Description: 1 = lymphoma (heterogeneous lymphoma pool spanning multiple lymphoma histologies – e.g., classical Hodgkin lymphoma combined with extranodal NK/T-cell lymphoma; or any-histology lymphoma pooled with solid-tumor and leukemia cohorts), 0 = solid tumor or other tumor type.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 = non-lymphoma tumor type (per source paper; e.g., NSCLC, GC/GEJ, ESCC, “Other” solid tumors in the Wang 2024 cohort, with NSCLC as the implicit reference when paired with the other Wang 2024 TUMTP_* indicators; or leukemia as the implicit reference in the Akbar 2025 cohort).
• Source aliases:
  • TTYPE1 (Wang 2024) – decompose into TUMTP_LYMPH = as.integer(TTYPE1 == 1). The Wang 2024 source paper uses a multi-level TTYPE factor with levels 1 = lymphoma, 2 = lung cancer (reference), 3 = other, 4 = GCGEJ, 5 = ESCC.
  • Categorical column “type of cancer” with level “Lymphoma” (Akbar 2025) – decompose into TUMTP_LYMPH = as.integer(cancer_type == "Lymphoma").
• Example models: Wang_2024_sugemalimab.R (exponential coefficient log(0.877) on baseline CL and log(0.879) on Vc), Akbar_2025_voriconazole.R (additive-fractional +1.91% effect on CL relative to leukemia reference; 95% CI spans zero).
• Notes: Distinct from TUMTP_HODGKIN_CLASSICAL (which is specifically classical Hodgkin lymphoma). Wang 2024 pools two lymphoma histologies (extranodal NK/T-cell lymphoma from CS1001-201 / NCT03595657 and classical Hodgkin lymphoma from CS1001-202 / NCT03505996) into a single lymphoma indicator; the indicator therefore captures a generic “hematologic-vs-solid-tumor” contrast rather than a histology-specific effect. Akbar 2025 uses a single “Lymphoma” category alongside leukemia, sarcoma, breast cancer, myeloma, and glioma in a heterogeneous-cancer TDM cohort. When a future paper studies a single lymphoma histology distinct from cHL, register a more specific canonical (e.g., TUMTP_ENKTL, TUMTP_NHL) rather than overloading this one. Document the per-paper histology composition in covariateData[[TUMTP_LYMPH]]$notes. Promoted from Scope: specific to Scope: general on 2026-05-09 alongside the Akbar 2025 voriconazole extraction so that any heterogeneous-cancer-cohort PK analysis can use this canonical name without scope-violation.

TUMTP_BREAST (canonical for breast-cancer tumor-type indicator)

• Description: 1 = breast cancer (any histology / receptor status), 0 = other tumor types.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 = all other tumor types (per source paper; in Lu 2022 the implicit reference is NSCLC, with colorectal cancer pooled into the reference because its CL effect was found insignificant relative to NSCLC).
• Source aliases:
  • TUMTP_BC – prior canonical name (pre-2026-06-19 standardization audit).
  • TUMTP (categorical column with levels including BC, NSCLC, CRC) – decompose into TUMTP_BREAST = as.integer(TUMTP == "BC").
• Example models: Lu_2022_patritumab.R (multiplicative fractional effect 0.811 on CLlin of DXd-conjugated antibody for breast-cancer patients vs the NSCLC reference; CRC effect was tested and found insignificant so CRC is pooled into the reference).
• Notes: Follows the TUMTP_HODGKIN_CLASSICAL / TUMTP_GASTRIC / TUMTP_SCLC decomposition pattern. Registers the breast-cancer arm of an oncology-cohort tumor-type contrast; pair with sister TUMTP_<GROUP> indicators (e.g., TUMTP_NSCLC, TUMTP_CRC) when a future paper retains separate effects for additional tumor types beyond the implicit reference.

TUMTP_TNBC (canonical for triple-negative breast cancer tumor-type indicator)

• Description: 1 = triple-negative breast cancer (estrogen-receptor, progesterone-receptor and HER2 negative), 0 = other tumor types.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 = all other tumor types (per source paper; in Terranova 2021 the reference is the most common ‘other tumor types’ category, with non-TNBC breast cancer pooled into it).
• Source aliases:
  • Tumor type: TNBC (Terranova 2021 berzosertib Figure 2 forest-plot label) – decompose a categorical tumor-type column into TUMTP_TNBC = as.integer(TUMTYPE == "TNBC").
• Example models: Terranova_2021_berzosertib.R (linear categorical multipliers 1 + e_tumtp_tnbc_<param> * TUMTP_TNBC of +0.42 on CL, +0.29 on V1 and -0.307 on V2 versus the ‘other tumor types’ reference; n = 33 / 240 = 13.8% of the cohort, mostly the basaloid-subtype TNBC expansion cohort of Study 001 Part C2).
• Notes: Receptor-status-defined subset of breast cancer. Use TUMTP_TNBC when a paper separates TNBC from other breast cancers; use TUMTP_BREAST when it pools all breast cancer. The two are not meant to be set together in one model: a paper that keeps both a TNBC and a non-TNBC breast group should pair TUMTP_TNBC with a separately registered non-TNBC indicator rather than overloading TUMTP_BREAST. A given subject can have at most one of the TUMTP_<GROUP> indicators set to 1; all-zero means the reference group.

TUMTP_CRC (canonical for colorectal-cancer tumor-type indicator)

• Description: 1 = colorectal cancer (any site / histology, including metastatic CRC), 0 = other tumor types.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 = all other tumor types (per source paper; in Zhang 2024 the reference is a healthy participant, obtained when TUMTP_CRC = 0 and every sister TUMTP_<GROUP> indicator is also 0).
• Source aliases:
  • TUM (Zhang 2024 tucatinib; integer tumor-type code where TUM == 1 is HER2+ mCRC, TUM == 0 is HER2+ mBC and TUM < 0 is a healthy participant) – decompose into TUMTP_CRC = as.integer(TUM == 1).
  • TUMTP (categorical column with levels including BC, NSCLC, CRC) – decompose into TUMTP_CRC = as.integer(TUMTP == "CRC").
• Example models: Zhang_2024_tucatinib.R (multiplicative fractional effect -0.705 on CL and -0.637 on relative bioavailability for HER2+ metastatic colorectal cancer versus the healthy-participant reference; the two partly offset, giving a net 18.7% lower apparent CL/F; n = 69 / 283 = 24.4% of the pooled cohort, all from study SGNTUC-017 / MOUNTAINEER), Yang_2026_encorafenib.R (linear categorical multiplier 1 + e_tumtp_crc_cl * TUMTP_CRC on the day-1 apparent clearance CL/F, coefficient -0.175 relative to a melanoma reference; n = 240 / 1310 = 18.3% of the pooled cohort, from the C4221001 and BEACON metastatic-CRC studies).
• Notes: Follows the TUMTP_HODGKIN_CLASSICAL / TUMTP_GASTRIC / TUMTP_SCLC decomposition pattern, and is the sister indicator anticipated by the TUMTP_BREAST entry. Registers the colorectal arm of an oncology-cohort tumor-type contrast. General scope because “colorectal cancer” is a stable histology-agnostic disease label rather than a per-paper pooled bucket; use TUMTP_OTHER instead when a paper collapses CRC into a heterogeneous residual pool (as Sathe 2024 / Sathe 2025 do), and pool CRC into the reference when a paper tests it and finds no effect (as Lu 2022 does). A given subject can have at most one of the TUMTP_<GROUP> indicators set to 1; all-zero means the reference group. Receptor status (e.g. HER2+) and RAS status are not encoded here – record them in covariateData[[TUMTP_CRC]]$notes when the source cohort is selected on them.

TUMTP_ESCC (canonical for oesophageal-squamous-cell-carcinoma tumor-type indicator)

• Description: 1 = oesophageal squamous cell carcinoma (ESCC), 0 = other tumor types.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 = all other tumor types (per source paper; in Wang 2024, the implicit reference is NSCLC when all the other TUMTP_* indicators are also 0).
• Source aliases:
  • TTYPE5 (Wang 2024) – decompose into TUMTP_ESCC = as.integer(TTYPE5 == 1).
• Example models: Wang_2024_sugemalimab.R (exponential coefficient log(0.99) on baseline CL and log(1.08) on Vc), Wang_2026_serplulimab.R (selects the ESCC stratum of a seven-histology joint fit: baseline CL0 0.178 L/day and Vc 3.48 L, the largest central volume of the seven; n = 389 / 2110 = 18.44% of the cohort, from the HLX10-007-EC301 Phase III trial; the source Figure 1 labels this row ‘EC’).
• Notes: Follows the TUMTP_HODGKIN_CLASSICAL / TUMTP_GASTRIC / TUMTP_SCLC decomposition pattern. Distinct from gastroesophageal-junction adenocarcinoma (which is captured by the broader TUMTP_GASTRIC indicator that pools GC and GEJ adenocarcinomas) – ESCC is a squamous-cell histology, not adenocarcinoma. Document the per-paper histology composition in covariateData[[TUMTP_ESCC]]$notes.

TUMTP_PCALCL (canonical for primary cutaneous anaplastic large-cell lymphoma indicator)

• Description: 1 = primary cutaneous anaplastic large-cell lymphoma (pcALCL), 0 = other tumor types.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 = all other tumor types (in Suri 2018, the non-pcALCL reference comprises Hodgkin lymphoma, systemic ALCL, mycosis fungoides, and other CD30+ hematologic malignancies pooled together).
• Source aliases:
  • PCALCL – used in Suri_2018_brentuximab.R. Suri 2018 reports the effect as a power-form multiplier cl_adc *= 0.728^TUMTP_PCALCL (pcALCL CL ~27% lower than non-pcALCL).
• Example models: Suri_2018_brentuximab.R (effect on ADC clearance only).
• Notes: Follows the TUMTP_HODGKIN_CLASSICAL / TUMTP_GASTRIC / TUMTP_SCLC decomposition pattern. Distinct from TUMTP_LYMPH (heterogeneous lymphoma pool) and TUMTP_HODGKIN_CLASSICAL (classical Hodgkin lymphoma). pcALCL is one of two histologies pooled into the broader CTCL category in Suri 2018 (alongside mycosis fungoides); the model singles out pcALCL because Suri 2018 backward elimination retained pcALCL as a separate effect on ADC clearance after exploring the broader CTCL contrast.

TUMTP_SARC (canonical for sarcoma tumor-type indicator)

• Description: 1 = sarcoma (any histology – soft-tissue or bone sarcoma pooled), 0 = other tumor types.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 = all other tumor types (per source paper; in Akbar 2025 the implicit reference is leukemia when paired with the other Akbar TUMTP_* indicators all = 0).
• Source aliases:
  • Categorical column “type of cancer” with level “Sarcoma” – decompose into TUMTP_SARC = as.integer(cancer_type == "Sarcoma"). Used in Akbar_2025_voriconazole.R.
• Example models: Akbar_2025_voriconazole.R (additive-fractional +18.5% effect on CL relative to leukemia reference).
• Notes: Follows the TUMTP_HODGKIN_CLASSICAL / TUMTP_GASTRIC / TUMTP_SCLC decomposition pattern. Akbar 2025 pools soft-tissue and bone sarcoma histologies into a single sarcoma category. Scope: specific because the reference category (leukemia in Akbar 2025) is paper-defined.

TUMTP_MYELO (canonical for multiple myeloma tumor-type indicator (used in heterogeneous-cancer pooled cohorts))

• Description: 1 = multiple myeloma, 0 = other tumor types.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 = all other tumor types (per source paper; in Akbar 2025 the implicit reference is leukemia when paired with the other Akbar TUMTP_* indicators all = 0).
• Source aliases:
  • Categorical column “type of cancer” with level “Myeloma” – decompose into TUMTP_MYELO = as.integer(cancer_type == "Myeloma"). Used in Akbar_2025_voriconazole.R.
• Example models: Akbar_2025_voriconazole.R (additive-fractional -2.33% effect on CL relative to leukemia reference; the 95% CI spans zero), Cheng_2021_avadomide.R (fractional effects CL/F x (1 + 0.309) and V2/F x (1 + 0.682) relative to the healthy-subject reference; n = 29 of 298).
• Notes: Distinct from the stub MM entry (which is reserved for multiple-myeloma-as-primary-indication PK studies; the MM definition lacks a complete schema and predates the TUMTP_* convention) and from DIS_SMM (smoldering multiple myeloma, an asymptomatic plasma-cell disorder). Use TUMTP_MYELO when the source paper pools multiple myeloma alongside other tumor types in a heterogeneous oncology cohort and treats cancer type as a many-level categorical covariate. Scope: specific because the reference category (leukemia in Akbar 2025) is paper-defined.

TUMTP_GLIO (canonical for glioma tumor-type indicator)

• Description: 1 = glioma (any grade / histology – e.g., glioblastoma, anaplastic astrocytoma, oligodendroglioma pooled), 0 = other tumor types.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 = all other tumor types (per source paper; in Akbar 2025 the implicit reference is leukemia when paired with the other Akbar TUMTP_* indicators all = 0; in Lacy 2018 the reference is healthy volunteer when paired with the other TUMTP_* indicators = 0).
• Source aliases:
  • Categorical column “type of cancer” with level “Glioma” – decompose into TUMTP_GLIO = as.integer(cancer_type == "Glioma"). Used in Akbar_2025_voriconazole.R.
  • POP (Lacy 2018 NONMEM categorical population indicator) – decompose into TUMTP_GLIO = as.integer(POP == "GB") (Lacy 2018 enrolled glioblastoma multiforme patients in Study XL184-201 dosed at 140 mg/day cabozantinib capsule).
• Example models: Akbar_2025_voriconazole.R (additive-fractional +8.81% effect on CL relative to leukemia reference; the 95% CI spans zero), Lacy_2018_cabozantinib.R (multiplicative fractional effect on CL/F = +0.216 and on Vc/F = -0.569 for GB patients relative to the healthy-volunteer reference; n = 39 GB patients in Study XL184-201), Gastonguay_2005_efaproxiral.R (categorical power-model multiplier on SLPp50 = 1.28 for cranial glioblastoma multiforme vs lung-cancer reference; CATP level 3; n = 72 GBM patients / 451 = 16.0% of radiation-therapy cohort), Nguyen_2019_cabozantinib.R, Nguyen_2019_cabozantinib_liver_dysfunction.R (the updated refit of the Lacy 2018 integrated cabozantinib model; unlike its predecessor, Nguyen 2019 Table 3 reports the effect as a direct MULTIPLICATIVE factor rather than a fractional change, so GB carries 1.2^TUMTP_GLIO on CL/F and 0.479^TUMTP_GLIO on Vc/F in the final model, and 1.2 / 0.448 in the liver-dysfunction variant; same n = 39 GB patients from Study XL184-201, now 2% of the enlarged 2023-subject cohort), Cheng_2021_avadomide.R (glioblastoma multiforme; fractional effect V2/F x (1 + 0.480) relative to the healthy-subject reference, the CL/F effect held at zero by the authors; the paper’s separate ‘brain cancer’ category is TUMTP_BRAIN; n = 44 of 298).
• Notes: Follows the TUMTP_HODGKIN_CLASSICAL / TUMTP_GASTRIC / TUMTP_SCLC decomposition pattern. Akbar 2025 reports a single “glioma” category without further subdivision by histology or grade. Scope: specific because the reference category (leukemia in Akbar 2025; healthy volunteer in Lacy 2018; lung cancer in Gastonguay 2005) is paper-defined.

TUMTP_CNS_PRIM (canonical for primary CNS tumour umbrella indicator)

• Description: 1 = primary central nervous system (CNS) tumour (an umbrella pooling glioma, medulloblastoma, ependymoma, brainstem glioma, atypical teratoid / rhabdoid tumour, and other tumours arising within the brain or spinal cord), 0 = non-primary-CNS tumour (per-source-paper reference cohort; e.g. sarcomas in Han 2015). Distinct from TUMTP_GLIO, which is glioma-only; use TUMTP_CNS_PRIM when the source paper’s binary tumour-type contrast pools glioma with non-glioma primary CNS histologies. The PRIM suffix distinguishes primary CNS tumours from metastatic CNS involvement (which is classified by primary site of origin under the other TUMTP_* entries). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 = non-primary-CNS tumour (paper-defined; for Han 2015 the complement cohort is the pooled paediatric sarcomas: refractory sarcomas AVF2771s + newly-diagnosed osteosarcoma AVF4117s + metastatic soft-tissue sarcoma BO20924).
• Source aliases:
  • Study-based indicator “Primary CNS tumour” as defined by paper narrative (Han 2015 Table 1 splits the paediatric cohort into two clinical groups: primary-CNS-tumour patients in AVF3842s and sarcoma patients in AVF2771s + AVF4117s + BO20924, with n = 76 in each group of the model-building population).
• Example models: Hu_2024_nivolumab.R (exponential effect on nivolumab baseline CL: exp(-0.578) in adults – an essentially all-glioblastoma cohort – and exp(-0.801) in pediatric subjects < 18 years, relative to an adult-melanoma reference; adult CNS-tumour subjects additionally have the time-varying-CL Emax term set to zero, and pediatric CNS-tumour subjects carry an Emax offset of +0.696), Hu_2024_ipilimumab.R (exponential effect on ipilimumab CL: exp(-0.661) in adults and exp(-0.668) in pediatric subjects < 18 years, relative to an adult-melanoma reference), Han_2015_bevacizumab.R (multiplicative fractional effect: CL x 0.725^TUMTP_CNS_PRIM and V1 x 0.854^TUMTP_CNS_PRIM relative to the paediatric-sarcoma reference; Han 2015 Table 2 rows ‘Primary CNS tumour on CL’ = 0.725 and ‘Primary CNS tumour on V1’ = 0.854 with RSE 4.3% and 3.7% respectively).
• Notes: Follows the TUMTP_CHL / TUMTP_GC / TUMTP_SCLC decomposition pattern. Use this canonical when the source paper reports a single umbrella indicator for primary CNS tumours as a group rather than resolving individual histologies (glioma / medulloblastoma / ependymoma). When a future paper resolves specific primary-CNS-tumour histologies separately, register the finer-grained canonical(s) alongside; the umbrella indicator can coexist with the specific ones because the pooling is a paper-level analytic decision. Scope: specific because the reference category is paper-defined.

TUMTP_NET (canonical for neuroendocrine-tumour tumor-type indicator)

• Description: 1 = neuroendocrine tumour (NET, including metastasised gastroenteropancreatic and other well-differentiated NETs), 0 = other tumour types (per-source-paper reference cohort). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 = other tumour types (paper-defined; in the Budiansah 2025 cohort the complement is meningioma, the other somatostatin-receptor-positive tumour type enrolled for peptide receptor radionuclide therapy).
• Source aliases:
  • Net / NETs – Kletting 2016 J Nucl Med Table 1 ‘Disease’ column, whose two levels are Net and Men (meningioma); decompose into TUMTP_NET = as.integer(Disease == "Net").
• Example models: Budiansah_2025_dotatate_pbpk.R (whole-body sst2 PBPK in which the tumour is the only tumour-type-specific region: the indicator switches the tumour interstitial fraction 0.3 vs 0.23, vascular fraction 0.1 vs 0.11, serum flow density 1.0 vs 0.9 mL/min/g and permeability-surface density 0.2 vs 0.31 mL/min/g).
• Notes: Follows the TUMTP_HODGKIN_CLASSICAL / TUMTP_GASTRIC / TUMTP_SCLC decomposition pattern. Distinct from TUMTP_MTC (medullary thyroid carcinoma), which is a specific neuroendocrine histology that a paper pooling all NETs would subsume; register the finer-grained canonical alongside when a source resolves NET subtypes. In radioligand-therapy papers the tumour type usually enters as a physiologic switch (perfusion, permeability, interstitial fraction) rather than as a covariate-effect coefficient on a PK parameter.

TUMTP_PANC (canonical for pancreatic-cancer tumor-type indicator)

• Description: 1 = pancreatic cancer (metastatic pancreatic adenocarcinoma), 0 = other tumor types (per-source-paper reference cohort). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 = all other tumor types. In Zhu_2013_ganitumab.R the complement is “other advanced solid cancers” (the pooled non-pancreatic advanced-solid-tumour cohort of Studies 2 and 3).
• Source aliases:
  • Categorical “cancer type” / “tumour type” column with a pancreatic level – decompose into TUMTP_PANC = as.integer(cancer_type == "pancreatic").
• Example models: Zhu_2013_ganitumab.R (ganitumab popPK; the single most significant covariate in the analysis – pancreatic cancer type raises typical CL 1.7-fold, from 0.679 to 1.154 L/day, and typical Vc 1.3-fold, from 3.85 to 5.13 L, relative to the non-pancreatic reference).
• Notes: Follows the TUMTP_GASTRIC / TUMTP_SCLC decomposition pattern; a subject may carry at most one TUMTP_<GROUP> indicator, and all-zero means the reference group. Note the boundary with TUMTP_OTHER: papers whose analysis plan buries pancreatic ductal adenocarcinoma inside a heterogeneous “other” residual bucket (e.g. Sathe_2024_sacituzumab.R, Sathe_2025_sacituzumab.R) must keep using TUMTP_OTHER – TUMTP_PANC is only for analyses that resolve pancreatic cancer as its own contrast.

TUMTP_LEUK (canonical for leukemia tumor-type indicator)

• Description: 1 = leukemia (any subtype – AML / ALL / CLL / CML pooled), 0 = other tumor types. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 = all other tumor types (per source paper). In Akbar 2025 leukemia is the implicit reference (column not used directly), but the canonical name is registered so that future papers that retain leukemia as a non-reference contrast can use it.
• Source aliases:
  • Categorical column “type of cancer” with level “Leukemia” – decompose into TUMTP_LEUK = as.integer(cancer_type == "Leukemia"). Implicit reference category in Akbar_2025_voriconazole.R (so the model file does not consume this column directly; it is registered for future heterogeneous-cancer-cohort analyses).
• Example models: Akbar_2025_voriconazole.R.
• Notes: Distinct from the more specific DIS_AML, DIS_BCPALL, DIS_CMML, DIS_MDS_AML entries – those are for leukemia-only or leukemia-vs-leukemia contrasts; TUMTP_LEUK is for heterogeneous-cancer pooled cohorts where leukemia is one of several tumor types and the analysis treats cancer type as a many-level categorical. Akbar 2025 had leukemia as 56.8% of the cohort and used it as the reference category. Scope: specific because the reference category in any source paper is paper-defined.

TUMTP_SOLID (canonical for solid-tumor (vs hematologic malignancy) tumor-type indicator)

• Description: 1 = patient with a solid-tumor malignancy, 0 = patient with a hematologic malignancy. Time-fixed per subject. For pooled oncology analyses that combine hematologic-malignancy and solid-tumor cohorts and treat tumor type as a two-level contrast.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 = hematologic malignancy (in Lin_2020_glasdegib.R, the pooled AML / MDS / other hematologic-malignancy patients of studies B1371001 and B1371003).
• Source aliases:
  • Solid – the indicator name in the Lin 2020 glasdegib final-model equations (Vp/F = 279.21 x (BWT/70) x (1 - 0.825 x Solid)); used in Lin_2020_glasdegib.R.
  • TMR (“tumor type”, “solid vs. heme tumor”) – used in Liva_2021_rec2282.R with the OPPOSITE orientation (haematologic = 1), applied as (1 - TUMTP_SOLID). The paper does not state which level is coded 1; the maintainers read the first-named level (solid) as the reference.
• Example models: Lin_2020_glasdegib.R (linear effects (1 + e x TUMTP_SOLID) on apparent peripheral volume, -0.825, and apparent intercompartmental clearance, -0.653; also selects the solid-tumor residual-error SD), Liva_2021_rec2282.R (linear fractional effect on the absorption lag time for haematologic malignancy, ALAG = TV * (1 + 0.942 * (1 - TUMTP_SOLID)), Liva 2021 Table 3 ‘TMR’; haematologic = multiple myeloma, lymphoma and AML, 40 of 56 patients; every patient of the OSU11130 trial had AML, so the authors note the effect may be a study effect).
• Notes: Distinct from DIS_CANCER, whose contrast is oncology vs NON-oncology (healthy volunteers or a non-oncology disease cohort); TUMTP_SOLID contrasts two oncology cohorts with each other. Member of the TUMTP_<type> family. Scope: specific because the complement group (which hematologic malignancies are pooled into the reference) is paper-defined.

TUMTP_BCL (canonical for B-cell lymphoma (pooled residual) tumor-type indicator)

• Description: 1 = B-cell lymphoma (BCL), 0 = other tumor types. Time-fixed per subject. In Gibiansky 2014 the BCL category is a pooled residual indolent-B-cell-lymphoma group that includes follicular lymphoma (FL was the primary indication in GAUDI; the four-level DIS column in the NONMEM control stream splits B-cell histologies into CLL = 1, BCL = 2 (residual indolent B-cell-lymphoma pool including FL), DLBCL = 3, MCL = 4).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 = all other tumor types (in Gibiansky 2014 the implicit reference is CLL when paired with TUMTP_DLBCL and TUMTP_MCL all = 0; the residual indolent-B-cell-lymphoma pool in this paper is dominated by follicular lymphoma).
• Source aliases:
  • DIS (Gibiansky 2014; integer code with DIS == 2 flagging BCL) – decompose into TUMTP_BCL = as.integer(DIS == 2).
• Example models: Gibiansky_2014_obinutuzumab.R (effect on time-dependent clearance decay rate kdes via the composite (TUMTP_BCL + TUMTP_DLBCL + TUMTP_MCL) (any-NHL effect; ratio 2.08) and on both time-dependent CL_T and steady-state CL_inf via the composite (TUMTP_BCL + TUMTP_DLBCL) (shared BCL/DLBCL effect; ratio 0.834 in the reverse direction, i.e., 16.6% lower CL than CLL)).
• Notes: Follows the TUMTP_HODGKIN_CLASSICAL / TUMTP_GASTRIC / TUMTP_SCLC decomposition pattern. Distinct from TUMTP_LYMPH (a broader heterogeneous lymphoma pool that lumps cHL with NHL histologies) – TUMTP_BCL is specifically B-cell lymphoma and pairs with sibling TUMTP_DLBCL and TUMTP_MCL for histology-specific contrasts within the NHL family. When a future paper studies follicular lymphoma in isolation (rather than pooled into BCL), register a more specific canonical (e.g., TUMTP_FL) rather than overloading this one.

TUMTP_ATLL (canonical for adult T-cell leukemia/lymphoma indicator)

• Description: 1 = adult T-cell leukemia/lymphoma (ATLL), 0 = other tumor types. ATLL is an aggressive mature T-cell neoplasm caused by human T-cell lymphotropic virus type 1 (HTLV-1) and is endemic to south-western Japan, the Caribbean basin, west Africa and parts of South America. Although ATLL is formally a peripheral T-cell lymphoma (PTCL) subtype, its histopathology, etiology, clinical presentation and prognosis differ enough that trials and population analyses routinely carry it as a stratum distinct from the other PTCL subtypes – so a cohort containing both needs this indicator alongside a PTCL reference (or vice versa) rather than a single pooled lymphoma flag.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0. The complement is paper-defined and must be recorded per model: in Inoue_2025_valemetostat.R the reference is peripheral T-cell lymphoma (with DIS_BCELLNHL and DIS_HEALTHY carrying the remaining two levels of a four-level population-type covariate), while the companion exposure-response models invert the contrast and carry a TUMTP_PTCL indicator against an ATLL reference.
• Source aliases:
  • ATLL – Inoue 2025 Table 2 row CL/F ~ ATLL and Table 1 disease type/subtype.
  • ATL – the shorter spelling used in some Japanese-sponsor datasets for the same entity. No value transformation.
• Example models: Inoue_2025_valemetostat.R (multiplicative effect on unbound valemetostat clearance, 0.828 versus a PTCL reference; the paper concludes disease type has no clinically meaningful effect because the point estimate and its uncertainty sit almost entirely inside the 0.8-1.25 reference range).
• Notes: Follows the TUMTP_<type> decomposition pattern alongside TUMTP_PCALCL, TUMTP_DLBCL and TUMTP_MCL. Distinct from TUMTP_LYMPH (heterogeneous lymphoma pool) and from TUMTP_OTHER_NHL, whose registered meaning is a non-follicular non-DLBCL residual histology category that explicitly includes T-cell NHL and therefore must not be used to encode an ATLL-versus-PTCL contrast. Distinct too from DIS_BCELLNHL, which carries the B-cell lymphoma stratum of the same cohorts.

TUMTP_PTCL (canonical for peripheral T-cell lymphoma tumor-type indicator)

• Description: 1 = peripheral T-cell lymphoma (PTCL), 0 = other tumor types. Time-fixed per subject. PTCL here means the pooled non-ATLL mature T-cell lymphoma group – angioimmunoblastic T-cell lymphoma, PTCL not otherwise specified, anaplastic large-cell lymphoma and the rarer subtypes together – as used when a cohort carries PTCL and ATLL as two strata of one patient-type covariate.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0. The complement is paper-defined and must be recorded per model. In the Inoue_2025_valemetostat_* exposure-response family the reference is adult T-cell leukemia/lymphoma, so TUMTP_PTCL is the inverse contrast of the TUMTP_ATLL used by the companion population PK model on the same cohort.
• Source aliases:
  • PTCL – Inoue 2025 Tables S5-S10 row Patient type: PTCL.
  • Patient type (two-level categorical with levels ATLL and PTCL) – decompose as TUMTP_PTCL = as.integer(patient_type == "PTCL").
• Example models: Inoue_2025_valemetostat_anemia.R, Inoue_2025_valemetostat_neutropenia.R, Inoue_2025_valemetostat_thrombocytopenia.R, Inoue_2025_valemetostat_teae_grade3.R, Inoue_2025_valemetostat_dose_reduction.R, Inoue_2025_valemetostat_dose_interruption.R (all six safety exposure-response models carry it as both a main effect on the logit and an interaction on the exposure slope, against an ATLL reference).
• Notes: Follows the TUMTP_<type> pattern. The paired-inverse relationship with TUMTP_ATLL is deliberate and is a property of the source analysis, not of the register: Inoue 2025 fits the population PK model with ATLL as the indicator and a PTCL reference, then fits the exposure-response models on a PTCL + ATLL subset with the contrast reversed. A model must carry whichever direction its source paper estimated – do not silently re-base one onto the other, because the two are not numerically interchangeable once interaction terms are present. Distinct from TUMTP_LYMPH (heterogeneous lymphoma pool, includes B-cell) and from the finer efficacy-model subtype indicators TUMTP_PTCL_NOS, TUMTP_ALCL and TUMTP_PTCL_OTHER, which decompose the PTCL group itself.

TUMTP_PTCL_NOS (canonical for peripheral T-cell lymphoma not-otherwise-specified indicator)

• Description: 1 = peripheral T-cell lymphoma, not otherwise specified (PTCL-NOS), 0 = other tumor types. Time-fixed per subject. PTCL-NOS is the diagnosis of exclusion within the mature T-cell lymphomas: nodal PTCL that does not meet criteria for AITL, ALCL or any other defined entity.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0. In Inoue_2025_valemetostat_orr_bicr.R the reference level of the four-level subtype covariate is angioimmunoblastic T-cell lymphoma (AITL), denoted by all three indicators being 0; AITL has no canonical of its own because no shipped model carries it as a column.
• Source aliases:
  • PTCL, NOS – Inoue 2025 Table S4 row PTCL subtypes: PTCL, NOS.
  • PTCL-NOS, PTCL NOS – equivalent spellings. No value transformation.
• Example models: Inoue_2025_valemetostat_orr_bicr.R (odds ratio 0.662 on the response logit and interaction 0.164 on the exposure slope, versus an AITL reference).
• Notes: Follows the TUMTP_<type> pattern. Part of the three-indicator decomposition of the Inoue 2025 efficacy PTCL-subtype covariate alongside TUMTP_ALCL and TUMTP_PTCL_OTHER, with AITL as the unmodelled reference level. Distinct from TUMTP_PTCL, which pools all of these subtypes into one stratum contrasted against ATLL in the companion safety models.

TUMTP_ALCL (canonical for anaplastic large-cell lymphoma tumor-type indicator)

• Description: 1 = anaplastic large-cell lymphoma (ALCL), 0 = other tumor types. Time-fixed per subject. Covers systemic ALCL of either ALK-positive or ALK-negative status; the source analyses that use this indicator do not separate the two.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0. In Inoue_2025_valemetostat_orr_bicr.R the reference level of the four-level subtype covariate is angioimmunoblastic T-cell lymphoma (AITL), denoted by all three indicators being 0; AITL has no canonical of its own because no shipped model carries it as a column.
• Source aliases:
  • ALCL – Inoue 2025 Table S4 row PTCL subtypes: ALCL.
• Example models: Inoue_2025_valemetostat_orr_bicr.R (odds ratio 0.923 on the response logit and interaction 0.0157 on the exposure slope, versus an AITL reference).
• Notes: Follows the TUMTP_<type> pattern. Distinct from TUMTP_PCALCL, which is primary cutaneous ALCL – a separate entity with a markedly better prognosis that is staged and treated differently. Do not merge the two: a cohort reporting ALCL without qualification in a systemic-PTCL trial means systemic ALCL and belongs here, while a cutaneous-lymphoma cohort’s ALCL belongs in TUMTP_PCALCL. If a future source separates ALK-positive from ALK-negative systemic ALCL, register parallel canonicals rather than overloading this one.

TUMTP_PTCL_OTHER (canonical for residual ‘other PTCL subtypes’ indicator)

• Description: 1 = a peripheral T-cell lymphoma subtype outside the named strata of the source’s subtype decomposition, 0 = otherwise. Time-fixed per subject. A study-defined residual category: its membership is whatever the source did not name explicitly, so the exact set of histologies differs between papers and must be recorded per model.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0. In Inoue_2025_valemetostat_orr_bicr.R the named strata are AITL (reference), PTCL-NOS and ALCL, so this indicator carries every other enrolled PTCL histology.
• Source aliases:
  • All other subtypes – Inoue 2025 Table S4 row PTCL subtypes: All other subtypes.
  • Other PTCL – equivalent phrasing in other sources.
• Example models: Inoue_2025_valemetostat_orr_bicr.R (odds ratio 0.879 on the response logit and interaction -0.155 on the exposure slope, versus an AITL reference).
• Notes: Follows the TUMTP_<type> pattern. Residual categories are only comparable within a single decomposition. Because membership is defined by exclusion, two papers’ TUMTP_PTCL_OTHER columns mean different things whenever their named strata differ, so a model must record its own paper’s named strata in covariateData[[TUMTP_PTCL_OTHER]]$notes. The same caution applies to TUMTP_OTHER and TUMTP_OTHER_NHL. Distinct from TUMTP_PTCL, which is the whole PTCL group rather than its unnamed remainder.

TUMTP_DLBCL (canonical for diffuse large B-cell lymphoma indicator)

• Description: 1 = diffuse large B-cell lymphoma (DLBCL), 0 = other tumor types. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 = all other tumor types (in Gibiansky 2014 the implicit reference is CLL when paired with TUMTP_BCL and TUMTP_MCL all = 0).
• Source aliases:
  • DIS (Gibiansky 2014; integer code with DIS == 3 flagging DLBCL) – decompose into TUMTP_DLBCL = as.integer(DIS == 3).
• Example models: Gibiansky_2014_obinutuzumab.R (effect on kdes via the any-NHL composite indicator; effect on CL_T and CL_inf via the shared BCL/DLBCL composite indicator), Lu_2017_polatuzumab_neuropathy.R (multiplicative effect on the grade >= 2 peripheral neuropathy hazard via exp(theta_DLBCL * TUMTP_DLBCL) with theta_DLBCL = -0.0697, SE 0.365; the median hazard ratio of DLBCL vs FL is 0.931 per Figure 3), Cheng_2021_avadomide.R (fractional effects CL/F x (1 - 0.647) and V2/F x (1 + 0.476) relative to the healthy-subject reference, with PCNSL and the residual NHL group coded as separate indicators; n = 60 DLBCL of 298).
• Notes: Follows the TUMTP_HODGKIN_CLASSICAL / TUMTP_GASTRIC / TUMTP_SCLC decomposition pattern. Distinct from TUMTP_LYMPH (broader lymphoma pool) and TUMTP_PCALCL (primary cutaneous anaplastic large-cell lymphoma; a CD30+ T-cell-lineage entity unrelated to DLBCL). DLBCL is the most common high-grade B-cell-NHL subtype; the Gibiansky 2014 cohort had only 30 DLBCL patients (4.4%), so a single estimated effect on CL is shared with BCL (a much larger pooled group; see TUMTP_BCL notes).

TUMTP_MCL (canonical for mantle cell lymphoma indicator)

• Description: 1 = mantle cell lymphoma (MCL), 0 = other tumor types. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 = all other tumor types (in Gibiansky 2014 the implicit reference is CLL when paired with TUMTP_BCL and TUMTP_DLBCL all = 0).
• Source aliases:
  • DIS (Gibiansky 2014; integer code with DIS == 4 flagging MCL) – decompose into TUMTP_MCL = as.integer(DIS == 4).
• Example models: Gibiansky_2014_obinutuzumab.R (effect on kdes via the any-NHL composite indicator; separate effect on CL_T and CL_inf via the standalone MCL indicator (ratio 1.75, i.e., 75% higher CL than CLL)).
• Notes: Follows the TUMTP_HODGKIN_CLASSICAL / TUMTP_GASTRIC / TUMTP_SCLC decomposition pattern. Distinct from TUMTP_LYMPH (broader lymphoma pool). The Gibiansky 2014 cohort had only 20 MCL patients (2.9%); the paper reports the highest obinutuzumab CL among the four B-cell-malignancy histologies for MCL, consistent with the highest CD20 expression density on MCL B-cells relative to the other histologies.

TUMTP_FL (canonical for follicular lymphoma indicator)

• Description: 1 = follicular lymphoma (FL), 0 = other tumor types. Time-fixed per subject. FL is the most common indolent non-Hodgkin lymphoma (NHL) subtype, derived from germinal-center B cells and characterized by the t(14;18) BCL2-IGH translocation in most cases.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 = all other tumor types (in Lu 2017 the implicit reference is the union of DLBCL and other non-FL NHL when paired with TUMTP_DLBCL = 0 and TUMTP_OTHER_NHL = 0; the Lu 2017 covariate model uses FL itself as the reference category, so the indicator does not appear explicitly in model – both TUMTP_DLBCL and TUMTP_OTHER_NHL are 0 for FL patients).
• Source aliases:
  • Categorical “tumor histology” column with level FL (Lu 2017) – decompose into TUMTP_FL = as.integer(tumor_histology == "FL").
• Example models: Lu_2017_polatuzumab_neuropathy.R (FL is the implicit reference category; the model file declares TUMTP_FL in covariatesDataExcluded to record FL as the reference without referencing it in model()).
• Notes: Follows the TUMTP_HODGKIN_CLASSICAL / TUMTP_GASTRIC / TUMTP_SCLC decomposition pattern. Distinct from TUMTP_BCL (a broader pooled-residual indolent-B-cell-lymphoma indicator in Gibiansky 2014 that lumps FL with other indolent histologies); use TUMTP_FL when the source paper names follicular lymphoma in isolation.

TUMTP_OTHER_NHL (canonical for ‘other non-FL non-DLBCL NHL’ tumor-histology indicator)

• Description: 1 = a non-follicular-lymphoma (non-FL) non-diffuse-large-B-cell-lymphoma (non-DLBCL) NHL histology (e.g., mantle cell lymphoma, marginal-zone lymphoma, small-lymphocytic lymphoma, transformed FL, chronic lymphocytic leukemia, T-cell NHL, etc., pooled together in the paper’s residual histology category), 0 = FL or DLBCL or any non-NHL tumor. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (FL or DLBCL or any non-NHL tumor; in Lu 2017 the implicit reference is FL when paired with TUMTP_DLBCL = 0).
• Source aliases:
  • Categorical “tumor histology” column with level not in {FL, DLBCL} (Lu 2017 “others” category) – decompose into TUMTP_OTHER_NHL = as.integer(!tumor_histology %in% c("FL", "DLBCL")).
• Example models: Lu_2017_polatuzumab_neuropathy.R (multiplicative effect on hazard of grade >= 2 peripheral neuropathy via the exp(theta_otherNonFL * TUMTP_OTHER_NHL) proportional-hazard term with theta_otherNonFL = 0.688, SE 0.758; the median hazard ratio of “other non-FL” vs FL is 1.98 per Figure 3 with very wide uncertainty given the small “others” sub-cohort).
• Notes: Captures the residual NHL-histology bucket in papers that decompose tumor histology into a three-level categorical (FL reference, DLBCL, otherNonFL). Distinct from TUMTP_BCL (Gibiansky 2014 four-level B-cell-lymphoma decomposition: CLL, BCL, DLBCL, MCL – where BCL is the indolent-B-cell-lymphoma residual including FL). When a future paper retains a different pool of non-FL non-DLBCL histologies, document the per-paper histology composition in covariateData[[TUMTP_OTHER_NHL]]$notes rather than overloading the canonical.

TUMTP_PCNSL (canonical for primary CNS lymphoma tumor-type indicator)

• Description: 1 = primary central nervous system lymphoma (PCNSL), an extranodal non-Hodgkin lymphoma (almost always diffuse large B-cell histology) confined to the brain, spinal cord, leptomeninges or eyes; 0 = otherwise. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0. In Cheng_2021_avadomide.R the typical-value reference is the healthy subject (every TUMTP_* indicator = 0).
• Source aliases:
  • PCNSL level of a categorical tumor-type column (Cheng 2021 Table 1 ‘Tumor’ row) – decompose into TUMTP_PCNSL = as.integer(tumor == "PCNSL").
• Example models: Cheng_2021_avadomide.R (fractional effects CL/F x (1 - 0.692) and V2/F x (1 + 0.846) relative to healthy subjects; n = 5 of 298).
• Notes: Follows the TUMTP_<type> pattern. Distinct from TUMTP_CNS_PRIM, the umbrella for primary CNS tumours that pools glioma with other brain and spinal-cord histologies, and from TUMTP_DLBCL: a paper that codes PCNSL as its own level alongside DLBCL treats the two as mutually exclusive, so a PCNSL patient has TUMTP_DLBCL = 0 even though the histology is usually diffuse large B-cell.

TUMTP_NHL (canonical for non-Hodgkin lymphoma tumor-type indicator)

• Description: 1 = non-Hodgkin lymphoma (NHL), 0 = otherwise. Time-fixed per subject. When the same model also codes named NHL subtypes as their own mutually exclusive levels (e.g. TUMTP_DLBCL, TUMTP_PCNSL), TUMTP_NHL = 1 only for the NHL patients outside those subtypes, so the indicator is then a paper-defined remainder and its membership must be recorded per model.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0. In Cheng_2021_avadomide.R the typical-value reference is the healthy subject.
• Source aliases:
  • NHL level of a categorical tumor-type column (Cheng 2021 Table 1 ‘Tumor’ row, alongside separate DLBCL and PCNSL levels) – decompose into TUMTP_NHL = as.integer(tumor == "NHL").
• Example models: Cheng_2021_avadomide.R (NHL other than DLBCL and PCNSL, histologies not enumerated; fractional effect V2/F x (1 + 0.344) relative to healthy subjects, the CL/F effect held at zero by the authors; n = 30 of 298).
• Notes: Follows the TUMTP_<type> pattern and the TUMTP_LYMPH entry’s recommendation to register an NHL-specific canonical rather than overload the broader lymphoma pool. Distinct from TUMTP_OTHER_NHL, whose definition excludes follicular lymphoma as well as DLBCL: use that entry only when the paper’s residual NHL bucket is defined as non-FL non-DLBCL. Distinct from TUMTP_LYMPH, which also pools Hodgkin lymphoma. As with TUMTP_PTCL_OTHER, residual membership is comparable only within one decomposition, so record the named strata in covariateData[[TUMTP_NHL]]$notes.

TUMTP_BRAIN (canonical for ‘brain cancer’ tumor-type indicator)

• Description: 1 = brain cancer as a paper-defined tumor-type level, 0 = otherwise. Time-fixed per subject. Use it when the source reports a “brain cancer” or “brain tumor” category without naming its histologies; when the same model codes named brain histologies separately (e.g. TUMTP_GLIO, TUMTP_PCNSL), TUMTP_BRAIN holds only the remainder.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0. In Cheng_2021_avadomide.R the typical-value reference is the healthy subject.
• Source aliases:
  • Brain cancer level of a categorical tumor-type column (Cheng 2021 Table 1 ‘Tumor’ row, alongside separate GBM and PCNSL levels) – decompose into TUMTP_BRAIN = as.integer(tumor == "Brain cancer").
• Example models: Cheng_2021_avadomide.R (brain cancer other than GBM and PCNSL; fractional effect V2/F x (1 + 0.415) relative to healthy subjects, the CL/F effect held at zero by the authors; n = 6 of 298).
• Notes: Follows the TUMTP_<type> pattern. Distinct from TUMTP_CNS_PRIM, which is an explicit umbrella for primary CNS tumours including glioma, and from TUMTP_GLIO (glioma). Whether the category includes brain metastases is not stated in the founding paper; record what the source says in covariateData[[TUMTP_BRAIN]]$notes.

LINE_1L (canonical for first-line-therapy indicator)

• Description: 1 = first-line therapy (1L) / treatment-naive, 0 = second-line or greater (2L+) / relapsed-or-refractory.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (2L+, second-line or greater / relapsed-refractory).
• Source aliases:
  • LINE (categorical column with levels 1L, 2L, 3L+,…) – decompose into LINE_1L = as.integer(LINE == "1L").
  • RRFN (relapsed/refractory flag; treatment-naive corresponds to RRFN == 0) – used in Lu_2019_polatuzumab.R. Decompose: LINE_1L = as.integer(RRFN == 0).
• Example models: Sanghavi_2020_ipilimumab.R (exponential coefficient -0.0949 on CL), Lu_2019_polatuzumab.R (multiplicative effects on V1 = 1.20, kdes = 3.38, CL_T = 3.53, FRAC_NS = 0.756; the same pooled-trial NHL cohort mixes 415 R/R and 45 first-line patients), Shulgin_2020_ici_trae34_covariate_mbma.R (study-arm level; the source codes the opposite-polarity Line2+, formed as 1 - LINE_1L, which reduces the anti-CTLA-4 exposure slope of the grade 3/4 trAE logit by 0.003 per unit normalized exposure; mixed-line cohorts were excluded), Zhang_2022_ici_irae_mbma.R, Zhang_2022_ici_irae_grade3_mbma.R (study-arm level; the source codes the opposite-polarity line2+, formed as 1 - LINE_1L, an additive -0.4757 on the any-grade and -0.6285 on the grade >= 3 irAE logit; mixed-line cohorts were rounded to 0 or 1 by majority).
• Notes: Promoted to scope: general on 2026-04-26 after Lu 2019 polatuzumab vedotin ratified the same 1L vs 2L+ binarization that Sanghavi 2020 ipilimumab introduced. The two papers use different indicator semantics (Sanghavi reports the effect as exp(-0.0949 * LINE_1L) and Lu reports theta^LINE_1L with theta < or > 1 depending on the parameter); both reduce to the same canonical 0/1 column. If a future paper requires finer resolution (separate effects for 2L vs 3L+), add a parallel LINE_2L canonical rather than overloading this one. LINE_3L and LINE_4L_PLUS are the first such parallel canonicals, registered 2026-09-11.

LINE_3L (canonical for third-line-therapy indicator)

• Description: 1 = the analyzed treatment is the subject’s third line of therapy (3L), i.e. the subject received exactly two prior anticancer regimens; 0 = otherwise. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0. Paired with LINE_4L_PLUS to decompose a three-level prior-therapy covariate; both 0 denotes second-line therapy (one prior regimen), which is the reference level in Inoue_2025_valemetostat_*.
• Source aliases:
  • Number of prior regimens: 2 – Inoue 2025 Tables S4-S10. Note the off-by-one: two prior regimens means the study treatment is the third line. See the Notes below.
  • Number of prior lines of therapy: 3 / LINE == "3L" – sources that report lines directly map with no arithmetic.
• Example models: Inoue_2025_valemetostat_orr_bicr.R, Inoue_2025_valemetostat_anemia.R, Inoue_2025_valemetostat_neutropenia.R, Inoue_2025_valemetostat_thrombocytopenia.R, Inoue_2025_valemetostat_teae_grade3.R, Inoue_2025_valemetostat_dose_reduction.R, Inoue_2025_valemetostat_dose_interruption.R (all seven exposure-response models carry it as a main effect on the logit and an interaction on the exposure slope, against a one-prior-regimen / second-line reference).
• Notes: THE LINE NUMBER IS ONE MORE THAN THE PRIOR-REGIMEN COUNT. A source column counting prior regimens must be incremented before it is mapped onto this family: N prior regimens means line N + 1. So LINE_3L = as.integer(n_prior_regimens == 2) and LINE_4L_PLUS = as.integer(n_prior_regimens > 2), while a reference level of one prior regimen is second-line therapy with both indicators 0. A data assembler who maps n_prior_regimens == 3 onto LINE_3L has introduced a silent off-by-one that shifts every subject by one line and is invisible in any summary statistic of the indicator itself. Record which convention the source used in covariateData[[LINE_3L]]$source_name. Naming ratified by the maintainers on 2026-09-11, who chose the treatment-line vocabulary over a NREG_2 / NREG_GT2 prior-regimen-count form so that one clinical concept stays in one family with LINE_1L; the cost of that choice is exactly the off-by-one documented here, and the mapping N prior regimens -> line N + 1 is the standard oncology convention but is not universal – some trials count regimens rather than lines, and the two diverge when a regimen switch occurs within a line or when maintenance therapy is counted separately. Verify against the source’s own definition before assembling data.

LINE_4L_PLUS (canonical for fourth-line-or-later-therapy indicator)

• Description: 1 = the analyzed treatment is the subject’s fourth or later line of therapy (4L+), i.e. the subject received more than two prior anticancer regimens; 0 = otherwise. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0. Paired with LINE_3L; both 0 denotes second-line therapy (one prior regimen) in Inoue_2025_valemetostat_*.
• Source aliases:
  • Number of prior regimens: >2 – Inoue 2025 Tables S4-S10. Off-by-one applies: more than two prior regimens means fourth line or later.
  • LINE == "4L+" / Number of prior lines of therapy: >=4 – direct mapping, no arithmetic.
• Example models: Inoue_2025_valemetostat_orr_bicr.R, Inoue_2025_valemetostat_anemia.R, Inoue_2025_valemetostat_neutropenia.R, Inoue_2025_valemetostat_thrombocytopenia.R, Inoue_2025_valemetostat_teae_grade3.R, Inoue_2025_valemetostat_dose_reduction.R, Inoue_2025_valemetostat_dose_interruption.R (all seven carry it as a main effect on the logit and an interaction on the exposure slope).
• Notes: See LINE_3L for the full off-by-one warning, the N prior regimens -> line N + 1 mapping rule and the 2026-09-11 naming ratification; the same caveats apply here. This indicator is open-ended on the right, so it pools 4L, 5L and beyond – if a future source estimates separate effects for those, register a further parallel canonical (LINE_5L_PLUS, or explicit LINE_4L / LINE_5L) rather than redefining this one, following the precedent LINE_1L set.

NIVO_1Q3W (canonical for nivolumab 1 mg/kg every 3 weeks co-administration indicator)

• Description: 1 = ipilimumab co-administered with nivolumab 1 mg/kg every 3 weeks; 0 = otherwise (monotherapy or any other nivolumab regimen).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no nivolumab or any non-1Q3W nivolumab regimen).
• Source aliases:
  • NIVO_REGIMEN (categorical column with levels none, 0.3 mg/kg Q3W, 1 mg/kg Q2W, 1 mg/kg Q3W, 3 mg/kg Q2W, 3 mg/kg Q3W) – decompose into NIVO_1Q3W = as.integer(NIVO_REGIMEN == "1 mg/kg Q3W").
• Example models: Sanghavi_2020_ipilimumab.R (exponential coefficient 0.0950 on ipilimumab CL).
• Notes: Paired with NIVO_3Q2W in the Sanghavi 2020 final model; both decomposed indicators are 0 for ipilimumab monotherapy. Other nivolumab regimens (0.3 mg/kg Q3W, 1 mg/kg Q2W, 3 mg/kg Q3W) were tested but not retained in the final model and collapse into the reference 0 group.

NIVO_3Q2W (canonical for nivolumab 3 mg/kg every 2 weeks co-administration indicator)

• Description: 1 = ipilimumab co-administered with nivolumab 3 mg/kg every 2 weeks; 0 = otherwise (monotherapy or any other nivolumab regimen).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no nivolumab or any non-3Q2W nivolumab regimen).
• Source aliases:
  • NIVO_REGIMEN (categorical column) – decompose into NIVO_3Q2W = as.integer(NIVO_REGIMEN == "3 mg/kg Q2W").
• Example models: Sanghavi_2020_ipilimumab.R (exponential coefficient 0.191 on ipilimumab CL).
• Notes: Paired with NIVO_1Q3W; same reference grouping convention.

COMBO_NIVO (canonical for any-regimen nivolumab combination-therapy indicator)

• Description: 1 = ipilimumab co-administered with any nivolumab regimen, 0 = ipilimumab monotherapy.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (ipilimumab monotherapy).
• Source aliases:
  • COMBO – used in Sanghavi_2020_ipilimumab.R. Equivalently derivable from NIVO_REGIMEN as COMBO_NIVO = as.integer(NIVO_REGIMEN != "none").
• Example models: Sanghavi_2020_ipilimumab.R (additive effect -0.202 on the Emax parameter of the time-varying CL function).
• Notes: Distinct from the per-regimen NIVO_1Q3W / NIVO_3Q2W indicators on baseline CL: COMBO_NIVO aggregates across all nivolumab regimens and acts on the time-varying-CL Emax parameter, whereas the per-regimen indicators act on baseline (time-zero) CL.

BMBLAST_PCT (canonical for baseline percentage of blasts in bone marrow)

• Description: Baseline percentage of nucleated bone-marrow cells that are blasts, from the screening / baseline bone-marrow aspirate or biopsy. Time-fixed baseline value. A marrow tumor-burden measure in AML / MDS analyses.
• Units: %
• Type: continuous
• Scope: specific
• Reference category: n/a – used with linear centering (1 + slope x (BMBLAST_PCT - ref)). Reference value observed: 38.2% (Lin_2020_glasdegib.R, the median among the hematologic-malignancy patients of the population PK dataset).
• Source aliases:
  • BPBL (baseline percentage blasts in bone marrow) – Lin 2020 glasdegib population PK source-column name; used in Lin_2020_glasdegib.R.
• Example models: Lin_2020_glasdegib.R (linear effect -0.004 per percentage point on CL/F, centered at 38.2%), Lin_2020_glasdegib_treatment.R (covariatesDataExcluded), Lin_2020_glasdegib_decitabine.R.
• Notes: Distinct from BLSTPB (percentage blasts in PERIPHERAL BLOOD – a different specimen) and from BLSTABL (absolute peripheral-blood blast count). Not collected in solid-tumor patients; a pooled hematologic + solid-tumor analysis must state what value the solid-tumor subjects carry (Lin 2020 imputes missing continuous covariates at the population median, which makes the centered effect vanish). Scope: specific because the covariate is only meaningful in myeloid-malignancy analyses; promote to general if a second paper retains it.

BLSTPB (canonical for baseline percentage of blasts in peripheral blood)

• Description: Baseline percentage of circulating peripheral-blood leukocytes that are blasts (immature lymphoid/myeloid precursor cells). Time-fixed baseline value. A relative measure of circulating tumour burden in B-cell leukemia population PK analyses, where it proxies the size of the target (CD22) sink available to an anti-CD22 antibody-drug conjugate.
• Units: %
• Type: continuous
• Scope: specific
• Reference category: n/a – used with power scaling (BLSTPB / <ref>)^exponent. Reference value observed: 5.25% (Garrett_2019_inotuzumab.R, the pooled-ALL median used in the published equations; the paper’s Results prose rounds the same percentile to “5%” in two places and Table 1 reports the ALL median as 4.00%, but 5.25 is the value the final-model equations and the deposited control stream both center on).
• Source aliases:
  • BLSTPB – used in Garrett_2019_inotuzumab.R (the source column name in that paper’s NONMEM $INPUT).
• Example models: Garrett_2019_inotuzumab.R (power exponent -0.0401 on cl_exp_kdes, the decay coefficient of the empirical time-dependent clearance, for B-cell ALL patients only; the effect is gated off for B-cell NHL by multiplying the exponent by DIS_BCPALL).
• Notes: Distinct from BLSTABL, the ABSOLUTE blast count in the same specimen (10^9 counts/L), which is the column the successor Wu_2024_inotuzumab.R reanalysis retains in place of this one – the two are different quantities on different scales and must not be interconverted or aliased to each other. Also distinct from blasts measured in bone marrow (different specimen) and from the CD22-positive subset (CD22PPB in the Garrett 2019 source data, screened but not retained). Percentage of peripheral-blood blasts is collected only in leukemia cohorts; in pooled leukemia + lymphoma analyses the source data carries a missing-data sentinel for the lymphoma subjects (Garrett 2019 Online Resource 3: IF(BLSTPB.EQ.-99) KDESBLSTPB = 1), which nlmixr2lib models reproduce by gating the exponent on the disease indicator rather than by requiring the sentinel value. Scope: specific because the covariate is only meaningful in B-cell-leukemia population PK analyses; promote to general if a second paper retains it.

BLSTABL (canonical for baseline absolute blast counts in peripheral blood)

• Description: Baseline absolute count of blasts (immature lymphoid/myeloid precursor cells) circulating in peripheral blood. Time-fixed baseline value.
• Units: 10^9 counts/L (equivalently 10^9 counts; reported as “x 10^9 counts” in Wu 2024 Table 2).
• Type: continuous
• Scope: specific
• Reference category: n/a – used with power scaling (BLSTABL / <ref>)^exponent. Reference value observed: 0.352 x 10^9 counts (Wu 2024 Table 3, BCP-ALL median).
• Source aliases:
  • BLSTABL – used in Wu_2024_inotuzumab.R.
• Example models: Wu_2024_inotuzumab.R (power exponent -0.0484 on kdes for BCP-ALL patients only; the effect is gated off for B-cell NHL by multiplying the exponent by DIS_BCPALL).
• Notes: Distinct from blasts in bone marrow (different specimen) and from BLSTPB (percentage of blasts in peripheral blood, used by the predecessor Garrett_2019_inotuzumab.R adult model – a different quantity on a different scale; do not interconvert or alias the two). Not applicable for B-cell NHL patients in pooled BCP-ALL + NHL analyses (Wu 2024 retains the effect only in BCP-ALL patients via the DIS_BCPALL gate). When supplying BLSTABL for an NHL subject, set the value to the BCP-ALL reference (0.352) so the gated power term evaluates to 1 numerically. Scope: specific because the covariate is most meaningful in B-cell-leukemia population PK analyses; promote to general if a second paper retains it.

COMBO_RG (canonical for anti-CD20 (rituximab or obinutuzumab) combination-therapy indicator)

• Description: 1 = polatuzumab vedotin co-administered with rituximab OR obinutuzumab, 0 = single-agent polatuzumab vedotin (or any other regimen lacking an anti-CD20 partner). Time-fixed per subject in the source paper’s analysis cohort.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (single-agent polatuzumab vedotin or no anti-CD20 partner).
• Source aliases:
  • COMBO (categorical: 0 = single agent, 1 = + rituximab, 2 = + obinutuzumab) – used in Lu_2019_polatuzumab.R. Decompose: COMBO_RG = as.integer(COMBO == 1 | COMBO == 2). The Lu 2019 NONMEM separately defines RTX = as.integer(COMBO == 1) and GA101 = as.integer(COMBO == 2) and applies effects as theta^(RTX + GA101); because RTX and GA101 are mutually exclusive, RTX + GA101 takes values {0, 1} and the effect collapses to theta^COMBO_RG.
• Example models: Lu_2019_polatuzumab.R (multiplicative effects on CL_INF = 0.844, kdes = 0.932, FRAC_NS = 0.709).
• Notes: Rituximab and obinutuzumab both bind CD20 on B cells (rituximab is a Type I anti-CD20 mAb, obinutuzumab a glycoengineered Type II), so co-administration is hypothesized to alter polatuzumab vedotin disposition through depletion of CD79b+ target B cells. The Lu 2019 final model fits a single combined effect rather than separate rituximab- and obinutuzumab-specific effects. Scope: specific because the relevant combination partners (CD20-directed mAbs) are tied to NHL pathway; if a future paper distinguishes rituximab from obinutuzumab combinations, register COMBO_R and COMBO_G separately rather than overloading this canonical.

COMBO_DURVA (canonical for durvalumab combination-therapy indicator)

• Description: 1 = the analyzed therapeutic mAb is co-administered with durvalumab (anti-PD-L1 IgG1), 0 = monotherapy.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (monotherapy).
• Source aliases:
  • COMB – used in Hwang_2022_tremelimumab.R ($INPUT NM-TRAN data item; control-stream switch IF(COMB.EQ.0) selects monotherapy parameters and IF(COMB.EQ.1) selects combination-therapy parameters).
• Example models: Hwang_2022_tremelimumab.R (selects between monotherapy and combination-with-durvalumab values of the time-varying-CL Tmax and lambda parameters).
• Notes: Parallels COMBO_NIVO but for durvalumab rather than nivolumab co-administration. Acts on the time-varying-CL component (Tmax and lambda); baseline CL is shared between monotherapy and combination groups in Hwang 2022.

COMBO_LEN_DEX (canonical for lenalidomide plus dexamethasone combination-therapy indicator)

• Description: 1 = the analyzed therapeutic mAb (or other agent under PK study) is co-administered with the lenalidomide + low-dose-dexamethasone (Ld) backbone, 0 = monotherapy or any non-Ld regimen. Lenalidomide is an immunomodulatory imide (IMiD) that activates natural killer cells; dexamethasone is an immunosuppressant glucocorticoid. The Ld backbone is a standard combination partner in multiple-myeloma and other hematologic-malignancy regimens.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (monotherapy or any non-Ld regimen). When a paper reports the reference patient as “with Ld coadministration” (as Ide 2020 does), the model still stores the canonical 0/1 column and applies the effect as exp(theta * (COMBO_LEN_DEX - 1)) so that COMBO_LEN_DEX = 1 yields factor 1 (paper’s reference) and COMBO_LEN_DEX = 0 activates the effect.
• Source aliases:
  • LENDEX – in Ide_2022_elotuzumab.R the source flag is 1 for both the Ld and the Pd studies (STUDY != 204011), so COMBO_LEN_DEX = LENDEX - POMDEX there.
  • LENDEX (1 = with Ld, 0 = without Ld; in Ide 2020 derived from STUDY != 204011 because study 204011 was the Ld-free elotuzumab-monotherapy cohort) – used in Ide_2020_elotuzumab.R.
  • COMBO_LD (retired canonical name; renamed to COMBO_LEN_DEX on 2026-04-27 for clarity).
• Example models: Ide_2022_elotuzumab.R (with COMBO_POM_DEX: the monotherapy contrast CLMono = 0.825 and KINTMono = 9.78 applies when both indicators are 0, the Pd contrast when COMBO_POM_DEX = 1; Ld is the reference), Ide_2020_elotuzumab.R (multiplicative effects: CLLd = 0.74 on nonspecific CL, encoded as exp(log(0.74) * (COMBO_LEN_DEX - 1)); KINTLd = 10.1 on the second-order target-mediated elimination rate from the peripheral compartment, encoded as exp(log(10.1) * (COMBO_LEN_DEX - 1))).
• Notes: Specific scope because the canonical’s mechanistic relevance is hematologic-malignancy-domain-bound (multiple myeloma and related plasma-cell or B-cell disorders). Distinct from COMBO_BELAMAF (which pools Ld with bortezomib-dex and pomalidomide-dex into a single broader “any-combination” belantamab indicator); COMBO_LEN_DEX is the per-backbone Ld-only flag. If a future paper distinguishes “Ld-only” from a broader “any-IMiD-plus-dex” backbone with separate effects, register a parallel canonical (e.g., COMBO_PD for pomalidomide-dex, COMBO_VD for bortezomib-dex). Sign of the exponential coefficient is paper-dependent: Ide 2020 reports CLLd = 0.74 so the Ld-coadministration arm has 26% lower nonspecific CL than the Ld-free arm, but KINTLd = 10.1 so the Ld arm has 10x higher second-order target-mediated elimination – both are mechanistically interpretable (dexamethasone suppresses non-specific catabolic clearance; lenalidomide-activated NK cells increase target-cell-binding-mediated elimination).

COMBO_POM_DEX (canonical for pomalidomide plus dexamethasone combination-therapy indicator)

• Description: 1 = the analyzed therapeutic mAb (or other agent under PK study) is co-administered with the pomalidomide + low-dose-dexamethasone (Pd) backbone, 0 = monotherapy or any non-Pd regimen. Pomalidomide is an immunomodulatory imide (IMiD) like lenalidomide; Pd is a standard relapsed/refractory multiple-myeloma backbone after lenalidomide and a proteasome inhibitor.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (monotherapy or any non-Pd regimen). Mutually exclusive with COMBO_LEN_DEX: a subject on Ld has COMBO_LEN_DEX = 1, COMBO_POM_DEX = 0, on Pd 0, 1, on monotherapy 0, 0.
• Source aliases:
  • POMDEX (1 = the Pd study CA204-125 / ELOQUENT-3, derived as STUDY == 204125) – used in Ide_2022_elotuzumab.R.
• Example models: Ide_2022_elotuzumab.R (multiplicative factors relative to the Ld reference: CLPomDex = 0.811 on nonspecific CL and KINTPomDex = 0.487 on the second-order peripheral target-mediated elimination rate).
• Notes: The per-backbone parallel of COMBO_LEN_DEX anticipated in that entry’s Notes; named with the same <IMiD>_DEX form rather than the COMBO_PD shorthand, for the same clarity reason COMBO_LD was renamed. Distinct from COMBO_BELAMAF, which pools Bor-Dex, Len-Dex and Pom-Dex into one belantamab indicator. Watch the source encoding: Ide 2022’s control stream keeps its LENDEX flag equal to 1 for the Pd study too (it is really an any-IMiD-backbone flag, STUDY != 204011), so there LENDEX = COMBO_LEN_DEX + COMBO_POM_DEX and POMDEX = COMBO_POM_DEX.

COMBO_BELAMAF (canonical for any-combination belantamab mafodotin therapy indicator)

• Description: 1 = belantamab mafodotin administered as part of a combination regimen (with bortezomib + dexamethasone, lenalidomide + dexamethasone, or pomalidomide + dexamethasone) in the relapsed/refractory multiple myeloma setting; 0 = belantamab mafodotin monotherapy.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (belantamab mafodotin monotherapy).
• Source aliases:
  • COMBO (when the source dataset uses a generic combination flag for belantamab mafodotin pooled regimens) – used in Papathanasiou_2025_belantamab.R.
• Example models: Papathanasiou_2025_belantamab.R (multiplicative factor theta = 1.44 on the Imax parameter of the time-varying CL function – combination therapy increases the steady-state CL reduction from 33.2 % to 44.0 %).
• Notes: Pools the three combination backbones tested in DREAMM-6 / DREAMM-7 / DREAMM-8 (Bor-Dex, Len-Dex, Pom-Dex) into a single binary because Papathanasiou 2025 reports no meaningful per-backbone difference in cycle-1 ADC exposure. If a future paper tests per-backbone combination effects, register dedicated indicators (COMBO_BELAMAF_BORDEX, etc.) rather than overloading this aggregate.

KG (canonical for subject-specific tumour-growth rate constant from a prior IPP fit)

• Description: Empirical-Bayes posterior estimate of the subject-specific tumour-size first-order growth rate constant carried over from an upstream tumour-size-dynamics population model. Supplied per subject in the dataset and used directly inside a downstream model (e.g., overall-survival hazard) that integrates the tumour-size ODE inline conditional on each subject’s growth/death rates.
• Units: internal scaled rate; the Zecchin 2016 OS model carries the source convention KG / 1000 * tumorSize in the SLD ODE so the column units are (1/day) * 1000 as published in the source NONMEM run. Document per-model via covariateData[[KG]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – used directly inside the SLD ODE (see Zecchin_2016_survival.R).
• Source aliases:
  • KG (NONMEM $INPUT column in DDMODEL00000218; identical column shipped in the bundle’s Simulated_OS.csv).
• Example models: Zecchin_2016_survival.R (Zecchin 2016 OS model, DDMODEL00000218).
• Notes: Specific scope because the column is the empirical-Bayes output of a particular upstream fit (modellib('Zecchin_2016_tumorovarian') / DDMODEL00000217). When this OS model is used standalone, the user must supply KG per subject – typically by first fitting the SLD model and extracting the per-subject empirical-Bayes posterior. The internal /1000 scaling is preserved verbatim from the source $DES block to maintain numerical equivalence with the published estimates.

KD0 (canonical for subject-specific carboplatin-related tumour-death rate constant from a prior IPP fit)

• Description: Empirical-Bayes posterior estimate of the subject-specific carboplatin-driven tumour-size death rate constant carried over from an upstream tumour-size-dynamics population model. Pairs with the time-varying AUC_CARBO covariate inside the SLD ODE term KD0 * AUC_CARBO * tumorSize.
• Units: internal scaled rate; the Zecchin 2016 OS model carries the source convention KD0 / 1000 * AUC_CARBO * tumorSize so the column units are (1/day per AUC unit) * 1000 as published in the source NONMEM run. Document per-model via covariateData[[KD0]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – used directly inside the SLD ODE (see Zecchin_2016_survival.R).
• Source aliases:
  • KD0 (NONMEM $INPUT column in DDMODEL00000218).
• Example models: Zecchin_2016_survival.R (Zecchin 2016 OS model, DDMODEL00000218).
• Notes: Specific scope because the column is tied to a specific drug (carboplatin) and a specific upstream IPP fit. The internal /1000 scaling is preserved verbatim from the source $DES block.

KD1 (canonical for subject-specific gemcitabine-related tumour-death rate constant from a prior IPP fit)

• Description: Empirical-Bayes posterior estimate of the subject-specific gemcitabine-driven tumour-size death rate constant carried over from an upstream tumour-size-dynamics population model. Pairs with the time-varying AUC_GEM covariate inside the SLD ODE term KD1 * AUC_GEM * tumorSize.
• Units: internal scaled rate; the Zecchin 2016 OS model carries the source convention KD1 / 100 * AUC_GEM * tumorSize so the column units are (1/day per AUC unit) * 100 as published in the source NONMEM run. Document per-model via covariateData[[KD1]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – used directly inside the SLD ODE (see Zecchin_2016_survival.R).
• Source aliases:
  • KD1 (NONMEM $INPUT column in DDMODEL00000218).
• Example models: Zecchin_2016_survival.R (Zecchin 2016 OS model, DDMODEL00000218).
• Notes: Specific scope because the column is tied to a specific drug (gemcitabine) and a specific upstream IPP fit. The internal /100 scaling is preserved verbatim from the source $DES block.

IBASE (canonical for subject-specific baseline tumour-size estimate from a prior IPP fit)

• Description: Empirical-Bayes posterior estimate of the subject-specific baseline sum-of-longest-diameters (SLD) tumour size carried over from an upstream tumour-size-dynamics population model. Used both to set the SLD ODE initial state (tumorSize(0) <- IBASE * 1000 in the Zecchin 2016 model: source convention multiplies by 1000 to convert the internal value to mm) and to scale the time-varying tumour-size ratio (mmbas <- IBASE * 1000) inside the OS hazard.
• Units: internal scaled length; the Zecchin 2016 OS model carries the source convention IBASE * 1000 = mm so the column itself is in metres (1 m = 1000 mm). Document per-model via covariateData[[IBASE]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – used as the SLD ODE initial condition and as the denominator in (tumorSize - mmbas) / mmbas.
• Source aliases:
  • IBASE (NONMEM $INPUT column in DDMODEL00000218; identical column shipped in the bundle’s Simulated_OS.csv with values typically in the 0.04-0.50 m range).
• Example models: Zecchin_2016_survival.R (Zecchin 2016 OS model, DDMODEL00000218).
• Notes: Distinct from the canonical TUM_SLD column. TUM_SLD carries the measured baseline tumour size in mm (used to compute the time-fixed NSLD0 = TUM_SLD / 70 covariate term in the Zecchin 2016 OS hazard), whereas IBASE carries the empirical-Bayes fitted baseline from the upstream SLD model (used to initialise the integrated SLD trajectory and to define the time-varying TSR(t) reference). The two are correlated but not equal because the upstream IPP fit smooths measurement noise away from the observed SLD0. Specific scope because the column is the empirical-Bayes output of a specific upstream model fit and the internal *1000 unit-conversion is tied to the source NONMEM coding convention.

NEW_LESION (canonical for time-varying new-lesion appearance indicator)

• Description: Time-varying binary indicator of whether a new (non-target) RECIST lesion has appeared since enrolment. 1 = new lesion present at the current observation time; 0 = no new lesion as of the current observation time. Once NEW_LESION flips to 1 it stays 1 for subsequent observation times in that subject (a step-function flag, not a transient pulse).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no new lesion appeared as of the current time).
• Source aliases:
  • NWLS – prior canonical name (pre-2026-06-19 readability standardization) and NONMEM $INPUT column in DDMODEL00000218; the bundle’s Simulated_OS.csv re-labels the same column NWLSCOV. Downstream consumers should map NWLSCOV -> NEW_LESION.
• Example models: Zecchin_2016_survival.R (Zecchin 2016 OS model, DDMODEL00000218; multiplicative effect on the Weibull hazard via exp(e_nwls_haz * NEW_LESION) with e_nwls_haz = 1.23 per Output_real_OS.lst FINAL TH5 / Table 2 of Zecchin 2016).
• Notes: Specific scope because the column encodes a per-paper RECIST-style binary that is supplied by the dataset; it is not a generic “any new lesion” indicator the user can populate from routine clinical data without an explicit lesion-appearance imaging schedule. The Zecchin 2016 OS model uses NEW_LESION directly (no time-gating), which is faithful to the simulated dataset shipped in the DDMORE bundle. The source Output_real_OS.lst (the listing on the original real dataset) gates the indicator with an additional TNWLS (lesion-appearance-time) column not shipped in the bundle’s simulated dataset; the two encodings are functionally equivalent when the dataset’s NEW_LESION column is constructed as a 0/1 step that flips at the lesion-appearance time. The bundle’s simulated dataset uses the simpler step-function form, and that is the form the nlmixr2lib model expects.

MM_NIGG (canonical for non-IgG multiple myeloma immunoglobulin-type indicator)

• Description: 1 = patient with non-IgG-secreting multiple myeloma (e.g., IgA, IgD, IgE, IgM, light-chain-only / Bence Jones, or non-secretory MM), 0 = patient with IgG-secreting multiple myeloma.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (IgG MM).
• Source aliases:
  • Ig_type – used in Fau_2020_isatuximab.R. Values 0 / 1 with the same orientation as the canonical (1 = non-IgG MM).
• Example models: Fau_2020_isatuximab.R (exponential effect on the steady-state linear CL CLinf with coefficient -0.751, and on the time-varying-CL half-time KCL with coefficient -0.931), Xu_2020_daratumumab.R (additive shift (1 + 0.806 * (1 - MM_NIGG)) on linear CL – Xu 2020 parameterises with non-IgG MM as reference, so an IgG MM patient receives an 80.6% higher linear CL than a non-IgG MM patient; canonical column semantics 1 = non-IgG / 0 = IgG are preserved), Miao_2023_teclistamab.R (multiplicative factors 0.689^MM_NIGG on the time-independent clearance component cl_exp_inf and 0.295^MM_NIGG on the time-dependent component cl_exp_component; the only covariate Miao 2023 retained on both clearance arms, anchored to IgG MM as in Fau 2020), Luo_2021_daratumumab.R (additive shift (1 + 1.26 * (1 - MM_NIGG)) on linear CL of the joint SC + IV daratumumab model – non-IgG MM reference as in Xu 2020, so an IgG MM patient has 2.26-fold the linear CL).
• Notes: Within-disease (multiple-myeloma) immunoglobulin-subtype stratifier. The mechanistic rationale (Fau 2020) is that endogenous IgG monoclonal protein in IgG-MM patients competes with the therapeutic IgG mAb for FcRn-mediated salvage, raising the therapeutic mAb’s catabolic clearance; non-IgG-MM patients lack that competition and exhibit lower therapeutic-mAb clearance. Distinct from the disease-state indicators (DIS_SMM = smoldering MM); applies only after a multiple-myeloma diagnosis is established. Scope: specific because the comparison is a within-MM stratifier rather than a cross-population indicator. Reference category at the model level (which value of MM_NIGG corresponds to TVCL = base) varies between papers: Fau 2020 anchors to 0 (IgG MM) and Xu 2020 anchors to 1 (non-IgG MM); the canonical column orientation (1 = non-IgG) is fixed across papers and the per-model covariateData[[MM_NIGG]]$reference_category field records which anchor each model uses.

ISS_II (canonical for International Staging System stage II indicator in multiple myeloma)

• Description: 1 = baseline International Staging System (ISS) stage II multiple myeloma, 0 = otherwise (ISS stage I or stage III). Time-fixed per subject. The ISS is the standard three-level prognostic staging system for multiple myeloma, derived from serum beta-2-microglobulin (B2M) and serum albumin (ALB): stage I = B2M < 3.5 mg/L and ALB >= 3.5 g/dL; stage III = B2M >= 5.5 mg/L; stage II = neither stage I nor stage III. Paired with ISS_III to reproduce the three-level ordinal effect with two mutually-exclusive level indicators (stage I is the reference level, both indicators = 0).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0; family reference level is ISS stage I (ISS_II = 0 and ISS_III = 0).
• Source aliases:
  • ISS (ordinal categorical column with levels I / II / III, or integer 1 / 2 / 3) – decompose: ISS_II = as.integer(ISS == "II").
• Example models: Miao_2023_teclistamab.R (multiplicative factor 1.31 on the time-independent clearance component cl_exp_inf, encoded as 1.31^ISS_II per Miao 2023 Table 2 footnote a).
• Notes: Level indicators, NOT cumulative thresholds – unlike the ECOG_GE1 / ECOG_GE2 pair, where ECOG_GE2 = 1 implies ECOG_GE1 = 1, ISS_II and ISS_III are mutually exclusive and an ISS stage III patient has ISS_II = 0. This matches the printed form of the Miao 2023 covariate equation, in which the stage II and stage III factors are applied as 1.31^(ISS = II) x 1.67^(ISS = III) so that exactly one of the two factors is active for any patient. Papers that report ISS as a single ordinal covariate with one estimated slope should still supply both indicators. B2M and ALB are the constituent laboratory values, but ISS stage is a derived clinical stage and is not an alias of either. Distinct from the revised ISS (R-ISS), which adds cytogenetics and LDH; a paper using R-ISS needs its own columns. Scope: specific because the staging system is defined only within multiple myeloma.

ISS_III (canonical for International Staging System stage III indicator in multiple myeloma)

• Description: 1 = baseline International Staging System (ISS) stage III multiple myeloma (serum beta-2-microglobulin >= 5.5 mg/L), 0 = otherwise (ISS stage I or stage II). Time-fixed per subject. Paired with ISS_II; see that entry for the full staging definition and the family encoding.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0; family reference level is ISS stage I (ISS_II = 0 and ISS_III = 0).
• Source aliases:
  • ISS (ordinal categorical column with levels I / II / III, or integer 1 / 2 / 3) – decompose: ISS_III = as.integer(ISS == "III").
• Example models: Miao_2023_teclistamab.R (multiplicative factor 1.67 on the time-independent clearance component cl_exp_inf, encoded as 1.67^ISS_III per Miao 2023 Table 2 footnote a).
• Notes: Mutually exclusive with ISS_II – see the ISS_II notes for why this family uses level indicators rather than the cumulative >= encoding used by ECOG_GE1 / ECOG_GE2. Worse ISS stage is associated with higher teclistamab clearance and therefore lower exposure (Miao 2023), the opposite direction from the usual disease-burden intuition for a T-cell engager; Miao 2023 attributes the association to disease status rather than target binding.

TUM_EGFR_MUT (canonical for tumour EGFR mutation indicator)

• Description: Binary indicator of tumour-cell EGFR (epidermal growth factor receptor) mutational status as assessed in the source paper. 1 = EGFR-mutant tumour (a sensitising kinase-domain mutation such as exon 19 deletion or L858R, detected by sequencing or an approved companion diagnostic); 0 = EGFR wild-type tumour. Time-fixed per subject – a somatic tumour-genotype call made at diagnosis, not a germline genotype. In non-small-cell lung cancer, EGFR-mutant tumours are a distinct clinical entity treated first-line with tyrosine-kinase inhibitors, and they respond markedly less well to PD-1 / PD-L1 checkpoint blockade than EGFR wild-type tumours.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (EGFR wild type). Wild type is the most frequent category in every reported NSCLC cohort, so it is also the natural modelling reference.
• Source aliases:
  • EGFR status / EGFR mutation status – Chatterjee 2016 main-article Methods and supplementary Table S3 row label. Values reported as “wild type” / “mutant” / “unknown”; map “mutant” to 1, “wild type” to 0, and carry the unknown stratum in the companion column TUM_EGFR_MUT_MISSING.
• Example models: Chatterjee_2016_pembrolizumab.R (additive effect -1.81 on the logit of the treatment-sensitive tumour fraction f in a bi-exponential tumor-size model of pembrolizumab in advanced NSCLC, i.e. an EGFR-mutant tumour has f = 0.181 against f = 0.574 for wild type – a 3.2-fold lower responding fraction, matching the paper’s post-hoc EBE ratio).
• Notes: Member of the TUM_<MARKER> somatic-tumour-genotype family (TUM_TP53_MUT, TUM_1P19Q_CODEL, TUM_IGHV_MUT), NOT of the SNP_<GENE>_RS<rsid> germline-pharmacogenomics family: this column encodes a tumour-cell mutation that changes tumour biology and treatment sensitivity, not a host genotype affecting drug disposition. Distinct from the covariate CRCL’s eGFR alias – “EGFR” the receptor gene and “eGFR” the estimated glomerular filtration rate are unrelated concepts that collide on the all-caps spelling, so always spell the receptor column with the TUM_ prefix. Specific scope because the marker is meaningful only for tumour types in which EGFR mutation is an established driver (predominantly NSCLC). Document the detection assay per-model in covariateData[[TUM_EGFR_MUT]]$notes.

TUM_EGFR_MUT_MISSING (canonical for tumour-EGFR-status-missing indicator)

• Description: Binary indicator for a subject whose tumour EGFR mutational status was not determined. 1 = EGFR status unknown (no genotyping performed, or an inconclusive result); 0 = EGFR status known, in which case the call is carried in TUM_EGFR_MUT. Companion to TUM_EGFR_MUT, for source papers that retain “unknown EGFR” as a distinct modelled category with its own estimated coefficient rather than imputing those subjects into the wild-type reference.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (EGFR status known). When TUM_EGFR_MUT_MISSING == 1, set TUM_EGFR_MUT = 0 and gate the mutant term inside model so the unknown subject receives only its own coefficient.
• Source aliases:
  • EGFR "Unknown" – Chatterjee 2016 supplementary Table S3 category label and the EGFR_2 coefficient row of supplementary Table S6.
• Example models: Chatterjee_2016_pembrolizumab.R (additive effect +1.66 on the logit of the treatment-sensitive tumour fraction f, relative to the EGFR wild-type reference; 26 of 505 subjects, supplementary Table S3).
• Notes: Follows the missing-data-indicator pattern established by ADA_MISSING, CYP2C9_MISSING and HEPIMP_MOD_OR_MISSING; sibling of PDL1_TUM_MISSING. Use only when the source paper explicitly fits a separate coefficient for the EGFR-unknown subgroup. As with every missingness coefficient, the estimate describes who was left ungenotyped rather than any tumour biology – Chatterjee 2016’s +1.66 implies a HIGHER responding tumour fraction in the unknown stratum than in wild type, which is a selection artefact of which patients were sent for genotyping and must not be extrapolated.

TUM_ALK_MUT (canonical for tumour ALK rearrangement indicator)

• Description: Binary indicator of tumour-cell anaplastic lymphoma kinase (ALK) rearrangement status as assessed in the source paper. 1 = ALK-positive tumour (an ALK gene fusion, most commonly EML4-ALK, detected by fluorescence in-situ hybridisation, immunohistochemistry or next-generation sequencing); 0 = not ALK-positive, which pools ALK wild-type tumours, tumours positive for a different driver such as ROS1, tumours of unknown status, and healthy participants who have no tumour at all. Time-fixed per subject – a somatic tumour-genotype call made at diagnosis, not a germline genotype. ALK rearrangements occur in roughly 2%-8% of non-small-cell lung cancers and define the population treated with ALK tyrosine-kinase inhibitors.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: Set by the source model rather than by the register. In Liu_2024_saf189s.R the reference cell of the covariate model is the ALK-POSITIVE, ALK-inhibitor-naive patient, so 1 is the reference there; in a model contrasting an ALK-positive subgroup against a wild-type majority, 0 is the natural reference. Always state which applies in covariateData[[TUM_ALK_MUT]]$reference_category.
• Source aliases:
  • ALK Mutation (values Positive / Negative or Other) – Liu 2024 Supplementary Table 1 row label. Despite the label, the lesion is a rearrangement rather than a point mutation; the _MUT suffix follows the TUM_<MARKER>_MUT family convention rather than describing the lesion class.
  • ALK+ / ALK-positive – narrative labels.
  • ALKPOT = 5 – Liu 2024’s “others (ROS1+ patients or unknown)” level, which is (1 - TUM_ALK_MUT) * (1 - DIS_HEALTHY): a not-ALK-positive SUBJECT WHO IS A PATIENT. That level is a MIXED bucket pooling genuinely ROS1-positive patients with patients of unknown status, so it cannot be decomposed into a clean ROS1 indicator and no TUM_ROS1_MUT column is implied by it.
• Example models: Liu_2024_saf189s.R (gates the multiplicative power-form effect on apparent central volume 0.917^((1 - TUM_ALK_MUT) * (1 - DIS_HEALTHY)), i.e. V/F 8.3% lower in ROS1-positive-or-unknown patients than in the ALK-positive ALK-inhibitor-naive reference; the DIS_HEALTHY gate is required because the 24 healthy volunteers are also not ALK-positive and carry their own separate effect).
• Notes: Member of the TUM_<MARKER> somatic-tumour-genotype family (TUM_EGFR_MUT, TUM_TP53_MUT, TUM_1P19Q_CODEL, TUM_IGHV_MUT), NOT of the SNP_<GENE>_RS<rsid> germline-pharmacogenomics family. Because the 0 cell pools several biologically distinct groups, a model that needs to separate them must combine this column with the companion indicators that identify those groups (DIS_HEALTHY for participants without a tumour, and a ROS1 indicator if a source paper ever estimates one) rather than reinterpreting the 0 cell. Distinct from PRIOR_ALKI, which is a treatment-history covariate applying WITHIN the ALK-positive stratum: a subject may be TUM_ALK_MUT = 1 with PRIOR_ALKI either 0 or 1.

TUM_TP53_MUT (canonical for tumour TP53 / p53 mutation indicator)

• Description: Binary indicator of tumour-cell TP53 mutational status as assessed in the source paper. 1 = TP53 mutant tumour (typically a missense mutation detected by sequencing, or p53 protein overexpression by immunohistochemistry used as a surrogate for TP53 missense mutation per Gillet et al. J Neurooncol 2014); 0 = TP53 wild-type tumour. Time-fixed per subject (a somatic tumour-genotype call made at diagnosis, not a germline genotype).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (TP53 wild-type).
• Source aliases:
  • “p53 mutation” / “p53 mutated” (Mazzocco 2015, paper text; p53 overexpression by IHC used as a surrogate for TP53 missense mutation).
• Example models: Mazzocco_2015_temozolomide.R (exponential effect on the temozolomide tumour-cell-death rate constant gamma: gamma = gamma0 * exp(beta_p53 * TUM_TP53_MUT) with beta_p53 = log(0.143 / 0.254) = -0.574; TP53-mutant LGG tumours are 44% less sensitive to TMZ than TP53-wild-type tumours).
• Notes: Specific scope because the column encodes a somatic tumour-genotype call (mechanism = altered DNA-damage response in tumour cells) rather than a germline pharmacogenomic variant. Distinct from the SNP_<GENE>_<RSID> family, which encodes inherited host germline genotypes affecting drug PK; TUM_TP53_MUT encodes a tumour-cell mutation and only makes mechanistic sense for drugs whose effect depends on a functional p53 pathway in the tumour. The Mazzocco 2015 cohort assayed p53 status by IHC overexpression; future extractions that use direct TP53 sequencing should still record their values under this canonical and document the assay method in covariateData[[TUM_TP53_MUT]]$notes.

TUM_BRAF_MUT (canonical for tumour BRAF mutation indicator)

• Description: Binary indicator of tumour-cell BRAF (B-type Raf proto-oncogene, serine/threonine kinase) mutational status as assessed in the source paper. 1 = BRAF-mutant tumour (in melanoma cohorts this is overwhelmingly the V600E / V600K activating substitution; some source datasets pool a BRAF translocation into the same level); 0 = BRAF wild-type tumour. Time-fixed per subject. Member of the TUM_<GENE>_MUT somatic-genotype family alongside TUM_EGFR_MUT, TUM_ALK_MUT and TUM_TP53_MUT.
• Units: (binary)
• Type: binary
• Scope: general (oncology)
• Reference category: 0 (BRAF wild-type, the majority class in unselected advanced-melanoma cohorts).
• Source aliases:
  • BRAFMN – Chatterjee 2017 NONMEM input column, coded 1 = mutation/translocation, 2 = wild type, -99 = unknown; the indicator is TUM_BRAF_MUT = as.integer(BRAFMN == 1).
  • “BRAF mutation status” / “B-type Raf (BRAF) mutation status” – Chatterjee 2017 main-article Methods and supplementary Table 5A row label.
• Example models: Chatterjee_2017_pembrolizumab_consolidated.R (fractional-deviation effect on the tumour growth rate of a bi-exponential tumour-size model: kgrowth * (1 + 0.702 * TUM_BRAF_MUT), i.e. the resistant fraction of a BRAF-mutant melanoma grows about 70% faster than wild type; supplementary Table 4C BRAF_1 on kgrowth, |RSE| 37.9%).
• Notes: A source paper that carries a separate “unknown” BRAF level must be read carefully before assuming it needs a companion missingness column. Chatterjee 2017 maps unknown status to the wild-type reference multiplier inside $PRED (IF(BRAFMN.EQ.-99) KGROWBRAFMN = 1), so the unknown stratum is pooled into the reference and TUM_BRAF_MUT = 0 reproduces the source exactly; only 11 of 1366 subjects (0.81%) were affected. Register a TUM_BRAF_MUT_MISSING companion (following TUM_EGFR_MUT_MISSING) only for a source that estimates a distinct coefficient for the unknown stratum. General scope because BRAF genotype is a standard stratification factor in melanoma, thyroid and colorectal oncology analyses. Do not conflate this tumour-genotype column with a BRAF-inhibitor co-medication indicator (CONMED_<INN>), which records what the patient received rather than what the tumour carries.

TUM_1P19Q_CODEL (canonical for tumour 1p/19q chromosomal codeletion indicator)

• Description: Binary indicator of tumour-cell 1p/19q chromosomal codeletion status. 1 = tumour carries the combined loss of the short arm of chromosome 1 (1p) and the long arm of chromosome 19 (19q); 0 = non-codeleted tumour (intact 1p and/or 19q). Time-fixed per subject (a somatic tumour-genotype call made at diagnosis, typically by fluorescence in-situ hybridisation, comparative genomic hybridisation, or loss-of-heterozygosity assay).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-codeleted; intact 1p and/or 19q).
• Source aliases:
  • “1p/19q codeletion” / “1p/19q codeleted” (Mazzocco 2015, paper text).
• Example models: Mazzocco_2015_temozolomide.R (exponential effect on the damaged-quiescent-to-proliferative repair rate constant kQpP: kQpP = kQpP0 * exp(beta_1p19q * TUM_1P19Q_CODEL) with beta_1p19q = log(0.00807 / 0.00947) = -0.160; 1p/19q-codeleted LGG tumours have 15% lower kQpP than non-codeleted tumours, consistent with longer reported duration of response in codeleted patients).
• Notes: Specific scope because the column encodes a brain-tumour-specific somatic chromosomal alteration with mechanistic relevance to DNA-repair capacity in glioma cells; it is not a generic “any tumour-chromosomal-alteration” indicator. The 1p/19q codeletion is a defining molecular feature of oligodendrogliomas (per the 2016 WHO classification of CNS tumours) and is mutually exclusive with TP53 missense mutation in the Mazzocco 2015 cohort (Ricard 2007, ref 12 of Mazzocco 2015).

TUM_IGHV_MUT (canonical for immunoglobulin heavy-chain variable region (IGHV) mutational status)

• Description: Binary indicator of the somatic hypermutation status of the immunoglobulin heavy-chain variable region (IGHV) gene rearrangement in a mature B-cell clone, most commonly assayed in chronic lymphocytic leukaemia (CLL). 1 = IGHV-mutated (the clone carries >= 2% divergence from the closest germline IGHV sequence); 0 = IGHV-unmutated (< 2% divergence). Time-fixed per subject – IGHV status is fixed at the point of clonal transformation and does not change with therapy. Note the prognostic polarity: unlike most mutation indicators, the MUTATED state is the FAVOURABLE one in CLL (IGHV-mutated disease has an indolent course, IGHV-unmutated disease is aggressive), so a positive coefficient on this column usually corresponds to a lower disease burden.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (IGHV-unmutated).
• Source aliases:
  • IMPIGVHMS (Ibrahim 2023 nlmixr control stream, Supporting Information S4 run100: igvh <- IMPIGVHMS) – the IMP prefix records that missing categorical covariates (n <= 6) were mode-imputed per Ibrahim 2023 Appendix S1 section 2. Note the source column transposes the VH / HV letters relative to the standard IGHV abbreviation.
  • “IGHV mutational status” / “IGVH-mutated” / “IGVH-unmutated” (Ibrahim 2023 paper text).
• Example models: Ibrahim_2023_ibrutinib_leukocyte_spd.R (selects the baseline-SPD typical value: spdbase = spdbase_unm * (1 - TUM_IGHV_MUT) + spdbase_m * TUM_IGHV_MUT, giving 48.9 cm^2 for unmutated vs 19.5 cm^2 for mutated – a 2.5-fold higher baseline lymph-node burden in IGHV-unmutated patients).
• Notes: Specific scope; a somatic B-cell-clone genotype call rather than a host germline variant, so it belongs to the TUM_<MARKER> family (TUM_TP53_MUT, TUM_1P19Q_CODEL) and NOT to the SNP_<GENE>_RS<rsid> germline-pharmacogenomics family. Encoded as an effect on a disease-burden parameter rather than on a PK parameter, which is the typical use: IGHV status is a disease-biology marker, not a metabolism marker. When a source paper reports the covariate with the opposite polarity (an “IGHV-unmutated = 1” column, which some CLL papers use because unmutated is the high-risk group), convert on ingestion with TUM_IGHV_MUT = 1 - <source column> and record the transformation in covariateData[[TUM_IGHV_MUT]]$notes.

TUM_17P_DEL (canonical for deletion(17p) chromosomal abnormality indicator)

• Description: Binary indicator of the deletion(17p) chromosomal abnormality in a tumour / leukaemic clone. 1 = del(17p) present, 0 = absent. The deleted region carries the TP53 locus, so del(17p) is functionally a TP53-loss marker and defines the highest-risk cytogenetic subgroup in chronic lymphocytic leukaemia (10-year overall survival 29%; Ibrahim 2023 Discussion). Time-fixed per subject; a baseline cytogenetic call, typically by fluorescence in-situ hybridisation (FISH).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no del(17p)).
• Source aliases:
  • del(17p) / “deletion (17p)” (Ibrahim 2023 paper text and Table 2 footnote f).
• Example models: Ibrahim_2023_ibrutinib_competing_risk.R (multiplicative effect on the alive-to-death transition rate of the competing-risk multistate model: exp(1.42 * TUM_17P_DEL), a hazard ratio of 4.14).
• Notes: Specific scope; a somatic chromosomal deletion, so it belongs to the TUM_<MARKER> family alongside TUM_1P19Q_CODEL (the other registered chromosomal-alteration indicator) and NOT to the germline SNP_<GENE>_RS<rsid> family. Distinct from TUM_TP53_MUT: TUM_TP53_MUT records a point mutation / IHC-overexpression call at the TP53 locus, whereas TUM_17P_DEL records loss of the whole chromosome arm carrying it. The two are frequently but not always concordant (del(17p) and TP53 mutation can occur independently, and papers that assay both usually report them as separate columns), so do not alias one onto the other – register a paper’s del(17p) column here and its TP53 sequencing / IHC column under TUM_TP53_MUT. Ibrahim 2023 also screened deletion(11q) and deletion(13q) as candidate covariates but retained neither in any final model, so no canonicals are registered for them.

TUM_CYTOGENETIC_HIGH_RISK (canonical for multiple-myeloma high-risk cytogenetics indicator)

• Description: Binary indicator of high-risk cytogenetics in the plasma-cell clone of multiple myeloma at baseline. 1 = high risk, 0 = standard risk or cytogenetic risk not available. The TOURMALINE-MM1 trial publication (Moreau 2016, N Engl J Med 374:1621) defined high risk as del(17p), t(4;14) or t(14;16); Srimani 2022 itself prints only the high / standard / not-available categories (Table 1).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (standard cytogenetic risk, or risk unavailable).
• Source aliases:
  • BCYABCAT – Srimani 2022 Table 2 covariate code (“kR (BCYABCAT)”).
  • high cytogenetic risk / standard cytogenetic risk (or risk unavailable) – Srimani 2022 Equation 6 case labels.
• Example models: Srimani_2022_ixazomib_mprotein.R (linear multiplicative effect on the M-protein response rate constant, kR * (1 + 0.590 * TUM_CYTOGENETIC_HIGH_RISK); 20.8% of the efficacy dataset was high risk).
• Notes: Member of the TUM_<MARKER> family alongside TUM_17P_DEL and TUM_1P19Q_CODEL. It is a COMPOSITE of several abnormalities, so a source that reports del(17p) on its own belongs under TUM_17P_DEL, not here; record in the per-model notes which abnormalities the source’s high-risk definition includes, because the definition varies between trials (some add 1q21 gain or t(14;20)). Unavailable risk is pooled with the reference in Srimani 2022; a source that fits a separate “not available” level needs its own indicator. Lin_2020_glasdegib_exposure.R / _treatment.R carry an unregistered CYTO_POOR for the acute-myeloid-leukaemia poor-risk cytogenetic category, which uses a different (ELN / MRC) classification and is not an alias of this entry.

TUM_JAK2_V617F_VAF (canonical for JAK2 V617F somatic variant allele frequency (allele burden))

• Description: Variant allele frequency of the somatic JAK2 V617F (c.1849G>T, p.Val617Phe) gain-of-function point mutation in peripheral blood or bone marrow, expressed as a PER CENT in 0-100. Operationally the fraction of sequenced alleles at the locus carrying the variant, and biologically a proxy for the proportion of haematopoiesis carried by the driver-mutant clone. The defining molecular lesion of the JAK2-mutant myeloproliferative neoplasms – polycythaemia vera (where over 95% of patients carry it), essential thrombocythaemia and primary myelofibrosis. Unlike a germline genotype this is a clone-size measurement: it varies continuously, it changes under cytoreductive or interferon therapy, and its reduction is itself a recognised efficacy endpoint (molecular response). Carry the value BASELINE-fixed per subject when the source uses it as a predictor, and as a time-varying column (or a modelled state) only when the source explicitly does so.
• Units: percent (0-100). Document per-model via covariateData[[TUM_JAK2_V617F_VAF]]$units, and state the scale explicitly: a source reporting the 0-1 fraction needs either the column multiplied by 100 or the coefficient divided by it.
• Type: continuous
• Scope: general
• Reference category: n/a – continuous. Typically entered UNCENTRED, in which case the model intercept is the value at a burden of zero, an extrapolated anchor rather than an observed state (a JAK2-unmutated patient does not belong to the fitted population). Where a source centres on a cohort median or quartile, record the centring value in the per-model notes.
• Source aliases:
  • JAK2 V617F (%) (Qin 2025 Table 1 row label).
  • “baseline JAK2 V617F allele burden” / “JAK2 V617F allele burden” (Qin 2025 Results 3.4 and Methods 2.2).
  • JAK2VAF, VAF, “mutant allele burden” – common NONMEM / case-report-form spellings in the myeloproliferative-neoplasm literature.
• Example models: Qin_2025_ropeginterferon_anemia.R (baseline allele burden per cent, entered linearly and uncentred on the logit of maintenance-phase anemia occurrence with coefficient 0.041351 per percentage point, an odds ratio of 1.0422 per point and 1.51 per 10 points; the only covariate retained by any of the three significant exposure-safety regressions in that paper), Qin_2025_ropeginterferon_jak2_week24.R, Qin_2025_ropeginterferon_jak2_week52.R.
• Notes: General scope – the same assay and the same quantity recur across every JAK2-mutant myeloproliferative-neoplasm programme, independent of drug. Belongs to the TUM_<MARKER> somatic-tumour-genetics family rather than the germline SNP_<GENE>_RS<rsid> family, and founds a TUM_<GENE>_<VARIANT>_VAF sub-family for continuous somatic clone fractions. Three boundaries matter. (1) Not a SNP_ entry. JAK2 V617F does have an rsid (rs77375493), so SNP_JAK2_RS77375493 would be mechanically well-formed under the auto-approve family, but every registered SNP_* column carries a germline GENOTYPE – an active-allele count (SNP_ORM1_RS17650), a T-allele count (SNP_ABCB1_RS1045642_T_COUNT) or a carrier indicator (SNP_ICAM1_RS1799969). A value of 61.2 under a SNP_ name would be silently ambiguous between an allele count and a per cent, in exactly the dimension that matters. (2) Not an alias of the binary TUM_<GENE>_MUT indicators (TUM_EGFR_MUT, TUM_TP53_MUT, TUM_IGHV_MUT, TUM_ALK_MUT), all of which are 0/1 presence calls; this is their continuous sibling, and a paper reporting only mutated-versus-not should use a TUM_JAK2_MUT indicator instead. (3) Not CTDNA, which is baseline circulating tumour DNA burden in mutant molecules per mL of plasma – a different analyte, a different assay and a different unit; note that a fuzzy canonical lookup of “baseline JAK2 V617F allele burden” returns CTDNA as a confident but wrong match. The variant token is kept in the name because JAK2 exon 12 mutations are a separate assayable abnormality in polycythaemia vera (Qin 2025 Discussion notes them explicitly), so a future exon-12 column can register as its own sibling rather than overloading this one. Naming ratified by the maintainers with this extraction.

WHO_PS (canonical for ordinal-numeric performance-status integer (WHO / ECOG))

• Description: World Health Organization (WHO) performance-status integer score, identical-by-construction with the Eastern Cooperative Oncology Group (ECOG) PS per the Oken 1982 cross-walk (Am J Clin Oncol 5:649-655). Scale: 0 = fully active / asymptomatic, 1 = ambulatory and able to do light work, 2 = ambulatory and capable of all self-care but unable to work, 3 = bedridden > 50% of waking hours, 4 = completely disabled / bedridden, 5 = dead. Used as a continuous-numeric covariate when the source paper enters the integer score directly into a linear / exponential / power covariate model (e.g. CL ~ (1 - theta_PS * WHO_PS) or CL ~ exp(theta_PS * WHO_PS)) rather than binarizing into ECOG_GE1 / ECOG_GE2. Time-fixed per subject in the typical baseline-PS application; document time-varying use in per-model notes.
• Units: (integer score; document the empirical range and reference value in covariateData[[WHO_PS]]$notes)
• Type: continuous (semantically ordinal but treated as continuous in the covariate model)
• Scope: general
• Reference category: n/a – used with linear or power scaling. Reference value observed: WHO_PS = 0 (Leger 2004 reference patient implicit; the typical CL formula (1 - 0.12 * WHO_PS) evaluates to 1 at WHO_PS = 0).
• Source aliases:
  • PS – WHO performance status integer; used in Leger_2004_topotecan.R and historically in Mould 2002 and Gallo 2000 topotecan popPK papers (cited by Leger 2004) using the same linear-ordinal form.
  • ECOG – equivalent ECOG performance status integer.
  • PERF – common NONMEM column header for the ordinal performance-status integer.
• Example models: Leger_2004_topotecan.R (linear-ordinal effect (1 - 0.12 * WHO_PS) on the renal-plus-non-renal CL formula (theta1 + theta2 * CrCl_Lh) * (1 - e_who_ps_cl * WHO_PS); PS distribution in cohort 0 / 1 / 2 / 3 = 79 / 98 / 11 / 2 patients).
• Notes: Distinct from ECOG_GE1 / ECOG_GE2 (binary indicators used when the source paper collapses the ordinal score). Use WHO_PS when the paper retains the integer score in a linear / exponential / power covariate form. If a future paper provides the ordinal integer column but binarizes the effect downstream, derive ECOG_GE1 = as.integer(WHO_PS >= 1) and ECOG_GE2 = as.integer(WHO_PS >= 2); do not duplicate the same patient-level integer under two columns. WHO and ECOG PS are operationally identical and cross-walked one-to-one per Oken 1982 (Am J Clin Oncol 5:649-655); the canonical name WHO_PS is chosen because the founding example (Leger 2004) names it as WHO performance status. Scope general because ordinal-linear PS is the dominant historical NONMEM covariate-model form for oncology popPK papers and is likely to recur in future extractions.

GEMOX (canonical for gemcitabine-then-oxaliplatin sequence indicator)

• Description: 1 = gemcitabine 30-min IV infusion administered first, immediately followed by oxaliplatin 120-min IV infusion on the same study day; 0 = otherwise (gemcitabine alone or oxaliplatin-then-gemcitabine). Time-fixed per cycle.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 = gemcitabine monotherapy when paired with OXGEM = 0; the combination-vs-monotherapy reference is encoded jointly with OXGEM.
• Source aliases:
  • GEMOX – used in Jiang_2008_gemcitabine.R.
• Example models: Jiang_2008_gemcitabine.R (multiplicative factor 0.65 on apparent dFdU central volume V_C,dFdU/F when GEMOX = 1; Jiang 2008 page 330 covariate equation: 0.65^GEMOX).
• Notes: Paired with OXGEM as the second arm of a 3-level oxaliplatin-coadministration-sequence categorical decomposition (gem-alone / gem-then-ox / ox-then-gem). The two indicators are mutually exclusive: GEMOX + OXGEM <= 1. Follows the NIVO_1Q3W / NIVO_3Q2W decomposition pattern. Scope: specific because the sequence-specific effect is tied to Jiang’s gemcitabine + oxaliplatin study design; promote to general if a second paper retains the same pair of indicators.

OXGEM (canonical for oxaliplatin-then-gemcitabine sequence indicator)

• Description: 1 = oxaliplatin 120-min IV infusion administered first, immediately followed by gemcitabine 30-min IV infusion on the same study day; 0 = otherwise (gemcitabine alone or gemcitabine-then-oxaliplatin). Time-fixed per cycle.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 = gemcitabine monotherapy when paired with GEMOX = 0.
• Source aliases:
  • OXGEM – used in Jiang_2008_gemcitabine.R.
• Example models: Jiang_2008_gemcitabine.R (multiplicative factor 0.54 on apparent dFdU central volume V_C,dFdU/F when OXGEM = 1; Jiang 2008 page 330 covariate equation: 0.54^OXGEM).
• Notes: Paired with GEMOX (see that entry for the joint encoding). The Jiang 2008 cohort distribution is gem-alone n=31, gem-then-ox n=38, ox-then-gem n=25; the larger V_C,dFdU/F reduction observed for ox-then-gem (factor 0.54) vs gem-then-ox (factor 0.65) is hypothesised by the authors to reflect order-dependent tissue-uptake effects of dFdU.

TUM_CELLS0 (canonical for subject-specific baseline tumour burden expressed as an absolute cell count)

• Description: Per-subject baseline number of malignant cells in the modelled compartment, in absolute cells (not mm, mm^3, or SLD). Used to initialise a mechanistic tumour-cell ODE state whose currency is cell count rather than a caliper- or imaging-derived size (tumor(0) <- TUM_CELLS0). In Minucci 2024 the value is the fitted initial malignant-B-cell burden obtained by trust-region optimisation against the individual CAR T-cell cellular-kinetic trajectory, because the IM19 phase I trial (Ying 2021) reported no direct longitudinal tumour-burden measurement.
• Units: cells (count)
• Type: continuous
• Scope: specific
• Reference category: n/a – used directly as the tumour-cell ODE initial condition.
• Source aliases:
  • N0_tumorCells – column in the Minucci 2024 Supplementary Material Table 1 (1).XLSX, sheet params-case-study-final; used in Minucci_2024_CART_qsp.R.
• Example models: Minucci_2024_CART_qsp.R (13 relapsed/refractory NHL patients; fitted values span 1.0e5 to 1.0e7 cells, with F0109 / F0110 / F0111 / F0123 pinned at the 1e5 lower optimisation bound and F0104 / F0106 / F0107 / F0118 / F0119 at the 1e7 upper bound).
• Notes: Deliberately distinct from the imaging/caliper tumour-burden family. TUMSZ and TUM_SLD carry measured sizes in mm; TUM_VOL carries a caliper volume in mm^3; IBASE carries an empirical-Bayes fitted baseline SLD from an upstream tumour-size model. TUM_CELLS0 is the cell-count analogue, appropriate for mechanistic / QSP models that track individual malignant cells and their receptor burden, where no length or volume conversion is defined by the source. Same fitted-input class as KG / KD0 / KD1 / IBASE (subject-specific values carried over from a prior fit rather than measured at baseline), hence scope: specific – a downstream user cannot populate this column from routine clinical data without re-running the source optimisation. Note that several Minucci 2024 values sit exactly on the optimiser’s bounds, so they are identifiability-limited rather than precisely estimated; treat the column as a per-subject model input, not as a clinical measurement.

NDIV (canonical for the fitted number of divisions per activated cell in an expansion-driven cell-therapy model)

• Description: Per-subject number of cell divisions that an activated (antigen-stimulated) therapeutic cell undergoes before differentiating into the effector phenotype. Drives the expansion amplification factor 2^NDIV applied to the activated-cell-to-effector flux. Non-integer values are expected and meaningful: the parameter is fitted on a continuous scale and represents a population-average division count, not a per-cell integer.
• Units: divisions (dimensionless count, continuous-valued)
• Type: continuous
• Scope: specific
• Reference category: n/a – multiplicative expansion parameter.
• Source aliases:
  • ndiv – column in the Minucci 2024 Supplementary Material Table 1 (1).XLSX, sheet params-case-study-final (the sheet’s units cell reads # cells, which is a labelling error: the description column and the paper’s Methods 2.3 both define it as a division count); used in Minucci_2024_CART_qsp.R.
• Example models: Minucci_2024_CART_qsp.R (fitted per-patient values 15.88 to 27.70 divisions across the 13 IM19 patients).
• Notes: Scope: specific because the value is the output of a per-subject fit to that subject’s own cellular-kinetic trajectory, not an independently measurable patient characteristic. Minucci 2024 global sensitivity analysis identifies NDIV as the single most influential parameter for CAR T-cell Cmax, contributing more than 80 percent of the variance – so the column is load-bearing and must not be defaulted. Although the paper classifies it as a drug-product (manufacturing) property, it is carried as a per-subject covariate because CAR T-cell products are manufactured from each patient’s own cells. Promote to general only if a second cell-therapy extraction retains an identically-defined division-count input.

FMEM (canonical for the fitted effector-to-memory conversion fraction in a cell-therapy model)

• Description: Per-subject fraction of effector therapeutic cells that convert to the long-lived memory phenotype at the point of effector-cell loss; the remaining 1 - FMEM truly die. Governs the long-term persistence phase of cellular kinetics.
• Units: fraction (0 to 1)
• Type: continuous
• Scope: specific
• Reference category: n/a – multiplicative branching fraction on the effector-loss flux.
• Source aliases:
  • fmem – column in the Minucci 2024 Supplementary Material Table 1 (1).XLSX, sheet params-case-study-final; used in Minucci_2024_CART_qsp.R.
• Example models: Minucci_2024_CART_qsp.R (fitted per-patient values spanning nine orders of magnitude, 2.91e-10 to 0.1, across the 13 IM19 patients).
• Notes: Scope: specific because the value is a per-subject fit output rather than a measured characteristic. Three of the 13 Minucci 2024 patients (F0110, F0111, F0123) sit exactly on the 0.1 upper optimisation bound and three more (F0119, F0125, F0126) are numerically indistinguishable from zero, so the column is identifiability-limited at both ends; the paper interprets the high-FMEM patients as those with no clear contraction phase in their cellular kinetics. Distinct from a residual-error or variance term: this is a structural branching fraction supplied per subject as data.

FCD8TDP (canonical for the CD8-positive fraction of an infused cell-therapy drug product)

• Description: Fraction of the infused therapeutic cell dose that is CD8-positive; the remaining 1 - FCD8TDP is CD4-positive. Characterises the composition of the administered product rather than the patient, and is used to split a single total cell dose between the CD8 and CD4 arms of a phenotype-resolved cellular-kinetic model.
• Units: fraction (0 to 1)
• Type: continuous
• Scope: specific
• Reference category: n/a – dose-splitting fraction.
• Source aliases:
  • fCD8Tdp – column in the Minucci 2024 Supplementary Material Table 1 (1).XLSX, sheet params-case-study-final, derived from the per-patient CD4:CD8 ratio of the infused IM19 product reported in Ying 2021; used in Minucci_2024_CART_qsp.R.
• Example models: Minucci_2024_CART_qsp.R (per-patient values 0.290 to 0.943 across the 13 IM19 patients; drives amt_CD8 = FCD8TDP * WT * dose_per_kg and amt_CD4 = (1 - FCD8TDP) * WT * dose_per_kg in the event table, and is exposed in model() as the CD8_dose_frac output).
• Notes: A drug-product characterisation covariate, not a patient covariate – the closest existing analogues are the FORM_<drug>_<formulation> formulation indicators, but those are categorical product selectors whereas this is a continuous composition fraction. Scope: specific because autologous cell-therapy product composition is measured per manufacturing run and has no analogue in small-molecule or monoclonal-antibody popPK. Note that Minucci 2024 assumes 100 percent product viability (fViable = 1), so FCD8TDP alone determines the CD8/CD4 dose split; a source reporting a viability below 1 would need that factor carried separately rather than folded into this column.

Laboratory / disease-activity

ALBR (canonical for serum albumin normalized to the laboratory’s upper limit of normal)

• Description: Serum albumin normalized to the laboratory’s upper limit of normal (albumin_observed / ULN_albumin).
• Units: (unitless ratio)
• Type: continuous
• Scope: specific
• Reference category: n/a – used as a power term (ALBR / <ref>)^exponent. Reference 0.78 used in Xu 2019 (corresponds to a median serum albumin of 38 g/L at a typical ULN of ~48.7 g/L).
• Source aliases: none.
• Example models: Xu_2019_sarilumab.R.
• Notes: Xu 2019 normalizes to each site’s ULN so that values across multiple labs with different reference ranges can be pooled. Scoped specific because the ULN-normalization convention is tied to the Xu 2019 analysis plan; future papers using the same ratio should either add themselves to the example_models list or promote this entry to general.

Inflammatory-bowel-disease disease-activity covariates

SCORE_CALPRO (canonical for fecal calprotectin)

• Description: Fecal calprotectin, a gut-inflammation biomarker (baseline or time-fixed per subject unless a paper explicitly uses a time-varying value).
• Units: mg/kg stool (equivalent to ug/g). Document per-model via covariateData[[SCORE_CALPRO]]$units.
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (SCORE_CALPRO / ref)^exponent. Reference 700 mg/kg used in Rosario 2015 (overall population median).
• Source aliases:
  • CALPRO – prior canonical name (pre-2026-06-19 SCORE_ family standardization).
• Example models: Rosario_2015_vedolizumab.R (reference 700 mg/kg; exponent +0.0310 on linear clearance CLL).
• Notes: Common IBD severity biomarker (inflammation of the gut epithelium). Assays typically report in ug/g stool; 1 ug/g = 1 mg/kg. Document per-model whether baseline-only or time-varying in covariateData[[SCORE_CALPRO]]$notes.

SCORE_CDAI (canonical for Crohn’s Disease Activity Index)

• Description: Crohn’s Disease Activity Index composite score. Higher values indicate more active disease; <150 remission, 150-219 mild, 220-450 moderate, >450 severe. Defined only for patients with a CD diagnosis; set to the reference value (or gate via the indicator) for UC patients.
• Units: (score, 0-600)
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (SCORE_CDAI / ref)^exponent. Reference 300 used in Rosario 2015 (typical moderate-CD score).
• Source aliases:
  • CDAI – prior canonical name (pre-2026-06-19 SCORE_ family standardization).
• Example models: Rosario_2015_vedolizumab.R (reference 300; exponent -0.0515 on CLL gated by IBD_CD so the effect applies only to CD patients), Yu_2022_crohns_cdai150_mbma.R, Yu_2022_crohns_cdai100_mbma.R, Yu_2022_crohns_cdai70_mbma.R, Yu_2022_crohns_dcdai_mbma.R (study-arm MEAN baseline CDAI in a model-based meta-analysis; power form (SCORE_CDAI / ref)^theta on every drug effect with exponents -5.22, -3.21, -2.03 and -2.05 and references 303.19, 307.60, 301.56 and 306.285, each the arm mean of that endpoint’s analysis dataset).
• Notes: Mutually exclusive with SCORE_PMAYO in pooled UC+CD populations: each patient has exactly one disease-activity score (SCORE_CDAI for CD, partial Mayo for UC). Gate via the IBD_CD indicator when pooling.

SCORE_SESCD (canonical for Simple Endoscopic Score for Crohn’s Disease)

• Description: Simple Endoscopic Score for Crohn’s Disease (SES-CD), the standardised ileocolonoscopic severity instrument. Four variables (ulcer size, ulcerated surface, affected surface, presence of narrowing) are each scored 0-3 in each of five bowel segments, giving a total of 0-56. Higher values indicate more severe endoscopic disease. Baseline value unless a model states otherwise. Defined only for patients with a CD diagnosis; set to the reference value (or gate via the indicator) for UC patients.
• Units: (score, 0-56)
• Type: continuous
• Scope: general
• Reference category: n/a – a continuous regressor. Both centred-exponential forms exp(theta * (SCORE_SESCD - ref)) and power scaling (SCORE_SESCD / ref)^exponent occur; reproduce the source form exactly, because the intercept or typical value absorbs the centring. Reference 12 used in Moein 2025 (the CD baseline median in that pooled analysis, and the value UC subjects are assigned so the effect cancels).
• Source aliases:
  • SES-CD, SESCD, SES_CD – standard spellings of the instrument in source tables and figures.
• Example models: Moein_2025_etrolizumab.R (exponential effect on CL, exp(0.00656 * (SCORE_SESCD - 12)), roughly 0.7% higher clearance per score point; UC subjects are scored at the reference 12 so the term cancels for them), Moein_2025_etrolizumab_induction_endorem.R, Moein_2025_etrolizumab_maintenance_endorem.R (log-odds shift on the endoscopic-remission logit, centred at the same reference 12).
• Notes: The endoscopic counterpart of the symptom-based SCORE_CDAI, and the CD analogue of the UC endoscopic instrument SCORE_MAYO_E; a paper reporting both SES-CD and CDAI is reporting two genuinely different constructs (objective mucosal damage versus patient-reported activity) and both belong in the model that retained them. Mutually exclusive with SCORE_MAYO_E in pooled UC+CD populations: each patient has exactly one endoscopic score. Gate via the IBD_CD indicator when pooling, or – as Moein 2025 does – assign UC subjects the reference value so a centred effect cancels identically. Where a model’s covariate-effect percentiles are computed over CD patients only (Moein 2025 Figure 1), record that in the per-model covariateData[[SCORE_SESCD]]$notes, because the covariate’s distribution is not the pooled one. Ratified canonically by the maintainers on 2026-09-11 alongside the Moein 2025 etrolizumab extraction as a well-formed member of the already-populated SCORE_<instrument> family.

SCORE_PMAYO (canonical for partial Mayo score)

• Description: Partial Mayo score for ulcerative colitis (sum of stool-frequency, rectal-bleeding, and physician-global subscores, range 0-9). Higher values indicate more active disease. Defined only for patients with a UC diagnosis; gate via the IBD_CD indicator when pooling UC+CD.
• Units: (score, 0-9)
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (SCORE_PMAYO / ref)^exponent. Reference 6 used in Rosario 2015 (typical moderate-UC score).
• Source aliases:
  • PMAYO – prior canonical name (pre-2026-06-19 SCORE_ family standardization).
• Example models: Rosario_2015_vedolizumab.R (reference 6; exponent +0.0408 on CLL gated by (1 - IBD_CD) so the effect applies only to UC patients).
• Notes: The partial Mayo score excludes the endoscopy subscore (the full Mayo score is 0-12). Mutually exclusive with SCORE_CDAI in pooled UC+CD populations.

SCORE_MAYO_E (canonical for baseline Mayo endoscopic subscore)

• Description: Mayo endoscopic subscore at baseline for ulcerative colitis, integer 0-3 (0 = normal / inactive disease, 1 = mild, 2 = moderate, 3 = severe). The endoscopic subscore is one of the four components of the full Mayo score (0-12); the partial Mayo score (SCORE_PMAYO) excludes it. Time-fixed per subject.
• Units: (score, 0-3)
• Type: categorical
• Scope: general
• Reference category: depends on per-model encoding – papers that include the full 0-3 range typically reference Mayo 0 or 1 (mild), while papers restricted to moderate-to-severe UC (the typical biologic-induction-therapy population) reference Mayo 2 or 3. Document the per-model reference category in covariateData[[SCORE_MAYO_E]]$reference_category.
• Source aliases:
  • MAYO_E – prior canonical name (pre-2026-06-19 standardization audit).
  • MPRE – used in Faelens_2021_infliximab.R (NONMEM column for “Mayo endoscopic score pre-induction”). The Faelens 2021 dataset additionally codes a “missing” sentinel MPRE = -99; treat this as out-of-domain when applying the model and document per-model.
• Example models: Faelens_2021_infliximab.R (categorical effect on KE: separate typical KE for Mayo 1 / Mayo 2 / Mayo 3; reference category Mayo 2).
• Notes: Distinct from the full Mayo score (0-12) and the partial Mayo score (SCORE_PMAYO, 0-9). The endoscopic subscore alone is the core inclusion criterion in many UC induction-therapy popPK datasets (typically Mayo 2 or 3 = moderate-to-severe disease). Mutually compatible with SCORE_PMAYO in datasets that report both.

SCORE_SLEDAI (canonical for Systemic Lupus Erythematosus Disease Activity Index)

• Description: Systemic Lupus Erythematosus Disease Activity Index score, a weighted composite of 24 clinical and laboratory features scoring 1-8 points each and summed to a total of 0-105. Higher values indicate more active SLE disease. Time-fixed at baseline in most SLE trials but can be time-varying when measured at follow-up. In model-based meta-analyses of SLE, the per-arm mean baseline SLEDAI (rather than an individual score) is the modelled covariate.
• Units: (score, 0-105)
• Type: continuous
• Scope: general
• Reference category: n/a – used centered on a per-paper reference value (10.5 in Goteti 2024, the across-arm mean baseline SLEDAI across the 81 study arms). Enters via (1 + beta * (SCORE_SLEDAI - reference)) on latent disease activity.
• Source aliases:
  • SLEDAI – printed name in Goteti 2024 Table 2 / Table 3 / equations; same orientation, no transformation.
  • SLEDAI (mean) – Goteti 2024 covariate-summary column header.
• Example models: Goteti_2024_SLE_mbma.R (per-arm mean baseline SLEDAI; centered at 10.5; multiplicative shifts of +0.208 per unit on latent mu_ijk and +0.255 per unit on latent delta_ijk).
• Notes: SLEDAI has several closely related scoring variants (SLEDAI-2K, SELENA-SLEDAI); the numeric range and interpretation are similar and papers use SLEDAI as an umbrella label. When a paper uses a specific variant, document in covariateData[[SCORE_SLEDAI]]$notes. Distinct from disease-specific composite outcomes like BICLA / SRI / LLDAS / CLASI (which are treatment-response composites, not disease-activity scores). Follows the SCORE_* family pattern (SCORE_EASI, SCORE_MGADL, SCORE_CDAI, etc.); canonical name is SCORE_SLEDAI without a B or BL prefix to match the AGE / WT / ALB pattern where baseline vs time-varying status is recorded in notes rather than the column name.

ENDO_ULCER (canonical for endoscopically active luminal disease at baseline)

• Description: 1 = mucosal ulcerations confirmed at baseline ileocolonoscopy / endoscopy (endoscopically active luminal disease), 0 = no mucosal ulcerations at baseline. Time-fixed (assessed at study entry).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no mucosal ulcerations at baseline).
• Source aliases: none known.
• Example models: Aguiar_2021_ustekinumab.R (Aguiar 2021 Table 3; covariate on baseline fecal calprotectin FC0: 213 mg/kg with ulcers vs 102 mg/kg without).
• Notes: IBD-specific structural-disease-activity indicator, distinct from the symptom-driven SCORE_CDAI / SCORE_PMAYO and the biomarker-driven SCORE_CALPRO / CRP. Useful as a covariate on baseline biomarkers (FC, CRP) and as a stratifier for simulations of biochemical remission.

Inflammatory-bowel-disease diagnosis

IBD_CD (canonical for Crohn’s disease indicator)

• Description: 1 = Crohn’s disease (CD) diagnosis, 0 = ulcerative colitis (UC) diagnosis. Used to gate pooled UC+CD population models where some parameters or covariate effects differ by diagnosis.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (ulcerative colitis).
• Source aliases:
  • DIAGNOSIS, DX (categorical "UC"/"CD") – derive IBD_CD = as.integer(DX == "CD").
• Example models: Rosario_2015_vedolizumab.R (two typical-CLL switch between UC and CD; gates SCORE_PMAYO and SCORE_CDAI effects; multiplicative +1% effect on Vc).
• Notes: Rosario 2015 models separate typical CLL for UC vs CD and gates the partial-Mayo (UC-only) and SCORE_CDAI (CD-only) disease-activity covariates via this indicator.

Concomitant / prior medication

CONMED_AD (canonical for concomitant Alzheimer’s-symptomatic medication)

• Description: 1 = subject is on concomitant Alzheimer’s-symptomatic medication (typically a cholinesterase inhibitor and/or memantine) at the observation, 0 = not on such medication. Time-fixed in the Conrado 2014 source dataset (one indicator per subject) and treated as such here; future models that need a time-varying form may register a successor canonical or document time-varying use in covariateData[[CONMED_AD]]$notes.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 1 (on concomitant Alzheimer’s-symptomatic medication) in the Conrado 2014 model – the source paper labels CONMED_AD = 1 as the “most common” category and centres the slope-effect coefficient on this group, so the multiplicative factor on disease-progression slope is 1 for CONMED_AD = 1 and 1 + e_sl_conmed_ad_off for CONMED_AD = 0. The non-standard “most-common-as-reference” convention is preserved from the source for traceability; future Alzheimer’s models that adopt the more common “off-treatment as reference” convention should register a successor canonical (CONMED_AD_TREATED) rather than overloading this one.
• Source aliases:
  • COMED2 – used in Conrado_2014_alzheimer.R (DDMORE Foundation Model Repository entry DDMODEL00000290). The 2 suffix in the source distinguishes this binary from upstream COMED and PRIMCOMED columns (free-text concomitant-medication entries) in the same NONMEM input dataset; the binary COMED2 flag is what enters the model.
• Example models: Conrado_2014_alzheimer.R (multiplicative factor on the typical-value disease-progression slope: slope_factor = 1 + (1 - CONMED_AD) * e_sl_conmed_ad_off with e_sl_conmed_ad_off = -0.302, i.e. ~30% smaller progression slope for the off-treatment reference cohort).
• Notes: The source .mod does not specify the symptomatic-medication class beyond the binary flag; the Conrado 2014 publication context (CAMD ADAS-Cog disease-progression dataset, 2014) makes cholinesterase-inhibitor / memantine the dominant interpretation. Treating CONMED_AD = 1 as the reference category is unusual relative to the rest of the CONMED_* family (CONMED_PARA, CONMED_NSAID, CONMED_AZA, etc.) which all use 0 = not-on as reference; the inversion is preserved here only because the source paper’s coefficient was estimated with the “most common = 1” convention.

CONMED_AED (canonical for any concomitant antiepileptic drug coadministration)

• Description: 1 = subject is on at least one concomitant antiepileptic drug (AED) other than the modelled AED at the PK observation, 0 = on the modelled AED as monotherapy. Time-varying when concurrent AEDs cycle on / off across study occasions, time-fixed when the source paper analyses chronic-maintenance cohorts whose concurrent therapy is unchanged across the analysis window.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (modelled-AED monotherapy).
• Source aliases:
  • CO_AED – used in Yukawa_1990_phenytoin.R (paper’s CO indicator inverted: source CO = 1 if PHT alone, theta_co otherwise; canonical CONMED_AED = 1 - CO_indicator so 0 is the PHT-monotherapy reference).
• Example models: Yukawa_1990_phenytoin.R (multiplicative ^CONMED_AED factor on Vmax (1.08) and Km (1.32) for chronic phenytoin patients on at least one of phenobarbital, carbamazepine, valproate, primidone, clonazepam, sultiame, ethotoin, ethosuximide, acetazolamide, or diazepam), Yukawa_2002_clonazepam_pediatric.R (paired with CONMED_AED_GE2 to recover the paper’s 3-tier drug-interaction factor on clonazepam CL/F).
• Notes: Generic concomitant-AED indicator covering the heterogeneous mix of older AEDs (PB, CBZ, VPA, primidone, clonazepam, etc.) studied alongside the modelled drug; the per-paper list of qualifying AEDs must be documented in covariateData[[CONMED_AED]]$notes. Distinct from drug-specific concomitant-AED indicators (e.g., a future CONMED_PB for concomitant phenobarbital alone) which would warrant separate canonicals when a paper distinguishes effects by AED class. When a paper distinguishes AED-count tiers (monotherapy vs +1 AED vs +>=2 AEDs) rather than the pooled mono-vs-anything contrast, pair CONMED_AED with the sibling [[CONMED_AED_GE2]] binary; the >=2-AEDs tier is CONMED_AED = 1 AND CONMED_AED_GE2 = 1, and data assemblers must enforce that CONMED_AED_GE2 = 1 implies CONMED_AED = 1. Follows the CONMED_* family pattern (CONMED_AZA, CONMED_NSAID, etc.).

CONMED_AED_GE2 (canonical for >=2 concomitant antiepileptic drugs)

• Description: 1 = subject is on at least two concomitant antiepileptic drugs (AEDs) other than the modelled AED at the PK observation, 0 = on at most one concomitant AED (i.e., modelled-AED monotherapy or modelled AED + one concomitant AED). Sibling binary to [[CONMED_AED]] for AED-count-tier covariate models; the two binaries together encode a 3-level tier (monotherapy / +1 / +>=2). Time-varying when concurrent AEDs cycle on / off across study occasions, time-fixed when the source paper analyses chronic-maintenance cohorts whose concurrent therapy is unchanged across the analysis window.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (zero or one concomitant AED). The >=2-tier reference (CONMED_AED_GE2 = 0 with CONMED_AED = 1) is “1 concomitant AED”; the absolute reference (CONMED_AED_GE2 = 0 with CONMED_AED = 0) is “monotherapy”.
• Source aliases:
  • CONMED_AED_GE2 – used in Yukawa_2002_clonazepam_pediatric.R (paper’s Polytherapy (B) bucket per Table I and Table II groupings: any record with >=2 AEDs alongside clonazepam).
• Example models: Yukawa_2002_clonazepam_pediatric.R (paired with [[CONMED_AED]] inside model() to recover the paper’s 3-tier drug-interaction factor: monotherapy DIF = 1, +1 AED DIF = 1.18, +>=2 AEDs DIF = 2.12 * TBW^(-0.119)).
• Notes: Specific scope; the binary is only meaningful in papers that distinguish a >=2-AEDs tier from a +1-AED tier. Promote to general scope when a second AED-count-tier paper ratifies the name. Mutual-consistency constraint: CONMED_AED_GE2 = 1 MUST imply CONMED_AED = 1. A record asserting CONMED_AED_GE2 = 1 and CONMED_AED = 0 is malformed and the model() body will silently treat it as the >=2 tier (since the indicator decomposition uses CONMED_AED * (1 - CONMED_AED_GE2) for the +1 tier and CONMED_AED_GE2 for the >=2 tier); data assemblers should validate the pair before passing to rxSolve(). Distinct from drug-specific CONMED_<drug> indicators (CONMED_CBZ, CONMED_VPA, CONMED_PB) which capture a specific AED rather than a count tier; when a future paper distinguishes BOTH the count AND the specific drugs, the two families coexist.

CONMED_ABX (canonical for any-other-antibiotic composite coadministration indicator)

• Description: 1 = subject is receiving at least one concomitant systemic antibiotic other than the antibiotic being modelled at the PK observation, 0 = no other concomitant antibiotic. A composite pooled umbrella indicator for anti-infective popPK analyses in which the source paper records only whether additional antibacterial therapy was co-administered and never names the individual agents, so no per-INN indicator can be assigned. Time-varying in principle (co-therapy starts and stops during an admission); time-fixed when the source paper captures the flag once at treatment start.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant antibiotic other than the modelled drug).
• Source aliases:
  • COMED – used in Zieck_2023_ceftazidime.R (Supplementary File S1 $INPUT column; enters $PK as FLAG1 via IF(COMED.EQ.1)FLAG1=1). Note that COMED is a generic NONMEM column name meaning “co-medication”; map it to CONMED_ABX only when the source paper defines it specifically as other-antibiotic use, and to the appropriate per-INN or per-class canonical otherwise.
  • Concomitant other antibiotic use – Zieck 2023 Table 2 row label.
• Example models: Zieck_2023_ceftazidime.R (power-of-coefficient effect on CL: CL_typ * 1.56^CONMED_ABX, i.e. +56% clearance when another antibiotic is co-administered; 27/40 patients exposed).
• Notes: Distinct from the named-agent indicators CONMED_MER (meropenem), CONMED_GEN (gentamicin), CONMED_CIP (ciprofloxacin) and CONMED_FUSIDIC (fusidic acid), which model a specific co-administered antibiotic. Use CONMED_ABX only when the source paper itself pools all other antibacterials under a single unnamed binary; use the per-INN indicators when the paper identifies and estimates effects for specific agents. Structurally parallel to [[CONMED_IMMUNOMOD]], the register’s other “source paper does not separate the individual agents” composite. An ABX = 1 flag is frequently confounded with infection severity and with the breadth of empiric coverage, so a positive clearance effect should be read as an empirical marker of a sicker / more heavily co-treated subset rather than as a mechanistic drug-drug interaction – Zieck 2023’s Discussion explicitly makes this caveat about its own 1.56 estimate (“we could not find a physiological explanation for this association… it cannot be ruled out that the identification of this association is based on coincidence”).

CONMED_ABI (canonical for concomitant abiraterone coadministration indicator)

• Description: 1 = subject is coadministered abiraterone acetate (oral CYP17A1 inhibitor used in metastatic castration-resistant prostate cancer; also a clinically relevant CYP2D6 / CYP2C8 inhibitor) at the PK observation, 0 = no abiraterone. Per-subject in Yoshida 2021 because abiraterone coadministration was a study-design factor (every patient in study GO27983 received 1000 mg abiraterone QD with prednisone / prednisolone 5 mg BID; six of 21 patients in JO29655 received abiraterone in stage 2; the other three studies enrolled no abiraterone arms).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no concomitant abiraterone).
• Source aliases:
  • ABIRATER – used in Yoshida_2021_ipatasertib.R (Yoshida 2021 NONMEM control stream variable name; same per-subject 0/1 orientation).
• Example models: Yoshida_2021_ipatasertib.R (linear-additive effects: -18.5% on apparent ipatasertib parent CL/F when abiraterone is present (Table 3 theta_CLI,Abi = -0.185); on the apparent M1 bioavailability the abiraterone effect is +61.5% but ONLY at multiple-dose state (Table 4 theta_FM1,Abi = 0.615 applied as (1 + 0.615 * MULTI_DOSE * CONMED_ABI) so that the same single-dose effect was zero per the source NONMEM control stream)).
• Notes: Specific scope until a second model legitimately ratifies abiraterone as a covariate; promote to general once that happens. Distinct from CONMED_AZOLE (CYP3A4 inhibitor used for fungal indications) and from CONMED_STEROID (the prednisone / prednisolone component that is always co-administered with abiraterone in mCRPC); when a model explicitly separates abiraterone vs the steroid backbone, use both covariates. Yoshida 2021 also reports that the precise mechanism of the abiraterone-on-ipatasertib effect is unknown (the standard CYP3A4 / CYP2C8 / CYP2D6 routes were ruled out by in-vitro and DDI studies), so the indicator captures the empirical exposure shift rather than a defined enzyme-inhibition mechanism.

CONMED_ALTEPLASE (canonical for concomitant intravenous alteplase (r-tPA) coadministration indicator)

• Description: 1 = subject received concomitant intravenous alteplase (recombinant tissue plasminogen activator, r-tPA; intravenous thrombolysis, IVT) before or concurrent with the index PK measurement window, 0 = no concomitant alteplase. Time-fixed at the index event in stroke / reperfusion contexts where IVT is administered in the emergency window prior to endovascular thrombectomy. Distinct from the broader / class-level “any thrombolytic” indicator (which would also include tenecteplase, reteplase, urokinase, streptokinase); the canonical name is drug-specific because the published popPK literature ties the covariate to alteplase pharmacokinetics directly rather than to the broader tissue-plasminogen-activator class.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant alteplase / IVT).
• Source aliases:
  • IVT – intravenous thrombolysis indicator; used in Vos_2025_iminobiotin.R (Vos 2025 Supplemental Table S8 uses “IVT” interchangeably with “alteplase” because the Dutch standard-of-care thrombolytic for LVO stroke in 2022-2023 was alteplase exclusively; the same column form IVT = 0/1 would translate to CONMED_ALTEPLASE in any nlmixr2lib reuse).
• Example models: Vos_2025_iminobiotin.R (multiplicative effect on 2-iminobiotin clearance: cl *= (1 + e_alteplase_cl * CONMED_ALTEPLASE) with e_alteplase_cl = 0.6469, i.e. concomitant alteplase increases 2-IB CL by ~65% from 9.29 L/h to 15.3 L/h in adult LVO-stroke patients; founding example).
• Notes: General scope because alteplase is the dominant thrombolytic agent in acute ischemic stroke worldwide (with tenecteplase emerging as the principal alternative); future popPK models that combine concomitant tenecteplase as a covariate should register a sibling CONMED_TENECTEPLASE canonical rather than reusing this one (the two agents differ in plasmin-generation kinetics, fibrin specificity, and elimination half-life and so should not be pooled). The Vos 2025 mechanism for the +65% CL effect on 2-iminobiotin is not firmly established: the paper’s Discussion explicitly rules out plasmin-mediated cleavage (2-IB is a small molecule, not a peptide like nerinetide), so the indicator captures an empirical exposure shift rather than a defined biochemical mechanism.

CONMED_AMIO (canonical for concomitant amiodarone coadministration indicator)

• Description: 1 = subject is coadministered amiodarone (Class III antiarrhythmic; CYP3A4 / CYP2C9 / P-gp inhibitor) during the study, 0 = no concomitant amiodarone. Time-varying when amiodarone start / stop events are captured; the Xia 2024 source treats amiodarone as time-fixed at the analysis baseline.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant amiodarone).
• Source aliases:
  • CM1 – used in Xia_2024_warfarin.R (Xia 2024 Figure 1 / Table 2 covariate-screening variable labelling: CM1 = 1 indicates combined amiodarone, 0 no combination).
• Example models: Xia_2024_warfarin.R (piecewise multiplicative effect on warfarin EC50: ec50 *= (1 + e_amio_ec50 * CONMED_AMIO) with e_amio_ec50 = -0.602, i.e. amiodarone reduces EC50 by ~60% in the Han Chinese cohort).
• Notes: Amiodarone is the canonical CYP2C9 / CYP3A4 inhibitor that potentiates warfarin’s anticoagulant effect in clinical practice; the Xia 2024 cohort prevalence was 25.7% (Table 1). The per-paper definition (any amiodarone use vs current loading-dose use vs steady-state use) should be documented in covariateData[[CONMED_AMIO]]$notes.

CONMED_AMINO (canonical for concomitant aminosalicylate therapy)

• Description: 1 = on concomitant aminosalicylate (5-aminosalicylic acid / mesalamine / mesalazine / olsalazine / sulfasalazine etc.) therapy at the PK observation, 0 = not.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant aminosalicylate).
• Source aliases: AMINO – used in Rosario_2015_vedolizumab.R.
• Example models: Rosario_2015_vedolizumab.R (power-form on CLL: CLL * 1.02^CONMED_AMINO).
• Notes: Covers the full aminosalicylate class (5-ASA is the single active moiety shared by most agents); use CONMED_AMINO rather than CONMED_5ASA unless the source paper explicitly restricts the indicator to 5-ASA monotherapy.

CONMED_ARIPIPRAZOLE (canonical for concomitant aripiprazole coadministration indicator)

• Description: 1 = subject is co-prescribed aripiprazole (second-generation antipsychotic; CYP2D6 and CYP3A4 substrate, and a competitive inhibitor of CYP2D6 at clinical concentrations) at the PK observation, 0 = no concomitant aripiprazole. Time-fixed at the analysis baseline in the founding source, which draws comedication status from a retrospective medical log; time-varying encoding is appropriate wherever aripiprazole start / stop events are captured.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no concomitant aripiprazole).
• Source aliases:
  • ARI – used in Zhang_2024_olanzapine.R (Zhang 2024 Equation 6: “ARI was aripiprazole, when patients took aripiprazole, ARI was 1, otherwise ARI was 0”; same 0/1 orientation).
• Example models: Zhang_2024_olanzapine.R (linear-shift effect on apparent oral clearance: cl *= (1 + e_ari_cl * CONMED_ARIPIPRAZOLE) with e_ari_cl = -0.392, i.e. concomitant aripiprazole lowers olanzapine CL/F by 39.2%, equivalently a with:without clearance ratio of 0.608:1 as stated in the Discussion and plotted in Figure 1H; founding example).
• Notes: Aripiprazole appears in the popPK literature on both sides of the DDI – as the victim drug (Knights_2015_aripiprazole.R, Koue_2007_aripiprazole.R, Kim_2008_aripiprazole.R, where CYP2D6 phenotype is the covariate on aripiprazole’s own clearance, encoded via CYP2D6_PM) and, here, as the perpetrator on a co-prescribed CYP2D6 substrate. Use CONMED_ARIPIPRAZOLE only for the perpetrator role; a model of aripiprazole itself does not need a covariate for the drug being modelled. Distinct from CYP2D6_PM: the comedication indicator captures a phenoconversion-like exposure shift in patients of unrecorded genotype, whereas CYP2D6_PM captures the germline poor-metaboliser phenotype, and a study measuring both should carry both columns. Specific scope until a second model registers aripiprazole as a covariate; promote to general once that happens. Zhang 2024’s cohort prevalence was 4 of 65 patients (6.2%, Table 2), so the -0.392 estimate rests on a small exposed subgroup – its bootstrap 95% interval [-0.535, -0.194] is correspondingly wide and should be quoted alongside the point estimate.

CONMED_AVD (canonical for brentuximab vedotin + AVD (adriamycin/doxorubicin, vinblastine, dacarbazine) combination indicator)

• Description: 1 = subject is receiving brentuximab vedotin in combination with the AVD chemotherapy backbone (adriamycin a.k.a. doxorubicin, vinblastine, dacarbazine) for newly diagnosed advanced-stage Hodgkin lymphoma; 0 = otherwise (single-agent brentuximab vedotin). Encodes the A+AVD frontline regimen as a study-design covariate on ADC clearance.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (single-agent brentuximab vedotin – no AVD coadministration).
• Source aliases:
  • DOX – used in Zhou_2025_brentuximab.R (the NONMEM dataset uses the doxorubicin-administration flag as the AVD-coadministration indicator because doxorubicin is given on the same days as the other AVD agents in this regimen).
• Example models: Zhou_2025_brentuximab.R (power-form effect on ADC clearance: CL * 2.12^CONMED_AVD – ADC clearance is ~2.1-fold higher under A+AVD vs single-agent BV).
• Notes: Distinct from CONMED_CHEMO (which is nivolumab + platinum-based chemotherapy). The A+AVD regimen is the standard chemotherapy backbone for frontline classical Hodgkin lymphoma; promote to general scope if a second BV paper reports the same A+AVD covariate effect with a comparable encoding.

CONMED_AZA (canonical for concomitant azathioprine)

• Description: 1 = on concomitant azathioprine at the PK observation, 0 = not on azathioprine.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant azathioprine).
• Source aliases: AZA – used in Rosario_2015_vedolizumab.R.
• Example models: Rosario_2015_vedolizumab.R (power-form on CLL: CLL * 0.998^CONMED_AZA; effect ~= null).
• Notes: Thiopurine immunomodulator used as maintenance therapy in IBD. Standard convention is baseline-use-only, but time-varying use is permitted; document per-model.

CONMED_AZITHROMYCIN (canonical for concomitant azithromycin coadministration indicator)

• Description: 1 = azithromycin coadministered with the victim-drug dose record, 0 = no concomitant azithromycin. Azithromycin is a macrolide antibiotic and a P-glycoprotein inhibitor with little CYP3A inhibition, so its DDI effects are usually attributed to intestinal P-gp.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant azithromycin).
• Source aliases:
  • AZIFL – used in Dave_2019_venetoclax.R (Dave 2019 PK control stream IF(AZIFL.EQ.1) TVFA = THETA(11) ; Azithromycin).
• Example models: Dave_2019_venetoclax.R (multiplicative factor 0.65 on venetoclax relative bioavailability, i.e. 35% lower F, Dave 2019 Table 2 ‘Azithromycin on F1’; healthy-volunteer DDI study V, Agarwal 2018 Adv Ther 35:2015; indicator set on the venetoclax dose record).
• Notes: Auto-approved member of the CONMED_<INN> family. Record per-model whether the effect is placed on bioavailability or on clearance, and whether it is time-varying (on the dose record only) or per-subject.

CONMED_CLARITHROMYCIN (canonical for concomitant clarithromycin coadministration indicator)

• Description: 1 = clarithromycin coadministered with the victim drug (at the dosing regimen the source studied), 0 = no concomitant clarithromycin. Clarithromycin is a macrolide antibiotic and a strong mechanism-based CYP3A inhibitor (and P-glycoprotein inhibitor), so its DDI effects are usually placed on the apparent clearance of CYP3A substrates.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant clarithromycin).
• Source aliases:
  • CLR – used in Kurosawa_2021_bedaquiline.R (Kurosawa 2021 Methods equation CL/F = CLpop * (1 + theta)^CLRi, CLRi = 0 no coadministration, 1 coadministration).
• Example models: Kurosawa_2021_bedaquiline.R (power-form effect on bedaquiline apparent clearance, (1 + e_conmed_clarithromycin_cl)^CONMED_CLARITHROMYCIN with e_conmed_clarithromycin_cl = -0.37, i.e. 37% lower CL/F with steady-state clarithromycin 500 mg every 12 hours; Kurosawa 2021 Table S1).
• Notes: Auto-approved member of the CONMED_<INN> family. Record per-model the clarithromycin regimen (dose, interval, whether steady state was reached before the victim dose) and whether the indicator is time-varying. Distinct from CONMED_AZITHROMYCIN (a macrolide with little CYP3A inhibition) and CONMED_ERYTHROMYCIN (a moderate CYP3A inhibitor), and from the pooled class indicators CONMED_CYP3A4_INH*.

CONMED_AZOLE (canonical for concomitant azole antifungal therapy (CYP3A4/P-gp inhibitor))

• Description: 1 = patient coadministered an azole antifungal (itraconazole, voriconazole, fluconazole, ketoconazole, posaconazole, isavuconazole, or another systemic azole) during the observation interval, 0 = no concomitant azole antifungal. Time-varying per subject because azole exposure starts and stops during the observation period.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant azole antifungal).
• Source aliases:
  • AZOLE – used in Kirubakaran_2022_tacrolimus.R.
  • CZ – used in Liu_2022_tacrolimus.R and Liu_2022_tacrolimus_cyp3a5.R (Liu 2022 merges voriconazole, itraconazole and posaconazole into one variable because few patients received any single agent; exponential effect exp(-0.40 * CONMED_AZOLE) on tacrolimus CL in pediatric HSCT recipients, -0.39 in the CYP3A5 subpopulation model).
• Example models: Kirubakaran_2022_tacrolimus.R (state-dependent typical CL/F: 21.1 L/h without azole and 4.2 L/h with azole, an 80% reduction; also a state-dependent BSV magnitude on CL/F: 61% CV without azole vs 89.5% CV with azole), Li_2019_cyclosporine.R (source column TAF, triazole antifungals only – voriconazole, itraconazole, posaconazole, fluconazole pooled; fractional effect (1 - 0.36 * CONMED_AZOLE) on intravenous cyclosporine CL in children after allo-HSCT).
• Notes: Azoles are mechanism-based CYP3A4 and P-glycoprotein inhibitors with different inhibitor potencies (itraconazole > voriconazole > fluconazole). The per-model covariateData[[CONMED_AZOLE]]$notes must document (1) which azoles are pooled into the indicator, (2) any post-cessation lag (Kirubakaran 2022 carries CONMED_AZOLE = 1 for one week after azole discontinuation to allow tacrolimus apparent clearance to stabilize given itraconazole’s long half-life), and (3) whether the indicator is a baseline-only proxy or a true time-varying flag.

CONMED_BLEOMYCIN (canonical for concomitant bleomycin coadministration indicator)

• Description: 1 = subject is receiving bleomycin as a component of the concomitant chemotherapy regimen during the observation interval, 0 = no concomitant bleomycin. Bleomycin is a glycopeptide antitumour antibiotic used in germ-cell-tumour, Hodgkin-lymphoma and squamous-cell regimens; roughly 65% of a dose is eliminated renally, so it is co-eliminated with (and potentially competes with or perturbs the renal handling of) other renally cleared cytotoxics such as methotrexate. Time-varying in principle because bleomycin dosing is scheduled on specific protocol days relative to the index drug.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no concomitant bleomycin).
• Source aliases:
  • BLM – used in Zhao_2025_methotrexate.R (Eq. 5 writes the term as e^BLM with BLM = 0.08 when combined with BLM, otherwise = 0; the register splits that into a 0/1 indicator times an estimated coefficient).
• Example models: Zhao_2025_methotrexate.R (exponential effect on MTX clearance, exp(0.08 * CONMED_BLEOMYCIN) = 1.083-fold; bleomycin is given on day 2 of a regimen whose methotrexate + vincristine dose falls on day 1).
• Notes: Specific scope until a second model reuses the indicator. The direction found by Zhao 2025 – bleomycin accelerates methotrexate clearance – is not mechanistically established; the paper states plainly that “the influence of BLM on MTX remains unexplored” and lists enzyme induction, transporter upregulation, oxidative-stress modulation and detoxifying-enzyme activation as candidate explanations. A per-model covariateData[[CONMED_BLEOMYCIN]]$notes should record the timing of bleomycin relative to the index drug, because a covariate flagged at the subject level can conflate a genuine interaction with a regimen-cohort difference when the two drugs are not concurrent in plasma.

CONMED_BOSENTAN (canonical for concomitant bosentan coadministration indicator)

• Description: 1 = bosentan (a dual endothelin-receptor antagonist and an inhibitor of NTCP, BSEP, MRP3 and MRP4) is co-administered or co-incubated, 0 = it is not. Bosentan is a standard probe perpetrator in bile-acid and hepatic-transporter interaction work because its BSEP inhibition is the mechanism proposed for its clinical hepatotoxicity.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 = bosentan absent. Carries no magnitude information: where the model needs the perpetrator exposure, that enters through a concentration column (CONC_INHIBITOR_MEDIUM_UNBOUND_UM / CONC_INHIBITOR_CELL_UM in vitro, or a CP_<drug>_<units> column in vivo) and this indicator selects the matching potency parameters.
• Source aliases: none standardized; source papers name the drug in prose or in a treatment-arm label.
• Example models: Guo_2016_taurocholate_schh.R (selects the bosentan column of the Table 1 inhibitory potencies: NTCP 27 uM, OATP1B1 18 uM, BSEP 32.5 uM, MRP3 42 uM), FergusonSells_2022_tadalafil.R (in vivo use as a CYP3A INDUCER: background bosentan therapy in pulmonary arterial hypertension raises tadalafil CL/F. The paper’s reference category is the bosentan-TAKING patient – typical CL/F 3.23 L/h, with a fractional change of -0.418 for patients not taking bosentan – so the coefficient enters as (1 + e * (1 - CONMED_BOSENTAN)) with the canonical 1 = bosentan coding unchanged; ambrisentan users take 0).
• Notes: General scope; a well-formed member of the CONMED_<INN> family using the full INN rather than an abbreviation, per the register’s anti-abbreviation principle. Mutually exclusive with the other inhibitor indicator in any model that selects a single perpetrator’s potency set; setting all such indicators to 0 must give the uninhibited condition.

CONMED_BUSULFAN (canonical for concomitant busulfan coadministration indicator)

• Description: 1 = the subject’s transplant conditioning regimen contains busulfan, 0 = it does not. Busulfan is a bifunctional alkylating agent given as a myeloablative backbone in haematopoietic stem cell transplant conditioning, and it is known to alter both its own clearance and that of co-administered agents. Note that in conditioning regimens busulfan is typically given on days that precede the co-administered drug’s dose (in the founding example, days -9 to -6 or -9 to -7 against a melphalan dose on day -5, -4 or -3), so the indicator marks a carried-over regimen effect – regimen-related organ dysfunction and oxidative-stress-mediated pathways are the mechanisms the founding paper proposes – rather than simultaneous circulating exposure. A model that needs the magnitude of the busulfan exposure should carry a separate exposure column (the founding cohort targeted a cumulative busulfan AUC of 75,000 ugh/L for BuMel and 56,250 ugh/L for BuMelThio) rather than overloading this binary.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 = no concomitant busulfan. Effect form in Choi 2026 is a multiplicative factor on clearance: cl *= 0.846 when CONMED_BUSULFAN == 1, i.e. a 15.4% clearance reduction and a corresponding ~18% rise in melphalan exposure.
• Source aliases:
  • RegimenB – the Choi 2026 table and equation label for the concomitant-busulfan indicator; same binary quantity, no value transformation.
• Example models: Choi_2026_melphalan.R (paediatric autologous HSCT melphalan popPK; the binary distinguishes the busulfan-containing BuMel and BuMelThio regimens, 11 of 20 patients, from the MEC, BEAM and FluMel regimens).
• Notes: Auto-approved as a well-formed CONMED_<INN> member during the Choi 2026 melphalan extraction (PMC13275562). Specific scope because the column so far means “busulfan is somewhere in this conditioning regimen”, which is only well defined for transplant-conditioning cohorts; promote to general if a second paper uses it for ordinary concomitant-medication coadministration. Distinct from CONMED_ALKYLATOR-style class composites (not registered) and from the several thiotepa / etoposide / carboplatin regimen components that the founding paper screened alongside it and did not retain – each regimen drug needs its own indicator rather than a single pooled regimen code, because the founding stepwise analysis tested them separately. Do not confuse a busulfan indicator used as a covariate on some other drug’s PK (this column) with a busulfan popPK model’s own dose or exposure columns; the library already carries busulfan models (LongBoyle_2015_busulfan.R, Schreib_2024_busulfan.R) where busulfan is the analyte rather than a comedication.

CONMED_CARBOPLATIN (canonical for concomitant carboplatin coadministration indicator)

• Description: 1 = the modelled drug’s observation was collected while the subject was concurrently receiving carboplatin as the co-administered antineoplastic, 0 = no concomitant carboplatin. Carboplatin is a second-generation platinum alkylating agent given predominantly as a doublet backbone with a taxane. A per-INN member of the CONMED_* family, used when a source paper resolves carboplatin as its own co-administration stratum rather than pooling it into a generic platinum ([[CONMED_PLATIN]]) or chemotherapy-backbone ([[CONMED_CHEMO]]) indicator.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (not on concomitant carboplatin). The substantive meaning of the zero level is set by the source paper’s stratification and must be documented per model – in Faessel_2019_pevonedistat.R the zero level pools pevonedistat monotherapy with pevonedistat plus azacitidine, docetaxel or gemcitabine, none of which affected pevonedistat clearance.
• Source aliases:
  • ConCarboTax – used in Faessel_2019_pevonedistat.R (Faessel 2019 Table 3 displayed equation CL = theta1 * (BSA/1.73)^theta15 * (1 + theta7 * ConCarboTax)). The source column is a single indicator for the carboplatin + paclitaxel doublet, so it decomposes to CONMED_CARBOPLATIN * CONMED_PACLITAXEL; in that analysis dataset the two agents were only ever given together, so the product reproduces the source indicator exactly.
• Example models: Faessel_2019_pevonedistat.R (multiplicative fractional effect on pevonedistat clearance: cl *= (1 + (-0.441) * CONMED_CARBOPLATIN * CONMED_PACLITAXEL), i.e. a 44.1% clearance reduction and a corresponding 1.79-fold rise in AUC, Faessel 2019 Table 3, SE 8.8%, bootstrap 95% CI -0.51 to -0.352).
• Notes: Auto-approved as a well-formed CONMED_<INN> member during the Faessel 2019 pevonedistat extraction (PMC6848912). Carries no dose information – where a model needs the carboplatin dose or exposure, that enters through [[DOSE_CARBOPLATIN_MGKG]] or [[AUC_CARBO]], which are separate columns with separate meanings. Distinct from [[CONMED_PLATIN]], which pools any platinum (cisplatin, carboplatin, oxaliplatin, nedaplatin, satraplatin), and from [[CONMED_CHEMO]], which pools any chemotherapy backbone including paclitaxel + carboplatin; use this per-INN column only when the source paper’s covariate is carboplatin-specific and separable from other cytotoxics. Papers frequently estimate a single coefficient for a doublet rather than for either agent alone. The register’s convention is to keep one column per INN and to form the doublet in model() as the product of the two indicators, rather than registering a fused CONMED_CARBOPLATIN_PACLITAXEL name – this keeps each agent greppable and lets a later paper that does separate them reuse the same columns. Record the doublet construction in each model’s covariateData notes. The mechanism of the pevonedistat clearance reduction is unknown and neither carboplatin nor paclitaxel is an established drug-drug-interaction perpetrator (Faessel 2019 section 4.2).

CONMED_CISPLATIN (canonical for concomitant cisplatin coadministration indicator)

• Description: 1 = the subject received cisplatin together with the modelled drug during the modelled treatment period, 0 = no concomitant cisplatin. A per-INN member of the CONMED_* family, used when a source paper resolves cisplatin as its own co-administration stratum rather than pooling it into a generic platinum ([[CONMED_PLATIN]]) indicator.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant cisplatin). In Arshad_2020_fluorouracil.R the zero level is 5-FU monotherapy given with radiotherapy (colorectal and anal cancer); the one level is 5-FU plus cisplatin 20 mg/m^2/day for 5 days (oesophageal cancer), so the indicator is confounded with tumour site in that dataset.
• Source aliases:
  • 5FUcomb / 5FUmono – Arshad 2020 Table 3 (Slopecomb, Slopemono) and Fig. 3; the paper names the treatment groups rather than a data column.
• Example models: Arshad_2020_fluorouracil.R (selects between two separately estimated linear drug-effect slopes on leukocyte proliferation: 2.82 L/mg with cisplatin, 1.17 L/mg without, Arshad 2020 Table 3 bootstrap medians).
• Notes: Carries no dose information – a model that needs the cisplatin dose or exposure takes it through a separate column. Distinct from [[CONMED_PLATIN]], which pools any platinum agent, and from [[CONMED_CARBOPLATIN]].

CONMED_CBZ (canonical for concomitant carbamazepine coadministration indicator)

• Description: 1 = subject is taking carbamazepine (CBZ) as a concomitant antiepileptic drug at the PK observation, 0 = no concomitant carbamazepine. Carbamazepine is a strong CYP3A4 / UGT / P-gp inducer that increases the apparent clearance of co-administered drugs. Time-varying when carbamazepine starts / stops within the observation window; time-fixed when the source paper analyses chronic-maintenance cohorts whose AED therapy is stable across the analysis window.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant carbamazepine).
• Source aliases:
  • CBZ – used in Schoemaker_2017_brivaracetam.R (paper covariate CBZ for carbamazepine coadministration), Hashimoto_1994_zonisamide.R (paper covariate CBZ for carbamazepine coadministration on zonisamide Vmax) and Lee_2024_topiramate.R (Lee 2024 Table S1 theta4, additive enzyme-induction term on topiramate CL/F).
• Example models: Schoemaker_2017_brivaracetam.R (multiplicative effect on apparent oral clearance: cl *= (1 + 0.479 * CONMED_CBZ); +47.9% relative to no-CBZ reference, corresponding to ~32% lower brivaracetam exposure, Schoemaker 2017 Table 1), Hashimoto_1994_zonisamide.R (multiplicative power-form effect on Vmax of Michaelis-Menten zonisamide elimination: vmax *= 1.13^CONMED_CBZ; +13% relative to no-CBZ reference, Hashimoto 1994 Table II theta2), Lee_2024_topiramate.R (additive effect in absolute units: cl <- (exp(lcl) + e_cbz_cl * CONMED_CBZ + ...) * ... with e_cbz_cl = 0.703 L/h on a 1.45 L/h monotherapy intercept, i.e. +48% relative to no-CBZ reference, Lee 2024 Table S1 / Discussion).
• Notes: Drug-specific CONMED_* indicator anticipated in the [[CONMED_AED]] notes; used when a paper estimates a separate CBZ-induction effect distinct from the pooled EIAED / AED class effect. Distinct from the broader [[CONMED_EIAED]] (any enzyme-inducing AED) and [[CONMED_AED]] (any concomitant AED). When a paper distinguishes individual AEDs separately (Schoemaker 2017: PB, CBZ, VPA each carry their own coefficient; Lee 2024: PHT, CBZ, OXC, PB each carry their own coefficient), use the drug-specific canonicals [[CONMED_PB]], CONMED_CBZ, [[CONMED_OXC]], [[CONMED_PHT]], [[CONMED_VPA]] rather than collapsing into the class-level indicator. The functional form of the induction effect is per-paper and not a property of this column: it is multiplicative / fractional in Schoemaker 2017, a power form in Hashimoto 1994, and additive in absolute L/h in Lee 2024 – each model’s model() body is authoritative, so do not assume a fractional coefficient when reading a CONMED_CBZ effect size across models. Hashimoto 1994 (zonisamide) tested phenytoin and valproate coadministration on zonisamide PK but found no significant effect (page 325, “data not shown”) and retained only CBZ in the final model, consistent with the AED-specific decomposition rationale.

CONMED_CLB (canonical for concomitant clobazam coadministration indicator)

• Description: 1 = subject is taking clobazam (CLB) as a concomitant antiepileptic drug at the PK observation, 0 = no concomitant clobazam. Clobazam is a 1,5-benzodiazepine used as adjunctive therapy in epilepsy; unlike the enzyme-inducing AEDs it is not an inducer, and where a source paper resolves a separate clobazam effect the reported direction has been a decrease in the apparent oral clearance of the co-administered drug. Time-varying when clobazam starts / stops within the observation window; time-fixed when the source paper analyses chronic-maintenance cohorts whose AED therapy is stable across the analysis window.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant clobazam).
• Source aliases:
  • CLB – used in Ogusu_2014_valproic_acid.R (Ogusu 2014 Eq. 7 exponent CLB, defined in the text immediately below the equation as “CLB = 1 if CLB was co-administered, and was otherwise 0”).
• Example models: Ogusu_2014_valproic_acid.R (power-of-binary multiplicative factor on apparent oral clearance: cl <- ... * e_clb_cl^CONMED_CLB with e_clb_cl = 0.906; concomitant clobazam lowers valproic acid CL/F by 9.4% relative to the no-clobazam reference, Ogusu 2014 Table 2 / Eq. 7, bootstrap 95% CI 0.852-0.970), Nakashima_2015_valproic_acid.R (power-form effect on apparent oral clearance of valproic acid: cl *= 0.906^CONMED_CLB, i.e. -9.4% CL/F relative to the no-clobazam reference; Ogusu 2014 Table 2), Nakashima_2015_valproic_acid_le18y.R, Nakashima_2015_valproic_acid_ge19y.R.
• Notes: Drug-specific CONMED_* indicator in the antiepileptic cluster alongside [[CONMED_CBZ]], [[CONMED_PB]], [[CONMED_PHT]], [[CONMED_OXC]], [[CONMED_VPA]] and [[CONMED_LCM]]. Register it separately rather than folding clobazam into [[CONMED_EIAED]] or [[CONMED_AED]]: clobazam is not an enzyme inducer, and in Ogusu 2014 its coefficient has the opposite sign to the CBZ / PB / PHT induction terms estimated in the same equation (0.906 versus 1.19 / 1.12 / 1.43), so pooling it into an inducer-class indicator would cancel real signal. The mechanism of the clearance reduction is not established; Ogusu 2014 Discussion cites a prior paediatric report of the same direction and states that “the underlying mechanism remains unknown”. As with the other AED indicators, the functional form of the effect is per-paper (power-of-binary in Ogusu 2014) and is not a property of this column – read each model’s model() body rather than assuming a fractional coefficient.

CONMED_COBICISTAT (canonical for concomitant cobicistat (CYP3A4 inhibitor / PK-booster) coadministration indicator)

• Description: 1 = subject is receiving concomitant cobicistat (COBI), a pharmacokinetic enhancer with no intrinsic antiviral activity used to boost co-administered antiretrovirals; 0 = no cobicistat. Cobicistat is a potent mechanism-based CYP3A4 inhibitor and also inhibits P-glycoprotein, BCRP, OATP1B1/1B3 and the renal transporters MATE1 and OCT2, so the indicator flags reduced CYP3A4-mediated clearance and/or increased bioavailability of the boosted co-administered drug. Time-fixed in fixed-dose-combination cohorts; time-varying when boosting starts / stops within the observation window.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no cobicistat coadministration).
• Source aliases:
  • Cobicistat – used in Thoueille_2023_tenofovir_reduced.R and its sibling models (paper covariate row “theta_Cobicistat”, Thoueille 2023 Table 2 / Table S2 / Table S3).
  • COBI, c – the standard antiretroviral-literature abbreviations; c appears as the trailing boosting token in fixed-dose-combination names (e.g., “elvitegravir/c”, “darunavir/c”).
• Example models: Thoueille_2023_tenofovir_reduced.R, Thoueille_2023_tenofovir_full.R (additive fractional effect on tenofovir alafenamide relative bioavailability: f(depot) <- exp(lfdepot + etalfdepot) * (1 + e_cobi_fdepot * CONMED_COBICISTAT), with e_cobi_fdepot = 1.11 in the reduced model and 1.15 in the full model, i.e. F rises from 1 to ~2.1-2.2), Westra_2025_osimertinib.R, Westra_2025_osimertinib_brownbase.R (multiplicative power-form effect on osimertinib apparent clearance: cl <- ... * e_cobi_cl^CONMED_COBICISTAT, with e_cobi_cl = 0.704 in the de novo model and 0.678 in the Brown-2017-based companion, i.e. a 29.6 / 32.2 percent CL/F reduction; cobicistat is used here as a deliberate cost-minimisation booster for an oncology drug rather than for an antiretroviral).
• Notes: Follows the antiretroviral three/four-letter abbreviation convention already used by [[CONMED_RTV]] (ritonavir), [[CONMED_EFV]] (efavirenz), [[CONMED_NVP]] (nevirapine) and [[CONMED_LPV]] (lopinavir). Cobicistat and ritonavir are the two clinically used PK boosters and both are potent CYP3A4 inhibitors, but they are not interchangeable columns: ritonavir has intrinsic protease-inhibitor activity and a broader induction / inhibition profile (including CYP2D6 and several UGTs), so a paper that boosts with cobicistat must use this canonical rather than overloading [[CONMED_RTV]]. The two founding use cases sit at opposite ends of the confounding question and are worth contrasting: Westra 2025 estimates the coefficient from a within-subject sequential design (each patient sampled at steady state first without and then with cobicistat), so its value is a clean drug-drug-interaction estimate, whereas the Thoueille 2023 coefficient absorbs a dose difference as well. Per-model covariateData[[CONMED_COBICISTAT]]$notes must state whether the indicator is confounded with dose level: in Thoueille 2023 every cobicistat-treated subject received tenofovir alafenamide 10 mg and every other subject 25 mg (cobicistat was forced into the model specifically to absorb that dose difference into relative bioavailability), so the coefficient there is a combined dose-normalization-plus-boosting effect rather than a pure drug-drug-interaction effect. Papers that test cobicistat as a P-glycoprotein perpetrator alongside a separate P-gp-inhibitor class flag should keep the two columns distinct – Thoueille 2023 deliberately excluded cobicistat from its P-gp-inhibitor covariate analysis to avoid double-counting.

CONMED_COLISTIMETHATE (canonical for concomitant intravenous colistimethate sodium coadministration indicator)

• Description: 1 = the subject was receiving concomitant intravenous colistimethate sodium (colistin methanesulfonate, the inactive prodrug of colistin / polymyxin E) during the observation window; 0 = no concomitant colistimethate. Colistimethate is a reserve polymyxin used against multi-resistant Gram-negative organisms and is itself renally eliminated and nephrotoxic, so the indicator flags a comedication whose presence marks both a sicker patient and a competing renal-handling load on the modelled drug. Time-varying when the polymyxin course starts or stops within the observation window; time-fixed when the source assigns it per patient over a short inpatient course.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant colistimethate).
• Source aliases:
  • COL – used in Rancic_2024_meropenem.R (Rancic 2024 Table 3 row “Effect of COL”, theta 15). The same paper’s prose alternates between “colistimethate” (Abstract, Results equation) and “colistin” (Table 1 row, Table 3 footnote) for one indicator; they are the prodrug and the active moiety respectively, and only one column was collected.
• Example models: Rancic_2024_meropenem.R (additive shift +1.28 L/h on meropenem clearance in critically ill adults; Rancic 2024 Table 3, 7 of 101 patients exposed).
• Notes: Member of the CONMED_<INN> binary concomitant-medication family, and the direct sibling of CONMED_VANCOMYCIN: both are reserve antibiotics that Rancic 2024 retained as additive shifts on meropenem clearance, and the paper’s own reading is that neither is a mechanistic drug-drug interaction – they mark the augmented renal clearance of patients sick enough to need combination reserve therapy (Discussion). The two indicators are not mutually exclusive: 3 of the 101 patients received both, so a dataset must be able to set both to 1 and let the additive shifts sum. Named for the administered prodrug rather than for colistin because that is what is given intravenously and what a prescribing record captures; a future paper that instead flags oral or inhaled colistimethate, or that separates polymyxin B, should register its own name rather than overloading this one (Ma_2026_colistinSulfate.R and Jacobs_2016_colistin.R model colistin as the analyte, not as a comedication, and a model must never use this indicator for its own analyte). Distinct from the class-level CONMED_ABX composite, which is the right column only when a paper does not resolve individual agents.

CONMED_CSA (canonical for concomitant cyclosporine (CsA) coadministration indicator)

• Description: 1 = subject is receiving cyclosporine (CsA) as the concomitant calcineurin inhibitor at the PK observation; 0 = no concomitant cyclosporine (in mycophenolate mofetil studies the typical alternative comparator is tacrolimus). Cyclosporine inhibits the MRP2 (ABCC2) biliary efflux transporter, which decreases biliary excretion of MPAG into the gut lumen and therefore decreases enterohepatic recirculation (EHC) of mycophenolic acid; cyclosporine also displaces mycophenolic acid from plasma protein binding sites. Time-varying when cyclosporine starts / stops within the observation window; time-fixed when the source paper analyses subjects randomized to one CNI regimen for the entire study window (e.g., de Winter 2009).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant cyclosporine; in MPA studies the reference is typically a tacrolimus-based regimen).
• Source aliases:
  • CsA – used in de Winter 2009 (binary indicator on the MPAG-to-gallbladder transport rate constant k57 to encode cyclosporine inhibition of MRP2-mediated biliary efflux); the canonical column is CONMED_CSA with the same value semantics.
• Example models: deWinter_2009_mycophenolic_acid.R (multiplicative power-form effect on the fMPAG-to-gallbladder rate constant k57: k57 <- exp(lk57) * e_csa_k57^CONMED_CSA with e_csa_k57 = 0.002, so cyclosporine cotreatment reduces k57 from 0.0796 1/h to 0.000159 1/h, suppressing EHC by ~99.8% relative to the tacrolimus reference; Table 2 / Eq. 9 of de Winter 2009).
• Notes: Distinct from a generic concomitant-immunosuppressant indicator: this canonical is specifically cyclosporine vs another CNI / no CNI. In renal-transplant popPK studies that randomise patients between cyclosporine and tacrolimus immunosuppression (the de Winter 2009 design), CONMED_CSA = 1 and CONMED_TAC (if registered later for tacrolimus-as-perpetrator effects) would be mutually exclusive at the patient level. Cyclosporine and tacrolimus differ in their effect on MPA pharmacokinetics: cyclosporine inhibits MRP2 (decreasing EHC of MPAG -> MPA) while tacrolimus does not, so co-administered MPA exposure is typically lower under cyclosporine than under tacrolimus for the same MMF dose. Data assemblers can derive CONMED_CSA = as.integer(cni_drug == "cyclosporine") from a raw CNI_DRUG text column.

CONMED_CSA_DOSE (canonical for total daily dose of co-administered ciclosporin (cyclosporine A))

• Description: Total daily dose of concomitant ciclosporin (cyclosporine A) at the current observation time. The continuous-dose companion to the binary CONMED_CSA: use CONMED_CSA when the source paper models cyclosporine cotreatment as a yes/no contrast (typically against a tacrolimus comparator arm) and CONMED_CSA_DOSE when it models a graded dose-response. In mycophenolic acid (MPA) popPK, ciclosporin inhibits the MRP2 (ABCC2) biliary efflux transporter, reducing biliary excretion of MPAG into the gut lumen and therefore suppressing enterohepatic recirculation of MPA; a higher ciclosporin dose consequently raises apparent MPA clearance.
• Units: mg/day
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as a centered power term (CONMED_CSA_DOSE / ref)^exponent. Reference value observed: 450 mg/day (van Hest 2005 as transcribed in the Maizaud 2025 S1 File [MAIN] block, pow(CIC/450, 0.31); the mrgsolve simulation default for the column is 300 mg/day). 0 is not a usable value in the power form – unlike CONMED_NAL_DOSE / CONMED_BUP_DOSE, whose Emax form vanishes at 0, a centered power term is singular at zero. A model needing to represent ciclosporin-free subjects must pair this column with the binary CONMED_CSA and gate the term.
• Source aliases:
  • CIC – Maizaud 2025 S1 File mrgsolve [PARAM] @annotated @covariates column name for the van Hest 2005 model (“Ciclosporine daily dose (mg)”).
  • CSA_DOSE / CYA_DOSE – common NONMEM column-name forms.
• Example models: vanHest_2005_mycophenolic_acid.R (centered power effect (CONMED_CSA_DOSE / 450)^0.31 on MPA apparent clearance CL/F; every subject in the van Hest 2005 cohort received ciclosporin, so the companion binary CONMED_CSA is 1 throughout and is not carried separately), Checchio_2017_psoriasis_pasi_landmark_mbma.R (arm-level STEP indicator in a psoriasis PASI-responder meta-analysis: ciclosporin is one of the three traditional oral agents that Checchio 2017 models as a single-step offset because per-patient titration left insufficient literature dose-response information, so only CONMED_CSA_DOSE > 0 is read. Note the units differ from the founding model – Checchio 2017 records the arm’s 2.5-5 mg/kg/day clinical range rather than a milligram daily total – which is harmless only because the column acts as a step there; a value must never be carried between the two models).
• Notes: Well-formed member of the CONMED_<drug>_DOSE daily-dose family (siblings CONMED_NAL_DOSE, CONMED_BUP_DOSE, CONMED_ATV_DOSE, CONMED_ATORVASTATIN_DOSE); per that family’s Notes, the CONMED_<drug>_DOSE shape replaces the earlier DOSE_<drug> / DOSE_<drug>_<unit> names. Specific scope because the exponent is tied to the van Hest 2005 MPA analysis; promote to general if a second paper ratifies a graded ciclosporin-dose effect in an independent programme. Distinct from CONMED_CSA, the binary cotreatment indicator, which remains the right canonical for cyclosporine-vs-tacrolimus contrasts such as deWinter_2009_mycophenolic_acid.R.

CONMED_CHEMO (canonical for anti-PD-(L)1 mAb + chemotherapy combination indicator)

• Description: 1 = subject is receiving an anti-PD-(L)1 monoclonal antibody in combination with platinum-based chemotherapy (gemcitabine + cisplatin, pemetrexed + cisplatin, paclitaxel + carboplatin, or platinum-doublet); 0 = otherwise. Encodes chemotherapy coadministration as a study-design covariate on the antibody’s CL.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no chemotherapy coadministration – monotherapy or, where applicable, a non-chemotherapy combination such as anti-PD-1 + anti-CTLA-4).
• Source aliases:
  • CHEMO – used in Zhang_2019_nivolumab.R.
  • MONOTR – used in Kuchimanchi_2024_dostarlimab.R (the paper’s structural-equation indicator for monotherapy; the canonical column carries the inverse value, i.e. CONMED_CHEMO = 1 - MONOTR, so the canonical column is 1 when the patient is on combo-chemotherapy).
  • comb (three-level combination-therapy column; levels 1 and 2 both carry a chemotherapy backbone in Abegesah 2025) – decompose into CONMED_CHEMO = as.integer(comb >= 1).
• Example models: Zhang_2019_nivolumab.R (exponential effect on baseline CL: exp(-0.104) ~= 0.90 fold, i.e. ~9.7% lower CL relative to monotherapy), Kuchimanchi_2024_dostarlimab.R (multiplicative effect on baseline CL: 1 - 0.0779 = 0.922, i.e. 7.79% lower CL on dostarlimab + carboplatin/paclitaxel relative to dostarlimab monotherapy), Abegesah_2025_durvalumab.R (multiplicative effect on baseline CL of 1 - 0.163 = 0.837 for durvalumab + gemcitabine/cisplatin relative to durvalumab monotherapy; paired with CONMED_TREMELIMUMAB so that the chemotherapy-without-tremelimumab stratum is CONMED_CHEMO * (1 - CONMED_TREMELIMUMAB)), Zierhut_2016_hcc_antiangiogenic_os_mbma.R (the first use OUTSIDE the anti-PD-(L)1 setting and the first at the study-arm rather than the patient level: an additive log(0.960) shift on ln(median overall survival) in an advanced-hepatocellular-carcinoma antiangiogenic-therapy MBMA, where the column flags the 13 of 68 arms that gave concomitant chemotherapy alongside the antiangiogenic or placebo treatment), Xiang_2021_bemarituzumab.R (second use outside the anti-PD-(L)1 setting: an anti-FGFR2b antibody given with mFOLFOX6 (5-fluorouracil, leucovorin, oxaliplatin); exponential effect on linear CL, exp(-0.200) = 0.819, i.e. 18.1% lower CL than monotherapy. The paper’s indicator is ‘combotherapy/study’ because every combination patient came from one study (FIGHT), so the effect is study-confounded; source column Therapy), Shulgin_2020_ici_trae34_covariate_mbma.R (study-arm level in an immune checkpoint inhibitor adverse-event MBMA: +1.1 on the logit of the grade 3/4 trAE rate, applied only through the product with TRT_ANTIPD1, so chemotherapy added to anti-CTLA-4 monotherapy has no effect), Zhang_2022_ici_irae_mbma.R, Zhang_2022_ici_irae_grade3_mbma.R (study-arm level in a non-small cell lung cancer irAE MBMA; the source fits ONE coefficient for chemotherapy OR targeted therapy, so the models combine this column with CONMED_NONCHEMO_OTHER as max(CONMED_CHEMO, CONMED_NONCHEMO_OTHER): +0.9093 on the any-grade and +0.8790 on the grade >= 3 irAE logit).
• Notes: Promoted from specific to general scope on 2026-04-27 after the Kuchimanchi 2024 dostarlimab + carboplatin/paclitaxel analysis ratified the same pooling convention (any chemotherapy backbone collapsed into a single binary indicator). The two papers use different functional forms for the effect on CL – Zhang 2019 uses exp(theta * CONMED_CHEMO) (exponential) and Kuchimanchi 2024 uses (1 + theta * CONMED_CHEMO) (multiplicative); these are different parameterisations of the same underlying study-design indicator and the canonical column meaning is unchanged. Document the per-model functional form in covariateData[[CONMED_CHEMO]]$notes.

CONMED_LOCOREGIONAL (canonical for concomitant locoregional (liver-directed) therapy indicator)

• Description: 1 = a concomitant locoregional, liver-directed procedure was administered alongside the systemic therapy under study; 0 = systemic therapy alone. Pools the intra-arterial, radioembolic and ablative modalities used in hepatocellular carcinoma – transarterial chemoembolization (TACE), selective internal radiation therapy (SIR-Spheres / Y-90 radioembolization) and cryoablation – under one study-design indicator, because trials report them interchangeably as “locoregional therapy” and meta-analytic datasets rarely resolve which was used.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no concomitant locoregional procedure; systemic therapy alone).
• Source aliases:
  • LOC / Concomitant LOC – Zierhut 2016 Results “Final model” equation, Table 1 and Table 3.
• Example models: Zierhut_2016_hcc_antiangiogenic_os_mbma.R (additive log(1.42) shift on the ln(median overall survival) response of an advanced-hepatocellular-carcinoma antiangiogenic-therapy MBMA: an arm receiving concomitant locoregional therapy has 1.42-fold the median overall survival of one that does not. The single largest covariate effect in that model, but estimated from only 7 of 68 arms).
• Notes: MBMA study-arm-level study-design covariate. CONCOMITANT only – an arm whose patients had locoregional therapy before enrolment is not flagged here, a distinction that matters in advanced hepatocellular carcinoma where prior TACE is near-universal; register a separate PRIOR_LOCOREGIONAL if a future source estimates that effect. Modality-class rather than per-agent member of the CONMED_ family, following CONMED_CHEMO and CONMED_PLATIN, which likewise name a class of concomitant therapy instead of a single INN; the per-agent CONMED_<INN> columns remain the right choice when a source resolves the actual drug. Scope: specific until a second source adopts the same pooling; promote to general at that point. Where a source does separate TACE from radioembolization from ablation with distinct coefficients, register those as their own indicators rather than overloading this column.

CONMED_DABIGATRAN (canonical for chronic dabigatran coadministration indicator)

• Description: 1 = subject was on chronic oral dabigatran etexilate (direct thrombin inhibitor) at presentation, 0 = not. Time-fixed at the analysis baseline in periprocedural settings, where the relevant fact is what the patient had been taking before the procedure rather than an ongoing coadministration; time-varying when start / stop events are captured. Because dabigatran is routinely withheld for one to two elimination half-lives before an invasive procedure, a model using this column must document the withholding interval in covariateData[[CONMED_DABIGATRAN]]$notes – the indicator then denotes recent-and-stopped rather than ongoing exposure, and the two have materially different pharmacology.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no chronic dabigatran).
• Source aliases:
  • PRADAX – used in Edrich_2015_heparin.R (the Edrich 2015 supplement $PK group indicator, named for the Pradaxa trade name; GROUPMOD = NEITHER*1 + COUMLO*THETA(3) + COUMHI*THETA(4) + PRADAX*THETA(5)).
  • DABI, DABIGATRAN, group D – other spellings in periprocedural-anticoagulation data extracts and prose.
• Example models: Edrich_2015_heparin.R (founding example; multiplicative effect on the heparin-sensitivity coefficient k_ACT, multiplier 1.08 relative to no chronic anticoagulant – i.e. essentially no effect, which is the paper’s conclusion: dabigatran withheld about 27 h (two half-lives) leaves heparin sensitivity indistinguishable from no anticoagulation).
• Notes: Well-formed member of the auto-approved CONMED_<INN> family. Distinct from CONMED_DABIGATRAN_DOSE, which is a per-arm total daily dose in a VTE-prophylaxis dose-response meta-analysis, not a subject-level indicator; a model that resolves the actual dose should use that column instead of this one. Mutually exclusive with CONMED_WARFARIN in sources that stratify patients by a single chronic anticoagulant; where both are zero the subject is in the no-chronic-anticoagulant reference group, so a model must not additionally register a CONMED_NONE indicator – the reference is recovered as 1 - CONMED_WARFARIN - CONMED_DABIGATRAN.

CONMED_WARFARIN (canonical for chronic warfarin coadministration indicator)

• Description: 1 = subject was on chronic oral warfarin (vitamin-K-antagonist anticoagulant) at presentation, 0 = not. Time-fixed at the analysis baseline in periprocedural settings; time-varying when start / stop events are captured. Warfarin’s anticoagulant effect persists for days after the last dose because it acts by depleting the carboxylated vitamin-K-dependent clotting factors rather than by inhibiting a target directly, so unlike a direct oral anticoagulant this indicator remains informative even when the drug has been withheld. Pair it with INR_BASE when a source stratifies warfarin patients by degree of anticoagulation achieved.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no chronic warfarin).
• Source aliases:
  • COUMLO / COUMHI – used in Edrich_2015_heparin.R, where the Edrich 2015 supplement $PK block carries TWO warfarin indicators (named for the Coumadin trade name) splitting the warfarin arm at the last pre-procedural INR: COUMLO is warfarin with an unintentionally low INR < 2, COUMHI is warfarin with INR >= 2 within the last 3 days. Reconstruct both from CONMED_WARFARIN together with INR_BASE rather than registering two separate indicator columns, so the INR cut point stays visible in model() where a reader can check it against the source.
  • WARF, WARFARIN, VKA, group W – other spellings in periprocedural-anticoagulation data extracts and prose. VKA may pool warfarin with acenocoumarol or phenprocoumon; register a separate pooled-class indicator if a source estimates one effect across the class.
• Example models: Edrich_2015_heparin.R (founding example; multiplicative effect on the heparin-sensitivity coefficient k_ACT, multiplier 1.92 for INR < 2 and 2.50 for INR >= 2 relative to no chronic anticoagulant – i.e. warfarin patients are roughly twice as heparin-sensitive, which is that paper’s central finding).
• Notes: Well-formed member of the auto-approved CONMED_<INN> family. Distinct from CONMED_WARFARIN_DOSE, which is a per-arm total daily dose in a VTE-prophylaxis dose-response meta-analysis, not a subject-level indicator. Mutually exclusive with CONMED_DABIGATRAN in sources that stratify patients by a single chronic anticoagulant; see that entry’s note on recovering the reference group. Do not repurpose this column to mean “warfarin is the drug being modelled” – in a warfarin popPK or K-PD model warfarin is the analyte, not a comedication, and the register’s warfarin-pharmacology covariates (INR_BASE, NPT_BASE, VKORC1_1639G_COUNT, the SNP_CYP2C9_* family) are what such a model needs.

CONMED_DECITABINE (canonical for concomitant decitabine (hypomethylating-agent) chemotherapy backbone indicator)

• Description: 1 = subject is receiving decitabine as the hypomethylating-agent (HMA) chemotherapy backbone during the study window; 0 = subject is on a different chemotherapy backbone (e.g. low-dose cytarabine) or no HMA at all. Used as a treatment-arm indicator in head-to-head trials that compare a hedgehog-pathway inhibitor (or other targeted agent) layered on a decitabine backbone vs an alternative backbone in older adults with AML or high-risk MDS who are ineligible for intensive chemotherapy.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (subject is not on a decitabine backbone; typically paired with CONMED_GLASDEGIB = 1 in the only source (Lin 2020), where the decitabine arm carries both indicators = 1 and the LDAC arms have CONMED_DECITABINE = 0).
• Source aliases:
  • Derived from the protocol-defined treatment arm in Lin_2020_glasdegib_decitabine.R (BRIGHT AML 1003 Phase 1b Arm B: glasdegib + decitabine, n = 7).
• Example models: Lin_2020_glasdegib_decitabine.R (paired with CONMED_GLASDEGIB to identify the three trial arms; multiplicative hazard reduction 1 - 0.618 * (CONMED_GLASDEGIB * CONMED_DECITABINE) on the exponential overall-survival hazard).
• Notes: Specific scope because the only source is Lin 2020 BRIGHT AML 1003 Phase 1b Arm B (5 AML + 2 MDS, n = 7). The reported CI of the glasdegib + decitabine hazard reduction is wide (-95.0% to -28.6%) due to the small sample. Auto-approved member of the CONMED_<INN> family.

CONMED_DHA (canonical for the dihydroartemisinin parasite-clearance coverage window)

• Description: 1 = the record falls inside a window during which coadministered dihydroartemisinin (DHA, the artemisinin component of an artemisinin-based combination therapy) is deemed to eliminate any blood-stage Plasmodium infection that would otherwise become detectable; 0 = outside such a window. Time-varying at the record level, not a subject-level attribute. The indicator exists because artemisinins kill roughly 10^4 parasites per 48 h asexual cycle, so a standard three-day course clears not only the infections present when it starts but also those emerging from the liver during the preceding several days, before they could reach the microscopy detection limit. In a time-to-event model of malaria reinfection the window therefore sets the hazard of a detected infection to zero. The window is anchored to the treatment round’s dose times and extends backwards in time from the first dose, so it cannot be derived inside model() from tad() or any other causal time function – it must be supplied as a column in the event table, computed from the planned round start times.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (outside the dihydroartemisinin coverage window; the drug-free state in which the baseline hazard and the partner drug’s concentration-effect term act alone).
• Source aliases:
  • DHA_EFF – used in Chotsiri_2019_piperaquine.R with reversed polarity: Chotsiri 2019 Eq. 4 defines DHA_EFF = 0 inside the window and 1 outside, because it is written as a multiplicative factor on the hazard. The canonical follows the CONMED_<INN> convention that 1 means “the comedication is acting”, so CONMED_DHA = 1 - DHA_EFF and the model recovers the paper’s factor as dhaeff = 1 - CONMED_DHA.
• Example models: Chotsiri_2019_piperaquine.R (seasonal malaria chemoprevention with dihydroartemisinin-piperaquine; the window runs from 6 days before the first dose of each three-day round to 24 h after its last dose, i.e. t = -144 h to t = +72 h relative to that round’s first dose, and multiplies the baseline hazard of new P. falciparum infection by 1 - CONMED_DHA).
• Notes: Auto-approved member of the CONMED_<INN> family. Record the paper’s own window definition in covariateData[[CONMED_DHA]]$notes per model, because the lead-in length is derived from an assumed parasite multiplication rate and liver-emergence biomass and so varies between papers (Chotsiri 2019 assumes ~10-fold per 48 h from ~10^5 parasites against a ~10^8 microscopy detection limit, giving 6 days). Distinct from a plain “received an artemisinin combination therapy” subject-level flag: this is an interval attached to each dosing round. Use a sibling CONMED_<INN> for the other artemisinin derivatives (artesunate, artemether) rather than widening this entry, since their elimination and parasite-reduction-ratio profiles differ.

CONMED_DIURETIC (canonical for concomitant diuretic indicator; class composition is paper-specific)

• Description: 1 = subject is coadministered a diuretic during the study window (or, for per-time-point datasets, at the current time), 0 = no concomitant diuretic. Used in popPK / popPK-PD analyses of renally cleared substrates because diuretic-driven volume contraction and shifts in renal tubular handling can change apparent clearance, and because thiazide and loop diuretics are anti-uricosuric (raise serum urate) and so directly modify urate-related PD endpoints. The class membership pooled into the indicator is paper-specific and MUST be enumerated in covariateData[[CONMED_DIURETIC]]$notes per model – the observed range runs from any-class (loop + thiazide + potassium-sparing + osmotic + carbonic-anhydrase, Kleiber 2017) through thiazide + loop + spironolactone (Stocker 2012) to thiazide + loop only, deliberately excluding potassium-sparing agents (Wright 2016). Composite class indicator following the CONMED_STEROID / CONMED_AED / CONMED_AZOLE pattern (pooled class indicator) rather than the CONMED_<INN> per-drug pattern (CONMED_SPIRON, CONMED_FUROSEMIDE,…).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant diuretic of the paper-specific class set). Effect is typically encoded multiplicatively as Theta^CONMED_DIURETIC (Kleiber 2017 Eq 6 / Wright 2016 power form) or as a linear deviation 1 + e_conmed_diuretic_<param> * CONMED_DIURETIC (Stocker 2012 form).
• Source aliases:
  • DIURETIC – used in Kleiber_2017_clonidine.R (paper’s pooled indicator combining furosemide intermittent + furosemide infusion + spironolactone + bumetanide; source column DIURETIC in the NONMEM dataset).
  • DIUR – used in Stocker_2012_oxypurinol.R (Stocker 2012 dataset column; pools thiazide + loop + spironolactone; n = 72 of 155 gouty patients).
  • diuretic – used in Wright_2016_allopurinol.R and Wright_2013_allopurinol.R (Wright narrative term; pools thiazide + loop diuretics only, excludes spironolactone / amiloride; n = 44 of 133 gouty patients in Wright 2016).
• Example models: Kleiber_2017_clonidine.R (any-class; multiplicative effect on CL: diur_cl = 0.659^CONMED_DIURETIC; per Table 3 / Table 4 CL is reduced by 34.1% when any diuretic is active), Stocker_2012_oxypurinol.R (thiazide + loop + potassium-sparing; multiplicative linear-deviation effect on apparent CL/Fm: cl *= (1 + (-0.294) * CONMED_DIURETIC), i.e. -29.4% apparent oxypurinol clearance; Stocker 2012 Table 3 theta7), Wright_2016_allopurinol.R (thiazide + loop only, no potassium-sparing; multiplicative power-form effect on both CL/F_oxy and baseline urate U0: cl *= 0.740^CONMED_DIURETIC (-26%) and rbase *= 1.14^CONMED_DIURETIC (+14%); Wright 2016 Table 3 thetadiuretic and thetaE0_diuretic), Wright_2013_allopurinol.R (fractional effect 0.61 on renal CL_oxy via e_conmed_diuretic_cl_oxy_renal^CONMED_DIURETIC; Wright 2013 Table 2).
• Notes: Composite indicator with paper-specific class membership. Users simulating across models that share this column but differ in class membership MUST populate it according to each paper’s own definition – Wright 2016 deliberately excludes potassium-sparing diuretics because they tend to be uricosuric (lower serum urate) and therefore have an opposite-direction PD effect from the anti-uricosuric thiazide / loop agents, whereas Stocker 2012 and Kleiber 2017 pool them in. Distinct from CONMED_SPIRON (spironolactone-only per-drug indicator, used in Yukawa_1996_digoxin.R), a member of the CONMED_<INN> family; the two canonicals coexist when a paper distinguishes specific drugs. When a paper requires class-resolved encoding (separate thiazide / loop / K-sparing indicators), register sibling canonicals (e.g. CONMED_THIAZIDE, CONMED_LOOP_DIUR) rather than overloading this one. CONMED_DIUR is retired: do not reintroduce it.

CONMED_FUROSEMIDE (canonical for concomitant furosemide (loop diuretic) coadministration indicator)

• Description: 1 = subject is coadministered furosemide during the study window (or, for per-time-point datasets, at the current time), 0 = no concomitant furosemide. Screened in popPK analyses of renally cleared substrates and of electrolytes, because loop-diuretic inhibition of the Na-K-2Cl cotransporter in the thick ascending limb both contracts plasma volume and abolishes the lumen-positive potential that drives paracellular magnesium and calcium reabsorption. In observational obstetric and critical-care cohorts the indicator is frequently a marker of the disease state that prompted the prescription (volume overload, oliguria, declining renal function) rather than of a direct drug interaction, and its estimated effect will then have the sign of the underlying disease rather than of the diuretic’s own pharmacology; state which interpretation the source paper supports in covariateData[[CONMED_FUROSEMIDE]]$notes.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant furosemide). Effect is encoded either as a fractional-change term 1 + e_conmed_furosemide_<param> * CONMED_FUROSEMIDE (Deng 2024 form) or multiplicatively as Theta^CONMED_FUROSEMIDE.
• Source aliases:
  • furosemide – Deng 2024 narrative and equation term; used in Deng_2024_magnesiumSulfate.R.
  • frusemide – the former British Approved Name for the same drug; treat as identical.
• Example models: Deng_2024_magnesiumSulfate.R (16 of 51 Chinese women with preeclampsia, 31.37%; the sole categorical covariate retained, and unusually it enters BOTH disposition parameters – cl *= (1 + (-0.16) * CONMED_FUROSEMIDE) and vc *= (1 + (-0.25) * CONMED_FUROSEMIDE), i.e. -16% clearance and -25% volume, both raising serum magnesium. Deng 2024 presents this as the first report of a furosemide effect on magnesium popPK and attributes it to furosemide being prescribed for the volume overload and renal impairment of progressing preeclampsia, so the indicator is confounded with disease severity by construction and should not be transported to a population in which furosemide is used for another reason).
• Notes: Named-drug member of the CONMED_<INN> family, registered separately from the class-level CONMED_DIURETIC per the standing instruction in that entry’s Notes (“When a paper requires class-resolved encoding … register sibling canonicals rather than overloading this one”) and in CONMED_MANNITOL’s Notes (“Register CONMED_FUROSEMIDE and other named diuretics separately”). Use CONMED_FUROSEMIDE when the source paper names furosemide specifically; use CONMED_DIURETIC when it pools a class and enumerates the membership. The two coexist, exactly as CONMED_SPIRON coexists with CONMED_DIURETIC; a paper that resolves furosemide out of a broader diuretic indicator may legitimately carry both columns.

CONMED_PIMOBENDAN (canonical for concomitant pimobendan coadministration indicator)

• Description: 1 = subject is coadministered pimobendan during the study window (or, for per-time-point datasets, at the current time), 0 = no concomitant pimobendan. Pimobendan is an inodilator combining calcium sensitisation of cardiac troponin C with phosphodiesterase-3 inhibition, licensed for canine myxomatous mitral valve disease and dilated cardiomyopathy; it is the most frequently prescribed positive inotrope in small-animal cardiology and therefore a routine screening covariate in veterinary cardiac popPK analyses. Two distinct mechanisms motivate screening it: as a positive inotrope it raises cardiac output and so may change hepatic and splanchnic blood flow (altering absorption and first-pass extraction of a coadministered drug), and as a PDE3 inhibitor it shares the cyclic-nucleotide-phosphodiesterase target family with PDE5 inhibitors such as sildenafil, though the two isoenzymes are distinct. In observational veterinary cohorts the indicator is frequently a marker of the cardiac disease that prompted the prescription rather than of a direct interaction; state which interpretation the source paper supports in covariateData[[CONMED_PIMOBENDAN]]$notes.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant pimobendan).
• Source aliases:
  • pimobendan – Yata 2026 Methods 2.6 and Results narrative term (listed there as one of two binary categorical covariates screened, alongside concurrent diuretic administration).
• Example models: Yata_2026_sildenafil_dog.R.
• Notes: Named-drug member of the CONMED_<INN> family, following the same registration pattern as CONMED_FUROSEMIDE and CONMED_SPIRON. Distinct from the class-level CONMED_DIURETIC: pimobendan is not a diuretic, and a paper that pools cardiac co-medications into a single indicator should register a separate class canonical rather than overloading either. Veterinary cardiac cohorts commonly carry both CONMED_PIMOBENDAN and CONMED_DIURETIC because the two are prescribed for overlapping but non-identical indications.

CONMED_DOXORUBICIN (canonical for concomitant doxorubicin (anthracycline) chemotherapy backbone indicator)

• Description: 1 = subject is receiving (or has received during the relevant observation window) doxorubicin as the anthracycline component of an adjuvant or neoadjuvant chemotherapy regimen; 0 = subject is receiving a different anthracycline (or no anthracycline at all). Used in cardiac-biomarker PD models to differentiate doxorubicin-driven myocardial damage from epirubicin-driven damage (doxorubicin has approximately twice the per-mg cardiotoxic effect of epirubicin at clinically equivalent oncologic exposures).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (subject is not on doxorubicin). When CONMED_DOXORUBICIN and CONMED_EPIRUBICIN are used together as a pair of complementary indicators (as in de Vries Schultink 2018), both being 0 means no anthracycline.
• Source aliases:
  • ANTH_TYPE = 'doxorubicin' – used in deVriesSchultink_2018_anthracycline_troponinT.R (the source paper records anthracycline as a categorical with levels {‘doxorubicin’, ‘epirubicin’}; the canonical register splits this into a pair of binary indicators following the CONMED_<drug> precedent rather than the categorical ANTH_TYPE).
• Example models: deVriesSchultink_2018_anthracycline_troponinT.R (used jointly with CONMED_EPIRUBICIN to switch the K-PD SLOPE parameter for troponin T between doxorubicin (reference) and epirubicin; epirubicin = 0.524-fold of doxorubicin effect, per de Vries Schultink 2018 Table 2).
• Notes: Specific scope until a second cardiotoxicity or anthracycline-effect model legitimately reuses this indicator. Distinct from CONMED_PLDH (PEGylated liposomal doxorubicin) and from CONMED_AVD (the doxorubicin component of the AVD backbone in Hodgkin lymphoma); use those canonicals for product-formulation or regimen-backbone semantics. When the source dataset’s anthracycline column is categorical (ANTH_TYPE), derive the canonical indicators as CONMED_DOXORUBICIN = (ANTH_TYPE == 'doxorubicin'), CONMED_EPIRUBICIN = (ANTH_TYPE == 'epirubicin').

CONMED_EPIRUBICIN (canonical for concomitant epirubicin (anthracycline) chemotherapy backbone indicator)

• Description: 1 = subject is receiving (or has received during the relevant observation window) epirubicin as the anthracycline component of an adjuvant or neoadjuvant chemotherapy regimen; 0 = subject is receiving a different anthracycline (or no anthracycline at all). The pairing of CONMED_EPIRUBICIN with CONMED_DOXORUBICIN captures the cardiotoxicity-relevant choice within the anthracycline class; epirubicin is less cardiotoxic per mg than doxorubicin (lifetime cumulative dose threshold approximately 950 mg/m^2 vs 550 mg/m^2 for doxorubicin).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (subject is not on epirubicin). When CONMED_DOXORUBICIN and CONMED_EPIRUBICIN are used together (as in de Vries Schultink 2018), both being 0 means no anthracycline.
• Source aliases:
  • ANTH_TYPE = 'epirubicin' – used in deVriesSchultink_2018_anthracycline_troponinT.R. See the matching entry for CONMED_DOXORUBICIN.
• Example models: deVriesSchultink_2018_anthracycline_troponinT.R (used jointly with CONMED_DOXORUBICIN to switch the K-PD SLOPE parameter; epirubicin = 0.524-fold of the doxorubicin reference effect on troponin T).
• Notes: Specific scope until a second cardiotoxicity or anthracycline-effect model legitimately reuses this indicator. The two indicators (CONMED_DOXORUBICIN and CONMED_EPIRUBICIN) are not mutually exclusive in principle (a subject receiving both anthracyclines could have both set to 1), but in practice the de Vries Schultink 2018 cohort assigned exactly one anthracycline per subject.

CONMED_ECULIZUMAB (canonical for concurrent eculizumab (complement C5 inhibitor) coadministration indicator)

• Description: 1 = subject is receiving eculizumab concurrently with the modelled drug at the current time; 0 = the subject is not on eculizumab at the current time. Time-varying within a subject: it turns on and off as the co-treatment period starts and stops, which is what distinguishes it from the baseline-status sibling CONMED_ECULIZUMAB_BL.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no concurrent eculizumab; monotherapy with the modelled drug at the current time).
• Source aliases:
  • ECU – Crass 2024 ESM Table 1 LDH control stream $INPUT column.
  • C5I, ECULI, ECZ – plausible alternative NONMEM $INPUT forms; document the per-model source column in covariateData[[CONMED_ECULIZUMAB]]$source_name.
• Example models: Crass_2024_pegcetacoplan_ldh.R (additive shift of +0.783 on the logit maximal fractional LDH suppression, applied only within the baseline-eculizumab stratum per the source guard IF (BECU.EQ.1) TVEMAX = THETA(4) + THETA(8)*ECU; lifts the maximal LDH suppression from inverse-logit(-1.39) = 20.0% during pegcetacoplan monotherapy to inverse-logit(-0.607) = 35.3% during eculizumab + pegcetacoplan dual therapy, matching the 35.0% quoted in Crass 2024 Sect. 3.4.2. Non-zero only during the PEGASUS run-in and transition dual-therapy windows, Crass 2024 Table 1 footnote a).
• Notes: Eculizumab is a humanised monoclonal antibody against complement component C5 and the long-standing standard of care in paroxysmal nocturnal hemoglobinuria and atypical haemolytic uraemic syndrome; because it blocks terminal complement it profoundly suppresses intravascular haemolysis markers (notably serum lactate dehydrogenase), so it is a load-bearing covariate for any PD model whose endpoint is a haemolysis biomarker. Well-formed member of the auto-approved CONMED_<INN> family. Distinct from the disease-state canonical DIS_PNH, which records the indication rather than the comedication, and from CONMED_ECULIZUMAB_BL, which is the time-fixed baseline status.

CONMED_ECULIZUMAB_BL (canonical for eculizumab (complement C5 inhibitor) treatment status at baseline)

• Description: 1 = subject was receiving eculizumab at study entry (complement C5-inhibitor experienced); 0 = subject was eculizumab-naive at study entry. Time-fixed per subject – the value is frozen at baseline and does not change even if eculizumab is later withdrawn or co-administered, which is what distinguishes it from the time-varying sibling CONMED_ECULIZUMAB.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (eculizumab-naive at baseline).
• Source aliases:
  • BECU – Crass 2024 ESM Table 1 Hb and LDH control stream $INPUT column (“baseline eculizumab”).
  • PRIORC5I, BC5I, ECUBL – plausible alternative NONMEM $INPUT forms.
• Example models: Crass_2024_pegcetacoplan_ldh.R (selects BOTH the baseline-LDH stratum and the maximal-suppression stratum: eculizumab-naive patients have typical baseline exp(7.56) = 1920 U/L with 91.7% maximal suppression, and baseline-eculizumab patients exp(5.52) = 249 U/L with 20.0% maximal suppression, because terminal-complement blockade has already suppressed intravascular haemolysis; encoded as an indicator-weighted mixture of the two stratum parameters rather than an additive shift, because the source control stream uses IF(BECU.EQ.1) TVBLDH = THETA(2) full-replacement switches), Crass_2024_pegcetacoplan_hemoglobin.R (fractional effect on the maximal hemoglobin response emax * (1 + e_conmed_eculizumab_bl_emax * CONMED_ECULIZUMAB_BL) with the coefficient entered as (0 FIX) in the source $THETA block, which is why Crass 2024 Table 2 reports the same 51% Emax for both strata; retained with fixed(0) so the structural form and the fixed status stay visible).
• Notes: The _BL suffix follows the register’s established baseline-value convention (HGB_BL, TRAST_BL, PLAQUE_BL, INS_BL,…) applied to a member of the auto-approved CONMED_<INN> family. Register the pair whenever a source distinguishes prior/baseline C5-inhibitor exposure from concurrent co-treatment; conflating them loses a real modelling distinction (Crass 2024 estimates both, and they act on different parameters with different signs of practical consequence). Scope: specific because the reference category is defined by the source trial’s enrolment criteria.

CONMED_EFV (canonical for concomitant efavirenz indicator)

• Description: 1 = subject is receiving efavirenz (EFV)-based antiretroviral therapy as the third agent in a combination ART regimen; 0 = subject is on the comparator regimen specified by the source paper (typically a protease-inhibitor-based regimen such as standard lopinavir/ritonavir 4:1). Efavirenz is a CYP3A and UGT inducer, so the indicator is used to flag PXR-mediated induction of metabolic clearance for co-administered antiretrovirals.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-EFV reference regimen, paper-defined; e.g., LPV/r 4:1 in Tikiso 2021).
• Source aliases:
  • EFV – used in Tikiso_2021_abacavir.R (the dataset’s paper-defined indicator, 1 = on EFV-based ART, 0 = on standard LPV/r 4:1).
  • efavirenz – used in Courlet_2021_amlodipine.R (Courlet 2021 Table 2 ‘theta efavirenz’; 1 = on efavirenz, 0 = ARVs devoid of CYP3A4 interaction potential in people living with HIV).
  • efavirenz – used in Jaiswal_2025_dordaviprone.R (Jaiswal 2025 Figure 3 forest plot arm, efavirenz 600 mg once daily as the index moderate CYP3A4 inducer; relative CYP3A4 activity 2.08 back-solved from the published dordaviprone AUC ratio of 0.349). Here efavirenz is a probe perpetrator in a DDI simulation rather than a component of an ART regimen, so the reference is simply “no efavirenz”. Same usage in Huth_2019_siponimod.R (Huth 2019 Table 3 column ‘Moderate 3A4 inducer’).
• Example models: Tikiso_2021_abacavir.R (multiplicative effect on apparent oral clearance: cl *= (1 + 0.120 * CONMED_EFV); +12.0% relative to the LPV/r 4:1 reference in HIV-infected children on combination ART), Svensson_2013_bedaquiline.R (multiplicative factor on apparent CL_BDQ and CL_M2: cl_eff = cl_base * 2.07^CONMED_EFV and on apparent CL_M3: cl_eff = cl_base * 1.12^CONMED_EFV; healthy-volunteer DDI study with no-co-medication reference and 1-week onset lag from start of 600 mg once-nightly EFV), Hoglund_2015_lumefantrine.R (multiplicative linear-deviation effect on lumefantrine apparent CL/F: cl *= (1 + 0.726 * CONMED_EFV); +72.6% relative to the no-ART reference in HIV-infected Ugandan adults, Hoglund 2015 Table 2 attributing the effect to CYP3A4 induction), Hoglund_2015_artemether.R (multiplicative linear-deviation effect on artemether relative bioavailability F: fdepot_typ *= (1 + (-0.715) * CONMED_EFV); -71.5% relative to the no-ART reference, Hoglund 2015 Table 3), Kay_2020_lumefantrine.R (TWO independent linear-deviation effects, on lumefantrine apparent CL/F: cl *= (1 + 0.982 * CONMED_EFV) (+98.2%, CYP3A4 induction) and on the first-order absorption rate constant: ka *= (1 + 0.484 * CONMED_EFV) (+48.4%); relative to the no-ART reference in HIV-infected Ugandan children with malaria (ASTMH 2020 poster 2167 Table 1)), Kay_2022_lumefantrine.R (TWO independent linear-deviation effects: on apparent oral lumefantrine CL/F cl *= (1 + 0.982 * CONMED_EFV) (+98.2 %, CYP3A4 induction) and on the first-order absorption rate constant ka *= (1 + 0.484 * CONMED_EFV) (+48.4 %); HIV-uninfected children are the reference (CONMED_EFV = 0) in a pediatric Ugandan malaria cohort on six-dose Coartem Dispersible, Kay 2022 Table 2 theta_7 and theta_10), Bukkems_2021_raltegravir.R (linear additive effect on typical bioavailability F: f_typ *= (1 + e_efv_fdepot * CONMED_EFV) with e_efv_fdepot = -0.167; efavirenz-associated raltegravir F is 17% lower than the no-efavirenz reference, consistent with efavirenz-mediated UGT1A1 induction of raltegravir glucuronidation, Bukkems 2021 Table 2 ‘Factor change in F efavirenz co-administration’), Adeojo_2024_levonorgestrel.R (TWO independent linear-deviation effects in a meta-analytic fit to pooled trial mean profiles: on systemic clearance cl *= (1 + 0.7235 * CONMED_EFV) (CL 5.86 -> 10.1 L/h, +72.4%, CYP3A4 induction) and on oral bioavailability fdepot *= (1 + (-0.3632) * CONMED_EFV) (F_oral 0.837 -> 0.533, -36.3%, additionally reflecting gut-wall CYP3A4); reference is no efavirenz and no antiretroviral therapy, and every efavirenz-exposed arm used efavirenz 600 mg orally once daily; coefficients derived from the two absolute values published in Adeojo 2024 Supplementary Table S2), Jaiswal_2025_dordaviprone.R (efavirenz 600 mg once daily as a probe moderate CYP3A4 inducer in a PBPK drug-drug-interaction simulation, with a no-efavirenz reference; the effect enters through a single relative-CYP3A4-activity term of 2.08 that scales gut and hepatic first-pass extraction and systemic clearance together, back-solved from the published dordaviprone AUC ratio of 0.349), Huth_2019_siponimod.R (efavirenz as the index moderate CYP3A4 inducer in a Simcyp DDI-prediction paper; relative CYP3A4 activity 2.870 back-solved from the six published genotype AUC ratios of Huth 2019 Table 3, with relative CYP2C9 activity held at exactly 1 because the Methods state the efavirenz model was built as a pure CYP3A4 inducer), Li_2018_guanfacine.R (efavirenz as a probe moderate CYP3A4 inducer in a Simcyp drug-drug-interaction simulation, with a no-efavirenz reference; the flag alone carries the 400 mg once-daily effect and is paired with DOSE_EFAVIRENZ_MG to select the 600 mg level, the two relative CYP3A4 activities being 1.777 and 2.858 back-solved from the published guanfacine AUC ratios of 0.58 and 0.33), Ezuruike_2018_ethinylestradiol.R (efavirenz 600 mg once daily as the index moderate CYP3A4 inducer in a Simcyp drug-interaction analysis, with a no-efavirenz reference; the effect enters through a single relative-CYP3A4-activity term of 2.58 that scales gut-wall intrinsic clearance, hepatic intrinsic clearance and systemic clearance together, back-solved from the published ethinylestradiol steady-state AUC ratio of 0.614 in Ezuruike 2018 Table 4 [Table 3 in print]), Salinger_2019_pretomanid.R (multiplicative power-form effects on BOTH apparent oral clearance, e_efv_cl^CONMED_EFV with e_efv_cl = 2.17, and relative bioavailability, e_efv_fdepot^CONMED_EFV with e_efv_fdepot = 1.24. Combined with the model’s separate HIV_POS effects on the same two parameters, this reproduces Salinger 2019’s reported 46% reduction in pretomanid exposure for HIV-positive subjects taking efavirenz: Cavg falls from 2.4 to 1.3 ug/mL. Time-varying, because the source used on / off comedication histories rather than a baseline flag), Wang_2020_delamanid.R (multiplicative effect on apparent oral clearance: cl *= 1.35^CONMED_EFV; efavirenz given to 22 of 48 HIV-positive MDR-TB patients on antiretroviral therapy, reference = no efavirenz, i.e. HIV-negative patients and HIV-positive patients on other antiretrovirals), Courlet_2021_amlodipine.R (linear multiplicative effect on amlodipine apparent CL/F: cl *= (1 + 1.40 * CONMED_EFV), a 2.4-fold CL/F and a 59% lower AUC0-24 relative to people living with HIV on ARVs without CYP3A4 interaction potential; Courlet 2021 Table 2), Francis_2020_lumefantrine.R (fractional effect on lumefantrine apparent CL/F cl *= (1 + 0.899 * CONMED_EFV), +89.9% (47% lower AUC) in a pooled WWARN meta-analysis of 10 studies; reference pools no ART, nevirapine- and dolutegravir-based ART), Nassar_2022_midazolam.R (a SINGLE 400 mg oral efavirenz dose as a CYP3A activator in healthy volunteers, reference = pooled placebo; five separately estimated fractional changes in intravenous midazolam clearance by time since the efavirenz dose, cl *= (1 + CONMED_EFV * e_conmed_efv_cl_t<a>to<b>), +15.0% in (0, 2] h, +59.1% (2, 3], +56.0% (3, 4], +28.5% (4, 5], +33.3% (5, 6] h and no effect afterwards; Nassar 2022 Table 1, time windows located by T_CONMED)
• Notes: Specific scope because the comparator regimen (LPV/r 4:1 in Tikiso 2021) is paper-defined; future ART population-PK models that test EFV-vs-other contrasts should extend the example list when the comparator matches, or register a finer-grained sibling indicator otherwise. The reference category is paper-defined and is NOT always another ART regimen: Hoglund 2015, Kay 2020/2022 and Adeojo 2024 use a no-efavirenz (in Adeojo 2024, no-ART) reference rather than an active comparator, so read each example’s stated reference before transferring a coefficient.

CONMED_EFV_MD (canonical for multiple-dose (steady-state) efavirenz indicator, single-dose efavirenz reference)

• Description: 1 = the record was collected while the subject was at steady state on chronic efavirenz (typically 600 mg once daily for at least ~2 weeks, long enough for CYP3A / CYP2B6 induction to be established); 0 = the record was collected after a single dose of efavirenz. Unlike the parent canonical CONMED_EFV, both levels are efavirenz-exposed – the contrast is induction-established vs induction-not-yet-established, not efavirenz vs no efavirenz. Use this indicator for the sequential single-dose-then-chronic DDI design in which a probe drug is administered twice, once alongside the first efavirenz dose and once at efavirenz steady state, so that the perpetrator’s own autoinduction is held constant between arms and only the induction state differs. Typically occasion-level (a subject contributes both levels), not subject-level.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (single-dose efavirenz).
• Source aliases:
  • EFV – used in Collins_2025_midazolam.R with reversed polarity: the source dataset codes EFV = 0 for the multiple-dose condition and EFV = 1 for the single-dose condition, so CONMED_EFV_MD = 1 - EFV. Collins 2025 Results states this explicitly (‘In the NONMEM control stream, EFV = 0 represents the multiple-dose efavirenz condition, and EFV = 1 represents the single-dose condition… The reference (baseline) state is thus the single-dose efavirenz condition’), and the deposited control stream (Supporting Information File S5) confirms it with L0 = 0 / IF(EFV.EQ.0) L0 = 1;multiple dose efavirenz.
• Example models: Collins_2025_midazolam.R (THREE independent effects from one indicator, all taken from the deposited final control stream: a linear-deviation effect on midazolam clearance cl *= (1 + 0.652 * CONMED_EFV_MD) (+65.2%, hepatic CYP3A induction), a linear-deviation effect on the first-order absorption rate constant ka *= (1 + 0.255 * CONMED_EFV_MD) (+25.5%), and a wholesale switch of oral bioavailability from a fixed 0.5 on the single-dose occasion to a logit-scale estimated expit(-0.511) = 0.375 on the multiple-dose occasion. The bioavailability term is the only parameter in that model carrying between-subject variability on one occasion but not the other, because the control stream places ETA(10) inside the multiple-dose branch).
• Notes: Do not fold this into CONMED_EFV: a model that used CONMED_EFV = 0 for the single-dose arm would assert that arm was efavirenz-free, which is false and would make the coefficient non-transferable to any study with a genuine no-efavirenz reference. The distinction matters quantitatively – Collins 2025 Discussion notes that a single efavirenz dose may already raise midazolam clearance somewhat, so a coefficient estimated against a single-dose reference understates the full efavirenz effect relative to one estimated against a drug-free reference. Also distinct from MULTI_DOSE_PT, which flags the multiple-dose phase of the analyte in a patient study; here the analyte (the CYP3A probe) is given as a single dose on both occasions and it is the perpetrator whose dosing history changes. Future papers using the same sequential single-dose/steady-state perpetrator design with a different inducer should register a parallel CONMED_<INN>_MD sibling rather than overloading this entry.

CONMED_EPI (canonical for concomitant epinephrine (adrenaline) coadministration indicator)

• Description: 1 = subject received epinephrine (INN: epinephrine; alternate INN adrenaline) coadministered with the parent drug during the study window; 0 = subject received the parent drug alone. Used to encode the local vasoconstrictor effect of epinephrine on intraperitoneal (or perineural) drug distribution: alpha-1 vasoconstriction of the peritoneal (or local) vessels reduces splanchnic blood flow and slows the transfer of the parent drug out of the injection compartment, while systemic beta-1 / beta-2 effects (positive inotropic / chronotropic action and peripheral vasodilatation) increase the apparent volume of distribution of the parent drug.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no coadministered epinephrine).
• Source aliases:
  • EPI – used in Royer_2011_cisplatin.R (Royer 2011 Table 2 covariate on IPCL and V).
• Example models: Royer_2011_cisplatin.R (intraperitoneal epinephrine in the cisplatin IP bath: multiplicative fractional coefficient -0.531 on intraperitoneal clearance IPCL (53.1% decrease) and +0.805 on central volume V (80.5% increase); the paper dichotomises epinephrine because plasma Pt concentrations were similar across the three epinephrine dose levels of 1, 2, and 3 mg/L in the IP bath).
• Notes: Specific scope until a second popPK / PD extraction with coadministered epinephrine reuses this indicator (e.g., a perineural local-anaesthetic + epinephrine PK model). Auto-approved member of the CONMED_<INN> family. The per-model covariateData[[CONMED_EPI]]$notes field should record the paper’s specific epinephrine dosing (route, concentration in the bath / mL, whether dichotomised or dose-response) since the mechanism differs across administration routes.

CONMED_EIAED (canonical for concomitant enzyme-inducing antiepileptic drug indicator)

• Description: 1 = subject is taking at least one enzyme-inducing antiepileptic drug (EIAED) such as carbamazepine, phenobarbital, or phenytoin during the study; 0 = no EIAED coadministration. EIAEDs induce hepatic metabolism (CYP3A4/UGT-mediated pathways) and increase the clearance of co-administered antiepileptic drugs and their active metabolites.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no EIAED coadministration).
• Source aliases:
  • MED (with the value convention inverted: source codes 1 = absence of EIAEDs and 0 = presence; canonical inverts this so 1 = presence and 0 = absence) – used in Rodrigues_2017_oxcarbazepine.R.
  • Inducer (same value convention as the canonical: 1 = on an enzyme-inducing comedication, 0 = not) – used in Bamgboye_2026_topiramate.R.
• Example models: Rodrigues_2017_oxcarbazepine.R (exponential effect on MHD apparent clearance: cl_mhd *= exp(e_eiaed_cl_mhd * (1 - CONMED_EIAED)); CL_MHD is 29.3% higher with EIAEDs vs without, encoded as +0.257 on the absence-indicator in the source paper), Bamgboye_2026_topiramate.R (power effect on intravenous topiramate central clearance: cl *= e_eiaed_cl ^ CONMED_EIAED with e_eiaed_cl = 1.63, i.e. CL is 63% higher on an enzyme inducer).
• Notes: Per-model covariateData[[CONMED_EIAED]]$notes should list the specific EIAEDs counted as “EIAED = 1” since inclusion criteria vary across antiepileptic-drug studies. Rodrigues 2017 counts carbamazepine, phenobarbital, and phenytoin as EIAEDs; other AEDs in the dataset (vigabatrin, clobazam, valproic acid, clonazepam, lamotrigine, diazepam, ethosuccimide, progabide) are not. Bamgboye 2026 counts carbamazepine, phenytoin, and oxcarbazepine – note that oxcarbazepine is a weaker inducer than the other two, so the EIAED set is not identical across models even though the canonical column name is. The source paper uses an “absence of EIAED” indicator (MED = 1 if no EIAED, MED = 0 if EIAED present); the canonical column inverts this so that the 1 group is “on EIAED”, matching the convention for other CONMED_* indicators.

CONMED_ERA (canonical for concomitant endothelin-receptor-antagonist monotherapy indicator in PAH)

• Description: 1 = patient is on a concomitant endothelin-receptor-antagonist (ERA) PAH therapy (e.g., bosentan, macitentan, ambrisentan) alone – i.e. on an ERA but NOT also on a phosphodiesterase type 5 inhibitor; 0 = otherwise. Time-fixed per subject for the PK/PD analysis (PAH comedication at baseline, stable dose required before selexipag start).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (PAH-comedication-naive OR on PDE5 inhibitor alone OR on both ERA+PDE5; mutually exclusive with CONMED_PDE5I and CONMED_ERA_PDE5I).
• Source aliases:
  • PAHCOMED category 1 (“ERA only”) – decomposed from the categorical PAHCOMED source column with levels {naive, ERA, PDE5, ERA+PDE5} in Krause_2017_selexipag.R.
• Example models: Krause_2017_selexipag.R (multiplicative effect on the ACT-333679 elimination rate constant: km *= (1 + 0.15 * CONMED_ERA); +15% relative to the PAH-comedication-naive reference, Krause 2017 Table 1), Axelsen_2024_selexipag_adult.R, Axelsen_2024_selexipag_pediatric.R (exponential effect km *= exp(0.164 * CONMED_ERA) adult / exp(0.186 * CONMED_ERA) pediatric, Axelsen 2024 Table 1 beta_km(COPAH1_1); source column COPAH1).
• Notes: Used together with CONMED_PDE5I and CONMED_ERA_PDE5I to decompose a four-level PAH-comedication categorical (naive / ERA-only / PDE5-only / ERA-and-PDE5) into three orthogonal mutually-exclusive binary indicators, with the PAH-comedication-naive group as the reference (all three indicators = 0). The four-level categorical encoding preserves the source paper’s stratum-specific coefficient set (Krause 2017 reports a separate categorical b-coefficient for each non-reference stratum rather than independent class-level effects). Specific scope because the indicator’s semantics are tied to the GRIPHON study’s PAH-comedication taxonomy. The registered models are all selexipag: Krause 2017 is the original Monolix GRIPHON fit and Axelsen 2024 is the same model re-estimated in NONMEM plus a Bayesian pediatric update. Note the two encode the categorical effect differently – Krause as (1 + b * I) and Axelsen as exp(b * I) – because Axelsen’s NONMEM control stream (Supplementary Table S3) adds each beta to the MU-referenced log-scale parameter before exponentiating. The coefficients are numerically close between the two papers, so the two forms agree to within ~1-2%; do not assume either form when adding a new model, read the source’s own transformation.

CONMED_ERA_PDE5I (canonical for concomitant ERA + PDE5-inhibitor combination indicator in PAH)

• Description: 1 = patient is on both a concomitant endothelin-receptor-antagonist (ERA) and a phosphodiesterase type 5 inhibitor (PDE5I) as PAH therapy; 0 = otherwise. Time-fixed per subject for the PK/PD analysis.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (PAH-comedication-naive OR on ERA alone OR on PDE5I alone; mutually exclusive with CONMED_ERA and CONMED_PDE5I).
• Source aliases:
  • PAHCOMED category 3 (“ERA and PDE5 inh.”) – decomposed from the categorical PAHCOMED source column with levels {naive, ERA, PDE5, ERA+PDE5} in Krause_2017_selexipag.R.
• Example models: Krause_2017_selexipag.R (multiplicative effect on the ACT-333679 elimination rate constant: km *= (1 + 0.37 * CONMED_ERA_PDE5I); +37% relative to the PAH-comedication-naive reference, Krause 2017 Table 1; the combined stratum has its own categorical coefficient distinct from the sum of the ERA-only and PDE5I-only effects), Axelsen_2024_selexipag_adult.R, Axelsen_2024_selexipag_pediatric.R (exponential effect km *= exp(0.358 * CONMED_ERA_PDE5I) adult / exp(0.368 * CONMED_ERA_PDE5I) pediatric, Axelsen 2024 Table 1 beta_km(COPAH3_1); source column COPAH3. This is the dominant stratum in the pediatric cohort – only 3 of 59 participants received no PAH comedication at all).
• Notes: Used together with CONMED_ERA and CONMED_PDE5I to decompose a four-level PAH-comedication categorical (naive / ERA-only / PDE5-only / ERA-and-PDE5) into three orthogonal mutually-exclusive binary indicators. The standalone combined-stratum coefficient (rather than a product of class-level indicator effects) preserves Krause 2017’s full categorical-effect parameterisation. Specific scope because the indicator’s semantics are tied to the GRIPHON study’s PAH-comedication taxonomy.

CONMED_EZE (canonical for concomitant ezetimibe coadministration indicator)

• Description: 1 = patient is taking ezetimibe (with or without other lipid-lowering comedication), 0 = not on ezetimibe.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (not on ezetimibe).
• Source aliases:
  • Derived from an ezetimibe-identifier column in the source.
• Example models: Kuchimanchi_2018_evolocumab.R (multiplicative effect 1.20 on Vmax: Vmax * 1.20^CONMED_EZE; labeled “Statin + ezetimibe exponent” in Kuchimanchi 2018 Table 3 because ~99% of ezetimibe users in the dataset were also on a conmed_statin, so the effect effectively captures combination therapy), Kuchimanchi_2018_evolocumab_ldlc.R (same PK-layer effect plus a multiplicative exponent 0.768 on baseline LDL-C in the exposure-response layer, Table 4), Kakara_2014_atorvastatin.R, Kakara_2014_pitavastatin.R, Kakara_2014_rosuvastatin.R (additive +0.109 contribution to the indirect-response Imax inhibition fraction INH on the LDL-C synthesis rate Kin: INH = Imax * DOSE / (ID50 + DOSE) + 0.109 * CONMED_EZE; Kakara 2014 Table 2 INH_EZT), Jadhav_2023_bempedoicAcid.R, Jadhav_2023_bempedoicAcid_ldlc.R (cleanly separated concomitant-ezetimibe indicator: proportional shift CL/F * (1 + (-0.0934) * CONMED_EZE) in the PK layer, Jadhav 2023 Table 2, and Imax * (1 + 0.190 * CONMED_EZE) in the indirect-response LDL-C layer, Table 3; Jadhav 2023 also carries a separate prior-therapy column [[PRIOR_EZE]], so this indicator is concomitant use only and is not a statin-combination proxy).
• Notes: Scope: specific because Kuchimanchi 2018 interprets the ezetimibe indicator as a combination-therapy marker rather than a pure ezetimibe effect. Future popPK/PD models with cleaner ezetimibe separation should add themselves here or register a more specific canonical.

CONMED_FLUCLOXACILLIN (canonical for concomitant flucloxacillin coadministration indicator)

• Description: 1 = patient is coadministered flucloxacillin (typically 6-12 g daily intravenously, or oral equivalent) alongside the modelled drug during the observation interval; 0 = no concomitant flucloxacillin. Flucloxacillin is an isoxazolyl penicillin (beta-lactamase-stable) used against beta-lactamase-producing staphylococci and streptococci. It is a pregnane X receptor (PXR) agonist: flucloxacillin binds PXR, which co-activates a PXR / retinoid X receptor complex on the XREM / PXRE enhancer and upregulates CYP3A4, CYP2C9 and CYP2C19 over roughly 2 days to 2 weeks. It is also about 95 % albumin bound and has been investigated as an albumin-binding-displacement perpetrator. Time-varying per subject, because flucloxacillin therapy starts and stops within the observation window and the induction effect has a documented onset lag.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant flucloxacillin).
• Source aliases:
  • flucloxacillin – Abdullah-Koolmees 2024 prose and Table 5 heading; same orientation as the canonical.
• Example models: AbdullahKoolmees_2024_voriconazole_pbpk.R (gates a step reduction of all three voriconazole CYP Michaelis constants – CYP2C19 9.3 to 3.72 uM, CYP3A4 834.7 to 181.45 uM, CYP2C9 20 to 13.33 uM – from a model-parameterised onset time tind fixed at 24 h, representing PXR-mediated CYP induction; the resulting rise in intrinsic hepatic clearance reproduces the undetectable voriconazole trough observed in the index case), AbdullahKoolmees_2024_posaconazole_pbpk.R (declared and carried at its reference value only – posaconazole is 83 % excreted unchanged and its minor UGT1A4 route was assumed unaffected by PXR upregulation, so the covariate has no structural effect in that model and documents the paper’s explicit negative finding).
• Notes: Distinct from CONMED_ABX (any-other-antibiotic composite) and from CONMED_FUSIDIC (a different staphylococcal agent whose interaction mechanism is CYP3A4 inhibition and albumin displacement rather than PXR-mediated induction) – flucloxacillin’s documented perpetrator mechanism is enzyme induction, so the sign of its effect on a CYP substrate’s clearance is opposite. The per-model covariateData[[CONMED_FLUCLOXACILLIN]]$notes must document the onset lag applied, because the induction is not instantaneous: published case reports place the fall in azole concentrations 2-7 days after flucloxacillin is started, while mechanistic CYP3A4 upregulation is reported to take 2 days to 2 weeks. Where a model implements the induction as a step change, record the step time.

CONMED_FLUOROURACIL (canonical for concomitant fluorouracil (5-FU) coadministration indicator)

• Description: 1 = patient is coadministered fluorouracil (5-FU), conventionally with leucovorin (LV) as a biochemical modulator, alongside the modelled drug; 0 = the modelled drug is given as monotherapy. 5-FU is a fluoropyrimidine antimetabolite whose active anabolites inhibit thymidylate synthase and are incorporated into RNA and DNA; it is cleared predominantly by dihydropyrimidine dehydrogenase. Registered as a coadministration indicator because 5-FU/LV is a standard partner for irinotecan-class and platinum-class cytotoxics, so the same column recurs across oncology popPK analyses that test whether the partner regimen perturbs the modelled drug’s disposition.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (modelled drug given as monotherapy).
• Source aliases:
  • treatment contain 5FU / treatment contains 5FU – Adiwijaya 2017 Tables S6 and S7 theta row labels, stated relative to “treatment do not contain 5FU”; same orientation as the canonical.
  • Coadministration with 5-FU – Adiwijaya 2017 Table S3 covariate-structure row.
• Example models: Adiwijaya_2017_irinotecan_liposomal.R (exponential effects of +0.075 on total-irinotecan CL and -1.53e-04 on SN-38 CL; both are small and the paper draws no drug-interaction conclusion from them, but the full-covariate approach retains them in the final model).
• Notes: Distinct from CONMED_CAPECITABINE should one be registered: capecitabine is an oral 5-FU prodrug and a patient on it is not receiving the intravenous 5-FU/LV regimen this indicator describes. When a source reports 5-FU and leucovorin as a single combination arm – the usual case – record that in the per-model covariateData[[CONMED_FLUOROURACIL]]$notes rather than adding a separate leucovorin column, because the two are not separately identifiable.

CONMED_OXALIPLATIN (canonical for concomitant oxaliplatin coadministration indicator)

• Description: 1 = patient receives oxaliplatin alongside the modelled drug; 0 = no oxaliplatin. Oxaliplatin is a third-generation platinum cytotoxic that forms DNA adducts and is cleared largely by non-enzymatic biotransformation and renal excretion of platinum species. It is the third component of the FOLFIRINOX / NALIRIFOX regimens, so it recurs as a partner drug in irinotecan-class popPK analyses.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (modelled drug given without oxaliplatin).
• Source aliases:
  • TRTOXA – Brendel 2021 final NONMEM control stream (Supplementary Material); TRTOXA = 1 is the effect category. Same orientation as the canonical.
  • Oxaliplatin administration / Oxaliplatin coadministration – Brendel 2021 Table 2 row labels.
• Example models: Brendel_2021_irinotecan_liposomal.R (factor 1.339 on total-irinotecan clearance and 0.656 on SN-38 clearance; oxaliplatin was always given with 5-FU/LV in the one contributing study, NCT02551991).
• Notes: A regimen-partner indicator, not an oxaliplatin exposure: models that need the oxaliplatin concentration as a PD driver use CP_OXA_MGL or CONC_OXA_UM instead. When the source’s only oxaliplatin arm also received 5-FU/LV, the two are not separately identifiable from other arms that received 5-FU/LV alone; record that in the per-model covariateData[[CONMED_OXALIPLATIN]]$notes rather than adding an interaction column.

CONMED_FLUOXETINE (canonical for concomitant fluoxetine (strong CYP2D6 inhibitor) coadministration indicator)

• Description: 1 = subject coadministered fluoxetine during the observation interval, 0 = no concomitant fluoxetine. Fluoxetine is a selective serotonin reuptake inhibitor and an FDA index strong CYP2D6 inhibitor; both fluoxetine and its long-lived metabolite norfluoxetine inhibit CYP2D6, and norfluoxetine additionally inhibits CYP3A4, so the interaction persists for weeks after the last dose. Per-model notes should record which pathway the source attributes the interaction to. Distinct from the class-level [[CONMED_CYP2D6_INH]] (any CYP2D6 inhibitor), which is the right column when the source pools inhibitors into one indicator rather than singling fluoxetine out. Time-fixed per subject in a therapeutic-drug-monitoring cohort where comedication is recorded once; time-varying when a study spans on / off fluoxetine periods.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant fluoxetine).
• Source aliases:
  • FLU – used in Zhang_2024_aripiprazole.R (Zhang 2024 Results 3.2: “For patients taking fluoxetine, FLU was 1; otherwise, FLU was 0”).
  • Fluoxetine hydrochloride capsule – the concomitant-medication row label in Zhang 2024 Table 2.
• Example models: Zhang_2024_aripiprazole.R (founding example; multiplicative fractional effect on apparent oral clearance only, cl * (1 + (-0.286) * CONMED_FLUOXETINE) per Zhang 2024 Eq. (6), a 28.6% clearance reduction that the paper summarises as a with:without fluoxetine clearance ratio of 0.714:1 and attributes to CYP2D6 inhibition of the main aripiprazole metabolic route).
• Notes: Auto-approved member of the CONMED_<INN> family (INN = fluoxetine). Registered as a per-drug indicator rather than folded into [[CONMED_CYP2D6_INH]] for the same reason [[CONMED_FLUCONAZOLE]] and [[CONMED_ITRACONAZOLE]] are kept apart from the pooled CYP3A4-inhibitor columns: a source that estimates an effect for one named inhibitor is measuring that drug’s perpetrator strength, and pooling it into a class indicator would discard exactly that contrast. Use the class-level column only when the source itself pools. When the exposed subgroup is small, say so in the per-model notes – in the founding example the coefficient rests on 3 of 119 patients, which the model file records explicitly.

CONMED_FOSRAVUCONAZOLE (canonical for concomitant fosravuconazole (E1224) coadministration indicator)

• Description: 1 = subject is randomised to a treatment arm that combines the modelled drug with fosravuconazole; 0 = the modelled drug alone. Fosravuconazole (development code E1224) is the water-soluble monolysine phosphate prodrug of ravuconazole, a broad-spectrum triazole antifungal that is rapidly hydrolysed to the active moiety after oral dosing. It has been developed by DNDi and Eisai as a candidate for Chagas disease because ravuconazole inhibits the Trypanosoma cruzi sterol 14-alpha-demethylase (CYP51) required for ergosterol biosynthesis. Like other systemic triazoles it also interacts with mammalian cytochrome P450 enzymes, so it is a plausible drug-drug-interaction perpetrator on the modelled drug.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (modelled drug given as monotherapy).
• Source aliases:
  • E1224 – used in Assmus_2025_benznidazole.R and Assmus_2025_benznidazole_qpcr.R (BENDITA, NCT03378661; the development code E1224 is used throughout the paper’s tables and figure legends in place of the INN).
• Example models: Assmus_2025_benznidazole.R (proportional increase in benznidazole apparent clearance: cl * (1 + e_conmed_fosravuconazole_cl * CONMED_FOSRAVUCONAZOLE) with e_conmed_fosravuconazole_cl = 0.177, i.e. CL/F is 17.7 % higher on combination therapy; Assmus 2025 Table 2, dOFV = -15.7), Assmus_2025_benznidazole_qpcr.R (same PK-layer effect; also tested as a predictor in the exposure-response layer, where it was not significant).
• Notes: Specific scope because the only source is the BENDITA trial (Assmus 2025). Auto-approved member of the CONMED_<INN> family. Distinct from the class-level [[CONMED_AZOLE]] indicator, which pools systemic azole antifungals used as incidental comedication and is oriented around CYP3A4 / P-gp inhibition (Kirubakaran 2022 tacrolimus: an 80 % CL/F reduction). Fosravuconazole here is a protocol-assigned combination study drug, and its measured effect on benznidazole runs in the opposite direction (a modest clearance increase), so collapsing it into CONMED_AZOLE would misstate both the direction and the provenance of the effect; a future paper that genuinely treats fosravuconazole as incidental azole comedication should use CONMED_AZOLE instead. In BENDITA the combination arms also differ from the monotherapy arms in benznidazole dose and schedule, so the indicator is confounded with regimen; Assmus 2025 nonetheless estimates it as a single arm-level covariate on CL/F and judges the resulting effect (< 20 %) not clinically relevant. Encoded as a time-fixed per-subject indicator because the paper models arm assignment, not a time-varying coadministration window; a future paper reporting a start/stop schedule can carry a time-varying indicator in the same column and should record that in covariateData[[CONMED_FOSRAVUCONAZOLE]]$notes.

CONMED_FUSIDIC (canonical for concomitant fusidic acid coadministration indicator)

• Description: 1 = patient is coadministered oral fusidic acid (typically 500 mg three times daily) alongside the modelled drug during the observation interval; 0 = no concomitant fusidic acid. Fusidic acid is a bacteriostatic protein-synthesis-inhibitor antibiotic used against staphylococcal infections; it is extensively plasma-protein bound (about 97 % to albumin), is orally bioavailable, and has been implicated as a CYP3A4 inhibitor and albumin-binding perpetrator in multiple drug-drug-interaction reports.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no concomitant fusidic acid).
• Source aliases:
  • fusidic acid – used in Marsot_2017_rifampicin.R (Marsot 2017 Table 1 / Table 2 stratifies typical CL/F and V/F by fusidic acid coadministration; the canonical column CONMED_FUSIDIC collapses the two strata into a single binary indicator).
• Example models: Marsot_2017_rifampicin.R (log-scale multiplicative effects on CL/F and V/F: cl = exp(lcl + e_conmed_fusidic_cl * CONMED_FUSIDIC + etalcl) with e_conmed_fusidic_cl = log(5.1 / 13.7) recovers Table 2’s stratified typical values 13.7 L/h at CONMED_FUSIDIC = 0 and 5.1 L/h at CONMED_FUSIDIC = 1; the analogous encoding on V/F recovers 61.1 L vs 23.8 L).
• Notes: Specific scope because the only source is Marsot 2017 (staphylococcal osteoarticular infections; 10 / 62 patients coadministered fusidic acid at 500 mg three times daily). Auto-approved member of the CONMED_<INN> family. The Marsot 2017 Discussion attributes the interaction to fusidic acid inhibition of CYP3A4 and to displacement of rifampicin from albumin binding, but notes that the mechanism remains uncertain; downstream models that use CONMED_FUSIDIC for a different modelled drug should document the mechanism hypothesis in covariateData[[CONMED_FUSIDIC]]$notes. When a future paper reports a time-varying fusidic acid coadministration schedule (start / stop within the observation window), the same canonical column can carry the time-varying indicator; the per-model notes should record whether the encoding is time-fixed or time-varying.

CONMED_GLASDEGIB (canonical for concomitant glasdegib (smoothened inhibitor) coadministration indicator)

• Description: 1 = subject is in a glasdegib-containing treatment arm (glasdegib added to a chemotherapy backbone such as low-dose cytarabine (LDAC) or decitabine); 0 = subject is in the chemotherapy-backbone-alone comparator arm (no glasdegib). Glasdegib is an oral smoothened (SMO) inhibitor of the Hedgehog signaling pathway, approved in the United States in combination with LDAC for older adults with newly diagnosed AML who are ineligible for intensive chemotherapy.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (LDAC alone or chemotherapy-backbone-alone comparator arm). In Lin_2020_glasdegib_treatment.R the reference is the LDAC alone arm (BRIGHT AML 1003 Phase 2). In Lin_2020_glasdegib_decitabine.R the reference is also LDAC alone, paired with CONMED_DECITABINE to identify the three-arm structure of the exploratory pooled analysis.
• Source aliases:
  • Derived from the protocol-defined treatment arm in Lin_2020_glasdegib_treatment.R and Lin_2020_glasdegib_decitabine.R (BRIGHT AML 1003, NCT01546038).
• Example models: Lin_2020_glasdegib_treatment.R (binary indicator on the exponential overall-survival hazard; the published equation uses the complementary indicator LDAC_alone = 1 - CONMED_GLASDEGIB so the model file derives ldac_alone <- 1 - CONMED_GLASDEGIB inside model() to preserve the paper’s parameter values verbatim), Lin_2020_glasdegib_decitabine.R (paired with CONMED_DECITABINE in the three-arm exploratory analysis).
• Notes: Specific scope because the only source is the BRIGHT AML 1003 trial (Lin 2020). Auto-approved member of the CONMED_<INN> family. The Lin 2020 treatment-response equation uses LDAC_alone as the binary covariate (1 = LDAC alone, 0 = glasdegib + LDAC) with parameter theta_ldac_alone = 1.376; the canonical-column form (1 = glasdegib + LDAC) preserves the published value via the in-model 1 - CONMED_GLASDEGIB mapping rather than refitting the parameter against a re-encoded covariate.

CONMED_GCSF (canonical for concomitant granulocyte colony-stimulating factor (G-CSF) treatment indicator)

• Description: 1 = subject received granulocyte colony-stimulating factor (G-CSF; filgrastim, pegfilgrastim, lenograstim, or another G-CSF product) as concomitant medication during the observation period; 0 = did not receive G-CSF. Used in chemotherapy-induced myelosuppression popPK/PD analyses either (a) as an exclusion filter, or (b) as a covariate on Friberg-style bone-marrow lifespan-model parameters (typically MTT, kprol, Slope, and/or Gamma) to capture the effect of prophylactic or reactive G-CSF support on neutrophil recovery kinetics.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant G-CSF treatment).
• Source aliases:
  • G-CSF – van Hasselt 2013 paper narrative and Table 2 / Table 3 header notation.
  • GCSF – common NONMEM column-name form used by Friberg-style myelosuppression implementations.
• Example models: vanHasselt_2013_eribulin.R (proportional / dichotomous effect on MTT with theta = 0.883 – G-CSF shortens MTT by ~12% – and on SLOPE with theta = 1.3 – G-CSF increases the linear drug-effect coefficient by ~30%. Encoded as mtt = ... * e_gcsf_mtt^CONMED_GCSF and slope = ... * e_gcsf_slope^CONMED_GCSF), Suri_2019_brentuximab_fn_adc.R, Suri_2019_brentuximab_fn_mmae.R, Suri_2019_brentuximab_neutropenia.R, Suri_2019_brentuximab_teae.R (primary prophylactic G-CSF as a fixed additive logit shift in landmark exposure-safety logistic models for brentuximab vedotin + AVD; odds ratios 0.44, 0.41, 0.32 and 0.19, matching the source’s stated 55-81% reduction).
• Notes: Time-varying per subject in principle (G-CSF is given as a course over specific days after chemotherapy), but van Hasselt 2013 treats it as a subject-level indicator in the covariate model (subject received G-CSF at any point during the observation period). Per-model covariateData[[CONMED_GCSF]]$notes should record whether the paper’s encoding is “ever received G-CSF” (subject-level, time-fixed) or per-cycle time-varying. Distinct from CSF1 (colony-stimulating factor 1 / M-CSF concentration, a target-engagement biomarker in anti-CSF-1R mAb models) – the two concepts share the abbreviation “CSF” but are otherwise unrelated.

CONMED_H2RA (canonical for concomitant H2-receptor-antagonist use)

• Description: 1 = patient on concomitant histamine H2-receptor-antagonist therapy (e.g., ranitidine, famotidine), 0 = no CONMED_H2RA use. Captures another class of gastric-pH-modifying co-medication that may reduce bioavailability of pH-sensitive orally administered drugs.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no CONMED_H2RA use).
• Source aliases:
  • H2 – used in Goel_2016_Sonidegib.R (Goel 2016 dataset; defined as “significant” CONMED_H2RA use, i.e. duration of CONMED_H2RA use >= 80% of the PK assessment phase).
• Example models: Goel_2016_Sonidegib.R (multiplicative effect on F: 0.996^CONMED_H2RA, no clinically meaningful effect; reported alongside CONMED_PPI for completeness).
• Notes: Per-model covariateData[[CONMED_H2RA]]$notes must document the operational definition (Goel 2016: >= 80% of PK assessment phase). Distinct from CONMED_PPI.

CONMED_ANTACID (canonical for concomitant antacid use)

• Description: 1 = patient on concomitant antacid therapy (typically over-the-counter aluminum-, magnesium-, or calcium-based salts such as calcium carbonate, magnesium carbonate, magnesium hydroxide, or aluminum hydroxide), 0 = no CONMED_ANTACID use. Captures the third class of gastric-pH-modifying co-medication alongside CONMED_PPI and CONMED_H2RA; primarily used to test the impact of transient gastric-pH elevation on the bioavailability of pH-sensitive orally administered drugs.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no CONMED_ANTACID use).
• Source aliases:
  • ANTACIDF – used in Kemal_2026_nemtabrutinib.R (Kemal 2026 NONMEM control stream in the supplement; time-varying regressor coded 1 = concomitant antacid at the observation, 0 otherwise).
• Example models: Kemal_2026_nemtabrutinib.R (multiplicative effect on F: (1 + -0.0360 * CONMED_ANTACID), i.e. 3.6% lower F under antacid coadministration; RSE 72.8%, 95% CI includes zero – retained in the full covariate model).
• Notes: Per-model covariateData[[CONMED_ANTACID]]$notes should document the operational definition (per-observation time-varying indicator vs subject-level ever-vs-never) and the class of antacids pooled into the = 1 category, since inclusion criteria vary by study. Distinct from CONMED_PPI (proton-pump inhibitors: irreversible H+/K+-ATPase inhibition, sustained gastric-pH elevation) and CONMED_H2RA (histamine H2 receptor antagonists: reversible histamine-mediated acid-secretion blockade). Antacids act by direct chemical neutralization of gastric acid rather than by inhibiting acid secretion, giving them a shorter and more localized pH-elevating effect than the other two classes.

CONMED_PHOSBINDER (canonical for concomitant phosphate-binding agent use)

• Description: 1 = patient on a concomitant phosphate-binding agent (oral agents that sequester dietary phosphate in the gastrointestinal lumen to lower serum phosphate – e.g. sevelamer, lanthanum carbonate, calcium acetate, calcium carbonate, sucroferric oxyhydroxide), 0 = no phosphate-binder use. Typically a subject-level ever-vs-never indicator. Two distinct clinical settings generate this covariate: management of hyperphosphatemia in chronic kidney disease, and management of the on-target hyperphosphatemia that is a class effect of fibroblast growth factor receptor (FGFR) inhibitors.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant phosphate-binder use).
• Source aliases:
  • BINDER – used in Gong_2023_pemigatinib.R (Gong 2023 NONMEM control stream in Appendix S1; IF(BINDER.EQ.0) CLBINDER = 1 / IF(BINDER.EQ.1) CLBINDER = (1 + THETA(7))). Gong 2023 Table 1 footnote b defines it as “phosphate-binding agents”.
  • BINDER – also the source column in Ji_2022_pemigatinib.R (Ji 2022 Equation I; 1 = used, 0 = not used, time-varying by visit).
• Example models: Gong_2023_pemigatinib.R (multiplicative effect on CL/F: (1 + -0.155 * CONMED_PHOSBINDER), i.e. 15.5% lower apparent clearance with phosphate-binder coadministration; %RSE 14.6, 95% CI -19.9% to -11.1%; the paper states the effect is statistically but not clinically significant, raising simulated Cmax,ss by roughly 6%), Ji_2022_pemigatinib.R (binder use is the paper’s reference level: CL/F is multiplied by (1 + 0.141 * (1 - CONMED_PHOSBINDER)), i.e. 1.141-fold higher without a binder, a 12.4% lower CL/F with one).
• Notes: Per-model covariateData[[CONMED_PHOSBINDER]]$notes should document the operational definition (subject-level ever-vs-never vs per-record time-varying) and which agents were pooled into the = 1 category, since these vary by study. Distinct from CONMED_ANTACID even though some compounds appear in both classes (calcium carbonate, calcium acetate, aluminium and magnesium salts): the two canonicals encode different therapeutic intents and different mechanisms of action on the drug of interest. CONMED_ANTACID is a gastric-pH-modifying covariate tested alongside CONMED_PPI and CONMED_H2RA for effects on the absorption of pH-sensitive drugs; CONMED_PHOSBINDER is a luminal-chelation covariate whose relevant mechanism is binding of the co-administered drug in the gut, and which in FGFR-inhibitor programmes is also a marker of on-treatment hyperphosphatemia management. When a source paper reports both, register both.

CONMED_SEVELAMER_BIXALOMER, CONMED_CALCIUM_CARBONATE, CONMED_LANTHANUM_CARBONATE, CONMED_FERRIC_CITRATE, CONMED_SUCROFERRIC_OXYHYDROXIDE (canonical for agent-specific concomitant phosphate-binder indicators)

• Description: 1 = the named phosphate binder was coadministered at this record; 0 = not. Five members of the CONMED_<INN> family sharing one register entry because they share one role: each is an oral phosphate binder that can chelate a coadministered drug in the gut lumen, and the source estimates a separate effect per agent. CONMED_SEVELAMER_BIXALOMER pools the two synthetic polymer binders (sevelamer hydrochloride and bixalomer); the other four name one metal-based binder each (calcium carbonate, lanthanum carbonate, ferric citrate hydrate, sucroferric oxyhydroxide). Time-varying when built from daily administration records.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant exposure to the named agent).
• Source aliases:
  • SBUSE, CUSE, LUSE, FUSE, SCUSE – used in Takada_2022_roxadustat.R (Takada 2022 Table 4 rows ‘SBUSE on F1’, ‘CUSE on F1’, ‘LUSE on F1’, ‘FUSE on F1’, ‘SCUSE on F1’, in that order).
• Example models: Takada_2022_roxadustat.R (exponentiated-switch factors on roxadustat relative bioavailability F1: 0.744 sevelamer/bixalomer, 0.931 calcium carbonate, 0.969 lanthanum carbonate, 0.744 ferric citrate, 0.837 sucroferric oxyhydroxide; multiplied together when several binders are taken; founding example).
• Notes: Use these when a source resolves the phosphate-binder effect by agent; use the class-level CONMED_PHOSBINDER when the source pools all binders into one indicator. A model that needs “any binder” from the agent indicators derives it inside model() (e.g. 1 - prod(1 - indicator)) rather than requiring a redundant CONMED_PHOSBINDER column, so the two cannot disagree. CONMED_CALCIUM_CARBONATE encodes calcium carbonate taken as a phosphate binder (luminal chelation); calcium carbonate taken as an antacid belongs under CONMED_ANTACID (see that entry). CONMED_FERRIC_CITRATE is ferric citrate taken as a phosphate binder, distinct from oral iron supplementation. The pooled CONMED_SEVELAMER_BIXALOMER name follows the source’s grouping ‘due to the same mechanism of action and similar effects on bioavailability’; register CONMED_SEVELAMER or CONMED_BIXALOMER separately if a later source separates them.

CONMED_CLOPIDOGREL (canonical for concomitant clopidogrel coadministration indicator)

• Description: 1 = the subject was receiving concomitant clopidogrel; 0 = not. Clopidogrel’s acyl-glucuronide metabolite is a strong time-dependent CYP2C8 inhibitor, so the indicator is the usual way to screen a CYP2C8-substrate drug’s clearance for that interaction.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant clopidogrel).
• Source aliases:
  • clopidogrel (CYP2C8 inhibitor) – Takada 2022 Table 2 candidate covariate on CL/F (prose name; no data-column code published).
• Example models: Takada_2022_roxadustat.R (declared in covariatesDataExcluded: screened on roxadustat CL/F and not retained; founding example).
• Notes: Member of the CONMED_<INN> family. Registered as a screened-only name; no current model carries an effect for it. Distinct from the class-level CONMED_ANTIPLATELET.

CONMED_HYALURONIDASE (canonical for coadministered recombinant human hyaluronidase (rHuPH20) subcutaneous-delivery adjuvant)

• Description: 1 = the subcutaneous dose was coadministered with recombinant human hyaluronidase (rHuPH20; marketed as the ENHANZE drug product, EDP, and co-formulated in products such as SC trastuzumab, SC rituximab, SC daratumumab and pertuzumab/trastuzumab fixed-dose combination), 0 = no hyaluronidase. rHuPH20 transiently depolymerises hyaluronan in the subcutaneous extracellular matrix, raising the volume and rate that can be injected at one site and altering the absorption of the coadministered drug. Time-fixed per dosing occasion: a trial that gives some doses with and some without the adjuvant carries it per dose record, not per subject.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no hyaluronidase coadministration).
• Source aliases:
  • EDP – used in Mahomed_2025_CAP256V2LS.R (Mahomed 2025 Table 3 “EDP – No / Yes” and Supplementary Table S1, where the EDP arms are marked with an asterisk; the paper writes it out as “a recombinant human hyaluronidase (ENHANZE Drug Product; EDP)”).
• Example models: Mahomed_2025_CAP256V2LS.R (founding example; three separate multiplicative effects from Supplementary Table S6 – clearance x 0.378, central volume x 0.41, and an effective relative-bioavailability factor of 0.452 applied on the logit scale as logit(0.712 * 0.452) - logit(0.712)).
• Notes: The covariate records the presence of the adjuvant, not the route: pair it with the model’s own IV-vs-SC dosing structure rather than treating it as a route indicator. Whether the estimated effects extend to intravenous records is a per-paper modelling choice that the model file must state – Mahomed 2025 applies its clearance and central-volume effects to every record and simulates hypothetical IV +EDP scenarios (Supplementary Tables S3-S5), even though every EDP dose actually administered in CAPRISA 012B was subcutaneous, so the systemic effects are identified only from the SC arms. A study that instead co-formulates hyaluronidase into a single fixed product with no comparator arm has no contrast to estimate and should not register the column at all. Distinct from FORM_<drug>_<formulation> indicators, which encode a physical dosage form of the drug itself; rHuPH20 is a separate coadministered enzyme with its own approval and its own INN.

CONMED_IFNB1A (canonical for concomitant interferon beta-1a coadministration indicator)

• Description: 1 = patient coadministered subcutaneous interferon beta-1a (Rebif or equivalent recombinant IFN beta-1a product) during the observation interval, 0 = no concomitant IFN beta-1a. Time-varying per subject because the source studies enrol both monotherapy and IFN beta-1a combination periods.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant IFN beta-1a).
• Source aliases:
  • IFNB1A – used in Savic_2017_cladribine.R.
• Example models: Savic_2017_cladribine.R (multiplicative effect on cladribine non-renal clearance: cl_nonrenal *= (1 + e_ifn_clnr * CONMED_IFNB1A) with e_ifn_clnr = 0.21, i.e. a 21% increase in non-renal CL when coadministered with IFN beta-1a).
• Notes: Captured at the dose-record level in Savic 2017 (multiple-dose study 26486 alternated between cladribine alone and cladribine + IFN beta-1a periods). The interaction mechanism is not definitively established in the source paper; Savic 2017 discusses that the observed effect could alternatively be modelled on bioavailability or interpreted as a period-effect / interoccasion variability artefact. Future cladribine + immunomodulator extractions should reuse this canonical when IFN beta-1a is the specific concomitant agent; register a sibling canonical (e.g. CONMED_IFNB1B, CONMED_IFNALPHA) if a different interferon species is intended.

CONMED_IFNALPHA (canonical for concomitant interferon alpha coadministration indicator)

• Description: 1 = patient coadministered systemic interferon alpha (any subtype; recombinant IFN-alpha-2a, IFN-alpha-2b, or pegylated IFN-alpha) during the observation interval, 0 = no concomitant IFN-alpha. Time-fixed when the source paper analyses cohorts whose IFN-alpha exposure is unchanged across the analysis window (per-trial protocol); time-varying when IFN-alpha cycling on/off is recorded at the observation level.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant IFN-alpha).
• Source aliases:
  • IFNa – used in Han_2016_bevacizumab.R (Han 2016 Table 3 categorical indicator on bevacizumab CL; the 102 IFN-alpha-coadministered patients were enrolled in renal cell carcinoma study BO17705 per Han 2016 Table 1 and Table 2 row ‘Concomitant treatment’).
• Example models: Han_2016_bevacizumab.R (multiplicative effect on bevacizumab CL: cl *= exp(e_ifna_lcl * CONMED_IFNALPHA) with e_ifna_lcl = log(0.844) = -0.170, i.e. CL is 15.6% lower when IFN-alpha is coadministered).
• Notes: Sibling canonical to CONMED_IFNB1A (interferon beta-1a) under the CONMED_<INN> family; the existing CONMED_IFNB1A entry explicitly anticipated this CONMED_IFNALPHA name for the alpha-interferon species. Per-paper alpha-subtype detail (IFN-alpha-2a vs -2b vs pegylated) is documented in covariateData[[CONMED_IFNALPHA]]$notes where available; pooled here because the source paper (Han 2016) does not separately enumerate subtypes. Distinct from CONMED_IFNB1A/CONMED_IFNB1B (beta-interferon species).

CONMED_IL6RI (canonical for concomitant interleukin-6 receptor inhibitor (class) coadministration indicator)

• Description: 1 = subject received a concomitant interleukin-6 receptor (IL-6R) inhibitor – tocilizumab, sarilumab, or satralizumab – pooled as a class, 0 = no IL-6R inhibitor. A class-level rather than per-INN indicator, because papers that test this covariate are generally testing the pharmacological consequence of blocking IL-6R signalling (reversal of IL-6-mediated suppression of CYP3A4 / CYP2C8 and downregulation of acute-phase proteins such as alpha-1-acid glycoprotein and CRP), not the identity of the antibody. Document which agents, at what doses, and whether the flag is baseline-only or time-varying in each model’s covariateData[[CONMED_IL6RI]]$notes.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant IL-6R inhibitor).
• Source aliases:
  • IL6INHIB – NONMEM control-stream column name used in Said_2025_imatinib.R (Said 2025 Data S1 $INPUT; KD = EXP(THETA(3) + THETA(7)*IL6INHIB + ETA(3)) and CLM = EXP(THETA(10) + IL6INHIB*THETA(13) + ETA(5))).
• Example models: Said_2025_imatinib.R (two retained log-additive effects in a joint imatinib / N-desmethyl-imatinib popPK model: the imatinib-AAG dissociation constant Kd is multiplied by 1.7 and apparent DM-imatinib clearance CLm/(Fm x F1) by 0.46 when an IL-6R inhibitor preceded imatinib; the class pools 8 mg/kg IV tocilizumab and 400 mg IV sarilumab given as single doses on ICU admission. The same covariate was tested on unbound imatinib clearance and NOT retained – Said 2025 Data S1 holds that coefficient at (0) FIX).
• Notes: Class-level scope follows the established CONMED_<CLASS> precedents in this family (CONMED_AZOLE, CONMED_ABX, CONMED_AED, CONMED_SULFONYLUREA, CONMED_ERA), which likewise pool mechanism-equivalent agents. If a future paper estimates separate tocilizumab and sarilumab effects, register per-INN siblings (CONMED_TOCILIZUMAB, CONMED_SARILUMAB) and keep this canonical for the pooled-class case; do not use both a per-INN indicator and this class indicator on the same parameter in one model. Distinct from IL6 (the measured serum interleukin-6 concentration, a continuous covariate) – note that IL-6R blockade typically raises measured total IL-6 because the antibody-receptor complex slows IL-6 clearance, so the two covariates can move in opposite directions and must not be treated as substitutes. Also distinct from DIS_CRITILL and DIS_COVID19, with which it is frequently collinear in ICU cohorts: in Said 2025 every IL-6R-inhibitor patient was a ventilated COVID-19 ICU patient, so the two effects are separable only because the pooled dataset also contains ICU COVID-19 patients who received no IL-6R inhibitor. General scope because IL-6R blockade is a mechanism-defined, widely-used intervention whose disease-drug-drug-interaction effect on CYP substrates and protein binding is expected to recur.

CONMED_IMMUNOMOD (canonical for any-immunomodulator (purine analogue or methotrexate) composite indicator)

• Description: 1 = subject is on at least one concomitant immunomodulator (any of: azathioprine (AZA), 6-mercaptopurine (6-MP), or methotrexate (MTX)) at the PK observation, 0 = no concomitant immunomodulator from this class. A composite pooled indicator used in inflammatory-bowel-disease popPK analyses where the source paper does not separate the individual thiopurine vs MTX effects on therapeutic-mAb (typically anti-TNF) clearance.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant immunomodulator from the AZA / 6-MP / MTX class).
• Source aliases:
  • IMM – used in Frymoyer_2017_infliximab.R (pooled purine-analogue-or-MTX indicator per Table 1 footnote of Frymoyer 2017: “Concomitant immunomodulation refers to purine-analogue or methotrexate.”).
• Example models: Frymoyer_2017_infliximab.R (power-of-coefficient effect on CL: CL_typ * 0.863^CONMED_IMMUNOMOD, i.e. -13.7% when on an immunomodulator).
• Notes: Distinct from the per-drug indicators CONMED_AZA, CONMED_MP, and CONMED_MTX, which model the effect of each immunomodulator separately. Use CONMED_IMMUNOMOD only when the source paper itself pools the three under a single binary; use the per-drug indicators when the source paper estimates separate effects. Future IBD popPK extractions that pool thiopurines + MTX into a single indicator should reuse this canonical and extend the example list.

CONMED_IPI_1Q6W (canonical for nivolumab + ipilimumab 1 mg/kg q6w combination indicator)

• Description: 1 = subject is receiving nivolumab in combination with ipilimumab 1 mg/kg every 6 weeks (continuous maintenance); 0 = otherwise.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (any non-1Q6W regimen – monotherapy, chemotherapy combination, or another ipilimumab schedule).
• Source aliases:
  • IPI1Q6W – used in Zhang_2019_nivolumab.R.
• Example models: Zhang_2019_nivolumab.R (exponential effect on baseline CL: exp(0.159) ~= 1.17 fold increase relative to monotherapy).
• Notes: Paired with CONMED_IPI_3Q3W. See the CONMED_IPI_3Q3W note for how the other ipilimumab schedules collapse into the reference group.

CONMED_IPI_3Q3W (canonical for nivolumab + ipilimumab 3 mg/kg q3w combination indicator)

• Description: 1 = subject is receiving nivolumab in combination with ipilimumab 3 mg/kg every 3 weeks (4-dose induction); 0 = otherwise. Encodes the high-intensity ipilimumab combination regimen as a study-design covariate on nivolumab CL.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (any non-3Q3W regimen – monotherapy, chemotherapy combination, or another ipilimumab schedule).
• Source aliases:
  • IPI3Q3W – used in Zhang_2019_nivolumab.R.
• Example models: Zhang_2019_nivolumab.R (exponential effect on baseline CL: exp(0.227) ~= 1.25 fold increase relative to monotherapy), Hu_2024_nivolumab.R (exponential effect on baseline CL: exp(0.349) ~= 1.42 fold increase relative to monotherapy – the largest single covariate effect on nivolumab CL in that pooled adult + pediatric model).
• Notes: Paired with CONMED_IPI_1Q6W; both indicators can coexist in one population, but a single subject has at most one set to 1 in the Zhang 2019 cohort. The remaining ipilimumab schedules (1 mg/kg q3w x 4 induction, 1 mg/kg q12w) had no statistically significant effect on nivolumab CL and were therefore folded into the reference (0) group along with monotherapy, leaving only IPI3Q3W and IPI1Q6W as named non-reference indicators.

CONMED_IPI_ANY (canonical for any-ipilimumab-coadministration indicator)

• Description: 1 = subject is receiving nivolumab in combination with any ipilimumab regimen (regardless of dose or schedule); 0 = nivolumab monotherapy or nivolumab + chemotherapy. Encodes the “is there ipilimumab in the regimen” question as a single binary covariate, distinct from the regimen-specific CONMED_IPI_3Q3W and CONMED_IPI_1Q6W indicators above.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no ipilimumab coadministration).
• Source aliases:
  • IPICO – used in Zhang_2019_nivolumab.R.
• Example models: Zhang_2019_nivolumab.R (additive effect on the time-varying-CL Emax parameter: Emax += -0.0668 when CONMED_IPI_ANY = 1), Hu_2024_nivolumab.R (additive effect on the same parameter: Emax += -0.124 when CONMED_IPI_ANY = 1).
• Notes: Logically the union of the regimen-specific indicators (CONMED_IPI_3Q3W, CONMED_IPI_1Q6W, CONMED_IPI_1Q3W, plus the unmodeled 1 mg/kg q12w schedule). Zhang 2019 uses it on the time-varying Emax (a different structural parameter from baseline CL), which is why it coexists with the regimen-specific indicators on baseline CL rather than substituting for them. Hu 2024 uses it the same way, on the Emax of the pooled adult + pediatric nivolumab model.

CONMED_IPI_1Q3W (canonical for nivolumab + ipilimumab 1 mg/kg q3w combination indicator)

• Description: 1 = subject is receiving nivolumab in combination with ipilimumab 1 mg/kg every 3 weeks (4-dose induction); 0 = otherwise. The low-intensity counterpart of [[CONMED_IPI_3Q3W]], typically the flip of the nivolumab-3 / ipilimumab-1 versus nivolumab-1 / ipilimumab-3 induction pairing.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (any non-1Q3W regimen – monotherapy, chemotherapy combination, or another ipilimumab schedule).
• Source aliases:
  • COMBO level 1 (Hu 2024 Appendix File S2 NONMEM control stream) – decompose into CONMED_IPI_1Q3W = as.integer(COMBO == 1).
• Example models: Hu_2024_nivolumab.R (exponential effect on baseline CL: exp(0.0973) ~= 1.10 fold increase relative to monotherapy; 95% CI -0.00235 to 0.208).
• Notes: Completes the CONMED_IPI_<dose>Q<interval>W schedule family alongside CONMED_IPI_3Q3W and CONMED_IPI_1Q6W. In Zhang 2019 this schedule was folded into the reference group because its CL effect was not significant, so the two papers’ reference groups differ; a model that uses this canonical must state which schedules its own reference pools. The regimen-agnostic union is [[CONMED_IPI_ANY]].

CONMED_NIVO_1Q3W (canonical for ipilimumab + nivolumab 1 mg/kg q3w combination indicator)

• Description: 1 = subject is receiving ipilimumab in combination with nivolumab 1 mg/kg every 3 weeks; 0 = otherwise. This is the mirror image of the CONMED_IPI_* family: it encodes the nivolumab arm of the same nivolumab + ipilimumab regimens as a covariate on ipilimumab disposition rather than on nivolumab disposition.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (any non-1Q3W regimen – ipilimumab monotherapy, another nivolumab schedule, or a non-nivolumab combination).
• Source aliases:
  • COMBO level 1 (Hu 2024 Appendix File S3 NONMEM control stream) – decompose into CONMED_NIVO_1Q3W = as.integer(COMBO == 1).
• Example models: Hu_2024_ipilimumab.R (exponential effect on ipilimumab CL: exp(0.0417) ~= 1.04 fold increase relative to ipilimumab monotherapy; 95% CI -0.00318 to 0.0866).
• Notes: Paired with [[CONMED_NIVO_3Q3W]]. Do not substitute a CONMED_IPI_* indicator when the modelled analyte is ipilimumab – the two families name the coadministered drug, so the perpetrator/victim orientation must match the model’s analyte. A single subject has at most one of the CONMED_NIVO_* indicators set to 1.

CONMED_NIVO_3Q3W (canonical for ipilimumab + nivolumab 3 mg/kg q3w combination indicator)

• Description: 1 = subject is receiving ipilimumab in combination with nivolumab 3 mg/kg every 3 weeks; 0 = otherwise. Mirror image of the CONMED_IPI_* family (see [[CONMED_NIVO_1Q3W]]).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (any non-3Q3W regimen – ipilimumab monotherapy, another nivolumab schedule, or a non-nivolumab combination).
• Source aliases:
  • COMBO level 2 (Hu 2024 Appendix File S3 NONMEM control stream) – decompose into CONMED_NIVO_3Q3W = as.integer(COMBO == 2).
• Example models: Hu_2024_ipilimumab.R (exponential effect on ipilimumab CL: exp(0.316) ~= 1.37 fold increase relative to ipilimumab monotherapy; 95% CI -0.0390 to 0.670, RSE 57.3%).
• Notes: Paired with [[CONMED_NIVO_1Q3W]].

CONMED_BRENTUXIMAB (canonical for concomitant brentuximab vedotin coadministration indicator)

• Description: 1 = subject coadministered brentuximab vedotin (anti-CD30 antibody-drug conjugate, typically 1.8 mg/kg every 3 weeks), 0 = no concomitant brentuximab vedotin. Use for single-agent brentuximab vedotin coadministration; the multi-drug regimen indicator [[CONMED_AVD]] covers brentuximab vedotin given as part of the A+AVD chemotherapy backbone.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant brentuximab vedotin).
• Source aliases:
  • COMBO level 4 / BVCO (Hu 2024 Table 4 and Appendix File S2) – decompose into CONMED_BRENTUXIMAB = as.integer(COMBO == 4).
• Example models: Hu_2024_nivolumab.R (exponential effect on nivolumab baseline CL: exp(0.132) ~= 1.14 fold increase relative to nivolumab monotherapy; 95% CI -0.0572 to 0.282; the regimen is brentuximab vedotin 1.8 mg/kg q3w x 4 with nivolumab 3 mg/kg q3w in the pediatric / young-adult classical Hodgkin lymphoma study CA209744).
• Notes: Follows the CONMED_<INN> naming pattern. Scope: general because brentuximab vedotin is a standard Hodgkin lymphoma backbone likely to recur as a coadministration covariate in other checkpoint-inhibitor and ADC popPK analyses.

CONMED_ITRACONAZOLE (canonical for concomitant itraconazole (strong CYP3A4 / P-gp inhibitor) coadministration indicator)

• Description: 1 = subject coadministered itraconazole during the observation interval (typically 200 mg once-daily oral or higher), 0 = no concomitant itraconazole. Distinct from the broader [[CONMED_AZOLE]] indicator (any azole antifungal) and from [[CONMED_CYP3A4_INH]] (any CYP3A4 inhibitor) when a paper singles out itraconazole as an index victim / perpetrator; itraconazole is used clinically as a probe strong CYP3A4 + P-glycoprotein inhibitor in dedicated DDI studies. Per-subject time-varying when a DDI study spans on / off itraconazole periods; time-fixed on the on-treatment day of a fixed-sequence DDI arm.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant itraconazole).
• Source aliases:
  • INTR – used in vandenBerg_2021_uprifosbuvir_pbpk.R (van den Berg 2021 supplement input column INTR = 0 alone / 1 itraconazole, per PN001 Group F and PN010).
  • Itraconazole – used in Zhou_2025_tacrolimus.R (Zhou 2025 Table 1 counts 38 of 988 trough records on itraconazole; exponential coefficient e_itra_cl = -0.87 on tacrolimus CL/F, a 57.98% reduction – the largest of the three azoles in that cohort, consistent with itraconazole being the most lipophilic mould-active azole and the most potent CYP3A4 inhibitor after ketoconazole).
  • itraconazole – used in Jaiswal_2025_dordaviprone.R (Jaiswal 2025 Table 2 / Table S6, multiple-dose itraconazole with a single 125 mg dordaviprone dose; relative CYP3A4 activity 0.138 back-solved from the published AUC ratio of 4.62).
• Example models: vandenBerg_2021_uprifosbuvir_pbpk.R (multiplicative factors on eight gut / hepatic parameters per van den Berg 2021 Table 3: CL_int,G x 0.444, CL_int,H x 0.797, ALAG2 x 0.145, KA2 x 0.499, Ktr1 x 0.166, KM6g x 0.294, FrM6g x 0.174, KelM6g x 0.0290; the combined effect of these factors is a delay in uprifosbuvir Tmax and a large increase in uprifosbuvir plasma exposure without a corresponding rise in M5 / M6, consistent with itraconazole shunting uprifosbuvir away from gut CYP3A4 / P-gp-mediated first-pass metabolism toward hepatic conversion), Comisar_2025_rimegepant.R (two multiplicative fractional effects per Comisar 2025 Table 3: -0.743 on CL/F, a 74.3% clearance decrease giving a 3.9-fold rise in simulated steady-state AUCtau and supporting the label recommendation to avoid strong CYP3A4 inhibitors, and -0.351 on the transit rate constant ktr, a 35.1% slowing judged not clinically meaningful because ka is the slower of the two absorption steps; paired with the new sibling CONMED_FLUCONAZOLE (-0.427 on CL/F) so the model carries the strong and the moderate CYP3A4 probe inhibitor as separate covariates), Jaiswal_2025_dordaviprone.R (PBPK drug-interaction simulation; the effect enters through a single relative-CYP3A4-activity term of 0.138 that scales gut and hepatic first-pass extraction and systemic clearance together, back-solved from the published dordaviprone AUC ratio of 4.62), Huth_2019_siponimod.R (relative CYP3A4 activity 0.209, back-solved from the six published siponimod genotype AUC ratios; CYP2C9 activity held at 1 because Huth 2019 classifies itraconazole as CYP3A4-selective).
• Notes: Auto-approved member of the CONMED_<INN> family (INN = itraconazole). Distinct from the class-level [[CONMED_AZOLE]] (Kirubakaran 2022 tacrolimus, which pools itraconazole with other systemic azoles as a common CYP3A4 / P-gp inhibitor class effect) because dedicated itraconazole DDI studies – particularly those where itraconazole is used as the strong probe inhibitor in a Phase 1 mechanistic study – carry substrate-specific factors on multiple gut and hepatic parameters that would be diluted by pooling with weaker azoles. Future dedicated itraconazole DDI popPK extractions should reuse this canonical and extend the example-models list. Promote to a drug-specific variant if a paper distinguishes multiple itraconazole regimens; keep as the class-neutral indicator otherwise.

CONMED_ERYTHROMYCIN (canonical for concomitant erythromycin (moderate CYP3A4 inhibitor) coadministration indicator)

• Description: 1 = subject coadministered erythromycin during the observation interval (typically 500 mg four times daily oral), 0 = no concomitant erythromycin. Erythromycin is the FDA index moderate CYP3A4 inhibitor and acts through mechanism-based (time-dependent) inactivation rather than reversible competition. Distinct from the class-level [[CONMED_CYP3A4_INH_MOD]] indicator: use this canonical when a paper singles erythromycin out by name as the index moderate inhibitor in a dedicated DDI simulation or study, and the class indicator when the paper pools several moderate inhibitors into one covariate.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant erythromycin).
• Source aliases:
  • erythromycin – used in Jaiswal_2025_dordaviprone.R (Jaiswal 2025 Figure 3 forest plot arm, erythromycin 500 mg four times a day).
• Example models: Jaiswal_2025_dordaviprone.R (multiplicative effect through a single relative-CYP3A4-activity term that scales gut and hepatic first-pass extraction and systemic clearance together; the activity 0.372 was back-solved from the published dordaviprone AUC ratio of 2.68), Huth_2019_siponimod.R (relative CYP3A4 activity 0.312, back-solved from the six published siponimod genotype AUC ratios).
• Notes: Auto-approved member of the CONMED_<INN> family (INN = erythromycin). Sibling of [[CONMED_FLUCONAZOLE]], [[CONMED_CIMETIDINE]] and [[CONMED_RIFAMPICIN]], registered alongside the Jaiswal 2025 dordaviprone extraction, which simulates one arm per named index CYP3A4 modulator rather than pooling by potency tier. Per-model notes should record the erythromycin regimen, because the magnitude of a mechanism-based inhibitor’s effect depends strongly on dose and duration.

CONMED_FLUCONAZOLE (canonical for concomitant fluconazole (moderate CYP3A4 inhibitor) coadministration indicator)

• Description: 1 = subject coadministered fluconazole during the observation interval (typically 200 mg once daily oral), 0 = no concomitant fluconazole. Fluconazole is an FDA index moderate CYP3A4 inhibitor and is also a moderate CYP2C9 / CYP2C19 inhibitor, so per-model notes should record which pathway the source paper attributes the interaction to. Distinct from the class-level [[CONMED_AZOLE]] (any azole antifungal) and from [[CONMED_CYP3A4_INH_MOD]] (any moderate CYP3A4 inhibitor) when a paper singles out fluconazole as the index moderate-inhibitor probe in a dedicated DDI arm. Per-subject time-varying when a DDI study spans on / off fluconazole periods; time-fixed on the on-treatment day of a fixed-sequence DDI arm.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant fluconazole).
• Source aliases:
  • Fluconazole use – used in Comisar_2025_rimegepant.R (Comisar 2025 Table 2 row label).
  • fluconazole – used in Jaiswal_2025_dordaviprone.R (Jaiswal 2025 Figure 3 forest plot arm, fluconazole 200 mg once daily).
• Example models: Comisar_2025_rimegepant.R (founding example; multiplicative fractional effect on apparent clearance only, cl * (1 + (-0.429) * CONMED_FLUCONAZOLE) per Comisar 2025 Table 2 ‘Fluconazole use on CL/F’, a 42.9% clearance reduction giving a 1.75-fold AUC increase. The paired strong-inhibitor arm in the same analysis carries CONMED_ITRACONAZOLE with a larger clearance effect (-0.744) plus an additional effect on the transit rate constant, which fluconazole does not have – the two azoles are deliberately NOT pooled), Jaiswal_2025_dordaviprone.R (relative CYP3A4 activity 0.410, back-solved from the published dordaviprone AUC ratio of 2.48), Huth_2019_siponimod.R (the dual case: relative CYP3A4 activity 0.339 AND relative CYP2C9 activity 0.509, back-solved jointly; the CYP2C9 arm is what makes this the one perpetrator whose predicted siponimod AUC ratio is roughly flat across genotypes, 2.09-2.49), Jiang_2021_ivosidenib.R (multiplicative 0.59-fold change in ivosidenib CL/F).
• Notes: Auto-approved member of the CONMED_<INN> family (INN = fluconazole). Sibling of [[CONMED_ITRACONAZOLE]]: the two are registered separately, rather than under a shared azole or CYP3A4-inhibitor class canonical, because dedicated phase 1 DDI programmes commonly run a moderate (fluconazole) and a strong (itraconazole) probe in the same study and estimate distinct effect sizes – pooling them into one indicator would discard exactly the contrast the DDI arm was designed to measure. Use the class-level [[CONMED_AZOLE]] or [[CONMED_CYP3A4_INH_MOD]] only when the source itself pools. Also a sibling of [[CONMED_ERYTHROMYCIN]], [[CONMED_CIMETIDINE]] and [[CONMED_RIFAMPICIN]], added with the Jaiswal 2025 dordaviprone extraction: fluconazole and erythromycin are both index moderate inhibitors but produce measurably different victim exposures there (dordaviprone AUC ratios 2.48 and 2.68 respectively), a second argument for per-drug indicators over a pooled moderate-inhibitor covariate whenever the source reports each arm separately.

CONMED_FLUVOXAMINE (canonical for concomitant fluvoxamine (weak CYP3A4 and CYP2C9 inhibitor) coadministration indicator)

• Description: 1 = subject coadministered fluvoxamine during the observation interval (typically 50-100 mg once daily oral), 0 = no concomitant fluvoxamine. Fluvoxamine is a selective serotonin reuptake inhibitor and a broad-spectrum CYP inhibitor: a strong CYP1A2 and CYP2C19 inhibitor, and additionally a weak inhibitor of CYP3A4 and CYP2C9. It is therefore the usual index weak dual CYP3A4/CYP2C9 probe in drug-interaction simulations, and per-model notes must record which isoform the source attributes the effect to – a fluvoxamine arm in a CYP1A2-substrate paper is a completely different effect size from one in a CYP2C9-substrate paper. Distinct from any class-level pooled SSRI indicator and from the reserved CONMED_CYP3A4_INH_WEAK.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant fluvoxamine).
• Source aliases:
  • fluvoxamine – used in Huth_2019_siponimod.R (Huth 2019 Table 3 column header ‘Weak 3A4/2C9 inhibitor’; Simcyp V16 compound-file perpetrator, no Ki printed).
• Example models: Huth_2019_siponimod.R (two multiplicative relative-enzyme-activity terms back-solved jointly from the six published genotype AUC ratios: relative CYP3A4 activity 0.909 and relative CYP2C9 activity 0.665, each scaling its own clearance arm, with the CYP3A4 term additionally rescaling gut-wall and hepatic first-pass extraction. This is the only perpetrator in that paper whose predicted AUC ratio decreases as CYP2C9 function worsens – 1.42 in CYP2C91/1 down to 1.12 in 3/3 – because inhibiting CYP2C9 buys nothing in a subpopulation that barely uses it while the CYP3A4 inhibition is only weak).
• Notes: Auto-approved member of the CONMED_<INN> family (INN = fluvoxamine). Sibling of [[CONMED_ITRACONAZOLE]], [[CONMED_KETOCONAZOLE]], [[CONMED_ERYTHROMYCIN]], [[CONMED_FLUCONAZOLE]], [[CONMED_CIMETIDINE]] and [[CONMED_RIFAMPICIN]], the per-drug index-perpetrator indicators that DDI-prediction papers simulate one arm at a time. Registered as its own canonical rather than under [[CONMED_CYP3A4_INH]] because fluvoxamine’s dual CYP3A4 + CYP2C9 action is precisely why a paper picks it, and a CYP3A4-only class indicator cannot express the second mechanism.

CONMED_CIMETIDINE (canonical for concomitant cimetidine (weak CYP3A4 inhibitor) coadministration indicator)

• Description: 1 = subject coadministered cimetidine during the observation interval (typically 400 mg twice daily oral), 0 = no concomitant cimetidine. Cimetidine is an FDA index weak CYP3A4 inhibitor and is additionally an H2-receptor antagonist and an inhibitor of renal organic-cation transport, so a cimetidine arm can perturb a victim drug through gastric pH and renal secretion as well as through CYP3A4; per-model notes must record which mechanism the source attributes the effect to. Distinct from the class-level [[CONMED_H2RA]] (any H2-receptor antagonist, used when the source is interested in the gastric-pH effect) and from the reserved CONMED_CYP3A4_INH_WEAK.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant cimetidine).
• Source aliases:
  • cimetidine – used in Jaiswal_2025_dordaviprone.R (Jaiswal 2025 Figure 3 forest plot arm, cimetidine 400 mg twice a day).
• Example models: Jaiswal_2025_dordaviprone.R (relative CYP3A4 activity 0.739, back-solved from the published dordaviprone AUC ratio of 1.42; the source attributes the effect entirely to weak CYP3A4 inhibition, since dordaviprone renal clearance is zero and the pH effect was separately shown to be null in the rabeprazole arm).
• Notes: Auto-approved member of the CONMED_<INN> family (INN = cimetidine). When a future source uses cimetidine as a probe for renal organic-cation-transport inhibition rather than for CYP3A4, reuse this same column and disambiguate in per-model notes; do not register a mechanism-specific variant.

CONMED_MOXIFLOXACIN (canonical for concomitant moxifloxacin coadministration indicator)

• Description: 1 = subject is receiving moxifloxacin (a fluoroquinolone antibacterial) as a regimen partner during the PK observation period, 0 = no concomitant moxifloxacin. Time-fixed when the source analyses randomised regimen arms with a stable combination for the whole treatment period; time-varying when the partner is started or stopped mid-observation.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant moxifloxacin).
• Source aliases:
  • MOX – Salinger 2019 supplemental Table S3 $INPUT column; same orientation as the canonical.
• Example models: Salinger_2019_pretomanid.R (a main effect on apparent oral clearance, e_moxifloxacin_cl^CONMED_MOXIFLOXACIN with e_moxifloxacin_cl = 0.967, plus two interaction products – e_moxifloxacin_pyrazinamide_cl^(CONMED_MOXIFLOXACIN * CONMED_PYRAZINAMIDE) = 0.733 and the bedaquiline-containing triple = 1.32 – and the matching pair of interaction effects on relative bioavailability).
• Notes: Well-formed member of the auto-approved CONMED_<INN> family. Full INN spelling is used rather than the source’s 3-letter MOX, matching the CONMED_LOPINAVIR / CONMED_ATAZANAVIR precedent for keeping the CONMED_ namespace unambiguous. In multidrug tuberculosis regimens this indicator is normally one factor of a product of partner indicators rather than a standalone effect, because the regimens are given as fixed combinations (PaMZ, BPaMZ) and the individual partner contributions are not separately identifiable; encode the interaction as the literal product of indicators the source’s control stream forms, and name the coefficient with every participating drug token (e_moxifloxacin_pyrazinamide_cl). Distinct from a fluoroquinolone-class indicator, which is not registered.

CONMED_MPLA (canonical for coadministered monophosphoryl lipid A (TLR4 agonist) indicator)

• Description: 1 = monophosphoryl lipid A (MPLA; a detoxified lipid A derivative and toll-like receptor 4 agonist used as an immunostimulant and vaccine adjuvant) has been given alongside the antibacterial, 0 = no MPLA. In the founding model MPLA is a single 2.0 mg/kg intraperitoneal dose given together with oral amoxicillin 12 h after infection, and the column is TIME-VARYING: 0 before the MPLA dose and 1 from it onward, because the published lung-kill effect only reproduces when it switches on at treatment rather than at infection.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no MPLA).
• Source aliases:
  • MPLA – Franck 2021 binary covariate (‘MPLA coadministration (yes/no)’, Supplementary Section S2; MPLA_lung, kMPLA,spleen, MPLA_TTE and FC_AMX+MPLA in Tables 1 and 2).
• Example models: Franck_2021_amoxicillin_mouse.R (PK interaction only: CL/F = 124 + (-0.145 mL/h/ug) * DOSE_AMOXICILLIN_UG * CONMED_MPLA), Franck_2021_amoxicillin_mpla_mouse.R (the same PK interaction plus a 1.40-fold kkill in lung, a 3.71 1/h kill in spleen and a -1.32 log hazard ratio on survival).
• Notes: Member of the auto-approved CONMED_<INN> family. MPLA has no INN; the abbreviation is used throughout the source literature and in the licensed adjuvant systems (AS01, AS04), following the non-INN CONMED_HYALURONIDASE precedent. Only one MPLA dose level was studied in the founding paper and MPLA PK was not measured, so the column is a presence indicator, not an exposure; a future model with MPLA dose-response should register a DOSE_MPLA_<UNITS> sibling rather than overload this flag.

CONMED_PYRAZINAMIDE (canonical for concomitant pyrazinamide coadministration indicator)

• Description: 1 = subject is receiving pyrazinamide (a first-line antitubercular) as a regimen partner during the PK observation period, 0 = no concomitant pyrazinamide. Time-fixed for randomised fixed-combination regimen arms.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant pyrazinamide).
• Source aliases:
  • PZA – Salinger 2019 supplemental Table S3 $INPUT column; same orientation as the canonical.
• Example models: Salinger_2019_pretomanid.R (enters only through interaction products with CONMED_MOXIFLOXACIN and CONMED_BEDAQUILINE: e_moxifloxacin_pyrazinamide_cl = 0.733 and e_moxifloxacin_pyrazinamide_fdepot = 0.925 on the PaMZ regimen, and e_bedaquiline_moxifloxacin_pyrazinamide_cl = 1.32 with e_bedaquiline_moxifloxacin_pyrazinamide_fdepot = 1.25 on BPaMZ; no pyrazinamide-alone effect was retained).
• Notes: Well-formed member of the auto-approved CONMED_<INN> family. A CONMED_PYRAZINAMIDE = 1 record with no retained main effect is normal and not an encoding error – see the CONMED_MOXIFLOXACIN Notes on why fixed-combination tuberculosis regimens identify products of partner indicators rather than individual main effects. Distinct from DOSE_PYRAZINAMIDE_* should a future model need the partner’s dose level rather than its presence.

CONMED_BEDAQUILINE (canonical for concomitant bedaquiline coadministration indicator)

• Description: 1 = subject is receiving bedaquiline (a diarylquinoline ATP-synthase inhibitor) as a regimen partner during the PK observation period, 0 = no concomitant bedaquiline. Time-fixed for randomised fixed-combination regimen arms. Bedaquiline is a CYP3A4 substrate, so the indicator flags a possible metabolic interaction as well as a regimen contrast.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant bedaquiline).
• Source aliases:
  • BDQ – Salinger 2019 supplemental Table S3 $INPUT column; same orientation as the canonical.
• Example models: Salinger_2019_pretomanid.R (enters only through the bedaquiline + moxifloxacin + pyrazinamide triple product: e_bedaquiline_moxifloxacin_pyrazinamide_cl = 1.32 on apparent oral clearance and e_bedaquiline_moxifloxacin_pyrazinamide_fdepot = 1.25 on relative bioavailability. Because the effect is gated on the product, bedaquiline given WITHOUT moxifloxacin and pyrazinamide – the BPaL regimen of the Nix-TB study – carries no effect, which is how the model encodes Salinger 2019’s finding that bedaquiline and linezolid together had little impact on pretomanid exposure).
• Notes: Well-formed member of the auto-approved CONMED_<INN> family. The gating detail above is load-bearing and easy to get wrong: a data assembler that sets CONMED_BEDAQUILINE = 1 for a BPaL cohort is correct, and the model must be the thing that declines to apply an effect, not the data. Distinct from STUDY_C208, the bedaquiline registration-trial cohort indicator, which is a study flag rather than a comedication flag. Linezolid and clofazimine are further tuberculosis regimen partners that Salinger 2019 carried in its analysis data set; neither retained an effect, so neither is registered here.

CONMED_RIFAMPICIN (canonical for concomitant rifampicin (strong CYP3A4 inducer) coadministration indicator)

• Description: 1 = subject coadministered rifampicin (rifampin) during the observation interval (typically 600 mg once daily oral), 0 = no concomitant rifampicin. Rifampicin is the FDA index strong CYP3A4 inducer and also induces CYP2C9 / CYP2C19 / UGT and P-glycoprotein, so per-model notes should record which pathways the source attributes the effect to. Because induction requires enzyme turnover, the indicator is time-varying with a multi-day onset and offset when a source models the induction time course explicitly; time-fixed in fixed-sequence DDI arms that dose to steady-state induction before the victim dose. Distinct from [[CONMED_RPT]] (rifapentine at full CYP3A4 induction) and from the class-level [[CONMED_CYP3A4_IND]].
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant rifampicin).
• Source aliases:
  • rifampicin / rifampin – both spellings appear in source manuscripts (INN is rifampicin, USAN is rifampin); standardize the column name to CONMED_RIFAMPICIN. Used in Jaiswal_2025_dordaviprone.R (Jaiswal 2025 Figure 3 forest plot arm, rifampicin 600 mg once daily).
• Example models: Jaiswal_2025_dordaviprone.R (relative CYP3A4 activity 3.13, back-solved from the published dordaviprone AUC ratio of 0.167; this is the one modulator whose held-out Cmax ratio the reduction reproduces poorly, over-predicting 0.328 as 0.404), Huth_2019_siponimod.R (dual inducer: relative CYP3A4 activity 4.679 and relative CYP2C9 activity 2.181, back-solved jointly from the six published siponimod genotype AUC ratios; Huth 2019 separately shows that omitting the CYP2C9 induction raises the predicted AUC ratio from 0.32 to 0.59, which is what identifies the CYP2C9 arm), Nassar_2022_midazolam.R (time-resolved induction of intravenous midazolam clearance during 600 mg oral rifampicin every 24 h for 2 days in healthy volunteers, reference = pooled placebo: no effect up to 22 h after the first dose, then five fractional changes cl *= (1 + CONMED_RIFAMPICIN * e_conmed_rifampicin_cl_t<a>to<b>) of +27.9% (22, 24] h, +17.6% (24, 26], +23.7% (26, 28], +46.7% (28, 30] and +10.2% after 30 h; Nassar 2022 Table 1, time windows located by T_CONMED).
• Notes: Auto-approved member of the CONMED_<INN> family (INN = rifampicin). Sources that model the induction onset / washout time course should carry the covariate as time-varying rather than registering a separate steady-state-induction canonical.

CONMED_RIFABUTIN (canonical for concomitant rifabutin (moderate CYP3A4 inducer) coadministration indicator)

• Description: 1 = subject coadministered rifabutin during the observation interval (typically 300 mg once daily oral), 0 = no concomitant rifabutin. Rifabutin is the FDA index moderate CYP3A4 inducer and, like rifampicin, also induces UGT and P-glycoprotein, so per-model notes should record which pathway the source attributes the effect to. As with the rest of the rifamycin family the indicator is time-varying with a multi-day induction onset and offset when a source models the time course explicitly, and time-fixed in fixed-sequence DDI arms that dose to steady-state induction before the sampling visit.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant rifabutin).
• Source aliases:
  • Rifabutin co-administration – Kawuma 2023 Table 1 covariate row; used in Kawuma_2023_dolutegravir.R (rifabutin 300 mg once daily for 14 days, arm B of NCT01231542).
• Example models: Kawuma_2023_dolutegravir.R (linear fractional effect on the apparent central volume of distribution: vc * (1 + (-0.331) * CONMED_RIFABUTIN), i.e. a 33.1% decrease. The authors also fitted rifabutin on clearance (+36%), on bioavailability alone (not significant), and jointly on clearance and bioavailability (+41% and +32%); the volume effect was selected on goodness-of-fit, plausibility and BIC, and the mechanism is hypothesised to be induction of P-glycoprotein restricting tissue distribution rather than the metabolic induction that drives the rifampicin effect).
• Notes: Auto-approved member of the CONMED_<INN> family (INN = rifabutin). Registered separately from [[CONMED_RIFAMPICIN]] and [[CONMED_RPT]] (rifapentine) rather than under a pooled rifamycin-class or [[CONMED_CYP3A4_IND]] canonical, for the same reason the moderate and strong CYP3A4 inhibitor probes are kept apart: the whole point of a rifabutin arm is that it is a less potent inducer than rifampicin and is being assessed as a substitute for it, so a study that runs both arms is designed to measure exactly the contrast a pooled indicator would erase. In the founding model the two indicators do not even land on the same parameter – rifampicin acts on clearance and rifabutin on central volume. Use the class-level [[CONMED_CYP3A4_IND]] or [[CONMED_CYP3A4_IND_MOD]] only when the source itself pools inducers.

CONMED_VORICONAZOLE (canonical for concomitant voriconazole (strong CYP3A inhibitor) coadministration indicator)

• Description: 1 = subject coadministered voriconazole during the observation interval (typically 200 mg twice daily oral); 0 = no concomitant voriconazole. Voriconazole is both a reversible and a time-dependent CYP3A inhibitor and is the dominant tacrolimus / calcineurin-inhibitor perpetrator in solid-organ transplant care, where antifungal prophylaxis and treatment are routine. Time-varying when the observation record spans on / off voriconazole periods (the usual case in transplant TDM datasets); time-fixed in a fixed-sequence DDI arm.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant voriconazole).
• Source aliases:
  • Voriconazole – used in Pei_2023_tacrolimus.R (Pei 2023 Supplement 2.1.1: “Categorical covariates, such as SEX (male = 1 and female = 2) and Voriconazole (co-administration = 1 and none = 0), were included in the analysis using indicator variables”).
  • Voriconazole – used in Zhou_2025_tacrolimus.R (Zhou 2025 Table 1 counts 741 of 988 trough records on voriconazole; the azole antifungal drug (AFD) covariate carries a separate coefficient per agent).
• Example models: Pei_2023_tacrolimus.R (exponential effect on apparent oral clearance: exp(e_vori_cl * CONMED_VORICONAZOLE) with e_vori_cl = -0.64, i.e. a 47% reduction in tacrolimus CL/F on voriconazole; Pei 2023 Table S4), Zhou_2025_tacrolimus.R (exponential effect on apparent oral clearance: exp(e_vori_cl * CONMED_VORICONAZOLE) with e_vori_cl = -0.48, i.e. a 38.21% reduction in tacrolimus CL/F in Chinese lung-transplant recipients; Zhou 2025 Table 2 final model, back-transformed factor 0.62 in Eq. 4 – closely reproducing the 36.2% reduction reported by Cai et al. in an independent lung-transplant cohort), Jiang_2021_ivosidenib.R (multiplicative 0.64-fold change in ivosidenib CL/F; estimated alongside separate fluconazole and posaconazole coefficients).
• Notes: Auto-approved member of the CONMED_<INN> family (INN = voriconazole). Distinct from the class-level [[CONMED_AZOLE]] (Kirubakaran 2022 tacrolimus, which pools voriconazole with the other systemic azoles) and from [[CONMED_CYP3A4_INH_STRONG]]: register the drug-specific indicator when a paper singles voriconazole out with its own coefficient, because the magnitude of voriconazole’s tacrolimus interaction is much larger than the pooled azole class effect (Pei 2023’s own PBPK arm predicts a 5.80-fold rise in tacrolimus AUC). Siblings [[CONMED_ITRACONAZOLE]] and [[CONMED_POSACONAZOLE]]; use all three together when a paper resolves a separate coefficient per mould-active azole (Zhou 2025 lung-transplant tacrolimus). When a source instead supplies the perpetrator’s measured concentration rather than an on/off flag, use [[CONC_VORI_NGML]].

CONMED_VORICONAZOLE_ORAL, CONMED_VORICONAZOLE_IV (canonical for concomitant voriconazole split by the perpetrator’s administration route, carried as separate indicators)

• Description: A mutually exclusive pair splitting concomitant voriconazole by the route the perpetrator was given. CONMED_VORICONAZOLE_ORAL = 1 when the subject received oral voriconazole; CONMED_VORICONAZOLE_IV = 1 when the subject received intravenous voriconazole. Both are 0 when no voriconazole is given. A record with both set to 1 is not a valid state. The route qualifier refers to the co-administered voriconazole, never to the route of the modelled (victim) drug, which keeps its own ROUTE_* columns.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 for both (no concomitant voriconazole).
• Source aliases:
  • FLAGMED = 4 (voriconazole oral) and FLAGMED = 5 (voriconazole i.v.) – Nassar 2022 ESM NONMEM control stream perpetrator flag; CONMED_VORICONAZOLE_ORAL = (FLAGMED == 4), CONMED_VORICONAZOLE_IV = (FLAGMED == 5).
• Example models: Nassar_2022_midazolam.R (founding example; a single 400 mg voriconazole dose, orally or as a 2-hour intravenous infusion, 2 h into a microdosed intravenous midazolam infusion. Each route carries its own set of time-interval fractional changes in midazolam clearance, cl *= (1 + CONMED_VORICONAZOLE_ORAL * e_conmed_voriconazole_oral_cl_t<a>to<b> + CONMED_VORICONAZOLE_IV * e_conmed_voriconazole_iv_cl_t<a>to<b>), with the windows located by T_CONMED. Oral: -39.9% in the first hour, about -30% to 4 h, then -54% to -71%; intravenous: -11% to -17% to 4 h, then -36% to -61% (Nassar 2022 Table 1). The routes differ most in the first hours, which the authors attribute to the whole absorbed oral dose reaching the liver at once).
• Notes: Registered as a pair rather than as a route qualifier on [[CONMED_VORICONAZOLE]] because the source estimates a complete, separate set of effects per route; the two arms’ trajectories diverge early and converge late, so pooling them into one indicator would erase exactly the contrast the design was built to measure. Use the unqualified [[CONMED_VORICONAZOLE]] when a paper reports a single voriconazole effect irrespective of route. Shape precedent: the [[CONMED_RIFAMPICIN_SD]] / _MD pair (mutually exclusive qualified indicators of one perpetrator) and [[CONMED_QPRL_ORAL]] (a route-qualified concomitant indicator). Other perpetrators studied by route in the same design should register a parallel CONMED_<INN>_ORAL / CONMED_<INN>_IV pair. ### CONMED_LETERMOVIR (canonical for concomitant letermovir coadministration indicator)
• Description: 1 = the subject was receiving letermovir (cytomegalovirus prophylaxis after haematopoietic stem cell transplantation, typically 480 mg/day oral or intravenous) at the observation; 0 = no concomitant letermovir. Time-varying at the sample level.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant letermovir).
• Source aliases:
  • LMV – used in Suetsugu_2021_voriconazole.R (Suetsugu 2021 Eq. 4: ‘LMVi is concomitant letermovir use in subject i (used = 1, not used = 0)’).
• Example models: Suetsugu_2021_voriconazole.R (multiplicative effect on voriconazole Vmax, e_conmed_letermovir_vmax^CONMED_LETERMOVIR with 1.72; letermovir induces CYP2C19 / CYP2C9).
• Notes: Auto-approved member of the CONMED_<INN> family (INN = letermovir). General scope because letermovir is routine post-HSCT prophylaxis and a known perpetrator for voriconazole, tacrolimus and ciclosporin. Distinct from [[CMV_PROPHY_PRIOR]] (prior prophylaxis with an unrecorded agent), and from [[TX_HCT]].

CONMED_METHYLPREDNISOLONE (canonical for concomitant methylprednisolone coadministration indicator)

• Description: 1 = the subject was receiving systemic methylprednisolone at the observation; 0 = not. Time-varying at the sample level.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant methylprednisolone).
• Source aliases:
  • mPSL – used in Suetsugu_2021_voriconazole.R (Suetsugu 2021 Eq. 4: ‘mPSLi is concomitant methylprednisolone use in subject i (used = 1, not used = 0)’).
• Example models: Suetsugu_2021_voriconazole.R (multiplicative effect on voriconazole Vmax, e_conmed_methylprednisolone_vmax^CONMED_METHYLPREDNISOLONE with 1.30), Xie_2025_midazolam.R (covariatesDataExcluded).
• Notes: Auto-approved member of the CONMED_<INN> family (INN = methylprednisolone). Use this drug-specific indicator when a paper singles methylprednisolone out with its own coefficient; use the class-level [[CONMED_STEROID]] when corticosteroids are pooled, and [[PRED_DOSE]] when the dose magnitude enters the model.

CONMED_POSACONAZOLE (canonical for concomitant posaconazole (strong CYP3A4 / P-gp inhibitor) coadministration indicator)

• Description: 1 = subject coadministered posaconazole during the observation interval (oral suspension 200 mg three times daily, or delayed-release tablet 300 mg once daily after loading); 0 = no concomitant posaconazole. Posaconazole is a strong CYP3A4 inhibitor and, with voriconazole and itraconazole, one of the three mould-active azoles routinely used for antifungal prophylaxis and treatment in solid-organ transplant and haematology care. Time-varying when the observation record spans on / off posaconazole periods (the usual case in transplant TDM datasets); time-fixed in a fixed-sequence DDI arm.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant posaconazole).
• Source aliases:
  • Posaconazole – used in Zhou_2025_tacrolimus.R (Zhou 2025 Methods ‘General clinical data’ records azole antifungal drugs (AFDs) as a combined-medication category, and the final model estimates a separate AFD coefficient per agent; Table 1 counts 157 of 988 trough records on posaconazole).
  • POS – used in Chen_2022_tacrolimus.R (Chen 2022 Results, text below Eq. 7: “When a patient was treated with posaconazole, the POS value was set to 1; otherwise, the POS value was set to 0”).
• Example models: Zhou_2025_tacrolimus.R (exponential effect on apparent oral clearance: exp(e_posa_cl * CONMED_POSACONAZOLE) with e_posa_cl = -0.31, i.e. tacrolimus CL/F multiplied by 0.73 – Zhou 2025 reports the back-transformed factor as 0.74 in Eq. 4 and the reduction as 26.30% in the Discussion; Zhou 2025 Table 2 final model), Jiang_2021_ivosidenib.R (multiplicative 0.65-fold change in ivosidenib CL/F), Chen_2022_tacrolimus.R (fractional effect on tacrolimus CL/F: (1 + e_conmed_posaconazole_cl * CONMED_POSACONAZOLE) with e_conmed_posaconazole_cl = -0.57, i.e. a CL/F ratio of 1:0.43 in children with Crohn’s disease undergoing HSCT; Chen 2022 Table 2 and Eq. 6; formulation not reported).
• Notes: Auto-approved member of the CONMED_<INN> family (INN = posaconazole). Sibling to [[CONMED_VORICONAZOLE]] and [[CONMED_ITRACONAZOLE]]; register the drug-specific indicator rather than the class-level [[CONMED_AZOLE]] whenever a paper resolves a separate coefficient per azole, because the three agents differ substantially in CYP3A4 inhibitory potency (Zhou 2025 Discussion, citing equimolar-potency ranking ketoconazole > itraconazole > voriconazole > fluconazole, and finding tacrolimus CL/F reductions of 57.98% for itraconazole, 38.21% for voriconazole and 26.30% for posaconazole in the same cohort). The three drug-specific indicators are mutually exclusive in the Zhou 2025 dataset (no recipient received two mould-active azoles concurrently), but the canonical does not assume exclusivity – a model that needs to forbid overlap should say so in covariateData[[CONMED_POSACONAZOLE]]$notes. Formulation matters: the oral suspension has markedly lower and more variable bioavailability than the delayed-release tablet, so the magnitude of the tacrolimus interaction is formulation-dependent (Zhou 2025 Discussion: ‘the patients enrolled in our study mainly used posaconazole as suspension, which shows smaller impact than delayed-release tablets’); record the formulation used in the per-model notes. Distinct from [[FORM_POSA_AB]] and [[STUDY_POSA_PHASE3]], which are covariates of posaconazole as the victim drug in the van Iersel 2018 posaconazole popPK analysis rather than indicators of posaconazole as a perpetrator.

CONMED_CASPOFUNGIN (canonical for concomitant caspofungin coadministration indicator)

• Description: 1 = subject coadministered caspofungin (an echinocandin antifungal) during the observation interval; 0 = no concomitant caspofungin. Time-varying per subject because antifungal courses start and stop. Caspofungin is not a meaningful CYP3A inhibitor, so unlike the azoles ([[CONMED_AZOLE]]) it is not expected to lower CYP3A-substrate clearance; a fitted effect usually reflects the clinical state of patients needing empirical echinocandin therapy.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant caspofungin).
• Source aliases:
  • CPFG – used in Liu_2022_tacrolimus.R and Liu_2022_tacrolimus_cyp3a5.R (Liu 2022 Table 3 ‘theta CPFG, CL’). Same orientation, no value transformation.
• Example models: Liu_2022_tacrolimus.R (exponential effect exp(0.16 * CONMED_CASPOFUNGIN) on tacrolimus CL, i.e. 17% higher CL, in pediatric HSCT recipients; Liu 2022 Eq. 4), Liu_2022_tacrolimus_cyp3a5.R (exp(0.04 * CONMED_CASPOFUNGIN); Eq. 6).
• Notes: Auto-approved member of the CONMED_<INN> family (INN = caspofungin).

CONMED_AMLODIPINE (canonical for concomitant amlodipine coadministration indicator)

• Description: 1 = subject is receiving amlodipine at the observation; 0 = not. Amlodipine is a dihydropyridine calcium-channel blocker and a weak CYP3A4 inhibitor, and is one of the most common antihypertensives in solid-organ transplant care, where post-transplant hypertension is near-universal. Time-varying when the record spans start / stop of amlodipine; time-fixed when the source paper records only a per-subject yes/no.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant amlodipine).
• Source aliases:
  • Amlodipine – used in Xiang_2025_tacrolimus_fpg.R (Xiang 2025 Table 1 co-medication row and Supplementary Table 3 covariate screen).
• Example models: Xiang_2025_tacrolimus_fpg.R (declared in covariatesDataExcluded: amlodipine reached the forward-inclusion threshold on baseline fasting plasma glucose, dOFV = -8.613, p < 0.01, but failed backward elimination at p < 0.001 and carries no published point estimate, so it is documented rather than used).
• Notes: Auto-approved member of the CONMED_<INN> family (INN = amlodipine). Distinct from a class-level calcium-channel-blocker indicator and from [[CONMED_CYP3A4_INH]] / [[CONMED_CYP3A4_INH_MOD]]: register the drug-specific indicator when a paper singles amlodipine out with its own screening step or coefficient. Amlodipine’s own CYP3A4 inhibition is weak relative to the azoles, so a significant amlodipine term in a tacrolimus or calcineurin-inhibitor model is as likely to be a marker of blood-pressure or graft-function status as a direct drug-drug interaction.

CONMED_WUZHI (canonical for concomitant Wuzhi capsule (Schisandra sphenanthera extract) coadministration indicator)

• Description: 1 = subject is receiving the Wuzhi capsule at the observation; 0 = no concomitant Wuzhi capsule. Wuzhi capsule is a standardised ethanolic extract of Schisandra sphenanthera fruit (principal lignan schisantherin A, with deoxyschisandrin and schisandrin B) marketed in China and routinely co-prescribed with tacrolimus in solid-organ transplant care as a deliberate pharmacokinetic booster: its lignans inhibit CYP3A4/CYP3A5 and P-glycoprotein, raising tacrolimus exposure and allowing a lower tacrolimus dose. Time-varying when the observation record spans on / off Wuzhi periods; time-fixed when the source paper analyses a fixed co-medication assignment.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant Wuzhi capsule).
• Source aliases:
  • WZ – used in Xiang_2025_tacrolimus.R (Xiang 2025 Eq. 6: “WZ = 1 if WZ is present; otherwise = 0”) and Chen_2020b_tacrolimus.R (Chen 2020 text below Table 1: “for patients co-administered WZ, the WZ value = 1, otherwise WZ value = 0”).
• Example models: Xiang_2025_tacrolimus.R (exponential effect on apparent oral clearance: exp(e_conmed_wuzhi_cl * CONMED_WUZHI) with e_conmed_wuzhi_cl = -0.211, i.e. a 19% reduction in tacrolimus CL/F on Wuzhi capsule; Xiang 2025 Table 2 and Eq. 6), Xiang_2025_tacrolimus_fpg.R, Xiang_2025_tacrolimus_egfr.R (the same effect carried as a fixed PK parameter in the sequential PK/PD steps), Chen_2020b_tacrolimus.R (fractional effect on CL/F: (1 + e_conmed_wuzhi_cl * CONMED_WUZHI) with e_conmed_wuzhi_cl = -0.108; Chen 2020 Table 3 and Eq. 7), Chen_2020a_tacrolimus.R (fractional effect on apparent oral clearance: (1 + e_conmed_wuzhi_cl * CONMED_WUZHI) with e_conmed_wuzhi_cl = -0.290, i.e. a CL/F ratio of 1:0.71 in pediatric lupus nephritis; Chen 2020 Table II and equation vi).
• Notes: Member of the CONMED_<agent> family; the token is the product name rather than an INN because Wuzhi capsule is a multi-constituent herbal preparation with no single INN. Distinct from the class-level [[CONMED_CYP3A4_INH]] and from [[CONMED_CYP3A4_INH_MOD]]: register the product-specific indicator when a paper singles the Wuzhi capsule out with its own coefficient, as the preparation is dosed and regulated as a discrete product in Chinese transplant practice rather than as a member of a pooled inhibitor class. Because Wuzhi capsule is prescribed in order to raise tacrolimus exposure, it is confounded with genotype in observational cohorts – Xiang 2025 notes it is “often prescribed for patients with CYP3A51/1 but rarely for patients with CYP3A53/3” – so a model carrying both this indicator and [[CYP3A5_EXPR]] should be simulated over the covariate combinations that actually occur in the source cohort.

CONMED_SDT (canonical for concomitant Sipjeondaebo-tang (Shi-Quan-Da-Bu-Tang) polyherbal decoction indicator)

• Description: 1 = subject received Sipjeondaebo-tang (SDT) over the pretreatment window preceding the observation; 0 = vehicle control. SDT – Shi-Quan-Da-Bu-Tang in Chinese, Juzen-taiho-to in Japanese – is a ten-herb decoction (Ginseng Radix Alba, Atractylodis Rhizoma Alba, Poria Sclerotium, Glycyrrhizae Radix, Angelicae Gigantis Radix, Cnidii Rhizoma, Rehmanniae Radix Preparata, Paeoniae Radix, Astragali Radix, Cinnamomi Cortex) prescribed in Korean, Chinese and Japanese traditional medicine for post-illness debility and widely co-administered with cytotoxic chemotherapy as an antiemetic and anti-anorexic adjuvant. It is a perpetrator of herb-drug interactions on the absorption axis rather than the clearance axis: repeated dosing slowed oral absorption of tegafur and, more consequentially, cut absorption of the co-formulated dihydropyrimidine-dehydrogenase inhibitor gimeracil, so that 5-FU was degraded faster. Duration is load-bearing – a single SDT dose produced no significant change, while seven daily doses did, so the indicator should record the repeated-dose regimen a paper actually studied rather than any SDT exposure.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (vehicle control; 1% sodium carboxymethylcellulose in the founding model).
• Source aliases:
  • Con / Pretre – Kim 2017 Table 2 parameter subscripts distinguishing the control and SDT-pretreated estimates (Ka,Con vs Ka,Pretre), which is how that paper encodes the arm rather than as a named data column.
  • SDT – Kim 2017 figure legends and Results prose.
• Example models: Kim_2017_tegafur_rat.R (selects between separately estimated control and SDT-pretreated values of the tegafur absorption rate constant Ka, the pre-systemic 5-FU-precursor formation rate constant Ka,Met, and the apparent 5-FU clearance CL5FU/F; Kim 2017 Table 2, doi:10.3390/molecules22091488).
• Notes: Member of the CONMED_<agent> family; like [[CONMED_WUZHI]] the token is the preparation name rather than an INN, because a multi-constituent herbal decoction has no single INN. Distinct from [[CONMED_WUZHI]] in mechanism as well as composition: Wuzhi capsule is a deliberate CYP3A/P-gp-inhibiting exposure booster, whereas SDT’s demonstrated effect is reduced absorption of a co-administered agent. Where a model carries the vehicle itself as a covariate, note that the founding paper found the control vehicle (carboxymethylcellulose) was not inert for gimeracil across single- versus multiple-dose control groups, so an SDT indicator does not absorb vehicle effects.

CONMED_LCM (canonical for lacosamide (monotherapy or coadministration) indicator)

• Description: 1 = subject is on lacosamide (LCM) at the observation, 0 = not on lacosamide. In an active-controlled monotherapy trial (SP0993) the indicator identifies the LCM arm (the CBZ-CR arm gets 0). In adjunctive-treatment cohorts the same canonical would flag concomitant lacosamide use. Time-varying is permitted when a study spans on / off transitions; time-fixed in monotherapy parallel-group trials.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (not on lacosamide; in monotherapy head-to-head trials this is the comparator arm, e.g., carbamazepine controlled-release in Lindauer 2017 SP0993).
• Source aliases:
  • TYPE – used in Lindauer_2017_lacosamide_dropout.R and Lindauer_2017_lacosamide_seizure.R (Lindauer 2017 SP0993 encoding: TYPE = 1 for the lacosamide arm, 0 for the CBZ-CR arm; same value semantics as the canonical CONMED_LCM column).
• Example models: Lindauer_2017_lacosamide_dropout.R (multiplicative log-hazard shift on the dropout hazard: hazard *= exp(-0.138 * CONMED_LCM), HR = 0.871 vs the CBZ-CR arm, Lindauer 2017 Table 2 Coeff_TYPE), Lindauer_2017_lacosamide_seizure.R (gates the drug-specific AUC covariate contributions – AUC_LCM centred deviation is active only when CONMED_LCM = 1, AUC_CBZ centred deviation active only when CONMED_LCM = 0 – and gates the AGE covariate effect on the first-seizure hazard which is LCM-only per Lindauer 2017 Section 3.4).
• Notes: Auto-approved member of the CONMED_<INN> family. Follows the CONMED_* concomitant-medication pattern (CONMED_CBZ, CONMED_EFV, CONMED_AZA, etc.). Semantically the indicator captures “is the subject taking lacosamide at this record”; in a monotherapy parallel-group trial that equates to arm membership, and in adjunctive or cross-over designs it captures the on / off status.

CONMED_KETAMINE (canonical for concomitant ketamine anaesthesia indicator)

• Description: 1 = subject received ketamine (typically intramuscular, as a short-acting dissociative anaesthetic for restraint or sampling) over the observation interval, 0 = no ketamine. The column exists because ketamine is a perpetrator on two fronts at once: it is a CYP3A4/CYP2B6/CYP2C9 substrate and competitive inhibitor, and – more important preclinically – it slows gastrointestinal transit and alters regional blood flow, so it can move absorption-phase parameters (transit rate, lag, Tmax) as much as it moves clearance. In laboratory-animal PK this is a procedural covariate, not a therapeutic co-medication: animals are anaesthetised to permit blood sampling, so the indicator usually partitions studies or sites rather than individuals. Time-varying where a study spans anaesthetised and unanaesthetised sampling occasions; time-fixed (effectively a site / protocol indicator) where a whole study used one regime.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no ketamine anaesthesia).
• Source aliases:
  • Anesthesia – the Reddy 2015 Tables 1 and 4 column heading, whose levels are Ketamine, None and Saffan; the canonical binary is 1 for the Ketamine level. A study using a different anaesthetic (Saffan in Reddy 2015 Study 4, a pregnanedione preparation with a quite different effect on Tmax) must not be folded into this column – give it its own indicator or exclude it, as Reddy 2015 did.
• Example models: Reddy_2015_oseltamivir_ferret.R (declared in covariatesDataExcluded: the source screened anaesthesia by Kruskal-Wallis test on individual post-hoc parameter estimates and found Kt, Ka, CLt, Vc and Vp all significantly shifted, but never estimated a ketamine coefficient – it reported two stratified parameter sets and carried only the ketamine one forward, so the covariate is documented rather than structurally applied).
• Notes: Auto-approved member of the CONMED_<INN> family (INN = ketamine). Because the anaesthetic is given to enable the sampling, the indicator is frequently confounded with site, investigator, dosing vehicle and animal colony; the founding example says so explicitly. Record that confounding in the per-model notes rather than implying a causal ketamine effect. Distinct from [[ETSEVO]] (end-tidal sevoflurane concentration, a graded inhalational-depth covariate) and from [[T_ANAESTHESIA]] (time from anaesthesia induction to the modelled treatment); use those when the source measures anaesthetic depth or timing instead of exposure to a named agent.

CONMED_LPV (canonical for concomitant lopinavir co-administration indicator)

• Description: 1 = subject is receiving concomitant lopinavir as part of an antiretroviral regimen at the observation, 0 = not on lopinavir. Captures the LPV-ritonavir DDI on ritonavir apparent oral clearance: when LPV/r is co-administered, ritonavir CL/F is roughly 2.7-fold higher than when ritonavir is given without lopinavir (Kappelhoff 2005 attributes the increase to lopinavir-driven enzyme induction superimposed on ritonavir’s own CYP3A4 inhibition). Distinct from the joint rifampicin + super-boosted-LPV/r 4:4 indicator CONMED_RIF_LPVR4 and from the cholesterol-surrogate use in Archary_2018_lopinavir.R; this canonical is the binary “is the subject on lopinavir at all” flag, not a regimen-specific contrast.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant lopinavir).
• Source aliases:
  • LPV – used in Kappelhoff_2005_ritonavir.R (paper Results equation following Table 2: CL/F = 10.5 * 2.72^LPV).
• Example models: Kappelhoff_2005_ritonavir.R (multiplicative power-form effect on apparent oral ritonavir CL/F: cl = exp(lcl + etalcl) * e_lpv_cl^CONMED_LPV with e_lpv_cl = 2.72; 36 of 186 subjects on LPV/r in the source cohort), Hoglund_2015_lumefantrine.R (TWO independent linear-deviation effects: on lumefantrine apparent CL/F cl *= (1 + (-0.621) * CONMED_LPV) (-62.1 %, CYP3A4 inhibition by ritonavir) and on desbutyl-lumefantrine apparent CL/F cl_desbutlum *= (1 + 3.92 * CONMED_LPV) (+392 %; mechanism unknown, Hoglund 2015 Discussion); Table 2), Hoglund_2015_artemether.R (TWO independent linear-deviation effects: on artemether apparent CL/F cl *= (1 + 0.328 * CONMED_LPV) (+32.8 %) and on dihydroartemisinin apparent CL/F cl_dha *= (1 + 1.43 * CONMED_LPV) (+143 %; attributed to CYP2B6 / 2C induction by lopinavir, Hoglund 2015 Discussion); Table 3), Kay_2020_lumefantrine.R (TWO independent linear-deviation effects, on lumefantrine apparent CL/F: cl *= (1 + (-0.514) * CONMED_LPV) (-51.4 %, CYP3A4 inhibition by the ritonavir booster) and on the first-order absorption rate constant: ka *= (1 + (-0.212) * CONMED_LPV) (-21.2 %); relative to the no-ART reference in HIV-infected Ugandan children with malaria (ASTMH 2020 poster 2167 Table 1)), Francis_2020_lumefantrine.R (three fractional effects of lopinavir-ritonavir-based ART on lumefantrine: CL/F x (1 - 0.501), F x (1 + 0.672), ka x (1 - 0.476); ~3.4-fold higher AUC).
• Notes: Follows the CONMED_* concomitant-medication pattern (CONMED_AZA, CONMED_RIF, CONMED_EFV, CONMED_CBZ, CONMED_AMIO, etc.). The Kappelhoff 2005 cohort tested concomitant saquinavir and indinavir on the same parameter via the same univariate procedure; neither was retained, so this canonical captures the only co-PI covariate kept in that paper’s final model. Time-fixed within an evaluated regimen in the source cohort (the dataset records baseline LPV/r vs non-LPV/r assignment); a time-varying form is permitted for cohorts with on / off regimen transitions and should be documented in covariateData[[CONMED_LPV]]$notes.

CONMED_METFORMIN (canonical for concomitant metformin co-administration indicator)

• Description: 1 = subject is on concomitant metformin during the modelled treatment period, 0 = not. Time-fixed in source datasets where metformin is a study-design “add-on” arm (e.g., Retlich 2015 Study 4 add-on-to-metformin design); permits a time-varying form for cohorts with on/off metformin transitions, document per-model via covariateData[[CONMED_METFORMIN]]$notes.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant metformin).
• Source aliases: none known.
• Example models: Retlich_2015_linagliptin.R (1 = study 4 add-on-to-metformin cohort; multiplicative effect on linagliptin relative bioavailability F: +69% F for metformin co-administration vs the monotherapy reference; the effect is attributed to a metformin – linagliptin drug-drug interaction consistent with a separately published DDI study, Graefe-Mody 2009).
• Notes: Follows the CONMED_* concomitant-medication pattern (AZA / MP / MTX / AMINO / NSAID / PARA / AD / RITUX / AED / CHEMO / EIAED / EFV / AZOLE). Metformin is a widely-co-prescribed first-line T2DM oral antidiabetic; future T2DM-popPK / -DDI extractions should reuse this canonical.

CONMED_MP (canonical for concomitant 6-mercaptopurine)

• Description: 1 = on concomitant 6-mercaptopurine (6-MP), 0 = not.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant 6-MP).
• Source aliases: MP – used in Rosario_2015_vedolizumab.R.
• Example models: Rosario_2015_vedolizumab.R (power-form on CLL: CLL * 1.04^CONMED_MP).
• Notes: Second thiopurine immunomodulator used in IBD maintenance; typically mutually exclusive with CONMED_AZA for a given subject.

CONMED_SULFONYLUREA (canonical for concomitant sulfonylurea (class) co-administration indicator)

• Description: 1 = subject is on a concomitant sulfonylurea (class indicator pooling glyburide / glibenclamide / glipizide / glimepiride / gliclazide / tolbutamide / glipizide-ER / chlorpropamide and other oral insulin-secretagogue sulfonylureas), 0 = not. Time-fixed in source datasets where the sulfonylurea is a study-design background-therapy arm (e.g., the Baron 2016 metformin + sulfonylurea cohort); permits a time-varying form when a study tracks on / off sulfonylurea transitions.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant sulfonylurea).
• Source aliases:
  • SU – used in Baron_2016_empagliflozin.R (Baron 2016 Methods and Table S2 ‘background sulfonylurea’ cohort; multiplicative effects on the model-predicted baseline FPG and on Gmax in Table S3).
• Example models: Baron_2016_empagliflozin.R (multiplicative effect on BFPG: bfpg_su^CONMED_SULFONYLUREA = 1.01^...; on Gmax: gmax_su^CONMED_SULFONYLUREA = 1.27^... – sulfonylurea co-treatment increases empagliflozin’s maximal FPG-lowering effect, attributed by the authors to possible improvement in beta-cell function).
• Notes: Class-level pooling indicator analogous to CONMED_AZOLE, CONMED_NSAID, CONMED_STATIN, CONMED_AED, etc. Future T2DM popPK / PD extractions that need to distinguish individual sulfonylureas (e.g., for a DDI sub-analysis on a CYP2C9-metabolized member) should register a per-INN sibling canonical (CONMED_GLYBURIDE, CONMED_GLIPIZIDE,…) and keep CONMED_SULFONYLUREA for the class-pooled summary.

CONMED_PIOGLITAZONE (canonical for concomitant pioglitazone co-administration indicator)

• Description: 1 = subject is on concomitant pioglitazone (thiazolidinedione (TZD) class oral antidiabetic), 0 = not. Time-fixed in source datasets where pioglitazone is a study-design background-therapy arm; permits a time-varying form when a study tracks on / off transitions.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant pioglitazone).
• Source aliases:
  • PIO – used in Baron_2016_empagliflozin.R (Baron 2016 Methods and Study 7 background pioglitazone cohort; multiplicative effects on the model-predicted baseline FPG and on Gmax in Table S3).
• Example models: Baron_2016_empagliflozin.R (multiplicative effect on BFPG: bfpg_pio^CONMED_PIOGLITAZONE = 0.999^...; on Gmax: gmax_pio^CONMED_PIOGLITAZONE = 1.02^... – both effects are not statistically significant in the source analysis but are retained per the full-covariate-modelling approach).
• Notes: Follows the per-INN CONMED_<INN> pattern (CONMED_METFORMIN, CONMED_LPV, CONMED_ATAZANAVIR, CONMED_RIF, CONMED_PROBENECID, etc.). Distinct from a putative future CONMED_TZD (class indicator) – the thiazolidinedione class additionally contains rosiglitazone (largely withdrawn) and lobeglitazone (Korean market); pool a TZD class indicator only if a future paper aggregates them.

CONMED_MTX (canonical for concomitant methotrexate)

• Description: 1 = on concomitant methotrexate, 0 = not.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant methotrexate).
• Source aliases: MTX – used in Rosario_2015_vedolizumab.R.
• Example models: Rosario_2015_vedolizumab.R (power-form on CLL: CLL * 0.983^CONMED_MTX).
• Notes: Immunomodulator used especially in CD maintenance. Generic concomitant-MTX indicator that may also appear in non-IBD models; start as scope: general.

CONMED_NSAID (canonical for concomitant NSAID use)

• Description: 1 = on concomitant non-steroidal anti-inflammatory drug (NSAID) therapy at baseline, 0 = not.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant NSAID use; typical patient).
• Source aliases:
  • NSAID – used in Li_2019_abatacept.R.
• Example models: Li_2019_abatacept.R (exponential effect on CL: CL * exp(0.0640 * CONMED_NSAID); ~6.6% higher CL, not clinically relevant per Li 2019).
• Notes: Baseline-use-only in Li 2019; time-varying use is permitted, document per-model. Follows the CONMED_* concomitant-medication pattern established for IBD models (AZA / MP / MTX / AMINO).

CONMED_OPIOID (canonical for concomitant mu-opioid analgesic indicator)

• Description: 1 = record taken while the subject was on a concomitant mu-opioid receptor agonist analgesic (any agent of the class), 0 = not. Class-level indicator in the same sense as CONMED_NSAID / CONMED_PPI; time-varying per record in the founding model.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant opioid analgesic).
• Source aliases:
  • mu-opioids – Xiong 2022 Table 3 row label (‘mu-Opioids covariate on CLpar/F’).
• Example models: Xiong_2022_tepotinib.R (fractional effect on tepotinib apparent clearance, CL/F * (1 - 0.167 * CONMED_OPIOID); 24% of tepotinib observations were under concomitant opioids, and the paper attributes the effect to opioid slowing of gastrointestinal motility, with no clinically meaningful change in AUC).
• Notes: Use a CONMED_<INN> indicator instead when a paper resolves the individual opioid. Distinct from DIS_OUD (opioid use disorder disease state) and from the Mann 2022 opioid-receptor PD inputs (CAR_OPIOID, L_OPIOID_pM, OPIOID_ID), which carry receptor occupancy or ligand identity rather than a co-medication flag.

CONMED_NVP (canonical for concomitant nevirapine indicator)

• Description: 1 = subject is receiving nevirapine (NVP) – typically as the non-nucleoside reverse-transcriptase inhibitor (NNRTI) third agent in an HIV combination ART regimen – coadministered with the modelled antiretroviral; 0 = subject is not on nevirapine. Nevirapine is a CYP3A inducer; the indicator is used to flag PXR-mediated induction of metabolic clearance for coadministered CYP3A-cleared antiretrovirals (e.g., the protease inhibitor lopinavir).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no nevirapine; paper-defined ART backbone excluding NVP).
• Source aliases:
  • N – in-equation indicator used by Jullien 2006 (final covariate submodel: CL/F =... * 1.34^N with N = 1 if nevirapine combined with lopinavir).
  • NVP – standard 3-letter HIV abbreviation in NONMEM control streams from pediatric-ART cohorts; same orientation, no value transformation.
• Example models: Jullien_2006_lopinavir.R (exponential effect on CL/F: cl *= exp(log(1.34) * CONMED_NVP); +34% CL/F when nevirapine is coadministered), Hoglund_2015_lumefantrine.R (multiplicative linear-deviation effect on lumefantrine relative bioavailability F: fdepot_typ *= (1 + (-0.248) * CONMED_NVP); -24.8% relative to the no-ART reference in HIV-infected Ugandan adults, Hoglund 2015 Table 2), Hoglund_2015_artemether.R (two independent linear-deviation effects: on artemether relative bioavailability F fdepot_typ *= (1 + (-0.663) * CONMED_NVP) (-66.3 %) and on apparent dihydroartemisinin CL/F cl_dha *= (1 + (-0.445) * CONMED_NVP) (-44.5 %); Hoglund 2015 Table 3), Svensson_2014_bedaquiline_nvp.R (power-form effects on apparent CL of bedaquiline and its M2 metabolite at full induction: cl *= 0.915 ^ CONMED_NVP and cl_m2 *= 1.05 ^ CONMED_NVP; Svensson 2014 Supplementary Table S1b ‘EFF NVP on BDQ CL = 0.915’ and ‘EFF NVP on M2 CL = 1.05’), Kay_2022_lumefantrine.R (TWO independent linear-deviation effects: on apparent oral lumefantrine CL/F cl *= (1 + 0.0191 * CONMED_NVP) (+1.91 %; not statistically significant, 95% CI -0.324 to 0.362) and on the first-order absorption rate constant ka *= (1 + (-0.0589) * CONMED_NVP) (-5.89 %; not statistically significant, 95% CI -0.207 to 0.0891); HIV-uninfected children are the reference (CONMED_NVP = 0) in a pediatric Ugandan malaria cohort, Kay 2022 Table 2 theta_9 and theta_12), Kay_2020_lumefantrine.R (TWO independent linear-deviation effects, on lumefantrine apparent CL/F: cl *= (1 + 0.0191 * CONMED_NVP) (+1.9%) and on the first-order absorption rate constant: ka *= (1 + (-0.0589) * CONMED_NVP) (-5.9%); both retained in the final model despite 95% CIs spanning zero (ASTMH 2020 poster 2167 Table 1)).
• Notes: Scope promoted from specific to general on 2026-06-27 alongside the Svensson 2014 bedaquiline-NVP extraction (the fourth model ratifying the canonical: Jullien 2006, Hoglund 2015 lumefantrine, Hoglund 2015 artemether, Svensson 2014). Sister indicator to the registered CONMED_EFV (efavirenz) entry, applying to the second commonly co-administered NNRTI. Future ART popPK models that test an NVP-vs-comparator contrast should extend the example list when the comparator matches.

CONMED_PARA (canonical for concomitant paracetamol (acetaminophen) use)

• Description: 1 = on concomitant paracetamol (acetaminophen) at the observation, 0 = not.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant paracetamol).
• Source aliases:
  • PCM – used in Plan_2012_pain.R (DDMORE Foundation Model Repository entry DDMODEL00000194).
• Example models: Plan_2012_pain.R (additive shift on the placebo-effect logit of the typical pain-score lambda: phl = logit(TVLAM) + 0.364 * CONMED_PARA; mean pain score ~9% higher on the 0-10 Likert scale during paracetamol use).
• Notes: Distinct from CONMED_NSAID – paracetamol is not classed as an NSAID (no anti-inflammatory mechanism, distinct AE profile). Time-varying use is permitted; the daily Likert measurements in Plan 2012 carry PCM as a per-observation flag. Follows the CONMED_* concomitant-medication pattern established for IBD models (AZA / MP / MTX / AMINO) and CONMED_NSAID (Li 2019).

CONMED_PEGIFN_RBV (canonical for concomitant peginterferon-alfa plus ribavirin backbone indicator)

• Description: 1 = the study drug is co-administered with the peginterferon-alfa plus ribavirin (pegIFN/RBV) backbone; 0 = it is not. Encodes the legacy interferon-containing hepatitis C regimen as a treatment-design covariate. The two agents are registered as one indicator rather than two because pegIFN and RBV are given together as a fixed backbone and are never separately identifiable in an HCV trial cohort.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 – paper-defined complement. For Osawa 2018 the reference is the all-oral DUAL regimen (daclatasvir + asunaprevir), stated there as TXref = asunaprevir.
• Source aliases:
  • TX (treatment description) – Osawa 2018 source-column convention; the Table 1 levels are “daclatasvir + PegIFN/RBV” (-> 1) and “daclatasvir + asunaprevir (DUAL)” (-> 0).
• Example models: Osawa_2018_daclatasvir.R (exponential effect -0.122 on daclatasvir CL/F, i.e. apparent clearance ~11% lower on the pegIFN/RBV backbone than on the DUAL regimen; 21.1% of the cohort).
• Notes: Check the reference level before reusing the coefficient’s sign. This indicator’s complement is not a no-treatment group but whatever comparator regimen the paper ran, and in the direct-acting-antiviral era that comparator is itself an active all-oral combination. Osawa 2018’s reference is the DUAL regimen, so its coefficient is a pegIFN/RBV-versus-DUAL contrast, not a pegIFN/RBV-versus-nothing one; a paper comparing pegIFN/RBV against monotherapy would produce a differently-signed number for the same column. Put the comparator in covariateData[[CONMED_PEGIFN_RBV]]$notes. Composite-regimen member of the CONMED_ family, following CONMED_AVD (brentuximab vedotin + AVD) and CONMED_CHEMO (anti-PD-(L)1 + platinum chemotherapy) in naming a multi-agent backbone with a single indicator. Distinct from CONMED_IFNALPHA (non-pegylated interferon alfa alone) and CONMED_IFNB1A (interferon beta-1a, a multiple-sclerosis therapy); distinct also from DOSE_RBV_MGD and T_PEGIFN, which quantify the ribavirin dose and the planned pegIFN course duration rather than flagging coadministration. Specific scope because the comparator regimen is paper-defined; promote to general once a second paper ratifies the column against a comparable reference.

CONMED_PAROXETINE (canonical for concomitant paroxetine coadministration indicator)

• Description: 1 = subject is co-prescribed paroxetine (SSRI antidepressant; a potent mechanism-based inhibitor of CYP2D6 and itself a CYP2D6 substrate) at the PK observation, 0 = no concomitant paroxetine. Time-fixed at the analysis baseline in the founding source, which draws comedication status from a retrospective medical record system; time-varying encoding is appropriate wherever paroxetine start / stop events are captured.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no concomitant paroxetine).
• Source aliases:
  • PAR – used in Zhang_2025_olanzapine.R (Zhang 2025 Equation 6: “PAR represents paroxetine; when patients took paroxetine, PAR was 1, otherwise PAR was 0”; same 0/1 orientation).
• Example models: Zhang_2025_olanzapine.R (linear-shift effect on apparent oral clearance: cl *= (1 + e_par_cl * CONMED_PAROXETINE) with e_par_cl = -0.289, i.e. concomitant paroxetine lowers olanzapine CL/F by 28.9%, equivalently a with:without clearance ratio of 0.711:1 as stated in the Results and Discussion and plotted in Figure 1H; founding example).
• Notes: Paroxetine appears in the popPK literature on both sides of the DDI – as the victim drug (Kim_2015_paroxetine.R, Shigetome_2025_paroxetine.R, where paroxetine’s own disposition is modelled) and, here, as the perpetrator on a co-prescribed CYP2D6 substrate. Use CONMED_PAROXETINE only for the perpetrator role; a model of paroxetine itself does not need a covariate for the drug being modelled. Distinct from CYP2D6_PM: the comedication indicator captures a phenoconversion-like exposure shift in patients of unrecorded genotype, whereas CYP2D6_PM captures the germline poor-metaboliser phenotype, and a study measuring both should carry both columns. Parallel in construction to CONMED_ARIPIPRAZOLE (Zhang_2024_olanzapine.R), the same group’s companion analysis of olanzapine in schizophrenia; the two models differ in perpetrator drug and population but share the structural model and the linear-shift covariate form. Specific scope until a second model registers paroxetine as a covariate; promote to general once that happens. Zhang 2025’s cohort prevalence was 18 of 72 patients (25.0%, Table 2), so the -0.289 estimate rests on a reasonably sized exposed subgroup, but its bootstrap 95% interval [-0.420, -0.078] still reaches close to no effect and should be quoted alongside the point estimate.

CONMED_PACLITAXEL (canonical for concomitant paclitaxel coadministration indicator)

• Description: 1 = the modelled drug’s observation was collected while the patient was concurrently receiving paclitaxel as the co-administered antineoplastic, 0 = not on concomitant paclitaxel. A per-INN member of the CONMED_* family, used when a source paper resolves paclitaxel as its own co-administration stratum rather than pooling it into a generic chemotherapy-backbone indicator.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (not on concomitant paclitaxel). The substantive meaning of the zero level is set by the source paper’s stratification and must be documented per model – in Zuo_2024_apatinib.R the all-zero reference is “apatinib co-administered with an immune-checkpoint-inhibitor monoclonal antibody”, not “no co-medication”.
• Source aliases:
  • CM (category B) – used in Zuo_2024_apatinib.R (Zuo 2024 Table 1 footnote defines four mutually exclusive co-administration groups A/B/C/D; B is paclitaxel).
• Example models: Zuo_2024_apatinib.R (multiplicative effect on apparent clearance: CL/F * 0.58^CONMED_PACLITAXEL, i.e. 42% lower CL/F on concomitant paclitaxel relative to the anti-PD-1 mAb reference group; Zuo 2024 Table 2, RSE 9%, bootstrap 95% CI 0.52-0.67), FernandezTeruel_2024_capivasertib.R (the effect is on the time-dependent-clearance term rather than on baseline CL/F: a 2.15-fold multiplier on the negative LOG-scale magnitude of capivasertib’s sigmoidal auto-inhibition of CL/F, lcl_time_max * (1 + 1.15 * CONMED_PACLITAXEL), which shrinks the maximal inhibition from 17% to 2% at 400 mg and so raises steady-state CL/F by about 20%. Direction is therefore OPPOSITE to Zuo 2024. The reference level pools capivasertib monotherapy and capivasertib + fulvestrant, fulvestrant not being a significant covariate. Paclitaxel was given only in the BEECH study (90 of 441 patients), so the effect is confounded with study, and Fernandez-Teruel 2024 Discussion cautions that BEECH’s sparse post-144 h sampling limits the ability to resolve the time-dependent PK; the effect is being re-examined in the phase III CAPItello-290 trial).
• Notes: Distinct from PRIOR_TAXANE, which records taxane exposure before study entry rather than concurrent co-administration, and from CONMED_CHEMO, which pools any chemotherapy backbone (including paclitaxel + carboplatin) into a single binary for anti-PD-(L)1 antibody popPK. Use CONMED_PACLITAXEL only when the source paper estimates a paclitaxel-specific effect separable from other cytotoxics; use CONMED_CHEMO when the paper itself collapses the backbone. Forms a mutually exclusive indicator set with [[CONMED_ANTINEO_OTHER]] and [[CONMED_ANTINEO_NONE]] in Zuo 2024; a subject in the reference stratum carries 0 on all three. The paclitaxel direction in Zuo 2024 is an empirical TDM association (paclitaxel and apatinib are both CYP3A4 substrates and the taxane co-treatment cohort is the sickest stratum), and the authors do not claim a mechanistic DDI.

CONMED_ANTINEO_OTHER (canonical for concomitant non-taxane cytotoxic antineoplastic composite indicator)

• Description: 1 = the modelled drug’s observation was collected while the patient was concurrently receiving a cytotoxic antineoplastic agent that the source paper pools into an unresolved “other” stratum rather than naming individually, 0 = not in that stratum. A composite umbrella indicator for oncology popPK analyses that resolve one or two co-administered agents by name and collapse the remainder; the exact class membership is paper-specific and must be recorded in covariateData[[CONMED_ANTINEO_OTHER]]$notes.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (not in the pooled other-cytotoxic stratum). As with the other members of the set, the substantive meaning of the zero level is set by the source paper’s stratification and is documented per model.
• Source aliases:
  • CM (category C) – used in Zuo_2024_apatinib.R, where the pooled stratum is platinum agents, capecitabine, and the tegafur/gimeracil/oteracil potassium combination (S-1).
• Example models: Zuo_2024_apatinib.R (multiplicative effect on apparent clearance: CL/F * 1.60^CONMED_ANTINEO_OTHER, i.e. 60% higher CL/F relative to the anti-PD-1 mAb reference group; Zuo 2024 Table 2, RSE 27%, bootstrap 95% CI 1.35-1.90).
• Notes: Structurally parallel to [[CONMED_ABX]], the register’s other “the source paper does not separate the individual agents” composite: use it only when the paper itself pools the agents, and use per-INN indicators (CONMED_PACLITAXEL, CONMED_DOXORUBICIN, CONMED_EPIRUBICIN, CONMED_DECITABINE,…) whenever the paper estimates a named-agent effect. Distinct from CONMED_CHEMO, which is specifically the anti-PD-(L)1 mAb + platinum-doublet study-design indicator with its own fixed class definition. Because the pooled membership varies between papers, a positive or negative effect on clearance should be read as an empirical marker of the co-treated subset rather than as a transferable drug-drug-interaction estimate. Forms a mutually exclusive indicator set with [[CONMED_PACLITAXEL]] and [[CONMED_ANTINEO_NONE]] in Zuo 2024.

CONMED_ANTINEO_NONE (canonical for antineoplastic-monotherapy indicator (no concomitant antineoplastic))

• Description: 1 = the modelled antineoplastic was given as monotherapy, with no concomitant antineoplastic agent at the observation; 0 = at least one concomitant antineoplastic agent was co-administered. Used when a source paper’s co-administration covariate is a multi-level categorical whose reference level is an active combination rather than monotherapy, so that “monotherapy” needs its own indicator column rather than being recovered as the all-zero baseline.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (some concomitant antineoplastic present). Note that this canonical deliberately does not assume monotherapy is the reference; that is the situation it exists to handle.
• Source aliases:
  • CM (category D) – used in Zuo_2024_apatinib.R (Zuo 2024 Table 1 footnote group D, “Monotherapy”).
• Example models: Zuo_2024_apatinib.R (multiplicative effect on apparent clearance: CL/F * 1.38^CONMED_ANTINEO_NONE, i.e. 38% higher CL/F relative to the anti-PD-1 mAb reference group; Zuo 2024 Table 2, RSE 38%, bootstrap 95% CI 1.14-1.65).
• Notes: Scoped to concomitant antineoplastic therapy only – a patient on apatinib monotherapy who is also taking, say, an antiemetic or an antihypertensive still carries CONMED_ANTINEO_NONE = 1. Distinct from CONMED_STATIN_MONO and CONMED_ERA / CONMED_PDE5I, which are monotherapy indicators for a co-medication class (the patient is on that class alone); this canonical says the modelled drug is unaccompanied. Where a paper does use monotherapy as its reference level, encode the combination strata instead (as Zhang_2019_nivolumab.R and Kuchimanchi_2024_dostarlimab.R do with CONMED_CHEMO) and do not introduce this column. Forms a mutually exclusive indicator set with [[CONMED_PACLITAXEL]] and [[CONMED_ANTINEO_OTHER]] in Zuo 2024; exactly one of the four Zuo strata applies to any observation, and the reference stratum (immune-checkpoint-inhibitor mAb co-administration) carries 0 on all three.

CONMED_PB (canonical for concomitant phenobarbital coadministration indicator)

• Description: 1 = subject is taking phenobarbital (PB) – including primidone, which is metabolised to phenobarbital and is conventionally pooled with PB – as a concomitant antiepileptic drug at the PK observation, 0 = no concomitant phenobarbital / primidone. Phenobarbital is a broad-spectrum CYP and UGT inducer that increases the apparent clearance of co-administered drugs. Time-varying when phenobarbital starts / stops within the observation window; time-fixed when the source paper analyses chronic-maintenance cohorts whose AED therapy is stable.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant phenobarbital or primidone).
• Source aliases:
  • PB – used in Schoemaker_2017_brivaracetam.R (paper covariate PB for phenobarbital or primidone coadministration; the source pools primidone with phenobarbital because primidone is metabolised to phenobarbital) and Lee_2024_topiramate.R (Lee 2024 Table S1 theta7, additive enzyme-induction term on topiramate CL/F; that paper does not state whether primidone was pooled with phenobarbital).
• Example models: Schoemaker_2017_brivaracetam.R (multiplicative effect on apparent oral clearance: cl *= (1 + 0.408 * CONMED_PB); +40.8% relative to no-PB reference, corresponding to ~29% lower brivaracetam exposure, Schoemaker 2017 Table 1), Lee_2024_topiramate.R (additive effect in absolute units: cl <- (exp(lcl) + e_pb_cl * CONMED_PB + ...) * ... with e_pb_cl = 0.376 L/h on a 1.45 L/h monotherapy intercept, i.e. +26% relative to no-PB reference and the smallest of that model’s four enzyme-inducing coefficients, Lee 2024 Table S1 / Discussion).
• Notes: Drug-specific CONMED_* indicator anticipated in the [[CONMED_AED]] notes; used when a paper estimates a separate phenobarbital-induction effect distinct from the pooled EIAED / AED class effect. Per-model covariateData[[CONMED_PB]]$notes should document whether primidone is pooled with phenobarbital (Schoemaker 2017 pools them because primidone is metabolised to phenobarbital; Lee 2024 is silent on the point). Distinct from the broader [[CONMED_EIAED]] (any enzyme-inducing AED) and [[CONMED_AED]] (any concomitant AED). When a paper distinguishes individual AEDs separately, use the drug-specific canonicals [[CONMED_CBZ]], CONMED_PB, [[CONMED_OXC]], [[CONMED_PHT]], [[CONMED_VPA]] rather than collapsing into the class-level indicator.

CONMED_PHT (canonical for concomitant phenytoin coadministration indicator)

• Description: 1 = subject is taking phenytoin (PHT) as a concomitant antiepileptic drug at the PK observation, 0 = no concomitant phenytoin. Phenytoin is a strong, broad-spectrum CYP (2C9, 2C19, 3A4) and UGT inducer and is generally the most potent enzyme-inducing antiseizure medication in routine use, so it typically carries the largest induction coefficient when a paper decomposes the enzyme-inducing class into individual drugs. Time-varying when phenytoin starts / stops within the observation window; time-fixed when the source paper analyses chronic-maintenance cohorts whose antiseizure therapy is stable across the analysis window.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant phenytoin).
• Source aliases:
  • PHT – used in Lee_2024_topiramate.R (Lee 2024 Table S1 theta6, additive enzyme-induction term on topiramate CL/F).
• Example models: Lee_2024_topiramate.R (additive effect on apparent oral clearance in absolute units: cl <- (exp(lcl) + e_pht_cl * CONMED_PHT + ...) * ... with e_pht_cl = 1.02 L/h on a 1.45 L/h monotherapy intercept, i.e. +70% relative to no-PHT reference, Lee 2024 Table S1 / Discussion; the largest of the four enzyme-inducing coefficients in that model).
• Notes: Drug-specific CONMED_* indicator anticipated in the [[CONMED_AED]] notes; used when a paper estimates a separate phenytoin-induction effect distinct from the pooled EIAED / AED class effect. Distinct from the broader [[CONMED_EIAED]] (any enzyme-inducing AED) and [[CONMED_AED]] (any concomitant AED). When a paper distinguishes individual AEDs separately, use the drug-specific canonicals [[CONMED_CBZ]], [[CONMED_PB]], [[CONMED_OXC]], CONMED_PHT and [[CONMED_VPA]] rather than collapsing into the class-level indicator. Distinct from [[DOSE_PHT_MGKGD]], which carries the phenytoin daily dose per kg as a continuous covariate rather than a coadministration indicator – use this binary canonical when the source paper models only the presence of phenytoin, and DOSE_PHT_MGKGD when it models the phenytoin dose level. Hashimoto 1994 (zonisamide) tested phenytoin coadministration but found no significant effect and retained only CBZ in the final model, so that model carries no CONMED_PHT column.

CONMED_PRM (canonical for concomitant primidone coadministration indicator)

• Description: 1 = patient is taking concomitant primidone, 0 = not. Primidone is a barbiturate-class antiepileptic that is metabolised to phenobarbital, so it induces the same hepatic pathways and is routinely grouped with phenobarbital and phenytoin as a “phenobarbital-like” enzyme inducer.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no primidone coadministration).
• Source aliases:
  • “primidone” – Gidal 2018 Table S1 comedication row and the Table S-2 footnote defining the composite phenobarbital-like inducer flag. The source publishes no column name.
• Example models: Gidal_2018_eslicarbazepine.R (one of the three constituent indicators of that paper’s composite phenobarbital-like flag, rebuilt inside model as 1 - (1 - CONMED_PB) * (1 - CONMED_PHT) * (1 - CONMED_PRM); the flag adds 1.24 L/h to eslicarbazepine apparent oral clearance and 12.0 L to apparent volume).
• Notes: Member of the CONMED_<INN> family; siblings CONMED_PB (phenobarbital) and CONMED_PHT (phenytoin) complete the phenobarbital-like inducer group. Registered as three separate indicators rather than as one group canonical because decomposed binary indicators keep the column meaning unambiguous, let a user encode a real regimen, and let a consuming model rebuild whatever grouping its source paper used. Note that a “phenobarbital-like” group of this kind conventionally EXCLUDES carbamazepine even though carbamazepine is also an inducer, because papers that carry the group usually give carbamazepine its own dose-dependent term (see DOSE_CBZ_MGD); CONMED_EIAED is therefore the wrong canonical for such a flag. Primidone is typically the sparsest of the three – 0.8% of the founding cohort – which is what makes the grouping necessary for estimability.

CONMED_LEV (canonical for concomitant levetiracetam coadministration indicator)

• Description: 1 = patient is taking concomitant levetiracetam, 0 = not. Levetiracetam binds synaptic vesicle protein 2A, is cleared largely by renal excretion and non-hepatic hydrolysis, and is neither a significant inducer nor a significant inhibitor of hepatic drug metabolism, so it is commonly screened in antiepileptic-drug interaction analyses as an expected null.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no levetiracetam coadministration).
• Source aliases:
  • “levetiracetam” – Gidal 2018 Table S1 comedication row. The source publishes no column name.
• Example models: Gidal_2018_eslicarbazepine_headache.R (carried in covariatesDataExcluded: screened as a candidate covariate and explicitly not retained for any of that paper’s three treatment-emergent-adverse-event endpoints, and separately reported to have no clinically noteworthy effect on eslicarbazepine exposure in the population PK analysis).
• Notes: Member of the CONMED_<INN> family. Registered from a screened-and-not-retained sighting rather than from a retained effect, which is a legitimate reason to register: a documented null needs a canonical name as much as a retained effect does, and recording it in covariatesDataExcluded preserves the provenance of the covariate screen without carrying a convention warning. A future model that DOES retain a levetiracetam effect should reuse this entry and add itself to the example list.

CONMED_OXC (canonical for concomitant oxcarbazepine coadministration indicator)

• Description: 1 = subject is taking oxcarbazepine (OXC) as a concomitant antiepileptic drug at the PK observation, 0 = no concomitant oxcarbazepine. Oxcarbazepine is the 10-keto analogue of carbamazepine and, via its active monohydroxy metabolite, is a weaker and more selective inducer than carbamazepine itself (principally CYP3A4 / UGT, without the potent autoinduction and broad CYP induction of CBZ), so it typically carries a smaller induction coefficient than [[CONMED_CBZ]] when a paper decomposes the enzyme-inducing class into individual drugs. Time-varying when oxcarbazepine starts / stops within the observation window; time-fixed when the source paper analyses chronic-maintenance cohorts whose antiseizure therapy is stable across the analysis window.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant oxcarbazepine).
• Source aliases:
  • OXC – used in Lee_2024_topiramate.R (Lee 2024 Table S1 theta5, additive enzyme-induction term on topiramate CL/F).
• Example models: Lee_2024_topiramate.R (additive effect on apparent oral clearance in absolute units: cl <- (exp(lcl) + e_oxc_cl * CONMED_OXC + ...) * ... with e_oxc_cl = 0.419 L/h on a 1.45 L/h monotherapy intercept, i.e. +29% relative to no-OXC reference, Lee 2024 Table S1 / Discussion).
• Notes: Drug-specific CONMED_* indicator anticipated in the [[CONMED_AED]] notes; used when a paper estimates a separate oxcarbazepine-induction effect distinct from the pooled EIAED / AED class effect. Must not be collapsed into [[CONMED_CBZ]] despite the close structural relationship between the two drugs: oxcarbazepine is a materially weaker inducer, and Lee 2024 estimates the two coefficients separately (OXC 0.419 L/h vs CBZ 0.703 L/h, a 1.7-fold difference), so overloading the carbamazepine column would misattribute the induction magnitude. Distinct from the broader [[CONMED_EIAED]] (any enzyme-inducing AED) and [[CONMED_AED]] (any concomitant AED). When a paper distinguishes individual AEDs separately, use the drug-specific canonicals [[CONMED_CBZ]], [[CONMED_PB]], CONMED_OXC, [[CONMED_PHT]] and [[CONMED_VPA]].

CONMED_ZNS (canonical for concomitant zonisamide coadministration indicator)

• Description: 1 = subject is taking zonisamide (ZNS) as a concomitant antiepileptic drug at the PK observation, 0 = no concomitant zonisamide. Zonisamide is a sulfonamide antiseizure drug cleared largely by CYP3A4-mediated reduction and by glucuronidation; unlike the classical enzyme-inducing antiseizure medications it is a victim rather than a perpetrator of induction, and is not itself an appreciable inducer or inhibitor at therapeutic concentrations. It therefore belongs in a model as a tested-and-usually-null comedication indicator, which is exactly how the founding source uses it. Time-varying when zonisamide starts / stops within the observation window; time-fixed when the source paper analyses chronic-maintenance cohorts whose antiseizure therapy is stable across the analysis window.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant zonisamide).
• Source aliases:
  • treatment group (LZ) – used in Munana_2018_levetiracetam_dog.R (Munana 2018 codes zonisamide coadministration as one level of a single three-level treatment group categorical covariate, L / LP / LZ, rather than as a standalone column; CONMED_ZNS and [[CONMED_PB]] together carry that factor with the L group as the shared reference).
• Example models: Munana_2018_levetiracetam_dog.R (multiplicative effect on the log apparent volume of distribution of extended-release levetiracetam in epileptic dogs: vc <- exp(lvc + e_conmed_pb_vc * CONMED_PB + e_conmed_zns_vc * CONMED_ZNS + etalvc) with e_conmed_zns_vc = 0, i.e. V/F = 0.15 L/kg in both the zonisamide and the levetiracetam-alone groups, Munana 2018 Table 1 theta V row).
• Notes: Drug-specific CONMED_* indicator in the family anticipated by the [[CONMED_AED]] notes; use it when a paper estimates a zonisamide effect separately rather than pooling it into [[CONMED_AED]]. Zonisamide is not an enzyme-inducing antiseizure medication, so a dataset that carries this column must not set [[CONMED_EIAED]] on the strength of it; the drug-specific inducer canonicals are [[CONMED_CBZ]], [[CONMED_PB]], [[CONMED_OXC]], [[CONMED_PHT]] and [[CONMED_VPA]]. The founding effect size is exactly zero and that is a published result, not a placeholder: Munana 2018 retained the three-level treatment factor on V/F and its zonisamide level reproduced the reference volume to the two significant figures Table 1 prints, against a 2.58-fold rise on phenobarbital in the same fit. Reading the null as “the covariate was dropped” would lose the paper’s stated conclusion that “coadministration of zonisamide was not shown to contribute to the variability”. Founding use is veterinary (dog); nothing about the column is species-specific.

CONMED_PLDH (canonical for concomitant PEGylated liposomal doxorubicin combination indicator)

• Description: 1 = subject is receiving the modelled drug in combination with PEGylated liposomal doxorubicin (PLDH; Doxil / Caelyx), 0 = otherwise (no concomitant PLDH). Encodes the PLDH co-administration regimen as a study-design covariate on clearance.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no concomitant PEGylated liposomal doxorubicin).
• Source aliases:
  • PLDH – used in Schmitt_2018_vinflunine.R (Schmitt 2018 NONMEM dataset; 1 = vinflunine + PEGylated liposomal doxorubicin combination cohort).
• Example models: Schmitt_2018_vinflunine.R (power-form effect on vinflunine clearance: cl *= 0.865^CONMED_PLDH – vinflunine apparent CL is reduced to 86.5% of single-agent value when coadministered with PLDH, Schmitt 2018 Table 2 and explicit CL formula on p.1607).
• Notes: PEGylated liposomal doxorubicin (PLDH) is a long-circulating doxorubicin formulation used in oncology combination regimens. The Schmitt 2018 vinflunine analysis is the first paper in this register to use PLDH as a popPK combination covariate; promote to general scope when a second paper ratifies the same encoding. Distinct from CONMED_AVD (brentuximab vedotin + AVD chemotherapy backbone, which includes free doxorubicin rather than the PEGylated liposomal formulation) and from PRIOR_ANTHRACYCLINE_DOSE (cumulative prior anthracycline exposure, not concomitant).

CONMED_PLATIN (canonical for any concomitant platinum-containing chemotherapy indicator)

• Description: 1 = subject received at least one cycle of platinum-containing chemotherapy (cisplatin, carboplatin, oxaliplatin, nedaplatin, satraplatin) concurrently with the modelled treatment during the analysis window; 0 = no concomitant platinum during that window. The concomitant counterpart of [[PRIOR_PLATIN]], which is exposure before the analysed treatment starts. Time-fixed per subject when the source paper collapses the whole assessment window into a single indicator (the usual case, because platinum cycles recur on a fixed schedule); time-varying in principle if a paper flags each cycle.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant platinum-containing chemotherapy).
• Source aliases:
  • PCNT (dichotomized at > 0) – used in Agoram_2006_darbepoetin_alfa_cia_hemoglobin.R. The source column is the integer count of platinum-containing chemotherapy cycles administered during the assessment; Agoram 2006 Equation 14 uses it only through the indicator X = 1 if PCNT > 0, else X = 0. Per the standing count-covariate policy the count is decomposed into binary indicators rather than carried as an integer.
• Example models: Agoram_2006_darbepoetin_alfa_cia_hemoglobin.R (multiplicative power effect on the hemoglobin-stimulation S50 of darbepoetin alfa: ec50 *= 1.86^CONMED_PLATIN, i.e. an 86% higher half-maximal stimulating concentration – a blunted erythropoietic response – in patients on concomitant platinum; asymptotic 95% CI (0.982, 2.74) overlaps the NULL value of 1, so the paper reports it as an effect it lacks information about rather than a demonstrated one).
• Notes: Distinct from [[PRIOR_PLATIN]] on timing (concomitant vs pre-treatment) – do not substitute one for the other, because a myelosuppressive concomitant regimen and a historical exposure are different mechanisms. Distinct from [[CONMED_CHEMO]], which is specifically the anti-PD-(L)1 mAb + platinum-doublet study-design indicator with its own fixed class definition, and from [[CONMED_ANTINEO_OTHER]], which pools platinum with non-platinum cytotoxics. When a paper distinguishes platinum-cycle-count tiers rather than the pooled any-vs-none contrast, pair this with the sibling [[CONMED_PLATIN_GT2]]; data assemblers must enforce that CONMED_PLATIN_GT2 = 1 implies CONMED_PLATIN = 1. Ratified canonically on 2026-09-03 alongside the Agoram 2006 darbepoetin alfa CIA extraction.

CONMED_PLATIN_GT2 (canonical for > 2 cycles of concomitant platinum-containing chemotherapy)

• Description: 1 = subject received more than two cycles of concomitant platinum-containing chemotherapy during the analysis window; 0 = two or fewer cycles (including none). Sibling binary to [[CONMED_PLATIN]] for platinum-cycle-count-tier covariate models; the two binaries together encode a 3-level tier (none / 1-2 cycles / > 2 cycles).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (at most two concomitant platinum cycles). The tier reference (CONMED_PLATIN_GT2 = 0 with CONMED_PLATIN = 1) is “1-2 cycles”; the absolute reference (CONMED_PLATIN_GT2 = 0 with CONMED_PLATIN = 0) is “no concomitant platinum”.
• Source aliases:
  • PCNT (dichotomized at > 2) – used in Agoram_2006_darbepoetin_alfa_cia.R and Agoram_2006_darbepoetin_alfa_cia_hemoglobin.R. Agoram 2006 Equation 12 defines X = 1 if PCNT > 2, else X = 0, stating this is the split at which “the greatest difference in mean clearance was observed”.
• Example models: Agoram_2006_darbepoetin_alfa_cia.R (multiplicative power effect on darbepoetin alfa clearance: cl *= 0.737^CONMED_PLATIN_GT2, a 26.3% clearance reduction in the > 2-cycle group, bootstrap 95% CI (0.597, 0.940) excluding the NULL value of 1), Agoram_2006_darbepoetin_alfa_cia_hemoglobin.R (same effect, carried at its fixed value inside the sequentially fitted PD model’s PK layer).
• Notes: Specific scope; the binary is only meaningful in papers that distinguish a > 2-cycle tier from a 1-2-cycle tier. Promote to general scope when a second platinum-cycle-count paper ratifies the name. Mutual-consistency constraint: CONMED_PLATIN_GT2 = 1 MUST imply CONMED_PLATIN = 1; a record asserting the reverse is malformed. Structurally parallel to the [[CONMED_AED]] / [[CONMED_AED_GE2]] pair, which decomposes a concomitant-antiepileptic count into the same two-indicator shape for the same reason. Note that Agoram 2006 dichotomizes the identical underlying PCNT count at two different thresholds in the same publication – > 2 for the PK model’s clearance effect and > 0 for the PD model’s S50 effect – so a data set supporting both models must carry both indicator columns. Ratified canonically on 2026-09-03 alongside the Agoram 2006 darbepoetin alfa CIA extraction.

CONMED_PDE5I (canonical for concomitant PDE5-inhibitor monotherapy indicator in PAH)

• Description: 1 = patient is on a concomitant phosphodiesterase type 5 inhibitor (PDE5I) PAH therapy (e.g., sildenafil, tadalafil) alone – i.e. on a PDE5 inhibitor but NOT also on an endothelin-receptor antagonist; 0 = otherwise. Time-fixed per subject for the PK/PD analysis (PAH comedication at baseline, stable dose required before selexipag start).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (PAH-comedication-naive OR on ERA alone OR on both ERA+PDE5; mutually exclusive with CONMED_ERA and CONMED_ERA_PDE5I).
• Source aliases:
  • PAHCOMED category 2 (“PDE5 inh.”) – decomposed from the categorical PAHCOMED source column with levels {naive, ERA, PDE5, ERA+PDE5} in Krause_2017_selexipag.R.
• Example models: Krause_2017_selexipag.R (multiplicative effect on the ACT-333679 elimination rate constant: km *= (1 + 0.07 * CONMED_PDE5I); +7% relative to the PAH-comedication-naive reference, Krause 2017 Table 1. The PDE5-only coefficient is statistically not significant (p = 0.19) but retained in the final model so the four-level PAH-comedication categorical is preserved end-to-end), Axelsen_2024_selexipag_adult.R, Axelsen_2024_selexipag_pediatric.R (exponential effect km *= exp(0.0633 * CONMED_PDE5I) adult / exp(0.0495 * CONMED_PDE5I) pediatric, Axelsen 2024 Table 1 beta_km(COPAH2_1); source column COPAH2. Imprecise in both fits – RSE 96.1% adult, 116% pediatric – and retained for the same reason).
• Notes: Used together with CONMED_ERA and CONMED_ERA_PDE5I to decompose a four-level PAH-comedication categorical (naive / ERA-only / PDE5-only / ERA-and-PDE5) into three orthogonal mutually-exclusive binary indicators. Specific scope because the indicator’s semantics are tied to the GRIPHON study’s PAH-comedication taxonomy.

CONMED_PPI (canonical for concomitant proton-pump inhibitor use)

• Description: 1 = patient on concomitant proton-pump inhibitor (CONMED_PPI) therapy, 0 = no CONMED_PPI use. Captures gastric-pH-elevating co-medication that can reduce the bioavailability of solubility-limited (typically weakly basic) orally administered drugs.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no CONMED_PPI use).
• Source aliases:
  • CONMED_PPI – used in Goel_2016_Sonidegib.R (Goel 2016 dataset; defined as “significant” CONMED_PPI use, i.e. duration of CONMED_PPI use >= 80% of the PK assessment phase).
  • PPI – source-paper column name; used in Shu_2024_posaconazole.R (Shu 2024 dataset; subject-level ever-versus-never indicator, 11 of 62 HSCT patients, with no minimum-duration threshold stated).
  • PPI – also the source column in Wang_2021_mycophenolic_acid.R (Wang 2021 heart transplant dataset; recorded per PK sampling cycle, omeprazole or pantoprazole by either the intravenous or the oral route).
  • PPI – also the source column in Ji_2022_pemigatinib.R (Ji 2022 Equation II; time-varying by visit).
• Example models: Goel_2016_Sonidegib.R (multiplicative effect on F: 0.696^CONMED_PPI, ~30% lower bioavailability under CONMED_PPI coadministration), Shu_2024_posaconazole.R (multiplicative effect on apparent volume: 3.832335^CONMED_PPI on V/F for posaconazole oral suspension; mechanistically a bioavailability effect that surfaces on V/F because the dataset is oral-only, so F is not separately identifiable – the same physiology as the Goel 2016 effect on F, relocated by the parameterisation), Wang_2021_mycophenolic_acid.R (multiplicative effect on F: 0.724^CONMED_PPI for mycophenolic acid after oral mycophenolate mofetil, a 27.6% fall in bioavailability), Chan_2020_fenebrutinib.R (time-varying; multiplicative effects on CL/F (x0.663), on the transit MTT (x0.835, and x2.26 more when fed) and on relative bioavailability (x0.657, and x0.693 more when fed)), Ji_2022_pemigatinib.R (PPI use is the paper’s reference level on Vc/F: Vc/F is multiplied by (1 - 0.244 * (1 - CONMED_PPI)), so Vc/F is 32% higher with a PPI than without).
• Notes: Per-model covariateData[[CONMED_PPI]]$notes must document the operational definition (e.g., Goel 2016 requires CONMED_PPI use covering >= 80% of the PK assessment window; other studies may use a simpler ever-vs-never indicator or a per-record time-varying flag). Distinct from CONMED_H2RA (H2-receptor antagonist) which acts on gastric pH via a different mechanism. On an apparent-parameter (CL/F, V/F) model fitted to oral-only data the PPI effect can legitimately appear on V/F or CL/F rather than on F; document which parameter carries it, because the sign convention flips (reduced F raises both apparent parameters).

CONMED_PANCRELIPASE (canonical for concomitant pancreatic-enzyme-replacement (pancrelipase) administration indicator)

• Description: 1 = a dose of pancreatic enzyme replacement therapy (PERT; INN pancrelipase – the porcine-derived lipase / protease / amylase combination marketed as Creon, Zenpep, Pancreaze, Pertzye and others) was administered as part of the modeled occasion; 0 = no enzymes administered on that occasion. Occasion-level rather than subject-level, because the same pancreatic-insufficient subject typically contributes both enzyme and no-enzyme occasions in a crossover fat-absorption study. The indicator flags whether exogenous lipase was available to hydrolyse dietary or test-meal triglyceride, and therefore acts on the bioavailability of any lipase-dependent analyte; a free fatty acid that does not require lipase is expected to show little or no effect.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no pancreatic enzymes administered with the meal / test dose).
• Source aliases: none known; source datasets typically encode the enzyme occasion with ad-hoc names (e.g., ENZ, ENZYME, PERT) or via an occasion / arm categorical.
• Example models: Mascarenhas_2015_pentadecanoic_triheptadecanoic.R (multiplicative bioavailability factor in the malabsorption blood test: for the lipase-dependent analyte heptadecanoic acid, F rises from 0.0292 without enzymes to 0.606 with enzymes – a >20-fold increase – while for the lipase-independent free fatty acid pentadecanoic acid F is essentially unchanged, 1.07 without versus 0.877 with; the contrast between the two analytes is the test read-out).
• Notes: Record the enzyme product and lipase-unit dose in covariateData[[CONMED_PANCRELIPASE]]$notes, because PERT products differ in protease:lipase ratio and in enteric-coating behaviour; Mascarenhas 2015 standardised on Creon 20 at 80 000-140 000 lipase units and reported no additional bioavailability gain above the 80 000-unit standard dose, so the binary indicator (rather than a continuous lipase-unit column) is the appropriate encoding for that dataset. Where the timing of the enzyme dose relative to the meal is itself modeled, pair this indicator with the CONMED_PANCRELIPASE_PRE30 / _POST30 / _POST60 qualifier family below, whose all-zero reference is “enzymes taken with the meal”. Distinct from AMYL and LIPASE, which are endogenous serum enzyme-activity measurements rather than an administered-drug indicator.

CONMED_PANCRELIPASE_PRE30, CONMED_PANCRELIPASE_POST30, CONMED_PANCRELIPASE_POST60 (canonical for pancreatic-enzyme dose-timing qualifiers relative to the meal)

• Description: Mutually exclusive binary qualifiers recording when the pancreatic enzyme dose was taken relative to the start of the meal or test dose, for occasions on which enzymes were in fact taken (CONMED_PANCRELIPASE = 1). CONMED_PANCRELIPASE_PRE30 = 1 for enzymes taken 30 minutes before the meal; CONMED_PANCRELIPASE_POST30 = 1 for 30 minutes after; CONMED_PANCRELIPASE_POST60 = 1 for 60 minutes after. All three are 0 when enzymes are taken with the meal, which is the reference timing. Occasion-level.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: all three = 0 (enzymes taken with the meal, at meal initiation).
• Source aliases: none known; Mascarenhas 2015 writes them as the indicator variables t1, t2 and t3 in its final-model equation block.
• Example models: Mascarenhas_2015_pentadecanoic_triheptadecanoic.R (multiplicative factors on heptadecanoic-acid bioavailability relative to enzymes-with-the-meal: 0.911 for _PRE30, 0.829 for _POST30 and 0.78 for _POST60, encoded in the power form e_pre30_fdepot_hepta^CONMED_PANCRELIPASE_PRE30 * ... so the reference timing contributes a factor of exactly 1; the paper found no corresponding timing effect on the lipase-independent analyte pentadecanoic acid, so these qualifiers act on the _hepta analyte only).
• Notes: Decomposed binaries rather than a single signed offset-in-minutes column, per the standing “decompose, don’t use an integer count / ordinal” policy – the published effects are not monotone in a way a linear time term would capture (_PRE30 at 0.911 and _POST30 at 0.829 straddle the reference asymmetrically), and each level carries its own estimate and confidence interval. Extend the family (CONMED_PANCRELIPASE_PRE60, etc.) rather than repurposing an existing member if a future study adds timing arms.

CONMED_PROBENECID (canonical for concomitant probenecid co-administration indicator)

• Description: 1 = subject coadministered probenecid (organic-anion transport inhibitor; multidrug resistance protein (MRP) family modulator with documented activity at the blood-brain barrier, blood-CSF barrier, and renal tubular secretion sites), 0 = no concomitant probenecid. Time-varying per subject because probenecid exposure can start and stop within the observation period.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no probenecid coadministration).
• Source aliases:
  • Paper-specific day-2 treatment-arm indicator – used in Xie_2000_m3g_rat.R (the paper distinguishes a probenecid-treatment arm from a control arm by experimental day; the canonical column is 1 during the day-2 probenecid co-infusion and 0 otherwise, including on day 1 of the probenecid arm before the probenecid loading dose).
• Example models: Xie_2000_m3g_rat.R (multiplicative exponential effect on the unbound BBB influx clearance CL_u,in: cluin *= exp(e_conmed_probenecid_cluin * CONMED_PROBENECID) with e_conmed_probenecid_cluin = log(0.17 / 0.11) = 0.4353, i.e. a 1.55-fold increase in CL_u,in into rat brain ECF under probenecid co-administration; CL_u,out and the intercompartmental brain clearance Q_br are not statistically affected by probenecid in the paper’s model selection and so are not paired with this canonical), Landersdorfer_2009_gemifloxacin.R (binary treatment-arm covariate that captures the static / non-mechanism-specific probenecid effects on gemifloxacin absorption rate (Ka 0.839 -> 0.897 1/h; +6.9%), absorption lag (Tlag 0.223 -> 0.129 h; -42%), and non-renal clearance (CL_NR 25.2 -> 21.0 L/h; -16.7%); the mechanism-specific competitive inhibition of renal tubular secretion is encoded separately via the time-varying CP_PRB_MGL covariate), Hamren_2008_tesaglitazar.R (multiplicative effect on parent CLrt and metabolite Vmax under probenecid co-administration: (1 - 0.75 * CONMED_PROBENECID) reduces both by 75% during the three-dose probenecid window in four healthy-control subjects; the shared estimate is encoded as two canonical-named parameters e_probenecid_cl_renal and e_probenecid_vmax_gluc to keep each effect’s naming canonical), Stocker_2012_oxypurinol.R (multiplicative linear-deviation effect on apparent CL/Fm: cl *= (1 + 0.383 * CONMED_PROBENECID), i.e. +38.3% apparent oxypurinol clearance with concomitant probenecid in adults with gout, consistent with the paper’s mechanism narrative of probenecid inhibiting renal tubular reabsorption of oxypurinol; Stocker 2012 Table 3 theta8), Ujihira_2025_glycochenodeoxycholicAcidSulfate.R (multiplicative effect on the hepatobiliary clearance arm of the endogenous OATP1B3 / OAT3 biomarker GCDCA-S: cl_nonren *= (1 + e_prob_clh * CONMED_PROBENECID) with e_prob_clh = 1/1.7 - 1 = -0.4118, encoding the paper’s estimated fold reduction X = 1.7 attributed to probenecid’s weak OATP1B3 inhibition; the concentration-dependent OAT3 inhibition of the renal arm is encoded separately through the coupled probenecid PK states and Ki,u,OAT3 rather than through this indicator).
• Notes: Scope: specific because the only source is a preclinical microdialysis rat BBB-transport paper (Xie 2000) and the column meaning is intrinsically tied to the day-2 probenecid co-infusion design rather than a general clinical-coadministration indicator. Per-model covariateData[[CONMED_PROBENECID]]$notes must document the dose / regimen of probenecid used (Xie 2000: 70 umol/kg IV loading dose plus 70 umol/kg/h constant infusion in male Sprague-Dawley rats), the parameter the indicator modifies, and any per-subject vs per-record time-varying convention.

CONMED_QPRL_ORAL (canonical for concomitant oral quinapril co-administration indicator)

• Description: 1 = subject coadministered oral quinapril (angiotensin-converting enzyme inhibitor; competitive substrate of the intestinal H+/oligopeptide carrier PEPT1 used by orally absorbed beta-lactams and a substrate of the renal anionic transport system shared by many beta-lactams), 0 = no oral quinapril coadministration. Captures the specific oral-quinapril coadministration condition under which the Padoin 1998 cephalexin / quinapril drug-drug interaction (DDI) on intestinal absorption and renal tubular secretion of cephalexin is observed. Time-fixed per subject in the founding source (parallel-group design); time-varying use in future per-occasion DDI studies is permitted.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no oral quinapril coadministration; reference includes cephalexin alone, cephalexin with intra-arterial quinapril, and quinapril-naive subjects).
• Source aliases:
  • Paper-specific group indicator – used in Padoin_1998_cephalexin_rat.R (the paper’s final model specification is Ka_j = Ka2 and CL_j = CL2 when oral quinapril is coadministered with oral cephalexin; the canonical column is 1 for group 5 (cephalexin GT + quinapril GT) and 0 for groups 1, 2, 3, and 4).
• Example models: Padoin_1998_cephalexin_rat.R (multiplicative exponential effect on both absorption rate Ka and elimination clearance CL: ka <- exp(lka + e_conmed_qprl_oral_ka * CONMED_QPRL_ORAL + etalka) with e_conmed_qprl_oral_ka = log(0.177 / 0.249) = -0.3413 (~29% lower Ka under oral quinapril coadministration, paper Table 4); cl <- exp(lcl + e_conmed_qprl_oral_cl * CONMED_QPRL_ORAL + etalcl) with e_conmed_qprl_oral_cl = log(0.640 / 0.810) = -0.2356 (~21% lower CL under oral quinapril coadministration, paper Table 4). The CL effect was found significant only in the oral-cephalexin + oral-quinapril group (paper found no DDI on CL when cephalexin was given intra-arterially, attributed to higher cephalexin renal concentrations outcompeting quinapril at the carrier)…………………
• Notes: Scope: specific because the only source is a single preclinical rat popPK paper (Padoin 1998) and the column meaning is intrinsically tied to the oral cephalexin + oral quinapril DDI design rather than a general quinapril-coadministration indicator. Per-model covariateData[[CONMED_QPRL_ORAL]]$notes must document the dose / regimen of quinapril used (Padoin 1998: 0.8 mg/kg single oral dose via gastric tube, 15 min before cephalexin, in male Wistar rats), the parameters the indicator modifies, and any per-subject vs per-record convention. The paper’s full specification distinguishes intra-arterial vs oral quinapril (no DDI on cephalexin CL was observed for either intra-arterial cephalexin with intra-arterial quinapril or intra-arterial cephalexin with oral quinapril); the canonical CONMED_QPRL_ORAL = 0 collapses all three non-DDI conditions (no quinapril, intra-arterial quinapril, oral quinapril paired with intra-arterial cephalexin) into the reference category because the model predicts the same CL and Ka values for each. Future popPK extractions of beta-lactam / ACE-inhibitor DDIs that test oral quinapril should reuse this canonical; ACE-inhibitor coadministration with a different perpetrator drug (e.g., enalapril, lisinopril) should register a separate canonical with the perpetrator’s INN in the name.

CONMED_RIF (canonical for concomitant rifampicin co-administration (acute or chronic))

• Description: 1 = subject is within a paper-defined rifampicin co-administration window for the modeled effect (either acute single-dose use as an OATP1B inhibitor or perfusion-style perpetrator, or chronic daily-dose use after the post-induction equilibrium lag is reached); 0 = subject is not on rifampicin or is in the inactive pre-effect window. Time-varying per subject. The semantics of “effect active” is paper-specific: for chronic-induction substrate popPK (Svensson 2014 bedaquiline), the indicator switches on after several days of daily 600 mg rifampicin dosing to reach the post-induction CYP3A4 equilibrium; for acute single-dose DDI popPK (Barnett 2018 coproporphyrin I / rosuvastatin), the indicator switches on at the rifampicin dose time and stays on for the within-occasion sampling window, with no induction lag because the mechanism of interest is competitive OATP1B inhibition rather than enzyme induction.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no rifampicin co-administration or pre-effect lag).
• Source aliases:
  • RIF – used in Svensson_2014_bedaquiline.R (paper-defined indicator switching on at day 3 of 600 mg daily rifampicin co-administration; the 3-day lag was selected by NONMEM objective-function search over 1-8 day candidates and chosen as the best fit per the paper’s Methods). Also used in Barnett_2018_coproporphyrin_I.R and Barnett_2018_rosuvastatin.R as the binary period-level covariate that captures Barnett 2018 Table 1’s reductions of Vcpi (CPI) and V1 / V2 / Q (RSV) during the rifampicin phase of the three-occasion crossover study. Also used in Gatti_1996_dapsone.R (Gatti 1996 used the source-paper symbol R; the indicator switches on for the 7 of 53 patients on rifampin co-administration for at least 2 weeks before blood sampling, so the chronic CYP3A4-induction effect was at equilibrium).
• Example models: Svensson_2014_bedaquiline.R (multiplicative factor on apparent CL_BDQ and CL_M2: cl_eff = cl_base * 4.78^CONMED_RIF; 4.78-fold induction of apparent clearance at chronic full induction), Barnett_2018_coproporphyrin_I.R (multiplicative factor on Vcpi: vc <- exp(lvc + etalvc) * (1 + e_rif_vc * CONMED_RIF) with e_rif_vc = -0.4841 capturing the Vcpi 6.59 L -> 3.4 L reduction in the acute rifampicin phase), Barnett_2018_rosuvastatin.R (multiplicative factors on V1, Q, and V2 with the same encoding form, capturing V1 430 -> 2.98 L, Q 45.3 -> 5.03 L/h, V2 865 -> 128 L reductions during the acute rifampicin phase), Gatti_1996_dapsone.R (multiplicative factor SHARED between apparent CL/F and V/F: cl = exp(lcl + etalcl) * (1 + e_rif_cl_vc * CONMED_RIF), vc = exp(lvc) * (1 + e_rif_cl_vc * CONMED_RIF) with e_rif_cl_vc = 0.696; 69.6% increase in both CL/F and V/F driven primarily by a first-pass / bioavailability effect, with the shared theta enforced by likelihood-ratio test against a non-shared parameterisation), Zhang_2012_lopinavir_ritonavir.R (multiplicative factors on LPV CL/F (+71.0%), RTV CL/F (+36.0%), LPV F (-20.0%), and RTV F (-45.0% at the 100 mg reference RTV dose) captured by the same (1 + e_rif_* * CONMED_RIF) encoding; the CONMED_RIF flag also gates the secondary dose-dependent boost on RTV F ((1 + e_dose_f_rtv * (DOSE - 100))) so the dose-effect is identifiable only within the rifampicin-coadministered arm), Zhang_2013_lopinavir_ritonavir.R (per-population multiplicative shifts on apparent CL of LPV adults +58% / children +48% and RTV adults +34% / children +22%, and per-population log-additive shifts on relative bioavailability of LPV and RTV with separate adult-only and child-only RIF effects), Comisar_2025_zavegepant.R (fractional reduction of true systemic clearance: (1 + e_conmed_rif_cl * CONMED_RIF) with e_conmed_rif_cl = -0.411, i.e. -41.1% CL per Comisar 2025 Table 3. This is the sign-inverted case relative to the other example models: rifampicin decreases rather than increases zavegepant clearance, which Comisar 2025 attributes to OATP1B3 and NTCP transporter inhibition outweighing CYP3A induction for a compound whose primary route is hepatobiliary. Encoders should therefore not assume a positive induction coefficient for this covariate), Choules_2024_enfortumab.R, Choules_2024_brentuximab.R (multiplicative factor on the MMAE payload clearance of an antibody-drug conjugate: cl_mmae *= e_rif_cl_mmae^CONMED_RIF with e_rif_cl_mmae = 1/0.47 = 2.12766, i.e. a 2.13-fold increase in MMAE clearance reproducing the published MMAE AUC geometric mean ratio of 0.47 in Choules 2024 Table 5; the indicator is a step at full induction after 7 days of 600 mg daily rifampin, and it lumps the CYP3A4-induction and P-gp-induction components of the interaction into one clearance multiplier), Kengo_2025_atazanavir.R, Kengo_2025_ritonavir.R (chronic full induction after two weeks of 600 mg daily rifampicin in the DERIVE trial; the indicator is paired with REGI_BID to build the source’s three-level regimen categorical, so the SAME covariate carries a different clearance fold-change on once-daily than on twice-daily ATV/r – 3.05 versus 2.03 for atazanavir – while ritonavir takes one fold-change of 2.12 across both arms. Rifampicin also cuts bioavailability, by 52.5% for atazanavir and 68.8% for ritonavir on once-daily ATV/r, and slows atazanavir absorption by 67.3% on ka. Doubling the ATV/r dosing frequency fully restores atazanavir bioavailability and partially restores ritonavir’s; the paper found no additional effect of raising rifampicin from 600 to 1200 mg daily, so the indicator stays binary), Kengo_2025_atazanavir_a5231.R (same paper’s unboosted-atazanavir re-fit on ACTG A5231, where rifampicin acts on bioavailability (-55.3%) and on ka (-70.8%, fixed) but NOT on clearance, because that fit attributes its 2.13-fold clearance increase to the absence of ritonavir rather than to rifampicin – a useful reminder that the covariate’s target parameters are per-model), Wang_2019_lamotrigine.R (fractional multiplicative factor on apparent oral clearance of lamotrigine: cl *= (1 + 0.647 * CONMED_RIF), a 64.7% increase attributed to induction of UGT1A4 / UGT2B7 glucuronidation rather than of CYP3A4. Wang 2019 defines no induction-onset lag because every sample was drawn at steady state after at least 7 days on the same dose, so the indicator is a plain per-observation comedication flag at post-induction equilibrium; the effect is almost exactly cancelled by concurrent valproate, 0.614 * 1.647 = 1.011), Francis_2020_lumefantrine.R (fractional effect on lumefantrine apparent CL/F cl *= (1 + 1.42 * CONMED_RIF), +142% (59% lower AUC) with chronic rifampicin-based antituberculosis treatment at full induction; 1 = on rifampicin throughout the artemether-lumefantrine course).
• Notes: Distinct from CONMED_RIF_LPVR4 (which is a compound indicator for rifampicin + super-boosted LPV/r 4:4 used by Tikiso 2021 pediatric abacavir popPK and carries a non-CL effect). The acute-vs-chronic semantic distinction is per-paper – document the exact window (rifampicin-dose time and effect duration; lag days to full induction) in covariateData[[CONMED_RIF]]$notes so downstream simulations can construct the correct event-table indicator. Future popPK extractions that hold rifampicin co-administration as a binary indicator should reuse this canonical; if a future model needs a paper-specific time-decaying induction trajectory rather than a step indicator, register a separate canonical (CONMED_RIF_INDUCTION_TIME). Specific scope because the effect-window definition is paper-specific.

CONMED_RPT (canonical for concomitant rifapentine co-administration at full CYP3A4 induction)

• Description: 1 = subject is on concomitant rifapentine (typically daily 600 mg oral rifapentine) and has reached the post-induction equilibrium state on the perpetrator drug’s enzyme-induced expression; 0 = subject is not on rifapentine or is still inside the pre-induction lag window.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no rifapentine co-administration or pre-induction lag).
• Source aliases:
  • RPT – used in Svensson_2014_bedaquiline.R (paper-defined indicator switching on at day 3 of 600 mg daily rifapentine co-administration; same lag as the sibling CONMED_RIF arm of the same paper).
• Example models: Svensson_2014_bedaquiline.R (multiplicative factor on apparent CL_BDQ and CL_M2: cl_eff = cl_base * 3.96^CONMED_RPT; 3.96-fold induction of apparent clearance at full induction).
• Notes: Sibling canonical to CONMED_RIF. Rifapentine and rifampicin are both CYP3A4-inducing rifamycins with similar mechanisms; the per-paper magnitude differs (rifapentine often slightly less induction than rifampicin per the Svensson 2014 fit). Like CONMED_RIF, this is a step indicator at full induction with the paper-specific lag documented in covariateData[[CONMED_RPT]]$notes. Specific scope until a second model ratifies the canonical; ratified on 2026-05-21 alongside the Svensson 2014 extraction.

CONMED_RIF_LPVR4 (canonical for concomitant rifampicin-based antitubercular treatment with super-boosted lopinavir/ritonavir 4:4 indicator)

• Description: 1 = subject is receiving rifampicin-based antitubercular treatment together with super-boosted lopinavir/ritonavir 4:4 (extra ritonavir added to standard 4:1 LPV/r to counter rifampicin-driven LPV induction); 0 = subject is on the comparator regimen specified by the source paper (typically standard LPV/r 4:1 without rifampicin in Tikiso 2021). Used to flag the combined induction effect of rifampicin (PXR-mediated UGT induction) plus extra ritonavir on co-administered antiretroviral PK.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-RIF + standard LPV/r 4:1 reference).
• Source aliases:
  • RIF – used in Tikiso_2021_abacavir.R (the dataset’s paper-defined indicator, 1 = on rifampicin-based TB treatment with super-boosted LPV/r 4:4, 0 = standard LPV/r 4:1 or EFV).
• Example models: Tikiso_2021_abacavir.R (multiplicative effect on bioavailability: f_depot *= (1 + (-0.294) * CONMED_RIF_LPVR4); -29.4% relative to the LPV/r 4:1 reference).
• Notes: Specific scope because the joint rifampicin + super-boosted-LPV/r contrast is paper-defined; the underlying induction is plausibly rifampicin-driven (the paper’s discussion notes that LPV concentrations were similar with vs without super-boosting, weakening a separate ritonavir contribution), but the canonical column captures the joint indicator the paper modeled. Future TB/HIV co-treatment models that test the same regimen contrast should extend the example list rather than register a new canonical.

CONMED_RITUX (canonical for concomitant rituximab combination therapy)

• Description: 1 = on concomitant rituximab combination therapy (with or without backbone chemotherapy), 0 = not on rituximab.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant rituximab; single-agent or non-rituximab combination).
• Source aliases:
  • RITUX – used in Wu_2024_inotuzumab.R.
• Example models: Wu_2024_inotuzumab.R (additive fractional change on CL1: CL1 * (1 + (-0.132) * CONMED_RITUX) ~= 13% lower CL1 with concomitant rituximab).
• Notes: Wu 2024 Table 3 footnote b explicitly flips the reference category vs. the predecessor Garrett 2019 adult model: in Garrett 2019 the reference was “with rituximab” (RITUX = 0 meant on-rituximab), whereas in Wu 2024 the reference is “without rituximab” (RITUX = 0 means no concomitant rituximab). Future models that pool an analogous rituximab-combination cohort with a single-agent reference should use this canonical with the Wu 2024 sign convention; if a paper retains the Garrett 2019 reverse-coded convention, document the value transformation in covariateData[[CONMED_RITUX]]$notes (CONMED_RITUX = 1 - source$RITUX) rather than registering a second canonical. Li_2017_nhl_pfs_mbma.R is the first consumer at the study-arm rather than the subject level (the cohort’s regimen includes rituximab) and the first to need the reverse-coded transformation in practice: its Table S1 tabulates No rituximab on lambda, so the model multiplies the coefficient by (1 - CONMED_RITUX) while the column keeps the canonical polarity. Siblings for the regimen levels that paper pools alongside rituximab: CONMED_CHOP, CONMED_BENDAMUSTINE, CONMED_NONCHEMO_OTHER.

CONMED_CHOP (canonical for concomitant CHOP or CHOP-like chemotherapy backbone indicator)

• Description: 1 = the regimen includes CHOP (cyclophosphamide, doxorubicin/hydroxydaunorubicin, vincristine, prednisone) or a CHOP-like variant, 0 = otherwise.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no CHOP or CHOP-like backbone).
• Source aliases:
  • CHOP/CHOP-like – Li 2017 Table 1 Type of treatment block and Table S1 coefficient row.
• Example models: Li_2017_nhl_pfs_mbma.R (study-arm-level; log-linear coefficient +0.819 on the Weibull scale lambda of progression-free survival, i.e. exp(0.819) = 2.27-fold longer median PFS time).
• Notes: A regimen-composite indicator in the CONMED_<regimen> sub-family alongside CONMED_AVD and CONMED_EOX, not a single-INN column. CHOP-like is deliberately pooled with CHOP because that is how the founding source modelled it: Li 2017 Methods defines CHOP-like as a regimen sharing 3 of the 4 CHOP drugs, and no finer decomposition is recoverable from the paper. A model needing a strict-CHOP-only column must register a separate canonical rather than narrowing this one. Not mutually exclusive with CONMED_RITUX: an R-CHOP arm carries both at 1 and receives both coefficients additively. Distinct from CONMED_DOXORUBICIN and CONMED_EPIRUBICIN, which record a single anthracycline rather than the four-drug regimen.

CONMED_BENDAMUSTINE (canonical for concomitant bendamustine coadministration indicator)

• Description: 1 = the regimen includes bendamustine (a bifunctional alkylating agent with purine-analogue features, used in indolent NHL and CLL), 0 = otherwise.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no bendamustine).
• Source aliases:
  • Bendamustine – Li 2017 Table 1 Type of treatment block and Table S1 coefficient row.
• Example models: Li_2017_nhl_pfs_mbma.R (study-arm-level; log-linear coefficient +1.39 on the Weibull scale lambda of progression-free survival, i.e. exp(1.39) = 4.01-fold longer median PFS time – the largest treatment effect in that model).
• Notes: Standard CONMED_<INN> member. In the founding model this is the smallest treatment stratum (11 of 155 cohorts, 7%), which is why its coefficient is also the least stable: it moves from 1.39 in the final model to 0.526 (RSE 54%) in that paper’s complete-covariate sensitivity analysis. Treat the point estimate accordingly.

CONMED_NONCHEMO_OTHER (canonical for residual ‘other non-chemotherapy anticancer drug’ coadministration indicator)

• Description: 1 = the regimen includes at least one non-chemotherapy anticancer agent that the source did not model as its own level, 0 = otherwise. A residual-bucket indicator whose membership is defined by the source’s own enumeration.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no other non-chemotherapy drug).
• Source aliases:
  • Other drugs – Li 2017 Table 1 Type of treatment block and Table S1 Other drugs on lambda.
• Example models: Li_2017_nhl_pfs_mbma.R (study-arm-level; log-linear coefficient +0.813 on the Weibull scale lambda of progression-free survival, i.e. exp(0.813) = 2.25-fold longer median PFS time; present in 60 of 155 cohorts, 38.7%), Zhang_2022_ici_irae_mbma.R, Zhang_2022_ici_irae_grade3_mbma.R (study-arm level; membership is the cohorts Zhang 2022 Supplementary Table S4 labels ‘target’ therapy given with the immune checkpoint inhibitor, agents not named individually; shares one coefficient with CONMED_CHEMO through max(CONMED_CHEMO, CONMED_NONCHEMO_OTHER)).
• Notes: Residual-bucket indicator, following the TUMTP_OTHER / TUMTP_PTCL_OTHER pattern on the covariate side. Its membership is source-specific and MUST be enumerated per model in covariateData[[CONMED_NONCHEMO_OTHER]]$notes – in Li 2017 it pools temsirolimus, GM-CSF, enzastaurin, celecoxib, tositumomab, alisertib, lenalidomide, pidilizumab, idelalisib, flavopiridol, cladribine, ibrutinib, everolimus, SAM486A, dacetuzumab, bevacizumab, galiximab, obinutuzumab, inotuzumab, thalidomide, epratuzumab, dexamethasone and ofatumumab under one coefficient, and explicitly EXCLUDES rituximab and bortezomib because those carry their own levels. Because one estimate spans that whole list, the column must never be read as a class effect for any single agent. Distinct from CONMED_CHEMO, which means any chemotherapy backbone rather than a non-chemotherapy residual.

CONMED_RTV (canonical for concomitant ritonavir (CYP3A4 inhibitor / PK-booster) coadministration indicator)

• Description: 1 = subject is receiving concomitant ritonavir (RTV), typically at low “booster” doses (100 mg twice daily) as a pharmacokinetic enhancer of co-administered HIV protease inhibitors or other CYP3A4-metabolised antiretrovirals; 0 = no ritonavir. Ritonavir is a potent CYP3A4 inhibitor and P-glycoprotein modulator, so the indicator flags reduced CYP3A4-mediated clearance (and potential bioavailability changes) of the perpetrator-sensitive co-administered drug.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant ritonavir).
• Source aliases:
  • PI / PIs – used in Csajka_2004_indinavir.R (the dataset’s binary protease-inhibitor / ritonavir-presence indicator; all 177 RTV-positive patients in Csajka 2004 received ritonavir specifically, so the column captures concomitant ritonavir).
• Example models: Csajka_2004_indinavir.R (multiplicative effect on apparent oral clearance: cl *= (1 + e_rtv_cl * CONMED_RTV) with e_rtv_cl = -0.63, i.e. ritonavir reduces indinavir CL/F by ~63% relative to the no-RTV reference; Csajka 2004 Table 3), Colombo_2006_atazanavir.R (linear-deviation effect on apparent oral clearance: cl = exp(lcl) * (1 + e_rtv_cl * CONMED_RTV) with e_rtv_cl = -0.46; a 46% reduction in atazanavir CL/F when RTV is coadministered), Mukherjee_2018_amlodipine.R (the DDI-ratio-calibrated encoding: a PBPK paper that publishes only predicted exposure RATIOS, not perpetrator kinetics, so the indicator carries two back-solved multipliers - 1.3783 on relative bioavailability and 0.6045 on clearance - chosen jointly to reproduce the paper’s own predicted amlodipine Cmax ratio of 1.42 and AUC-infinity ratio of 2.28 under ritonavir 100 mg once daily. Splitting the activity across bioavailability and clearance rather than lumping it into clearance is what lets one binary column match a Cmax ratio and an AUC ratio that differ by 1.6-fold), Kengo_2025_atazanavir_a5231.R (the sign-inverted encoding of the same effect, and the reason the register keeps the coefficient’s published polarity rather than normalising it: Kengo 2025 Table S3 fixes atazanavir clearance at the ritonavir-BOOSTED DERIVE value and estimates a 2.13-fold increase attributable to the ABSENCE of ritonavir, so the model applies 1 + (e_rtv_cl - 1) * (1 - CONMED_RTV) with e_rtv_cl = 2.13 and the reference category is 1, not 0. Note that 1/2.13 = 0.47, i.e. a 53% clearance reduction on ritonavir, which is the same order as the Csajka and Colombo estimates read in the usual direction).
• Notes: Follows the CONMED_* concomitant-medication pattern (AED / AMIO / AZA / AZOLE / CBZ / EFV / etc.). Use CONMED_RTV for the binary “is ritonavir co-administered” question; for continuous ritonavir exposure as a covariate use the separate AUC_RTV canonical (e.g. Dickinson_2009_atazanavir.R). The two canonicals are not synonyms – AUC_RTV carries dose-response information that a binary CONMED_RTV flag intentionally collapses. Distinct from CONMED_RIF_LPVR4, which is a joint rifampicin + super-boosted-LPV/r indicator.

CONMED_SGLT2I (canonical for concomitant SGLT2 inhibitor coadministration indicator)

• Description: 1 = subject coadministered a sodium-glucose cotransporter 2 inhibitor (canagliflozin, dapagliflozin, empagliflozin, ertugliflozin, or another systemic SGLT2 inhibitor) during the study, 0 = no concomitant SGLT2 inhibitor. The per-model notes field should document the per-study cumulative-exposure threshold if one is applied (e.g., FIDELIO-DKD used “long-term use” = “SGLT2 inhibitor present for at least 50% of the on-finerenone treatment period”). Distinct from the CINH plasma SGLT-inhibitor concentration regressor used in renal-glucose-reabsorption QSP models – that canonical carries continuous time-varying plasma exposure of the SGLT2 inhibitor itself, whereas CONMED_SGLT2I is a binary drug-drug-interaction indicator for the SGLT2i as a coadministered drug.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant SGLT2 inhibitor coadministration).
• Source aliases:
  • SGLT / SGLT_INH0 / GLYCSI_CATN = 2 – used in vandenBerg_2021_finerenone.R (the FIDELIO-DKD popPK NONMEM control stream’s three-level SGLT categorical with 2 flagging long-term SGLT2i use).
• Example models: vandenBerg_2021_finerenone.R (multiplicative effect 1.10^CONMED_SGLT2I on apparent CL/F AND on relative bioavailability F1 inversely; long-term SGLT2 inhibitor coadministration in the FIDELIO-DKD analysis reduces steady-state finerenone AUC by ~17.1% per Table 2 / Figure 3 of the source paper).
• Notes: Follows the CONMED_<concept> concomitant-medication pattern. When a model needs the on-treatment change in SGLT2 inhibitor use rather than its absolute level, pair this column with CONMED_SGLT2I_BASE. SGLT2 inhibitors are a class of anti-diabetic agents with cardio-renal protective effects; concomitant use is increasingly common in chronic kidney disease and type 2 diabetes populations, where they overlap therapeutically with mineralocorticoid receptor antagonists (e.g., finerenone). The mechanism by which SGLT2i coadministration modestly alters finerenone PK is hypothesised by van den Berg 2021 to relate to SGLT2i-mediated partial reversal of uremic-toxin-induced CYP3A4 inhibition (Discussion paragraph 4.1 ‘The small effect of SGLT2i use on clearance might also be attributable to this phenomenon, as at least pre-clinical data indicate that SGLT2i use may partly reverse this pathophysiology’). Distinct from DIS_DIAB (diabetes-mellitus comorbidity) and from CINH (continuous plasma SGLT2-inhibitor concentration regressor).

CONMED_SGLT2I_BASE (canonical for SGLT2 inhibitor use at treatment start)

• Description: 1 = the subject was already using a sodium-glucose cotransporter 2 inhibitor at the start of study treatment (randomisation), 0 = not using one at that time. Time-fixed per subject. Exists so that a model carrying the TIME-VARYING CONMED_SGLT2I flag can apply a drug-drug-interaction effect to the on-treatment CHANGE in SGLT2 inhibitor use, theta * (CONMED_SGLT2I - CONMED_SGLT2I_BASE), rather than to its absolute level. That centring is required whenever the model’s estimated baseline state already embeds the SGLT2 inhibitor’s effect for subjects who were on one at randomisation – applying the raw flag would double-count it.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (SGLT2-inhibitor-naive at treatment start).
• Source aliases:
  • SGLTSTART – Goulooze 2022 convention in both the UACR and eGFR control streams.
• Example models: Goulooze_2022_finerenone_uacr.R, Goulooze_2022_finerenone_egfr.R (both apply the acute SGLT2 inhibitor effects as theta * (CONMED_SGLT2I - CONMED_SGLT2I_BASE); the eGFR model applies its CHRONIC-slope SGLT2i effect to the raw CONMED_SGLT2I flag instead, so a model may legitimately use both forms).
• Notes: Member of the _BASE baseline-suffix family alongside CRCL_BASE; the _BASE suffix marks the time-fixed treatment-start value of a column that is otherwise time-varying. Do not use this column as a substitute for CONMED_SGLT2I when a model only needs the time-fixed baseline status and never the time course – use CONMED_SGLT2I with a note that it is time-fixed in that analysis.

CONMED_SPART (canonical for spartalizumab (PDR001, anti-PD-1) coadministration indicator)

• Description: 1 = the analyzed therapeutic mAb is coadministered with spartalizumab (PDR001, anti-PD-1 IgG4), 0 = no spartalizumab coadministration. Time-fixed per subject in source analyses to date.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no spartalizumab coadministration – monotherapy or combination with non-spartalizumab agents such as a hypomethylating agent).
• Source aliases:
  • HASPDR – used in Xu_2023_MBG453.R (Monolix supplement Appendix S2; the source describes the column as “this patient HAS received PDR001 [spartalizumab, anti PD-1 mAb]”).
• Example models: Xu_2023_MBG453.R (exponential effect on CL: exp(0.0194 * CONMED_SPART); not statistically significant in the full covariate model but retained because Xu 2023 used the full-covariate-model approach).
• Notes: Parallels COMBO_NIVO (ipilimumab + nivolumab) and COMBO_DURVA (durvalumab combinations) but for spartalizumab. Promote to general scope if a second paper reports a spartalizumab-coadministration covariate with comparable encoding.

CONMED_SPIRON (canonical for concomitant spironolactone coadministration indicator)

• Description: 1 = subject is coadministered spironolactone (aldosterone-receptor antagonist; potassium-sparing diuretic; inhibitor of renal P-glycoprotein and of CYP3A4) during the study, 0 = no concomitant spironolactone. The covariate is not tied to one victim drug: it has been reported as a clearance or volume covariate for digoxin (renal tubular P-glycoprotein secretion), vancomycin (volume of distribution in neonates) and cyclosporin (CYP3A4 / P-glycoprotein-mediated apparent oral clearance).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant spironolactone).
• Source aliases:
  • SPI – used in Zhou_2010_digoxin.R (Zhou 2010 Table 1 and Table 7 column label).
  • spirolactone – used in Wang_2019_cyclosporin.R (Wang 2019 spells the drug “spirolactone” in the title, equations and Tables II-III; the rINN is spironolactone).
• Example models: Zhou_2010_digoxin.R (multiplicative linear-deviation form on Cl/F: cl *= (1 - 0.412 * CONMED_SPIRON), i.e. ~41% lower Cl/F with concomitant spironolactone in the older Chinese CHF cohort; Zhou 2010 Table 7), MarquesMinana_2010_vancomycin.R (multiplicative linear-deviation form on weight-normalised V_d: vc *= (1 - 0.344 * CONMED_SPIRON), i.e. ~34% smaller V_d with concomitant spironolactone in the 70-neonate Marques-Minana 2010 cohort; Table 3 theta4), Wang_2019_cyclosporin.R (multiplicative linear-deviation form on CL/F: cl *= (1 + CONMED_SPIRON * -0.265), i.e. ~26.5% lower apparent oral clearance with concomitant spironolactone in 18 Chinese children with refractory nephrotic syndrome; Wang 2019 Table III).
• Notes: 32 of 119 subjects (27%) in Zhou 2010 were coadministered spironolactone; 11 of 18 (61%) in Wang 2019. The clinical rationale in Zhou 2010 (Discussion) is that spironolactone inhibits the renal-tubular secretion of digoxin via competition for the renal P-glycoprotein transporter, raising digoxin steady-state concentration; Wang 2019 attributes the cyclosporin interaction to the same transporter plus CYP3A4. Scope was promoted from specific to general when the third independent popPK paper (Wang 2019) reported the covariate with the same binary, linear-deviation-multiplicative encoding as the two earlier entries, satisfying the promotion criterion originally recorded here. All three published effects are negative on the affected parameter, but the sign is not part of the canonical definition – record it per model.

CONMED_STATIN (canonical for concomitant conmed_statin (HMG-CoA reductase inhibitor) therapy)

• Description: 1 = patient coadministered a conmed_statin (HMG-CoA reductase inhibitor) during the study, 0 = no conmed_statin coadministration.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no conmed_statin coadministration).
• Source aliases: none known.
• Example models: Martinez_2019_alirocumab.R (additive effect on linear clearance CLL: CLL = TVCLL + COV1*(WT-82.9) + COV2*CONMED_STATIN; +0.00644 L/h when conmed_statin is coadministered), Wright_2025_colchicine.R (multiplicative fractional effect on colchicine clearance only: cl <- ... * (1 - (1 - e_conmed_statin_cl) * CONMED_STATIN) with e_conmed_statin_cl = 0.66, a ~30% reduction; any concomitant statin at any dose counts, the retained cohort being atorvastatin (n = 12) and simvastatin (n = 4), Wright 2025 Table 2 footnote c), Nicolas_2019_alirocumab.R (additive effect of +0.408 on the LDL-C Emax of the Part II indirect response model, +16.8% at median covariates; note this paper’s definition is BROADER than its own Part I companion’s – Nicolas 2019 Table 2 footnote c counts any statin at any dose, whereas Martinez_2019_alirocumab.R excludes rosuvastatin >= 20 mg/day and atorvastatin >= 40 mg/day, so the same column name carries different memberships in the two halves of one pooled analysis and must be rebuilt per model).
• Notes: Per-model covariateData[[CONMED_STATIN]]$notes must document which statins and dose thresholds are included in the “CONMED_STATIN = 1” category, since inclusion criteria vary by study. Martinez 2019 codes CONMED_STATIN = 1 for coadministration of rosuvastatin (< 20 mg/day), atorvastatin (< 40 mg/day), or simvastatin (any dose); other conmed_statin regimens are coded as 0.

CONMED_STATIN_MONO (canonical for concomitant statin-monotherapy indicator)

• Description: 1 = patient is on a conmed_statin and no other lipid-lowering comedication (conmed_statin monotherapy), 0 = not on conmed_statin monotherapy (either on no lipid-lowering therapy or on a multi-drug lipid-lowering combination such as conmed_statin + ezetimibe).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (not on conmed_statin monotherapy).
• Source aliases:
  • Derived from a conmed_statin-identifier column in the source (any of atorvastatin, rosuvastatin, simvastatin, lovastatin, pravastatin, pitavastatin, fluvastatin) with an AND over “no other lipid-lowering comedication”.
• Example models: Kuchimanchi_2018_evolocumab.R (multiplicative effect 1.13 on Vmax: Vmax * 1.13^CONMED_STATIN_MONO), Budha_2015_rg7652.R (time-varying within-subject multiplicative log-effect on the indirect-response Ksyn anchor for LDL-C: Ksyn = Kdeg * LDLC * exp(-0.648 * CONMED_STATIN_MONO(t)); same parameter dynamically reproduces both the ~45% lower pre-dose baseline LDL-C in atorvastatin-pretreated cohorts and the LDL rebound after statin cessation, Budha 2015 Table II ‘Statin on screening LDLc’ = -0.648), Kuchimanchi_2018_evolocumab_ldlc.R (same PK-layer effect plus an exponent 0.797 on baseline LDL-C and 0.937 on Emax in the exposure-response layer, Table 4).
• Notes: Scope: specific because both registered example models narrowly define the statin covariate as monotherapy only – Kuchimanchi 2018 (“patients on a conmed_statin only and no other comedication”); Budha 2015 (atorvastatin 40 mg daily as the sole concomitant lipid-lowering therapy in the Phase 1 statin cohorts). Mutually compatible with CONMED_EZE: a subject on conmed_statin+ezetimibe has CONMED_STATIN_MONO = 0 and CONMED_EZE = 1; a subject on conmed_statin alone has CONMED_STATIN_MONO = 1 and CONMED_EZE = 0; a subject on no lipid-lowering therapy has both 0. Per-model temporal grain (time-fixed per subject vs time-varying within subject) is documented in covariateData[[CONMED_STATIN_MONO]]$notes per model. Future popPK/PD models that adopt a broader “any conmed_statin” definition should register a separate CONMED_STATIN or CONMED_STATIN canonical rather than reusing this name.

CONMED_TGLOWER (canonical for composite background triglyceride-lowering therapy indicator)

• Description: 1 = patient is on one or any combination of background triglyceride-lowering therapies, 0 = on none. The composite is deliberately class-agnostic: the constituent classes are fibrates, omega-3 fatty acids (including icosapent ethyl / Vascepa and fish oil), and statins. Named for the therapeutic INTENT the source papers screen – triglyceride lowering – rather than for any single molecule or class.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no background triglyceride-lowering therapy).
• Source aliases:
  • TG-lowering therapy – Ousey 2026 Table 2 covariate row name.
  • Background TG-lowering therapies – Ousey 2026 Table 1 row name.
• Example models: Ousey_2026_plozasiran.R (multiplicative exponential effect on the APOC3-on-triglyceride half-maximal level of a cascading K-PD indirect-response model: ki50_apoc3 = 17.4 * exp(-1.79 * CONMED_TGLOWER), i.e. 2.91 mg/dL on therapy vs 17.4 mg/dL off therapy, so concomitant therapy potentiates rather than antagonises the drug effect; 55/75 patients (73.3%) on therapy, Ousey 2026 Table 2 and Table 1 footnote a).
• Notes: Composite drug-class conmed indicator, in the same family as CONMED_ABX (“any-other-antibiotic composite”), CONMED_DIURETIC (“class composition is paper-specific”), CONMED_AZOLE and CONMED_IMMUNOMOD. Use this name when a source reports only a single pooled Yes/No triglyceride-lowering-therapy row and fits exactly one coefficient for it; per-model class composition is documented in covariateData[[CONMED_TGLOWER]]$notes. Do NOT decompose into per-class indicators unless the source actually reports and fits them separately – Ousey 2026 explicitly declines to attribute the effect (“there was no definitive evidence to indicate which, if any, specific TG-lowering therapy was primarily responsible”). Distinct from and compatible with the narrower registered lipid conmeds: CONMED_STATIN (any statin), CONMED_STATIN_MONO (statin with no other lipid-lowering comedication – mutually exclusive in practice with a multi-class CONMED_TGLOWER = 1), the CONMED_STATIN_LI/MI/HI intensity strata, CONMED_EZE, and the CONMED_<INN> binaries. Also distinct in intent from an LDL-focused umbrella: a future paper screening a broad lipid-lowering composite aimed at LDL-C rather than triglycerides should register a separate canonical rather than reusing this name.

CONMED_LIPIDLOWER (canonical for composite antihyperlipidemic / lipid-lowering therapy indicator)

• Description: 1 = patient is receiving any antihyperlipidemic (lipid-lowering) medication, pooled across drug classes; 0 = receiving none. The composite is deliberately class-agnostic and is the broad LDL-focused umbrella that CONMED_TGLOWER’s notes anticipate: use it when a source reports a single pooled “antihyperlipidemic” / “lipid-lowering therapy” Yes/No row and fits exactly one coefficient for it.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no antihyperlipidemic therapy).
• Source aliases:
  • Antihyperlipidemic – Elhefnawy 2023 Table 1 / Table 3 covariate row name.
  • Lipid-lowering therapy, LLT – common registry and clinical-dataset labels for the same pooled indicator.
• Example models: Elhefnawy_2023_recurrent_ischemic_stroke_nondm.R (multiplicative log-linear effect on the Gompertz hazard of recurrent ischemic stroke among patients without diabetes: exp(-0.835 * CONMED_LIPIDLOWER), adjusted hazard ratio 0.433, i.e. a 43.3% reduction in the recurrence hazard; Elhefnawy 2023 Cureus 15(12):e50794 Table 3), Elhefnawy_2023_recurrent_ischemic_stroke_dm.R.
• Notes: Composite drug-class conmed indicator in the same family as CONMED_ABX, CONMED_DIURETIC, CONMED_AZOLE, CONMED_IMMUNOMOD and CONMED_TGLOWER. Per-model class composition is documented in covariateData[[CONMED_LIPIDLOWER]]$notes; registry sources frequently do not enumerate the constituent agents at all, in which case say so rather than assuming statins only. Distinct from CONMED_STATIN (any statin specifically), CONMED_STATIN_MONO (statin with no other lipid-lowering comedication), the CONMED_STATIN_LI/MI/HI intensity strata, CONMED_EZE and the CONMED_<INN> binaries – reach for one of those whenever the source actually identifies the molecule or class, and reserve CONMED_LIPIDLOWER for the undifferentiated pooled row. Distinct in intent from CONMED_TGLOWER, which names a triglyceride-lowering composite: a source that separates the two intents should use both columns. Also distinct from the disease-state flag DIS_HYPERLIP, which records the diagnosis rather than its treatment; the two frequently appear together and are not redundant (Elhefnawy 2023 fits both in the same hazard, in opposite directions).

CONMED_STATIN_LI, CONMED_STATIN_MI, CONMED_STATIN_HI (canonical for concomitant statin-intensity stratum indicators)

• Description: Three mutually exclusive binary indicators identifying the therapeutic-intensity stratum of a patient’s concomitant statin regimen, following the low- / moderate- / high-intensity classification of the 2018 ACC/AHA cholesterol guideline (Grundy et al., J Am Coll Cardiol 2019;73:e285-e350). CONMED_STATIN_LI = 1 for a low-intensity regimen (roughly < 30% expected LDL-C lowering; e.g. simvastatin 10 mg, pravastatin 10-20 mg, lovastatin 20 mg, fluvastatin 20-40 mg, pitavastatin 1 mg). CONMED_STATIN_MI = 1 for a moderate-intensity regimen (~30-49% expected lowering; e.g. atorvastatin 10-20 mg, rosuvastatin 5-10 mg, simvastatin 20-40 mg, pravastatin 40-80 mg, lovastatin 40 mg, fluvastatin 40 mg BID or XL 80 mg, pitavastatin 2-4 mg). CONMED_STATIN_HI = 1 for a high-intensity regimen (>= 50% expected lowering; e.g. atorvastatin 40-80 mg, rosuvastatin 20-40 mg). A patient not on any concomitant statin has all three set to 0.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 for all three simultaneously = no concomitant statin therapy. At most one of the three may be 1 for a given patient/record; a model that references more than one of them multiplies the corresponding effects, so a data set that sets two to 1 is malformed.
• Source aliases:
  • Low-intensity statin, Moderate-intensity statin, High-intensity statin – Jadhav 2023 Table 3 covariate row names.
  • STATINT / STATIN_INTENSITY (a single ordinal 0/1/2/3 column) – decompose into the three binaries rather than carrying the ordinal, per the register’s decomposed-indicator convention.
• Example models: Jadhav_2023_bempedoicAcid_ldlc.R (proportional shifts on the indirect-response Imax of -0.238 / -0.302 / -0.424 and on baseline LDL-C of -0.159 / -0.268 / -0.293 for low / moderate / high intensity; Jadhav 2023 Table 3), Nicolas_2019_alirocumab.R.
• Notes: Distinct from CONMED_STATIN (any concomitant statin, undifferentiated by intensity) and from CONMED_STATIN_MONO (statin monotherapy, i.e. a statin with no other lipid-lowering comedication) – the intensity strata say nothing about whether other lipid-modifying therapies are co-administered, and the monotherapy flag says nothing about dose intensity, so the three families are orthogonal and may coexist in one model. Also distinct from the CONMED_<INN> binaries (CONMED_ATORVASTATIN, CONMED_SIMVASTATIN,…), which identify the specific molecule irrespective of dose, and from the CONMED_<drug>_DOSE continuous family. The intensity classification is dose- and molecule-dependent, so a source paper’s mapping from regimen to stratum must be recorded per model in covariateData[[CONMED_STATIN_MI]]$notes when it departs from the ACC/AHA table.

CONMED_ATORVASTATIN (canonical for concomitant atorvastatin coadministration indicator)

• Description: 1 = patient is coadministered atorvastatin at the observation / dose record, 0 = no concomitant atorvastatin. Molecule-specific binary companion to the class-level CONMED_STATIN flag and the intensity strata.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant atorvastatin).
• Source aliases:
  • Atorvastatin – Jadhav 2023 Table 2 covariate row name.
  • ATORVA, ATOR – common dataset abbreviations. The three-letter form ATV is NOT an acceptable alias: it is reserved for the atazanavir indicator CONMED_ATAZANAVIR (see the CONMED_<drug>_DOSE notes on that collision), which is why this entry spells out the full INN.
• Example models: Jadhav_2023_bempedoicAcid.R, Jadhav_2023_bempedoicAcid_ldlc.R (proportional shift on relative oral bioavailability F1 * (1 + 0.142 * CONMED_ATORVASTATIN), i.e. 14.2% higher relative bioavailability of bempedoic acid with concomitant atorvastatin relative to the F1 = 1 no-atorvastatin anchor; Jadhav 2023 Table 2 and footnote b).
• Notes: Distinct from CONMED_ATORVASTATIN_DOSE (continuous mg/day), which should be used instead when a model scales an effect with the atorvastatin dose rather than switching on its presence.

CONMED_SIMVASTATIN (canonical for concomitant simvastatin coadministration indicator)

• Description: 1 = patient is coadministered simvastatin at the observation / dose record, 0 = no concomitant simvastatin. Molecule-specific binary companion to the class-level CONMED_STATIN flag and the intensity strata.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant simvastatin).
• Source aliases:
  • Simvastatin – Jadhav 2023 Table 2 covariate row name.
  • SIMVA, SMV – common dataset abbreviations.
• Example models: Jadhav_2023_bempedoicAcid.R, Jadhav_2023_bempedoicAcid_ldlc.R (proportional shift on apparent central volume Vc/F * (1 + (-0.154) * CONMED_SIMVASTATIN), i.e. 15.4% lower Vc/F with concomitant simvastatin; Jadhav 2023 Table 2).
• Notes: Distinct from CONMED_SMV_DOSE (continuous mg/day).

CONMED_STEROID (canonical for systemic corticosteroid administration indicator)

• Description: 1 = patient is on systemic corticosteroid therapy at the observation (or, depending on the paper’s encoding, during the time interval the observation summarises), 0 = no systemic corticosteroid administration. Supports two temporal grains depending on the source paper: (i) time-fixed per subject, capturing baseline / chronic concurrent corticosteroid use in diseases where background steroid use is standard of care (SLE, severe asthma); (ii) time-varying per record, capturing acute corticosteroid pulses for relapse / flare treatment in diseases where steroids are administered as a per-event course (multiple sclerosis acute relapse, autoimmune flares).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no systemic corticosteroid administration in the current record / interval).
• Source aliases:
  • BSTEROID – used in Narwal_2013_sifalimumab.R and Zheng_2016_sifalimumab.R (time-fixed baseline-use form).
  • STEROID – used in VelezdeMendizabal_2013_multipleSclerosis.R (time-varying per-monthly-record acute-administration form).
• Example models: Narwal_2013_sifalimumab.R (time-fixed multiplicative on CL: CL * (1 + 0.195 * CONMED_STEROID)), Zheng_2016_sifalimumab.R (time-fixed multiplicative on CL (1 + 0.11 * CONMED_STEROID) and on V1 (1 - 0.09 * CONMED_STEROID) in the SLE phase IIb cohort, which was ~85% conmed_steroid-treated at baseline), VelezdeMendizabal_2013_multipleSclerosis.R (time-varying per-monthly-record switch of the first-order Markov coefficient from theta_pdv to theta_pdv_s when a corticosteroid course was given for a clinical MS relapse that month).
• Notes: Distinct from PRICORT, which is strictly a prior (pre-study) indicator. CONMED_STEROID covers both concurrent chronic corticosteroid use at / from study baseline and per-record acute corticosteroid pulses; the per-model covariateData[[CONMED_STEROID]]$notes field documents the temporal grain (time-fixed vs time-varying) the source paper used. When a future paper needs CONMED_STEROID and PRICORT jointly, both can coexist on the same subject. The name STEROID_BL was used as an alias in earlier register drafts and is retired; use CONMED_STEROID for all future models.

CONMED_DEXAMETHASONE (canonical for dexamethasone co-medication indicator)

• Description: 1 = subject received dexamethasone as co-medication in the window the source paper defines, 0 = not. In the founding model the window is CRS prophylaxis with the first four epcoritamab doses in Cycle 1 (premedication before each dose and on the following days), and the indicator is time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no dexamethasone in the defined window). In Li 2026, patients premedicated with prednisolone and not dexamethasone are 0.
• Source aliases:
  • CRS prophylaxis (dexamethasone/IV fluids) – Li 2026 Table 1 four-level category; the dexamethasone half is this column, the IV-fluids half is CONMED_IV_FLUIDS.
• Example models: Li_2026_epcoritamab.R (with CONMED_IV_FLUIDS, forms the paper’s three-level Cycle 1 prophylaxis covariate on S50 of the Grade >= 2 CRS hazard: exactly one of the two raises S50 3.28-fold, both 5.10-fold).
• Notes: Member of the CONMED_<INN> family. Narrower than CONMED_STEROID (any systemic corticosteroid): use this column when a source distinguishes dexamethasone from other corticosteroids, as Li 2026 does (prednisolone users are not counted). Record the exposure window (prophylaxis vs treatment, which cycle) in covariateData[[CONMED_DEXAMETHASONE]]$notes. Distinct from COMBO_LEN_DEX / COMBO_POM_DEX, which name dexamethasone as part of an anti-myeloma combination regimen.

CONMED_IV_FLUIDS (canonical for prophylactic intravenous fluid hydration indicator)

• Description: 1 = subject received intravenous fluid hydration as co-medication in the window the source paper defines, 0 = not. In the founding model the window is CRS prophylaxis with the first four epcoritamab doses in Cycle 1; time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no IV fluids in the defined window).
• Source aliases:
  • CRS prophylaxis (dexamethasone/IV fluids) – Li 2026 Table 1 four-level category (see CONMED_DEXAMETHASONE).
• Example models: Li_2026_epcoritamab.R (combined with CONMED_DEXAMETHASONE into the three-level Cycle 1 prophylaxis covariate on S50 of the Grade >= 2 CRS hazard).
• Notes: Class-level member of the CONMED_ family (like CONMED_STEROID): the source does not specify the fluid, volume or rate. A binary indicator, distinct from the volume canonicals FLUID_IN_24H, PFA and CUM_FLUID_BAL_PCT; register a volume- or rate-valued sibling if a paper models the amount of fluid given. IV fluids given to manage (rather than prevent) an adverse event were not counted in Li 2026.

CONMED_THEOPHYLLINE (canonical for concomitant theophylline / aminophylline coadministration indicator)

• Description: 1 = subject was taking theophylline or aminophylline as background co-medication, 0 = not taking either. Aminophylline is a theophylline-ethylenediamine salt that dissociates to theophylline in vivo, so source papers that record the two together are recording one exposure and a single indicator is correct. Time-fixed per subject when the source captures baseline background therapy (the usual case in asthma / COPD trials, where the drug is a maintenance controller).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no theophylline or aminophylline co-medication).
• Source aliases:
  • CTHEO – used in Jin_2025_benralizumab_fev1.R; defined in Jin 2025 Resource 1 as ‘CTHEO Theophylline/aminophylline use (0: no, 1: yes)’ and appearing in Resource 4 as the row label Beta_FEVB (CTHEO_1).
• Example models: Jin_2025_benralizumab_fev1.R (log-scale multiplicative effect of -0.089 on baseline pre-bronchodilator FEV1, i.e. an 8.5% lower baseline in users), Jin_2025_benralizumab.R.
• Notes: Specific scope. In severe-asthma models this covariate usually behaves as a disease-severity marker rather than a pharmacological interaction: theophylline is a late-line controller, so its users have worse baseline lung function, and the effect direction (lower baseline FEV1) reflects selection into the therapy rather than any effect of theophylline on the modelled drug. Do not reinterpret a negative coefficient as a drug-drug interaction without the source saying so. Distinct from FORM_THEO_APNECUT, which is a formulation indicator for models in which theophylline itself is the analyte, and from the CONMED_<INN> CYP-modulator entries (CONMED_AZOLE, CONMED_CBZ, …), which are registered for their metabolic-interaction role.

CONMED_STEROID_SPARING (canonical for steroid-sparing immunosuppression-protocol indicator)

• Description: 1 = subject was assigned a steroid-sparing immunosuppressive protocol (corticosteroids administered for a short period – typically <= 7-14 days post-transplant – as opposed to a continuous-steroid regimen); 0 = subject was on a continuous-corticosteroid regimen. Time-fixed per subject (the protocol is assigned at the transplantation date and does not change during the analysis window). Operationally a centre-level attribute in source datasets where the protocol is determined by the transplanting centre (Passey 2011) or a per-subject protocol decision in other datasets.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (continuous-corticosteroid regimen; non-sparing protocol).
• Source aliases:
  • steroid sparing centre – used in Passey_2011_tacrolimus.R (Passey 2011 Methods: “Centres were designated as using a steroid sparing immunosuppressive regimen if they administered steroids for <= 7 days post transplant”).
• Example models: Passey_2011_tacrolimus.R (power-of-binary-indicator multiplicative factor on apparent oral clearance: e_steroid_spare_cl ^ CONMED_STEROID_SPARING with e_steroid_spare_cl = 0.70; steroid-sparing patients have 30% lower apparent oral tacrolimus CL/F than continuous-steroid patients; Passey 2011 Discussion attributes the effect to reduced CYP3A induction in the absence of ongoing corticosteroid therapy).
• Notes: Companion to CONMED_STEROID (which captures concurrent / baseline corticosteroid USE; CONMED_STEROID_SPARING captures the protocol-level decision to MINIMIZE corticosteroid use). The two coexist in the same dataset when needed: a steroid-sparing patient still has CONMED_STEROID = 1 during days 1-7 post-transplant (the short-duration administration window) and CONMED_STEROID = 0 thereafter. The Passey 2011 binary indicator collapses both phases into a single time-invariant per-subject attribute via the centre-level assignment; document the per-model temporal interpretation in covariateData[[CONMED_STEROID_SPARING]]$notes. Distinct from PRICORT (pre-study corticosteroid history) and from HCT_COND_RIC (reduced-intensity conditioning regimen for HSC transplantation, which is a different protocol axis). Future models that distinguish the specific duration of steroid administration (e.g. 7 days vs 14 days vs 30 days) should register companion canonicals rather than overloading CONMED_STEROID_SPARING.

CONMED_STEROID_TOPICAL (canonical for concomitant TOPICAL corticosteroid therapy indicator)

• Description: 1 = concomitant topical corticosteroid (TCS) therapy applies, 0 = it does not. Registered for dermatology models, where TCS is the background standard of care and the decision to permit it is a first-order determinant of measured response. Supports two temporal / hierarchical grains, which must be documented per model: (i) TRIAL-ARM-LEVEL, where the indicator records whether the study protocol PERMITTED background TCS in that arm – the grain a model-based meta-analysis of published aggregate data operates at; (ii) SUBJECT-LEVEL, where it records an individual patient’s actual TCS use, time-fixed at baseline or time-varying per record.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant topical corticosteroid therapy / a monotherapy trial design).
• Source aliases:
  • I_TCS – used in Serrano_2026_atopicDermatitis_placebo_mbma.R (Serrano 2026 Results 3.2: “I_TCS is a binary indicator taking a value of 1 for trials with TCS concomitant therapy and 0 otherwise (trials without concomitant therapy)”).
• Example models: Serrano_2026_atopicDermatitis_placebo_mbma.R (trial-arm-level design flag; additive +0.774 logit-unit shift in the maximum placebo effect on the EASI-75 responder rate, i.e. a 1.84-fold higher EASI-75 placebo rate at Week 12; 14 of 41 placebo arms permitted TCS).
• Notes: Distinct from CONMED_STEROID, which is SYSTEMIC corticosteroid administration recorded per subject. The route and the usual grain both differ and the two columns must never be substituted for one another – topical corticosteroids act locally on the lesion being scored, whereas systemic corticosteroids are a disease-severity / immunosuppression marker in the autoimmune and transplant models CONMED_STEROID was registered for. Sibling of CONMED_STEROID_SPARING (a transplant protocol-level minimisation decision) in the CONMED_STEROID* family; a subject may legitimately carry CONMED_STEROID and CONMED_STEROID_TOPICAL simultaneously. When the indicator is a trial-design permission rather than observed use, say so explicitly in covariateData[[CONMED_STEROID_TOPICAL]]$notes: an arm that permitted TCS is not an arm in which every patient used it, and a future subject-level extraction should not read a permission-grain coefficient as a use-grain one. A future paper that needs to distinguish TCS potency class (e.g. low / mid / high-potency) or a mandated versus as-needed regimen should register companion canonicals rather than overloading this one. Companion to CONMED_TCI should a topical-calcineurin-inhibitor indicator be needed.

CONMED_TELMISARTAN (canonical for concomitant telmisartan coadministration indicator)

• Description: 1 = telmisartan (an angiotensin II receptor blocker and an inhibitor of OATP1B1, NTCP, BSEP, MRP3 and MRP4) is co-administered or co-incubated, 0 = it is not. Telmisartan is a standard probe perpetrator in hepatic-transporter interaction work because it inhibits uptake and efflux simultaneously yet is not hepatotoxic, which makes it a useful negative control for the interplay of multiple transporters.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 = telmisartan absent. Carries no magnitude information: the perpetrator exposure enters through a concentration column and this indicator selects the matching potency parameters.
• Source aliases: none standardized; source papers name the drug in prose or in a treatment-arm label.
• Example models: Guo_2016_taurocholate_schh.R (selects the telmisartan column of the Table 1 inhibitory potencies: NTCP 60 uM, OATP1B1 0.44 uM, BSEP 16.1 uM, MRP3 60 uM).
• Notes: General scope; a well-formed member of the CONMED_<INN> family using the full INN. Mutually exclusive with the other inhibitor indicator in any model that selects a single perpetrator’s potency set. Telmisartan is also extensively bound inside hepatocytes (cellular unbound fraction 0.09-0.13 in the founding example), so a model using this indicator together with CONC_INHIBITOR_CELL_UM must be explicit about whether the supplied intracellular concentration is total or unbound.

CONMED_UGT_INH (canonical for concomitant UGT-inhibitor coadministration indicator (pooled))

• Description: 1 = subject is receiving at least one drug the source paper classifies as a UGT (uridine 5’-diphospho-glucuronosyltransferase) inhibitor at the PK observation, 0 = not on any such drug. Class-level pool paralleling CONMED_EIAED (pooled enzyme-inducing antiepileptic drugs); used when a paper aggregates multiple mechanism-related UGT-inhibiting comedications into a single indicator because per-drug sample sizes are too small to identify separate effects. Time-varying in principle; time-fixed for chronic-maintenance cohorts.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant UGT inhibitor).
• Source aliases:
  • Inh / Inhibitors – used in Milosheska_2016_lamotrigine.R (paper Table 4 footnote *Valproic acid or sertraline (Inh); the paper pools valproic acid (n = 13) and sertraline (n = 2) because both were hypothesised to act via UGT2B7 competitive inhibition and the per-drug sample sizes were too small to estimate distinct effects).
• Example models: Milosheska_2016_lamotrigine.R (multiplicative effect on parent lamotrigine apparent clearance: cl *= (1 - 0.579 * CONMED_UGT_INH); carriers have -57.9% lower CL relative to the no-inhibitor reference, Milosheska 2016 Table 4 row Co-treatment with inhibitors).
• Notes: The per-paper list of drugs counted as UGT inhibitors MUST be documented in covariateData[[CONMED_UGT_INH]]$notes because the pooling criterion varies across studies (Milosheska 2016 pools valproic acid + sertraline; a future paper might pool valproic acid + probenecid + sertraline + fluvoxamine). Distinct from the drug-specific canonical CONMED_VPA (concomitant valproate alone). When a paper reports enough per-drug data to identify separate effects, use the drug-specific canonicals (CONMED_VPA, CONMED_SERTRALINE,…) rather than collapsing into CONMED_UGT_INH. Analogous to CONMED_EIAED (pooled enzyme-inducing AEDs) which serves the same role for the inducer side.

CONMED_TREMELIMUMAB (canonical for concomitant tremelimumab coadministration indicator)

• Description: 1 = subject is receiving tremelimumab (anti-CTLA-4 IgG2 kappa) concomitantly with the analyte under PK study, 0 = no tremelimumab coadministration. Used as a treatment-arm indicator in pooled checkpoint-inhibitor analyses that combine monotherapy and anti-PD-(L)1 + anti-CTLA-4 doublet or triplet arms.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no tremelimumab coadministration – monotherapy or a non-tremelimumab combination).
• Source aliases:
  • comb (three-level combination-therapy column; level 2 in Abegesah 2025, “durvalumab, tremelimumab and chemotherapy”) – decompose into CONMED_TREMELIMUMAB = as.integer(comb == 2).
• Example models: Abegesah_2025_durvalumab.R (multiplicative effect on durvalumab CL of (1 - 0.0929), i.e. -9.3% relative to durvalumab monotherapy; paired with CONMED_CHEMO to reconstruct the paper’s three-level comb categorical).
• Notes: The mirror image of COMBO_DURVA, which marks durvalumab coadministration when the analyte is something else – the two columns describe the same AstraZeneca doublet from opposite sides and both may be needed in a registry that carries Hwang_2022_tremelimumab.R (tremelimumab analyte, durvalumab partner) alongside Abegesah_2025_durvalumab.R (durvalumab analyte, tremelimumab partner). Do not use one for the other. When a paper’s combination column is a three-or-more-level categorical that crosses tremelimumab with a chemotherapy backbone, store the levels as separate canonical columns (CONMED_TREMELIMUMAB and CONMED_CHEMO) and reconstruct the mutually exclusive strata inside model(), rather than inventing a single composite canonical: in Abegesah 2025 no tremelimumab-without-chemotherapy patients exist, so CONMED_TREMELIMUMAB = 1 implies CONMED_CHEMO = 1 and the chemotherapy-only stratum is CONMED_CHEMO * (1 - CONMED_TREMELIMUMAB). Document that crossing in the per-model covariateData notes, because a future cohort containing a tremelimumab-without-chemotherapy arm would make the two columns independent.

CONMED_TPM (canonical for concomitant topiramate coadministration indicator)

• Description: 1 = subject is taking topiramate (TPM) as a concomitant antiepileptic drug at the observation, 0 = no concomitant topiramate. Topiramate is a broad-spectrum antiseizure medication acting partly through voltage-gated sodium-channel blockade – the same principal mechanism as valproate, carbamazepine and phenytoin – so in valproate efficacy models its coefficient is commonly interpreted as mechanistic redundancy rather than a PK interaction. It is a weak and dose-dependent CYP3A4 inducer, so it is not routinely classed with the strong enzyme-inducing AEDs. Time-varying when topiramate starts / stops within the observation window; time-fixed when the source paper analyses chronic-maintenance cohorts whose antiseizure therapy is stable across the analysis window.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant topiramate).
• Source aliases:
  • TPM – used in Nakashima_2015_valproic_acid.R (Nakashima 2015 Eq 6 and Table 3 slope row TPM).
• Example models: Nakashima_2015_valproic_acid.R (modifier of the exposure slope of a logistic exposure-response model for an over-50% reduction in seizure frequency: e_tpm_slope = -1.73 per 100 ug/mL of predicted trough valproate, i.e. concomitant topiramate makes the concentration-response slope more negative; Nakashima 2015 Table 3).
• Notes: Drug-specific CONMED_* indicator anticipated in the [[CONMED_AED]] notes; use it when a paper estimates a separate topiramate effect distinct from the pooled EIAED / AED class effect. Distinct from the broader [[CONMED_EIAED]] and [[CONMED_AED]]; topiramate’s induction is weak and dose-dependent, so it must not be folded into CONMED_EIAED without a source statement that the paper classed it that way. Distinct from [[DOSE_TPM_MGD]], which carries the topiramate daily dose as a continuous covariate – use this binary canonical when the source paper models only the presence of topiramate. Note the direction of the founding effect: it acts on the SLOPE rather than the intercept, so its impact grows with valproate exposure; the founding authors read the CBZ / PHT / TPM effects together as evidence that adding a second sodium-channel blocker to valproate provides no additional benefit (Nakashima 2015 Discussion).

CONMED_VPA (canonical for concomitant valproate (valproic acid) coadministration indicator)

• Description: 1 = subject is taking valproate (valproic acid, sodium valproate, divalproex) as a concomitant antiepileptic drug at the PK observation, 0 = no concomitant valproate. Valproate is a broad-spectrum AED that inhibits UGT and CYP2C9, and chronic use is associated with weight gain. Time-varying when valproate starts / stops within the observation window; time-fixed when the source paper analyses chronic-maintenance cohorts whose AED therapy is stable.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant valproate).
• Source aliases:
  • VPA – used in Schoemaker_2017_brivaracetam.R (paper covariate VPA for valproate coadministration).
  • Regimen – used in Damnjanovic_2023_lamotrigine.R. Damnjanovic 2023 Table 2(b) names the covariate after the therapeutic regimen rather than the drug, but its own footnote resolves it to valproate presence: “factor for the relationship between Cl and therapeutic regimen (whether existence of valproic acid)”. No value transformation: the cohort is entirely on dual therapy, so “regimen contains valproate” and “concomitant valproate” are the same indicator. The reference category is the valproate-FREE stratum, fixed by the Discussion statement that valproate coadministration decreases lamotrigine clearance, which requires the negative coefficient to attach to the valproate-present stratum.
• Example models: Schoemaker_2017_brivaracetam.R (multiplicative effect on apparent oral clearance: cl *= (1 - 0.101 * CONMED_VPA); -10.1% relative to no-VPA reference, corresponding to ~11% higher brivaracetam exposure, Schoemaker 2017 Table 1), Damnjanovic_2023_lamotrigine.R (exponential effect on apparent oral clearance of lamotrigine, cl *= exp(-0.61 * CONMED_VPA) = a 45.7% reduction, encoding valproate’s inhibition of lamotrigine glucuronidation in a paediatric dual-therapy cohort), Wang_2019_lamotrigine.R (fractional multiplicative effect on apparent oral clearance of lamotrigine, cl *= (1 + (-0.386) * CONMED_VPA) = a 38.6% reduction, in 89 Chinese adults and children with epilepsy; the indicator is genuinely time-varying here because comedication status was recorded at each TDM sampling time point rather than once per subject).
• Notes: Drug-specific CONMED_* indicator anticipated in the [[CONMED_AED]] notes. Schoemaker 2017 retained the VPA effect in the final model even though it did not formally meet the SCM inclusion criteria (forward p < 0.01) because quantifying its contribution was considered informative; the authors note the apparent VPA-induced exposure rise may be confounded with VPA-driven weight / fat gain in chronic users. Distinct from the broader [[CONMED_AED]] (any concomitant AED). When a paper distinguishes individual AEDs separately, use the drug-specific canonicals [[CONMED_CBZ]], [[CONMED_PB]], CONMED_VPA rather than collapsing into the class-level indicator.

CONMED_VANCOMYCIN (canonical for concomitant intravenous vancomycin coadministration indicator)

• Description: 1 = the subject was receiving concomitant intravenous vancomycin at this observation; 0 = no concomitant vancomycin. Vancomycin is a glycopeptide antibiotic with well-described nephrotoxic potential, and the indicator exists to capture nephrotoxin-mediated impairment of the modelled drug’s renal clearance – not vancomycin’s own PK. Time-varying when vancomycin starts or stops within the observation window, which is the usual case for an inpatient antibiotic course.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant vancomycin).
• Source aliases:
  • VAN – used in Downes_2022_tobramycin.R (Downes 2022 Table 2 theta VAN; the paper’s own covariate name).
• Example models: Downes_2022_tobramycin.R (exponentiated-switch factor e_conmed_vancomycin_cl^CONMED_VANCOMYCIN with e_conmed_vancomycin_cl = 0.708 on tobramycin CL, i.e. a 29.2% lower clearance with concomitant vancomycin; Downes 2022 Table 2 and Results, dichotomized Y/N as a time-varying covariate).
• Notes: Member of the CONMED_<INN> binary concomitant-medication family. Read the effect as a nephrotoxin-interaction marker, not as a mechanistic drug-drug interaction, and treat it as fragile. Downes 2022 entered it from an exploratory forward-selection screen of five nephrotoxic comedications, and it was the only one of the five retained; the estimate rests on 5 of 58 patients over 8 of 111 courses, carries a 19.1 %RSE, and its bootstrap 95% CI (0.462-1.24) crosses 1. The paper’s rationale is that its 48-hour observation window is shorter than the lag between kidney injury and a measurable serum-creatinine rise, so a nephrotoxin indicator can capture clearance impairment that the concurrent CRCL covariate has not yet registered – which is why the two coexist in that model rather than being redundant. The authors deliberately set the indicator to 0 for every simulated patient in their target-attainment analysis so as not to over-state the effect, and any user reproducing that analysis should do the same. Distinct from the class-level nephrotoxin composites Downes 2022 also tested and rejected (none versus >= 1 agent; 0-1 versus >= 2 agents), and distinct from RRT_HEMODIAL_STATUS and the CRCL family, which measure renal function rather than exposure to a nephrotoxin. Note that vancomycin appears in this register in two unrelated roles: here as a coadministered agent affecting another drug, and as the modelled analyte in Goti_2018_vancomycin.R / Jung_2024_vancomycin.R / Shen_2024_vancomycin.R / Chung_2013_vancomycin.R; a model must never use this indicator for its own analyte.

CONMED_TICARCLAV, CONMED_TMPSMX, CONMED_PIPTAZO, CONMED_ACYCLOVIR (canonical for concomitant ticarcillin-clavulanate / trimethoprim-sulfamethoxazole / piperacillin-tazobactam / acyclovir coadministration indicators)

• Description: 1 = the subject was receiving the named agent concomitantly at this observation; 0 = not. Four members of the CONMED_<INN> family that share a single register entry because they share a single role: each is a nephrotoxic comedication screened as a potential brake on a renally-cleared drug’s clearance. Time-varying when the agent starts or stops within the observation window.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant exposure to the named agent).
• Source aliases:
  • ticarcillin/clavulanate, trimethoprim-sulfamethoxazole, piperacillin-tazobactam, acyclovir – used in Downes_2022_tobramycin.R, which names them in prose rather than as short data-column codes.
• Example models: Downes_2022_tobramycin.R (all four declared in covariatesDataExcluded, not covariateData – see Notes; founding example).
• Notes: These four are registered as screened-only names and carry no effect in any current model. Downes 2022 dichotomized each Y/N as a time-varying covariate and offered it to a forward selection on tobramycin clearance requiring a >= 6.63 drop in OFV; only CONMED_VANCOMYCIN met the threshold, and the paper also tested and rejected two class-level composites (none versus >= 1 nephrotoxin; 0-1 versus >= 2). CONMED_ACYCLOVIR is a step weaker still: it is counted in the paper’s Table 1 nephrotoxin tally but is absent from the Methods forward-selection list, so it was never screened individually. They are registered so the composition of a published nephrotoxin screen is auditable and so a future paper that does retain one of them inherits a name rather than inventing one; a model that retains an effect should move the entry into its covariateData and add its own worked example here. CONMED_NSAID – the fifth agent in the same Downes 2022 screen – is registered separately above and predates this entry. Prefer these INN-specific names over the class-level CONMED_ABX when a paper resolves individual agents.

CONMED_VERTEPORFIN (canonical for concomitant verteporfin photodynamic therapy (PDT) coadministration indicator)

• Description: 1 = subject received one or more concomitant verteporfin photodynamic-therapy (PDT) procedures during the ranibizumab treatment period; 0 = no concomitant PDT. Time-fixed per subject in Xu 2013 (All covariates adopted their values measured at baseline). Verteporfin (Visudyne) is a benzoporphyrin-derivative photosensitizer activated by non-thermal 689 nm laser to induce localized choroidal vessel occlusion in neovascular age-related macular degeneration; the PDT procedure is administered 7 days before ranibizumab in the FOCUS combination protocol.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant verteporfin PDT).
• Source aliases:
  • PDT – used in Xu_2013_ranibizumab.R (Xu 2013 Fig 1 lists “concomitant PDT therapy” as a screened covariate; “Specifically in this case, PDT is verteporfin for injection.”).
• Example models: Xu_2013_ranibizumab.R (multiplicative reduction of vitreous elimination rate Ka: ka *= (1 - e_conmed_verteporfin_ka * CONMED_VERTEPORFIN) with e_conmed_verteporfin_ka = 0.353, i.e. 35.3% lower Ka on subjects with concomitant PDT; Xu 2013 Table 3, theta6 = 0.353 stored as a positive fractional-reduction magnitude under the header ‘Covariate multiplier for Ka’).
• Notes: Member of the CONMED_<INN> binary concomitant-medication family. Verteporfin PDT is the specific PDT modality in the Xu 2013 AMD analysis, so the INN-specific canonical name is used rather than a generic CONMED_PDT. Xu 2013 attributes the slower vitreous elimination in PDT-treated subjects to vessel occlusion and soft tissue scarring reducing choroidal drainage of ranibizumab from the vitreous humor.

CONMED_SILDENAFIL (canonical for concomitant sildenafil coadministration indicator)

• Description: Binary indicator for concomitant sildenafil coadministration. 1 = sildenafil was administered during this experimental occasion (in the Bender 2009 rat study, a 2 mg/kg bolus followed by a 6 h steady-state infusion), 0 = saline (no sildenafil). Per-occasion (not per-subject) in a crossover design.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (saline / no concomitant sildenafil).
• Source aliases:
  • SLDB – used in Bender_2009_pregabalin_rat_binary.R (Bender 2009 binary sildenafil-presence indicator).
• Example models: Bender_2009_pregabalin_rat_binary.R (proportional reduction in pregabalin CL when sildenafil is coadministered: cl *= (1 - e_sild_cl * CONMED_SILDENAFIL) with e_sild_cl = 0.302, i.e. 30.2% lower CL on the sildenafil occasion, Bender 2009 Table IV).
• Notes: Specific scope because the sildenafil-on-pregabalin-CL drug-drug-interaction effect is paper-specific (Bender 2009 used sildenafil as a PDE-5-inhibitor probe). Member of the CONMED_<drug> binary concomitant-presence family. Per-occasion in the Bender 2009 crossover (each rat received saline on one occasion and sildenafil on the other, separated by a >= 3-day washout). Companion to CONMED_SILDENAFIL_NMETAB_CC, the continuous N-desmethyl-sildenafil-metabolite-concentration covariate used in the saturable-inhibition variant of the same model.

CONMED_NAL_DOSE, CONMED_BUP_DOSE (canonical for daily dose of co-administered naltrexone / bupropion)

• Description: Time-varying daily dose of the named drug at the current observation time (suffix = INN lowercase abbreviation: nal naltrexone, bup bupropion). 0 = the named drug is not part of the regimen at this time; positive value = current total daily dose. Drives the combined dose- and time-dependent Emax drug effect in the naltrexone/bupropion fixed-dose-combination PD model.
• Units: mg/day
• Type: continuous
• Scope: specific
• Reference category: 0 (the drug is not given; the drug-effect term contributes 0).
• Source aliases:
  • DOSE_NAL_MGD (NAL, Sharma 2018 Eq. 4 daily naltrexone dose in mg) – earlier DOSE_<drug>_<unit> form used in Sharma_2018_naltrexone_bupropion.R before the CONMED_<drug>_DOSE rename. Maps to CONMED_NAL_DOSE.
  • DOSE_BUP_MGD (BUP, Sharma 2018 Eq. 4 daily bupropion dose in mg) – earlier DOSE_<drug>_<unit> form used in Sharma_2018_naltrexone_bupropion.R before the rename. Maps to CONMED_BUP_DOSE.
• Example models: Sharma_2018_naltrexone_bupropion.R (each daily-dose covariate enters the combined dose- and time-dependent Emax drug effect on body-weight kout: drug_term = CONMED_NAL_DOSE / (ed50nal + CONMED_NAL_DOSE) + CONMED_BUP_DOSE / (ed50bup + CONMED_BUP_DOSE), Sharma 2018 Eq. 4).
• Notes: Specific scope because the dose-effect amplitudes are tied to the Sharma 2018 naltrexone/bupropion DTPD body-weight analysis. Members of the CONMED_<drug>_DOSE daily-dose family (siblings CONMED_ATV_DOSE, CONMED_INH_DOSE, etc.); the CONMED_<drug>_DOSE shape replaces the earlier DOSE_<drug> / DOSE_<drug>_<unit> names (which conflated a covariate with a dose-amount column) per the naming audit. Companion to DIS_DIAB (the diabetes covariate in the same model; pre-2026-06-19 canonical name T2DM).

PRICORT (canonical for prior corticosteroid use indicator)

• Description: 1 = patient received systemic corticosteroid treatment prior to study entry, 0 = no prior corticosteroid use. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no prior corticosteroid use).
• Source aliases: none.
• Example models: Ma_2020_sarilumab_das28crp.R (multiplicative on DAS28-CRP Kout: Kout * theta^PRICORT), Ma_2020_sarilumab_anc.R (power-form on Emax: Emax * 0.819^PRICORT).
• Notes: Ma 2020 applies it as a multiplicative effect of the form param * theta^PRICORT in both DAS28-CRP and ANC PD models. Generally applicable clinical-history indicator.

PRIOR_ANTHRACYCLINE_DOSE (canonical for prior cumulative anthracycline dose)

• Description: Cumulative dose of anthracycline chemotherapy received by the subject prior to the first dose analysed in the current popPK / popPK-PD model, expressed in doxorubicin-equivalent body-surface-area-normalised mg/m^2. Time-fixed per subject for the analysis window (the running cumulative anthracycline dose at the first observed dose).
• Units: mg/m^2 (doxorubicin-equivalent)
• Type: continuous
• Scope: specific
• Reference category: n/a – typically used as a linear shift ((1 + theta * (PRIOR_ANTHRACYCLINE_DOSE - ref))) on a baseline parameter (e.g., baseline cardiac troponin I before the next anthracycline cycle). Reference values observed: 90 mg/m^2 (Kunarajah 2017, cohort median).
• Source aliases:
  • PCAMT – Kunarajah 2017 NM-TRAN convention (“Prior Cumulative Anthracyclines aMounT”; doxorubicin-equivalent mg/m^2).
• Example models: Kunarajah_2017_doxorubicin.R (linear shift on baseline cardiac troponin I: bl_cTnI * (1 + 0.00308 * (PRIOR_ANTHRACYCLINE_DOSE - 90)) – ~0.31% increase in baseline cTnI per 1 mg/m^2 of prior cumulative anthracycline exposure).
• Notes: Distinct from PRIOR_ANTICANCER (a binary modality indicator, 1 = any prior anticancer therapy) – PRIOR_ANTHRACYCLINE_DOSE carries the actual cumulative dose, restricted to the anthracycline drug class (doxorubicin, daunorubicin, epirubicin, idarubicin), and is the column needed when the source paper’s effect is dose-response in the prior-exposure regime rather than presence / absence. When a paper records anthracycline exposure as anthracycline-class-by-class doses and the model effect aggregates them, sum to a single doxorubicin-equivalent value before populating this column (use the published bone-marrow / cardiotoxicity isoeffective conversion factors). When a paper distinguishes the type of anthracycline (e.g., doxorubicin vs daunorubicin separately), register parallel canonicals (PRIOR_DOXORUBICIN_DOSE, PRIOR_DAUNORUBICIN_DOSE) rather than overloading this name. Scope: specific because the column meaning is intrinsically tied to anthracycline-class chemotherapy exposure; promote to general if a second paper ratifies the same definition.

PRIOR_ANTICANCER (canonical for prior anticancer therapy of any modality)

• Description: 1 = subject received any prior anticancer therapy (cytotoxic chemotherapy, radiotherapy, hormonal therapy, targeted therapy, immunotherapy, or surgical debulking with adjuvant intent) before the start of the analyzed treatment, 0 = treatment-naive. Broader than LINE_1L, which is specifically a treatment-line indicator restricted to systemic drug therapy lines. PRIOR_ANTICANCER captures the full clinical concept of prior cancer treatment exposure as used in cytotoxic-chemotherapy myelosuppression analyses, where any prior anticancer modality may have depleted the bone-marrow proliferating pool and therefore affects baseline ANC.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (treatment-naive; no prior anticancer therapy of any modality).
• Source aliases:
  • PC (Kloft 2006 / Netterberg 2017 NM-TRAN convention for “previous anticancer therapy”; values 0 = naive, 1 = had prior anticancer therapy) – used in Netterberg_2017_docetaxel.R.
• Example models: Netterberg_2017_docetaxel.R (multiplicative effect on baseline ANC: BACOV *= (1 + theta * PRIOR_ANTICANCER) with theta = -0.147; prior-anticancer patients have ~14.7% lower baseline ANC than treatment-naive patients).
• Notes: Distinct from LINE_1L (which is the inverse semantics for systemic-drug therapy lines only) and PRIOR_TNF / PRIOR_BIO (which are modality-specific to anti-TNF / biologic exposure in inflammatory-disease cohorts). Use PRIOR_ANTICANCER when the source paper’s covariate counts any anticancer modality (including radiotherapy and surgery) as prior exposure. When a future paper restricts the indicator to cytotoxic chemotherapy alone, use LINE_1L (with values inverted: paper’s PRIOR_CHEMO = 1 - LINE_1L). When a paper distinguishes prior chemotherapy from prior radiotherapy, register a parallel PRIOR_RADIATION canonical.

PRIOR_SYSTEMIC (canonical for prior systemic anticancer therapy indicator)

• Description: 1 = prior systemic anticancer therapy (cytotoxic chemotherapy, targeted therapy, immunotherapy or hormonal therapy) was received before the start of the analyzed treatment; 0 = systemically treatment-naive. Locoregional procedures, radiotherapy and surgery do NOT set this flag. Narrower than PRIOR_ANTICANCER, which counts any anticancer modality.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no prior systemic anticancer therapy). At the study-arm level the reference pools first-line arms with arms of mixed treatment-line composition – see Notes.
• Source aliases:
  • PTx – Zierhut 2016 Results “Final model” equation and Table 3.
  • All second line patients – Zierhut 2016 Table 1 row label for the arms that take the value 1.
• Example models: Zierhut_2016_hcc_antiangiogenic_os_mbma.R (additive log(0.933) shift on ln(median overall survival): a wholly pretreated advanced-hepatocellular-carcinoma arm has 93.3% of the median overall survival of an otherwise identical arm, a forced-in and non-significant effect whose 95% CI of 74.0-118% spans both benefit and harm).
• Notes: Registered because “prior systemic therapy” is the clinically load-bearing contrast in tumour types where locoregional and surgical pretreatment are near-universal – in advanced hepatocellular carcinoma almost every patient has had transarterial chemoembolization, so PRIOR_ANTICANCER (any modality, explicitly including radiotherapy and debulking surgery) would be close to constant and would not carry the intended meaning. Member of the PRIOR_<class> family (PRIOR_ANTICANCER, PRIOR_RADIATION, PRIOR_PLATIN, PRIOR_TAXANE, PRIOR_VINCA, PRIOR_ALKI, PRIOR_IPI, PRIOR_TNF, PRIOR_BIO, PRIOR_STATIN, PRIOR_EZE, PRIOR_OCTREOTIDE), sitting between the any-modality PRIOR_ANTICANCER and the agent-specific members. At the study-arm level in an MBMA the flag is all-or-nothing: 1 means the ENTIRE arm was pretreated, so arms of mixed treatment-line composition take 0 and the column is therefore NOT 1 - LINE_1L (which would also capture the mixed arms). Read alongside LINE_1L / LINE_3L / LINE_4L_PLUS, which resolve the treatment line as an ordinal rather than collapsing it to pretreated-vs-naive; use those when a source estimates per-line effects. Scope: specific until a second source adopts the systemic-only restriction.

PRIOR_ALKI (canonical for prior ALK-inhibitor therapy indicator)

• Description: 1 = the subject received an anaplastic lymphoma kinase (ALK) tyrosine-kinase-inhibitor therapy (crizotinib, ceritinib, alectinib, brigatinib, lorlatinib, ensartinib, or an equivalent) before enrolment in the current study; 0 = ALK-inhibitor-naive. Time-fixed per subject. Distinguishes previously-treated from treatment-naive ALK-positive non-small-cell-lung-cancer patients, a split that drives both response rate and, in some analyses, apparent disposition.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (ALK-inhibitor-naive). Set to 0 for subjects who are not ALK-positive at all (ROS1-positive, unknown status, or healthy participants); those groups carry their own indicators, so the ALK-pretreatment contrast stays within the ALK-positive stratum.
• Source aliases:
  • ALKPOT = 2, 3, 4 – Liu 2024 Table 2 and Supplementary Table 1. The source variable is a six-level “prior anti-cancer therapy in ALK+ patients” code whose levels 2 (intolerant to crizotinib as the sole prior treatment), 3 (one prior second- or third-generation ALK inhibitor) and 4 (at least two prior second- or third-generation ALK inhibitors) the paper itself collapses into one estimated effect; map those three levels to 1, level 1 to 0, and carry levels 5 and 6 on TUM_ALK_MUT and DIS_HEALTHY respectively.
  • ALK-inhibitor-naive / crizotinib-pretreated – narrative subgroup labels for the same split.
• Example models: Liu_2024_saf189s.R (multiplicative power-form effect on apparent central volume: 0.734^PRIOR_ALKI, i.e. V/F 26.6% lower in ALK-inhibitor-pretreated ALK+ patients than in ALK-inhibitor-naive ALK+ patients; reference category is the ALKPOT = 1 group, n = 115 of 317).
• Notes: Member of the PRIOR_<class> prior-therapy family (PRIOR_TNF, PRIOR_IPI, PRIOR_VINCA, PRIOR_PLATIN, PRIOR_TAXANE, PRIOR_STATIN, PRIOR_BIO, PRIOR_OCTREOTIDE). Narrower than PRIOR_ANTICANCER, which is prior anti-cancer therapy of ANY modality including chemotherapy and radiation; a patient may be PRIOR_ANTICANCER = 1 and PRIOR_ALKI = 0. Where a source paper resolves the pretreatment history further (one prior ALK inhibitor vs two or more, or which generation), record the finer levels in covariateData[[PRIOR_ALKI]]$notes rather than inventing sub-columns, unless the paper estimates a separate coefficient per level – in which case register the levels as their own indicators. The ROS1-inhibitor counterpart is a distinct concept and should be registered separately if a source paper estimates it.

PRIOR_RADIATION (canonical for prior radiotherapy exposure indicator)

• Description: 1 = subject received any prior radiotherapy (external-beam, brachytherapy, or radionuclide-targeted) before the start of the analyzed treatment; 0 = radiotherapy-naive. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (radiotherapy-naive).
• Source aliases:
  • “prior radiotherapy (yes/no)” (Lu 2017 narrative; coded as a binary 0/1 in the source NONMEM dataset) – used in Lu_2017_polatuzumab_neuropathy.R.
• Example models: Lu_2017_polatuzumab_neuropathy.R (multiplicative effect on the grade >= 2 peripheral neuropathy hazard via exp(theta_priorRadioTx * PRIOR_RADIATION) with theta_priorRadioTx = -7.94e-3, SE 0.319 – effectively no detectable effect given the large SE).
• Notes: Sibling of PRIOR_ANTICANCER (the broader any-modality indicator) and PRIOR_VINCA / PRIOR_PLATIN (chemotherapy-class-specific siblings). Use this canonical when the source paper distinguishes radiotherapy as a separate covariate from prior chemotherapy / hormonal / immunotherapy exposure. Pre-authorized as a parallel-sibling canonical in the PRIOR_ANTICANCER notes.

PRIOR_VINCA (canonical for prior vinca-alkaloid chemotherapy indicator)

• Description: 1 = subject received any prior vinca-alkaloid chemotherapy (vincristine, vinblastine, vinorelbine, vindesine) before the start of the analyzed treatment; 0 = vinca-naive. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (vinca-naive).
• Source aliases:
  • “prior vinca alkaloids treatment (yes/no)” (Lu 2017 narrative; coded as a binary 0/1 in the source NONMEM dataset) – used in Lu_2017_polatuzumab_neuropathy.R.
• Example models: Lu_2017_polatuzumab_neuropathy.R (multiplicative effect on the grade >= 2 peripheral neuropathy hazard via exp(theta_priorVinca * PRIOR_VINCA) with theta_priorVinca = -0.102, SE 0.469 – not detectable given the large SE; the paper hypothesized prior vinca-alkaloid exposure could sensitize patients to antimicrotubule-induced PN but did not find a clinically meaningful effect).
• Notes: Vinca alkaloids are microtubule-destabilising agents (the natural-product parent class of the vc-MMAE / vc-MMAF ADC payloads); prior exposure is a clinically meaningful pretreatment indicator in any analysis where antimicrotubule-toxicity history affects subsequent dosing decisions or AE risk. Sibling of PRIOR_ANTICANCER (broader any-modality indicator), PRIOR_RADIATION, PRIOR_PLATIN, and PRIOR_TAXANE (parallel chemotherapy-class siblings).

PRIOR_PLATIN (canonical for prior platinum-based chemotherapy indicator)

• Description: 1 = subject received any prior platinum-based chemotherapy (cisplatin, carboplatin, oxaliplatin, nedaplatin, satraplatin) before the start of the analyzed treatment; 0 = platinum-naive. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (platinum-naive).
• Source aliases:
  • “prior platinum-based treatment (yes/no)” (Lu 2017 narrative; coded as a binary 0/1 in the source NONMEM dataset) – used in Lu_2017_polatuzumab_neuropathy.R.
• Example models: Lu_2017_polatuzumab_neuropathy.R (multiplicative effect on the grade >= 2 peripheral neuropathy hazard via exp(theta_Platin * PRIOR_PLATIN) with theta_Platin = 0.159, SE 0.345 – not detectable given the large SE; the paper hypothesized prior platinum exposure could sensitize patients to subsequent antimicrotubule-induced PN given that platinum agents themselves cause sensory PN, but did not find a clinically meaningful effect).
• Notes: Platinum agents are DNA-crosslinking cytotoxics with peripheral-neuropathy as a well-known adverse-event profile (especially for cisplatin and oxaliplatin); prior exposure is therefore a defensible pretreatment indicator in any subsequent-PN-risk analysis. Sibling of PRIOR_ANTICANCER (broader any-modality indicator), PRIOR_RADIATION, PRIOR_VINCA, and PRIOR_TAXANE (parallel chemotherapy-class siblings). Distinct from AUC_CARBO (the continuous per-cycle carboplatin-AUC drug-exposure covariate, which is time-varying and used as a direct drug-exposure driver in tumor-dynamics models, e.g. Zecchin_2016_survival.R).

PRIOR_BIO (canonical for prior biologic exposure)

• Description: 1 = subject previously treated with any biologic (broader than PRIOR_TNF: includes anti-TNF agents plus anti-integrin, anti-IL-12/23, anti-IL-17, anti-IL-23, anti-IL-6, etc.), 0 = biologic-naive.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (biologic-naive).
• Source aliases:
  • bio-naive (Aguiar 2021 source-paper variable, with the indicator inverted: paper’s bio-naive = 1 - PRIOR_BIO). Effect coefficients in the source paper apply to bio-naive; in model derive bio_naive <- 1 - PRIOR_BIO to preserve the paper’s reported coefficient.
• Example models: Aguiar_2021_ustekinumab.R (Aguiar 2021 Table 2 footnote a; multiplicative effect on CL: factor (1 - 0.227 * (1 - PRIOR_BIO)), so bio-naive patients have ~23% lower CL than previously-exposed patients).
• Notes: Distinct from PRIOR_TNF (a strict subset). Use PRIOR_BIO when the source paper’s covariate counts any biologic as prior exposure (anti-TNF, anti-integrin, anti-IL-12/23, anti-IL-17, anti-IL-23, anti-IL-6, etc.); use PRIOR_TNF when the source paper specifically tested anti-TNF exposure. When the source paper uses the inverted “bio-naive” indicator (1 = naive), document the inversion in covariateData[[PRIOR_BIO]]$notes and apply 1 - PRIOR_BIO in model so the canonical column stores 1 = previously exposed.

PRIOR_TNF (canonical for prior anti-TNF biologic exposure indicator)

• Description: 1 = subject previously treated with an anti-TNF (tumor necrosis factor) inhibitor, 0 = TNF-naive.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (TNF-naive).
• Source aliases:
  • PRIORTNF (all caps, no underscore) – acceptable alternative spelling.
• Example models: Moein_2022_etrolizumab.R (multiplicative fractional effect on CL, +4.9%).
• Notes: Use when the source paper reports a binary “prior anti-TNF inhibitor” covariate on any PK parameter. Generally applicable across RA/PsA/IBD/axSpA biologic PK models.

PRIOR_IMMUNOMODULATORY_DRUG (canonical for prior immunomodulatory-drug (IMiD) therapy indicator)

• Description: 1 = subject previously treated with an immunomodulatory imide drug (IMiD: thalidomide, lenalidomide or pomalidomide) before the current regimen, 0 = IMiD-naive. A baseline treatment-history indicator in multiple-myeloma cohorts, where prior IMiD exposure marks a more heavily pretreated and less IMiD-responsive disease.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (IMiD-naive). Where a source reports its effect on the naive group (as Srimani 2022 does), the model applies it as (1 - PRIOR_IMMUNOMODULATORY_DRUG) and the typical value is the IMiD-exposed one.
• Source aliases:
  • PIMID – Srimani 2022 Table 2 / Table 3 covariate code (“Yss (PIMID)”, “B3x (PIMID)”).
  • Prior immunomodulatory drug therapy: Naive / Exposed – Srimani 2022 Table 1 row label.
  • PRIOR_IMID – common abbreviated spelling; same orientation, no value transformation.
• Example models: Srimani_2022_ixazomib_mprotein.R (linear multiplicative effect on the relative steady-state M-protein nadir, Yss * (1 - 0.427 * (1 - PRIOR_IMMUNOMODULATORY_DRUG)); 55.9% of the efficacy dataset IMiD-exposed), Srimani_2022_ixazomib_diarrhea.R (additive shift of -2.87 on the cumulative logits out of grade 3 diarrhea for IMiD-naive patients in a proportional-odds Markov model).
• Notes: Member of the PRIOR_<class> family (PRIOR_TAXANE, PRIOR_TNF, PRIOR_VINCA, PRIOR_PLATIN, PRIOR_ALKI), spelled out rather than abbreviated to PRIOR_IMID. Distinct from CONMED_IMMUNOMOD, which records CONCURRENT purine-analogue / methotrexate immunomodulator co-medication in inflammatory-disease biologic PK models, not prior IMiD therapy. A regimen that includes an IMiD on study (e.g. lenalidomide-dexamethasone background therapy) does not set this flag; it records exposure before study entry only.

PRIOR_IPI (canonical for prior ipilimumab treatment indicator)

• Description: 1 = subject previously treated with ipilimumab (anti-CTLA-4 monoclonal antibody) before the start of the current anti-PD-1 / anti-PD-L1 (or other) checkpoint-inhibitor regimen, 0 = ipilimumab-naive.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (ipilimumab-naive). Document per-paper how subjects with missing IPI status are pooled (Ahamadi 2017 pools “missing” with naive in the final coefficient; some papers may report a separate “missing” effect).
• Source aliases:
  • IPI (Ahamadi 2017; categorical with levels IPI-naive, IPI-treated, missing) – decompose into PRIOR_IPI = as.integer(IPI == "IPI-treated") and treat the missing category like naive unless the source paper retains a separate “missing” coefficient.
• Example models: Ahamadi_2017_pembrolizumab.R (proportional changes on CL of +14.0% and on Vc of +7.36% for IPI-treated relative to IPI-naive; “missing” 26.4% of cohort is pooled with naive in the canonical encoding because Table 3 reports only the naive-vs-treated coefficient).
• Notes: Distinct from PRIOR_ANTICANCER (any modality), PRIOR_BIO (any biologic), PRIOR_TNF (anti-TNF biologic). Use PRIOR_IPI when the source paper specifically tested prior ipilimumab exposure as a covariate; this is a common covariate in advanced-melanoma popPK analyses where ipilimumab was the standard-of-care immune-checkpoint inhibitor preceding PD-1 / PD-L1 entrants.

PRIOR_CART (canonical for prior chimeric antigen receptor (CAR) T cell therapy indicator)

• Description: 1 = subject received CAR T cell therapy (any autologous or allogeneic CAR T product) before the start of the current treatment, 0 = CAR-T-naive. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no prior CAR T cell therapy).
• Source aliases:
  • prior CAR T cell therapy (yes/no) – Li 2026 Table 1 and Table 2 (‘Effect of prior CAR T cell therapy = yes’); no data-column name is printed.
• Example models: Li_2026_epcoritamab.R (multiplies the maximum stimulated Grade >= 2 cytokine-release-syndrome hazard SMAX by exp(-0.964) = 0.382 in the epcoritamab repeated time-to-event model; 125 of 600 patients, 20.8%).
• Notes: Member of the PRIOR_<drug or class> prior-therapy family (PRIOR_IPI, PRIOR_TNF, PRIOR_BIO, PRIOR_SYSTEMIC, …). Common in models of T-cell-engaging bispecific antibodies and later-line lymphoma / myeloma therapies, where prior CAR T cell therapy marks a heavily pretreated population with altered T-cell fitness. Note the direction in the founding model: prior CAR T cell therapy LOWERED the CRS risk on epcoritamab, which the source describes as consistent with clinical observation, so do not assume a sign. Distinct from PRIOR_ANTICANCER (any modality) and PRIOR_SYSTEMIC (any systemic therapy); a patient with PRIOR_CART = 1 is also 1 on both of those.

PRIOR_STATIN (canonical for prior (pre-study) statin therapy indicator)

• Description: 1 = patient was on established statin (HMG-CoA reductase inhibitor) therapy before entering the study, 0 = statin-naive at study entry. Time-fixed per subject. Captures the pharmacological history that shapes the observed baseline lipid panel and the residual room for further LDL-C lowering, independent of whether a statin continues to be taken during the study.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (statin-naive at study entry).
• Source aliases:
  • Statin prior therapy – Jadhav 2023 Table 3 covariate row name.
  • PRIORSTAT, STATINPR – common dataset abbreviations.
• Example models: Jadhav_2023_bempedoicAcid_ldlc.R (proportional shifts on the indirect-response Imax of -0.373 and on baseline LDL-C of -0.296; Jadhav 2023 Table 3).
• Notes: Deliberately distinct from the concomitant-use family (CONMED_STATIN, CONMED_STATIN_MONO, CONMED_STATIN_LI / _MI / _HI): Jadhav 2023 screened “prior established LMTs” and “concomitant medication (low-, moderate-, or high-intensity statin or ezetimibe)” as separate covariate sets and retained both in the final model, so the two must be carried as separate columns. Statin-intolerance cohorts commonly have PRIOR_STATIN = 1 with all concomitant-statin indicators 0. Sibling of PRIOR_EZE.

PRIOR_EZE (canonical for prior (pre-study) ezetimibe therapy indicator)

• Description: 1 = patient was on established ezetimibe therapy before entering the study, 0 = ezetimibe-naive at study entry. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (ezetimibe-naive at study entry).
• Source aliases:
  • Ezetimibe prior therapy – Jadhav 2023 Table 3 covariate row name.
  • PRIOREZE – common dataset abbreviation.
• Example models: Jadhav_2023_bempedoicAcid_ldlc.R (proportional shift on baseline LDL-C of -0.0596; Jadhav 2023 Table 3).
• Notes: Distinct from CONMED_EZE (concomitant ezetimibe during the study). Jadhav 2023 Online Resource 5 reports a “Prior Treatment (Ezetimibe : No Ezetimibe)” contrast alongside a separate “Concomitant Treatment (Ezetimibe : No Ezetimibe)” contrast, confirming that the analysis dataset carried both columns and that the two effects act on different model parameters (prior use on baseline LDL-C; concomitant use on Imax and on bempedoic acid CL/F). Sibling of PRIOR_STATIN.

CONMED_NNRTI_IND (canonical for concomitant enzyme-inducing NNRTI indicator)

• Description: 1 = subject is coadministered an enzyme-inducing non-nucleoside reverse transcriptase inhibitor (efavirenz or nevirapine) at the observation, 0 = no concomitant enzyme-inducing NNRTI. Both efavirenz and nevirapine are CYP3A4 inducers; this pooled indicator collapses their effects on metabolite elimination of co-administered antiretrovirals into a single binary covariate.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no enzyme-inducing NNRTI; subject on NRTI-only or PI-based backbone).
• Source aliases:
  • NNI – used in Hirt_2006_nelfinavir.R (paper’s pooled “nonnucleosidic inhibitor” indicator: 1 if EFV or NVP, 0 otherwise; pooling justified by the paper’s finding that the two drugs’ inducer effects on M8 elimination were not significantly different and the two were never administered simultaneously in the cohort).
• Example models: Hirt_2006_nelfinavir.R (additive-form multiplicative effect on M8 apparent elimination rate kel_m8: kel_m8 *= (1 + 0.91 * CONMED_NNRTI_IND); KM0 is ~1.9-fold higher under EFV or NVP co-administration, consistent with CYP3A4 induction of M8 elimination).
• Notes: Pooling the two named NNRTIs follows the precedent of CONMED_EIAED (any enzyme-inducing AED) and CONMED_AZOLE (any azole antifungal). Distinct from CONMED_EFV (efavirenz alone, founded by Tikiso 2021): the pooled CONMED_NNRTI_IND indicator is used when the source paper did not separately identify EFV vs NVP effects. A future paper that needs to distinguish them should use CONMED_EFV (and a future sibling CONMED_NVP) rather than this pooled indicator. The “inducer” qualifier excludes non-inducing NNRTIs (rilpivirine, doravirine, etravirine); record the per-paper list of pooled drugs in covariateData[[CONMED_NNRTI_IND]]$notes. Specific scope until a second model ratifies the convention; at that point promote to general.

CONMED_PSEUDOEPHEDRINE (canonical for concomitant pseudoephedrine coadministration indicator)

• Description: 1 = subject is taking concomitant pseudoephedrine at the observation, 0 = not. Pseudoephedrine is a sympathomimetic decongestant that is renally excreted via active tubular secretion in addition to glomerular filtration, and competes with other tubularly-secreted drugs for proximal-tubule transporters. Time-varying when pseudoephedrine starts / stops within the observation window; in Van Wart 2004 the indicator is set per PK sample (PSEU_jk, sample-level).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant pseudoephedrine).
• Source aliases:
  • PSEU – used in VanWart_2004_garenoxacin.R (paper covariate PSEU_jk, sample-level indicator that the kth PK sample from the jth patient was collected during concomitant pseudoephedrine administration).
• Example models: VanWart_2004_garenoxacin.R (multiplicative effect on the structural CL/F covariate term: cl_structural *= (1 - 0.144 * CONMED_PSEUDOEPHEDRINE); -14.4% relative to the no-pseudoephedrine reference, Van Wart 2004 Table 4).
• Notes: Drug-specific CONMED_* indicator capturing competition for proximal-tubule active secretion; distinct from CYP-based and UGT-based induction or inhibition mechanisms encoded elsewhere. Renal clearance of pseudoephedrine itself is approximately 511-532 mL/min in a 70 kg adult (Van Wart 2004 Discussion), well above the glomerular filtration estimate (~120 mL/min), confirming active tubular secretion as the dominant elimination route. Other drugs that share this active-tubular-secretion route (levofloxacin, ciprofloxacin, gatifloxacin, gemifloxacin, garenoxacin) are candidates for the same indicator in future models. Distinct from CONMED_PROBENECID (anionic OAT inhibitor, mechanistically a perpetrator on anion-secreted substrates) – pseudoephedrine is the cationic / mixed-substrate competitor rather than a transporter inhibitor.

CONMED_ATAZANAVIR (canonical for concomitant atazanavir coadministration indicator)

• Description: 1 = subject is receiving atazanavir (ATV; an HIV-1 protease inhibitor) as a co-medication during the PK observation period, 0 = no concomitant atazanavir. Atazanavir is a UGT1A1 inhibitor (and weak CYP3A4 inhibitor) so the indicator is used to flag UGT1A1-mediated inhibition of co-administered antiretrovirals whose elimination depends on glucuronidation. Time-fixed when the source paper analyses cohorts whose ART regimen is stable across the analysis window; time-varying when atazanavir is added or stopped during the observation period.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no concomitant atazanavir).
• Source aliases:
  • ATV – used in ArabAlameddine_2012_raltegravir.R (paper Table 2 covariate symbol for the atazanavir coadministration indicator; same orientation as the canonical, 1 = on atazanavir).
• Example models: ArabAlameddine_2012_raltegravir.R (multiplicative additive shift on relative bioavailability in HIV-positive individuals: F_HIV+ *= (1 + e_atazanavir_fdepot * CONMED_ATAZANAVIR) with e_atazanavir_fdepot = 0.39; atazanavir-coadministered patients have ~39% higher relative bioavailability than non-atazanavir reference, consistent with ATV-mediated UGT1A1 inhibition of raltegravir glucuronidation), vonHentig_2009_saquinavir.R (power-of-binary multiplier on saquinavir CL/F: cl *= e_atazanavir_cl^CONMED_ATAZANAVIR with e_atazanavir_cl = 0.703, i.e. saquinavir CL/F is reduced to 70.3% when atazanavir 300 mg QD is coadministered, attributed to ATV-mediated CYP3A inhibition; von Hentig 2009 Table 2 full-model column), Bukkems_2021_raltegravir.R (linear additive effect on apparent oral clearance: cl *= (1 + e_atazanavir_cl * CONMED_ATAZANAVIR) with e_atazanavir_cl = -0.17; atazanavir-coadministered raltegravir CL is 17% lower than the no-atazanavir reference, consistent with ATV-mediated UGT1A1 inhibition of raltegravir glucuronidation, Bukkems 2021 Table 2 ‘Factor change in CL with atazanavir’), Tsirizani_2025_ritonavir.R (TWO effects carried by the same indicator on the RITONAVIR side of the interaction: (1 + e_atazanavir_fdepot * CONMED_ATAZANAVIR) on relative bioavailability with e_atazanavir_fdepot = 1.37, and (1 + e_atazanavir_cl * CONMED_ATAZANAVIR) on CL/F with e_atazanavir_cl = 0.207. This is the first entry in which ritonavir is the victim rather than the perpetrator: the boosting-dose ritonavir PK of 170 African children in the CHAPAS-4 trial has 137% higher relative bioavailability and 20.7% faster clearance in the atazanavir arm than in the darunavir reference arm. The reference is the darunavir arm, not a no-PI arm – the trial had no ritonavir-alone control, so the comparator here is a NAMED third protease inhibitor rather than the usual pooled no-atazanavir group; see CONMED_LOPINAVIR for the paired indicator that completes the three-level companion-PI covariate).
• Notes: Specific scope because the indicator is paper-defined as the presence of atazanavir within an HIV ART regimen; the comparator (no-atazanavir) pools all other backbones (PI without ATV, NNRTI-based, integrase-only, etc.) in every model except Tsirizani_2025_ritonavir.R, where the comparator is the single darunavir arm of a randomised trial. The full INN spelling ATAZANAVIR is used rather than the 3-letter abbreviation ATV because ATV collides with the legacy atorvastatin slot in the CONMED_<drug>_DOSE family (now spelled out as CONMED_ATORVASTATIN_DOSE); both surfaces use the full INN to keep the CONMED_ namespace unambiguous. Distinct from AUC_RTV (a continuous ritonavir AUC0-24 covariate used in Dickinson_2009_atazanavir.R to scale atazanavir CL/F via the booster’s UGT1A1 inhibition) and from CONMED_EFV (efavirenz, a CYP3A and UGT inducer that drives the opposite direction of effect on raltegravir bioavailability). When reading an effect size off this indicator, always check which drug is the analyte: the sign and the magnitude are properties of the victim-perpetrator pair, not of atazanavir alone.

CONMED_LOPINAVIR (canonical for concomitant lopinavir coadministration indicator)

• Description: 1 = subject is receiving lopinavir (LPV; an HIV-1 protease inhibitor, given only as the ritonavir-boosted co-formulation lopinavir/ritonavir) as a co-medication during the PK observation period, 0 = no concomitant lopinavir. Lopinavir is a CYP3A4 substrate and inhibitor; the indicator flags CYP3A4-mediated interaction with a co-administered antiretroviral, and in boosted-PI cohorts it doubles as the marker of a twice-daily rather than once-daily dosing schedule. Time-fixed when the source paper analyses cohorts whose ART regimen is stable across the analysis window; time-varying when lopinavir is added or stopped during the observation period.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no concomitant lopinavir). In Tsirizani_2025_ritonavir.R the reference is specifically the darunavir/ritonavir arm rather than a pooled no-lopinavir group.
• Source aliases:
  • PI_BCK_BONE == 3 – used in Tsirizani_2025_ritonavir.R, where the supplemental NONMEM control stream (Data S1) encodes the three-level companion-protease-inhibitor covariate as a single integer column and gates the lopinavir effects on IF(PI_BCK_BONE.EQ.3). Same orientation as the canonical once decomposed; a data assembler derives CONMED_LOPINAVIR = as.integer(PI_BCK_BONE == 3).
• Example models: Tikiso_2022_ethambutol.R (multiplicative effect on ethambutol relative bioavailability in African children with TB: (1 + e_lopinavir_fdepot * CONMED_LOPINAVIR) with e_lopinavir_fdepot = -0.320, i.e. 32% lower F on lopinavir/ritonavir-based ART, 88% of it super-boosted 4:4; source column LPVRTV > 0. The paper could not separate a P-gp-mediated interaction from HIV-related malabsorption or formulation differences), Tsirizani_2025_ritonavir.R (linear additive effect on relative bioavailability of boosting-dose ritonavir: (1 + e_lopinavir_fdepot * CONMED_LOPINAVIR) with e_lopinavir_fdepot = -0.234, i.e. ritonavir relative bioavailability is 23.4% lower in the lopinavir arm than in the darunavir reference arm. The indicator ALSO gates a second effect multiplicatively with the occasion indicator: e_evening_tlag^(CONMED_LOPINAVIR * oc1) with e_evening_tlag = 3.61, because the lopinavir arm was the only twice-daily arm and so the only one in which an evening-dose absorption-lag effect is identifiable).
• Notes: Specific scope pending a second model that ratifies the same semantics. The full INN spelling LOPINAVIR is used rather than the 3-letter abbreviation LPV, matching the CONMED_ATAZANAVIR precedent, to keep the CONMED_ namespace unambiguous. Paired with CONMED_ATAZANAVIR in Tsirizani_2025_ritonavir.R to encode a three-level randomised companion-PI covariate; the third level (darunavir) is the reference and is represented by BOTH indicators being 0, so no CONMED_DARUNAVIR column is registered. A future paper that needs darunavir as a non-reference level should register CONMED_DARUNAVIR as a sibling rather than re-orient either existing indicator. Because lopinavir is not marketed unboosted, a CONMED_LOPINAVIR == 1 record always implies concomitant ritonavir; a model whose analyte is a CYP3A4 substrate other than ritonavir cannot separate the two from this column alone.

CONMED_INOTROPE (canonical for concomitant inotrope / vasoactive coadministration indicator)

• Description: 1 = subject is receiving at least one inotropic / vasoactive agent (dopamine, dobutamine, epinephrine, norepinephrine, milrinone, vasopressin, or equivalent) at the time of the observation; 0 = not on any such agent. Time-varying in principle; typically captured per-occasion for popPK observations in critical-care / neonatal-ICU studies. Used as a hemodynamic / renal-perfusion surrogate when the pharmacology mechanism is reduced GFR due to hemodynamic instability rather than a direct CYP / transporter DDI.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (not on any inotrope). Effect form in Zhao 2014 is a direct multiplicative selection on CL: F_inotrope = theta8 = 0.708 when CONMED_INOTROPE == 1 and 1 otherwise; encoded in nlmixr2 as f_inotrope <- e_inotrope_cl ^ CONMED_INOTROPE with e_inotrope_cl = 0.708.
• Source aliases:
  • inotropic agents / inotropes – used in Zhao_2014_ciprofloxacin.R (Zhao 2014 Table 2 ‘Comedication’ row and Table 4 F_inotrope notation); same orientation as the canonical, no value transformation.
• Example models: Zhao_2014_ciprofloxacin.R (multiplicative selection on CL: 0.708 ^ CONMED_INOTROPE – 29.2% CL reduction in patients on inotropes), Trang_2019_plazomicin.R (plazomicin popPK; additive proportional selection on the SECOND peripheral volume rather than on clearance, vp2 * (1 + 3.90 * CONMED_INOTROPE), i.e. vasopressor use nearly quintuples Vp2. A second ratifying model, and a precedent that the effect need not be on CL: the paper’s own Vp1:vasopressor relationship was dropped after the bootstrap while the Vp2 one survived. Vasopressor was defined as adrenaline, dobutamine, etilefrine, isoproterenol, noradrenaline or norepinephrine at any time during the study).
• Notes: Specific scope pending corroboration by a second model that ratifies the same semantics. The per-paper drug list pooled into the indicator should be documented in covariateData[[CONMED_INOTROPE]]$notes for each model (Zhao 2014 does not enumerate the exact agents; the cohort is critical-care neonates so the practical mix is dominated by dopamine / dobutamine / epinephrine / norepinephrine / milrinone). Distinct from extracorporeal-support modalities (ECMO_PUMP_SPEED, etc.) – inotropes are a pharmacological intervention rather than a mechanical-support modality. Distinct from heart rate (HR) – inotropes may cause both reduced GFR and altered HR but the indicator captures the therapy itself rather than its hemodynamic consequence. If a future paper retains a finer-grained taxonomy (epinephrine-only vs dopamine-only vs vasopressor-only), register CONMED_EPI / CONMED_DOBU / CONMED_VASOPRESS as siblings rather than overload CONMED_INOTROPE with conflicting semantics.

CONMED_STANOZOLOL (canonical for concomitant stanozolol coadministration indicator)

• Description: 1 = subject is coadministered the anabolic-androgenic steroid stanozolol at the observation, 0 = not. Stanozolol is a synthetic 17-alpha-alkylated androgen used in aplastic anemia to stimulate erythropoiesis; it inhibits CYP3A4-mediated metabolism of ciclosporin (and several other CYP3A4 substrates). In Ni 2013 stanozolol coadministration was time-fixed per subject (78 of 102 children remained on stanozolol throughout the TDM observation window).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no stanozolol coadministration).
• Source aliases:
  • stanozolol indicator (no specific NONMEM column name reported in Ni 2013) – used in Ni_2013_ciclosporin.R.
• Example models: Ni_2013_ciclosporin.R (multiplicative power-of-binary effect on CL/F: cl *= e_conmed_stanozolol_cl ^ CONMED_STANOZOLOL with e_conmed_stanozolol_cl = 0.83, i.e. ~17% lower CL/F when stanozolol is coadministered relative to the no-stanozolol reference; Ni 2013 Table 2 ‘F comedication’).
• Notes: Anabolic-androgenic 17-alpha-alkylated steroid; mechanistically distinct from glucocorticoid corticosteroids (do not reuse [[CONMED_STEROID]]). Ni 2013 tested both prednisone and stanozolol in forward covariate selection and retained only stanozolol after backward elimination – the clinically relevant mechanism is CYP3A4 inhibition (Ni 2013 Discussion paragraph 5). A future popPK paper that tests stanozolol or related 17-alpha-alkylated anabolic androgens (oxandrolone, oxymetholone, danazol) as covariates would either ratify this canonical (promoting to general scope) or warrant a family-level canonical (e.g. CONMED_ANDROGEN) – defer that decision until a second paper appears.

CONMED_BUDESONIDE (canonical for inhaled-budesonide treatment-arm indicator)

• Description: 1 = subject is in the inhaled-budesonide treatment arm; 0 = subject is in the placebo arm or in an alternative-controller arm (e.g., nedocromil). Per-subject time-fixed binary indicator carrying the head-to-head randomized-arm assignment in an inhaled-corticosteroid trial where pulmonary function (FEV1) is modelled as a longitudinal disease-progression endpoint and the inhaled-corticosteroid effect modifies the typical-value FEV1 prediction. Distinct from a PK-driven exposure effect: the model treats the indicator as an additive on/off term and does NOT carry a budesonide PK compartment.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (placebo or alternative-controller arm; the complement composition is defined per-model in covariateData[[CONMED_BUDESONIDE]]$notes).
• Source aliases:
  • derived per subject from the CAMP-trial assigned arm – used in Wu_2014_FEV1_asthma.R (the CAMP-trial randomized-arm assignment with three levels {placebo, nedocromil, budesonide}; CONMED_BUDESONIDE = as.integer(arm == "budesonide"); nedocromil is pooled into the reference because the paper reports no significant nedocromil effect on FEV1).
• Example models: Wu_2014_FEV1_asthma.R (additive shift on the linear-scale FEV1 typical-value: FEV1 = exp(theta_age*AGE + theta_ht*HT - theta_int) + e_conmed_bud_fev1 * CONMED_BUDESONIDE with e_conmed_bud_fev1 = +0.103 L (paper Table 2) – budesonide-arm subjects have ~0.10 L higher FEV1 relative to the placebo+nedocromil reference, with an IIV SD of 0.129 L on the drug-effect term itself; the high shrinkage 67.3% reported in the paper reflects that only ~30% of the cohort carries the indicator).
• Notes: Sibling of the existing head-to-head drug-arm CONMED_* family and follows the CONMED_<INN> convention (full INN spelled out, matching CONMED_METFORMIN, CONMED_PROBENECID, CONMED_SILDENAFIL, CONMED_SPIRON). The contrast is between an active treatment arm and a reference arm in a randomized trial, encoded as a per-subject time-fixed binary. Specific scope because the budesonide-vs-{placebo, nedocromil} head-to-head is tied to the CAMP-trial design; future inhaled-corticosteroid-vs-placebo popPD models can extend this entry’s example_models list, while a contrast between a different ICS (e.g., fluticasone, mometasone) and budesonide should register a sibling canonical rather than reusing CONMED_BUDESONIDE. Distinct from TRT (Novakovic 2017 categorical 3-level cladribine-arm covariate) and TRT_PHASE (active-treatment-phase indicator); CONMED_BUDESONIDE is a per-subject randomized-arm indicator rather than a time-varying treatment-phase flag.

CONMED_UMECLIDINIUM (canonical for inhaled-umeclidinium co-administration (treatment-arm) indicator)

• Description: 1 = inhaled umeclidinium is co-administered with the modelled drug (as the single-inhaler FF/UMEC/VI triple therapy, or as a separate UMEC inhaler alongside FF/VI); 0 = the modelled drug is given without umeclidinium (e.g. FF/VI dual therapy). Per-subject time-fixed binary carrying the randomized-arm assignment in a COPD inhaled-combination trial.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: per-model; in Mehta_2020_fluticasoneFuroate.R the model reference is 1 (umeclidinium co-administered) and the effect is applied to 1 - CONMED_UMECLIDINIUM.
• Source aliases:
  • treatment (FF/VI) – used in Mehta_2020_fluticasoneFuroate.R (Mehta 2020 Eq. 2 categorical indicator CAT = 1 for the FF/VI dual-therapy arm, so CAT = 1 - CONMED_UMECLIDINIUM; the reference arms FF/UMEC/VI and FF/VI + UMEC both contain umeclidinium).
• Example models: Mehta_2020_fluticasoneFuroate.R (log-additive effect on fluticasone furoate CL/F: exp(0.351 * (1 - CONMED_UMECLIDINIUM)), i.e. 42% higher CL/F in the FF/VI arm).
• Notes: Follows the CONMED_<INN> convention. The indicator records arm assignment; it does not assert a mechanism (Mehta 2020 offers none, and the FF/VI arm came from a single study, IMPACT). Distinct from CONMED_UMECLIDINIUM_DOSE (per-arm total daily dose in a COPD dose-response meta-analysis).

CONMED_NNRTI (canonical for concomitant NNRTI (any non-nucleoside reverse transcriptase inhibitor) co-administration indicator)

• Description: 1 = subject is on at least one concomitant non-nucleoside reverse transcriptase inhibitor (NNRTI) – typically efavirenz, nevirapine, etravirine, rilpivirine, doravirine, or delavirdine – at the PK observation; 0 = no concomitant NNRTI. A pooled class indicator used in popPK analyses of CYP3A4 substrates (such as HIV protease inhibitors) where the source paper does not separate the individual NNRTI effects because the included NNRTIs are both PXR-mediated CYP3A4 inducers with a similar net effect on CYP3A4 substrate clearance.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant NNRTI).
• Source aliases:
  • NNRTI – used in Kappelhoff_2005_indinavir.R (Kappelhoff 2005 Table 2 footnote: “NNRTI is 1 when efavirenz or nevirapine are administered and 0 when they are not”). Kappelhoff 2005 specifically pooled efavirenz and nevirapine because “inclusion of separate effects for each drug did not improve goodness-of-fit.”
• Example models: Kappelhoff_2005_indinavir.R (multiplicative power-form effect on CL/F: cl *= 1.41^CONMED_NNRTI; concomitant NNRTI use increases indinavir apparent CL by 41%).
• Notes: Distinct from the per-drug indicator CONMED_EFV (efavirenz only) – use CONMED_EFV when the source paper estimates the efavirenz effect alone and CONMED_NNRTI when the source paper pools the class. Per-model covariateData[[CONMED_NNRTI]]$notes should list which specific NNRTIs are counted as “1” in the dataset (Kappelhoff 2005: efavirenz and nevirapine; future extractions including rilpivirine / doravirine / etravirine / delavirdine should document each one).

CONMED_RTV_AUC_12H (canonical for ritonavir AUC over the 0-12 h q12h dosing interval (BID ritonavir regimens))

• Description: Per-subject (time-fixed within an evaluated regimen) ritonavir AUC over a 12-hour dosing interval (q12h, BID ritonavir), used as a co-medication exposure covariate driving boosted-protease-inhibitor (lopinavir, atazanavir, darunavir, etc.) clearance in popPK models that account for ritonavir’s CYP3A4-inhibition effect on a BID regimen. In Crommentuyn 2005 the value is computed per-subject as DOSE_RTV / CL_RTV using individual Bayesian CL_RTV estimates from the Kappelhoff et al. 2005 ritonavir popPK model (Crommentuyn 2005 reference [19], Br J Clin Pharmacol 59:174-82), and feeds the lopinavir CL/F inverse-saturable form cl = exp(lcl) * (auc50 / (auc50 + CONMED_RTV_AUC_12H)) * IND with auc50 = 2.26 mg*h/L (Crommentuyn 2005 Methods Equations 1-2; Table 2).
• Units: mg*h/L (document per-model via covariateData[[CONMED_RTV_AUC_12H]]$units if a different exposure unit is reported).
• Type: continuous
• Scope: specific
• Reference category: n/a – enters via the inverse-saturable form auc50 / (auc50 + CONMED_RTV_AUC_12H) (Crommentuyn 2005) or via a centred power form (vonHentig 2009, Chen 2024). Reference values observed: 3.58 mgh/L (Crommentuyn 2005 cohort median across 122 HIV-1-infected adults on BID LPV/r; range 0.85-18.77), 6.70355 mgh/L (vonHentig 2009 cohort median), 12.2 mg*h/L (Chen 2024 centring constant for critically ill adults on nirmatrelvir/ritonavir; the higher value reflects the reduced ritonavir CL/F of an ICU cohort, whose companion ritonavir model gives a typical CL/F of 10.3 L/h against roughly 13 L/h in the ambulatory HIV cohorts).
• Source aliases:
  • AUC12h – printed name in Crommentuyn 2005 (Methods Equations 1-2; Results page 6; Table 2 row 3). The paper writes the subscript 12h to denote the BID dosing-interval AUC; the column name is registered with the 12h token in the column name to keep the q12h-BID convention visible without relying on subscripts.
  • AUCRIT – printed name in Chen 2024 (Equation 1, Table 3 row dCLdAUCRIT, Table 4 column header, Figure 1B). Chen 2024 writes the subscript RIT for ritonavir and reports the quantity as mg/L*h; the ritonavir regimen is 100 mg q12h throughout, so the value is the 12 h dosing-interval AUC and belongs to this canonical rather than the q24h AUC_RTV / CONMED_RTV_AUC.
• Example models: Crommentuyn_2005_lopinavir.R (lopinavir CL/F inverse-saturable dependence on CONMED_RTV_AUC_12H: cl = exp(lcl) * (auc50 / (auc50 + CONMED_RTV_AUC_12H)) * IND with auc50 = 2.26 mg*h/L and IND = 1 + 0.39 * CONMED_NNRTI), vonHentig_2009_saquinavir.R (saquinavir CL/F centred power dependence on CONMED_RTV_AUC_12H: cl = exp(lcl) * e_atazanavir_cl^CONMED_ATAZANAVIR * (CONMED_RTV_AUC_12H / 6.70355)^e_rtv_auc_12h_cl with e_rtv_auc_12h_cl = -0.403; centred at the cohort median 6.70355 mg*h/L so saquinavir CL/F equals the typical 60.4 L/h at the median ritonavir exposure when atazanavir is absent), Chen_2024_nirmatrelvir.R (nirmatrelvir CL/F centred power dependence on CONMED_RTV_AUC_12H alongside CRCL: cl = exp(lcl + etalcl) * (CRCL / 80)^e_crcl_cl * (CONMED_RTV_AUC_12H / 12.2)^e_rtv_auc_12h_cl with e_rtv_auc_12h_cl = -0.45, Chen 2024 Equation 1; the per-subject covariate value is computed by the paper’s Equation 2 as 100 / cl_ritonavir from the companion ritonavir model fit in the same paper, modellib('Chen_2024_ritonavir'), so the two models compose directly).
• Notes: Specific scope – column meaning is tied to ritonavir as the booster drug AND to the q12h-BID dosing-interval AUC convention. Sibling of AUC_RTV (which is the q24h once-daily form, Dickinson 2009 cohort median 7.52 mgh/L). The two canonicals capture chemically-identical ritonavir AUC but at different dosing-interval scopes; never overload them. For simulation users without observed per-subject ritonavir AUC, the Crommentuyn 2005 cohort median 3.58 mgh/L reproduces typical-value behaviour (the value at which the inverse-saturable form gives CL/F = 5.73 L/h in the lopinavir + ritonavir steady-state simulation, matching the paper’s reported per-cohort-median CL/F).

CONMED_RTV_CC (canonical for plasma concentration of co-administered ritonavir acting on another drug’s clearance)

• Description: Plasma concentration of ritonavir given as a CYP3A4-inhibiting pharmacokinetic booster, supplied as a covariate column rather than computed from a coupled ritonavir PK model, and driving the inhibition of the boosted drug’s apparent clearance (typically an Imax / IC50 form). The consuming model’s covariateData[[CONMED_RTV_CC]]$notes must state which concentration the value is: in the founding model it is the measured per-subject ritonavir trough, held constant over time, not a time-varying concentration.
• Units: mg/L (document per-model via covariateData[[CONMED_RTV_CC]]$units).
• Type: continuous
• Scope: specific
• Reference category: 0 (no ritonavir on board; the inhibition term vanishes and the boosted drug’s CL/F equals its unboosted value). Reference values observed: ritonavir troughs from < 0.02 to 1.5 mg/L in Alvarez 2021’s Covid-19 cohort on 100 mg ritonavir BID (no median printed).
• Source aliases:
  • CresRTV (Alvarez 2021 Methods Equation 1 and Table 2 legend; measured ritonavir trough concentration in mg/L, same orientation as the canonical, no value transformation). The Methods text prints the unit as mg/mL, a typo contradicted by the Table 2 legend and the fixed IC50 of 0.057 mg/L.
• Example models: Alvarez_2021_lopinavir.R (lopinavir CL/F inhibited by the ritonavir trough, cl = exp(lcl + etalcl) * (1 - imax * CONMED_RTV_CC / (ic50 + CONMED_RTV_CC)) with imax = 0.929 and ic50 = 0.057 mg/L fixed from Dickinson 2011).
• Notes: Member of the CONMED_<drug>_CC family (CONMED_RIF_CC, CONMED_PROPOFOL_CC, …) for the concentration of a co-administered named drug; RTV follows the existing ritonavir members CONMED_RTV, CONMED_RTV_AUC_12H. Distinct from the exposure-integral siblings CONMED_RTV_AUC_12H (q12h AUC) and AUC_RTV / CONMED_RTV_AUC (q24h AUC): a concentration and an AUC are not interchangeable, and an IC50 fitted against one cannot be applied to the other. Zhang_2012_lopinavir_ritonavir.R uses the same Imax form but solves ritonavir concentration from its own coupled PK model, so it needs no covariate column.

CONMED_SEVO (canonical for concomitant sevoflurane volatile-anaesthesia maintenance indicator)

• Description: 1 = subject received sevoflurane as the volatile anaesthetic for maintenance of general anaesthesia during the perioperative observation window (typically a propofol IV induction followed by sevoflurane gas maintenance), 0 = no sevoflurane maintenance (e.g., propofol-only TIVA, isoflurane, desflurane, or another agent). Time-fixed per subject in perioperative-PK / NMB-reversal models where sevoflurane is administered as a continuous maintenance gas from induction until reversal of neuromuscular blockade. Distinct from ETSEVO (continuous end-tidal sevoflurane concentration in vol %), which is the quantitative driver used in emergence-from-anaesthesia models such as Shin 2014; CONMED_SEVO is the per-subject on/off flag.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no sevoflurane; typically propofol-only TIVA or another volatile agent).
• Source aliases:
  • SEV – used in Kleijn_2011_sugammadex_rocuronium.R (Kleijn 2011 Table 4 covariate-effect rows “No sevoflurane: keoSEV = 1” and “Sevoflurane: keoSEV = 1 + theta”).
• Example models: Kleijn_2011_sugammadex_rocuronium.R (multiplicative effect on the rocuronium effect-compartment equilibration rate keo *= (1 + e_conmed_sevo_keo * CONMED_SEVO) with e_conmed_sevo_keo = -0.567 and on the rocuronium NMB EC50 ec50 *= (1 + e_conmed_sevo_ec50 * CONMED_SEVO) with e_conmed_sevo_ec50 = -0.395; sevoflurane simultaneously slows the central-to-effect-compartment equilibration and lowers the rocuronium concentration required for half-maximal neuromuscular blockade, reproducing the well-known sevoflurane potentiation of aminosteroid NMBAs).
• Notes: Distinct from ETSEVO (end-tidal sevoflurane volume percent, a continuous PD driver). Use CONMED_SEVO when the source paper encodes sevoflurane maintenance as a per-subject on/off indicator without a continuous end-tidal trajectory; use ETSEVO when the per-record end-tidal concentration is available and the model uses it directly. Follows the CONMED_* family pattern (CONMED_VPA, CONMED_RIF, CONMED_AMIO). Future perioperative models that use a similar binary indicator for isoflurane (CONMED_ISO) or desflurane (CONMED_DES) should register parallel canonicals rather than overloading this name.

CONMED_DOPA (canonical for concomitant dopamine (inotrope) coadministration indicator)

• Description: 1 = subject is receiving dopamine (an inotropic / vasoactive agent used for haemodynamic support in critically ill neonates, infants and adults) at the PK observation; 0 = no concomitant dopamine. Treated as time-varying per observation in the source paper (each PK record carries the dopamine flag corresponding to the dosing window).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant dopamine).
• Source aliases:
  • DOPA – used in Fuchs_2014_gentamicin.R (paper covariate name in Table 2; same orientation as the canonical, 1 = on dopamine, 0 = otherwise).
  • DA – used in Lin_2021_vancomycin.R (Lin 2021 Methods covariate list; coded 1 = dopamine not used, 2 = dopamine used, so CONMED_DOPA = DA - 1; model() rebuilds the source code as CONMED_DOPA + 1).
• Example models: Fuchs_2014_gentamicin.R (multiplicative linear effect on CL: 1 + (-0.120) * CONMED_DOPA, i.e. 12% lower CL on dopamine-coadministered neonates relative to no-dopamine reference; Fuchs 2014 Table 2 theta_CL_DOPA = -0.120, SE 22%), Lin_2021_vancomycin.R (additive-linear term in an additive-linear CL equation: +0.036 L/h * DA with DA = CONMED_DOPA + 1, i.e. dopamine raises CL by 0.036 L/h in adult ICU patients), Samb_2022_gentamicin.R (the Fuchs 2014 plasma layer carried with every parameter fixed, 1 + (-0.120) * CONMED_DOPA on CL, under a saliva compartment).
• Notes: Promoted to general scope when Lin_2021_vancomycin.R (adult ICU) reused the indicator with the same meaning as the neonatal Fuchs_2014_gentamicin.R. The two papers’ effects run in opposite directions (Fuchs lowers CL, Lin raises it), which is expected of a marker that partly tracks haemodynamic instability. Fuchs 2014 Discussion notes that the effect “might principally reflect cardiovascular instability in critically ill newborns” – so the dopamine flag is partly a marker of disease severity rather than a clean drug-drug-interaction effect. The mechanism on glomerular filtration in neonates remains debated (Fuchs 2014 Discussion citations [57-59]).

CONMED_PROPOFOL_CC (canonical for measured plasma concentration of co-administered propofol)

• Description: Time-varying measured plasma concentration of propofol administered concomitantly as the intravenous general-anaesthetic agent, supplied as a covariate column rather than computed from a coupled propofol PK model. Used as the perpetrator exposure in models where propofol modifies the disposition of another drug – Li 2024 found that measured propofol concentration explains the reduced norepinephrine clearance seen under general anaesthesia better than the accompanying fall in cardiac output does, consistent with propofol’s inhibition of the norepinephrine transporter. Set to 0 whenever no propofol is on board (awake phase, before induction, after full washout), which collapses the exponential effect term to 1.
• Units: ug/mL (document per-model via covariateData[[CONMED_PROPOFOL_CC]]$units if a different exposure unit is reported).
• Type: continuous
• Scope: specific
• Reference category: 0 (no propofol administered). Reference values observed: median measured propofol concentration during Eleveld-model target-controlled maintenance anaesthesia was 3.53 ug/mL in Li 2024’s healthy-volunteer cohort; a typical maintenance target is 3 ug/mL and a high target 6 ug/mL.
• Source aliases:
  • CPROP (Li 2024 Table 3 covariate-effect row “CPROP~CL” and Results 3.2; per-record measured propofol plasma concentration in ug/mL, same orientation as the canonical, no value transformation).
• Example models: Li_2024_norepinephrine.R (exponential effect on norepinephrine clearance, cl *= exp((e_conmed_propofol_cc_cl + etae_conmed_propofol_cc_cl) / 100 * CONMED_PROPOFOL_CC) with e_conmed_propofol_cc_cl = -3.57 per 100 ug/mL; the covariate replaced a binary awake-versus-anaesthesia session factor in the final model).
• Notes: Follows the registered CONMED_<drug>_CC family (CONMED_RIF_CC, CONMED_INH_CC, CONMED_MER_CC,…) for the time-varying concentration of a co-administered named drug, with the full INN spelled out as in CONMED_SILDENAFIL_NMETAB_CC because a PROP abbreviation would collide with propranolol and propranolol-like INNs. Distinct from CEFFECT, which carries the propofol effect-site concentration acting as the PD driver of propofol’s own anaesthetic effect (Koo_2012_propofol.R); CONMED_PROPOFOL_CC is a measured plasma concentration acting as a perpetrator on a different drug’s PK. Also distinct from the binary CONMED_SEVO / volatile-agent indicators, which carry no magnitude information.

CONMED_KETOCONAZOLE (canonical for concomitant ketoconazole (combined strong CYP3A4 and P-glycoprotein inhibitor) coadministration indicator)

• Description: 1 = subject coadministered ketoconazole during the observation interval (typically 400 mg once-daily oral, the standard index-inhibitor regimen), 0 = no concomitant ketoconazole. Ketoconazole is the classical probe combined strong CYP3A4 + P-glycoprotein inhibitor and is the perpetrator arm of many dedicated DDI studies and DDI-prediction PBPK analyses. Per-subject time-varying when a DDI study spans on / off ketoconazole periods; time-fixed on the on-treatment day of a fixed-sequence DDI arm.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant ketoconazole).
• Source aliases:
  • No source column name; Choules 2024 encodes the perpetrator arm as a separate simulation rather than as a dataset column, so the indicator is constructed by the model user.
• Example models: Choules_2024_enfortumab.R, Choules_2024_brentuximab.R (multiplicative factor on MMAE clearance: cl_mmae *= e_keto_cl_mmae^CONMED_KETOCONAZOLE with e_keto_cl_mmae = 1/1.38 = 0.72464 for enfortumab vedotin and 1/1.37 = 0.72993 for brentuximab vedotin, i.e. a 27-28% reduction in MMAE clearance reproducing the published MMAE AUC geometric mean ratios of 1.38 and 1.37 in Choules 2024 Table 5), Huth_2019_siponimod.R (relative CYP3A4 activity 0.084 – the most complete CYP3A4 inhibition in that paper – producing a predicted siponimod AUC ratio of 4.20 in the CYP2C93/3 subpopulation against 1.24 in 1/1).
• Notes: Auto-approved member of the CONMED_<INN> family (INN = ketoconazole). Distinct from the class-level [[CONMED_AZOLE]] (which pools ketoconazole with the other systemic azoles as a common inhibitor class effect) and from [[CONMED_CYP3A4_INH_STRONG]]: register the drug-specific indicator when a paper singles ketoconazole out with its own coefficient, which is the usual case because ketoconazole is chosen precisely for the combined CYP3A4 + P-gp mechanism rather than for CYP3A4 inhibition alone, so its effect size is not interchangeable with a CYP3A4-only strong inhibitor. Sibling to [[CONMED_ITRACONAZOLE]] and [[CONMED_VORICONAZOLE]]. When a model lumps the CYP3A4 and P-gp components of the interaction into a single clearance multiplier (as the Choules 2024 reductions do), say so in covariateData[[CONMED_KETOCONAZOLE]]$notes so a downstream user does not attribute the whole effect to CYP3A4. Frequently paired with a [[CONMED_RIF]] inducer arm in the same model.

CONMED_NUC (canonical for concomitant nucleos(t)ide analogue antiviral coadministration indicator)

• Description: 1 = subject (or, in a meta-analysis, the treatment arm) received a nucleos(t)ide analogue (NA) antiviral – entecavir, tenofovir disoproxil fumarate, tenofovir alafenamide, lamivudine, adefovir, telbivudine or another agent of the class – concomitantly with the index therapy; 0 = no concomitant nucleos(t)ide analogue. In chronic hepatitis B, NAs suppress HBV replication by inhibiting the viral reverse transcriptase and are the standard-of-care backbone that immunomodulatory agents are added to, so the indicator most often distinguishes combination therapy from monotherapy.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant nucleos(t)ide analogue; i.e. index-drug monotherapy).
• Source aliases:
  • Treatment group (Peg-IFNalpha vs Peg-IFNalpha + NA) – used in Hanan_2026_peginterferon_alfa_eot_mbma.R and Hanan_2026_peginterferon_alfa_24wk_mbma.R (Hanan 2026 Table 2 “Drug effect” rows; same orientation as the canonical, 1 = the combination arm).
• Example models: Hanan_2026_peginterferon_alfa_eot_mbma.R (selects the regimen-specific logit-scale intercept: drug_eff = logitp_mono * (1 - CONMED_NUC) + logitp_comb * CONMED_NUC, giving 8.5% vs 11.7% HBsAg loss at end of treatment in the reference population), Hanan_2026_peginterferon_alfa_24wk_mbma.R (same intercept-selection form; 11.1% vs 14.1% at 24 weeks post-treatment).
• Notes: Class-level member of the CONMED_ family, alongside CONMED_AZOLE, CONMED_ABX, CONMED_AED and CONMED_DIUR, which are likewise drug classes rather than single INNs; register a drug-specific sibling (CONMED_ENTECAVIR, CONMED_TENOFOVIR) when a paper resolves the class into individual agents with their own coefficients rather than overloading this column. The class token is NUC, not the NA abbreviation used throughout the hepatitis B literature, because NA is R’s missing-value literal and a column of that name is a latent bug in any downstream data-assembly step. Per-model covariateData[["CONMED_NUC"]]$notes must document (1) which agents are pooled into the indicator, and (2) the exposure window the flag refers to – in Hanan 2026 it marks NA coadministration during the Peg-IFNalpha course, while whether the protocol permitted NA to be continued after Peg-IFNalpha cessation is a separate covariate (screened in that paper’s 24-week model and not retained). Note also that in the founding models the indicator is a treatment-arm definition selecting between two estimated intercepts, not an additive covariate effect with its own coefficient; that is the form the authors estimated.

CONMED_TAF (canonical for concomitant emtricitabine/tenofovir alafenamide (FTC/TAF) nucleos(t)ide backbone indicator)

• Description: 1 = the subject’s antiretroviral nucleos(t)ide reverse-transcriptase-inhibitor backbone is emtricitabine + tenofovir alafenamide (FTC/TAF), given as a fixed-dose combination alongside the index antiretroviral; 0 = a different nucleos(t)ide backbone. In HIV treatment the index drug (an integrase inhibitor, NNRTI or protease inhibitor) is always paired with a two-drug nucleos(t)ide backbone, so the reference category is another backbone, never “no backbone” – which is what distinguishes this canonical from the class-level [[CONMED_NUC]], whose reference category is index-drug monotherapy. Per-model notes must state which backbone(s) the reference category pools.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (a non-FTC/TAF nucleos(t)ide backbone; the specific comparator is paper-specific and must be named in the per-model notes).
• Source aliases:
  • BB – used in Waalewijn_2024_dolutegravir.R (CHAPAS-4 NMTRAN $INPUT backbone column; IF (BB==1) BB_BIO = THETA(14) selects the FTC/TAF bioavailability, with THETA(4) = 1 FIX for the standard-of-care backbones).
• Example models: Waalewijn_2024_dolutegravir.R (linear additive effect on relative bioavailability: fdepot = exp(lfdepot + iov_fdepot) * (1 + e_taf_fdepot * CONMED_TAF) with e_taf_fdepot = -0.196, i.e. dolutegravir F falls 19.6% on an FTC/TAF backbone relative to a pooled standard-of-care reference of lamivudine/abacavir or lamivudine/zidovudine; Waalewijn 2024 Supplementary Table S3).
• Notes: A drug-specific sibling of the class-level [[CONMED_NUC]], registered per that entry’s instruction to resolve the class into individual agents when a paper gives them their own coefficients. Follows the antiretroviral three/four-letter abbreviation convention already used by [[CONMED_RTV]] (ritonavir), [[CONMED_EFV]] (efavirenz), [[CONMED_NVP]] (nevirapine), [[CONMED_LPV]] (lopinavir) and [[CONMED_COBICISTAT]]. The column flags the FTC/TAF combination product, not isolated tenofovir alafenamide exposure. Emtricitabine and tenofovir alafenamide are co-formulated and never separated in the trials that estimate this effect, so the coefficient is a combination-versus-combination contrast and must not be read as a TAF-specific drug-drug interaction; a paper that does separate them needs distinct columns. It is likewise distinct from a continuous TAF or tenofovir exposure covariate – Waalewijn 2024 tested TAF and tenofovir AUC as continuous replacements and found neither improved on the categorical indicator, the same binary-versus-continuous distinction drawn between [[CONMED_RTV]] and AUC_RTV. Register a CONMED_TDF sibling rather than overloading this column for tenofovir disoproxil fumarate backbones: the two tenofovir prodrugs differ substantially in plasma tenofovir exposure and in their interaction profiles.

CONMED_DOPA_AGONIST (canonical for concomitant dopamine-agonist coadministration indicator)

• Description: 1 = the subject is receiving a dopamine-receptor agonist as a concomitant anti-Parkinson medication; 0 = not. A drug-class indicator pooling ergot and non-ergot D2/D3 agonists used as levodopa-sparing adjuncts in Parkinson disease.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no concomitant dopamine agonist).
• Source aliases:
  • DOPA – printed name in Knebel 2012 (individual-parameter equations and Table II row labels, e.g. *theta8 DOPA[dopamine agonists yes]).
• Example models: Knebel_2012_istradefylline_offtime.R (multiplicative effect e_conmed_dopa_agonist_<param>^CONMED_DOPA_AGONIST on the baseline percentage OFF time E0, on EmaxP and on EmaxI; Knebel 2012 Table II thetas 8, 15 and 22).
• Notes: Specific scope. Member of the drug-class arm of the CONMED_<X> family (CONMED_STATIN, CONMED_CCB, CONMED_PPI, CONMED_NSAID, CONMED_AZOLE, CONMED_H2RA, CONMED_SGLT2I, CONMED_PDE5I, CONMED_IL6RI, CONMED_ERA, CONMED_CARBAPENEM, CONMED_DIURETIC). Deliberately NOT named CONMED_DOPA, which is already taken and means something different – that canonical is the concomitant dopamine (inotrope) indicator for haemodynamic support in critical care, whereas this one is the dopamine-agonist anti-Parkinson drug class. The name collision is exactly why the _AGONIST token is spelled out. Agents pooled under this indicator in Knebel 2012 Table III: ropinirole, pramipexole, pergolide, bromocriptine. The most prevalent concomitant medication in the founding cohort at 63 percent of patients. A model needing a single named agonist rather than the class should register CONMED_<INN> (e.g. CONMED_PRAMIPEXOLE) instead.

CONMED_COMTI (canonical for concomitant catechol-O-methyl transferase (COMT) inhibitor coadministration indicator)

• Description: 1 = the subject is receiving a catechol-O-methyl transferase inhibitor as a concomitant anti-Parkinson medication; 0 = not. COMT inhibitors block the peripheral O-methylation of levodopa and are given with levodopa/carbidopa to prolong its effect.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no concomitant COMT inhibitor).
• Source aliases:
  • COMT – printed name in Knebel 2012 (individual-parameter equations and Table II row labels, e.g. *theta9 COMT[COMT inhibitors yes]).
• Example models: Knebel_2012_istradefylline_offtime.R (multiplicative effect e_conmed_comti_<param>^CONMED_COMTI on the baseline percentage OFF time E0, on EmaxP and on EmaxI; Knebel 2012 Table II thetas 9, 16 and 23).
• Notes: Specific scope. Member of the drug-class arm of the CONMED_<X> family, following the enzyme-or-receptor-target-plus-I naming of CONMED_SGLT2I, CONMED_PDE5I and CONMED_IL6RI. Agents pooled under this indicator in Knebel 2012 Table III: entacapone, tolcapone. Present in 36 percent of the founding cohort. Knebel 2012 flags this covariate’s estimated effect as internally inconsistent and cautions against acting on it: the estimate increases the maximum istradefylline effect in istradefylline-treated patients while decreasing the maximum disease-progression/placebo-response effect in placebo-treated patients, both estimates carry wide bootstrap intervals, and a separate subgroup analysis of the same trials did not identify COMT-inhibitor use as a significant covariate (p = 0.939). Do not treat the founding model’s coefficients as a validated drug-interaction effect. A model needing a single named inhibitor rather than the class should register CONMED_<INN> (e.g. CONMED_ENTACAPONE) instead.

CONMED_ENTACAPONE (canonical for concomitant entacapone coadministration indicator)

• Description: 1 = the subject is receiving entacapone (a reversible peripheral catechol-O-methyl transferase inhibitor) together with levodopa; 0 = not. Per-subject time-varying when a crossover alternates entacapone-containing and entacapone-free levodopa treatment.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no entacapone).
• Source aliases:
  • No source column name; Senek 2020 encodes the treatment arm (LECIG, levodopa-entacapone-carbidopa intestinal gel, vs LCIG, levodopa-carbidopa intestinal gel).
• Example models: Senek_2020_levodopa.R (fractional shift on levodopa CL/F, cl *= 1 + e_conmed_entacapone_cl * exp(etae_conmed_entacapone_cl) * CONMED_ENTACAPONE, with e_conmed_entacapone_cl = -0.365 and an 11.4% CV IIV on the shift; Senek 2020 Table 2).
• Notes: Auto-approved member of the CONMED_<INN> family. Single-agent sibling of the class-level [[CONMED_COMTI]] (entacapone + tolcapone pooled); use this name when a paper estimates an entacapone-specific effect. In the founding model entacapone is co-formulated in the infused gel, so the indicator also carries any formulation difference between LECIG and LCIG that the source did not separate from the COMT inhibition.

CONMED_SELEGILINE (canonical for concomitant selegiline coadministration indicator)

• Description: 1 = the subject is receiving selegiline (a selective, irreversible monoamine oxidase type B inhibitor used as an anti-Parkinson adjunct) as a concomitant medication; 0 = not.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no concomitant selegiline).
• Source aliases:
  • SELG – printed name in Knebel 2012 (individual-parameter equations and Table II row labels, e.g. *theta10 SELG[selegiline yes]). Knebel 2012 misspells the drug as “selegine” once in the equation glossary.
• Example models: Knebel_2012_istradefylline_offtime.R (multiplicative effect e_conmed_selegiline_<param>^CONMED_SELEGILINE on the baseline percentage OFF time E0, on EmaxP and on EmaxI; Knebel 2012 Table II thetas 10, 17 and 24).
• Notes: Specific scope. Member of the CONMED_<INN> single-agent family (CONMED_ATORVASTATIN, CONMED_KETOCONAZOLE, CONMED_METFORMIN, CONMED_QUETIAPINE, CONMED_SILDENAFIL, CONMED_VENLAFAXINE,…). Present in only 13 percent of the founding cohort, the least prevalent of that model’s four concomitant-medication indicators and correspondingly the least precisely estimated – the EmaxP bootstrap interval runs from 8.60e-11 to 1.72, i.e. essentially uninformative. A future model pooling MAO-B inhibitors as a class (selegiline plus rasagiline plus safinamide) should register a CONMED_MAOBI class sibling rather than overload this single-agent name.

CONMED_SEMAGLUTIDE (canonical for concomitant semaglutide (GLP-1 receptor agonist) coadministration indicator)

• Description: 1 = the victim drug is dosed while the subject is on semaglutide (at the regimen the source studied, typically once-weekly SC semaglutide at steady state); 0 = no semaglutide. Semaglutide is a long-acting GLP-1 receptor agonist that delays gastric emptying, so its DDI effects on orally co-administered drugs are usually placed on the absorption parameters (ka, ktr, lag time) rather than on clearance.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant semaglutide).
• Source aliases:
  • placebo – used in Langeskov_2022_paracetamol.R and Langeskov_2022_atorvastatin.R (Langeskov 2022 Phoenix PML ka = tvka * (1 - dKadplacebo1 * (placebo == 1))). Despite its name the source column was coded 1 for semaglutide co-administration and 0 without (Section 3.2), so no transformation is needed.
• Example models: Langeskov_2022_paracetamol.R (fractional reduction ka * (1 - 0.525 * CONMED_SEMAGLUTIDE), Table 2), Langeskov_2022_atorvastatin.R (fractional reductions of ka by 0.829 and of ktr by 0.791, Table 3); both at steady-state once-weekly SC semaglutide 1.0 mg, per-occasion indicator in a crossover.
• Notes: Auto-approved member of the CONMED_<INN> family. Distinct from the semaglutide dose covariates DOSE_SEMAGLUTIDE_MG, DOSE_SEMAGLUTIDE_INJ_MG and DOSE_SEMAGLUTIDE_PO_MG, which carry an assigned semaglutide dose for models of semaglutide itself; this column flags semaglutide as the perpetrator in a DDI analysis of another drug. Record per-model the semaglutide regimen and whether steady state was reached before the victim dose. A future model pooling GLP-1 receptor agonists as a class should register a class sibling rather than overload this single-agent name.

CONMED_AMANTADINE (canonical for concomitant amantadine coadministration indicator)

• Description: 1 = the subject is receiving amantadine (an NMDA-receptor antagonist with weak dopaminergic activity, used as an anti-Parkinson adjunct and in particular against levodopa-induced dyskinesia) as a concomitant medication; 0 = not.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no concomitant amantadine).
• Source aliases:
  • AMAT – printed name in Knebel 2012 (individual-parameter equations and Table II row labels, e.g. *theta11 AMAT[amantadine yes]).
• Example models: Knebel_2012_istradefylline_offtime.R (multiplicative effect e_conmed_amantadine_<param>^CONMED_AMANTADINE on the baseline percentage OFF time E0, on EmaxP and on EmaxI; Knebel 2012 Table II thetas 11, 18 and 25).
• Notes: Specific scope. Member of the CONMED_<INN> single-agent family. Present in 26 percent of the founding cohort. Worth noting for anyone reusing this column alongside the companion Knebel_2012_istradefylline_dyskinesia.R model: amantadine is itself an anti-dyskinesia agent, so in a dyskinesia exposure-response analysis this indicator is a treatment for the endpoint rather than a neutral demographic covariate. Knebel 2012 screened it on the adverse-event endpoints but retained no covariates there (poor precision), so no such coefficient exists in the library; a future model that does estimate one should document the confounding explicitly.

CONMED_CZP (canonical for concomitant clonazepam coadministration indicator)

• Description: 1 = subject is taking clonazepam (CZP) as a concomitant antiepileptic drug at the observation, 0 = no concomitant clonazepam. Clonazepam is a 1,4-benzodiazepine anticonvulsant; it is not an enzyme inducer, so where it carries a coefficient the mechanism is usually pharmacodynamic (additive or antagonistic seizure control) rather than a clearance interaction. Time-varying when clonazepam starts / stops within the observation window; time-fixed when the source paper analyses chronic-maintenance cohorts whose antiseizure therapy is stable across the analysis window.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant clonazepam).
• Source aliases:
  • CZP – used in Nakashima_2015_valproic_acid.R (Nakashima 2015 Eq 6 and Table 3 intercept row CZP) and in Nakashima_2015_valproic_acid_ge19y.R (Nakashima 2015 Eq 8).
• Example models: Nakashima_2015_valproic_acid.R (additive log-odds shift on the intercept of a logistic exposure-response model for an over-50% reduction in seizure frequency: e_czp_logit = -1.18, i.e. concomitant clonazepam lowers the odds of response; Nakashima 2015 Table 3), Nakashima_2015_valproic_acid_ge19y.R (the same term with a much larger coefficient, -2.56, in the >= 19 y subgroup; Nakashima 2015 Eq 8).
• Notes: Drug-specific CONMED_* indicator anticipated in the [[CONMED_AED]] notes; use it when a paper estimates a separate clonazepam effect distinct from the pooled EIAED / AED class effect. Distinct from the broader [[CONMED_EIAED]] (any enzyme-inducing AED) and [[CONMED_AED]] (any concomitant AED); clonazepam is NOT enzyme-inducing and must not be folded into CONMED_EIAED. This is a PD (efficacy) covariate in its founding source, not a clearance covariate – clonazepam was not retained on valproate CL/F in the upstream PK model (Ogusu 2014), and Nakashima 2015 retained it only on the efficacy logit. The founding authors state the mechanism is unclear (Nakashima 2015 Discussion: “the mechanisms underlying the association between the co-administration of CZP and the probability of a more than 50% reduction in seizure frequency are currently unclear”), so do not read a directional pharmacological prior into the sign.

CONMED_ANTIPLATELET (canonical for concomitant antiplatelet indicator; class composition is paper-specific)

• Description: 1 = subject is receiving an antiplatelet agent (aspirin, clopidogrel, ticagrelor, prasugrel, ticlopidine, dipyridamole, cilostazol etc.) over the relevant observation window – or, for per-time-point datasets, at the current time; 0 = no concomitant antiplatelet. Used both as a bleeding / cardiovascular-outcome covariate in event-risk models and as a PK covariate for substrates that share metabolic or transporter pathways with the thienopyridines. The class membership pooled into the indicator is paper-specific and SHOULD be enumerated in covariateData[[CONMED_ANTIPLATELET]]$notes per model – sources range from aspirin-only through any-oral-antiplatelet. Composite class indicator following the CONMED_DIURETIC / CONMED_STATIN / CONMED_AED pooled-class pattern rather than the CONMED_<INN> per-drug pattern.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant antiplatelet of the paper-specific class set).
• Source aliases:
  • APLT – used in Elhefnawy_2023_recurrent_ischemic_stroke.R (Elhefnawy 2023 Table 1 / Table 3 row name; the indicator records an antiplatelet prescribed at discharge from the index-stroke admission, so it encodes prescription rather than adherence, and the paper does not enumerate which agents were pooled).
  • ANTIPLT, ASA – common clinical-dataset abbreviations; use the aspirin-specific CONMED_ASA form only if a future source restricts the indicator to aspirin monotherapy while also carrying a separate broader antiplatelet column.
• Example models: Elhefnawy_2023_recurrent_ischemic_stroke.R (log-linear effect on the recurrent-ischemic-stroke hazard: exp(-0.514 * CONMED_ANTIPLATELET), i.e. an adjusted hazard ratio of 0.59 – about a 40 percent reduction in recurrence hazard; Elhefnawy 2023 Table 3 theta8).
• Notes: Composite indicator with paper-specific class membership; users simulating across models that share this column but differ in class membership MUST populate it according to each paper’s own definition. Distinct from CONMED_ANTICOAG-style anticoagulant indicators (different mechanism and different bleeding profile) – register those separately rather than pooling them here, even though registry datasets sometimes combine the two under an “antithrombotic” heading. When a paper requires drug-resolved encoding, register sibling CONMED_<INN> canonicals (e.g. CONMED_CLOPIDOGREL) rather than overloading this one. Related but distinct from the pharmacogenetic canonical CYP2C19_S2_CARRIER, which modulates the activation of clopidogrel rather than recording whether an antiplatelet is being taken.

CONMED_ANTIANDROGEN (canonical for concomitant antiandrogen indicator; class composition is paper-specific)

• Description: 1 = subject received an antiandrogen (androgen-receptor antagonist) in the paper-defined window, 0 = not. Used in prostate-cancer disease-progression models of androgen-deprivation therapy, where a short antiandrogen course around LHRH-agonist initiation (to block the testosterone flare) or combined androgen blockade can deepen the PSA response. The class membership and the time window pooled into the indicator are paper-specific and MUST be enumerated in covariateData[[CONMED_ANTIANDROGEN]]$notes per model. Composite class indicator following the CONMED_DIURETIC / CONMED_ANTIPLATELET pattern rather than the CONMED_<INN> per-drug pattern.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no antiandrogen of the paper-specific class set in the paper-specific window).
• Source aliases:
  • AND / IND_AND – used in Zou_2020_leuprorelin.R (bicalutamide, enzalutamide, flutamide or nilutamide dispensed within 30 days of leuprorelin initiation; 33 of 264 subjects).
• Example models: Zou_2020_leuprorelin.R (exponential effect on the PSA kill rate: kse * exp(0.677 * CONMED_ANTIANDROGEN), a 96.8 percent higher kill rate; Zou 2020 Eq 17, Table 2).
• Notes: Spelled out rather than CONMED_NSAA (nonsteroidal antiandrogen); a paper whose class set is restricted to nonsteroidal agents still uses this canonical with the composition in the notes. Distinct from CONMED_CYPROTERONE (steroidal antiandrogen, per-drug indicator) and from CONMED_ABI (abiraterone, an androgen-synthesis inhibitor rather than a receptor antagonist); when a paper requires drug-resolved encoding, register sibling CONMED_<INN> canonicals (e.g. CONMED_BICALUTAMIDE) rather than overloading this one.

RHEUMATOID_FACTOR (canonical for serum rheumatoid factor concentration)

• Description: Serum rheumatoid factor (an autoantibody, predominantly IgM, directed against the Fc portion of IgG) concentration. Baseline value typical; document time-varying use in per-model notes.
• Units: U/mL or IU/mL (interchangeable in the clinical-PK literature). Document per-model via covariateData[[RHEUMATOID_FACTOR]]$units.
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling on a log-transformed value: (log(RHEUMATOID_FACTOR) / log(ref))^exponent (or, equivalently, the source-paper form (LRF / log(ref))^exponent where LRF = log(RHEUMATOID_FACTOR)). Reference value observed: 110 U/mL (Frey 2010, corresponding to LRF = 4.7 in the paper’s final-model equation).
• Source aliases:
  • LRF – log-transformed RF (natural log of the value in U/mL); Frey 2010 fits the covariate on the log scale and reports the reference as LRF = 4.7 (i.e., log(110) ~= 4.7). The canonical column carries the raw RF concentration in U/mL; the log transform is applied inside model.
  • RF – universal NONMEM/clinical-PK abbreviation; rejected as the canonical name on 2026-04-28 because the bare two-letter abbreviation is uncommon in published popPK papers and could be confused with other shortenings.
• Example models: Frey_2010_tocilizumab.R (U/mL, reference 110 U/mL == LRF = 4.7; small positive exponent +0.1 on linear CL applied to log(RHEUMATOID_FACTOR)).
• Notes: RF concentrations span several orders of magnitude across the rheumatoid-arthritis population (Frey 2010 observed range 15-11,800 U/mL across the four phase-III studies; reference paper: Frey 2010 Table I), motivating the log transform before power scaling. The mechanistic rationale (Frey 2010 Discussion, p764) is that RF – being an anti-IgG autoantibody – could in principle bind the Fc region of the therapeutic IgG monoclonal antibody and accelerate clearance, but the observed CL effect was small (-4.9% to +6.5% across the observed RF range) and the paper acknowledges that high RF concentrations may also reduce the assay’s ability to detect the drug, leading to an apparent CL increase.

BLPHYVAS (canonical for baseline physician’s global assessment VAS)

• Description: Baseline Physician’s Global Assessment of Disease Activity, 100-mm visual analogue scale (0 = no disease activity, 100 = maximum). Time-fixed per subject.
• Units: mm (0-100 VAS)
• Type: continuous
• Scope: general
• Reference category: n/a – used as a power term (BLPHYVAS / <ref>)^exponent. Reference 66 used in Ma 2020.
• Source aliases: none.
• Example models: Ma_2020_sarilumab_das28crp.R.
• Notes: One of the components of the DAS28 composite score; in Ma 2020 it appears as a baseline covariate on the DAS28-CRP disease-activity BASE rather than on the score itself. Applicable to any rheumatology model where baseline physician-assessed disease activity is used as a PK/PD covariate.

BLHAQ (canonical for baseline HAQ-DI score)

• Description: Baseline Health Assessment Questionnaire Disability Index (HAQ-DI; 0 = no disability, 3 = maximum disability). Time-fixed per subject.
• Units: unitless (0-3 composite score)
• Type: continuous
• Scope: general
• Reference category: n/a – used as a power term (BLHAQ / <ref>)^exponent. Reference 1.75 used in Ma 2020.
• Source aliases: none.
• Example models: Ma_2020_sarilumab_das28crp.R.
• Notes: Patient-reported disability score frequently used as a baseline covariate in rheumatoid-arthritis PK/PD analyses.

PAIN (canonical for patient-reported global pain visual analogue score)

• Description: Patient-reported global pain on a 100-mm visual analogue scale (PAIN; 0 = no pain, 100 = worst imaginable pain). Distinct from BLPHYVAS (the physician’s global assessment of disease activity). Both baseline and time-varying usages are covered; document per-model in covariateData[[PAIN]]$notes whether the column is baseline-only.
• Units: mm (0-100 VAS).
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (PAIN / ref)^exponent. Reference value observed: 60 in Frey 2013 (approximate dataset median across OPTION/TOWARD).
• Source aliases: none known (the PAIN column name is used directly in the Frey 2013 NONMEM control stream).
• Example models: Frey_2013_tocilizumab.R (baseline; power effect on the indirect-response BASE parameter with reference 60 and exponent 0.062), Mercier_2014_tramadol_tapentadol_mbma.R (arm-level baseline pain intensity on the 0-10 normalized scale used across the pooled MBMA trials – see per-model covariateData[[PAIN]]$notes for the 0-10 vs 0-100 scale convention).
• Notes: PAIN can take the value 0 in real cohorts (a patient who reports no pain at the assessment), which makes the bare power form (PAIN/60)^exp return 0. Frey 2013 documents an explicit 0.010 floor on PAIN in its Table 2 covariate range; the model file applies the same floor inside model so simulation under PAIN = 0 returns a finite BASE rather than collapsing to zero. Canonical name follows the EOS / SCORE_EASI convention (no BL prefix); baseline-vs-time-varying status is per-model. The canonical scale is 0-100 mm VAS; MBMA extractions that pool multiple heterogeneous pain scales (e.g., Mercier 2014) may carry PAIN on a 0-10 normalized scale – document the scale per-model in covariateData[[PAIN]]$notes so downstream users know to rescale before merging with a 0-100 mm cohort.

SWOL_28JOINT (canonical for 28-joint swollen joint count)

• Description: Swollen joint count on the 28-joint (DAS28) scale (integer 0-28; component of the DAS28 composite). Baseline value is typical; document time-varying use in per-model notes.
• Units: count (0-28)
• Type: continuous
• Scope: general
• Reference category: n/a – used as a shifted power term ((SWOL_28JOINT + 1)/(<ref> + 1))^exponent to avoid the zero-count edge case. Reference value observed: 16 in Li 2019 (approximate dataset median of the popPK cohort); 12 in Williams 2016 (median of DAS28cfb PK/PD dataset).
• Source aliases:
  • SWOL – used in Li_2019_abatacept.R (Li 2019 Methods abbreviation).
  • SJ28 – used in Williams_2016_rituximab_das28cfb.R (Williams 2016 Supplemental Methods g_i covariate function).
• Example models: Li_2019_abatacept.R (power effect on CL with exponent 0.0965; not clinically relevant per Li 2019), Williams_2016_rituximab_das28cfb.R (additive log-scale effect (SWOL_28JOINT - 12) on each of PMAX, kp, and Emax in the DAS28 change-from-baseline model).
• Notes: The _28JOINT suffix distinguishes this from the 66-joint swollen count (SWOL_66JOINT) used in the extended ACR joint set. Canonical name drops the BL prefix to match the SCORE_EASI / AGE / WT / ALB convention where baseline-vs-time-varying status is documented in covariateData notes rather than the column name.

TEND_28JOINT (canonical for 28-joint tender joint count)

• Description: Tender joint count on the 28-joint (DAS28) scale (integer 0-28; component of the DAS28 composite). Baseline value is typical; document time-varying use in per-model notes.
• Units: count (0-28)
• Type: continuous
• Scope: general
• Reference category: n/a – used with additive shift (TEND_28JOINT - <ref>) in log-scale coefficient models or shifted power ((TEND_28JOINT + 1)/(<ref> + 1))^exponent to avoid the zero-count edge case. Reference value observed: 16 in Williams 2016 (median of DAS28cfb PK/PD dataset).
• Source aliases:
  • TJ28 – used in Williams_2016_rituximab_das28cfb.R (Williams 2016 Supplemental Methods g_i covariate function).
• Example models: Williams_2016_rituximab_das28cfb.R (additive log-scale effect (TEND_28JOINT - 16) on each of PMAX, kp, and Emax in the DAS28 change-from-baseline model).
• Notes: Companion to SWOL_28JOINT — the tender-count analog of the swollen count on the 28-joint DAS28 subscale. Distinct from TEND_68JOINT which uses the 68-joint extended count.

SWOL_66JOINT (canonical for 66-joint swollen joint count)

• Description: Swollen joint count on the 66-joint extended (ACR) scale (integer 0-66; component of the ACR responder-rate criteria alongside TEND_68JOINT). Baseline value is typical; document time-varying use in per-model notes.
• Units: count (0-66)
• Type: continuous
• Scope: general
• Reference category: n/a – used with additive shift (SWOL_66JOINT - <ref>) in log-scale coefficient models. Reference value observed: 16 in Williams 2016 (median of the ACR dataset).
• Source aliases:
  • SJ66 – used in Williams_2016_rituximab_acr.R (Williams 2016 Supplemental Methods g_i covariate function).
• Example models: Williams_2016_rituximab_acr.R (additive log-scale effect (SWOL_66JOINT - 16) on PMAX and onset half-life in the cumulative-probit ACR responder-rate model).
• Notes: Register-anticipated: the existing SWOL_28JOINT entry’s Notes section explicitly reserves SWOL_66JOINT for the extended-joint scale used by ACR. Companion to TEND_68JOINT.

TEND_68JOINT (canonical for 68-joint tender joint count)

• Description: Tender joint count on the 68-joint extended (ACR) scale (integer 0-68; component of the ACR responder-rate criteria alongside SWOL_66JOINT). Baseline value is typical; document time-varying use in per-model notes.
• Units: count (0-68)
• Type: continuous
• Scope: general
• Reference category: n/a – used with additive shift (TEND_68JOINT - <ref>) in log-scale coefficient models. Reference value observed: 24 in Williams 2016 (median of the ACR dataset).
• Source aliases:
  • TJ68 – used in Williams_2016_rituximab_acr.R (Williams 2016 Supplemental Methods g_i covariate function).
• Example models: Williams_2016_rituximab_acr.R (additive log-scale effect (TEND_68JOINT - 24) on PMAX and onset half-life in the cumulative-probit ACR responder-rate model).
• Notes: Companion to SWOL_66JOINT. Distinct from TEND_28JOINT (28-joint DAS28 subscale).

PGA_PT (canonical for baseline patient’s global assessment of arthritis (VAS))

• Description: Patient’s own overall rating of arthritis disease activity on a 100-mm visual analogue scale (0 = no disease activity, 100 = maximum). Distinct from BLPHYVAS (the physician’s global assessment) and from PAIN (the patient’s pain-specific rating). Time-fixed per subject in the known example; document time-varying use in per-model notes.
• Units: mm (0-100 VAS)
• Type: continuous
• Scope: general
• Reference category: n/a – used with additive shift (PGA_PT - <ref>) in log-scale coefficient models or as a power term (PGA_PT / <ref>)^exponent. Reference value observed: 70 in Williams 2016 (median of both the DAS28cfb and ACR datasets).
• Source aliases:
  • PGA – used in Williams_2016_rituximab_das28cfb.R and Williams_2016_rituximab_acr.R (Williams 2016 Supplemental Methods g_i covariate function; the paper writes “PGA” for the patient’s global assessment).
• Example models: Williams_2016_rituximab_das28cfb.R (additive log-scale effect on each of PMAX, kp, and Emax), Williams_2016_rituximab_acr.R (additive log-scale effect on PMAX and onset half-life in the cumulative-probit ACR model).
• Notes: The _PT suffix disambiguates the patient’s global assessment from the physician’s global assessment (BLPHYVAS) — the paper-verbatim “PGA” abbreviation is ambiguous between the two in cross-literature usage.

CONMED_ANTILYMPHOCYTE (canonical for antilymphocyte agent use indicator)

• Description: 1 = the subject received an antilymphocyte agent (a class-level composite covering anti-thymocyte globulin, antilymphocyte globulin, alemtuzumab and equivalent lymphocyte-depleting antibodies) at any point recorded by the source analysis, 0 = no antilymphocyte agent recorded. Class-level composite in the spirit of CONMED_IMMUNOMOD and CONMED_STEROID, used when the source reports exposure to the class rather than to a named agent.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no antilymphocyte agent use).
• Source aliases:
  • Antilymphocyte use – printed row label in Sun 2025 Table S2 and Figure S3 panel m; Antilymphocyte use vs. none in Table S4.
• Example models: Sun_2025_maribavir_thrombocytopenia.R (coefficient -1.30, OR 0.271, 95% CI 0.101-0.730, p = 0.0098 – antilymphocyte exposure is associated with LOWER treatment-emergent thrombocytopenia risk in this cohort).
• Notes: Overlaps with, but is not the same column as, the HCT_TCD_* set. HCT_TCD_ALEMTUZUMAB / HCT_TCD_ATG record the graft T-cell-depletion PROTOCOL for the current transplant – a property of how the transplant was prepared – whereas CONMED_ANTILYMPHOCYTE records antilymphocyte drug exposure as a concomitant-medication fact. Sun 2025 reports both, with different denominators (108 of 238 subjects, 45.4%, for antilymphocyte use, against 83 of 238 across the in-vivo HCT_TCD_* levels), and fits them in different endpoint models – so they must not be aliased even though they are strongly correlated. The composite cannot be decomposed to named agents from the source, which reports only the class; a paper naming the agent should use CONMED_<INN> (e.g. CONMED_ALEMTUZUMAB) instead. Scope: specific.

CYP2A6 (canonical for CYP2A6 individual metabolic-activity score)

• Description: Continuous individual-level CYP2A6 metabolic-activity score. Same intent and documentation policy as the CYP2D6, CYP3A4, CYP2C19 and CYP2B6 continuous canonicals: one canonical column for any CYP2A6 phenotype proxy the source paper reports as a continuous number (probe-substrate metabolic ratio, model-based individual clearance, percent-of-reference enzyme activity, copy-number-corrected expression score, or an activity-score sum from *allele genotypes), with the per-model covariateData[[CYP2A6]]$units, description and notes recording which proxy is in force and the reference value the model centers on. CYP2A6 is the principal enzyme in nicotine C-oxidation to cotinine, so this column is the natural metabolic covariate for nicotine and coumarin popPK models. Time-invariant in the known example (a single baseline determination).
• Units: Paper-specific – document per-model (e.g., percent in Marchand 2017, where the value is baseline CYP2A6 activity expressed as a percentage).
• Type: continuous
• Scope: general
• Reference category: n/a (continuous). Models center on a population-typical value (29.2 percent in Marchand 2017); document the reference value per-model in covariateData[[CYP2A6]]$notes.
• Source aliases:
  • CYP2A6 activity – Marchand 2017 (paper Table 2 baseline covariate summary, Suppl. Table 1 primary-covariate rationale “CYP2A6 is the major CYP enzyme involved in nicotine metabolism”, and Table 4 covariate equations). Reported as a percentage; learning-dataset mean 31.5 (SD 18.2) percent.
• Example models: Marchand_2017_nicotine.R (two power-law effects centered on 29.2 percent: (CYP2A6 / 29.2)^0.322 on apparent nicotine clearance, so a doubling of CYP2A6 activity raises Cl/F by 25 percent; and (CYP2A6 / 29.2)^-0.401 on the background baseline nicotine concentration C0, so a doubling lowers C0 by 24 percent – Marchand 2017 Table 4).
• Notes: Fulfils part of the prospective-registration TODO recorded under CYP3A4 for the drug-metabolizing-CYP set (CYP1A2, CYP2A6, CYP2B6, CYP2C8, CYP2C9, CYP2C19, CYP2E1, CYP3A5), following the same continuous-individual-activity-score pattern that CYP2B6 used when it filled its own slot. A future paper reporting only discrete CYP2A6 phenotype labels (poor / intermediate / normal metabolizer) should register companion binary indicators on the CYP2D6_PM / CYP2C9_EM model rather than coercing labels into this column; a paper offering both should prefer this continuous column, per the reasoning recorded under CYP2D6. Note the opposite-sign pairing in the founding example: higher CYP2A6 activity raises clearance and therefore lowers the carried-over background concentration, which is a coherence check worth repeating when a future model puts this covariate on two parameters at once.

CYP2D6 (canonical for CYP2D6 individual metabolic-activity score)

• Description: Continuous individual-level CYP2D6 metabolic-activity score. The intent is a single canonical column for any CYP2D6 phenotype proxy that the source paper reports as a continuous number (probe-substrate model-based individual clearance, copy-number-corrected expression score, activity-score sum from *allele genotypes, etc.); the per-model covariateData[[CYP2D6]]$units, description, and notes document which proxy is in force and the population-median reference value used inside the model. Time-invariant in all known examples (germline genotype or one-time probe-substrate measurement).
• Units: Paper-specific – document per-model (e.g., ng/L in Ter Heine 2014 where the value is the dextromethorphan-probe model-based individual CYP2D6 clearance).
• Type: continuous
• Scope: general
• Reference category: n/a (continuous). Models center on a population median (e.g., 1560 ng/L in Ter Heine 2014); document the reference value per-model in covariateData[[CYP2D6]]$notes.
• Source aliases:
  • CYP2D6 – used directly in TerHeine_2014_tamoxifen.R.
  • AS (CYP2D6 activity score) – Ashraf 2024 (paper Section 3.2 and Table 1 footnote a). The CPIC consensus sum of per-allele activity values for the subject’s diplotype: 1 per normal-function allele (1, 2), 0.5 per decreased-function allele (9, 17, 29, 41), 0.25 per 10 allele, 0 per no-function allele (3, 4, 5, 6, 40), with duplicated normal-function alleles scoring 2.25, 3 or 4 depending on the partner allele. Observed values 0, 0.25, 0.5, 0.75, 1, 1.25, 1.5, 2, 2.25, 3, 4. No value transformation is needed – the activity score is used directly as the column value.
• Example models: TerHeine_2014_tamoxifen.R (power-law effect on the tamoxifen -> endoxifen formation clearance: (CYP2D6 / 1560)^e_CYP2D6_cl_endx), Ashraf_2024_codeine.R (exponential effect on the codeine -> morphine metabolic fraction referenced to activity score 2: f = exp(lfm_mor + eta + e_cyp2d6_fm_mor * (CYP2D6 - 2)) followed by the paper’s f / (1 + f) rescaling; e_cyp2d6_fm_mor = 1.00), Grzegorzewski_2022_dextromethorphan_pbpk.R (activity score 0-4 scaling a mechanistic Michaelis-Menten reaction rather than a clearance: hepatic CYP2D6 Vmax * CYP2D6 and Km * CYP2D6^-0.4, with CYP2D6 = 0 removing the pathway; no reference value).
• Notes: The activity-score-sum encoding anticipated by the original TODO below is now in use (Ashraf_2024_codeine.R), confirming that the single continuous canonical spans both a probe-substrate clearance (Ter Heine 2014, ng/L) and a genotype-derived CPIC activity score (Ashraf 2024, unitless 0-4); the per-model units and notes fields carry which proxy is in force, so no separate CYP2D6_ACTSCORE canonical is needed. Papers that report only the categorical PM / IM / NM / UM grouping still use the binary CYP2D6_PM / CYP2D6_EM indicators below rather than this column – Ashraf 2024 fitted both parameterisations and retained the continuous activity score because it explained more of the between-subject variability in the codeine-to-morphine metabolic ratio (33% unexplained vs 45% for the phenotype classes), so a paper that offers both should prefer the continuous column. A CYP2D6_PHENO_GROUP categorical companion remains unregistered and should be proposed when a future model genuinely needs the four-level grouping as a model term. The companion binary CYP2D6_PM entry below covers source papers that report only a poor-metabolizer-versus-not-poor-metabolizer dichotomy.

CYP2D6_PM (canonical for CYP2D6 poor-metabolizer phenotype indicator)

• Description: 1 = subject is a CYP2D6 poor metabolizer (genotype encoding no functional enzyme activity; e.g., homozygous 4/4, 5/5, or any compound combination of nonfunctional alleles); 0 = subject is an extensive, intermediate, or ultrarapid metabolizer (i.e., carrying at least one functional or reduced-function but not null allele). Time-fixed per subject (germline genotype-derived phenotype).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (extensive, intermediate, or ultrarapid metabolizer; carries at least one functional or reduced-function CYP2D6 allele).
• Source aliases:
  • 2D6PM – used in Knights_2015_aripiprazole.R (Knights 2015 Eq. 2 binary 2D6PM = 1 if CYP2D6 poor metabolizer, = 0 otherwise; in the supplement NONMEM control stream the dataset column CYP2D6EM is recoded as PM = (CYP2D6EM == 0) – i.e., the source dataset’s CYP2D6EM extensive-metabolizer indicator is value-inverted to produce the model’s PM indicator).
  • P1 (PM) / “poor metabolizer” – Sherwin 2012 (paper Table 2 mixture-model subpopulation P1; Table 1 phenotype assignment for genotyped subjects). Sherwin 2012 uses CYP2D6_PM paired with CYP2D6_EM to encode all three PM / IM / EM phenotypes; IM is the implicit reference (both indicators = 0).
• Example models: Knights_2015_aripiprazole.R (proportional-shift effect on apparent oral clearance: CL/F = TVCL * (1 + e_2d6pm_cl * CYP2D6_PM) with e_2d6pm_cl = -0.478; CYP2D6 poor metabolizers have 47.8% lower apparent oral clearance than non-poor-metabolizers; Knights 2015 Equation 2 and Figure 1B), Sherwin_2012_risperidone.R (phenotype-indicator covariate gating subpopulation-specific apparent oral clearance and metabolite-formation fraction in a one-compartment risperidone + (+/-)-9-hydroxyrisperidone mixture model: PM CL/F = 9.38 L/h, IM CL/F = 29.2 L/h, EM CL/F = 37.4 L/h at the 70 kg allometric reference; KF (fraction metabolized) PM = 0.16, EM = 0.13, IM = 1 fixed; Sherwin 2012 Table 2).
• Notes: Opposite-orientation companion to the CYP2C9_EM precedent (which encodes the extensive-metabolizer-equals-1 phenotype) – here the 1 indicates the poor-metabolizer end of the phenotype spectrum because the Knights 2015 source paper explicitly defines its binary as 2D6PM (PM = 1) and reports the coefficient sign relative to that orientation; preserving the PM-equals-1 orientation reproduces the paper’s reported coefficient sign and typical-value parameters directly. Distinct from the continuous CYP2D6 activity-score canonical above: use CYP2D6_PM when the source paper reports a categorical phenotype (PM / IM / EM); use CYP2D6 when the source reports a probe-derived continuous activity number. For three-level PM / IM / EM phenotype encoding (Sherwin 2012 mixture model), pair CYP2D6_PM with the companion CYP2D6_EM below following the SLCO1B1_HAP15_HET / SLCO1B1_HAP15_HOM two-binary-indicator pattern with IM as the implicit reference (both indicators = 0); this is consistent with the CYP2C9_EM register entry’s anticipated CYP2C9_PM companion. The Knights 2015 source dataset records CYP2D6EM (extensive-metabolizer indicator, 1 = EM, 0 = not-EM); the model’s PM indicator is derived by value inversion (CYP2D6_PM = 1 - CYP2D6EM after handling the supplement’s -99 missing-value sentinel as CYP2D6_PM = 0).

CYP2D6_EM (canonical for CYP2D6 extensive-metabolizer phenotype indicator)

• Description: 1 = subject is a CYP2D6 extensive metabolizer (genotype encoding full or near-full enzyme activity; e.g., 1/1 wild-type homozygote, or one functional allele paired with another functional allele); 0 = subject is an intermediate, poor, or ultrarapid metabolizer. Time-fixed per subject (germline genotype-derived phenotype).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (intermediate, poor, or ultrarapid metabolizer). When CYP2D6_EM is paired with CYP2D6_PM (Sherwin 2012 three-level encoding), the reference category specifically corresponds to the intermediate-metabolizer (IM) stratum: both CYP2D6_PM = 0 and CYP2D6_EM = 0 together indicate an IM subject.
• Source aliases:
  • P2 (EM) / “extensive metabolizer” – Sherwin 2012 (paper Table 2 mixture-model subpopulation P2; Table 1 phenotype assignment for genotyped subjects).
• Example models: Sherwin_2012_risperidone.R (phenotype-indicator covariate gating subpopulation-specific apparent oral clearance and metabolite-formation fraction in a one-compartment risperidone + (+/-)-9-hydroxyrisperidone mixture model; paired with CYP2D6_PM with IM as the implicit reference; Sherwin 2012 Table 2).
• Notes: Companion to the CYP2D6_PM canonical above; the paired-indicator pattern follows SLCO1B1_HAP15_HET / SLCO1B1_HAP15_HOM (two binary indicators encoding a three-level categorical with an implicit reference). When a source paper reports only a binary PM-vs-non-PM dichotomy (e.g., Knights 2015), CYP2D6_PM alone is sufficient and CYP2D6_EM is omitted from the model’s covariateData. When a source distinguishes all three PM / IM / EM phenotypes (Sherwin 2012), include both indicators with IM as the implicit reference. UM (ultrarapid metabolizer) – when present in a source cohort – must be handled by the paired CYP2D6_UM indicator registered below on the same pattern; note in particular that a source which POOLS extensive with ultrarapid metabolizers (Mao 2012) still needs CYP2D6_UM as a separate column, because CYP2D6_EM is defined as 0 for ultrarapid metabolizers and setting it to 1 for a URM subject would contradict this entry. Distinct from the continuous CYP2D6 activity-score canonical above and from CYP2D6_PM: use CYP2D6_EM when the source separately identifies the EM stratum (vs lumping IM + EM into a single non-PM group).

CYP2D6_IM (canonical for CYP2D6 intermediate-metabolizer phenotype indicator)

• Description: 1 = subject is a CYP2D6 intermediate metabolizer (genotype encoding reduced but non-zero enzyme activity; e.g. one non-functional allele paired with one reduced-function allele, or two reduced-function alleles); 0 = subject is an extensive, ultrarapid, or poor metabolizer. Time-fixed per subject (germline genotype-derived phenotype). Designed to be paired with CYP2D6_PM so that both indicators = 0 selects the POOLED extensive-plus-ultrarapid reference group.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0. When paired with CYP2D6_PM (the Nguyen 2025 two-indicator encoding), CYP2D6_PM = 0 and CYP2D6_IM = 0 together indicate an extensive OR ultrarapid metabolizer – the pooled reference against which both proportional shifts are reported.
• Source aliases:
  • IM – used in Nguyen_2025_valbenazine.R (Nguyen 2025 Supplemental Table S2 equation (5) binary IM; Table 1 theta 18 row label “Intermediate metabolizer on CLM”).
• Example models: Nguyen_2025_valbenazine.R (proportional-shift effect on the [+]-alpha-HTBZ metabolite clearance: CLM = 31.0 * (1 + e_cyp2d6_pm_cl_htbz * CYP2D6_PM) * (1 + e_cyp2d6_im_cl_htbz * CYP2D6_IM) with e_cyp2d6_im_cl_htbz = -0.281; intermediate metabolizers have 28.1% lower metabolite clearance and a 1.39-fold higher steady-state metabolite AUC than the pooled extensive-or-ultrarapid reference; Nguyen 2025 Table 1 and Supplemental Table S2).
• Notes: Third member of the CYP2D6_<phenotype> paired-binary-indicator family (CYP2D6_PM, CYP2D6_EM), following the pattern those entries document and the CYP2D6_EM entry explicitly anticipates for additional strata (“UM… must be handled by registering an additional paired CYP2D6_UM indicator on the same pattern”). Required rather than optional for the Nguyen 2025 encoding: reusing CYP2D6_EM inverted would be WRONG, because CYP2D6_EM = 0 also captures ultrarapid metabolizers and would therefore apply the intermediate-metabolizer clearance reduction to them, whereas the source paper pools ultrarapid WITH extensive in the reference group. Which pairing to use depends on how the source stratifies: CYP2D6_PM + CYP2D6_EM when IM is the implicit reference (Sherwin 2012), CYP2D6_PM + CYP2D6_IM when EM + UM is the pooled reference (Nguyen 2025). A source that resolves all four phenotypes will additionally need CYP2D6_UM (still unregistered; no model requires it yet). Note the assignment convention Nguyen 2025 Table 1 footnote a records: subjects whose genotype was inconclusive between intermediate and extensive were assigned to the INTERMEDIATE category, so CYP2D6_IM = 1 for those subjects. Distinct from the continuous CYP2D6 activity-score canonical, which is the right choice when a source reports a probe-derived continuous activity number rather than a categorical phenotype.

CYP2D6_UM (canonical for CYP2D6 ultrarapid-metabolizer phenotype indicator)

• Description: 1 = subject is a CYP2D6 ultrarapid metabolizer (genotype encoding greater-than-normal enzyme activity, arising from gene duplication or multiplication of a functional allele; e.g. *1/*1xN or *1/*2xN); 0 = subject is an extensive, intermediate, or poor metabolizer. Time-fixed per subject (germline genotype-derived phenotype).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (extensive, intermediate, or poor metabolizer). As with the other members of this family, what the all-indicators-zero state means depends on which siblings the model carries: paired with CYP2D6_EM and CYP2D6_PM alone (Mao 2012), all-zero is whatever stratum the source leaves as its intercept – in Mao 2012 that is the “genotype not collected” group rather than a phenotype, which the model’s covariateData notes must state explicitly.
• Source aliases:
  • URM – used in Mao_2012_vernakalant.R (Mao 2012 Table I CYP2D6 genotype rows and Table III “EM/URM” clearance stratum). Note that “URM” and “UM” are used interchangeably in the literature for this phenotype; the canonical follows the CYP2C19_UM / UGT2B17_UM spelling anticipated by the sibling entries.
• Example models: Mao_2012_vernakalant.R (carried as a distinct column even though Mao 2012 POOLS ultrarapid with extensive metabolizers in its patient clearance stratum – the covariate equation applies theta6 = 0.156 to (Patient & EM/URM), so CYP2D6_UM enters clearance only through the pooled term (1 - DIS_HEALTHY) * (CYP2D6_EM + CYP2D6_UM) and carries no coefficient of its own. It cannot be folded into CYP2D6_EM because that canonical is defined as 0 for ultrarapid metabolizers. All 4 URM subjects (1.3% of the 311 genotyped) were ACT IV patients; no healthy volunteer was a URM, so the paper’s healthy-metabolizer coefficient theta5 is defined for EM only).
• Notes: Fourth member of the CYP2D6_<phenotype> paired-binary-indicator family (CYP2D6_PM, CYP2D6_EM, CYP2D6_IM). Registration was pre-authorized by the CYP2D6_EM entry, which states that “UM (ultrarapid metabolizer) – when present in a source cohort – must be handled by registering an additional paired CYP2D6_UM indicator on the same pattern; this is not yet needed for any existing model and is deferred until a UM-aware paper is extracted.” Mao 2012 is that UM-aware paper. Which subset of the family a model carries depends on how the source stratifies: CYP2D6_PM alone for a binary PM-vs-non-PM dichotomy (Knights 2015), CYP2D6_PM + CYP2D6_EM with IM as the implicit reference (Sherwin 2012), CYP2D6_PM + CYP2D6_IM when EM + UM is the pooled reference (Nguyen 2025), and CYP2D6_PM + CYP2D6_EM + CYP2D6_UM when the source resolves the ultrarapid stratum separately or pools it with EM against a third reference group (Mao 2012). Adding CYP2D6_UM alongside CYP2D6_EM is what lets a model reproduce an EM+UM pooled effect without violating either canonical’s definition. Distinct from the continuous CYP2D6 activity-score canonical, which is the right choice when a source reports a probe-derived continuous activity number (an ultrarapid metabolizer typically maps to an activity score of 2 or above under the CPIC mapping) rather than a categorical phenotype.

CYP2D6_STAR10_HET (**canonical for CYP2D6*10 (rs1065852) heterozygote indicator**)

• Description: Binary genotype indicator for the CYP2D610 heterozygote group at rs1065852 (the c.100C>T transition in exon 1, P34S substitution). 1 = subject carries exactly one CYP2D610 (T) allele (genotype C/T); 0 = otherwise (the union of C/C wild-type homozygotes and T/T homozygous 10 carriers; the paired indicator CYP2D6_STAR10_HOM flags the T/T homozygous-mutant group). Time-fixed per subject (germline genotype). Distinct from the broader CYP2D6_PM / CYP2D6_EM phenotype canonicals above: CYP2D6_PM / CYP2D6_EM summarise the overall metabolic phenotype across all assayed CYP2D6 alleles, while CYP2D6_STAR10_HET / CYP2D6_STAR10_HOM resolve just the rs1065852 SNP and are the right canonical when the source paper genotypes only the 10 variant.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (C/C wild-type at rs1065852, when paired with CYP2D6_STAR10_HOM = 0). The reference group is the C/C wild-type stratum; CYP2D6_STAR10_HOM flags the T/T homozygous-mutant stratum.
• Source aliases:
  • CYP2D6*10 C/T – used in Pei_2016_iloperidone.R (Pei 2016 Results ‘Population pharmacokinetic (PPK) models’ encodes CYP2D6*10 C/C, C/T, T/T as integers 1, 2, 3; the C/T = 2 stratum maps to CYP2D6_STAR10_HET = 1).
• Example models: Pei_2016_iloperidone.R (multiplicative effect on the iloperidone -> M2 (P-95) formation rate constant K24: K24_CT = 0.693 * K24_typical relative to the C/C wild-type reference; Pei 2016 Table 3 and Results ‘Population pharmacokinetic (PPK) models’ equation (K24)_i = theta * 0.00649 * exp(eta_i) with theta = 0.693 for C/T).
• Notes: Follows the CYP3A5_STAR1_HET / CYP3A5_STAR1_HOM and SLCO1B1_HAP15_HET / SLCO1B1_HAP15_HOM precedents (paired binary indicators for a three-level germline genotype with an implicit wild-type reference). The _STAR10_ token names the variant-allele orientation (the *10 allele is the reduced-function variant of CYP2D6 that causes lower CYP2D6 metabolic activity); the C/C wild-type stratum is the implicit reference because the source-paper coefficient signs are reported relative to that reference. Pei 2016 pooled C/C and C/T as the reference for the K23 (M1 formation) effect (only T/T was distinguished on K23), so a model using these two canonicals will route the K23 effect through CYP2D6_STAR10_HOM alone while routing the K24 effect through both CYP2D6_STAR10_HET and CYP2D6_STAR10_HOM – this asymmetric per-parameter use is normal and is what the paired-indicator pattern is designed to express.

CYP2D6_STAR10_HOM (**canonical for CYP2D6*10 (rs1065852) homozygous-mutant indicator**)

• Description: Binary genotype indicator for the CYP2D610 homozygous-mutant group at rs1065852. 1 = subject carries two CYP2D610 (T) alleles (genotype T/T); 0 = otherwise (the union of C/C wild-type homozygotes and C/T heterozygotes; the paired indicator CYP2D6_STAR10_HET flags the C/T heterozygous group). Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (C/C wild-type at rs1065852, when paired with CYP2D6_STAR10_HET = 0). The reference group is the C/C wild-type stratum; CYP2D6_STAR10_HET flags the C/T heterozygous stratum.
• Source aliases:
  • CYP2D6*10 T/T – used in Pei_2016_iloperidone.R (Pei 2016 Results ‘Population pharmacokinetic (PPK) models’ encodes CYP2D6*10 C/C, C/T, T/T as integers 1, 2, 3; the T/T = 3 stratum maps to CYP2D6_STAR10_HOM = 1).
• Example models: Pei_2016_iloperidone.R (multiplicative effects on two iloperidone metabolite-formation rate constants: K23 (M1 formation) is multiplied by 1.34 in T/T subjects relative to the C/C + C/T pooled reference, and K24 (M2 formation) is multiplied by 0.492 in T/T subjects relative to the C/C wild-type reference; Pei 2016 Table 3 and Results ‘Population pharmacokinetic (PPK) models’ equations (K23)_i = theta * 0.00451 * exp(eta_i) with theta = 1.34 for T/T and (K24)_i = theta * 0.00649 * exp(eta_i) with theta = 0.492 for T/T).
• Notes: Paired with CYP2D6_STAR10_HET (see that entry’s Notes for the three-level decomposition rationale, the variant-orientation convention, and the asymmetric K23-vs-K24 use the paired indicators support in the Pei 2016 model). Allele frequencies for the CYP2D6*10 (rs1065852) variant in Chinese populations are 48-70% per the source paper’s Introduction; the Pei 2016 cohort (n = 70 Chinese schizophrenia patients) observed C/C 15.7%, C/T 60.0%, T/T 24.3%.

CYP3A4 (canonical for CYP3A4 / CYP3A4-and-CYP3A5 individual metabolic-activity score)

• Description: Continuous individual-level CYP3A4 (or combined CYP3A4 + CYP3A5) metabolic-activity score. Same intent and documentation policy as CYP2D6 above. Some sources measure CYP3A4 alone via probe substrate; others (Ter Heine 2014) report a combined CYP3A4/5 activity because the chosen probe (dextromethorphan N-demethylation) cannot distinguish CYP3A4 from CYP3A5. The per-model notes field documents whether the value is CYP3A4-only or the CYP3A4 + CYP3A5 combined score.
• Units: Paper-specific – document per-model (e.g., ng/L in Ter Heine 2014 where the value is the dextromethorphan-probe model-based individual CYP3A4/5 clearance).
• Type: continuous
• Scope: general
• Reference category: n/a (continuous). Models center on a population median (e.g., 44.7 ng/L in Ter Heine 2014); document the reference value per-model.
• Source aliases:
  • CYP3A4 – used directly in TerHeine_2014_tamoxifen.R (the column carries combined CYP3A4 + CYP3A5 activity per the source paper).
  • CYP3A4/5 – long form sometimes used in source manuscripts; standardize the column name to CYP3A4 and document the combined-isoform semantics in per-model notes.
• Example models: TerHeine_2014_tamoxifen.R (power-law effect on the tamoxifen -> endoxifen formation clearance: (CYP3A4 / 44.7)^e_CYP3A4_cl_endx).
• Notes: TODO – register the rest of the canonical drug-metabolizing-CYP set prospectively (CYP1A2, CYP2A6, CYP2B6, CYP2C8, CYP2C9, CYP2C19, CYP2E1, CYP3A5) using the same continuous-individual-activity-score pattern, so future popPK models that report CYP-probe-derived covariates can drop straight into the existing convention rather than re-deliberating the encoding each time. Coordinate with the consolidation TODO on CYP2D6 so categorical-vs-continuous encoding is handled uniformly across all CYPs.

CYP3A5_EXPR (canonical for CYP3A5 expresser status)

• Description: 1 = subject carries at least one functional CYP3A51 allele (genotype 1/1 or 1/3, equivalent to one or two A alleles at rs776746); 0 = homozygous CYP3A53/*3 (G/G at rs776746) – i.e., a nonexpresser. Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (CYP3A53/3 nonexpresser).
• Source aliases:
  • X – used in Bergmann_2014_tacrolimus.R (Bergmann 2014 Table 2 footnote: X = 1 for 1/1 and 1/3; X = 0 for 3/3).
  • CYP3A5 – used in Chen_2020b_tacrolimus.R (Chen 2020 text below Table 1: CYP3A5 value = 0 for 3/3, 1 for a CYP3A5*1 allele; multiplicative effect 1.61^CYP3A5 on CL/F).
  • CYP3A5 expresser – used in Storset_2014_tacrolimus.R (Storset 2014 Table 2 final theory-based model; *1/*1 and *1/*3 pooled as expressers because Storset 2014 had only n = 3 1/1 subjects).
• Example models: Bergmann_2014_tacrolimus.R (multiplicative effect on tacrolimus CL/F: theta_CYP3A5 ^ CYP3A5_EXPR, with theta_CYP3A5 = 1.60; expressers have 60% higher apparent oral clearance than nonexpressers), Storset_2014_tacrolimus.R (multiplicative effects on apparent plasma clearance: cl *= 1.30^CYP3A5_EXPR; and on oral bioavailability: fdepot *= 0.82^CYP3A5_EXPR; Storset 2014 Table 2 final theory-based model), Chen_2021b_tacrolimus.R (multiplicative effect on tacrolimus CL/F: 1 + (1.29 - 1) * CYP3A5_EXPR; Chen 2021 Table 2).
• Notes: Distinct from the SNP-pattern canonical SNP_<GENE>_<RSID> (which encodes “mutant allele presence” – 1 = at least one variant allele). For CYP3A5 the 3 allele (rs776746 G) is the variant that abolishes function, so a literal “mutant-allele-presence” indicator (1 = any G allele) would group 1/3 heterozygotes with the 3/3 nonexpressers, which is the opposite of the clinically meaningful expresser-vs-nonexpresser dichotomy used by every CYP3A5-aware popPK model. The CYP3A5_EXPR canonical preserves the expresser-equals-1 orientation directly. Future CYP3A5 papers using a 3/3 indicator (rather than 1 carrier) should still record their values under CYP3A5_EXPR and document the value inversion in notes (CYP3A5_EXPR = 1 - source_indicator); registering a parallel CYP3A5_NONEXPR is discouraged. The canonical name follows the <gene>_<phenotype> rather than the <gene>_<rsid> pattern because the column captures derived metabolic phenotype rather than raw genotype. Distinct from CYP3A4 (continuous individual-activity score for CYP3A4 / CYP3A4 + CYP3A5 combined): the binary CYP3A5_EXPR is the right fit for source papers that report only the rs776746 genotype, while the continuous CYP3A4 is for sources that report a probe-substrate-derived activity number. In solid-organ-transplant popPK / pharmacogenetics studies that genotype both the recipient and the donated graft separately (e.g., Moes 2016 liver-transplant tacrolimus), the recipient genotype goes into CYP3A5_EXPR and the donor genotype into the sibling canonical CYP3A5_EXPR_DONOR below.

CYP3A5_EXPR_DONOR (canonical for transplanted-graft donor CYP3A5 expresser status)

• Description: 1 = transplanted graft (typically liver, kidney, or other CYP3A5-expressing organ) was donated by a donor carrying at least one functional CYP3A51 allele (genotype 1/1 or 1/3 at rs776746); 0 = donor is homozygous CYP3A53/*3 (nonexpresser). Time-fixed per recipient (germline genotype of the donor at the time of transplantation). Sibling canonical to the recipient-genotype indicator CYP3A5_EXPR.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (donor CYP3A53/3 nonexpresser graft).
• Source aliases:
  • Donor CYP3A5*3 – used in Moes_2016_tacrolimus.R (Moes 2016 Methods; donor and recipient genotypes pooled with the four-level combination indicator C1 / C2 / C3 / C4 defined in Methods, where CYP3A5_EXPR_DONOR = 1 corresponds to graft from a *1-carrying donor).
  • CYP3A5 donor – used in Ji_2018_tacrolimus.R (Ji 2018 derives the donor-side genotype from the four-level combinational CYP3A5 group: REDE/RNDE -> 1; REDN/RNDN -> 0).
• Example models: Moes_2016_tacrolimus.R (categorical donor + recipient CYP3A5 combination effect on apparent oral CL/F: reference C1 = both nonexpressers; C2 = recipient *1 carrier + donor nonexpresser = +33%; C3 = recipient nonexpresser + donor *1 carrier = +33%; C4 = both *1 carriers = +71%; Moes 2016 Table 4 final model – the C2 / C3 / C4 levels are reconstructed inside model() from the two binary inputs CYP3A5_EXPR and CYP3A5_EXPR_DONOR), Ji_2018_tacrolimus.R (combinational categorical effect on tacrolimus CL/F that depends on both recipient and donor CYP3A5 status: multiplier 2.314 when the recipient is an expresser and the donor is an expresser (REDE), 1.523 when the recipient is an expresser and the donor is a nonexpresser (REDN), and 1.0 otherwise (RNDE / RNDN reference, which Ji 2018 merged because the two estimated effects were similar)).
• Notes: Genotyped only in transplant studies where donor DNA is recoverable (Moes 2016 obtained donor DNA from spleen / liver biopsies). When the same paper reports both recipient and donor genotypes separately, encode them as two binary inputs (CYP3A5_EXPR for the recipient, CYP3A5_EXPR_DONOR for the donor) rather than as a single four-level combination column, so the underlying recipient / donor biology is explicit in the dataset. Models that fit a categorical four-level combination effect (paper-Moes-style C1 / C2 / C3 / C4) reconstruct the levels inside model() from the two binary inputs, so the source-paper’s per-level coefficients remain the estimated quantities. The intestinal-CYP3A5 contribution (recipient genotype) and the hepatic-CYP3A5 contribution (donor genotype) act on different anatomic compartments of tacrolimus’s first-pass metabolism, which is why both donor and recipient genotypes are independently informative in liver-transplant tacrolimus PK.

CYP3A5_STAR1_HET (canonical for CYP3A51/3 heterozygote indicator)

• Description: Binary genotype indicator for the CYP3A51/3 heterozygote group. 1 = subject carries exactly one functional CYP3A51 allele at rs776746 (genotype 1/3); 0 = otherwise (the union of 3/3 nonexpressers and 1/*1 homozygotes; the paired indicator CYP3A5_STAR1_HOM flags the homozygous-expresser group). Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (CYP3A53/3 nonexpresser, when paired with CYP3A5_STAR1_HOM = 0). The reference group is the homozygous 3/3 nonexpresser stratum; CYP3A5_STAR1_HOM flags the homozygous-expresser stratum.
• Source aliases:
  • CYP3A5*1/*3 – used directly in Passey_2011_tacrolimus.R (Passey 2011 Table 3 reports a separate multiplicative factor of 1.70 for the heterozygote stratum vs the 3/3 reference).
• Example models: Passey_2011_tacrolimus.R (power-of-binary-indicator multiplicative factor on apparent oral clearance: e_cyp3a5_het_cl ^ CYP3A5_STAR1_HET with e_cyp3a5_het_cl = 1.70; 1/3 heterozygotes have ~70% higher apparent oral CL/F than 3/3 nonexpressers; paired with CYP3A5_STAR1_HOM and used jointly), Pei_2023_tacrolimus.R (exponential form, heart transplant), Zhou_2025_tacrolimus.R (exponential form exp(e_cyp3a5_het_cl * CYP3A5_STAR1_HET) with e_cyp3a5_het_cl = 0.63, i.e. a 1.87-fold higher CL/F in 1/3 heterozygotes; the first lung-transplant cohort to resolve all three CYP3A5 strata – Zhou 2025 Discussion notes that the only prior lung-transplant estimate (Cai et al.) distinguished 1/3 from 3/3 at only 1.30-fold and did not resolve 1/1 at all).
• Notes: Follows the SLCO1B1_HAP15_HET / SLCO1B1_HAP15_HOM precedent (paired binary indicators for a three-level genotype) rather than overloading CYP3A5_EXPR (which pools 1/1 and 1/3 into a single expresser indicator and is the right canonical when a source paper does the same pooling). Use the paired CYP3A5_STAR1_HET + CYP3A5_STAR1_HOM binaries when the source paper assigns a distinct typical-value covariate effect to each of the three CYP3A5 genotype strata (3/3, 1/3, 1/1). Passey 2011 motivated the three-level decomposition because the cohort (n = 681) had enough 1/1 homozygotes (72, 11%) to identify a distinct typical-value factor for that stratum, whereas earlier popPK papers (Bergmann 2014, Storset 2014) had too few 1/1 subjects (or the equivalent expresser-pooled treatment) to distinguish 1/1 from 1/3. The _STAR1_ token (rather than _STAR3_) names the functional-allele-presence orientation consistent with the parent CYP3A5_EXPR canonical’s expresser-equals-1 convention.

CYP3A5_STAR1_HOM (canonical for CYP3A51/1 homozygote indicator)

• Description: Binary genotype indicator for the CYP3A51/1 homozygote group. 1 = subject carries two functional CYP3A51 alleles at rs776746 (genotype 1/1); 0 = otherwise (the union of 3/3 nonexpressers and 1/*3 heterozygotes; the paired indicator CYP3A5_STAR1_HET flags the heterozygous group). Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (CYP3A53/3 nonexpresser, when paired with CYP3A5_STAR1_HET = 0). The reference group is the homozygous 3/3 nonexpresser stratum; CYP3A5_STAR1_HET flags the heterozygous-expresser stratum.
• Source aliases:
  • CYP3A5*1/*1 – used directly in Passey_2011_tacrolimus.R (Passey 2011 Table 3 reports a separate multiplicative factor of 2.00 for the homozygote-expresser stratum vs the 3/3 reference).
• Example models: Passey_2011_tacrolimus.R (power-of-binary-indicator multiplicative factor on apparent oral clearance: e_cyp3a5_hom_cl ^ CYP3A5_STAR1_HOM with e_cyp3a5_hom_cl = 2.00; 1/1 homozygotes have 100% higher apparent oral CL/F than 3/3 nonexpressers; paired with CYP3A5_STAR1_HET and used jointly), Pei_2023_tacrolimus.R (exponential form, heart transplant), Zhou_2025_tacrolimus.R (exponential form exp(e_cyp3a5_hom_cl * CYP3A5_STAR1_HOM) with e_cyp3a5_hom_cl = 1.00, i.e. a 2.72-fold higher CL/F in 1/1 homozygotes; identified from only 13 of 142 lung-transplant recipients (9.15%), so the estimate is the least precisely determined of the paper’s covariate effects at RSE 9.23% on the log scale).
• Notes: Paired with CYP3A5_STAR1_HET (see that entry’s Notes for the three-level decomposition rationale).

CYP2C9_EM (canonical for CYP2C9 extensive-metabolizer phenotype indicator)

• Description: 1 = subject is a CYP2C9 extensive metabolizer (genotype 1/1; wild-type homozygote with full enzyme activity); 0 = subject is a CYP2C9 intermediate or poor metabolizer (heterozygous or homozygous carrier of any reduced-function allele such as 2, 3, *13) OR has unknown phenotype (when paired with CYP2C9_PM_IM). Time-fixed per subject (germline genotype-derived phenotype).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (intermediate or poor metabolizer, reduced-function-allele carrier; OR unknown phenotype when paired with CYP2C9_PM_IM).
• Source aliases:
  • CYP2C9 phenotype – Jeong 2022 (paper Table 1 phenotype classification: *1/*1 = EM; *1/*3, *1/*13 = IM).
• Example models: Jeong_2022_torsemide.R (linear-deviation effect on apparent clearance and inter-compartmental clearance: CL/F = tvCL/F * (1 + 0.510 * CYP2C9_EM) and Q/F = tvQ/F * (1 + 0.365 * CYP2C9_EM); CYP2C9 extensive metabolizers have 51% higher apparent clearance and 36.5% higher apparent inter-compartmental clearance than intermediate metabolizers; Jeong 2022 Table 4 final Pop-PK model), Kleideiter_2017_cebranopadol.R (additive log shifts on CL: e_cyp2c9em_lcl = log(82.4 / 74.3) = 0.1037 and e_cyp2c9pmim_lcl = log(58.7 / 74.3) = -0.2353; reference category is unknown phenotype with both indicators = 0, the most common stratum; paired with CYP2C9_PM_IM), Kleideiter_2018_cebranopadol.R (multiplicative effect on CL applied as e_em_cl^CYP2C9_EM with e_em_cl = 82.4 / 74.3 = 1.109; EM subjects have about +11% CL vs the model’s CYP2C9 reference, which in this paper is the ‘unknown phenotype’ pool rather than IM/PM – only 38.3% of the analysis cohort had a known CYP2C9 phenotype, so the 0-level here pools unknown subjects together with PIM subjects whose effect is carried separately by the sibling CYP2C9_PIM canonical; Kleideiter 2018 Table 13).
• Notes: Follows the CYP3A5_EXPR precedent above (functional-allele-carrier = 1) rather than the SNP_<GENE>_<RSID> mutant-presence pattern, because (a) clinical CYP2C9 phenotype is reported as EM / IM / PM and the canonical name should mirror the clinically meaningful axis, and (b) Jeong 2022 chose IM as the model reference, so the EM-equals-1 orientation preserves the paper’s reported coefficient signs and typical-value parameters directly. The canonical name pools reduced-function alleles (2, 3, *13, etc.) into the 0 category because the population-PK literature typically does not separately resolve them; future papers that distinguish IM from PM should propose a paired companion (CYP2C9_PM) so the three-level EM / IM / PM phenotype can be encoded with two binary indicators on the SLCO1B1_HAP15_HET / SLCO1B1_HAP15_HOM pattern. Distinct from CYP2D6 / CYP3A4 (continuous-activity-score canonicals): use CYP2C9_EM when the source paper reports a discrete phenotype label; use a continuous CYP2C9 canonical (not yet ratified; see TODO on CYP3A4) when a future paper reports a probe-derived activity number.

CYP2C9_PM_IM (canonical for pooled CYP2C9 poor-or-intermediate-metabolizer phenotype indicator)

• Description: 1 = subject is a CYP2C9 poor metabolizer OR intermediate metabolizer (i.e., heterozygous or homozygous carrier of a reduced-function allele such as 2, 3, *13, pooled because the source paper did not distinguish PM from IM); 0 = subject is an extensive metabolizer OR has an unknown / unassayed CYP2C9 phenotype. Time-fixed per subject (germline genotype-derived phenotype where known). The companion canonical CYP2C9_EM carries the 1 = EM indicator; subjects with unknown CYP2C9 status appear with both CYP2C9_PM_IM = 0 and CYP2C9_EM = 0.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (extensive metabolizer OR unknown phenotype; the per-paper reference cohort that 0 represents is paper-defined). When paired with CYP2C9_EM, both indicators = 0 indicates the unknown-phenotype stratum and CYP2C9_EM = 1 with CYP2C9_PM_IM = 0 indicates the EM stratum.
• Source aliases:
  • CYP2C9 – Kleideiter 2017 (paper Table 13 row “CYP2C9 poor and intermediate metabolizers 58.7 L/h”).
  • CYP2C9_PIM – used in Kleideiter_2018_cebranopadol.R (the erratum-corrected re-extraction; same pooled poor-or-intermediate-metabolizer semantics, abbreviated PIM).
• Example models: Kleideiter_2017_cebranopadol.R (additive log shift on CL: e_cyp2c9pmim_lcl = log(58.7 / 74.3) = -0.2353; reduced apparent clearance vs the unknown-phenotype reference; paired with CYP2C9_EM to form the three-level stratification), Kleideiter_2018_cebranopadol.R (multiplicative CL ratio 58.7 / 74.3 = 0.790 applied as ratio^CYP2C9_PIM; paired with CYP2C9_EM, both 0 = the unknown-phenotype reference, Kleideiter 2018 Table 13).
• Notes: Pairs with CYP2C9_EM for the three-level “unknown / EM / PM-IM” stratification used in the Kleideiter cebranopadol model, where the unknown-phenotype reference (both indicators = 0) is the most common category (only 38.3% of the analysis cohort had a known CYP2C9 phenotype). PM and IM are pooled because the source covariate analysis does not separately resolve them; downstream papers that distinguish PM from IM should register a paired CYP2C9_PM canonical and split this group, encoding the three-level EM / IM / PM phenotype with two binary indicators on the SLCO1B1_HAP15_HET / SLCO1B1_HAP15_HOM pattern. Distinct from the CYP2C9_EM canonical’s Jeong-2022 use (where 0 = IM/PM, no unknown subjects, and the 0-level directly carries the PM/IM phenotype): in the Kleideiter cohort the 0-level of both CYP2C9_EM and CYP2C9_PM_IM pools all subjects not assigned that specific phenotype label (including the ‘unknown’ fraction), so per-model covariateData notes must document the reference complement.

CYP2C9_IM_AS15 (canonical for CYP2C9 intermediate-metabolizer, CPIC activity score 1.5, phenotype indicator)

• Description: 1 = subject is a CYP2C9 intermediate metabolizer whose CPIC genotype-predicted activity score (AS) is 1.5 – i.e. one normal-function allele plus one decreased-function allele, canonically *1/*2; 0 = any other CYP2C9 phenotype. Time-fixed per subject (germline genotype-derived phenotype). Pairs with CYP2C9_PM_IM_AS10 as two mutually-exclusive binary indicators; both 0 identifies the extensive-metabolizer reference stratum (AS 2.0, *1/*1 and *1/*9).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0. When paired with CYP2C9_PM_IM_AS10, both indicators = 0 is the extensive-metabolizer (AS 2.0) reference stratum, CYP2C9_IM_AS15 = 1 is the AS-1.5 stratum, and CYP2C9_PM_IM_AS10 = 1 is the AS-1.0 stratum.
• Source aliases:
  • IM with AS of 1.5 – Chen 2024 (paper Table 2 phenotype classification row; Table 4 parameter row CL/F_IM_1.5).
• Example models: Chen_2024_noscapine.R (additive log shift on apparent clearance: e_cyp2c9im_as15_lcl = log(531 / 958) = -0.5901; Chen 2024 Table 4 estimates a free typical value per stratum – CL/F_EM = 958 L/h, CL/F_IM_1.5 = 531 L/h, CL/F_PM_IM_1.0 = 343 L/h – so the extensive-metabolizer value is carried by lcl and the other two strata enter as log shifts, the Kleideiter_2017_cebranopadol.R pattern).
• Notes: This pair resolves CYP2C9 phenotype on the activity-score axis rather than the bare PM / IM label, because that is the axis on which the source model stratifies and because the two axes do not coincide: CPIC assigns AS 1.0 to both *1/*3 (an intermediate metabolizer) and *3/*3 (a poor metabolizer), which is why the AS-1.0 stratum is named PM_IM rather than PM. Distinct from the pooled CYP2C9_PM_IM canonical, whose 1-level is the union of the AS-1.5 and AS-1.0 strata; a paper that resolves the activity score should use this pair, and a paper that does not should use CYP2C9_EM / CYP2C9_PM_IM. Do not mix the two pairs in one model. A continuous activity-score covariate is not a substitute: in Chen 2024 the clearance ratios (531/958 = 0.554, 343/958 = 0.358) are not proportional to the activity-score ratios (0.75, 0.50), so discrete indicators are required to reproduce the published typical values. Extend to further strata as CYP2C9_<PHENOTYPE>_AS<score> (e.g. CYP2C9_PM_AS00) if a future cohort resolves them.

CYP2C9_PM_IM_AS10 (canonical for CYP2C9 poor-or-intermediate-metabolizer, CPIC activity score 1.0, phenotype indicator)

• Description: 1 = subject’s CYP2C9 CPIC genotype-predicted activity score (AS) is 1.0, pooling the intermediate metabolizers that reach AS 1.0 (*1/*3, *2/*2) with the poor metabolizers (*2/*3, *3/*3); 0 = any other CYP2C9 phenotype. Time-fixed per subject (germline genotype-derived phenotype). Pairs with CYP2C9_IM_AS15 as two mutually-exclusive binary indicators; both 0 identifies the extensive-metabolizer reference stratum (AS 2.0, *1/*1 and *1/*9).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0. When paired with CYP2C9_IM_AS15, both indicators = 0 is the extensive-metabolizer (AS 2.0) reference stratum.
• Source aliases:
  • PM & IM with AS of 1 – Chen 2024 (paper Table 2 phenotype classification row; Table 4 parameter row CL/F_PM_IM_1.0).
• Example models: Chen_2024_noscapine.R (additive log shift on apparent clearance: e_cyp2c9pmim_as10_lcl = log(343 / 958) = -1.0271; paired with CYP2C9_IM_AS15, both 0 = the extensive-metabolizer reference carried by lcl; Chen 2024 Table 4).
• Notes: The PM_IM element of the name records that this stratum deliberately pools poor metabolizers with the activity-score-1.0 intermediate metabolizers – Chen 2024 Sect. 3.1: “CYP2C91/3 and CYP2C92/2 as IMs with an activity score (AS) of 1, were classified together with the PMs into one group in this study because of the limited sample size of homozygous carriers”. A source paper that separately estimates PM and AS-1.0 IM effects should register a further split rather than reuse this canonical. This is a strict subset of the pooled CYP2C9_PM_IM canonical (which additionally absorbs the AS-1.5 *1/*2 subjects that CYP2C9_IM_AS15 carries here), so the two must not be used interchangeably: substituting CYP2C9_PM_IM for this column would silently mis-assign every *1/*2 subject (20.8% of the Chen 2024 cohort). Derivable from the per-allele count canonicals as CYP2C9_S3_COUNT >= 1 | CYP2C9_S2_COUNT == 2, but the indicators are preferred when the source paper declares a phenotype group and estimates a free typical value per group rather than an allele-dosage function. See the CYP2C9_IM_AS15 Notes for the full rationale.

CYP2C9_RH (canonical for CYP2C9 reduced-hydroxylator phenotype indicator)

• Description: 1 = subject was assigned the “reduced hydroxylator” CYP2C9 phenotype by the source study’s phenotyping scheme; 0 = subject was assigned any other CYP2C9 phenotype in that scheme (in the founding paper: intermediate metabolizer OR normal hydroxylator). Time-fixed per subject. Records a hydroxylation-capacity phenotype LABEL as the source study reported it, and deliberately makes no star-allele or PM / IM equivalence claim.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (any non-reduced-hydroxylator CYP2C9 phenotype; the per-paper composition of the 0-level must be documented in covariateData[[CYP2C9_RH]]$notes, since which labels the source pools into the reference is paper-defined).
• Source aliases:
  • CYPC – Kastrissios 2006 equation 7 (“CYPC is an indicator variable taking the value 0 for subjects with the intermediate/normal hydroxylator CYP 2C9 phenotype and the value 1 for those with the reduced hydroxylator phenotype”); phenotype code 5 in the paper’s shared 1-6 code legend, reported in Table II as CYP 2C9, 2/5/6 = 45/16/43.
• Example models: Kastrissios_2006_apricoxib.R (founding example; exponential effect on apparent oral clearance exp(e_cyp2c9_rh_cl * CYP2C9_RH) with e_cyp2c9_rh_cl = -0.163, i.e. reduced hydroxylators have 1 - exp(-0.163) = 15.0% lower CL/F, matching Kastrissios 2006 Discussion “Apparent oral clearance was decreased by 15% in individuals expressing the reduced hydroxylator CYP 2C9 phenotype relative to intermediate and normal phenotypes”).
• Notes: CYP2C9_PM_IM is specifically WRONG here: Kastrissios 2006 pools its INTERMEDIATE metabolizers (n = 45 of 104) into the reference group alongside normal hydroxylators, so reusing CYP2C9_PM_IM would move those 45 subjects into the 1-level and apply the -0.163 clearance effect to them. An inverted CYP2C9_EM does not reproduce the paper’s reference stratum either. The register’s CYP2C9_EM and CYP2C9_PM_IM Notes both anticipate a companion named CYP2C9_PM, and that slot was considered and deliberately NOT used: “reduced hydroxylator” is a 2006-era hydroxylation-phenotype label, and its observed frequency of 15% (Table II 16/104; Table I quotes 0.15 for Western populations) is an order of magnitude above true CYP2C9 poor-metabolizer prevalence in white populations (~1-2%) and closer to reduced-function-allele carriage, so naming it CYP2C9_PM would assert an equivalence the paper never states and would collide semantically with the CPIC activity-score pair already ratified from Chen 2024. Reach for CYP2C9_RH when a source reports hydroxylator-capacity labels; use CYP2C9_EM / CYP2C9_PM_IM for PM / IM / EM labels and the CYP2C9_*_AS* pair when the source resolves a CPIC activity score. Do not mix the three axes in one model. Note a minor internal inconsistency in the founding paper that does not affect the encoding: its shared footnote legend codes CYP2C9 as “1 = extensive… 2 = intermediate… 5 = reduced hydroxylator, 6 = normal hydroxylator” and Table II reports levels 2/5/6 present, while Table I groups the same data as “(1&6, 5)”; either way the 1-level is unambiguously the level-5 reduced-hydroxylator group (n = 16, 15%) and the 0-level is everything else.

CYP2B6_IM (canonical for CYP2B6 intermediate-metabolizer phenotype indicator)

• Description: 1 = subject is a CYP2B6 intermediate metabolizer defined by the combined 516G>T (rs3745274) | 983T>C (rs28399499) SNP-vector phenotype assignment used in the African paediatric antiretroviral literature: 516GT | 983TT (heterozygous loss-of-function at 516) or 516GG | 983TC (heterozygous loss-of-function at 983); 0 = any other phenotype (extensive, slow, or ultra-slow metabolizer; see paired canonicals CYP2B6_SM and CYP2B6_USM). Time-fixed per subject (germline genotype-derived phenotype). EM (516GG | 983TT) is the reference category when all three of CYP2B6_IM, CYP2B6_SM, and CYP2B6_USM are 0.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (any non-IM phenotype: EM, SM, or USM).
• Source aliases:
  • metabolizer status IM / IM – Bienczak 2016 (paper Methods ‘Covariate effects’ paragraph 2, Results ‘Population pharmacokinetics’ paragraph 3, and Table 3 final estimates; categorical metabolizer indicator entered as multiplicative effect on intrinsic clearance CLint with EM reference).
• Example models: Bienczak_2016_nevirapine.R (multiplicative log-additive effect on CLint: cl_meta <- exp(e_cyp2b6_im_cl * CYP2B6_IM + e_cyp2b6_sm_cl * CYP2B6_SM + e_cyp2b6_usm_cl * CYP2B6_USM), with e_cyp2b6_im_cl = log(1 - 0.17) = -0.186, giving the 17% lower CLint reported in Bienczak 2016 Table 3 / Results ‘Population pharmacokinetics’ paragraph 3).
• Notes: Sibling indicators CYP2B6_SM and CYP2B6_USM together encode the four-level EM / IM / SM / USM phenotype with three binary columns (EM = all three zero); follows the dummy-coding pattern used elsewhere in the register for multi-level categoricals (e.g. RACE_*, HEPIMP_MILD / HEPIMP_SEV / HEPIMP_MODSEV). The IM grouping pools the two distinct genotypes (516GT | 983TT and 516GG | 983TC) into a single indicator because Bienczak 2016 Results ‘Population pharmacokinetics’ paragraph 3 reports that ‘Using six rather than four 516G>T | 983T>C SNP-vector metabolizer groups reduced OFV by only 5 points (df = 2, P = 0.08) and was therefore not used.’ Distinct from the continuous CYP3A4 activity-score canonical (which captures probe-substrate-derived activity rather than SNP-vector phenotype) and from CYP3A5_EXPR (binary expresser indicator built on a single rs776746 genotype).

CYP2B6_SM (canonical for CYP2B6 slow-metabolizer phenotype indicator)

• Description: 1 = subject is a CYP2B6 slow metabolizer defined by the combined 516G>T (rs3745274) | 983T>C (rs28399499) SNP-vector phenotype assignment used in the African paediatric antiretroviral literature: 516TT | 983TT (homozygous loss-of-function at 516) or 516GT | 983TC (compound heterozygous loss-of-function at both); 0 = any other phenotype (extensive, intermediate, or ultra-slow metabolizer; see paired canonicals CYP2B6_IM and CYP2B6_USM). Time-fixed per subject (germline genotype-derived phenotype). EM (516GG | 983TT) is the reference category when all three of CYP2B6_IM, CYP2B6_SM, and CYP2B6_USM are 0.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (any non-SM phenotype: EM, IM, or USM).
• Source aliases:
  • metabolizer status SM / SM – Bienczak 2016 (paper Methods ‘Covariate effects’ paragraph 2, Results ‘Population pharmacokinetics’ paragraph 3, and Table 3 final estimates; categorical metabolizer indicator entered as multiplicative effect on intrinsic clearance CLint with EM reference).
• Example models: Bienczak_2016_nevirapine.R (multiplicative log-additive effect on CLint: cl_meta <- exp(e_cyp2b6_im_cl * CYP2B6_IM + e_cyp2b6_sm_cl * CYP2B6_SM + e_cyp2b6_usm_cl * CYP2B6_USM), with e_cyp2b6_sm_cl = log(1 - 0.50) = -0.693, giving the 50% lower CLint reported in Bienczak 2016 Table 3 / Results ‘Population pharmacokinetics’ paragraph 3).
• Notes: Sibling indicators CYP2B6_IM and CYP2B6_USM together encode the four-level EM / IM / SM / USM phenotype with three binary columns (EM = all three zero). See CYP2B6_IM Notes for the SNP-vector pooling rationale and the link back to the dummy-coding pattern used elsewhere in the register.

CYP2B6_USM (canonical for CYP2B6 ultra-slow-metabolizer phenotype indicator)

• Description: 1 = subject is a CYP2B6 ultra-slow metabolizer defined by the combined 516G>T (rs3745274) | 983T>C (rs28399499) SNP-vector phenotype assignment used in the African paediatric antiretroviral literature: 983CC homozygosity (irrespective of 516G>T genotype, i.e. 516GG | 983CC in Bienczak 2016’s principal grouping); 0 = any other phenotype (extensive, intermediate, or slow metabolizer; see paired canonicals CYP2B6_IM and CYP2B6_SM). Time-fixed per subject (germline genotype-derived phenotype). EM (516GG | 983TT) is the reference category when all three of CYP2B6_IM, CYP2B6_SM, and CYP2B6_USM are 0.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (any non-USM phenotype: EM, IM, or SM).
• Source aliases:
  • metabolizer status USM / USM – Bienczak 2016 (paper Methods ‘Covariate effects’ paragraph 2, Results ‘Population pharmacokinetics’ paragraph 3, and Table 3 final estimates; categorical metabolizer indicator entered as multiplicative effect on intrinsic clearance CLint with EM reference).
• Example models: Bienczak_2016_nevirapine.R (multiplicative log-additive effect on CLint: cl_meta <- exp(e_cyp2b6_im_cl * CYP2B6_IM + e_cyp2b6_sm_cl * CYP2B6_SM + e_cyp2b6_usm_cl * CYP2B6_USM), with e_cyp2b6_usm_cl = log(1 - 0.68) = -1.139, giving the 68% lower CLint reported in Bienczak 2016 Table 3 / Results ‘Population pharmacokinetics’ paragraph 3).
• Notes: Sibling indicators CYP2B6_IM and CYP2B6_SM together encode the four-level EM / IM / SM / USM phenotype with three binary columns (EM = all three zero). Bienczak 2016 is the first study to quantify the USM phenotype on nevirapine clearance (cohort prevalence 0.6%; Table 2 row 4); the rs28399499 (983T>C) loss-of-function allele is essentially absent from European-ancestry populations but reaches appreciable frequency in sub-Saharan African cohorts, so models on European or East Asian populations may report only EM / IM / SM (with CYP2B6_USM identically zero across the dataset). See CYP2B6_IM Notes for the SNP-vector pooling rationale.

CYP2B6 (canonical for CYP2B6 individual metabolic-activity score)

• Description: Continuous individual-level CYP2B6 metabolic-activity score. Same intent and documentation policy as the CYP2D6, CYP3A4, and CYP2C19 continuous canonicals above. Typically derived from *-allele diplotype-to-phenotype conversion following CPIC / PharmVar guidance (poor metabolizer -> 0, intermediate -> 0.5, normal / extensive -> 1, rapid -> 1.5, ultra-rapid -> 2), but the per-model notes field documents which mapping the source paper used and the underlying diplotype-to-phenotype table when non-standard. Distinct from the categorical phenotype-indicator canonicals CYP2B6_IM / CYP2B6_SM / CYP2B6_USM (Bienczak 2016 African-paediatric SNP-vector encoding): use the continuous CYP2B6 when the source paper reports a numeric activity score, and the categorical indicators when the source reports discrete phenotype labels.
• Units: Paper-specific – document per-model. Aruldhas 2021 (methadone) uses a 5-level activity score {0, 0.5, 1, 1.5, 2} for PM / IM / NM / RM / UM derived from CPIC efavirenz guideline diplotype assignments (Supplemental Table S1 for the paper’s per-diplotype mapping).
• Type: continuous
• Scope: general
• Reference category: n/a (continuous). Models center on the normal-metabolizer reference (activity score 1); document the reference value per-model. Aruldhas 2021 uses the standard centered-additive form CLF = CLF_ref * (1 + Theta_CYP2B6 * (CYP2B6 - 1)) *... with CLF_ref reported at activity score 1.
• Source aliases:
  • CYP2B6 activity score – Aruldhas 2021 (paper Methods ‘Covariate models’ and Table S1; used as a linear covariate on the fractional metabolite-formation clearance CLF for both R- and S-methadone).
• Example models: Aruldhas_2021_R_methadone.R, Aruldhas_2021_S_methadone.R (linear covariate on the fractional metabolite-formation clearance CLF: clf <- exp(lclf) * (1 + e_cyp2b6_clf * (CYP2B6 - 1)) * ... with e_cyp2b6_clf = 0.745 for R-methadone and 0.636 for S-methadone).
• Notes: Fulfils the prospective-registration TODO recorded under CYP3A4 for the drug-metabolizing-CYP set (CYP1A2, CYP2A6, CYP2B6, CYP2C8, CYP2C9, CYP2C19, CYP2E1, CYP3A5).

GSTA1_RM (canonical for GSTA1 rapid-metabolizer diplotype-group indicator)

• Description: 1 = subject’s GSTA1 promoter diplotype places them in the rapid-metabolizer group (higher predicted promoter activity, therefore higher glutathione-S-transferase A1 expression and faster conjugation of GSTA1 substrates such as busulfan); 0 = subject is a normal or poor metabolizer. Time-fixed per subject (germline genotype-derived phenotype). Designed to be paired with GSTA1_PM so that both indicators = 0 selects the normal-metabolizer reference group.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0. When paired with GSTA1_PM (the Nava 2018 two-indicator encoding), GSTA1_RM = 0 and GSTA1_PM = 0 together indicate a normal metabolizer – the reference against which both multiplicative clearance factors are reported.
• Source aliases:
  • GSTA1 group G1 / “rapid metabolizers” – Nava 2018 (paper Methods ‘PK analysis and genotyping’; Table 1 ‘GSTA1 diplotype group – Group 1’; Table 2 ‘GSTA1-group on CL (Reference G2) – G1’). The group is assigned from the diplotype built across four GSTA1 promoter loci – -69 (rs3957357), -513 (rs11964968), -631 (rs4715333) and -1142 (rs58912740) – and grouped by predicted promoter activity per the paper’s Supplemental Material Table S1 and Figures S1-S2. No value transformation is needed: the paper’s G1 membership maps directly to GSTA1_RM = 1.
• Example models: Nava_2018_busulfan.R (multiplicative factor on busulfan clearance: fgsta1 <- 1 + (e_gsta1_rm_cl - 1) * GSTA1_RM + (e_gsta1_pm_cl - 1) * GSTA1_PM with e_gsta1_rm_cl = 1.07; rapid metabolizers clear busulfan 7% faster than normal metabolizers).
• Notes: Member of the <GENE>_<phenotype> paired-binary-indicator family whose pattern the CYP2D6_PM / CYP2D6_EM / CYP2D6_IM / CYP2D6_UM and CYP2C9_EM / CYP2B6_IM entries document. The _RM suffix (rapid) is used rather than _UM (ultrarapid) because the source paper’s own three-level vocabulary is rapid / normal / poor; _UM remains reserved for sources that describe a gene-duplication-driven ultrarapid stratum. Which subset of the family a model carries depends on how the source stratifies: a source reporting only a poor-vs-not dichotomy needs GSTA1_PM alone; Nava 2018 resolves rapid and poor against a normal reference and so carries both. A GSTA1_NM normal-metabolizer indicator is deliberately NOT registered – the normal group is the implicit reference (both indicators = 0), exactly as IM is for the Sherwin 2012 CYP2D6_PM + CYP2D6_EM pairing – and should only be added if a future source makes normal a non-reference stratum. Distinct from GST_BL_NMOL_MIN_ML, which carries a measured baseline glutathione-S-transferase enzyme activity rather than a genotype-derived phenotype class; a paper offering both should prefer the continuous measurement.

GSTA1_PM (canonical for GSTA1 poor-metabolizer diplotype-group indicator)

• Description: 1 = subject’s GSTA1 promoter diplotype places them in the poor-metabolizer group (lower predicted promoter activity, therefore reduced glutathione-S-transferase A1 expression and slower conjugation of GSTA1 substrates such as busulfan); 0 = subject is a normal or rapid metabolizer. Time-fixed per subject (germline genotype-derived phenotype). Companion to GSTA1_RM above.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0. When paired with GSTA1_RM, both indicators = 0 indicates a normal metabolizer.
• Source aliases:
  • GSTA1 group G3 / “poor metabolizers” – Nava 2018 (paper Methods ‘PK analysis and genotyping’; Table 1 ‘GSTA1 diplotype group – Group 3’; Table 2 ‘GSTA1-group on CL (Reference G2) – G3’). Assigned from the same four-locus promoter diplotype as GSTA1_RM. No value transformation is needed.
• Example models: Nava_2018_busulfan.R (multiplicative factor on busulfan clearance with e_gsta1_pm_cl = 0.88; poor metabolizers clear busulfan 12% more slowly than normal metabolizers, and are the group the paper flags as being at higher risk of toxic busulfan exposure).
• Notes: See the GSTA1_RM entry above for the family pattern, the choice of the _RM suffix for the rapid stratum, why no GSTA1_NM companion is registered, and the distinction from the measured-activity canonical GST_BL_NMOL_MIN_ML. A source that genotypes only the -69C/T locus (rs3957357) resolves GSTA1*B homozygotes but cannot distinguish the *B1b haplotype or separate *A2 / *A3 from *A1 (Nava 2018 Discussion), so it cannot populate the rapid stratum; such a source should use GSTA1_PM alone, or the SNP-level SNP_GSTA1_RS3957357 canonical if one is registered for it in future.

SNP_GSTA1_RS3957356 (canonical for GSTA1 -52G>A (rs3957356) variant-carrier indicator)

• Description: Binary genotype indicator for the GSTA1 promoter single-nucleotide polymorphism -52G>A (rs3957356), the marker of the reduced-promoter-activity GSTA1*B haplotype. 1 = the variant was detected (at least one A allele); 0 = not detected. Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (variant not detected).
• Source aliases:
  • GSTA1 / “GSTA1 SNP” / “GSTA1 52G>A” – Dadkhah 2022 (Section 2.3 genotyping by the Ansari et al. method; Section 2.5 categorical covariates coded 0/1; Table 1 row ‘GSTA1 52G>A’ 28 of 37 = 75.7%; Section 3.3 equation (2) ‘for GSTA1’ / ‘for non-GSTA1’). The paper does not separate heterozygous from homozygous carriers. No value transformation is needed.
• Example models: Dadkhah_2022_busulfan.R (exponential effect on the clearance of the busulfan metabolite sulfolane: CLSu = 1.61 * exp(1.43 * SNP_GSTA1_RS3957356) L/h, i.e. carriers clear sulfolane 4.18-fold faster; the paper notes the direction is counterintuitive and attributes it to intermediate steps of the busulfan metabolic pathway).
• Notes: Member of the SNP_<GENE>_RS<rsid> family. Distinct from GSTA1_RM / GSTA1_PM, which classify a four-locus promoter diplotype into metabolizer groups (Nava 2018) and cannot be derived from this single locus; and from the -69C>T locus rs3957357, which is in strong linkage with rs3957356 but is a different SNP and needs its own entry if a source uses it. Distinct from GST_BL_NMOL_MIN_ML, the measured glutathione-S-transferase activity phenotype.

SNP_ORM1_RS17650 (canonical for ORM1 rs17650 active-allele count)

• Description: Continuous individual-level ORM1 rs17650 active-allele count: 0, 1, or 2 copies. The rs17650 SNP distinguishes the “F” (fast-migrating) and “S” (slow-migrating) allozymes of alpha-1 acid glycoprotein (AAG) encoded by ORM1. The variant F-phenotype allele binds methadone with lower affinity than the reference S phenotype (Aruldhas 2021 Discussion). Time-invariant (germline genotype). The paper uses the standard centered-additive form V2 = V2_ref * (1 + Theta_rs17650 * (n_active - 1)) with n_active = number of active alleles at rs17650.
• Units: (count, 0/1/2 alleles per subject)
• Type: continuous
• Scope: specific
• Reference category: n/a (continuous). Aruldhas 2021 centers the covariate at n_active = 1 (heterozygous), so V2_ref is the typical central-compartment volume in a heterozygous subject at the reference AAG concentration.
• Source aliases:
  • rs17650 / number of active alleles rs17650 – Aruldhas 2021 (paper Results ‘Covariate modeling on R methadone’ / ‘Covariate modeling on S methadone’ and Table 2).
• Example models: Aruldhas_2021_R_methadone.R, Aruldhas_2021_S_methadone.R (linear covariate on the central-compartment volume V2: vc <- vc_ref * (1 + e_snp_orm1_rs17650_vc * (SNP_ORM1_RS17650 - 1)) * (1 + e_aag_vc * (AAG - 94.76)); the effect is independent of and additive with the AAG-concentration effect on V2).
• Notes: Aruldhas 2021 reports that the addition of rs17650 to V2 improved the fit even after AAG concentration was already in the model (R-methadone dOFV = -6.5; S-methadone dOFV = -7.2), attributing this to differences in binding affinity between the F- and S- allozymes rather than to changes in AAG concentration. Rs1126801 (a companion SNP that further subdivides the F allele into F1 vs F2) was too rare in the Aruldhas 2021 study population (n = 2 subjects with the variant) to be included as a covariate. Scoped specific because rs17650 has only been reported as a covariate in the Aruldhas 2021 methadone analysis; future paediatric or adult popPK models of AAG-bound basic / lipophilic drugs may promote the scope to general.

SNP_CYP3A4_RS2246709 (canonical for CYP3A4 rs2246709 active-allele count)

• Description: Continuous individual-level CYP3A4 rs2246709 active-allele count: 0, 1, or 2 copies. rs2246709 is an intronic CYP3A4 variant that Aruldhas 2021 identified as significantly associated with reduced R- and S-methadone fractional clearance to EDDP after backward elimination. Time-invariant (germline genotype). The paper uses the standard centered-additive form CLF = CLF_ref * (1 + Theta_rs2246709 * (n_active - 1)).
• Units: (count, 0/1/2 alleles per subject)
• Type: continuous
• Scope: specific
• Reference category: n/a (continuous). Aruldhas 2021 centers the covariate at n_active = 1 (heterozygous), so CLF_ref is the typical fractional metabolite-formation clearance in a subject with heterozygous rs2246709 and CYP2B6 activity score 1.
• Source aliases:
  • rs2246709 / number of active alleles rs2246709 – Aruldhas 2021 (paper Results ‘Covariate modeling on R methadone’ and Table 2).
• Example models: Aruldhas_2021_R_methadone.R, Aruldhas_2021_S_methadone.R (linear covariate on the fractional metabolite-formation clearance CLF: clf <- clf_ref * (1 + e_cyp2b6_clf * (CYP2B6 - 1)) * (1 + e_snp_cyp3a4_rs2246709_clf * (SNP_CYP3A4_RS2246709 - 1)); e_snp_cyp3a4_rs2246709_clf = 0.450 for R-methadone (RSE 33.9%) and 1.68 for S-methadone (RSE 57.8%, retained as the final S-methadone model in preference to the initial rs11882424-based model because rs11882424 is less commonly genotyped in clinical laboratories)).
• Notes: Note that the very-large S-methadone coefficient (1.68) with wide RSE (57.8%) implies that the paper’s model can produce a physically-invalid (negative) CLF for S-methadone at n_active = 0 (wild-type homozygous) subjects: 1 + 1.68 * (0 - 1) = -0.68. This is a limitation of the paper’s parameter estimates at the extreme of the covariate range, not of the extraction; the vignette Assumptions and deviations section documents that users simulating a wild-type-homozygous rs2246709 subject should either use the alternative rs11882424-based S-methadone model reported in the paper’s Table 2 discussion or accept the paper’s noted limitation. Distinct from the standard SNP_<GENE>_RS<rsid>_<allele>_COUNT naming convention because Aruldhas 2021 does not specify the wild-type vs variant allele letters at rs2246709 in the paper text; the column carries the paper’s “number of active alleles” (which the paper treats as a symmetric additive count without labelling which allele is which). Scoped specific because rs2246709 has only been reported as a covariate in the Aruldhas 2021 methadone analysis.

CONMED_CYP2D6_INH (canonical for concomitant CYP2D6 inhibitor coadministration indicator)

• Description: 1 = subject coadministered any medication classified by the source paper as a CYP2D6 inhibitor during the study, 0 = no concomitant CYP2D6 inhibitor. The CYP2D6 sibling of CONMED_CYP3A4_INH, and distinct in the same way from the CYP2D6 continuous-activity-score and the CYP2D6_PM / CYP2D6_IM / CYP2D6_EM / CYP2D6_UM phenotype canonicals: those describe the subject’s intrinsic enzyme activity, whereas this column records a concomitant-medication (drug-drug-interaction) exposure. Use it when the source paper enters CYP2D6-inhibitor coadministration into the model as a binary indicator, regardless of which inhibitor strengths it pools into the 1 category.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no CYP2D6 inhibitor coadministration).
• Source aliases:
  • Patient using con med classified as CYP2D6 inhibitor – the indicator’s spelled-out name in the Riggs 2012 albinterferon covariate equations (p. 479).
  • Other source-dataset column names typically: CYP2D6I, CYP2D6INH, or a free-text concomitant-medication indicator. Document the source-column name per-model in covariateData[[CONMED_CYP2D6_INH]]$source_name.
• Example models: Riggs_2012_albinterferon.R (multiplicative effect 1.05 on apparent clearance of the albumin-interferon fusion protein albinterferon alfa-2b, per Table I row “CYP 2D6 inhibitor present”; the paper never enumerates which agents it classified into the class. Note that albinterferon is cleared by proteolysis rather than by cytochrome-P450 metabolism, so in that model the indicator is best read as a marker of comedication burden or of the condition prompting it rather than as a mechanistic interaction – the estimated effect’s bootstrap 95% CI, 0.979 to 1.12, sits inside the authors’ +/-25% no-clinical-relevance band).
• Notes: Per-model covariateData[[CONMED_CYP2D6_INH]]$notes must document which inhibitor strengths and which specific drug examples the source pooled into the = 1 category, since inclusion criteria vary by study; when the paper does not say, record that gap explicitly. Future models needing strength-stratified encoding should register companion canonicals (CONMED_CYP2D6_INH_STRONG, CONMED_CYP2D6_INH_MOD) rather than overloading this entry, mirroring the CONMED_CYP3A4_INH_STRONG / CONMED_CYP3A4_INH_MOD pattern. A CYP2D6-inducer sibling would be CONMED_CYP2D6_IND, following CONMED_CYP3A4_IND.

CONMED_CYP3A4_INH (canonical for concomitant CYP3A4 inhibitor coadministration indicator)

• Description: 1 = subject coadministered any CYP3A4 inhibitor during the study, 0 = no concomitant CYP3A4 inhibitor. Distinct from the CYP3A4 continuous-activity-score canonical above: CONMED_CYP3A4_INH captures concomitant-medication exposure (a drug-drug-interaction indicator), not intrinsic enzyme activity. Use this canonical when the source paper enters CYP3A4-inhibitor coadministration into the popPK model as a binary indicator, regardless of which inhibitor strengths (strong / moderate / weak) the paper pools into the 1 category.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no CYP3A4 inhibitor coadministration).
• Source aliases:
  • CYP3A4_INH – prior canonical name (pre-2026-06-19 CONMED_ prefix standardization).
  • Other source-dataset column names typically: CYP3AI, CYP3A4I, CYP3AINH, or a free-text concomitant-medication indicator. Document the source-column name per-model in covariateData[[CONMED_CYP3A4_INH]]$source_name.
• Example models: Yassen_2025_asundexian.R (proportional-shift effect on CL/F: (1 + e_cyp3a4_inh_cl * CONMED_CYP3A4_INH) with e_cyp3a4_inh_cl = -0.0531; the asundexian dataset pools weak + moderate CYP3A4 inhibitors into the CONMED_CYP3A4_INH = 1 category because strong inhibitors were a Phase II exclusion criterion).
• Notes: Per-model covariateData[[CONMED_CYP3A4_INH]]$notes must document which inhibitor strengths (strong / moderate / weak) and which specific drug examples are pooled into the CONMED_CYP3A4_INH = 1 category, since inclusion criteria vary by study. Future models that need stratified encoding (separate strong / moderate / weak indicators) should register companion canonicals (e.g. CONMED_CYP3A4_INH_STRONG, CONMED_CYP3A4_INH_MOD, CONMED_CYP3A4_INH_WEAK) rather than overloading CONMED_CYP3A4_INH. The complementary CYP3A4-inducer indicator follows the same pattern as CONMED_CYP3A4_IND.

CONMED_CYP3A4_INH_HI (canonical for concomitant CYP3A4 inhibitor (strong/moderate/weak) coadministration with high (>=50%) cumulative exposure during the on-treatment period)

• Description: 1 = subject coadministered any CYP3A4 inhibitor classified as strong, moderate, or weak (FDA/EMA classification) for at least 50% of the on-study treatment period; 0 = otherwise. Companion canonical to CONMED_CYP3A4_INH that stratifies by cumulative exposure intensity (rather than by inhibitor strength). Distinct from CONMED_CYP3A4_INH_LO which captures the complementary “lower-exposure / unclassified” category. The two indicators are mutually exclusive per subject (a subject is either in CONMED_CYP3A4_INH_HI = 1, CONMED_CYP3A4_INH_LO = 1, or both = 0); when both are 0 the subject was either never on a CYP3A4 inhibitor or was on one only briefly.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no high-cumulative-exposure CYP3A4 inhibitor coadministration; subject may still have CONMED_CYP3A4_INH_LO = 1).
• Source aliases:
  • CYPINH = 4 (strong, >=50%) or CYPINH = 6 (moderate, >=50%) or CYPINH = 8 (weak, >=50%) in the FIDELIO-DKD popPK NONMEM control stream’s 9-level CYPINH categorical variable; used in vandenBerg_2021_finerenone.R.
• Example models: vandenBerg_2021_finerenone.R (multiplicative effect 0.951^CONMED_CYP3A4_INH_HI on CL/F AND on F1 inversely, per paper’s NONMEM control stream).
• Notes: Companion canonical to the broader CONMED_CYP3A4_INH (any inhibitor coadministration, no exposure-intensity stratification). The pooled strong/moderate/weak grouping at the >=50% exposure threshold matches the FIDELIO-DKD analysis’s covariate-engineering choice (the study did not separate strong from moderate from weak inhibitors but did separate high-cumulative-exposure from lower-exposure). The accompanying inhibitor-strength-stratified canonicals CONMED_CYP3A4_INH_STRONG / CONMED_CYP3A4_INH_MOD / CONMED_CYP3A4_INH_WEAK mentioned in the CONMED_CYP3A4_INH Notes follow a different stratification axis (by strength, ignoring cumulative duration) and would coexist with the HI / LO duration-based pair without conflict.

CONMED_CYP3A4_INH_LO (canonical for concomitant CYP3A4 inhibitor coadministration in ‘other’ lower-exposure or unclassified category)

• Description: 1 = subject coadministered a CYP3A4 inhibitor in any of the following sub-categories: (a) unclassified inhibitor (strength not assignable to the strong / moderate / weak FDA/EMA tiers) at any duration, OR (b) strong / moderate / weak inhibitor present for LESS THAN 50% of the on-study treatment period; 0 = otherwise. Companion canonical to CONMED_CYP3A4_INH_HI that captures all CYP3A4-inhibitor exposure other than the high-cumulative-exposure-and-known-strength category. The two indicators are mutually exclusive per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no lower-exposure / unclassified CYP3A4 inhibitor coadministration; subject may still have CONMED_CYP3A4_INH_HI = 1).
• Source aliases:
  • CYPINH IN (1, 2, 3, 5, 7) in the FIDELIO-DKD popPK NONMEM control stream’s 9-level CYPINH categorical variable – corresponding to unclassified <50% (1), unclassified >=50% (2), strong <50% (3), moderate <50% (5), weak <50% (7). Used in vandenBerg_2021_finerenone.R.
• Example models: vandenBerg_2021_finerenone.R (multiplicative effect 0.996^CONMED_CYP3A4_INH_LO on CL/F AND on F1 inversely, per paper’s NONMEM control stream).
• Notes: Companion to CONMED_CYP3A4_INH_HI. The unclassified-inhibitor category pools any concomitant medication the source dataset’s medication-classification logic could not assign to a strong / moderate / weak tier (typically due to ambiguous drug-coding); these subjects would otherwise be lost to analysis if the encoding required a strength tier. Combining unclassified + <50%-exposure into a single “lower-impact” indicator was the FIDELIO-DKD analysis’s choice; the effect size (0.996) is much smaller than the high-exposure category (0.951).

CONMED_PGP_INH (canonical for concomitant P-glycoprotein inhibitor coadministration indicator)

• Description: 1 = subject coadministered a P-glycoprotein (P-gp / MDR1 / ABCB1) efflux-transporter inhibitor during the study, 0 = no concomitant P-gp inhibitor. P-gp inhibition raises the systemic exposure of P-gp substrates by reducing intestinal efflux (increasing oral bioavailability) and by reducing biliary and renal-tubular secretion (reducing apparent clearance). Use this canonical when the source paper enters P-gp-inhibitor coadministration into the popPK model as a binary indicator, regardless of which inhibitor potencies (potent / strong / moderate / weak) the paper pools into the 1 category.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no P-glycoprotein-inhibitor coadministration).
• Source aliases:
  • DDICYPSH – used in Mitra_2026_ziftomenib.R (Kura Oncology KOMET-001 + KO-MEN-003 NONMEM control-stream column; strong CYP3A4 inhibitor coadministration during a ziftomenib dose record; dominant driver in the R/R AML cohort is prophylactic antifungal azole use).
• Example models: Mitra_2026_ziftomenib.R (multiplicative effects: 0.459x on parent ziftomenib CL/F, 0.195x on KO-739 CL, 0.449x on KO-516 CL when CONMED_CYP3A4_INH_STRONG = 1; encoded as additive shifts on the log(CL) scale per the paper’s NONMEM PK block), Kemal_2026_nemtabrutinib.R (multiplicative effect on CL/F: (1 + -0.0119 * CONMED_CYP3A4_INH_STRONG), i.e. 1.2% lower CL/F under strong CYP3A4 inhibition; RSE 774%, 95% CI includes zero – retained in the full covariate model), Wada_2023_sparsentan.R (log-additive effect on CL/F: exp(-1.069 * CONMED_CYP3A4_INH_STRONG), a 66% reduction in apparent clearance and a 191.3% increase in steady-state AUC – the largest covariate effect in the model and the only one for which the paper suggests a dose adjustment may be warranted; paired with the sibling new canonical CONMED_CYP3A4_INH_MOD at -0.273), Thoueille_2023_tenofovir_full.R, Thoueille_2023_tenofovir_alafenamide.R (multiplicative fractional effect on apparent tenofovir clearance: (1 + e_conmed_pgp_inh_cl * CONMED_PGP_INH) with e_conmed_pgp_inh_cl = -0.121 and -0.116 respectively, i.e. ~12% lower CL/F, an effect the paper reports as independent of cobicistat coadministration).
• Notes: Companion canonical to CONMED_CYP3A4_INH (which pools any inhibitor strength into a single 0/1 indicator) and to CONMED_CYP3A4_INH_HI / CONMED_CYP3A4_INH_LO (which stratify by cumulative exposure duration rather than by inhibitor-strength category). Future extractions that need the sibling moderate- and weak-strength indicators should register CONMED_CYP3A4_INH_MOD and CONMED_CYP3A4_INH_WEAK following the same pattern.

CONMED_CYP3A4_INH_MOD (canonical for concomitant moderate CYP3A4 inhibitor coadministration indicator)

• Description: 1 = subject / dose-record with concomitant coadministration of a moderate CYP3A4 inhibitor (FDA/EMA classification of “moderate”: AUC increase >= 2-fold and < 5-fold for a sensitive CYP3A4 substrate; representative agents include cyclosporine, erythromycin, diltiazem, verapamil, fluconazole, and grapefruit juice), 0 = no moderate CYP3A4 inhibitor coadministration during the observation window (whether no CYP3A4 inhibitor at all, or a weak or strong inhibitor only). Time-varying per record. Strength-stratified companion to the broader CONMED_CYP3A4_INH and sibling to CONMED_CYP3A4_INH_STRONG; distinct from CONMED_CYP3A4_INH_HI / CONMED_CYP3A4_INH_LO, which stratify by cumulative exposure duration rather than by inhibitor-strength tier.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no moderate CYP3A4 inhibitor coadministration; a subject on a weak or strong inhibitor, or on none, falls here when using this indicator alone). When paired with CONMED_CYP3A4_INH_STRONG, both = 0 selects the pooled “no moderate or strong inhibitor” reference.
• Source aliases:
  • CYP3A4 inhibitor three- or four-level categorical column (none / weak / moderate / strong) – the moderate level maps to CONMED_CYP3A4_INH_MOD = 1. Used in Wada_2023_sparsentan.R, where the source dataset’s CYP3A4 inhibitor column has levels none / weak / moderate / strong and the paper estimates separate coefficients for the moderate and strong levels only, leaving weak pooled into the reference.
• Example models: Wada_2023_sparsentan.R (log-additive effect on CL/F: exp(-0.273 * CONMED_CYP3A4_INH_MOD), i.e. 24% lower CL/F under moderate CYP3A4 inhibition, which the paper translates to a 31.4% increase in steady-state AUC and a 16.0% increase in Cmax; paired with CONMED_CYP3A4_INH_STRONG at -1.069; cohort exposure 10.5% moderate), Jiang_2021_ivosidenib.R (pooled with CONMED_CYP3A4_INH_STRONG as max(MOD, STRONG) into the paper’s ‘other strong or moderate CYP3A4 inhibitors’ group, 0.92-fold CL/F, switched off when voriconazole, fluconazole or posaconazole is flagged because those carry their own coefficients).
• Notes: Per-model covariateData[[CONMED_CYP3A4_INH_MOD]]$notes must document which specific agents the source paper classified as moderate and what happens to the weak stratum (pooled into the reference, or carried in a separate CONMED_CYP3A4_INH_WEAK indicator). The remaining sibling CONMED_CYP3A4_INH_WEAK is registered below, founded by Willmann_2018a_rivaroxaban.R, which estimates all three strengths simultaneously as separate indicators.

CONMED_CYP3A4_INH_WEAK (canonical for concomitant weak CYP3A4 inhibitor coadministration indicator)

• Description: 1 = subject / dose-record with concomitant coadministration of a weak CYP3A4 inhibitor (FDA/EMA classification of “weak”: AUC increase >= 1.25-fold and < 2-fold for a sensitive CYP3A4 substrate; representative agents include cimetidine, ranitidine, amlodipine, fluvoxamine and amiodarone), 0 = no weak CYP3A4 inhibitor coadministration during the observation window (whether no CYP3A4 inhibitor at all, or a moderate or strong inhibitor only). Time-varying per record. Strength-stratified companion to the broader CONMED_CYP3A4_INH and sibling to CONMED_CYP3A4_INH_STRONG and CONMED_CYP3A4_INH_MOD; distinct from CONMED_CYP3A4_INH_HI / CONMED_CYP3A4_INH_LO, which stratify by cumulative exposure duration rather than by inhibitor-strength category.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no weak CYP3A4 inhibitor coadministration). When the full strength-stratified triple is used together, all three of CONMED_CYP3A4_INH_WEAK, CONMED_CYP3A4_INH_MOD and CONMED_CYP3A4_INH_STRONG = 0 selects the “no CYP3A4 inhibitor” reference.
• Source aliases:
  • WCYP3A4INH1 – the NONMEM control-stream column name in Willmann_2018a_rivaroxaban.R (Willmann 2018a Supplementary Data S9, run12 $INPUT). Its numeric suffix is a category code and carries no meaning beyond distinguishing the five comedication columns of that dataset.
• Example models: Willmann_2018a_rivaroxaban.R (multiplicative effect on rivaroxaban CL/F: 0.939^CONMED_CYP3A4_INH_WEAK, i.e. 6.1% lower apparent clearance under weak CYP3A4 inhibition; RSE 2.17%, bootstrap 95% CI 0.900-0.975; estimated simultaneously with CONMED_CYP3A4_INH_MOD at 0.863 and CONMED_CYP3A4_INH_STRONG at 0.978, and multiplied against CONMED_PGP_INH and CONMED_CYP3A4_IND), Jiang_2021_ivosidenib.R (the paper’s ‘mild CYP3A4 inhibitors’ group; multiplicative 1.04-fold change in ivosidenib CL/F).
• Notes: Per-model covariateData[[CONMED_CYP3A4_INH_WEAK]]$notes must document which specific agents the source paper classified as weak; Willmann 2018a does not name them. Use this canonical only when the source estimates the weak stratum as its own coefficient – most popPK analyses pool weak inhibitors into the reference (see the CONMED_CYP3A4_INH_MOD entry, where Wada_2023_sparsentan.R does exactly that), and reusing a coefficient across strength strata is not valid. A caution that generalises beyond this canonical: in the founding model the strong-inhibitor coefficient is effectively null (0.978) because strong inhibitors were a protocol exclusion and only 6 of 5,041 patients received one, so the strength ordering of the three estimated coefficients is NOT monotonic and must not be read as a dose-response in inhibitor potency.

CONMED_CYP3A4_PGP_INH (canonical for concomitant CYP3A4-OR-P-glycoprotein inhibitor coadministration indicator (pooled union of the two mechanisms))

• Description: 1 = subject coadministered a concomitant medication that inhibits CYP3A4 or P-glycoprotein (P-gp / MDR1 / ABCB1), or both; 0 = neither. This is the union of the two inhibition mechanisms carried in a single indicator, and it is deliberately distinct from CONMED_CYP3A4_INH (CYP3A4 inhibition only) and CONMED_PGP_INH (P-gp inhibition only). Use it when the source paper defines its comedication covariate as an “or” of the two mechanisms rather than testing them separately – a common choice for the direct oral anticoagulants (rivaroxaban, apixaban, edoxaban, dabigatran), whose product labelling frames the clinically actionable drug-drug interaction as combined CYP3A4-and-P-gp inhibition and whose real-world cohorts are dominated by agents (amiodarone, dronedarone, diltiazem, verapamil) that hit both pathways at once. Supplying a value derived from only one of the two mechanisms silently misclassifies any patient on a single-pathway inhibitor.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant CYP3A4 or P-gp inhibitor).
• Source aliases:
  • CYP3A4 or P-gp inhibitors – the source-paper category label used in Lai_2026_rivaroxaban.R (Lai 2026 Methods 2.2 and Table 1).
• Example models: Lai_2026_rivaroxaban.R (log-additive effect on CL/F: exp(-0.38 * CONMED_CYP3A4_PGP_INH), i.e. a 31.6% reduction in apparent clearance, which the paper’s Discussion rounds to “reduced CL/F by 32%”; the largest covariate effect in the model, with the tightest RSE of any parameter at 5.44%. 86 of 226 real-world Taiwanese patients with atrial fibrillation (38.1%) were positive; the agents were amiodarone (n = 56), dronedarone (n = 15), diltiazem (n = 12) and verapamil (n = 6), all four of which inhibit P-gp, with dronedarone / diltiazem / verapamil additionally moderate CYP3A4 inhibitors and amiodarone a weak one – so in this particular cohort the two mechanisms coincide for every positive patient even though the covariate is defined as their union), Cirincione_2018_apixaban.R, Cirincione_2018_apixaban_asian_subgroups.R (fractional-change effect on total apparent apixaban clearance, (1 + e_conmed_cyp3a4_pgp_inh_cl * CONMED_CYP3A4_PGP_INH) with coefficient -0.146, i.e. a 14.6% reduction. 718 of 4,385 subjects (16.4%) were positive. Two features of this ratification are worth carrying forward. It is genuinely time-varying at the observation record – the authors define the covariate as coadministration “at the time of the PK sample collection”, not ever-use, which is the orientation the canonical’s description assumes but which Lai 2026 did not exercise. And the strong and moderate inhibitor categories were pooled by necessity rather than by design: only 3 of 4,385 subjects contributed a sample while on a strong inhibitor, so the estimated effect is essentially a moderate-inhibitor effect wearing a union label. The paper says so itself, contrasting its 14.6% clearance reduction with the 99% and 40% AUC increases measured in the dedicated ketoconazole and diltiazem interaction studies, and advising that “greater emphasis should be placed on the phase I results” for clinical decisions about this interaction).
• Notes: Per-model covariateData[[CONMED_CYP3A4_PGP_INH]]$notes must list the specific agents the source paper pooled into the 1 category and, where the paper reports it, the share of positive patients each agent accounts for – the mechanism mix determines whether the estimated effect size transfers to a different comedication profile. Two boundaries are worth stating explicitly when using this canonical. First, it does not subsume a separately-modelled BCRP-inhibitor indicator: Lai 2026 screened concomitant BCRP inhibitors (benzbromarone, febuxostat, sulfasalazine) as their own covariate on CL/F and eliminated it, so those patients are not carried here. Second, the complementary inducer arm follows the existing CONMED_CYP3A4_IND pattern and is not part of this indicator. Register this union canonical rather than overloading CONMED_CYP3A4_INH, for the same reason CONMED_KETOCONAZOLE is kept distinct from CONMED_CYP3A4_INH_STRONG: an effect size estimated for combined CYP3A4 + P-gp inhibition is not interchangeable with one estimated for CYP3A4 inhibition alone. Promoted to general scope on the ratification by Cirincione_2018_apixaban.R, per this entry’s own standing instruction: two independent analyses, two different direct oral anticoagulants (rivaroxaban and apixaban), two different data sources (a real-world Taiwanese cohort and a pooled twelve-study clinical development programme), both defining the covariate as the union of the two inhibition mechanisms with the same 1 = exposed orientation. The two estimated effect sizes differ by a factor of two (31.6% versus 14.6% reduction in CL/F), which is the expected consequence of different comedication mixes and is exactly why the per-model notes requirement above is retained rather than relaxed.

CONMED_CYP3A4_IND (canonical for concomitant CYP3A4 inducer coadministration indicator)

• Description: 1 = subject coadministered any CYP3A4 inducer during the study, 0 = no concomitant CYP3A4 inducer. Sibling indicator to CONMED_CYP3A4_INH; both capture concomitant-medication exposure (a drug-drug-interaction indicator), not intrinsic enzyme activity (distinct from the CYP3A4 continuous-activity-score canonical above). Use this canonical when the source paper enters CYP3A4-inducer coadministration into the popPK model as a binary indicator, regardless of which inducer strengths (strong / moderate / weak) the paper pools into the 1 category.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no CYP3A4 inducer coadministration).
• Source aliases:
  • CYP3A4_IND – prior canonical name (pre-2026-06-19 CONMED_ prefix standardization).
  • Other source-dataset column names typically: CYP3AIND, CYP3A4IND, INDU, INDUCER, or a free-text concomitant-medication indicator. Document the source-column name per-model in covariateData[[CONMED_CYP3A4_IND]]$source_name.
• Example models: Gupta_2016_lenvatinib.R (multiplicative power-form effect on CL/F: 1.30^CONMED_CYP3A4_IND with e_cyp3a4_ind_cl = log(1.30) ~ 0.262; the Gupta dataset pools any concomitant CYP3A4 inducer reported in the per-subject medication log into the CONMED_CYP3A4_IND = 1 category, with n = 19 (2.4%) of the 779-subject pooled cohort flagged positive).
• Notes: Per-model covariateData[[CONMED_CYP3A4_IND]]$notes must document which inducer strengths (strong / moderate / weak) and which specific drug examples are pooled into the CONMED_CYP3A4_IND = 1 category, since inclusion criteria vary by study. Future models that need stratified encoding (separate strong / moderate / weak indicators) should register companion canonicals (e.g. CONMED_CYP3A4_IND_STRONG, CONMED_CYP3A4_IND_MOD, CONMED_CYP3A4_IND_WEAK) rather than overloading CONMED_CYP3A4_IND. Sibling canonical to CONMED_CYP3A4_INH.

CONMED_CYP3A4_IND_MOD (canonical for concomitant moderate CYP3A4 inducer coadministration indicator)

• Description: 1 = subject coadministered a moderate-strength CYP3A4 inducer (FDA / EMA classification) at the observation, 0 = no concomitant moderate CYP3A4 inducer. Strength-stratified companion to CONMED_CYP3A4_IND (pooled any-inducer indicator) and to the sibling canonicals CONMED_CYP3A4_IND_STRONG / CONMED_CYP3A4_IND_WEAK (reserved for future extractions using the same stratification convention). Use this canonical when the source paper enters moderate-strength CYP3A4 inducers as an isolated indicator rather than pooling them with strong or weak inducers.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant moderate CYP3A4 inducer; subjects on strong or weak inducers, or on no inducer, all fall here when using this indicator alone).
• Source aliases:
  • C3A4INDM – used in Kemal_2026_nemtabrutinib.R (Kemal 2026 NONMEM control stream; time-varying regressor 1 = moderate CYP3A4 inducer at the observation, 0 otherwise. In the Kemal 2026 cohort no subjects received a strong inducer, so the source dataset enters only the moderate-inducer stratum into the model).
• Example models: Kemal_2026_nemtabrutinib.R (multiplicative effect on CL/F: (1 + 0.0220 * CONMED_CYP3A4_IND_MOD), i.e. 2.2% higher CL/F under moderate CYP3A4 induction; RSE 458%, 95% CI includes zero – retained in the full covariate model).
• Notes: Companion to the pooled CONMED_CYP3A4_IND (any inducer, no strength stratification). Per-model covariateData[[CONMED_CYP3A4_IND_MOD]]$notes should document which specific inducers the source paper classified as moderate (typically per FDA guidance: efavirenz, bosentan, etravirine, phenobarbital, modafinil, dabrafenib) and the fraction of the cohort exposed. Sibling canonicals CONMED_CYP3A4_IND_STRONG and CONMED_CYP3A4_IND_WEAK are reserved for parallel future extractions and are named in the CONMED_CYP3A4_IND register entry.

CONMED_CYP3A4_IND_STRONG (canonical for concomitant strong CYP3A4 inducer coadministration indicator)

• Description: 1 = subject coadministered a strong CYP3A4 inducer (FDA / EMA classification, i.e. an inducer that reduces the AUC of a sensitive CYP3A4 substrate by >= 80%) at the observation, 0 = no concomitant strong CYP3A4 inducer. Strength-stratified companion to CONMED_CYP3A4_IND (pooled any-inducer indicator) and sibling of CONMED_CYP3A4_IND_MOD. Use this canonical when the source paper enters strong-strength CYP3A4 inducers as their own indicator rather than pooling them with moderate or weak inducers.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant strong CYP3A4 inducer; subjects on moderate or weak inducers, or on no inducer, all fall here when using this indicator alone).
• Source aliases:
  • Strong CYP3A4 inducer – Rich 2026 Table 1 row label (Rich_2026_momelotinib.R); the source dataset column name is not published.
• Example models: Rich_2026_momelotinib.R (log-additive effect on apparent momelotinib clearance: exp(log(2.01) * CONMED_CYP3A4_IND_STRONG), i.e. a 2.01-fold rise in CL/F and therefore a 50% fall in Cavg,ss, paired with the sibling CONMED_CYP3A4_IND_MOD at 1.39. Rich 2026 attributes the effect to joint induction of CYP3A / 2C8 / 2C19 rather than CYP3A4 alone, consistent with the 46% fall in momelotinib AUCinf seen in the dedicated rifampin drug-drug-interaction study).
• Notes: Reserved by name in the CONMED_CYP3A4_IND and CONMED_CYP3A4_IND_MOD register entries before this first use, so the name is not a fresh coinage. Per-model covariateData[[CONMED_CYP3A4_IND_STRONG]]$notes must document which specific inducers the source classified as strong (typically per FDA guidance: rifampicin, carbamazepine, phenytoin, St John’s wort, enzalutamide, mitotane) and the fraction of the cohort exposed; when the paper does not say, record that gap explicitly (Rich 2026 does not). Prefer the drug-specific CONMED_RIFAMPICIN when a study’s strong-inducer arm is a single named agent given as a probe; use this class indicator when the source pools whatever strong inducers patients happened to be taking.

CONMED_CYP3A4_IND_WEAK (canonical for concomitant weak CYP3A4 inducer coadministration indicator)

• Description: 1 = subject / record with concomitant coadministration of a weak CYP3A4 inducer (FDA / EMA classification: an inducer that reduces the AUC of a sensitive CYP3A4 substrate by >= 20% and < 50%), 0 = no concomitant weak CYP3A4 inducer. Strength-stratified companion to CONMED_CYP3A4_IND (pooled any-inducer indicator) and sibling of CONMED_CYP3A4_IND_MOD and CONMED_CYP3A4_IND_STRONG; the inducer-side counterpart of CONMED_CYP3A4_INH_WEAK.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant weak CYP3A4 inducer; subjects on moderate or strong inducers, or on no inducer, all fall here when using this indicator alone).
• Source aliases:
  • time-varying weak CYP3A4 inducer – Savic 2022 Results wording (Savic_2022_voxelotor.R); Table 2 row ‘CYP3A4 inducer on CL/F, exp TH’. The source dataset column name is not published.
• Example models: Savic_2022_voxelotor.R (time-varying; exponential effect on voxelotor CL/F, exp(0.39 * CONMED_CYP3A4_IND_WEAK), i.e. 47% higher apparent clearance, 6.1 to 9.1 L/h; 14 of 264 PK-evaluable patients exposed, no moderate-inducer users, strong inducers prohibited by protocol).
• Notes: Reserved by name in the CONMED_CYP3A4_IND and CONMED_CYP3A4_IND_MOD register entries before this first use, so the name is not a fresh coinage. Per-model covariateData[[CONMED_CYP3A4_IND_WEAK]]$notes must document which agents the source classified as weak and the fraction exposed; record the gap when the paper does not name them (Savic 2022 does not). Do not reuse a weak-stratum coefficient for moderate or strong inducers: Savic 2022’s Discussion expects strong inducers to reduce voxelotor exposure substantially, an effect its data could not estimate.

CONMED_OATP1B_INH (canonical for concomitant OATP1B1 / OATP1B3 inhibitor coadministration indicator)

• Description: 1 = subject coadministered an inhibitor of the hepatic uptake transporters OATP1B1 (SLCO1B1) and/or OATP1B3 (SLCO1B3) during the observation interval, 0 = no concomitant OATP1B1/1B3 inhibitor. Inhibiting OATP1B-mediated hepatic uptake raises the systemic exposure of OATP1B substrates by reducing first-pass hepatic extraction and hepatic clearance; depending on how the source paper parameterises it the effect may land on apparent clearance or on relative bioavailability. Use this canonical when the source enters OATP1B-inhibitor coadministration as a binary indicator, regardless of which inhibitor potencies it pools into the 1 category.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no OATP1B1/1B3-inhibitor coadministration).
• Source aliases:
  • OATP1B1/1B3 inhibitor – Rich 2026 Table 1 row label (Rich_2026_momelotinib.R); the source dataset column name is not published.
• Example models: Rich_2026_momelotinib.R (log-additive effect on momelotinib relative bioavailability: exp(log(1.64) * CONMED_OATP1B_INH), i.e. 64% higher momelotinib, M21 and total-active-moiety Cavg,ss, the effect landing on F_rel rather than on CL/F because it scales the whole absorbed dose; consistent with the 57% rise in momelotinib AUCinf under single-dose rifampin that identified momelotinib as an OATP1B1/1B3 substrate).
• Notes: Transporter-class member of the CONMED_<pathway>_INH family, following CONMED_PGP_INH (P-glycoprotein), which is the founding transporter-class precedent, and the enzyme-class CONMED_CYP3A4_INH / CONMED_CYP2D6_INH. OATP1B deliberately covers OATP1B1 and OATP1B3 jointly: clinical inhibitors are not selective between the two paralogues, papers almost always report them as a pair, and the register already uses “OATP1B” as the pathway label in the CP_RIF_UM and CP_GDC_UM entries. Register separate CONMED_OATP1B1_INH / CONMED_OATP1B3_INH canonicals only if a source genuinely resolves the two. Distinct from CP_RIF_UM / CP_GDC_UM, which carry an instantaneous perpetrator concentration driving a mechanistic competitive-inhibition term rather than a static on/off flag; use those when the model needs the graded time course. Per-model covariateData[[CONMED_OATP1B_INH]]$notes must list the agents the source pooled into the 1 category, or record that the source does not say (Rich 2026 does not).

APOE4_COUNT (canonical for APOE-epsilon4 allele count)

• Description: Continuous individual-level APOE-epsilon4 allele count: 0 = non-carrier, 1 = heterozygous (one epsilon4 allele), 2 = homozygous (two epsilon4 alleles). Time-invariant (germline genotype). Models in the Alzheimer’s-disease-progression literature treat the 0 / 1 / 2 count as a continuous effect on baseline cognitive score and / or disease-progression slope, with the population-mean carrier-allele count used as the centring value (e.g., 0.72 in the Conrado 2014 CAMD cohort).
• Units: (count, 0 / 1 / 2 alleles per subject; population-mean centring value documented per-model)
• Type: continuous
• Scope: general
• Reference category: n/a (continuous). Models centre on the dataset / population mean APOE-epsilon4 count; document the centring value per-model in covariateData[[APOE4_COUNT]]$notes.
• Source aliases:
  • APOE4C – used directly in Conrado_2014_alzheimer.R. The “C” suffix in the source distinguishes the cleaned continuous APOE-epsilon4 count column (0 / 1 / 2 with unknown recoded to the population mean) from the upstream raw APOE4 column (0 = non-carrier, 1 = heterozygous, 2 = homozygous, 3 = unknown).
• Example models: Conrado_2014_alzheimer.R (centring 0.72; multiplicative effect on baseline ADAS-Cog and on disease-progression slope: factor = 1 + e * (APOE4_COUNT - 0.72) with e_blapoe4 = 0.0372 on baseline and e_slapoe4 = 0.195 on slope).
• Notes: APOE-epsilon4 carrier status is the strongest established genetic risk factor for late-onset Alzheimer’s disease; the allele-count form (rather than a binary carrier indicator) is preferred when the source paper distinguishes heterozygous from homozygous carriers. Future models that report only a binary carrier indicator (any-epsilon4 vs none) should register a separate canonical (APOE4_CARRIER) rather than overloading APOE4_COUNT. The unknown category (often recorded as APOE4 = 3 in CDISC datasets) is conventionally recoded by the source paper to the population-mean count to avoid dropping subjects; document the recoding rule used per-model.

APOE4_HET (canonical for APOE-epsilon4 heterozygote indicator)

• Description: 1 = subject carries exactly one APOE-epsilon4 allele (heterozygous); 0 = otherwise (non-carrier or homozygote). Time-fixed per subject (germline genotype). Paired with APOE4_HOM to encode the three-level APOE-epsilon4 genotype as two binary indicators with the non-carrier genotype as the reference, following the SLCO1B1_HAP15_HET / SLCO1B1_HAP15_HOM and CYP3A5_STAR1_HET / CYP3A5_STAR1_HOM pattern that APOE4_COUNT names as the correct route when a paper distinguishes the two carrier genotypes non-additively.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 = APOE-epsilon4 non-carrier. APOE4_HET and APOE4_HOM are mutually exclusive; a non-carrier has both set to 0.
• Source aliases:
  • APOEGEN (Monolix / CDISC categorical genotype column with levels Non-Carrier / Heterozygous / Homozygous): derive APOE4_HET = as.integer(APOEGEN == "Heterozygous").
  • APOE4 (raw CDISC coding 0 / 1 / 2 / 3 for non-carrier / heterozygous / homozygous / unknown): derive APOE4_HET = as.integer(APOE4 == 1).
• Example models: Cao_2026_lecanemab.R (multiplicative log-scale effect on the A-beta monomer aggregation rate kAggAO: exp(0.15 * APOE4_HET + 0.18 * APOE4_HOM), i.e. +16.2% for heterozygotes and +19.7% for homozygotes relative to non-carriers).
• Notes: Use APOE4_HET + APOE4_HOM (not APOE4_COUNT) whenever the source paper estimates separate, non-additive heterozygote and homozygote coefficients – the additive allele-count form cannot represent a model in which the homozygote effect is not twice the heterozygote effect. Use APOE4_CARRIER when the paper collapses both carrier genotypes into a single binary contrast. A paper may use more than one of the three codings on different parameters (Cao 2026 uses the _HET / _HOM pair on an amyloid aggregation rate and APOE4_CARRIER on the cognitive-scale scaling factors), so the three canonicals coexist rather than supersede one another. Founding example: Cao_2026_lecanemab.R.

APOE4_HOM (canonical for APOE-epsilon4 homozygote indicator)

• Description: 1 = subject carries two APOE-epsilon4 alleles (homozygous); 0 = otherwise (non-carrier or heterozygote). Time-fixed per subject (germline genotype). Companion to APOE4_HET; see that entry for the encoding rationale.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 = APOE-epsilon4 non-carrier.
• Source aliases:
  • APOEGEN (categorical genotype column): derive APOE4_HOM = as.integer(APOEGEN == "Homozygous").
  • APOE4 (raw CDISC coding 0 / 1 / 2 / 3): derive APOE4_HOM = as.integer(APOE4 == 2).
• Example models: Cao_2026_lecanemab.R (log-scale coefficient 0.18 on kAggAO; observed frequency 13.4-16.5% across the ADNI, Study 201 and Study 301 cohorts).
• Notes: See APOE4_HET. APOE-epsilon4 homozygosity is the strongest common genetic risk factor for late-onset Alzheimer’s disease and is separately labelled in the anti-amyloid antibody programmes because of its association with amyloid-related imaging abnormalities (ARIA), so papers frequently report it as its own stratum rather than folding it into an allele count. Founding example: Cao_2026_lecanemab.R.

APOE4_CARRIER (canonical for binary APOE-epsilon4 carrier indicator)

• Description: 1 = subject carries at least one APOE-epsilon4 allele (heterozygous or homozygous); 0 = non-carrier. Time-fixed per subject (germline genotype). This is the binary collapse of the three-level genotype, for papers that fit a single carrier-vs-non-carrier contrast rather than distinguishing heterozygotes from homozygotes.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 = APOE-epsilon4 non-carrier.
• Source aliases:
  • tAPOECarrier (Monolix transformed binary carrier covariate) – used in Cao_2026_lecanemab.R.
  • APOEGEN / APOE4 (three-level genotype): derive APOE4_CARRIER = as.integer(APOE4_HET == 1 | APOE4_HOM == 1), equivalently as.integer(APOE4_COUNT >= 1).
• Example models: Cao_2026_lecanemab.R (multiplicative log-scale effects on the CDR-SB and ADAS-Cog scaling factors: exp(-0.21 * APOE4_CARRIER) on alpha_cdr and exp(-0.097 * APOE4_CARRIER) on alpha_adas).
• Notes: Prefer APOE4_COUNT when the paper models an additive per-allele effect, APOE4_HET + APOE4_HOM when it estimates separate non-additive genotype effects, and APOE4_CARRIER when it fits only the binary contrast. Founding example: Cao_2026_lecanemab.R.

NAT2_RAPID (canonical for NAT2 rapid (fast) acetylator phenotype indicator)

• Description: 1 = subject is an NAT2 (arylamine N-acetyltransferase 2) rapid (fast) acetylator, i.e. carries two functional NAT2 alleles (any combination of *4, *11, *12, *13); 0 = subject is an intermediate or slow acetylator. Time-fixed per subject (germline genotype-derived phenotype). Paired with NAT2_SLOW to encode the full three-level slow / intermediate / rapid acetylator phenotype with two binary indicators; in that pairing the joint state NAT2_SLOW = 0, NAT2_RAPID = 0 denotes an intermediate acetylator.
• Units: (binary)
• Type: binary
• Scope: general – the trimodal NAT2 acetylator phenotype is the dominant source of isoniazid clearance variability and is re-fit in essentially every isoniazid popPK analysis; the _SLOW / _RAPID indicator pair is the reusable encoding whenever a cohort has enough rapid acetylators to identify a distinct rapid typical value.
• Reference category: 0 (intermediate or slow acetylator). Which level the source paper’s covariate model treats as its reference is a separate question and varies between papers – record it per model in covariateData[[NAT2_RAPID]]$reference_category and select the reference inside model() from the two indicators.
• Source aliases:
  • NAT2 (categorical with values "slow" / "intermediate" / "rapid", or "fast" for the rapid level): derive NAT2_RAPID = as.integer(NAT2 %in% c("rapid", "fast")).
  • ACETYL_RAPID / ACETYL_FAST (rapid-acetylator indicator already in source datasets) – same orientation as the canonical, no transformation.
• Example models:
  • Vinnard_2017_isoniazid.R (three-level phenotype with the slow acetylator as the source paper’s reference; nat2_int <- (1 - NAT2_SLOW) * (1 - NAT2_RAPID) derives the intermediate indicator and CL/F is scaled by 1 + 0.63 * nat2_int + 1.65 * NAT2_RAPID, i.e. rapid acetylators clear isoniazid 2.65-fold faster than slow acetylators).
  • Seng_2015_isoniazid.R (three-level phenotype with the intermediate acetylator as the joint reference; selects between three typical-value clearances).
  • Chen_2022_isoniazid_nat2class.R (genotype-class form with the rapid (*4/*4, the paper’s wt/wt) group as the reference; nat2_score <- (1 - NAT2_RAPID) * (1 + NAT2_SLOW) rebuilds the paper’s 0 / 1 / 2 score for wt/wt, m/wt and m/m, which enters as exp(-0.55 * score) on CL/F and exp(-0.47 * score) on the fraction of CL/F forming acetylisoniazid). Only the *5, *6 and *7 SNPs were typed, so the paper’s wt/wt is *4/*4 by construction.
• Notes: Pre-named by the NAT2_SLOW register entry, whose Notes block reserved this name for “future papers that distinguish rapid from intermediate (separately from slow)” so the three-level phenotype can be encoded with two binary indicators on the SLCO1B1_HAP15_HET / SLCO1B1_HAP15_HOM and CYP3A5_STAR1_HET / CYP3A5_STAR1_HOM pattern. Do not use this canonical alone as a slow-versus-nonslow proxy – NAT2_RAPID = 0 pools intermediate with slow, which is the opposite pooling from the NAT2_SLOW-only convention (where 0 pools intermediate with rapid), and the two pool the clinically distinct group in each case. A model that carries only one of the pair is asserting that its cohort could not resolve the third level; a model that carries both must derive the intermediate indicator in model() rather than registering a third column. Distinct from any raw genotype-string column, which carries allele information rather than the derived metabolic phenotype. Rapid acetylators are the under-exposed group at standard isoniazid doses (the treatment-failure risk), mirroring CYP2D6_UM at the ultrarapid end of the CYP2D6 spectrum, whereas NAT2_SLOW = 1 marks the over-exposed hepatotoxicity-risk group.

NAT2_SLOW (canonical for NAT2 slow-acetylator phenotype indicator)

• Description: 1 = subject is an NAT2 (arylamine N-acetyltransferase 2) slow acetylator, defined by carrying two reduced-function NAT2 SNP alleles (homozygous variant for one or more of the canonical slow-acetylator SNPs rs1801279, rs1801280, rs1799930, rs1799931, OR heterozygous for two or more of those SNPs); 0 = subject is an intermediate or rapid acetylator (heterozygous for at most one of the canonical SNPs, or wild-type homozygous for all). Time-fixed per subject (germline genotype-derived phenotype). The intermediate and rapid (fast) phenotypes are pooled into the 0 category because the Horita 2018 source paper found no significant differences in t1/2, CL/F, or AUC0-8 between rapid and intermediate genotypes and combined them as the “nonslow” group; this pooling is the standard convention in the NAT2-aware antituberculosis-isoniazid popPK literature.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (nonslow: intermediate or rapid acetylator pooled). Reference category in the source-paper IIV model is the SLOW group; the canonical-orientation convention here is NAT2_SLOW = 1 for slow, consistent with the _SLOW naming, and the source-paper-reported coefficient signs map back to typical-value parameters via (1 - NAT2_SLOW) selection inside model where needed.
• Source aliases:
  • NAT2 (categorical with values "slow" / "intermediate" / "rapid" or 0 / 1 / 2): derive NAT2_SLOW = as.integer(NAT2 == "slow") (or as.integer(NAT2 == 0) depending on the source’s level coding); the intermediate and rapid levels collapse to NAT2_SLOW = 0.
  • NAT2_SS (slow-vs-not-slow indicator already in source datasets) – same orientation as the canonical, no transformation.
  • ACETYL_SLOW (slow-acetylator indicator) – same orientation as the canonical, no transformation.
• Example models: Horita_2018_isoniazid.R (selects between two typical-value clearances via lcl_slow * NAT2_SLOW + lcl_nonslow * (1 - NAT2_SLOW) and pairs each typical value with its own IIV variance; reproduces the source paper’s separate CL/F slow = 4.44 L/h and CL/F nonslow = 8.08 L/h typical-value estimates with separate omegas 0.105 and 0.230 respectively); Chen_2022_isoniazid_nat2class.R (with NAT2_RAPID; NAT2_SLOW = 1 marks the paper’s m/m group, two of the *5 / *6 / *7 alleles, score 2 in exp(-0.55 * score) on CL/F).
• Notes: The NAT2 (rs1208 / rs1041983 / rs1801279 / rs1801280 / rs1799929 / rs1799930 / rs1799931 / rs1208) gene encodes the cytosolic arylamine N-acetyltransferase 2 enzyme responsible for the major isoniazid metabolic pathway (acetylation to acetyl-isoniazid); slow acetylators have substantially reduced isoniazid clearance, higher Cmax, and higher AUC than intermediate or rapid acetylators, with documented impact on both efficacy (treatment failure in rapid acetylators given standard doses) and toxicity (hepatotoxicity in slow acetylators given high doses). The slow / intermediate / rapid trimodal phenotype is conventionally collapsed to slow vs nonslow in popPK models when the cohort lacks enough rapid acetylators to identify a distinct rapid typical value, OR when the rapid and intermediate phenotypes are statistically indistinguishable in the data (Horita 2018 cohort: 51 slow / 50 intermediate / 12 fast). Papers that distinguish rapid from intermediate (separately from slow) pair this indicator with the companion canonical NAT2_RAPID so the three-level phenotype is encoded with two binary indicators on the SLCO1B1_HAP15_HET / SLCO1B1_HAP15_HOM and CYP3A5_STAR1_HET / CYP3A5_STAR1_HOM patterns; in that pairing the joint state (both = 0) denotes an intermediate acetylator, and the intermediate indicator is derived inside model() rather than registered as a third column (see Vinnard_2017_isoniazid.R and Seng_2015_isoniazid.R). The _SLOW orientation (slow = 1) follows the clinically meaningful axis (slow acetylators are the at-risk group for isoniazid hepatotoxicity and the higher-AUC group for treatment outcomes), paralleling the CYP2D6_PM = 1 orientation for the poor-metabolizer end of the CYP2D6 phenotype spectrum. Distinct from any genotype-string column (which carries the raw allele information); NAT2_SLOW captures the derived metabolic phenotype only. Ratified canonically on 2026-05-26 alongside the Horita 2018 isoniazid extraction.

SNP_NAT2_RS1801280_C_COUNT (**canonical for NAT2*5 (341T>C, rs1801280) variant C-allele count**)

• Description: Individual-level count of NAT2 rs1801280 variant alleles (c.341T>C, I114T), the defining SNP of the NAT2*5 slow-acetylator allele family: 0 = 341T/T, 1 = T/C, 2 = C/C. Time-invariant (germline genotype). Where a study types only the *5-, *6- and *7-defining SNPs, this count equals the number of *5 alleles.
• Units: (count, 0/1/2 alleles per subject)
• Type: continuous
• Scope: general
• Reference category: n/a (continuous). A count of 0 is the natural reference.
• Source aliases:
  • M341 / *5 allele score – Chen 2022 Table 3 and Methods ‘Covariates’ method 3 (0 / 1 / 2 for w/w, m/w, m/m at the *5 allele). The paper writes the change as ‘C341 -> T’; the variant allele of *5 is 341C. Same orientation as the canonical, no value transformation.
  • NAT2 diplotype string ("*4/*5", "*5/*7", …): count the *5 alleles.
• Example models: Chen_2022_isoniazid.R (exponential per-allele effects exp(-0.77 * count) on isoniazid CL/F and exp(-0.72 * count) on the fraction of CL/F forming acetylisoniazid).
• Notes: Member of the SNP_<GENE>_RS<rsid>_<allele>_COUNT family, alongside SNP_NAT2_RS1799930_A_COUNT (*6) and SNP_NAT2_RS1799931_A_COUNT (*7). Use the per-SNP counts when a paper gives each slow allele its own effect; when it pools them into an acetylator phenotype use NAT2_SLOW / NAT2_RAPID instead. The three counts are not independent of those phenotypes: their sum is the number of slow alleles (0 rapid, 1 intermediate, 2 slow) under the usual assumption that each allele carries one defining SNP.

SNP_NAT2_RS1799930_A_COUNT (**canonical for NAT2*6 (590G>A, rs1799930) variant A-allele count**)

• Description: Individual-level count of NAT2 rs1799930 variant alleles (c.590G>A, R197Q), the defining SNP of the NAT2*6 slow-acetylator allele family: 0 = 590G/G, 1 = G/A, 2 = A/A. Time-invariant (germline genotype). Where a study types only the *5-, *6- and *7-defining SNPs, this count equals the number of *6 alleles.
• Units: (count, 0/1/2 alleles per subject)
• Type: continuous
• Scope: general
• Reference category: n/a (continuous). A count of 0 is the natural reference.
• Source aliases:
  • M590 / *6 allele score – Chen 2022 Table 3 and Methods ‘Covariates’ method 3. Same orientation as the canonical, no value transformation.
  • NAT2 diplotype string: count the *6 alleles.
• Example models: Chen_2022_isoniazid.R (exp(-0.60 * count) on isoniazid CL/F, exp(-0.45 * count) on the fraction forming acetylisoniazid).
• Notes: See SNP_NAT2_RS1801280_C_COUNT for the relationship to the NAT2 phenotype indicators.

SNP_NAT2_RS1799931_A_COUNT (**canonical for NAT2*7 (857G>A, rs1799931) variant A-allele count**)

• Description: Individual-level count of NAT2 rs1799931 variant alleles (c.857G>A, G286E), the defining SNP of the NAT2*7 slow-acetylator allele family: 0 = 857G/G, 1 = G/A, 2 = A/A. Time-invariant (germline genotype). Where a study types only the *5-, *6- and *7-defining SNPs, this count equals the number of *7 alleles.
• Units: (count, 0/1/2 alleles per subject)
• Type: continuous
• Scope: general
• Reference category: n/a (continuous). A count of 0 is the natural reference.
• Source aliases:
  • M870 / M803 / *7 allele score – Chen 2022 Table 3 row and footnote respectively; both misprint the 857 position given in the paper’s Methods. Same orientation as the canonical, no value transformation.
  • NAT2 diplotype string: count the *7 alleles.
• Example models: Chen_2022_isoniazid.R (exp(-0.29 * count) on isoniazid CL/F, exp(-0.14 * count) on the fraction forming acetylisoniazid).
• Notes: Chen 2022 estimates the smallest per-allele effect for *7 (one copy leaves CL/F at 74.8 percent of *4/*4, against 46.3 and 54.9 for *5 and *6). See SNP_NAT2_RS1801280_C_COUNT for the relationship to the NAT2 phenotype indicators.

FCGR3A_VV (canonical for FCGR3A 158 V/V homozygote indicator)

• Description: 1 = subject is homozygous for valine at amino-acid position 158 of the FcgammaRIIIa receptor (V/V), encoded by the rs396991 polymorphism in the FCGR3A gene; 0 = otherwise (heterozygote V/F or homozygote F/F pooled). The dominant V/V vs (V/F + F/F) grouping is the encoding used in the Aguiar 2021 source paper after testing dominant and recessive groupings during covariate model building.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 = V/F heterozygote or F/F homozygote (combined).
• Source aliases:
  • FCGR3A (genotype string, e.g., "V/V" / "V/F" / "F/F"): derive FCGR3A_VV = as.integer(FCGR3A == "V/V").
  • rs396991 (raw allele coding, often "AA" / "AC" / "CC" or "GG" / "GT" / "TT" depending on assay strand): map V allele -> 1, F allele -> 0 with the assay-specific allele convention; derive FCGR3A_VV = as.integer(genotype is V-homozygous).
• Example models: Aguiar_2021_ustekinumab.R (Aguiar 2021 Table 2 footnote; logit-scale effect on subcutaneous bioavailability F: 88.8% in V/V vs 71.0% in V/F + F/F).
• Notes: rs396991 (FCGR3A 158V>F) is a well-studied pharmacogenetic polymorphism affecting FcgammaRIIIa-IgG affinity and has been associated with response to several IgG monoclonal antibodies (rituximab, infliximab, ustekinumab). The V allele is the higher-affinity variant. Document the assay-strand allele convention used in the source paper in covariateData[[FCGR3A_VV]]$notes. Future models that use a recessive (F/F vs V/* combined) or codominant (additive 0/1/2) coding should register a separate canonical (e.g., FCGR3A_FF, FCGR3A_VCT for V-allele count) rather than overloading FCGR3A_VV.

DSBAL_TT (canonical for DsbA-L (GSTK1) rs1917760 T/T genotype indicator)

• Description: Indicator for the DsbA-L (GSTK1) rs1917760 -1308G>T T/T genotype; 1 = subject carries the T/T genotype, 0 = subject carries the G/G or G/T genotype (the pooled reference). Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (G/G or G/T, pooled).
• Source aliases:
  • DsbAL (three-level column 0 = G/G, 1 = G/T, 2 = T/T) – used in Oniki_2018_bmi.R; derive DSBAL_TT = as.integer(DsbAL == 2). G/G and G/T are pooled because the T/T state is the functional minor-allele genotype most strongly associated with elevated BMI.
• Example models: Oniki_2018_bmi.R (additive +1.5 kg/m^2 shift on the typical BMI for T/T carriers vs the pooled G/G-or-G/T reference, Oniki 2018 Eq. 1).
• Notes: General scope because the genotype is a stable germline marker; the GSTK1 (DsbA-L) rs1917760 polymorphism has been associated with adiposity / metabolic phenotypes. Pools the G/G homozygous and G/T heterozygous strata into the reference following the dominant-for-the-minor-allele encoding used by Oniki 2018; future extractions that resolve a separate G/T heterozygote effect should register a paired DSBAL_GT indicator on the SLCO1B1_HAP15_HET / SLCO1B1_HAP15_HOM precedent.

PNPLA3_CG (canonical for PNPLA3 rs738409 C/G heterozygote indicator)

• Description: Indicator for the PNPLA3 rs738409 c.444C>G (I148M) C/G heterozygote genotype; 1 = subject carries the C/G genotype, 0 = subject does not carry the C/G genotype. Paired with PNPLA3_GG to encode the three-level rs738409 genotype with two binary indicators (C/C is the reference when both are 0). Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (C/C wild-type, when PNPLA3_GG is also 0).
• Source aliases:
  • PNPLA3 (three-level column 0 = C/C, 1 = C/G, 2 = G/G) – used in Oniki_2018_nafld_risk.R; derive PNPLA3_CG = as.integer(PNPLA3 == 1).
• Example models: Oniki_2018_nafld_risk.R (multiplicative factor 0.761 on the (BMI50 - 17) half-saturation offset of the logit-of-NAFLD sigmoid for C/G heterozygotes vs the C/C reference, Oniki 2018 Eq. 4 / Figure 2c; closer to 1 than the G/G factor, consistent with an additive allele-dose effect).
• Notes: General scope because the genotype is a stable germline marker; PNPLA3 rs738409 (I148M) is the best-established common genetic risk variant for non-alcoholic fatty liver disease (NAFLD). Paired with PNPLA3_GG (homozygote) following the SLCO1B1_HAP15_HET / SLCO1B1_HAP15_HOM paired-binary precedent for a three-level genotype where each stratum carries a distinct typical-value effect.

PNPLA3_GG (canonical for PNPLA3 rs738409 G/G homozygote indicator)

• Description: Indicator for the PNPLA3 rs738409 c.444C>G (I148M) G/G homozygote genotype; 1 = subject carries the G/G genotype, 0 = subject does not carry the G/G genotype. Paired with PNPLA3_CG to encode the three-level rs738409 genotype with two binary indicators (C/C is the reference when both are 0). Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (C/C wild-type, when PNPLA3_CG is also 0).
• Source aliases:
  • PNPLA3 (three-level column 0 = C/C, 1 = C/G, 2 = G/G) – used in Oniki_2018_nafld_risk.R; derive PNPLA3_GG = as.integer(PNPLA3 == 2).
• Example models: Oniki_2018_nafld_risk.R (multiplicative factor 0.592 on the (BMI50 - 17) half-saturation offset of the logit-of-NAFLD sigmoid for G/G homozygotes vs the C/C reference, Oniki 2018 Eq. 4 / Figure 2c).
• Notes: General scope. Companion homozygote indicator to PNPLA3_CG; see PNPLA3_CG notes for the joint three-level usage and reference category.

UGT1A1_STAR28_HOM (**canonical for UGT1A1*28 7/7 homozygous-variant genotype indicator**)

• Description: Binary germline genotype indicator for the UGT1A128 homozygous-variant group. 1 = subject carries two 28 alleles, i.e. the 7/7 diplotype at the (TA)n dinucleotide repeat in the UGT1A1 promoter TATA box (rs8175347/rs3064744); 0 = otherwise (the union of 6/6 wild-type homozygotes and 6/7 heterozygotes). Time-fixed per subject. The extra TA repeat reduces UGT1A1 transcription, so 7/7 is the reduced-glucuronidation stratum; it is the genotype underlying Gilbert syndrome and the historical predictor of SN-38 exposure and hematological toxicity with nonliposomal irinotecan.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (UGT1A1*28 non-7/7, i.e. 6/6 or 6/7).
• Source aliases:
  • UGT1A1*28 == homozygous (vs UGT1A1*28 non-homozygous) – Adiwijaya 2017 Table S7 theta row label.
  • Non 7/7 / 7/7 – Adiwijaya 2017 Table 1 genotype-count rows.
• Example models: Adiwijaya_2017_irinotecan_liposomal.R (exponential effect of -1.46e-05 on SN-38 clearance – numerically indistinguishable from zero. The coefficient is retained only because the paper’s full-covariate approach kept it in the final model; the paper’s own reading is that UGT1A1*28 is not a significant covariate for SN-38 with liposomal irinotecan, the 7/7 clearance being 1.000-times the non-7/7 clearance, in deliberate contrast to nonliposomal irinotecan. The mechanistic explanation offered is that liposomal encapsulation throttles the SN-38 load reaching UGT enzymes, so even reduced-activity genotypes keep up.).
• Notes: Follows the <GENE>_STAR<allele>_<HET|HOM> shape established by UGT2B15_STAR2_HET / UGT2B15_STAR2_HOM; the star-allele form is used in preference to the SNP_<GENE>_RS<rsid> family because the source papers report diplotypes (6/6, 6/7, 7/7) rather than rsID genotypes, and because UGT1A1*28 has two rsIDs in circulation for the same repeat polymorphism. A companion UGT1A1_STAR28_HET should be registered by the first model that needs the 6/7 stratum separately; Adiwijaya 2017 pools 6/6 and 6/7 into the reference in its final model and only separates them in a sensitivity analysis (its Table S11), which found the 6/7 and 7/7 clearances within 0.0% and 2.7% of 6/6.

UGT2B7_211GG (**canonical for UGT2B7 211G>T (rs7438135 / UGT2B7*2 Ala71Ser) homozygous G/G genotype indicator**)

• Description: 1 = subject is homozygous for the wild-type (ancestral) guanine at nucleotide 211 of the UGT2B7 gene, corresponding to alanine at amino-acid position 71 (Ala71/Ala71) in the substrate-binding N-terminal half of the UGT2B7 enzyme; 0 = otherwise (211GT heterozygote or 211TT homozygote). Encoded by the UGT2B7 211G>T single-nucleotide polymorphism (rs7438135, also referred to as the UGT2B7*2 variant). Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (any non-G/G genotype: 211GT or 211TT).
• Source aliases:
  • UGT2B7 211GG / UGT2B7-211GG / 211GG – used in Yu_2017_mycophenolic_acid.R (paper Table 5 column headers; one of three binary group-membership indicators reconstructed from the paper’s ordinal column “UGT2B7 genotype”).
• Example models: Yu_2017_mycophenolic_acid.R (one of three UGT2B7-211 binary indicators that reconstruct the paper’s ordinal column UGT2B7 genotype in {GT=1, GG=2, TT=3}; the model() block computes ugt2b7_211_code = UGT2B7_211GT * 1 + UGT2B7_211GG * 2 + UGT2B7_211TT * 3 and applies V1/F = 14.7 + 7.72 * ugt2b7_211_code per Yu 2017 Table 3 / page 1574 final-model formula).
• Notes: rs7438135 (UGT2B7 c.211G>T, p.Ala71Ser) is a coding-region SNP in the UGT2B7 N-terminal substrate-binding half; the T (Ser71) allele is the variant. Allele frequencies vary by ethnicity (the T allele is most common in Asian populations in Yu 2017’s Chinese renal-transplant cohort: 51/101 GT + 8/101 TT = 59/101 PK evaluations carrying at least one T allele). Future papers that use a dominant (any-G vs T/T) or recessive (G/G vs any-T) grouping should still register the same three binary canonicals UGT2B7_211GG / UGT2B7_211GT / UGT2B7_211TT in the input data so that downstream simulations can reconstruct the paper’s chosen grouping. Distinct from UGT1A9 promoter / 5’-UTR variants (Yu 2017 also genotyped UGT1A9*22 but did not find it significant; no UGT1A9 canonical is being proposed here).

UGT2B7_211GT (**canonical for UGT2B7 211G>T (rs7438135 / UGT2B7*2 Ala71Ser) heterozygous G/T genotype indicator**)

• Description: 1 = subject is heterozygous at nucleotide 211 of the UGT2B7 gene, carrying one wild-type G allele (Ala71) and one variant T allele (Ser71); 0 = otherwise (211GG homozygote or 211TT homozygote). Encoded by the UGT2B7 211G>T single-nucleotide polymorphism (rs7438135, also referred to as the UGT2B7*2 variant). Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (any non-G/T genotype: 211GG or 211TT).
• Source aliases:
  • UGT2B7 211GT / UGT2B7-211GT / 211GT – used in Yu_2017_mycophenolic_acid.R (paper Table 5).
• Example models: Yu_2017_mycophenolic_acid.R (one of three UGT2B7-211 binary indicators; see UGT2B7_211GG notes for the paper’s ordinal-code reconstruction).
• Notes: Companion canonical to UGT2B7_211GG and UGT2B7_211TT; see UGT2B7_211GG notes for variant biology and allele-frequency context.

UGT2B7_211TT (**canonical for UGT2B7 211G>T (rs7438135 / UGT2B7*2 Ala71Ser) homozygous T/T genotype indicator**)

• Description: 1 = subject is homozygous for the variant thymine at nucleotide 211 of the UGT2B7 gene, corresponding to serine at amino-acid position 71 (Ser71/Ser71); 0 = otherwise (211GG homozygote or 211GT heterozygote). Encoded by the UGT2B7 211G>T single-nucleotide polymorphism (rs7438135, UGT2B7*2 variant). Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (any non-T/T genotype: 211GG or 211GT).
• Source aliases:
  • UGT2B7 211TT / UGT2B7-211TT / 211TT – used in Yu_2017_mycophenolic_acid.R (paper Table 5).
• Example models: Yu_2017_mycophenolic_acid.R (one of three UGT2B7-211 binary indicators; see UGT2B7_211GG notes for the paper’s ordinal-code reconstruction).
• Notes: Companion canonical to UGT2B7_211GG and UGT2B7_211GT; see UGT2B7_211GG notes for variant biology and allele-frequency context.

UGT2B7_M161CT (canonical for UGT2B7 -161C>T (rs7668258) heterozygous C/T genotype indicator)

• Description: 1 = subject is heterozygous at nucleotide -161 of the UGT2B7 gene promoter, carrying one wild-type C allele and one variant T allele; 0 = otherwise (M161CC homozygote or M161TT homozygote). Encoded by the UGT2B7 -161C>T single-nucleotide polymorphism (rs7668258). Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (any non-C/T genotype: M161CC or M161TT). The reference used in effect estimation is the M161CC homozygous wild-type stratum, encoded when all three UGT2B7_M161* indicators equal 0.
• Source aliases:
  • UGT2B7 -161CT / UGT2B7 -161 C/T / -161CT – used in Milosheska_2016_lamotrigine.R (paper Table 4 covariate equation UGT2B7 -161CT vs CC row).
• Example models: Milosheska_2016_lamotrigine.R (multiplicative effect on parent apparent clearance: cl *= (1 - 0.0358 * UGT2B7_M161CT); heterozygotes have -3.6% lower CL relative to the CC homozygous wild-type reference, Milosheska 2016 Table 4 row UGT2B7 -161C>T genotype CT vs CC).
• Notes: Companion canonical to UGT2B7_M161CC and UGT2B7_M161TT. See UGT2B7_M161CC for variant biology and the M-for-minus naming rationale.

UGT2B7_M161TT (canonical for UGT2B7 -161C>T (rs7668258) homozygous T/T genotype indicator)

• Description: 1 = subject is homozygous for the variant thymine at nucleotide -161 of the UGT2B7 gene promoter; 0 = otherwise (M161CC homozygote or M161CT heterozygote). Encoded by the UGT2B7 -161C>T single-nucleotide polymorphism (rs7668258). Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (any non-T/T genotype: M161CC or M161CT). The reference used in effect estimation is the M161CC homozygous wild-type stratum, encoded when all three UGT2B7_M161* indicators equal 0.
• Source aliases:
  • UGT2B7 -161TT / UGT2B7 -161 T/T / -161TT – used in Milosheska_2016_lamotrigine.R (paper Table 4 covariate equation UGT2B7 -161TT vs CC row).
• Example models: Milosheska_2016_lamotrigine.R (multiplicative effect on parent apparent clearance: cl *= (1 - 0.204 * UGT2B7_M161TT); homozygous variant carriers have -20.4% lower CL relative to the CC homozygous wild-type reference, Milosheska 2016 Table 4 row UGT2B7 -161C>T genotype TT vs CC).
• Notes: Companion canonical to UGT2B7_M161CC and UGT2B7_M161CT. Reduced UGT2B7 transcriptional activity attributable to the -161 T allele is one of the biological interpretations offered by Milosheska 2016 Discussion paragraph 5 for the lower lamotrigine glucuronidation observed in TT carriers.

UGT2B7_372AG (canonical for UGT2B7 372A>G (rs28365063) heterozygous A/G genotype indicator)

• Description: 1 = subject is heterozygous at nucleotide 372 of the UGT2B7 coding sequence, carrying one wild-type A allele and one variant G allele; 0 = otherwise (372AA homozygote or 372GG homozygote). Encoded by the UGT2B7 372A>G single-nucleotide polymorphism (rs28365063). Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (any non-A/G genotype: 372AA or 372GG). The reference used in effect estimation is the 372AA homozygous wild-type stratum, encoded when all three UGT2B7_372* indicators equal 0.
• Source aliases:
  • UGT2B7 372AG / UGT2B7 372 A/G / 372AG – used in Milosheska_2016_lamotrigine.R (paper Table 4 covariate equation UGT2B7 372AG vs AA row).
• Example models: Milosheska_2016_lamotrigine.R (multiplicative effect on parent apparent clearance: cl *= (1 + 0.194 * UGT2B7_372AG); heterozygotes have +19.4% higher CL relative to the AA homozygous wild-type reference, Milosheska 2016 Table 4 row UGT2B7 372 A > G genotype AG vs AA).
• Notes: Companion canonical to UGT2B7_372AA and UGT2B7_372GG. See UGT2B7_372AA for variant biology.

UGT2B7_372GG (canonical for UGT2B7 372A>G (rs28365063) homozygous G/G genotype indicator)

• Description: 1 = subject is homozygous for the variant guanine at nucleotide 372 of the UGT2B7 coding sequence; 0 = otherwise (372AA homozygote or 372AG heterozygote). Encoded by the UGT2B7 372A>G single-nucleotide polymorphism (rs28365063). Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (any non-G/G genotype: 372AA or 372AG). The reference used in effect estimation is the 372AA homozygous wild-type stratum, encoded when all three UGT2B7_372* indicators equal 0.
• Source aliases:
  • UGT2B7 372GG / UGT2B7 372 G/G / 372GG – used in Milosheska_2016_lamotrigine.R (paper Table 4 covariate equation UGT2B7 372GG vs AA row).
• Example models: Milosheska_2016_lamotrigine.R (multiplicative effect on parent apparent clearance: cl *= (1 + 1.17 * UGT2B7_372GG); homozygous variant carriers have +117% higher CL relative to the AA homozygous wild-type reference, Milosheska 2016 Table 4 row UGT2B7 372 A > G genotype GG vs AA).
• Notes: Companion canonical to UGT2B7_372AA and UGT2B7_372AG. The magnitude of the GG effect is by far the largest single-variant effect on lamotrigine apparent clearance identified in Milosheska 2016; the GG homozygote frequency in the paper’s Slovenian cohort was 2.0% (2 of 99 subjects), so per-model covariateData[[UGT2B7_372GG]]$notes should record the small-N caveat.

CYP2D6_PM_IM (canonical for pooled CYP2D6 poor-or-intermediate-metabolizer phenotype indicator)

• Description: 1 = subject is a CYP2D6 poor metabolizer OR intermediate metabolizer, pooled because the source paper tested the pooling and could not reject it; 0 = subject is an extensive metabolizer OR an ultrarapid metabolizer, likewise pooled. Time-fixed per subject (germline genotype-derived phenotype). This is the four-into-two collapse of the PM / IM / EM / UM phenotype onto a single binary axis carrying a single covariate coefficient.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (extensive OR ultrarapid metabolizer – the pooled high-activity stratum).
• Source aliases:
  • CYPD – Kastrissios 2006 equation 7 (“CYPD is an indicator variable taking the value 0 for subjects with the extensive/ultrafast CYP 2D6 phenotype and the value 1 for subjects with the poor/intermediate phenotype”); the paper’s Table II phenotype codes are 1 = extensive, 2 = intermediate, 3 = poor, 4 = ultrafast, so CYP2D6_PM_IM = 1 for codes 2 and 3 and 0 for codes 1 and 4.
• Example models: Kastrissios_2006_apricoxib.R (founding example; exponential effect on apparent oral clearance exp(e_cyp2d6_pmim_cl * CYP2D6_PM_IM) with e_cyp2d6_pmim_cl = -1.01, i.e. pooled poor-or-intermediate metabolizers have 1 - exp(-1.01) = 63.6% lower CL/F than the pooled extensive-or-ultrarapid reference, matching Kastrissios 2006 Discussion “63.6% lower apparent oral clearance”; Table IV row KCL/F-CYP2D6 (poor and intermediate metabolizer)).
• Notes: Exact name-for-name mirror of the already-ratified CYP2C9_PM_IM into the CYP2D6_<phenotype> binary-indicator family (CYP2D6_PM, CYP2D6_EM, CYP2D6_IM), registered because that family previously covered only the two pairings the CYP2D6_IM Notes enumerate – CYP2D6_PM + CYP2D6_EM with IM as the implicit reference (Sherwin 2012), and CYP2D6_PM + CYP2D6_IM with EM + UM as the pooled reference (Nguyen 2025). Kastrissios 2006 needs a third arrangement, PM and IM pooled INTO the 1-level, which neither pairing expresses with a single coefficient. Prefer this canonical over reusing CYP2D6_PM + CYP2D6_IM with a shared coefficient whenever the source paper estimated ONE theta for the pooled group: the pair would force a data assembler to split a group the paper deliberately never resolved, would admit the nonsensical state of both indicators = 1 (doubling the effect), and would misrepresent one estimated theta as two columns. Conversely, a source that estimates separate PM and IM effects must use the pair, not this canonical. The pooling in the founding paper is an explicitly tested modelling decision rather than a reporting accident: Kastrissios 2006 Table III run 12 “combine poor and intermediate CYP 2D6 status… Objective function not significantly different… indicating that poor and intermediate CYP 2D6 status can be combined” and run 17 “combine extensive and ultrafast CYP 2D6 status… not significantly different”, motivated by the Discussion’s note that intermediate and ultrafast metabolizers were fewer than 4% of treated subjects so their separate estimates “may have been unreliable”. Distinct from the continuous CYP2D6 activity-score canonical, which is the right choice when a source reports a probe-derived activity number.

ADA_POS (canonical for anti-drug antibody positive status indicator)

• Description: 1 = antidrug-antibody-positive, 0 = ADA-negative.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (ADA-negative).
• Source aliases:
  • ADA (semantically “ever positive”) – used in Zhu_2017_lebrikizumab.R. When translating from a paper that uses ADA as “ever positive,” verify the time-frame matches ADA_POS semantics before renaming.
  • ADA (time-varying positivity, primary covariate in Xu 2019) – used in Xu_2019_sarilumab.R.
  • NAB (neutralizing antibody positive – used in Petrov_2024_romiplostim.R). Strictly a subset of total ADA-positive (ADA antibodies that neutralize the drug’s biological effect). Document per-model when the source assay measured NAB only and the canonical column thus excludes binding-only ADA.
  • ATAPOSNEW (ADA-positive in the newer/updated-assay cohort) – used in Suri_2018_brentuximab.R as the modern-assay arm of a two-era ADA decomposition.
  • ADA_POSNEW (retired intermediate name; renamed to ADA_POS on 2026-04-29 for consistency across single- and multi-assay models).
• Example models: Clegg_2024_nirsevimab.R, Hu_2026_clesrovimab.R, Petrov_2024_romiplostim.R, Suri_2018_brentuximab.R (multi-assay; paired with ADA_POSOLD and ADA_MISSING; cl *= (1 + 0.125 * ADA_POS)), Xu_2019_sarilumab.R.
• Notes: In single-assay studies this is a straightforward binary. In studies pooling data across assay generations (different sensitivity / drug-tolerance characteristics), ADA_POS represents modern/current-assay positivity; companion indicators ADA_POSOLD and ADA_MISSING capture historical-assay-positive and missing-result sub-groups respectively. All three are mutually exclusive; reference is ADA-negative (all three = 0). Distinct from the continuous ADA quantities [[ADA_TITER]] (dimensionless dilution factor / assay titer) and [[CONC_ADA_NGML]] (calibrated mass concentration in ng/mL); a paper’s ADA_POS column is frequently derived from one of those by applying an assay cutoff, so when both are available prefer the continuous column if the final model estimated a continuous effect.

ADA_POSOLD (canonical for ADA-positive in older-assay study indicator)

• Description: 1 = subject is anti-drug-antibody-positive in a study that used the older lower-sensitivity, lower-drug-tolerance ADA assay; 0 = otherwise.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (ADA-negative or not in an older-assay study). Mutually exclusive with ADA_POS and ADA_MISSING.
• Source aliases:
  • ATAPOSOLD – used in Suri_2018_brentuximab.R.
• Example models: Suri_2018_brentuximab.R (multiplicative additive effect on ADC clearance: cl *= (1 + 0.177 * ADA_POSOLD)).
• Notes: Companion to ADA_POS (multi-assay form); see that entry’s Notes for the decomposition rationale. The “newer” vs “older” assay split is paper-specific (Suri 2018 newer assay: sensitivity 23.573 ng/mL, drug tolerance 25 ug/mL; older assay: sensitivity 4 ng/mL, drug tolerance 3,125 ng/mL). Time-varying once positive.

ADA_MISSING (canonical for ADA-result-missing indicator)

• Description: 1 = ADA value is missing (subject did not have a measured ADA result, distinct from a measured negative); 0 = ADA result is reported (positive or negative).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (ADA result reported). Mutually exclusive with ADA_POS and ADA_POSOLD.
• Source aliases:
  • ATAMISSING – used in Suri_2018_brentuximab.R.
• Example models: Suri_2018_brentuximab.R (multiplicative additive effect on ADC clearance: cl *= (1 + 0.192 * ADA_MISSING)).
• Notes: Used when a substantial fraction of the pooled cohort has no ADA measurement (Suri 2018: 205 of 380 patients) and the modeler retains ADA-missing as a separate level rather than collapsing missingness onto the ADA-negative reference. The non-zero positive estimate of e_adam_adc_cl indicates ADA-missing patients are not exchangeable with the ADA-negative reference – interpret with caution given the missingness mechanism is not random. Distinct from the continuous ADA quantities [[ADA_TITER]] and [[CONC_ADA_NGML]]; a missing-result indicator has no continuous analogue, so a model that carries a continuous ADA column and also has missing measurements needs an explicit per-model imputation rule rather than this flag.

ADA_TITER (canonical for continuous antidrug-antibody titer/titre)

• Description: Continuous antidrug-antibody titer/titre (time-varying; matched in time to the PK sample). Covers both the British-spelling reciprocal-dilution convention (ADA_TITRE, with ADA_TITRE = 1 for ADA-negative so log_e(1) = 0 cancels a log-linear effect) and the American-spelling linear-titer convention (ADA_TITER, with ADA_TITER = 0 for ADA-negative). The per-model covariateData[[ADA_TITER]]$description and notes must state which zero-encoding convention is in force so the covariate column cannot be misinterpreted.
• Units: Reciprocal dilution (e.g., 10, 20, 40,…, 2560) OR assay units (log2 or arbitrary) – document per-model in covariateData[[ADA_TITER]]$units.
• Type: continuous
• Scope: general
• Reference category: n/a – ADA-negative encoded per-model (see zero-encoding note).
• Source aliases:
  • ADA_TITRE – British spelling (reciprocal-dilution convention; 1 for negative).
  • ADA titre – British spelling long form.
  • ADAT – used in Moein_2022_etrolizumab.R (American linear-titer convention; 0 for negative).
• Example models: Jackson_2022_ixekizumab.R (reciprocal-dilution reference convention with ADA_TITER = 1 for negatives and (1 + coef * log_e(ADA_TITER)) on CL), Moein_2022_etrolizumab.R (linear-titer convention with ADA_TITER = 0 for negatives and exp(theta * ADA_TITER) on CL, per-unit-titer theta = 0.0365), Robbie_2012_palivizumab.R (reciprocal-dilution values 0/10/20/40/>=80 with category-specific multiplicative effects per titer bin; 0 = ADA negative reference).
• Notes: The prior separate ADA_TITRE (British, 1 = negative) and ADA_TITER (American, 0 = negative) canonicals were merged on 2026-04-20 into a single general-scope ADA_TITER. The zero-encoding convention is the load-bearing semantic and must be documented per-model. Distinct from ADA_POS (binary presence/absence); when the paper reports both, the final model usually keeps only one. Distinct from [[CONC_ADA_NGML]], a calibrated ADA MASS concentration in ng/mL from a drug-tolerant assay: that column has no zero-encoding at all (ADA-negatives carry real positive concentrations) and is dimensioned, whereas ADA_TITER is a dimensionless dilution factor or arbitrary assay unit. Do not use ADA_TITER for a ng/mL quantity – the zero-encoding semantic above would be false and the unit would be misstated. Imputation rules (LOCF / NOCB / baseline-as-negative) should be documented per-model.

ABPEG_<isotype> – canonical family for PRE-EXISTING anti-polyethylene-glycol antibody levels. Antibodies directed at the polyethylene-glycol (PEG) moiety of a PEGylated drug, not at the drug’s protein or small-molecule payload. The defining property, and the reason this family is separate from the ADA_* and CONC_ADA_* canonicals, is that these antibodies are pre-existing: they arise from environmental PEG exposure (cosmetics, food, excipients) and are present in a large fraction of PEG-naive subjects, so they can and do act on the first dose of drug ever administered. Treatment-emergent anti-drug antibody columns cannot, by construction. Isotype is load-bearing and is carried in the name (ABPEG_IGM, ABPEG_IGG), because papers routinely screen both and retain only one. Values are on the originating assay’s own arbitrary scale (typically a plate-reader mean fluorescence intensity), are measured in every subject regardless of positivity, and are strictly positive with no special encoding for antibody-negative subjects; positivity, where a paper reports it, is a derived dichotomy from an assay cut point. Store the LINEAR level and apply any log transform inside model(), so a cut point printed on the log scale is recorded in the model file on the linear scale next to its log-scale twin. Distinct from ADA_TITER (a dimensionless dilution factor whose load-bearing semantic is a 0- or 1-for-negative encoding, which is actively false here), from CONC_ADA_NGML / CONC_ADA_AUML (calibrated anti-drug-antibody concentrations against the drug itself), from ADA_POS (binary treatment-emergent positivity), and from IGG / IGM (total serum immunoglobulin, not antigen-specific).

ABPEG_IGM (canonical for pre-existing anti-polyethylene-glycol IgM antibody level)

• Description: Anti-PEG IgM antibody level on the originating assay’s arbitrary scale, measured before the first dose of a PEGylated drug. Member of the ABPEG_<isotype> family described above. Baseline (one value per subject), not time-varying, unless a source explicitly re-measures it per exposure phase.
• Units: assay-specific arbitrary scale (mean fluorescence intensity for the founding assay); document per-model via covariateData[[ABPEG_IGM]]$units.
• Type: continuous
• Scope: general
• Reference category: n/a – enters as a hockey stick above an assay cut point. Cut points observed: 3.67 on the linear scale (1.3 on the natural-log scale), estimated in NONMEM by Siebel 2022 and held fixed by Wurthwein 2025; 2 on the linear scale, the assay cut point of Khalil 2022.
• Source aliases:
  • anti-PEG IgM prior PIAd12 – Wurthwein 2025 ESM Tables S5-S9, the level in the sample drawn within 3 days before the first induction dose.
  • IgMprior / IgMprior PIAd12 – the same quantity in the Wurthwein 2025 running text and figure captions.
  • IGMPLOG (natural log of IGMP) / anti-PEG IgMprior – Siebel 2022 ESM Section 3 control stream and Table 3; store the linear level and take the log in model().
• Example models: Wurthwein_2025_pegasparaginase_hrPostinduction.R, Wurthwein_2025_pegasparaginase_r2ea.R (founding examples; hockey stick 1 + theta * (log(ABPEG_IGM) - 1.3) on the initial clearance of PEGylated asparaginase, active only above the cut point and gated to the first induction dose, theta 0.396 and 0.449 respectively); Siebel_2022_pegasparaginase.R (the paper that estimated the cut point, carried as the estimated hinge labpeg_igm_hinge = 1.30, theta 0.414).
• Notes: The founding paper illustrates why the effect is usually gated to the first dose: anti-PEG levels fall sharply once dosing starts (Wurthwein 2025 ESM Section 2.2 and Fig. S2, ‘A marked decrease in antibody levels from prior to after administration is apparent following the first dose’), so a baseline level is informative about the first exposure and progressively less so afterwards. Clinical relevance is dose-dependent and worth recording per-model: Wurthwein 2025 Section 4.2 judges the effect ‘of rather limited relevance given the currently used high PEG-ASNase dosages but might become more relevant when down-titrating dosages’.

ABPEG_IGG (canonical for pre-existing anti-polyethylene-glycol IgG antibody level)

• Description: Anti-PEG IgG antibody level on the originating assay’s arbitrary scale, measured before the first dose of a PEGylated drug. Member of the ABPEG_<isotype> family described above. Baseline, not time-varying.
• Units: assay-specific arbitrary scale (mean fluorescence intensity for the founding assay); document per-model via covariateData[[ABPEG_IGG]]$units.
• Type: continuous
• Scope: general
• Reference category: n/a – enters as a hockey stick above an assay cut point. Cut point observed: 8 on the linear scale (2.079 on the natural-log scale), the assay cut point of Khalil 2022.
• Source aliases:
  • anti-PEG IgG prior / IgGprior – Wurthwein 2025 ESM Tables S5, S6.
• Example models: Wurthwein_2025_pegasparaginase_hrPostinduction.R.
• Notes: Registered as a distinct member rather than folded into ABPEG_IGM because the two isotypes are separately assayed, carry different cut points, and are routinely tested head-to-head: Wurthwein 2025 ESM Table S6 fits both and keeps IgM (BIC 73239 versus 73259), concluding ‘the influence of anti-PEG IgGprior was less pronounced’. A model that screens IgG without retaining it should record it in covariatesDataExcluded, not covariateData.

Disease / treatment history

TRTPH_<phase> – canonical family for protocol-defined treatment-phase / treatment-block indicators. Mutually exclusive binary indicators identifying which named phase (block, course, element) of a multi-phase treatment protocol a record falls in. Time-varying per record. Intended for regimens whose protocol is divided into named blocks that differ materially in concomitant therapy, intensity, or physiological state – most commonly pediatric and adult oncology protocols (ALL, AML, lymphoma), but the concept generalises to any multi-block regimen. Source papers normally report this as a single multi-level categorical covariate (“treatment phase”, “protocol element”, “treatment block”) with one row per level in the parameter table; decompose it into one TRTPH_<phase> binary per non-reference level, following the register’s decomposed-binary convention for multi-level categoricals (ICU_ADM_*, RACE_*, TUMTP_*). The protocol’s induction (or first) phase is the reference category and is encoded as all indicators 0, so no TRTPH_ column is registered for it.

Each model MUST document the protocol name and the phase-to-column mapping in covariateData[[TRTPH_<phase>]]$notes, because the same phase word (“maintenance”, “intensification”) means different things across protocols. A future paper using a different protocol registers its own members under this prefix; if two protocols use the same phase word with materially different therapy, qualify the member with the protocol (TRTPH_ALL11_MAINT) rather than overloading the bare name. Distinct from TRT_PHASE, which is a binary double-blind active-treatment-on / off gate for placebo-controlled PD models, NOT a multi-level protocol-block categorical. Distinct from CYCLE (an integer dose-number / cycle counter within a single regimen) and from OCC (a sampling-occasion index for inter-occasion variability). Distinct from the CONMED_<INN> family: a TRTPH_<phase> column deliberately stands in for the entire concomitant-chemotherapy block of that phase, which is the right encoding when the individual agents are co-administered and cannot be separated – Kloos 2021 tested doxorubicin and methotrexate as individual covariates on top of the treatment phase and rejected both in backward elimination.

TRTPH_1B (canonical for DCOG ALL-11 protocol 1B treatment-phase indicator)

• Description: 1 = the record falls within protocol 1B of the Dutch Childhood Oncology Group ALL-11 acute-lymphoblastic-leukemia protocol; 0 = any other phase. Protocol 1B follows induction protocol 1A and comprises cyclophosphamide, cytarabine and 6-mercaptopurine plus a 1,500 IU/m^2 PEGasparaginase dose at day 40. Member of the TRTPH_<phase> family described above.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0, with all other TRTPH_ indicators 0 = protocol 1A induction.
• Source aliases: 1B – Kloos 2021 Table 2 / Table 3 treatment-phase row label.
• Example models: Kloos_2021_pegasparaginase.R (clearance effect FIXED to 1, i.e. pooled with the 1A reference, because only two patients were treated as high risk and the effect was not estimable reliably; Kloos 2021 Table 2 prints “1 (fix)”).
• Notes: A fixed-to-reference member is still worth registering: dropping the column would silently erase the fact that the authors tested the phase and could not estimate it. See the family preamble above for the protocol-documentation requirement.

TRTPH_CONSOLIDATION (canonical for consolidation treatment-phase indicator)

• Description: 1 = the record falls within the consolidation phase of the treatment protocol; 0 = any other phase or population. For the founding model (QuANTUM-First, AC220-A-U302, newly diagnosed FLT3-ITD-positive AML) consolidation is standard high-dose cytarabine plus quizartinib or placebo, allogeneic hematopoietic cell transplantation, or both, in patients achieving remission. Member of the TRTPH_<phase> family described above.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0. Note that the all-TRTPH_-indicators-0 state is protocol-defined and is NOT necessarily the protocol’s own induction phase: for Vaddady 2024 the reference is relapsed/refractory AML patients on quizartinib monotherapy, “for whom no distinct treatment phases were reported”, so induction carries its own indicator rather than being folded into the reference. Any model using this column must state in covariateData notes what its all-zero state means.
• Source aliases: PHASE == 2 – Vaddady 2024 final NONMEM control stream (Supporting Information); Consolidation – Vaddady 2024 Table 3 treatment-phase row label.
• Example models: Vaddady_2024_quizartinib.R (fractional change -0.192 on quizartinib relative bioavailability, RSE 15.6 percent, Equation 9; and +0.272 on the parent-to-metabolite conversion fraction fMET, RSE 13.3 percent, Equation 16).
• Notes: Bare (unqualified) phase name, per the family preamble’s qualify-only-on-collision default; ratified by the maintainers. A later protocol whose consolidation phase differs materially in background therapy should qualify its own member (TRTPH_<PROTOCOL>_CONSOLIDATION) rather than overloading this entry.

TRTPH_CONTINUATION (canonical for continuation treatment-phase indicator)

• Description: 1 = the record falls within the continuation phase of the treatment protocol; 0 = any other phase or population. For the founding model (QuANTUM-First, AC220-A-U302) continuation is single-agent quizartinib or placebo for up to 3 years in patients with blood-count recovery, i.e. no background chemotherapy. Member of the TRTPH_<phase> family described above.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0; see the reference-category note on TRTPH_CONSOLIDATION – the all-zero state is protocol-defined and for Vaddady 2024 is relapsed/refractory AML on monotherapy, not induction.
• Source aliases: PHASE == 3 – Vaddady 2024 final NONMEM control stream (Supporting Information); Continuation – Vaddady 2024 Table 3 treatment-phase row label.
• Example models: Vaddady_2024_quizartinib.R (fractional change +0.418 on quizartinib relative bioavailability, RSE 12.5 percent, Equation 9; and -0.249 on fMET, RSE 8.47 percent, Equation 16).
• Notes: Bare (unqualified) phase name; ratified by the maintainers. This is the phase with the highest dose-normalised quizartinib exposure (about 1.4-fold the relapsed/refractory reference); Vaddady 2024 confirmed by a VPC restricted to patients who entered continuation that the effect is phase-related rather than driven by selection of patients who survived to that phase.

TRTPH_INDUCTION (canonical for induction treatment-phase indicator)

• Description: 1 = the record falls within the induction phase of the treatment protocol; 0 = any other phase or population. For the founding model (QuANTUM-First, AC220-A-U302) induction is quizartinib or placebo plus intravenous cytarabine and an anthracycline (daunorubicin or idarubicin). Member of the TRTPH_<phase> family described above.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0. This member exists precisely because the family preamble’s default – fold the protocol’s own induction phase into the all-zero reference – does NOT hold universally: Vaddady 2024 sets the reference to relapsed/refractory AML patients on quizartinib monotherapy, so induction is a non-reference level needing its own indicator. Register TRTPH_INDUCTION only when induction is genuinely non-reference in the source model; when induction IS the reference, encode it as all indicators 0 and register no column, per the preamble.
• Source aliases: PHASE == 1 – Vaddady 2024 final NONMEM control stream (Supporting Information); Induction – Vaddady 2024 Table 3 treatment-phase row label.
• Example models: Vaddady_2024_quizartinib.R (fractional change -0.419 on quizartinib relative bioavailability, RSE 4.92 percent, Equation 9; and +0.715 on fMET, RSE 6.38 percent, Equation 16).
• Notes: Bare (unqualified) phase name, per the family preamble’s qualify-only-on-collision default; ratified by the maintainers. In parent-plus-metabolite models the phase effects on parent bioavailability and on the conversion fraction act in opposite directions by construction: the parent bioavailability effect is carried over to the metabolite, and the conversion-fraction phase effect exists to counterbalance that spurious carry-over (Vaddady 2024 Discussion).

TRTPH_M (canonical for DCOG ALL-11 protocol M treatment-phase indicator)

• Description: 1 = the record falls within protocol M of the DCOG ALL-11 protocol (6-mercaptopurine plus high-dose methotrexate, 5,000 mg/m^2/dose at days 8, 22, 36 and 50); 0 = any other phase. Member of the TRTPH_<phase> family described above.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0, with all other TRTPH_ indicators 0 = protocol 1A induction.
• Source aliases: M – Kloos 2021 Table 2 / Table 3 treatment-phase row label.
• Example models: Kloos_2021_pegasparaginase.R (multiplicative on CL: 0.87, RSE 5.2 percent, bootstrap 95 percent CI 0.80-0.95).

TRTPH_MR_INTENS (canonical for DCOG ALL-11 medium-risk-group intensification treatment-phase indicator)

• Description: 1 = the record falls within the medium-risk-group intensification phase of the DCOG ALL-11 protocol (dexamethasone, vincristine, 6-mercaptopurine, plus doxorubicin in TEL/AML1-negative patients or methotrexate in TEL/AML1-positive patients); 0 = any other phase. Member of the TRTPH_<phase> family described above.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0, with all other TRTPH_ indicators 0 = protocol 1A induction.
• Source aliases: MRG intens. / MR intensification – Kloos 2021 Table 2 / Table 3 treatment-phase row labels.
• Example models: Kloos_2021_pegasparaginase.R (multiplicative on CL: 0.89, RSE 5.2 percent, bootstrap 95 percent CI 0.82-0.98).

TRTPH_MR_MAINT (canonical for DCOG ALL-11 medium-risk-group maintenance treatment-phase indicator)

• Description: 1 = the record falls within the medium-risk-group maintenance phase of the DCOG ALL-11 protocol (dexamethasone, vincristine, methotrexate and 6-mercaptopurine); 0 = any other phase. Member of the TRTPH_<phase> family described above.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0, with all other TRTPH_ indicators 0 = protocol 1A induction.
• Source aliases: MRG maint. / MR maintenance – Kloos 2021 Table 2 / Table 3 treatment-phase row labels.
• Example models: Kloos_2021_pegasparaginase.R (multiplicative on CL: 0.81, RSE 3.9 percent, bootstrap 95 percent CI 0.75-0.86).

TRTPH_SR_IV (canonical for DCOG ALL-11 standard-risk-group protocol IV treatment-phase indicator)

• Description: 1 = the record falls within protocol IV of the standard-risk arm of the DCOG ALL-11 protocol (dexamethasone and vincristine plus a single individualized PEGasparaginase dose at day 1); 0 = any other phase. Member of the TRTPH_<phase> family described above.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0, with all other TRTPH_ indicators 0 = protocol 1A induction.
• Source aliases: SRG protocol IV / SR protocol IV – Kloos 2021 Table 2 / Table 3 treatment-phase row labels.
• Example models: Kloos_2021_pegasparaginase.R (multiplicative on CL: 0.81, RSE 6.5 percent, bootstrap 95 percent CI 0.73-0.90).

HCT_COND_RIC (canonical for reduced-intensity conditioning regimen indicator)

• Description: 1 = subject received reduced-intensity conditioning (RIC) chemotherapy prior to allogeneic hematopoietic cell transplantation, 0 = subject received myeloablative conditioning (MAC). Conditioning intensity is fixed per subject for the analysis window (the conditioning regimen was completed before transplantation, before any of the post-transplant tacrolimus PK observations). RIC regimens use lower-dose chemotherapy / radiotherapy to preserve some host haematopoiesis and rely on graft-versus-tumour effect for cytoreduction; MAC regimens deliver high-dose chemotherapy and / or total-body irradiation that fully ablates host marrow.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (myeloablative conditioning, MAC).
• Source aliases:
  • RIC (Dunlap 2025 NM-TRAN convention; binary 0 / 1) – used directly in Dunlap_2025_tacrolimus.R.
• Example models: Dunlap_2025_tacrolimus.R (Dunlap 2025 Table 2 reduced-covariate-model column; exponential effect on apparent oral clearance: cl *= 0.63 ^ HCT_COND_RIC, so RIC recipients have ~37% lower apparent oral tacrolimus clearance than MAC recipients).
• Notes: Conditioning regimen intensity has been reported to associate with post-transplant tacrolimus apparent clearance, likely via gut / hepatic CYP3A activity, GVHD-related inflammatory response, and post-transplant haematopoietic state. The paper-specific definition of “RIC” follows the source publication’s own classification (e.g., Dunlap 2025 follows the institutional protocol at UNCMC, which pools non-myeloablative conditioning regimens into the RIC category when assigning the binary indicator); document the source paper’s RIC criteria in covariateData[[HCT_COND_RIC]]$notes. When a future paper distinguishes a third intensity tier (non-myeloablative, NMA) as a separate covariate level rather than pooling NMA into RIC, register a parallel canonical (e.g. HCT_COND_NMA) instead of overloading HCT_COND_RIC. Scope: specific because the column is meaningful only for allo-HCT recipients.

HCT_PRIOR (canonical for prior haematopoietic cell transplantation indicator)

• Description: 1 = the subject had at least one prior haematopoietic cell transplant (HCT) before the index event of the analysis (a subsequent HCT, or – as in the CAR-T founding example – before an unrelated index therapy such as CAR-T infusion), 0 = no prior HCT (transplant-naive prior to the index event). A history of prior transplantation marks a heavily pre-treated recipient with greater cumulative marrow and organ injury. The prior transplant may be autologous or allogeneic; document the type in covariateData[[HCT_PRIOR]]$notes when the source specifies it (Mueller-Schoell 2021 records a prior autologous stem cell transplant).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no prior HCT; “naive” in the source’s wording).
• Source aliases:
  • Prior HCT treatment / Prior HCT vs. naive – printed row labels in Sun 2025 Figure S3 panels c and f, and Table S4.
  • Number of prior HCT – Sun 2025 Table S2 reports this as a count (0: 201, 1: 31, 2: 6 of 238); the modelled covariate is the binary >= 1 collapse of it.
  • Previous ASCT [yes/no] – Mueller-Schoell 2021 Supplementary Table S1; a prior autologous stem cell transplant before CAR-T infusion (7 of 19 patients). Same construct (subject had a prior haematopoietic cell transplant), here in a CAR-T rather than a transplant cohort.
• Example models: Sun_2025_maribavir_vomiting.R (coefficient -2.05, i.e. prior-HCT recipients have LOWER treatment-emergent vomiting risk), Sun_2025_maribavir_immunosuppressant_increase.R (coefficient +1.89, i.e. prior-HCT recipients are far more likely to record a raised immunosuppressant concentration), MuellerSchoell_2021_axicabtageneCiloleucel_qsp.R (a previous ASCT raises the baseline CAR-T maximum expansion rate Vmax1,base via the shared fractional-change term 1 + 2.53 * HCT_PRIOR, applied to both mixture subpopulations; Equations 10-11 and Table 1).
• Notes: The underlying source column is a COUNT (0, 1 or 2 prior transplants) that the analysis collapsed to a binary. Registered as binary because that is the form fitted; a future paper that fits the count itself should register HCT_N_PRIOR rather than overloading this name. Distinct from TX_HCT, which marks whether the subject is an HCT recipient at all (versus a solid-organ-transplant recipient or a non-transplant subject); in an all-HCT cohort TX_HCT is uniformly 1 while HCT_PRIOR still discriminates. Promoted from Scope: specific to Scope: general on the strength of a second, independent source (Mueller-Schoell 2021) using the column with identical value semantics in a non-transplant (CAR-T) cohort, where a prior autologous stem cell transplant before the index therapy is the fitted contrast. The generalising step is that the index event need not itself be an HCT – the column marks a prior HCT before whatever the analysis’s index therapy is; the transplant type and index event are documented per model.

HCT_TCELL_INFUSION (canonical for post-transplant T-cell infusion indicator)

• Description: 1 = the subject received an adoptive T-cell infusion (e.g. donor lymphocyte infusion or a CMV-specific cytotoxic T-lymphocyte product) after the current haematopoietic cell transplant, 0 = no such infusion, or not reported. The infusion is an immunotherapeutic attempt to reconstitute antiviral or anti-tumour cellular immunity, so it flags a recipient whose immune reconstitution was judged inadequate.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no post-HCT T-cell infusion given, or not reported – Sun 2025 Figure S3 panel d labels the reference level explicitly as No/not reported, pooling the 4 unreported subjects with the 225 negatives).
• Source aliases:
  • T-cell infusion given post-HCT – printed row label in Sun 2025 Table S3, Figure S3 panel d and Table S4.
• Example models: Sun_2025_maribavir_cmv_clearance_wk8.R (coefficient -2.86 on the primary efficacy logit – recipients needing a T-cell infusion are much less likely to clear CMV), Sun_2025_maribavir_diarrhea.R (coefficient +1.74).
• Notes: Only 9 of 238 subjects were positive in the founding cohort, so the coefficients are estimated on very few events and the confidence intervals are correspondingly wide (e.g. OR 5.72, 95% CI 1.15-28.4 for diarrhea). The not reported subjects are pooled INTO the reference by the source rather than fitted as their own level – the opposite convention from CD8_PP65_NR in the same paper – so a downstream user must set this column to 0, not NA, for unreported subjects. Member of the HCT_* transplant-history family alongside HCT_COND_RIC, HCT_PRIOR, HCT_NUCCELL and the HCT_TCD_* set. Scope: specific.

HCT_TCD_ALEMTUZUMAB, HCT_TCD_ATG, HCT_TCD_EXVIVO (canonical for the graft T-cell-depletion modality set)

• Description: Mutually exclusive binary indicators for how, if at all, donor T cells were depleted for the current allogeneic haematopoietic cell transplant. HCT_TCD_ALEMTUZUMAB = 1 for in-vivo depletion with alemtuzumab (anti-CD52), HCT_TCD_ATG = 1 for in-vivo depletion with anti-thymocyte globulin, HCT_TCD_EXVIVO = 1 for ex-vivo graft manipulation (e.g. CD34 selection or ex-vivo T-cell removal). All three are 0 for an unmanipulated, non-T-cell-depleted graft. T-cell depletion reduces graft-versus-host disease risk at the cost of slower immune reconstitution and higher opportunistic-infection risk, so the modality is a determinant of post-transplant CMV control.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 for all three indicators = no T-cell depletion (None; 135 of 238 subjects, 56.7%, in the founding cohort).
• Source aliases:
  • T-Cell depletion modality – printed row label in Sun 2025 Table S2 and Figure S3 panel l, with levels None, Alemtuzumab administration (in vivo), Anti-thymocyte globulin administration (in vivo), Ex-vivo T-cell depletion, Not reported.
• Example models: Sun_2025_maribavir_headache.R (alemtuzumab +1.77, ATG -0.882, ex-vivo +0.863 on the headache logit; only the alemtuzumab contrast is significant, OR 5.84, 95% CI 1.24-27.5).
• Notes: The source’s Not reported level is deliberately NOT registered as a fourth indicator. It has exactly ONE subject in the founding cohort (1 of 238, 0.4%), and the fitted coefficient is -12.7 with a standard error of 1460 and p = 0.993 – the signature of complete separation, i.e. a non-estimate produced by a level with no informative data. Shipping it would put a spurious exp(-12.7) = 3e-6 odds ratio into a model file. A downstream user with an unreported subject should set all three indicators to 0, placing that subject in the None reference, and should understand this as the deliberate approximation it is. In-vivo antibody depletion (ALEMTUZUMAB, ATG) and EXVIVO graft manipulation are registered as separate indicators rather than pooled into one HCT_TCD_ANY because Sun 2025 fits them separately and their estimated effects differ in SIGN (alemtuzumab +1.77 versus ATG -0.882), so pooling would cancel a real contrast. Related to but distinct from CONMED_ANTILYMPHOCYTE, which records antilymphocyte drug exposure at any point rather than the transplant’s graft-manipulation protocol; the two are correlated but not identical, and Sun 2025 fits them in different endpoint models. Scope: specific.

HCT_NUCCELL (canonical for the transplanted nucleated cell number for the current HCT)

• Description: The number of nucleated cells contained in the graft infused for the current haematopoietic cell transplant, i.e. the cell dose. A larger cell dose generally predicts faster engraftment and haematopoietic recovery.
• Units: cells, on the source’s own reporting scale. Sun 2025 estimates the coefficient per increment of 1000 cells against a reference of 400 cells, so a model using this column forms (HCT_NUCCELL - 400) / 1000 at the call site. Record the per-model unit convention in covariateData[[HCT_NUCCELL]]$units – graft cell doses are variously reported as x10^6/kg, x10^8/kg or absolute counts, and Sun 2025 does not state which scale its column uses.
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as a centred linear regressor on the logit. Reference value observed: 400 (Sun 2025 Figure S3 panel m, row 400 cells Reference).
• Source aliases:
  • Nucleated cell number / Nucleated cell number for current HCT (per increment of 1000 cells) – printed row labels in Sun 2025 Figure S3 panel m and Table S4.
• Example models: Sun_2025_maribavir_thrombocytopenia.R (coefficient +0.00689 per 1000-cell increment above the 400-cell reference; OR 1.01 per increment, p = 0.022).
• Notes: The absolute scale of this column is UNSTATED in the source. Sun 2025 prints 400 cells as the reference and per increment of 1000 cells as the coefficient basis, with no multiplier and no per-kg normalisation, which is not a physically plausible graft cell dose read literally (a real allogeneic graft contains on the order of 10^8-10^10 nucleated cells). The register carries the source’s own numbers verbatim rather than guessing a multiplier, and the centring at 400 with a per-1000 increment makes the model file self-consistent and reproducible against the published odds ratio whatever the underlying unit turns out to be. A downstream user supplying a real cell dose must first rescale it onto the source’s scale; the founding model’s covariateData notes say so explicitly. Scope: specific, and it should stay specific until a paper reports this covariate with an unambiguous unit.

T_HCT (canonical for time from the current haematopoietic cell transplant to the start of the modelled treatment)

• Description: Time elapsed between the current haematopoietic cell transplant and the start of the modelled drug treatment. A per-subject scalar, not a time-varying clock: it locates the treatment course relative to the transplant, so it is a proxy for how far into post-transplant immune reconstitution the subject was when treatment began.
• Units: h
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as an uncentred linear regressor. Sun 2025 estimates the coefficient per day, so a model divides the canonical hours by 24 at the call site, exactly as T_FIRSTDOSE prescribes for its own per-day coefficients.
• Source aliases:
  • Days from onset of current HCT to treatment – printed row label in Sun 2025 Table S3 (key secondary endpoint column).
• Example models: Sun_2025_maribavir_cmv_clearance_wk16.R (coefficient +0.0489 per day on the key-secondary efficacy logit; OR 1.05 per day, 95% CI 1.01-1.09, p = 0.0140 – subjects treated later after transplant were MORE likely to achieve sustained CMV clearance, consistent with greater elapsed immune reconstitution).
• Notes: Carried in hours to match the units$time = "h" convention of the PK models this family sits beside, so a published per-day coefficient is recovered by dividing by 24 at the call site rather than by storing days in the column – the same convention T_FIRSTDOSE documents. Distinct from T_FIRSTDOSE (time since the first dose of the modelled drug, a within-course clock that runs forward during the analysis) and from T_ENTRY (a study-entry offset on a shared integration axis): T_HCT is a fixed pre-treatment interval measured backwards from treatment start to an event that happened before the analysis window opened. Member of the auto-approved T_<event> family. Scope: specific; meaningful only in transplant cohorts.

AGVHD_LIVER (canonical for acute graft-versus-host disease – liver involvement indicator)

• Description: 1 = subject has documented evidence of acute graft-versus-host disease (aGvHD) involving the liver (any grade, I-IV) at the current model time; 0 = no documented liver aGvHD. Time-varying per subject in allo-HSCT cohorts: 0 before the first documented liver-aGvHD diagnosis and 1 from that time onward for the modeled observation window. Interpolation convention (Waterhouse 2024 Methods): next observation carried backward (NOCB), i.e., a subject with a positive liver-aGvHD diagnosis at any time during the study has AGVHD_LIVER = 1 from study start through the diagnosis time; use last-observation-carried-forward (LOCF) instead if the source paper documents that convention.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no documented liver aGvHD; the pooled reference includes both subjects who never developed liver aGvHD and time periods before diagnosis in subjects who eventually did).
• Source aliases:
  • LIVGVHD / GVHD_LIVER – Waterhouse 2024 NM-TRAN column for the time-varying liver-aGvHD indicator. Same orientation, no value transformation.
• Example models: Waterhouse_2024_vedolizumab.R (multiplicative power-form effect on CL: cl *= 1.05^AGVHD_LIVER, i.e., liver-aGvHD subjects show ~5% higher CL than the reference at the diagnosis time and thereafter; the paper’s Table 2 reports the estimate as 1.05 (0.834, 1.26) with 10.3% RSE, and the Results paragraph 3 concludes the 90% CI crosses 1 so the effect is not clinically meaningful).
• Notes: Sibling to AGVHD_SKIN and AGVHD_INTESTINE – a patient can have more than one organ involvement simultaneously (Table 1 footnote in Waterhouse 2024: “a patient could exhibit one or more types of GvHD”), so the three indicators are orthogonal binary covariates rather than mutually-exclusive levels of a categorical. Time-varying with the diagnosis-time boundary; supply per subject at every observation timestamp in the event dataset. Specific scope because the column is meaningful only for allo-HSCT cohorts. Distinct from TX_LIVER (liver-transplant recipient indicator, a solid-organ-transplant-cohort covariate). Future extractions distinguishing chronic-vs-acute GvHD should register sibling canonicals (e.g., CGVHD_LIVER) rather than overloading this name.

AGVHD_SKIN (canonical for acute graft-versus-host disease – skin involvement indicator)

• Description: 1 = subject has documented evidence of acute graft-versus-host disease (aGvHD) involving the skin (any grade, I-IV) at the current model time; 0 = no documented skin aGvHD. Time-varying per subject in allo-HSCT cohorts using the same NOCB interpolation convention as AGVHD_LIVER.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no documented skin aGvHD).
• Source aliases:
  • SKGVHD / GVHD_SKIN – Waterhouse 2024 NM-TRAN column for the time-varying skin-aGvHD indicator.
• Example models: Waterhouse_2024_vedolizumab.R (multiplicative power-form effect on CL: cl *= 1.03^AGVHD_SKIN; Waterhouse 2024 Table 2 reports 1.03 (0.940, 1.12) with 4.56% RSE; the 90% CI crosses 1 so the effect is not clinically meaningful).
• Notes: Sibling to AGVHD_LIVER and AGVHD_INTESTINE; see the AGVHD_LIVER entry for the shared multi-organ / orthogonal-indicator rationale.

AGVHD_INTESTINE (canonical for acute graft-versus-host disease – intestinal involvement indicator)

• Description: 1 = subject has documented evidence of acute graft-versus-host disease (aGvHD) involving the intestine / gastrointestinal tract (any grade, I-IV) at the current model time; 0 = no documented intestinal aGvHD. Time-varying per subject in allo-HSCT cohorts using the same NOCB interpolation convention as AGVHD_LIVER.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no documented intestinal aGvHD).
• Source aliases:
  • INTGVHD / GVHD_INTESTINE / GVHD_GI – Waterhouse 2024 NM-TRAN column for the time-varying intestinal-aGvHD indicator.
• Example models: Waterhouse_2024_vedolizumab.R (multiplicative power-form effect on CL: cl *= 1.07^AGVHD_INTESTINE; Waterhouse 2024 Table 2 reports 1.07 (0.870, 1.27) with 9.54% RSE; the 90% CI crosses 1 so the effect is not clinically meaningful).
• Notes: Sibling to AGVHD_LIVER and AGVHD_SKIN; see the AGVHD_LIVER entry for the shared multi-organ / orthogonal-indicator rationale. Vedolizumab’s mechanism (integrin alpha-4 beta-7 blockade) specifically targets gut-homing leukocyte trafficking, so AGVHD_INTESTINE is the mechanistically-motivated on-target-organ indicator in vedolizumab-for-GvHD-prophylaxis popPK studies; the paper notes the observed effect (~7% increase in CL at intestinal-aGvHD onset) is consistent with target-mediated drug disposition at the on-target site but the estimate is not statistically or clinically significant in this dataset.

DISEXT_EP (canonical for extensive colitis / pancolitis indicator)

• Description: 1 = extensive colitis or pancolitis disease extension, 0 = otherwise (any non-extensive disease extension, e.g. left-sided colitis or proctitis).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0. In papers that decompose the disease-extension categorical into both DISEXT_EP and DISEXT_OTHER, DISEXT_EP = 0 AND DISEXT_OTHER = 0 corresponds to the left-sided-colitis reference group; in papers that use a single binary indicator for extensive colitis, the reference is pooled non-extensive (left-sided + any other extension).
• Source aliases:
  • EXTCOL – used in Faelens_2021_infliximab.R (binary 0/1 for extensive colitis at baseline; no separate “other” category).
  • Derived from a multi-level DISEXT column in the source (levels: left-sided colitis, extensive/pancolitis, other): DISEXT_EP = as.integer(DISEXT == "extensive/pancolitis").
• Example models: Moein_2022_etrolizumab.R (paired with DISEXT_OTHER; multiplicative effect on CL, +8.2% vs. left-sided colitis), Faelens_2021_infliximab.R (single-binary encoding; multiplicative fold-change on V of 1.25 when DISEXT_EP = 1).
• Notes: Optionally paired with DISEXT_OTHER when the source paper decomposes a three-level disease-extension categorical (left-sided / extensive-pancolitis / other) into two indicators; the pairing is paper-specific and not required. Promoted from scope: specific to scope: general on 2026-04-27 because the binary “extensive colitis vs not” semantics generalize across UC popPK papers regardless of whether the original dataset additionally distinguished an “other” disease-extension category.

DISEXT_OTHER (canonical for ‘other disease extension’ indicator)

• Description: 1 = disease extension other than left-sided colitis or extensive/pancolitis, 0 = not.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (left-sided colitis, when paired with DISEXT_EP = 0).
• Source aliases: Derived from a multi-level DISEXT column: DISEXT_OTHER = as.integer(DISEXT == "other").
• Example models: Moein_2022_etrolizumab.R (multiplicative effect on CL, +18% vs. left-sided colitis; large uncertainty due to 2% prevalence), Moein_2025_etrolizumab.R (the successor pooled UC+CD model, which instead COLLAPSES this level with DISEXT_EP into a single “UC not left-sided colitis” category carrying one shared coefficient, and moves the effect from CL to an additive shift on the LOGIT of bioavailability: -0.244, giving F = 0.693 versus the 0.743 reference).
• Notes: Paired with DISEXT_EP; together they encode the three-level disease-extension categorical. The two levels are not always given separate coefficients: where the “other” group is very small (20 of 2312 subjects in Moein 2025), a paper may pool it with extensive/pancolitis and estimate one non-left-sided coefficient. Both columns are still required in that case, because the model forms the pooled indicator as DISEXT_EP + DISEXT_OTHER; record the pooling in the per-model covariateData notes so the shared estimate is not mistaken for two independent ones.

DISLOC_ILEUM (canonical for Crohn’s disease ileum-only location indicator)

• Description: 1 = Montreal disease location L1, terminal-ileal disease only; 0 = otherwise. Anatomical LOCATION of Crohn’s disease, recorded at diagnosis or baseline. First member of the DISLOC_ILEUM / DISLOC_COLON pair that decomposes the three-level Montreal location factor (L1 ileal / L2 colonic / L3 ileocolonic) into two binary indicators.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0. When paired, DISLOC_ILEUM = 0 AND DISLOC_COLON = 0 corresponds to the ileocolonic (L3) reference group, which is the modal category in most CD trial populations. A model that references a different level must say so in covariateData[[DISLOC_ILEUM]]$reference_category – Moein 2025 does exactly this for one of its six endpoints, referencing ileum-only rather than ileocolonic.
• Source aliases:
  • L1, Montreal L1, ileum only, ileal – standard spellings of the Montreal location category.
  • Derived from a multi-level disease-location column: DISLOC_ILEUM = as.integer(DISLOC == "ileum only").
• Example models: Moein_2025_etrolizumab_induction_endoimp.R, Moein_2025_etrolizumab_induction_endorem.R, Moein_2025_etrolizumab_maintenance_endoimp.R, Moein_2025_etrolizumab_maintenance_endorem.R (log-odds shift on the endoscopic-endpoint logit; which of the two indicators the paper retains flips by endpoint, so both columns are needed across the family).
• Notes: Paired with DISLOC_COLON; together they encode the three-level Montreal location categorical. Deliberately kept under a new DISLOC_ (disease LOCATION) prefix rather than folded into the existing DISEXT_ (disease EXTENT) prefix: DISEXT_EP / DISEXT_OTHER describe how far ulcerative colitis extends along the colon, whereas the Montreal L1/L2/L3 classification describes which region of the gut Crohn’s disease occupies. The two are different clinical concepts on different indications and a pooled UC+CD dataset can carry both, so conflating the prefixes would make the columns ambiguous. Montreal location is conventionally treated as time-fixed. The behaviour (B1/B2/B3) and upper-GI (L4) modifiers of the Montreal classification are orthogonal to location and would need their own canonicals if a model retained them. Ratified canonically by the maintainers on 2026-09-11 alongside the Moein 2025 etrolizumab extraction.

DISLOC_COLON (canonical for Crohn’s disease colon-only location indicator)

• Description: 1 = Montreal disease location L2, colonic disease only; 0 = otherwise. The colonic sibling of DISLOC_ILEUM, completing the two-indicator decomposition of the three-level Montreal location factor.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (ileocolonic L3, when paired with DISLOC_ILEUM = 0). See the DISLOC_ILEUM entry for the per-model exception.
• Source aliases:
  • L2, Montreal L2, colon only, colonic – standard spellings of the Montreal location category.
  • Derived from a multi-level disease-location column: DISLOC_COLON = as.integer(DISLOC == "colon only").
• Example models: Moein_2025_etrolizumab_induction_endoimp.R, Moein_2025_etrolizumab_induction_endorem.R, Moein_2025_etrolizumab_maintenance_endoimp.R, Moein_2025_etrolizumab_maintenance_endorem.R (log-odds shift on the endoscopic-endpoint logit, paired with DISLOC_ILEUM).
• Notes: See the DISLOC_ILEUM entry for the shared pairing, reference-category and prefix rationale. Colon-only Crohn’s disease is the category most often confused with ulcerative colitis on endoscopy, which is why the location covariate carries signal for endoscopic endpoints specifically: in Moein 2025 the two endoscopic endpoints retain a location effect while the symptom-based clinical-remission endpoints do not.

PREVENT_PRIMARY (canonical for primary-prevention cohort indicator)

• Description: 1 = the cohort (or, in a meta-analysis, the trial) enrolled a primary-prevention population, i.e. participants with no previous history of the clinical event the therapy is meant to prevent; 0 = it did not. First member of the mutually exclusive PREVENT_PRIMARY / PREVENT_SECONDARY / PREVENT_MIXED trio that decomposes a three-level prevention-category factor into binary indicators. Exactly one of the three is 1.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (not a primary-prevention cohort). Because the trio is exhaustive, a model must either omit one level as the implicit reference or, as in the founding example, carry all three and give each its own coefficient with no separate intercept term.
• Source aliases:
  • Prevention = Primary – Volkova 2023 Table 1 beta_2 sub-row label.
• Example models: Volkova_2023_lipidLowering_mace_mbma.R (selects the per-year age slope on the log risk ratio of unspecified stroke: +0.007/year in primary-prevention trials).
• Notes: Cardiovascular-prevention framing, but the primary/secondary distinction generalises to any preventive indication (stroke, fracture, exacerbation), so this is registered as general scope. Sibling canonicals PREVENT_SECONDARY and PREVENT_MIXED. The event whose history defines the category is indication-specific and must be documented per model in covariateData[[PREVENT_PRIMARY]]$notes – in Volkova 2023 it is a previous major adverse cardiac event.

PREVENT_SECONDARY (canonical for secondary-prevention cohort indicator)

• Description: 1 = the cohort enrolled a secondary-prevention population, i.e. participants with a previous history of the clinical event the therapy is meant to prevent; 0 = it did not. Second member of the PREVENT_* indicator trio; see PREVENT_PRIMARY.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (not a secondary-prevention cohort).
• Source aliases:
  • Prevention = Secondary – Volkova 2023 Table 1 beta_2 sub-row label.
• Example models: Volkova_2023_lipidLowering_mace_mbma.R (selects the per-year age slope on the log risk ratio of unspecified stroke: +0.010/year in secondary-prevention trials, a steeper loss of benefit with age than the primary-prevention slope).
• Notes: Sibling of PREVENT_PRIMARY and PREVENT_MIXED; mutually exclusive with both. Secondary-prevention cohorts carry a higher baseline event rate, so a model that uses this indicator for an effect-size contrast should not also be relied on for the control-arm event rate unless the source models that separately.

PREVENT_MIXED (canonical for mixed-or-unstated-prevention cohort indicator)

• Description: 1 = the cohort enrolled both primary- and secondary-prevention participants, or the source did not state the prevention category; 0 = otherwise. Third member of the PREVENT_* indicator trio; see PREVENT_PRIMARY.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (not a mixed / unstated cohort).
• Source aliases:
  • Prevention = Both – Volkova 2023 Table 1 beta_2 sub-row label.
• Example models: Volkova_2023_lipidLowering_mace_mbma.R (selects the per-year age slope on the log risk ratio of unspecified stroke: +0.011/year).
• Notes: Sibling of PREVENT_PRIMARY and PREVENT_SECONDARY. This level routinely pools two different things – genuinely mixed cohorts and cohorts whose prevention category was simply never reported. Volkova 2023 Section 2.2 states explicitly that “when prevention type was not explicitly indicated in the published sources, a mixed category was assigned to a study”, so 18 of its 54 trials sit in this level. Record per model whether the source distinguished the two, because a coefficient on this level is partly a missing-data indicator and not purely a population descriptor.

PRIOR_TAXANE (canonical for binary prior-taxane chemotherapy indicator)

• Description: 1 = subject received any prior taxane regimen (docetaxel, paclitaxel, cabazitaxel, etc.) before study entry, 0 = taxane-naive. Time-invariant within a subject (records treatment history at baseline). Oncology-pretreatment indicator: relevant for cohorts in which prior taxane exposure plausibly alters disease biology (e.g., advanced / castration-resistant prostate cancer, where taxane pretreatment is associated with more advanced disease and selects for taxane-resistant clones).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (taxane-naive).
• Source aliases:
  • PTAX – used in vanHasselt_2015_eribulin.R (van Hasselt 2015 paper notation; binary 0 / 1 for prior docetaxel pretreatment).
• Example models: vanHasselt_2015_eribulin.R (multiplicative effect on baseline serum PSA: psa0 = exp(lpsa0 + etalpsa0) * e_prior_taxane_psa0^PRIOR_TAXANE with e_prior_taxane_psa0 = 3.23 – prior-taxane patients have ~3.2x higher baseline PSA than taxane-naive patients, consistent with more advanced disease at study entry).
• Notes: Pairs with PRIOR_TAXANE_DAYS when a continuous duration-of-pretreatment is also relevant (e.g., van Hasselt 2015 uses both: PTAX on PSA0 and NTRT on KD). Distinct from CONMED_* (which is concomitant medication during the study, not pretreatment history) and from generic PRIOR_* chemotherapy indicators (which would warrant a separate canonical when a paper differentiates by drug class rather than collapsing to taxanes). Scope: specific because the population semantics (CRPC) and the “any prior taxane” pooling are tied to van Hasselt 2015; future papers that distinguish per-drug pretreatment (docetaxel vs paclitaxel vs cabazitaxel separately) should register parallel canonicals rather than overloading PRIOR_TAXANE.

PRIOR_TAXANE_DAYS (canonical for cumulative days of prior taxane treatment)

• Description: Cumulative number of days of prior taxane chemotherapy at study entry. 0 for taxane-naive patients (i.e., for any subject with PRIOR_TAXANE = 0). Time-invariant within a subject (records the pretreatment history at baseline).
• Units: days
• Type: continuous
• Scope: specific
• Reference category: n/a – normalised by a population reference value (van Hasselt 2015 uses the population median of 720 days among prior-taxane-pretreated patients) and entered as a power covariate (1 + PRIOR_TAXANE_DAYS / 720) ^ e_prior_taxane_days_<param>. The +1 inside the bracket makes the covariate effect collapse to a multiplier of 1 for taxane-naive patients (PRIOR_TAXANE_DAYS = 0) regardless of the estimated exponent.
• Source aliases:
  • NTRT – used in vanHasselt_2015_eribulin.R (van Hasselt 2015 paper notation; cumulative number of days of prior taxane treatment).
• Example models: vanHasselt_2015_eribulin.R (van Hasselt 2015 Eq. 4 power covariate on drug PSA inhibition rate KD0: kd0 = exp(lkd0 + etalkd0) * (1 + PRIOR_TAXANE_DAYS / 720)^e_prior_taxane_days_kd0 with e_prior_taxane_days_kd0 = -4.00 – KD0 decreases with longer prior-taxane exposure, encoding cross-resistance between docetaxel and eribulin via the shared microtubule-inhibition mechanism).
• Notes: Pairs with PRIOR_TAXANE (binary). The paper also considered an alternative continuous parameterisation in cycles of prior taxane (NCYCL, median 30 cycles) which was deemed slightly less informative (dOFV = -8 for NCYCL vs -10 for NTRT) and was not retained in the final model. If a future model needs the cycle-count form, register a parallel canonical (e.g. PRIOR_TAXANE_CYCLES) rather than overloading this one. Scope: specific because the 720-day normalisation reference is tied to the van Hasselt 2015 study population (post-docetaxel mCRPC patients).

T_DIAG_DIAB (canonical for time since type 2 diabetes diagnosis)

• Description: Time elapsed since clinical diagnosis of type 2 diabetes mellitus (T2DM) at study entry, in years. Continuous time-fixed covariate; supply the per-subject duration in years. T2DM-specific (Type 1 cohorts use a separate T_DIAG_T1D canonical if registered in the future).
• Units: years
• Type: continuous
• Scope: general
• Reference category: n/a – enters as a power covariate (T_DIAG_DIAB / 2)^theta on the model-predicted baseline FPG, on Gmax (the maximal SGLT2-inhibition-driven FPG-reduction), and on kHbA1cout. Reference value 2 years per Baron 2016 Tables S3 and S4.
• Source aliases:
  • DUR – used in Baron_2016_empagliflozin.R (Baron 2016 Methods column “duration of diabetes”; PK/PD dataset categorisation reported as <1 year / 1-5 years / >5 years in Table S2 with 58.5% of patients in the >5 years group).
• Example models: Baron_2016_empagliflozin.R (power-form effects centred at 2 years on BFPG: (T_DIAG_DIAB / 2)^0.0512, on Gmax: (T_DIAG_DIAB / 2)^0.0117, on kHbA1cout: (T_DIAG_DIAB / 2)^-0.577).
• Notes: Disease-progression marker complementary to the binary DIS_DIAB indicator – DIS_DIAB records the comorbidity at baseline (1/0) while T_DIAG_DIAB quantifies how long that diabetes has been present. The covariate is a time-since-event under the canonical T_<event> family. Distinct from T_NUT_SUPP (time on nutritional supplementation) and T_POST_ECMO (time after ECMO decannulation) which are different T_<event> siblings. For a baseline-only continuous power-form effect, a 0-year subject (newly diagnosed at study entry) yields (0/2)^theta = 0 – supply at least a small floor value (e.g. 0.1 year) for newly-diagnosed subjects in simulation; the floor convention should be documented in the per-model covariateData[[T_DIAG_DIAB]]$notes.

T_DIAG_CANCER (canonical for time since primary cancer diagnosis)

• Description: Time elapsed between the patient’s primary cancer diagnosis (histopathology / imaging confirmation date recorded in the clinical database) and study entry / re-baseline, in days. Continuous time-fixed covariate; supply the per-subject duration in days. Cancer-nonspecific (gastric, GEJC, NSCLC, breast, etc. all use this same canonical). Distinct from T_DIAG_DIAB (T2DM-specific; different disease and different reference form).
• Units: days
• Type: continuous
• Scope: general
• Reference category: n/a – entered as a power covariate (T_DIAG_CANCER / ref)^exponent on structural parameters. Reference values observed: 53 days (Terranova 2022 avelumab arm median at randomization for the JAVELIN Gastric 100 gastric-cancer cohort, per Supplementary Methods).
• Source aliases:
  • Tdiag – used in Terranova_2022_TGD_OS_gastric.R (Terranova 2022 Supplementary Methods equation form (Tdiag/53)^theta5 on the Gompertz tumor-growth rate constant Kg).
  • Log time since diagnosis – Terranova 2022 Table S2 OS TTE coefficient label; enters on log(median OS) via log(T_DIAG_CANCER) * beta.
• Example models: Terranova_2022_TGD_OS_gastric.R (power effect on Gompertzian Kg: Kg = tvKg * (T_DIAG_CANCER / 53)^-0.00291, reference 53 days; and additive linear-log effect on log-median OS: log_median_OS += 0.0436 * log(T_DIAG_CANCER)).
• Notes: Sibling of T_DIAG_DIAB under the canonical T_<event> family. Auto-approved without a separate naming ruling per the T_ policy on 2026-07-24. For a 0-day subject (diagnosed at study entry) the power form (0/53)^theta evaluates to 0 when the exponent is positive, which is a boundary case; supply at least a small floor value (e.g. 1 day) for such subjects in simulation. Distinct from post-treatment intervals (T_POST_ECMO, etc.) which are time-since-intervention rather than time-since-diagnosis.

T_DIAG_CD (canonical for time since Crohn’s disease diagnosis)

• Description: Duration of Crohn’s disease at study entry – time elapsed since the clinical diagnosis of Crohn’s disease – in years. Continuous and time-fixed. In a model-based meta-analysis it is the study-arm mean (or median) disease duration reported in the trial’s baseline table, not a patient value.
• Units: years
• Type: continuous
• Scope: general
• Reference category: n/a – enters as a power covariate (T_DIAG_CD / ref)^theta. Reference values observed: 9.317 years (Yu 2022 Delta CRP model) and 9.679 years (Yu 2022 Delta IBDQ model), each the arm mean of the model’s analysis-dataset sheet.
• Source aliases:
  • dur – Yu 2022 deposited dataset (Supplementary DataSheet 1) and model code.
  • disease duration, duration of Crohn's disease – Yu 2022 Table 1, Table 2 and Methods ‘Covariate’.
• Example models: Yu_2022_crohns_dcrp_mbma.R (exponent 4.95 on every drug effect: a longer disease duration gives a larger CRP decrease), Yu_2022_crohns_dibdq_mbma.R (exponent -8.98 on every drug effect: a longer disease duration gives a smaller IBDQ improvement).
• Notes: Sibling of T_DIAG_DIAB and T_DIAG_CANCER under the auto-approved T_<event> family. Crohn’s-disease-specific in meaning, but general in scope because disease duration carries the same definition in every Crohn’s disease model. Distinct from the Crohn’s disease indicator IBD_CD, which records the diagnosis itself; a pooled ulcerative-colitis cohort would need a separate T_DIAG_UC. Both founding exponents are steep, so the reference value is load-bearing: in Yu_2022_crohns_dibdq_mbma.R a 1% change in the reference moves the drug effect by about 9%, which is why each model states the dataset mean it normalises by. ### T_SYMPT_BPH (canonical for duration of benign prostatic hyperplasia symptoms)
• Description: Time elapsed between the onset of the patient’s BPH / lower urinary tract symptoms (as reported at the medical-history interview) and study entry. Continuous time-fixed covariate. Distinct from the time since the formal BPH DIAGNOSIS, which is typically shorter (D’Agate 2020 Table 1: median symptom duration 4 y versus median time since diagnosis 2.3 y) and should be a separate T_DIAG_BPH canonical if a model needs it.
• Units: years
• Type: continuous
• Scope: general
• Reference category: n/a – enters as a centred linear term 1 + e_t_sympt_bph_<param> * (T_SYMPT_BPH - <ref>). Reference value observed: 4 years (D’Agate 2020 pooled-population median).
• Source aliases:
  • BPHDUR – the $INPUT column in the D’Agate 2020 Supporting Information control stream and the Table 2 row label ‘BPHDUR on magnitude of placebo effect (duration of symptoms)’.
• Example models: DAgate_2020_bph_ipss_mbma.R (founding example; linear effect on the magnitude of the placebo effect on IPSS, DELTA_placebo * (1 - 0.025 * (T_SYMPT_BPH - 4))).
• Notes: Member of the auto-approved T_<event> family (sibling of T_DIAG_DIAB, T_DIAG_CANCER, T_MOTORCOMPL). The D’Agate 2020 Supporting Information labels the column ‘duration of BPH symptoms (months)’, but its Table 1 reports the same covariate in years (median 4, range 0-54.8) and the control stream centres it at 4.00, so years is the unit of the fitted coefficient.

T_PEGIFN (canonical for planned pegylated-interferon-alpha treatment-course duration)

• Description: Planned (protocol-specified) duration of the pegylated-interferon-alpha (Peg-IFNalpha) treatment course, in weeks. A design characteristic of the treatment arm rather than a patient attribute: it is the length of the course the arm was assigned, which in chronic hepatitis B ranges from the standard 48 weeks up to extended 96-week regimens and down to shortened 24-week ones. Time-fixed.
• Units: week
• Type: continuous
• Scope: general
• Reference category: n/a – continuous, entered as a centred deviation (T_PEGIFN - ref). Reference values observed: 48 weeks (Hanan 2026, the standard Peg-IFNalpha regimen and the reference stated in Table 2). Record the centring value in the model’s covariateData[["T_PEGIFN"]]$notes rather than assuming a library-wide default.
• Source aliases:
  • Peg-IFNalpha treatment duration (weeks) – used in Hanan_2026_peginterferon_alfa_eot_mbma.R and Hanan_2026_peginterferon_alfa_24wk_mbma.R (Hanan 2026 Table 2 covariate row “Peg-IFNalpha treatment duration (per week, ref.=48 weeks)”; same quantity in the same units, no value transformation).
• Example models: Hanan_2026_peginterferon_alfa_eot_mbma.R (linear centred effect on the logit-scale probability of HBsAg loss at end of treatment: logit_p += 0.0133 * (T_PEGIFN - 48), OR 1.013 per week), Hanan_2026_peginterferon_alfa_24wk_mbma.R (same form on the 24-week post-treatment endpoint: logit_p += 0.0303 * (T_PEGIFN - 48), OR 1.031 per week – more than twice the per-week effect seen at end of treatment).
• Notes: Auto-approved member of the T_<event> canonical family, alongside T_CPB (cardiopulmonary-bypass run duration) and T_PUMP (breast-milk expression session duration): the T_ prefix denotes a procedure- or treatment-related time covariate and, as in those siblings, the “event” is the Peg-IFNalpha course itself and the column carries its duration – not a time since or until it. It is not elapsed time on the model integration axis and it is not a time-varying covariate; both founding models are cross-sectional and have no time axis at all. In particular, in the founding paper’s 24-week post-treatment model, T_PEGIFN is the treatment-course length and is emphatically not the fixed 24-week post-cessation follow-up interval, which is part of the endpoint definition rather than a covariate. Units are weeks, matching how interferon regimens are universally prescribed and reported; convert to weeks rather than registering a day-scaled sibling. Drug-specific by design (Peg-IFNalpha), following the same convention as the DOSE_<drug>_<units> family – a different agent’s course duration should be registered as its own T_<drug> canonical rather than overloading this one.

T_PATCH_WEAR (canonical for transdermal patch application (wear) duration)

• Description: Time from application to removal of a transdermal patch, in hours, supplied on the patch dose records. A design characteristic of the application, typically 16 or 24 h for daily nicotine patches.
• Units: h
• Type: continuous
• Scope: general
• Reference category: n/a – continuous; used structurally rather than as a centred covariate.
• Source aliases:
  • DUR (the dataset column DUR2 renamed DUR in $INPUT) – Olsson Gisleskog 2021 ESM transdermal control stream (D5 = DUR - ALAG5).
• Example models: OlssonGisleskog_2021_nicotine_transdermal.R (sets the zero-order release duration to T_PATCH_WEAR minus the 4.06-h zero-order lag, so that release ends at patch removal).
• Notes: Auto-approved member of the T_<event> family; like T_CPB and T_PEGIFN the column carries the duration of the event (the application) rather than a time since it. It is not elapsed time on the integration axis.

BL_PN_GR1 (canonical for active baseline grade 1 peripheral neuropathy indicator)

• Description: 1 = subject had active grade 1 peripheral neuropathy (CTCAE 4.0 grade 1: asymptomatic; clinical or diagnostic observations only; intervention not indicated) at study entry; 0 = no active PN at baseline. Time-fixed at study entry. The covariate captures pre-existing low-grade PN typically arising from prior antimicrotubule or platinum chemotherapy in heavily pretreated oncology cohorts and is used to assess whether baseline PN sensitizes patients to subsequent antimicrotubule-induced PN.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no active PN at baseline).
• Source aliases:
  • “active grade 1 PN at study entry (yes/no)” (Lu 2017 narrative; coded as a binary 0/1 in the source NONMEM dataset) – used in Lu_2017_polatuzumab_neuropathy.R.
• Example models: Lu_2017_polatuzumab_neuropathy.R (multiplicative effect on the grade >= 2 peripheral neuropathy hazard via exp(theta_baselinePN * BL_PN_GR1) with theta_baselinePN = -0.222, SE 0.324 – not detectable given the large SE; the paper reports a sensitivity analysis in which this indicator was replaced by a broader “history of prior PN” indicator with similarly inconclusive results).
• Notes: Specific to oncology cohorts where grade 1 PN is permitted at study entry (per the source paper’s eligibility criteria). Distinct from PREV_AE_SCORE (the time-varying ordinal Markov-state AE-score covariate from Girard 2012 pimasertib, which conditions a current-observation outcome on the previous observation’s grade); BL_PN_GR1 is a time-fixed baseline indicator that does not update during the analysis window. Distinct from DIS_DPN (painful diabetic peripheral polyneuropathy disease-state indicator from chronic-pain cohorts); DIS_DPN flags a diabetes-related primary diagnosis used as a stratification covariate, while BL_PN_GR1 flags a pre-existing low-grade adverse-event status from any prior cause in an oncology trial. When a future paper uses a different baseline-PN grade threshold (e.g., grade <= 2 permitted) or a different AE category, register a parallel canonical (e.g., BL_PN_GR2, BL_FATIGUE_GR1) rather than overloading this one.

T_MOTORCOMPL (canonical for time since onset of levodopa-related motor response complications)

• Description: Elapsed time since the onset of levodopa-related motor response complications (wearing-off, ON-OFF fluctuations, dyskinesia) in a patient with Parkinson disease. A joint measure of disease duration and disease severity, distinct from time since diagnosis and from time since starting levodopa because complications emerge only after a period of successful levodopa treatment.
• Units: year
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as the normalised power term (T_MOTORCOMPL / 2.8)^exponent. The normalising constant 2.8 is the rounded cohort median (Knebel 2012 Table I: mean 3.74, median 2.76, range 0.04-29.9 years).
• Source aliases:
  • TOMC – printed name in Knebel 2012 (individual-parameter equations and Table II row labels, e.g. *(TOMC/2.8)^theta7).
• Example models: Knebel_2012_istradefylline_offtime.R (power effect on the baseline percentage OFF time E0, on the maximum disease-progression/placebo-response effect EmaxP, and on the maximum istradefylline effect EmaxI; Knebel 2012 Table II thetas 7, 14 and 21).
• Notes: Specific scope. Member of the T_<EVENT> elapsed-time family (T_DIAG_CANCER, T_DIAG_DIAB, T_ENTRY, T_FIRSTDOSE, T_CPB, T_ECMO, T_NUT_SUPP, T_POST_ECMO, T_POST_HEMODIAL, T_PUMP); T_DIAG_DIAB is the closest analogue, being likewise a disease-duration covariate in years. Units are years here rather than the days used by T_DIAG_CANCER, because the founding model’s normalising constant (2.8) and its published cohort statistics are all in years and rewriting them in days would misstate the source; record the unit per model via covariateData[[T_MOTORCOMPL]]$units. Knebel 2012 retained this covariate in preference to two collinear alternatives – time since diagnosis of Parkinson disease (TPD) and time since start of levodopa therapy (LYRS), which had pairwise correlations of at least 0.64 with it and gave similar estimates but a higher objective function. Those two are documented in the founding model’s covariatesDataExcluded rather than registered here, because Knebel 2012 publishes no effect estimate for either; a future Parkinson model that does estimate them should register T_DIAG_PD and T_LEVODOPA as siblings at that point.

T_DIAG_RA (canonical for time since rheumatoid arthritis diagnosis)

• Description: Time elapsed since the clinical diagnosis of rheumatoid arthritis at study entry (RA disease duration), in years. Continuous time-fixed covariate; per-subject, or the arm mean in a model-based meta-analysis.
• Units: year
• Type: continuous
• Scope: general
• Reference category: n/a – continuous. Leil 2021 centres it at 8.2 years (log-transformed, additive on the placebo Emax).
• Source aliases:
  • disease duration (years) – Leil 2021 Table 2 column label (arm mean).
• Example models: Leil_2021_rheumatoidArthritis_das28_mbma.R (additive -0.133 * log(T_DIAG_RA / 8.2) on the placebo Emax of the DAS28 change from baseline; longer-standing disease gives a smaller background-therapy response). Also documented (screened, not retained) in Chen_2025_methotrexate_das28_mbma.R, Chen_2025_methotrexate_acr20_mbma.R and Chen_2025_methotrexate_acr50_mbma.R.
• Notes: Sibling of T_DIAG_DIAB and T_DIAG_CANCER under the canonical T_<event> family. Must be strictly positive where it enters through log(); supply a small floor for arms of newly diagnosed patients.

Hypercholesterolemia biomarkers

PCSK9 (canonical for baseline unbound serum PCSK9 concentration)

• Description: Baseline unbound serum proprotein convertase subtilisin/kexin type 9 (PCSK9) concentration.
• Units: ng/mL (document per-model in covariateData[[PCSK9]]$units if a different unit – typically nM – is used in a given model; conversion uses a PCSK9 molecular weight of ~72 kDa, so 1 nM ~= 72 ng/mL).
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (PCSK9 / ref)^exponent. Reference values observed: 425 ng/mL (= 5.9 nM) in Kuchimanchi_2018_evolocumab.R (population median).
• Source aliases: none known.
• Example models: Kuchimanchi_2018_evolocumab.R (power exponent 0.194 on Vmax: Vmax * (PCSK9/425)^0.194), Nicolas_2019_alirocumab.R (source column FBSPCSK9, the pre-first-dose mean including screening draws; two ADDITIVE linear-deviation effects centred on the 265 ng/mL pooled median, +0.00156 per ng/mL on the LDL-C Emax and +0.00340 per ng/mL on the Hill coefficient – a linear-deviation rather than a power form, so the reference enters as a subtraction and not as a divisor).
• Notes: PCSK9 is the pharmacological target of anti-PCSK9 monoclonal antibodies (evolocumab, alirocumab, etc.); baseline PCSK9 drives the magnitude of target-mediated elimination and is a recurring covariate in anti-PCSK9 popPK models. Baseline (time-fixed) covariate; patients with missing baseline PCSK9 are typically excluded from analyses that include PCSK9 as a covariate.

Pharmacogenomic SNPs

Canonical pattern: SNP_<GENE>_<RSID>. Use one binary indicator per SNP genotype that the source paper tests as a model covariate. The SNP_ prefix makes the category unambiguous; the gene symbol disambiguates rsIDs grouped by gene; the rsID provides a globally unique identifier. Encoding follows the most common pharmacogenomic convention (also used by Papachristos 2020): 1 = mutant allele present (heterozygous or homozygous mutant); 0 = homozygous wild-type. When a paper uses a different encoding (e.g., per-allele dosage 0/1/2, dominant model with mutant homozygotes only, or recessive model), document the encoding explicitly in covariateData[[<COL>]]$notes and consider registering a separate canonical name. SNP indicators default to scope: specific because the parameter on which they act and the encoded reference category are tied to the source paper’s analysis plan; promote to general when a second paper ratifies identical semantics.

SNP_ABCG2_RS4148157 (canonical for ABCG2 rs4148157 variant indicator)

• Description: Binary genotype indicator for the ABCG2 (BCRP) rs4148157 single-nucleotide polymorphism (G > A; intron 11, population MAF ~0.10). 1 = at least one variant (A) allele present (heterozygous AG or homozygous AA carrier); 0 = homozygous wild-type (GG).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (homozygous wild-type GG).
• Source aliases:
  • GENECAT – Roberts 2016 (paper Results ‘Population Pharmacokinetic Analysis’ section: the individual-Ka equation Ka_i = Ka_pop * exp(theta_1 * GENECAT), with GENECAT = 1 for AG / AA and 0 for GG).
• Example models: Roberts_2016_topotecan.R (exponential effect on oral topotecan absorption rate constant Ka: Ka = Ka_pop * exp(1.06 * SNP_ABCG2_RS4148157); AG/AA carriers have Ka approximately 2.89x higher than GG homozygotes, corresponding to an observed Cmax approximately 1.7x higher).
• Notes: Time-fixed per subject (germline genotype). Variant carrier rate in the Roberts 2016 paediatric brain-tumour cohort: 19% (10 of 52 successfully genotyped patients; 9 AG heterozygotes plus 1 AA homozygote, with AA and AG pooled because only one AA homozygote was present). The rs4148157 variant is intronic and is in strong linkage disequilibrium with the better-characterised rs2231142 variant (Q141K, exon 6); Roberts 2016 Discussion proposes that rs4148157 may be acting as a surrogate marker for rs2231142 in this analysis. ABCG2 (also called BCRP, breast cancer resistance protein) is an intestinal and hepatic efflux transporter that influences oral bioavailability of topotecan and other substrates.

SNP_ABCG2_RS2231142_HOM (canonical for ABCG2 rs2231142 (Q141K) homozygous-variant indicator)

• Description: Binary genotype indicator for the ABCG2 (BCRP) rs2231142 single-nucleotide polymorphism (c.421C>A; exon 5; amino-acid Q141K / p.Gln141Lys; population MAF ~30-60% in East Asians, ~5-15% in Caucasians and African-Americans). 1 = homozygous variant (421A/A) carrier; 0 = otherwise (the union of heterozygous 421C/A and homozygous wild-type 421C/C). This is a recessive-model encoding: heterozygotes are pooled with wild-type homozygotes because Ueshima 2018 (the founding example) reported that only the 421A/A stratum had a distinct typical-value covariate effect on apixaban non-renal clearance. Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0. In a recessive-model (_HOM-only) encoding the 0-level is the union of homozygous wild-type 421C/C and heterozygous 421C/A. In a paired-indicator encoding (SNP_ABCG2_RS2231142_HET also present) the reference group is the 421C/C wild-type homozygote stratum, i.e. both indicators = 0.
• Source aliases:
  • ABCG2 421A/A – used in Ueshima_2018_apixaban.R (Ueshima 2018 Methods Eq. of the final-model CL/F: dichotomous parameter ABCG2 equals 1 for 421A/A and 0 for 421C/C or 421C/A).
  • rs2231142 TT – used in Jiang_2023_imatinib.R (Jiang 2023 reports the SNP on the opposite DNA strand as G>T, so the TT genotype is the 421A/A homozygous-variant stratum; Table 3 row rs2231142 TT).
• Example models: Ueshima_2018_apixaban.R (multiplicative factor on the non-renal arm of apparent oral clearance: cl_nonren = exp(lcl) * e_abcg2_homvar_cl_nonren^SNP_ABCG2_RS2231142_HOM with e_abcg2_homvar_cl_nonren = 0.341; 421A/A homozygotes have non-renal CL/F reduced by 65.9% relative to 421C/C or 421C/A; paper Table 4 theta6 = 0.341), Jiang_2023_imatinib.R (power-of-binary multiplicative factor on apparent oral clearance, paired with SNP_ABCG2_RS2231142_HET: e_abcg2_het_cl^SNP_ABCG2_RS2231142_HET * e_abcg2_hom_cl^SNP_ABCG2_RS2231142_HOM with e_abcg2_hom_cl = 0.976; TT homozygotes have CL/F 2.4% lower than GG wild-type homozygotes, imprecisely estimated on n = 6 TT subjects; paper Table 3), Toshimoto_2017_irinotecan_pbpk.R (same two efflux processes scaled to 51.3 percent of wild type; Toshimoto 2017 Supplementary Table 2A).
• Notes: Distinct from SNP_ABCG2_RS4148157 in two ways: (a) a different SNP (rs2231142 is the well-characterised coding Q141K variant in exon 5, whereas rs4148157 is an intronic variant in intron 11 that is in strong linkage disequilibrium with rs2231142 and used by Roberts 2016 as a surrogate marker), (b) a different genetic model (recessive 421A/A-only in Ueshima 2018 vs dominant any-A-allele in Roberts 2016 rs4148157). Strand orientation: papers report this SNP either as c.421C>A (dbSNP reference orientation, variant allele A) or as G>T on the opposite strand (variant allele T); the two are the same variant, so 421C/C = GG, 421C/A = GT, and 421A/A = TT. Always map the paper’s genotype labels onto the reference-orientation strata before assigning indicator values. Genetic-model orientations: use SNP_ABCG2_RS2231142_HOM alone for a recessive encoding (Ueshima 2018); pair it with SNP_ABCG2_RS2231142_HET when the source paper assigns a distinct typical-value effect to each of the three genotype strata (Jiang 2023), following the SLCO1B1_HAP15_HET / SLCO1B1_HAP15_HOM and CYP3A5_STAR1_HET / CYP3A5_STAR1_HOM precedents. The per-subject column value is identical in both encodings (1 iff the subject is 421A/A); only the model’s reference-category interpretation changes, which is why the paired form does not need a separate canonical for the homozygote. When a future paper uses a dominant-model encoding (any-A-allele = 1, pooling heterozygotes with homozygous variants into a single carrier group), register a paired companion canonical SNP_ABCG2_RS2231142_CARRIER so the two genetic-model orientations remain separate canonicals. The Q141K variant impairs ABCG2 plasma-membrane localisation and function; it is the most commonly studied ABCG2 pharmacogenetic SNP in popPK literature (substrates include apixaban, rosuvastatin, sulfasalazine, topotecan, methotrexate).

SNP_ABCG2_RS2231142_HET (canonical for ABCG2 rs2231142 (Q141K) heterozygote indicator)

• Description: Binary genotype indicator for the heterozygous stratum of the ABCG2 (BCRP) rs2231142 single-nucleotide polymorphism (c.421C>A; exon 5; amino-acid Q141K / p.Gln141Lys). 1 = subject carries exactly one variant allele (genotype 421C/A, equivalently GT on the opposite-strand reporting convention); 0 = otherwise (the union of 421C/C wild-type homozygotes and 421A/A homozygous-variant carriers; the paired indicator SNP_ABCG2_RS2231142_HOM flags the homozygous-variant group). Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (421C/C wild-type homozygote, when paired with SNP_ABCG2_RS2231142_HOM = 0). The reference group is the wild-type homozygote stratum; SNP_ABCG2_RS2231142_HOM flags the homozygous-variant stratum.
• Source aliases:
  • rs2231142 GT – used in Jiang_2023_imatinib.R (Jiang 2023 reports the SNP on the opposite DNA strand as G>T, so the GT genotype is the 421C/A heterozygous stratum; Table 3 row rs2231142 GT; the paper’s final-model equation writes the exponent as heterozygous).
• Example models: Jiang_2023_imatinib.R (power-of-binary multiplicative factor on apparent oral clearance: e_abcg2_het_cl^SNP_ABCG2_RS2231142_HET with e_abcg2_het_cl = 0.879; heterozygotes have CL/F 12.1% lower than GG wild-type homozygotes and therefore reach higher trough concentrations at the same dose; paper Table 3, RSE 5%, bootstrap 95% CI 0.785-0.968), Toshimoto_2017_irinotecan_pbpk.R (scales BCRP-mediated biliary excretion and enterocyte-to-lumen efflux of SN-38 to 87.5 percent of wild type, weighted by the 33 percent BCRP contribution to each of those two processes; Toshimoto 2017 Supplementary Table 2A).
• Notes: Paired with SNP_ABCG2_RS2231142_HOM (see that entry’s Notes for the strand-orientation convention, the recessive-vs-paired genetic-model distinction, and the three-level decomposition rationale). Follows the SLCO1B1_HAP15_HET / SLCO1B1_HAP15_HOM, CYP3A5_STAR1_HET / CYP3A5_STAR1_HOM, and CYP2D6_STAR10_HET / CYP2D6_STAR10_HOM precedents: two binary indicators encode a three-level germline genotype with the wild-type homozygote as the implicit reference (both indicators = 0). Distinct from a hypothetical SNP_ABCG2_RS2231142_CARRIER (not yet registered), which would pool heterozygotes with homozygous variants under a dominant model; use the paired indicators when the source paper resolves a distinct typical-value effect for the heterozygous stratum specifically. Jiang 2023 is the founding example and motivated the three-level decomposition because the East Asian cohort had abundant heterozygotes (45/85, 53%) but few homozygous variants (6/85, 7%) – the heterozygote effect (0.879) is precisely estimated while the homozygote effect (0.976) is not. Genotype distribution in the Jiang 2023 cohort of 85 Chinese postoperative GIST adults (Table 2): GG 34 (40%), GT 45 (53%), TT 6 (7%); allele frequency G 0.66 / T 0.34; Hardy-Weinberg p = 0.22.

SNP_ABCG2_RS2231142_T_COUNT (canonical for ABCG2 rs2231142 (Q141K) variant T-allele count)

• Description: Continuous individual-level count of ABCG2 (BCRP) rs2231142 variant alleles reported on the G>T strand convention: 0 = GG wild-type homozygote, 1 = GT heterozygote, 2 = TT variant homozygote. Equivalently 0 = 421C/C, 1 = 421C/A, 2 = 421A/A on the dbSNP c.421C>A reference orientation. Time-invariant (germline genotype). ABCG2 encodes breast cancer resistance protein, the efflux transporter that limits substrate absorption and facilitates renal and biliary excretion; the Q141K variant impairs plasma-membrane localisation and function. The count form carries the full three-genotype information in one column, so model() can derive a per-allele slope, a carrier indicator, or – as in the founding example – a term in a multi-SNP additive group score.
• Units: (count, 0/1/2 alleles per subject)
• Type: continuous
• Scope: general
• Reference category: n/a (continuous). A count of 0 (GG / 421C/C wild-type homozygote) is the natural reference.
• Source aliases:
  • ABCG2 rs2231142 (G > T) – Wei 2025 Results 3.2 and Methods ‘Grouping and combination of variants’. Same orientation as the canonical, no value transformation.
  • rs2231142 genotype string ("GG" / "GT" / "TT"): derive the count as the number of T alleles in the diplotype.
• Example models: Wei_2025_methotrexate_genotype.R (one of three variant-allele counts summed inside model() into a composite ABCC4-ABCG2-ADORA2A carrier indicator that multiplies methotrexate clearance by 0.91; this SNP is the worked example the source paper uses to fix its own grouping convention).
• Notes: Auto-approved member of the SNP_<GENE>_RS<rsid>_<allele>_COUNT family, following the SNP_ABCB1_RS1045642_T_COUNT, SNP_CYP2B6_RS3745274_T_COUNT and SNP_CYP4F2_RS2108622_T_COUNT precedents. Composable with the binary siblings SNP_ABCG2_RS2231142_HOM (recoverable as == 2) and SNP_ABCG2_RS2231142_HET (recoverable as == 1); the count is the more general encoding and should be preferred when the source resolves all three genotypes or needs an additive allele dose. Strand orientation: papers report this SNP either as c.421C>A (dbSNP reference, variant allele A) or as G>T on the opposite strand (variant allele T); these are the same variant, so always map the paper’s genotype labels onto the variant-allele count before assigning values – see the SNP_ABCG2_RS2231142_HOM Notes for the full mapping. Direction of effect in the founding paper: the T (variant) allele is associated with REDUCED methotrexate clearance, consistent with impaired BCRP-mediated efflux.

SNP_UGT1A1_RS8175347_HET (**canonical for UGT1A1 *28 heterozygote indicator**)

• Description: Binary genotype indicator for the UGT1A1 *28 promoter polymorphism (rs8175347): an extra TA dinucleotide in the TATA box of the UGT1A1 promoter, (TA)7TAA instead of the wild-type (TA)6TAA, which reduces transcription and therefore UGT1A1 glucuronidation capacity. 1 = subject carries exactly one *28 allele ((TA)6/(TA)7, commonly written *1/*28); 0 = otherwise (the union of *1/*1 wild-type homozygotes and *28/*28 homozygous-variant carriers, the latter flagged by the paired SNP_UGT1A1_RS8175347_HOM). Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (*1/*1 wild-type homozygote, when paired with SNP_UGT1A1_RS8175347_HOM = 0).
• Source aliases:
  • UGT1A1 *28, heterozygote column – Toshimoto 2017 Supplementary Table 2A.
  • *1/*28 – Toshimoto 2017 Methods, where the dominant and recessive genetic models are defined over the same three strata.
• Example models: Toshimoto_2017_irinotecan_pbpk.R (multiplies the UGT1A1-mediated SN-38 glucuronidation clearance in both liver and enterocyte by 0.602, i.e. heterozygotes retain 60.2 percent of wild-type glucuronidation capacity; Supplementary Table 2A).
• Notes: Paired with SNP_UGT1A1_RS8175347_HOM; see that entry’s Notes for the activity-ratio provenance caveat, which is load-bearing. *28 is the best-established pharmacogenetic determinant of irinotecan toxicity and is the variant behind Gilbert’s syndrome. Genotype frequencies in the Teft 2015 cohort that Toshimoto 2017 used to build its virtual population: *1/*1 44.66 percent, *1/*28 47.57 percent, *28/*28 7.77 percent. Distinct from the East Asian *6 allele (rs4148323, G71R), which is a coding rather than a promoter variant; if a future source models both, register *6 separately rather than reusing this column.

SNP_UGT1A1_RS8175347_HOM (**canonical for UGT1A1 *28 homozygous-variant indicator**)

• Description: Binary genotype indicator for the UGT1A1 *28 promoter polymorphism (rs8175347): an extra TA dinucleotide in the TATA box of the UGT1A1 promoter, (TA)7TAA instead of the wild-type (TA)6TAA, which reduces transcription and therefore UGT1A1 glucuronidation capacity. 1 = subject carries two *28 alleles ((TA)7/(TA)7, commonly written *28/*28); 0 = otherwise (the union of *1/*1 and *1/*28, the latter flagged by the paired SNP_UGT1A1_RS8175347_HET). Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (*1/*1 wild-type homozygote, when paired with SNP_UGT1A1_RS8175347_HET = 0).
• Source aliases:
  • UGT1A1 *28, homozygote column – Toshimoto 2017 Supplementary Table 2A.
• Example models: Toshimoto_2017_irinotecan_pbpk.R (multiplies the UGT1A1-mediated SN-38 glucuronidation clearance in both liver and enterocyte by 0.322; Supplementary Table 2A).
• Notes: Companion to SNP_UGT1A1_RS8175347_HET. Activity-ratio provenance caveat: Toshimoto 2017 Supplementary Table 2A footnote a states that the 60.2 / 32.2 percent activity ratios assigned to *28 were in fact measured for the UGT1A1 *6 polymorphism (Yamamoto 1998) and borrowed. A future extraction that needs a *28-specific in vitro activity ratio must not take these numbers as *28 measurements. The column itself is unaffected – it encodes genotype, not activity – but any model reusing it should carry its own sourced ratio.

SNP_SLCO1B1_RS4149056_HET (canonical for SLCO1B1 c.521T>C heterozygous (521T/C) indicator)

• Description: Binary genotype indicator for the SLCO1B1 (OATP1B1 hepatic uptake transporter) c.521T>C polymorphism (rs4149056, V174A / p.Val174Ala): 1 = subject carries the heterozygous (521T/C) genotype; 0 = otherwise (the paired SNP_SLCO1B1_RS4149056_HOM flags the other variant stratum). Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (wild-type homozygote, when paired with SNP_SLCO1B1_RS4149056_HOM = 0).
• Source aliases:
  • SLCO1B1 c.521T>C – Toshimoto 2017 Supplementary Table 2A.
• Example models: Toshimoto_2017_irinotecan_pbpk.R (multiplies the OATP1B1-mediated hepatic uptake clearance PSact,inf,h of BOTH SN-38 and SN-38G by 0.634; Supplementary Table 2A), Stillemans_2022_atorvastatin.R (atorvastatin CL/F = theta_CL * (1 + theta), theta = -0.402 for 521T/C; Table 3).
  • Beck_2022_elagolix.R (OATP1B1 ‘intermediate transporter’ = heterozygous 521T>C; raises elagolix relative bioavailability F1 by the fraction 0.421, i.e. F1 = 1.42 vs the extensive-transporter reference; Beck 2022 Table 3 footnote d). Paired there with SNP_SLCO1B1_RS4149056_MISSING for the non-genotyped stratum.
• Notes: Paired with SNP_SLCO1B1_RS4149056_HOM. c.521T>C is the classical reduced-function OATP1B1 variant (the defining SNP of the *5 allele and, in cis with c.388A>G, of the *15 / *17 haplotypes). Relationship to SLCO1B1_HAP15_HET / SLCO1B1_HAP15_HOM: those canonicals encode the phased *15 HAPLOTYPE (388G and 521C together); this column encodes the single c.521T>C SNP on its own. Use the haplotype canonicals when the source phases the two loci into *15, and this pair plus SNP_SLCO1B1_RS2306283_* when the source models the two SNPs as separate covariates – as Toshimoto 2017 does, combining their activities multiplicatively on the same transporter. The two encodings must not be mixed in one model. Toshimoto 2017 draws the joint genotype from the linkage-disequilibrium frequency table of its Supplementary Table 2B (Pasanen 2008) rather than treating the two loci as independent. Activity ratio 63.4 percent, from the authors’ unpublished in-house data (Supplementary Table 2A footnote b) – not from a citable source, so a future extraction should prefer its own published ratio.

SNP_SLCO1B1_RS4149056_HOM (canonical for SLCO1B1 c.521T>C homozygous-variant (521C/C) indicator)

• Description: Binary genotype indicator for the SLCO1B1 (OATP1B1 hepatic uptake transporter) c.521T>C polymorphism (rs4149056, V174A / p.Val174Ala): 1 = subject carries the homozygous-variant (521C/C) genotype; 0 = otherwise (the paired SNP_SLCO1B1_RS4149056_HET flags the other variant stratum). Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (wild-type homozygote, when paired with SNP_SLCO1B1_RS4149056_HET = 0).
• Source aliases:
  • SLCO1B1 c.521T>C – Toshimoto 2017 Supplementary Table 2A.
• Example models: Toshimoto_2017_irinotecan_pbpk.R (multiplies the OATP1B1-mediated hepatic uptake clearance PSact,inf,h of BOTH SN-38 and SN-38G by 0.268; Supplementary Table 2A), Stillemans_2022_atorvastatin.R (atorvastatin CL/F = theta_CL * (1 + theta), theta = -0.041 for 521C/C; Table 3).
  • Beck_2022_elagolix.R (OATP1B1 ‘poor transporter’ = homozygous-variant 521T>C; raises elagolix relative bioavailability F1 by the fraction 0.963, i.e. F1 = 1.96; Beck 2022 Table 3 footnote d).
• Notes: Companion to SNP_SLCO1B1_RS4149056_HET; see that entry’s Notes for the *15-haplotype relationship, the linkage-disequilibrium sampling and the unpublished-ratio caveat. Activity ratio 26.8 percent of wild type.

SNP_SLCO1B1_RS4149056_MISSING (canonical for SLCO1B1 c.521T>C genotype-missing indicator)

• Description: Binary indicator for a subject whose SLCO1B1 c.521T>C (rs4149056) genotype is not available (no pharmacogenetic sample, typically because the participant did not consent to testing). 1 = genotype missing; 0 = genotype known (in which case SNP_SLCO1B1_RS4149056_HET / SNP_SLCO1B1_RS4149056_HOM carry the genotype). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (genotype known). When SNP_SLCO1B1_RS4149056_MISSING == 1, SNP_SLCO1B1_RS4149056_HET and SNP_SLCO1B1_RS4149056_HOM must both be 0 so the genotype-indicator terms evaluate to zero for the non-genotyped subject.
• Source aliases:
  • Missing transporter – Beck 2022 Table 3 row ‘Missing transporter on F1’; Table 2 ‘OATP1B1 genotype status, Missing’ (545 of 2168, 25.1%).
• Example models: Beck_2022_elagolix.R (raises elagolix relative bioavailability F1 by the fraction 0.101, i.e. F1 = 1.10 vs the genotyped extensive-transporter reference; Beck 2022 Table 3 footnote d).
• Notes: Follows the missing-genotype indicator pattern of SNP_CYP2C19_RS3814637_MISSING, CYP2C9_MISSING and CYP2C19_MISSING: the source estimates a separate typical-value effect for the non-genotyped subgroup instead of imputing it as wild type. For a simulation of a fully genotyped population set it to 0 and draw the HET / HOM indicators from the genotype frequencies.

SNP_SLCO1B1_RS4149057_CC (canonical for SLCO1B1 c.571T>C homozygous CC genotype indicator)

• Description: Binary genotype indicator for the SLCO1B1 (OATP1B1 hepatic uptake transporter) c.571T>C polymorphism (rs4149057, L191L synonymous): 1 = subject carries the homozygous CC genotype; 0 = otherwise. Paired with SNP_SLCO1B1_RS4149057_TT; both indicators are 0 for the heterozygous TC group, which the founding source uses as the reference category. Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 together with SNP_SLCO1B1_RS4149057_TT = 0, i.e. the heterozygous TC group.
• Source aliases:
  • SNP3 SLCO1B1 571T>C (rs4149057), CC genotype – Yin 2021 Table 2 row CL_SNP3 (CC).
• Example models: Yin_2021_mitotane.R (multiplicative effect on apparent clearance of mitotane; the CC group has CL/F 0.753x the TC reference, Table 2 CL_SNP3 (CC) = 0.753).
• Notes: rs4149057 (c.571T>C, L191L) is a synonymous SLCO1B1 variant. Yin 2021 uses the heterozygous TC genotype as the reference (Eq. S5 sets the reference-category multiplier to 1) and estimates separate multipliers for the CC and TT homozygotes, hence the paired _CC / _TT indicators rather than a wild-type-anchored carrier/count column. This is distinct from the more commonly modelled reduced-function SLCO1B1 variants c.521T>C (SNP_SLCO1B1_RS4149056_*, rs4149056) and c.388A>G (SNP_SLCO1B1_RS2306283_*, rs2306283); rs4149057 is a different locus and the two must not be conflated. In the Yin 2021 cohort the effect direction is unusual – the CC homozygote lowers CL/F while the TT homozygote raises it 2.49-fold relative to TC – so the TC reference and the two-indicator encoding preserve the reported non-monotone pattern exactly.

SNP_SLCO1B1_RS4149057_TT (canonical for SLCO1B1 c.571T>C homozygous TT genotype indicator)

• Description: Binary genotype indicator for the SLCO1B1 (OATP1B1 hepatic uptake transporter) c.571T>C polymorphism (rs4149057, L191L synonymous): 1 = subject carries the homozygous TT genotype; 0 = otherwise. Paired with SNP_SLCO1B1_RS4149057_CC; both indicators are 0 for the heterozygous TC reference group. Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 together with SNP_SLCO1B1_RS4149057_CC = 0, i.e. the heterozygous TC group.
• Source aliases:
  • SNP3 SLCO1B1 571T>C (rs4149057), TT genotype – Yin 2021 Table 2 row CL_SNP3 (TT).
• Example models: Yin_2021_mitotane.R (multiplicative effect on apparent clearance of mitotane; the TT group has CL/F 2.49x the TC reference, Table 2 CL_SNP3 (TT) = 2.49).
• Notes: Companion to SNP_SLCO1B1_RS4149057_CC; see that entry’s Notes for the TC-reference encoding, the distinction from rs4149056 / rs2306283 and the non-monotone effect direction.

SNP_ERCC1_RS3212986_CC (canonical for ERCC1 rs3212986 (C8092A) homozygous wild-type C/C genotype indicator)

• Description: Binary genotype indicator for the ERCC1 (excision repair cross-complementation group 1, nucleotide-excision-repair DNA-repair gene) rs3212986 (C8092A, 3’-UTR) polymorphism: 1 = subject carries the homozygous wild-type C/C genotype; 0 = otherwise (A/C heterozygote or A/A homozygote). Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0, i.e. the A/C + A/A (variant-A-carrier) group.
• Source aliases:
  • ERCC1 (rs3212986) C/C – Cao 2022 final-model equation (If ERCC1 phenotype (rs3212986) = C/C, theta_ERCC1 = theta_2 (0.83); else 0).
• Example models: Cao_2022_pemetrexed.R (exponential effect on intercompartmental clearance Q of intravenous pemetrexed; exp(e_ercc1_q * SNP_ERCC1_RS3212986_CC) with e_ercc1_q = 0.83, so C/C homozygotes have about 2.3-fold higher Q than the A-carrier reference; Figure 2B).
• Notes: The variant allele at rs3212986 is A (the common / wild-type allele is C), so the covariate effect attaches to the wild-type C/C homozygote rather than to variant-allele carriage. A generic mutant-allele-presence indicator (SNP_ERCC1_RS3212986, 1 = any A allele) would group the A/C heterozygotes with the A/A homozygotes and flag exactly the reference stratum – the inverse of the reported effect – so the homozygote-specific _CC indicator is used (same wild-type-homozygote-anchored pattern as SNP_SLCO1B1_RS4149057_CC). Cao 2022 reports rs3212986 C/C homozygotes as having higher pemetrexed intercompartmental clearance than the two variant genotypes and speculates (Discussion) that ERCC1 genotype may modulate pemetrexed plasma-protein binding or tissue distribution; the mechanism is not established. Genotype counts in the cohort were A/A:A/C:C/C = 15:56:45 (Supplementary Table S2). Distinct from the other commonly investigated ERCC1 variant rs11615 (T19007C / N118N), which Cao 2022 genotyped but did not retain.

SNP_SLCO1B3_RS7311358_G_CARRIER (canonical for SLCO1B3 699A>G G-allele carrier indicator)

• Description: Binary indicator for carriage of the SLCO1B3 (OATP1B3 hepatic uptake transporter) c.699A>G polymorphism (rs7311358, I233M): 1 = subject carries at least one G allele (AG heterozygote or GG homozygote); 0 = AA. Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (AA, no G allele).
• Source aliases:
  • SNP2 SLCO1B3 699A>G (rs7311358) – Yin 2021 Table 2 row CL_SNP2 (AG/GG).
• Example models: Yin_2021_mitotane.R (multiplicative effect on apparent clearance of mitotane; G carriers have CL/F 0.601x the AA reference, i.e. 39.9% lower, Table 2 CL_SNP2 (AG/GG) = 0.601).
• Notes: Dominant (carrier) encoding, pooling AG and GG because Yin 2021 combined the minor-homozygote group with the heterozygote group (fewer than four minor homozygotes). In the Yin 2021 cohort rs7311358 was in 100% linkage disequilibrium with SLCO1B3 c.334G>T (rs4149117) and c.1557G>A (rs2053098); the authors retained rs7311358 as the variant with the strongest clinical annotation. Follows the SNP_<GENE>_RS<rsid>_<allele>_CARRIER pattern established by SNP_CYP1A2_RS762551_C_CARRIER. A future extraction that distinguishes AG from GG should add paired _HET / _HOM indicators.

SNP_SLCO1B1_RS2306283_HET (canonical for SLCO1B1 c.388A>G heterozygous (388A/G) indicator)

• Description: Binary genotype indicator for the SLCO1B1 (OATP1B1 hepatic uptake transporter) c.388A>G polymorphism (rs2306283, N130D / p.Asn130Asp): 1 = subject carries the heterozygous (388A/G) genotype; 0 = otherwise (the paired SNP_SLCO1B1_RS2306283_HOM flags the other variant stratum). Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (wild-type homozygote, when paired with SNP_SLCO1B1_RS2306283_HOM = 0).
• Source aliases:
  • SLCO1B1 c.388A>G – Toshimoto 2017 Supplementary Table 2A.
  • RS2306283 / RS230 (1 = G/G, 2 = A/G, 3 = A/A; derive HET = (RS2306283 == 2), HOM = (RS2306283 == 1)) – Wang 2019 Abstract and Section 3.3.
• Example models: Toshimoto_2017_irinotecan_pbpk.R (multiplies the OATP1B1-mediated hepatic uptake clearance PSact,inf,h of BOTH SN-38 and SN-38G by 1.61; Supplementary Table 2A), Wang_2019_methotrexate.R (multiplies methotrexate clearance by 0.647; Table 3 theta6 – a REDUCTION, the opposite direction to Toshimoto 2017).
• Notes: Paired with SNP_SLCO1B1_RS2306283_HOM. Direction of effect is opposite to most transporter variants: c.388A>G INCREASES OATP1B1 uptake activity (161 percent of wild type in heterozygotes, 221 percent in homozygotes per Toshimoto 2017 Supplementary Table 2A), so a carrier clears the substrate from blood into the liver FASTER. A model that assumes every variant reduces function will get the sign wrong. See SNP_SLCO1B1_RS4149056_HET Notes for the relationship to the SLCO1B1_HAP15_* haplotype canonicals, for the linkage-disequilibrium joint sampling with c.521T>C, and for the unpublished-ratio caveat (Supplementary Table 2A footnote b).

SNP_SLCO1B1_RS2306283_HOM (canonical for SLCO1B1 c.388A>G homozygous-variant (388G/G) indicator)

• Description: Binary genotype indicator for the SLCO1B1 (OATP1B1 hepatic uptake transporter) c.388A>G polymorphism (rs2306283, N130D / p.Asn130Asp): 1 = subject carries the homozygous-variant (388G/G) genotype; 0 = otherwise (the paired SNP_SLCO1B1_RS2306283_HET flags the other variant stratum). Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (wild-type homozygote, when paired with SNP_SLCO1B1_RS2306283_HET = 0).
• Source aliases:
  • SLCO1B1 c.388A>G – Toshimoto 2017 Supplementary Table 2A.
  • RS2306283 / RS230 (1 = G/G, 2 = A/G, 3 = A/A; derive HET = (RS2306283 == 2), HOM = (RS2306283 == 1)) – Wang 2019 Abstract and Section 3.3.
• Example models: Toshimoto_2017_irinotecan_pbpk.R (multiplies the OATP1B1-mediated hepatic uptake clearance PSact,inf,h of BOTH SN-38 and SN-38G by 2.21; Supplementary Table 2A), Wang_2019_methotrexate.R (multiplies methotrexate clearance by 0.805; Table 3 theta5 – a REDUCTION, the opposite direction to Toshimoto 2017).
• Notes: Companion to SNP_SLCO1B1_RS2306283_HET; see that entry’s Notes, in particular that this variant INCREASES rather than decreases OATP1B1 activity (221 percent of wild type).

SNP_AADAC_RS1803155_HOM (canonical for AADAC rs1803155 homozygous-variant (AA) indicator)

• Description: Binary genotype indicator for the AADAC (arylacetamide deacetylase) G>A polymorphism rs1803155: 1 = subject carries two variant A alleles (A/A); 0 = otherwise (the union of the G/A heterozygote and the G/G wild-type homozygote). Time-fixed per subject (germline genotype). AADAC is the microsomal serine esterase that deacetylates arylacetamide-containing drugs; for the rifamycins it catalyses the conversion of rifapentine to 25-desacetyl rifapentine and of rifampicin to 25-desacetyl rifampicin, so the variant is screened as a determinant of rifamycin clearance.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (G/A heterozygote or G/G wild-type homozygote, pooled).
• Source aliases:
  • AADAC rs1803155 – Francis 2019 Table 2 row AADAC rs1803155 (AA) effect on CL (%) and Table 3 row AADAC G > A rs1803155; used in Francis_2019_rifapentine.R.
• Example models: Francis_2019_rifapentine.R (multiplicative fractional shift on apparent oral clearance, (1 + e_snp_aadac_rs1803155_hom_cl * SNP_AADAC_RS1803155_HOM) with the coefficient -0.104, i.e. AA homozygotes clear rifapentine 10.4 percent more slowly than G carriers; Francis 2019 Table 2, delta-OFV 6.2, P = 0.013).
• Notes: The reference category pools G/A with G/G rather than contrasting each against wild type, and that pooling is a finding of the source rather than a modelling convenience: Francis 2019 Results state that the three genotypes were first estimated as separate groups, that the GG and GA effects came out similar and neither was statistically significant, and that only 3 of 162 genotyped patients were G/G, so no meaningful separate G/G estimate was obtainable. A future extraction with enough G/G subjects to separate the strata should register a paired SNP_AADAC_RS1803155_HET on the usual _HET / _HOM pattern rather than redefining this column’s reference category. Note also that the variant allele is the major allele in this population – frequency 0.82 in the South African cohort, with 65.4 percent A/A – against 0.50 to 0.64 reported in European American, African American, Korean and Japanese populations, so the “variant = 1” orientation does not mean “minority = 1” here. Sloan et al. found no comparable AADAC effect on rifampicin pharmacokinetics in Malawian patients; the source attributes the contrast to differing variant-genotype prevalence between African ethnic groups.

SNP_AGBL4_RS319952_GG (canonical for AGBL4 rs319952 homozygous GG genotype indicator)

• Description: Binary genotype indicator for the AGBL4 (ATP/GTP-binding protein-like 4, also known as CCP6) single-nucleotide polymorphism rs319952 (A>G). 1 = subject’s reported genotype is GG (homozygous for the G allele); 0 = otherwise (the pooled AA and AG groups). Time-fixed per subject (germline genotype). AGBL4 encodes a cytosolic carboxypeptidase that removes glutamate residues from the carboxy-terminus of tubulin; it has no known direct role in drug metabolism, and the association with bedaquiline clearance is empirical. Named by the reported genotype letters rather than by wild-type / variant status: the A allele is the reference (major) allele in the founding cohort, so GG is the minor-allele homozygote.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the AA and AG genotypes pooled; A is the reference allele).
• Source aliases:
  • rs319952 genotype G (vs A&AG) – Zou 2022 Results Equation 2 and Table 3 row theta rs319952 on CL/F; used in Zou_2022_bedaquiline.R.
• Example models: Zou_2022_bedaquiline.R (an additive shift of -1.4 L/h on typical apparent clearance, exp(lcl) * (GGT/28.9)^-0.476 + e_snp_agbl4_rs319952_gg_cl * SNP_AGBL4_RS319952_GG with e_snp_agbl4_rs319952_gg_cl = -1.4; Zou 2022 Table 3, RSE 27.1%). Note this is the one register SNP whose effect enters as an absolute L/h shift rather than a multiplicative / exponential factor, because the source parameterised the genotype effect that way (CL/F 1.4 L/h lower for GG).
• Notes: First AGBL4 entry in the register. The founding source (Zou 2022) genotyped AA/AG/GG = 31/39/13 of 99 Chinese MDR-TB patients, with 16 ungenotyped; the ungenotyped subjects were imputed to the most frequent class (AG) and therefore entered the fit with this indicator = 0. The AA+AG pooling (rather than a paired _HET / _HOM contrast) is the source’s own finding: rs319952 was retained as a GG-versus-AA&AG dichotomy in the backward-elimination step. A future extraction that resolves all three strata should register a paired SNP_AGBL4_RS319952_HET on the usual _HET / _HOM pattern rather than redefining this column’s reference. The association was reported for antituberculosis drug-induced liver toxicity in an Ethiopian GWAS but not reproduced in the Chinese Han population; its mechanistic link to bedaquiline clearance is unestablished.

SNP_ABCC2_RS717620_HET (canonical for ABCC2 c.-24C>T heterozygote indicator)

• Description: Binary genotype indicator for the ABCC2 (MRP2 canalicular and apical efflux transporter) c.-24C>T promoter polymorphism (rs717620), a 5’-untranslated-region variant associated with reduced MRP2 expression. 1 = subject carries exactly one variant allele (C/T); 0 = otherwise (the paired SNP_ABCC2_RS717620_HOM flags the T/T stratum). Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (C/C wild-type homozygote, when paired with SNP_ABCC2_RS717620_HOM = 0).
• Source aliases:
  • ABCC2 -24C>T – Toshimoto 2017 Supplementary Table 2A.
• Example models: Toshimoto_2017_irinotecan_pbpk.R (scales MRP2-mediated biliary excretion and enterocyte-to-lumen efflux to 66.6 percent of wild type, weighted by the MRP2 contribution to each compound: 50 percent for irinotecan, 33 percent for SN-38 and 100 percent for SN-38G).
• Notes: Paired with SNP_ABCC2_RS717620_HOM. The 66.6 / 33.3 percent activity ratios are recorded in Toshimoto 2017 Supplementary Table 2A as Assumption – the authors assumed a one-third loss of function per variant allele rather than measuring or citing one, so the ratios carry no experimental support and a future extraction should supply its own. Genotype frequencies in the Teft 2015 cohort: C/C 58.25 percent, C/T 32.04 percent, T/T 9.71 percent. Distinct from the SNP_ABCC4_* columns (a different ABC-C-subfamily transporter gene).

SNP_ABCC2_RS717620_HOM (canonical for ABCC2 c.-24C>T homozygous-variant indicator)

• Description: Binary genotype indicator for the ABCC2 (MRP2 canalicular and apical efflux transporter) c.-24C>T promoter polymorphism (rs717620), a 5’-untranslated-region variant associated with reduced MRP2 expression. 1 = subject carries two variant alleles (T/T); 0 = otherwise (the paired SNP_ABCC2_RS717620_HET flags the C/T stratum). Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (C/C wild-type homozygote, when paired with SNP_ABCC2_RS717620_HET = 0).
• Source aliases:
  • ABCC2 -24C>T – Toshimoto 2017 Supplementary Table 2A.
• Example models: Toshimoto_2017_irinotecan_pbpk.R (scales MRP2-mediated biliary excretion and enterocyte-to-lumen efflux to 33.3 percent of wild type, weighted by the per-compound MRP2 contribution).
• Notes: Companion to SNP_ABCC2_RS717620_HET; see that entry’s Notes for the Assumption provenance of the activity ratios and the cohort genotype frequencies.

SNP_ABCC4_RS2274407_G_COUNT (canonical for ABCC4 rs2274407 variant G-allele count)

• Description: Continuous individual-level count of ABCC4 rs2274407 variant alleles as defined on the T>G orientation: 0 = TT wild-type homozygote, 1 = TG heterozygote, 2 = GG variant homozygote. Time-invariant (germline genotype). ABCC4 encodes multidrug-resistance protein 4 (MRP4), an ATP-binding-cassette C-subfamily efflux transporter expressed across many tissues and tumours. The variant sits at the 3’ splice acceptor site of exon 8; the cited functional work reports that it does not alter MRP4 transport activity directly but may disrupt normal splicing of ABCC4 pre-mRNA.
• Units: (count, 0/1/2 alleles per subject)
• Type: continuous
• Scope: general
• Reference category: n/a (continuous). A count of 0 (TT wild-type homozygote) is the natural reference.
• Source aliases:
  • ABCC4 rs2274407 (T > G) – Wei 2025 Results 3.2, which is the definition this canonical follows.
  • ABCC4 G912T – the older cDNA nomenclature used in the Wei 2025 Discussion and in the cited Mesrian Tanha 2017 source, which describes the same variant on the opposite orientation and therefore calls the clearance-increasing allele “T”. These two labels invert each other: resolve any new source against the dbSNP reference orientation before assigning counts, and use the direction-of-effect statement rather than the letter as the check.
• Example models: Wei_2025_methotrexate_genotype.R (one of three variant-allele counts summed inside model() into a composite ABCC4-ABCG2-ADORA2A carrier indicator that multiplies methotrexate clearance by 0.91).
• Notes: Auto-approved member of the SNP_<GENE>_RS<rsid>_<allele>_COUNT family. Direction of effect in the founding paper: the G (variant) allele is associated with REDUCED methotrexate clearance – the Wei 2025 Discussion states the converse, that “the T allele was associated with increased MTX clearance”, which is the same statement under the T>G definition the same paper gives in Results 3.2. Because this variant is reported under two mutually inverted nomenclatures in the founding paper itself, a future extraction MUST confirm which orientation its source uses before reusing this column; a silent mis-assignment flips the sign of the effect. Distinct from SNP_ABCC2_* and SNP_ABCB1_* columns (different transporter genes).

SNP_ADORA2A_RS2298383_T_COUNT (canonical for ADORA2A rs2298383 variant T-allele count)

• Description: Continuous individual-level count of ADORA2A rs2298383 variant alleles on the C>T orientation: 0 = CC wild-type homozygote, 1 = CT heterozygote, 2 = TT variant homozygote. Time-invariant (germline genotype). ADORA2A encodes the adenosine A2A receptor; rs2298383 lies in a putative promoter region and has been associated with transcriptional regulation, with higher mRNA levels reported in CC-genotype carriers. Methotrexate’s anti-inflammatory and neurotoxic effects are partly adenosine-mediated, which is the mechanistic rationale for screening this receptor variant against methotrexate disposition.
• Units: (count, 0/1/2 alleles per subject)
• Type: continuous
• Scope: general
• Reference category: n/a (continuous). A count of 0 (CC wild-type homozygote) is the natural reference.
• Source aliases:
  • ADORA2A rs2298383 (C > T) – Wei 2025 Results 3.2. Same orientation as the canonical, no value transformation.
• Example models: Wei_2025_methotrexate_genotype.R (one of three variant-allele counts summed inside model() into a composite ABCC4-ABCG2-ADORA2A carrier indicator that multiplies methotrexate clearance by 0.91).
• Notes: Auto-approved member of the SNP_<GENE>_RS<rsid>_<allele>_COUNT family. Direction of effect in the founding paper: the T (variant) allele is associated with REDUCED methotrexate clearance (Wei 2025 Discussion: “the C allele was associated with increased MTX clearance”). This is a pharmacodynamic-receptor gene rather than a transporter or metabolising-enzyme gene, so a clearance association is not mechanistically obvious; the founding paper retains it only as one third of a composite score whose whole effect is a 9% clearance reduction, and explicitly concedes the effect “may have limited clinical significance”. Do not reuse this column as a standalone clearance covariate without re-establishing the association in the new source.

SNP_IL10_RS1800896_HET (canonical for IL-10 G-1082A (rs1800896) heterozygote indicator)

• Description: Binary genotype indicator for the heterozygous stratum of the IL10 rs1800896 promoter single-nucleotide polymorphism (-1082 G>A, reported as T>C on the opposite strand). 1 = subject carries exactly one variant allele (genotype C/T on the reporting strand used by the source); 0 = otherwise. Time-fixed per subject (germline genotype). IL-10 is a potent down-modulator of CYP3A enzyme activity, so carriers of the higher-IL-10-expression genotype show reduced CYP3A-mediated clearance.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (rs1800896 T/T homozygote on the source’s reporting strand).
• Source aliases:
  • rs1800896-TC – Pei 2023 Table S4 row label for the coefficient; the paper writes the locus as “IL-10 G-1082A (rs1800896)” in the text and tabulates the genotype counts under “IL-10 (rs1800896)” in Table 2.
• Example models: Pei_2023_tacrolimus.R (exponential effect on apparent oral clearance: exp(e_il10_het_cl * SNP_IL10_RS1800896_HET) with e_il10_het_cl = -0.35, i.e. heterozygotes have 0.70 times the tacrolimus CL/F of T/T homozygotes; Pei 2023 Table S4).
• Notes: A future paper resolving all three strata should add the paired SNP_IL10_RS1800896_HOM following the SNP_ABCG2_RS2231142_HET / _HOM and CYP3A5_STAR1_HET / _HOM precedents, or register a _VAR_COUNT sibling if it fits a per-allele model. Auto-approved member of the SNP_<GENE>_RS<rsid> family. Distinct from the other two IL-10 loci genotyped in Pei 2023 (rs1800871 and rs1800872, Table S3), neither of which entered the final model.

SNP_ICAM1_RS1799969 (canonical for ICAM-1 rs1799969 mutant indicator)

• Description: Binary genotype indicator for the ICAM1 rs1799969 single-nucleotide polymorphism (G > A; Gly241Arg / K469E in some references). 1 = at least one mutant (A) allele present (heterozygous or homozygous mutant); 0 = homozygous wild-type (GG).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (homozygous wild-type).
• Source aliases:
  • cat – Papachristos 2020 (the paper writes the indicator as cat in the CL covariate equation; no formal column name is given in the published narrative).
• Example models: Papachristos_2020_bevacizumab_pk.R, Papachristos_2020_bevacizumab_qss.R, Papachristos_2020_bevacizumab_pkpd.R (multiplicative effect on bevacizumab CL: CL * exp(-0.423 * SNP_ICAM1_RS1799969) in the PK and PK/PD models; CL * exp(-0.33 * SNP_ICAM1_RS1799969) in the binding QSS model – mutant carriers have lower CL and higher trough levels).
• Notes: Time-fixed per subject. Mutant carrier rate in the Papachristos 2020 mCRC cohort: 20% (Notes Table 1 of the paper). The biological mechanism by which the ICAM1 mutant slows bevacizumab clearance is unknown; the association is empirical and may be specific to mCRC.

SNP_VEGFA_RS1570360 (canonical for VEGF-A rs1570360 mutant indicator)

• Description: Binary genotype indicator for the VEGFA rs1570360 single-nucleotide polymorphism (-1154 G > A; promoter region). 1 = at least one mutant (A) allele present; 0 = homozygous wild-type (GG).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (homozygous wild-type).
• Source aliases:
  • cat1 – Papachristos 2020 (used as the first categorical indicator in the inter-compartmental clearance equation of the PK model; no formal column name in the narrative).
• Example models: Papachristos_2020_bevacizumab_pk.R (multiplicative effect on bevacizumab Q: Q * exp(0.378 * SNP_VEGFA_RS1570360) – mutant carriers have higher inter-compartmental clearance).
• Notes: Time-fixed per subject. Mutant carrier rate in the Papachristos 2020 mCRC cohort: 33%. The covariate is significant in the standalone PK model but does not appear in the binding QSS or PK/PD models because in those models the inter-compartmental clearance covariate effects are absorbed into the rs699947 effect on Q (PK/PD) or into the K_ss / BM0 effects (QSS).

SNP_VEGFA_RS699947 (canonical for VEGF-A rs699947 mutant indicator)

• Description: Binary genotype indicator for the VEGFA rs699947 single-nucleotide polymorphism (-2578 C > A; promoter region). 1 = at least one mutant (A) allele present; 0 = homozygous wild-type (CC).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (homozygous wild-type).
• Source aliases:
  • cat2 – Papachristos 2020 PK model (second categorical indicator on Q).
  • cat – Papachristos 2020 binding QSS model (effect on K_ss and BM0) and PK/PD model (effect on Q).
• Example models: Papachristos_2020_bevacizumab_pk.R (effect on Q: -0.429), Papachristos_2020_bevacizumab_qss.R (effect on K_ss: +1.22, on BM0: -0.851), Papachristos_2020_bevacizumab_pkpd.R (effect on Q: -0.414).
• Notes: Time-fixed per subject. Mutant carrier rate in the Papachristos 2020 mCRC cohort: 52%. The mutant allele is associated with lower baseline free VEGF-A levels and a higher in-vivo affinity (higher K_ss), consistent with reports that rs699947 mutants have prolonged overall survival on bevacizumab-based therapy.

SNP_SLCO1B1_RS11045819 (canonical for SLCO1B1 rs11045819 mutant indicator)

• Description: Binary genotype indicator for the SLCO1B1 rs11045819 single-nucleotide polymorphism (C > A; OATP1B1 transporter, exon 5, P155T). 1 = at least one mutant (A) allele present (heterozygous AC or homozygous AA); 0 = homozygous wild-type (CC). The Hennig 2015 cohort (n = 35 successfully genotyped of 44) reported 5 AC heterozygotes, 30 CC homozygotes, and 0 AA homozygotes, so the indicator is effectively heterozygous-vs-CC in that cohort.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (homozygous wild-type CC).
• Source aliases:
  • SLCO1B1 rs11045819 genotype – Hennig 2015 (paper text; the source NONMEM control stream is in the unrecovered AAC supplement, so the formal column name is not available).
• Example models: Hennig_2015_rifabutin.R (multiplicative effect on rifabutin bioavailability F: F * (1 + 0.304 * SNP_SLCO1B1_RS11045819) – AC carriers have ~30% higher rifabutin F than CC reference; dOFV = -6.5).
• Notes: Time-fixed per subject. Carrier rate in the Hennig 2015 South-African HIV/TB cohort: 14% (5 of 35 genotyped). SLCO1B1 encodes OATP1B1, a hepatic uptake transporter; rs11045819 has been associated with reduced rifampicin and lopinavir concentrations in prior studies but in Hennig 2015 was associated with INCREASED rifabutin bioavailability (note opposite direction of effect across rifamycins).

SNP_SLCO3A1_RS8027174 (canonical for SLCO3A1 rs8027174 variant indicator)

• Description: Binary genotype indicator for the SLCO3A1 rs8027174 single-nucleotide polymorphism (g.91941607G>T; intron variant; encodes the OATP3A1 organic anion transporter). 1 = at least one variant (T) allele present (heterozygous G/T or homozygous T/T); 0 = homozygous G/G. Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (homozygous G/G).
• Source aliases:
  • SLCO3A1 G>T – Stillemans 2021 (Table 3 row ‘SLCO3A1 G>T on V’; the NONMEM control stream is not published).
  • SLCO3A1 g.91941607G>T – Stillemans 2021 Eur J Clin Pharmacol (Table 1 genotype row; Table 2 row ‘theta SLCO3A1 on V’).
• Example models: Stillemans_2021_darunavir.R (categorical effect on darunavir V/F: V * (1 + 0.81 * SNP_SLCO3A1_RS8027174) – carriers have 81% larger apparent volume), Stillemans_2021b_darunavir.R (same categorical form in the reduced no-AAG model re-estimated on the merged learning + validation data: V * (1 + 0.697 * SNP_SLCO3A1_RS8027174)).
• Notes: In the Stillemans 2021 Belgian HIV cohort 18 of 123 genotyped subjects were G/T and none T/T (Table 2), so the indicator is effectively heterozygote-vs-G/G there. Distinct from the other SLCO3A1 SNP genotyped in that study, rs4294800 (g.91917464G>A), which was screened on CL and V but not retained.

SNP_SLC22A1_RS628031_HOM (canonical for SLC22A1 (OCT1) 1222G>A (rs628031) homozygous-variant indicator)

• Description: Binary genotype indicator for the homozygous-variant stratum of the SLC22A1 rs628031 single-nucleotide polymorphism (c.1222G>A; exon 7; amino-acid M408V / p.Met408Val; encodes OCT1, the hepatic organic cation uptake transporter). 1 = subject carries two variant alleles (genotype 1222A/A); 0 = otherwise (the union of 1222G/G wild-type homozygotes and 1222G/A heterozygotes). Recessive-model encoding: heterozygotes are pooled with wild-type homozygotes because the founding paper resolved a distinct typical-value effect only for the AA stratum. Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (pooled 1222G/G wild-type homozygotes and 1222G/A heterozygotes).
• Source aliases:
  • SLC22A1 -1222AA – Wang 2019 Table 4 row theta SLC22A1-1222 AA on CL/F and the final-model E_SLC22A1-1222AA factor; same orientation as the canonical, no value transformation. Used in Wang_2019_lamotrigine.R.
• Example models: Wang_2019_lamotrigine.R (fractional multiplicative factor on apparent oral clearance: cl *= (1 + (-0.525) * SNP_SLC22A1_RS628031_HOM), i.e. AA homozygotes clear lamotrigine 52.5% more slowly than the pooled GG + GA reference; Wang 2019 Table 4, %CV 29.5, bootstrap 95% CI -0.79 to -0.19).
• Notes: Auto-approved member of the SNP_<GENE>_RS<rsid> family. OCT1 mediates active uptake of lamotrigine into hepatocytes and human brain endothelial cells (Wang 2019 Discussion, citing Dickens 2012), so a loss-of-function variant reduces hepatic delivery and therefore apparent clearance – the direction observed in the founding model and in Shen 2016. Nomenclature trap: Wang 2019’s Discussion writes the variant as “1222A>G” while its own Tables 3 and 4 and Methods write “1222G>A”; the genotype frequencies settle it (GG 46.1%, GA 48.3%, AA 5.6% in 89 Han Chinese patients), so A is the minor/variant allele and AA is the rare homozygous-variant stratum. Distinct from the sibling SNP SLC22A1 1022C>T (rs2282143), which the same paper screened and rejected. Register a companion SNP_SLC22A1_RS628031_HET if a future source estimates a separate heterozygote effect, following the SLCO1B1_HAP15_HET / SLCO1B1_HAP15_HOM precedent.

SNP_ABCG2_RS2231137_HOM (canonical for ABCG2 (BCRP) 34G>A (rs2231137, V12M) homozygous-variant indicator)

• Description: Binary genotype indicator for the homozygous-variant stratum of the ABCG2 rs2231137 single-nucleotide polymorphism (c.34G>A; exon 2; amino-acid V12M / p.Val12Met; encodes breast cancer resistance protein, BCRP). 1 = subject carries two variant alleles (genotype 34A/A); 0 = otherwise (the union of 34G/G wild-type homozygotes and 34G/A heterozygotes). Recessive-model encoding. Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (pooled 34G/G wild-type homozygotes and 34G/A heterozygotes).
• Source aliases:
  • ABCG2 -34AA – Wang 2019 Table 4 row theta ABCG2-34 AA on V/F and the final-model E_ABCG2-34AA factor; same orientation as the canonical, no value transformation. Used in Wang_2019_lamotrigine.R.
• Example models: Wang_2019_lamotrigine.R (fractional multiplicative factor on apparent central volume of distribution: vc *= (1 + (-0.420) * SNP_ABCG2_RS2231137_HOM), i.e. AA homozygotes have V/F 42.0% lower than the pooled GG + GA reference; Wang 2019 Table 4, %CV 35.5, bootstrap 95% CI -0.65 to -0.14).
• Notes: Auto-approved member of the SNP_<GENE>_RS<rsid> family. Do not confuse with the rs2231142 (Q141K, c.421C>A) entries already in this register (SNP_ABCG2_RS2231142_HOM / _HET / _T_COUNT): rs2231137 is a different coding variant in exon 2, and the founding paper screened BOTH and retained only rs2231137 – “In our final model, ABCG2 421C>A had no effect on LTG concentrations, … whereas ABCG2 34G>A was found to be a significant covariate” (Wang 2019 Discussion). The 34A allele is linked to reduced ABCG2 translation and lower mRNA expression (Tamura 2007, Rudin 2008, cited in the founding paper), and lamotrigine is an established BCRP substrate (Romermann 2015). Founding cohort (Wang 2019 Results): GG 36.0%, AA 13.5% of 89 Han Chinese patients with epilepsy. Register SNP_ABCG2_RS2231137_HET or _A_COUNT companions if a future source resolves the heterozygous stratum or needs an additive allele dose.

SNP_ABCB1_RS2032582_HOM (canonical for ABCB1 / MDR1 2677G>T/A (rs2032582) homozygous 2677TT indicator)

• Description: Binary genotype indicator for the 2677T/T stratum of the ABCB1 rs2032582 single-nucleotide polymorphism (c.2677G>T/A; exon 21; amino-acid Ala893Ser (T allele) or Ala893Thr (A allele); encodes P-glycoprotein / MDR1). 1 = subject is homozygous 2677T/T; 0 = every other genotype. Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (any genotype other than 2677T/T – GG, GT, GA, TA and AA are all pooled).
• Source aliases:
  • MDR1 -2677TT – Wang 2019 Table 4 row theta MDR1-2677 TT + C3435 TT on V/F (the TT half of the paper’s combined two-SNP covariate) and Results ‘Frequencies of UGTs, MDR1, ABBC2, and SLC22A1 Genotyping Variants’. Used in Wang_2019_lamotrigine.R.
• Example models: Wang_2019_lamotrigine.R (multiplied inside model() with SNP_ABCB1_RS1045642_HOM to form the source’s joint 2677TT + 3435TT indicator, which then enters apparent volume as vc *= (1 + 1.390 * mdr1TTTT) – a 139% (2.39-fold) increase in V/F for subjects who are TT at both loci; Wang 2019 Table 4, %CV 43.0, bootstrap 95% CI 0.21-3.25).
• Notes: Auto-approved member of the SNP_<GENE>_RS<rsid> family. Tri-allelic locus: rs2032582 carries a G, a T and an A allele, so “2677TT” is one of six possible genotypes rather than the usual homozygous-variant half of a biallelic pair; the 2677A allele therefore falls into the 0 reference group. Record the per-paper genotype-to-indicator mapping in covariateData[[...]]$notes, since a source that pools TA with TT would need a different canonical. Distinct from ABCB1_HAP_TTT, which is the phased rs1128503 / rs2032582 / rs1045642 haplotype: Wang 2019 did not phase, it simply crossed two unphased genotype indicators after observing that 91.67% of its 2677TT carriers were also 3435TT, so the product-of-indicators encoding – not the haplotype canonical – is the faithful one. Compose, do not create a merged canonical: when a source combines two SNPs into one covariate variable, register (or reuse) the single-rsid indicators and form the joint stratum inside model(), following the composite-score precedent in Wei_2025_methotrexate_genotype.R. Register SNP_ABCB1_RS2032582_HET or a _T_COUNT companion if a future source resolves the heterozygous stratum or needs an additive allele dose.

SNP_ABCB1_RS1045642 (canonical for ABCB1 rs1045642 (c.3435C>T) mutant allele carrier indicator)

• Description: Binary genotype indicator for the ABCB1 rs1045642 single-nucleotide polymorphism (c.3435C>T; exon 26; synonymous Ile1145Ile; encodes P-glycoprotein / MDR1 efflux transporter). 1 = subject carries at least one T (mutant) allele (heterozygous CT or homozygous TT; pooled because TT homozygote frequency was 1 of 262 in the Bisaso 2014 cohort); 0 = homozygous wild-type (CC). Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (homozygous wild-type CC).
• Source aliases:
  • ABCB13435 – used in Bisaso_2014_albumin.R (paper text “ABCB1c.3435C>T mutation” and Figure 3 stratification “ABCB13435==0 stands for ABCB1c.3435CC while ABCB13435==1 stands for ABCB1c.3435CT and ABCB1c.3435TT”; same orientation as the canonical).
• Example models: Bisaso_2014_albumin.R (multiplicative additive shift on baseline albumin secretion rate Q0: Q0 = exp(lq0) * (1 + e_snp_abcb1_rs1045642_q0 * SNP_ABCB1_RS1045642) with e_snp_abcb1_rs1045642_q0 = 0.167; T-carriers have 16.7% higher Q0 than CC wild-type, equivalent to the paper text’s “16% higher” framing).
• Notes: Distinct from ABCB1_HAP_TTT (which is the multi-SNP haplotype across rs1128503 / rs2032582 / rs1045642 jointly – a different concept even though rs1045642 is one of the three contributing SNPs); use ABCB1_HAP_TTT when the source paper reports a phased haplotype, and SNP_ABCB1_RS1045642 when the source paper reports the single c.3435C>T SNP alone. Heterozygote and homozygote T-carriers are pooled in Bisaso 2014 because the TT cohort was n = 1; extractions that instead assign the effect to the homozygous-variant stratum alone use the sibling SNP_ABCB1_RS1045642_HOM (registered alongside Gu_2025_rivaroxaban.R, which pools CC and CT into its reference group), and a source estimating a separate heterozygote effect would add the matching SNP_ABCB1_RS1045642_HET, following the SLCO1B1_HAP15_HET / SLCO1B1_HAP15_HOM precedent. ABCB1 c.3435C>T is associated with altered P-glycoprotein expression and has been linked in the literature to predisposition to ART and rifampicin-based anti-TB drug-induced liver injury (Yimer 2011, cited in Bisaso 2014 Discussion).

SNP_ABCB1_RS1045642_HOM (canonical for ABCB1 rs1045642 (c.3435C>T) homozygous-variant indicator)

• Description: Binary genotype indicator for the homozygous-variant stratum of the ABCB1 rs1045642 single-nucleotide polymorphism (c.3435C>T; exon 26; synonymous Ile1145Ile; encodes the P-glycoprotein / MDR1 efflux transporter). 1 = subject carries two variant alleles (genotype c.3435TT, written AA in the dbSNP A/G orientation); 0 = otherwise (the union of c.3435CC wild-type homozygotes and CT heterozygotes). Time-fixed per subject (germline genotype). Use this canonical – not the carrier-orientation SNP_ABCB1_RS1045642 – when the source paper assigns the covariate effect to the homozygous-variant group alone and pools heterozygotes with wild-type homozygotes.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (pooled c.3435CC wild-type homozygotes and CT heterozygotes).
• Source aliases:
  • AA – Gu 2025 Results section 3.2 final-model equation (AA = 1 for AA genotype of rs1045642, AA = 0 for other genotypes of rs1045642), and Table 3 row CL_A642. Gu 2025 genotypes rs1045642 on the A/G strand, so its AA stratum is the c.3435TT homozygous-variant group; same orientation as the canonical, no value transformation.
• Example models: Gu_2025_rivaroxaban.R (log-additive effect on apparent clearance: exp(-0.204 * SNP_ABCB1_RS1045642_HOM), equivalently the multiplicative factor 0.815 reported as CL_A642 in Gu 2025 Table 3, so AA subjects clear rivaroxaban about 18.5% more slowly than the pooled non-AA reference; Gu 2025 screened the SNP first as a four-category variable, found only the AA stratum separated, and reclassified it as this two-category covariate), Wang_2019_lamotrigine.R (used ONLY as one factor of a two-SNP product: multiplied inside model() with SNP_ABCB1_RS2032582_HOM to build Wang 2019’s combined MDR1-2677TT + C3435TT indicator, which raises apparent lamotrigine volume by 139%. The 3435TT indicator carries no independent effect in that model).
• Notes: The homozygous half of the paired-indicator decomposition anticipated in the SNP_ABCB1_RS1045642 notes, following the SLCO1B1_HAP15_HET / SLCO1B1_HAP15_HOM, CYP2D6_STAR10_HET / CYP2D6_STAR10_HOM and CYP3A5_STAR1_HET / CYP3A5_STAR1_HOM precedents. The companion SNP_ABCB1_RS1045642_HET indicator is deliberately NOT registered here because Gu 2025 pools heterozygotes into the reference group and estimates no separate heterozygote effect; register it when a source paper estimates one. Distinct from SNP_ABCB1_RS1045642 (1 = any T-allele carrier, het or hom) and from ABCB1_HAP_TTT (the phased three-SNP haplotype). Beware the nomenclature trap in the founding paper: the Gu 2025 Discussion calls the AA stratum the “wild genotype”, which contradicts both the standard dbSNP orientation (A = the c.3435T variant allele) and the paper’s own allele frequencies (G = 0.60, A = 0.40 in its NVAF cohort, matching the Han-Chinese c.3435C frequency); the canonical follows the genotype coding of the printed equation rather than the prose label.

SNP_ABCB1_RS1045642_T_COUNT (canonical for ABCB1 3435C>T (rs1045642) T-allele count)

• Description: Continuous individual-level count of ABCB1 c.3435C>T (rs1045642) T alleles: 0 = CC homozygous wild type, 1 = CT heterozygous, 2 = TT homozygous variant. Time-invariant (germline genotype). ABCB1 encodes P-glycoprotein / MDR1, the efflux transporter of which lenalidomide is a weak substrate; the c.3435C>T synonymous variant (exon 26, Ile1145Ile) alters P-gp expression. The count form carries the full three-genotype information in one column, so model() can derive a per-allele slope, a carrier indicator, or a genotype-specific coefficient depending on the source paper’s parameterisation.
• Units: (count, 0/1/2 alleles per subject)
• Type: continuous
• Scope: general
• Reference category: n/a (continuous). A count of 0 (CC homozygous wild type) is the natural reference and is the reference stratum in Liang_2024_lenalidomide.R.
• Source aliases:
  • COVR_ABA – Liang 2024 Results final-model equation (V/F_i (L) = 29.1 x Exp(0.554 + theta_ABA x COVR_ABA_i + theta_CHFY x COVR_CHFY_i), where “theta_ABA was the coefficient corresponding to each genotype of ABCB1 3435 C > T, COVR_ABA_i was the individual covariate value for the i-th individual”). Same orientation as the canonical, no value transformation.
  • ABCB1 3435C>T genotype string ("CC" / "CT" / "TT"): derive the count as the number of T alleles in the diplotype.
• Example models: Liang_2024_lenalidomide.R (log-additive effect on the apparent oral volume of distribution, with a GENOTYPE-SPECIFIC coefficient multiplied by the allele count: the log-scale contribution to V/F is 0.0151 * 1 for CT and 0.335 * 2 = 0.670 for TT, relative to CC. TT subjects therefore have a 95% larger V/F than CC subjects while CT subjects are essentially indistinguishable from CC at +1.5%; Liang 2024 Table 4 rows “ABCBA 3435 C > T = CT” and “= TT”).
• Notes: Distinct from the binary siblings SNP_ABCB1_RS1045642 (1 = any T-allele carrier, reference CC) and SNP_ABCB1_RS1045642_HOM (1 = TT only, reference CC + CT), and from the phased three-SNP haplotype ABCB1_HAP_TTT. Both binary siblings are recoverable from this column (> 0 and == 2 respectively), so a future extraction may use whichever matches its source paper’s parameterisation; the count is the more general encoding and should be preferred when the source resolves all three genotypes. Unlike the SNP_ABCB1_RS1045642_GA / _GG pair, this column is safely composable with the C/T-oriented siblings above, because Liang 2024 reports genotypes in the c.3435C>T cDNA orientation directly (CC / CT / TT) rather than in the dbSNP A/G strand orientation, so no wild-type / variant ambiguity arises. Follows the SNP_CYP2B6_RS3745274_T_COUNT and SNP_CYP4F2_RS2108622_T_COUNT precedents, and is an auto-approved member of the SNP_<GENE>_RS<rsid> family. Founding cohort (Liang 2024 Table 3, n = 51 Han Chinese patients): CC 19 (37.2%), CT 21 (41.2%), TT 11 (21.6%); T-allele frequency 41.1%, Hardy-Weinberg P = 0.27. Note that the founding paper’s effect is strongly non-additive in allele dose – the heterozygote coefficient is near zero and the homozygote coefficient is 22-fold larger – which is why the founding model carries two coefficients rather than a single per-allele slope; do not assume a linear per-allele effect when reusing this column.

SNP_ABCB1_RS4148738_HET (canonical for ABCB1 rs4148738 heterozygote (CT) indicator)

• Description: Binary genotype indicator for the heterozygous stratum of the ABCB1 rs4148738 single-nucleotide polymorphism (intronic; encodes P-glycoprotein / MDR1, the intestinal efflux transporter that limits dabigatran etexilate absorption). 1 = subject carries exactly one C and one T allele (genotype CT); 0 = otherwise (the union of CC and TT homozygotes). Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (non-CT, i.e. the pooled CC + TT homozygote group). Note this is a heterozygote-versus-both-homozygotes contrast, not the more usual heterozygote-versus-wild-type-homozygote contrast, because the founding paper carried separate CC / CT / TT dummy columns and only the CT dummy survived stepwise covariate modelling.
• Source aliases:
  • ABCB1CT – used in Yang_2024_dabigatran.R (Yang 2024 Supplementary Material $INPUT column list, which also carries the unused ABCB1C and ABCB1T dummies; the retained covariate block is IF(ABCB1CT.EQ.1) V2ABCB1CT = ( 1 + THETA(12))).
• Example models: Yang_2024_dabigatran.R (linear proportional deviation on the apparent central volume of distribution: vc = exp(lvc + etalvc) * (1 + e_abcb1_het_vc * SNP_ABCB1_RS4148738_HET) with e_abcb1_het_vc = 0.38; CT heterozygotes have 38% larger V2/F than the pooled CC + TT reference; Yang 2024 Table 4 row “rs4148738 on V2”, RSE 45.5%, bootstrap 95% CI 0.04-0.55).
• Notes: Distinct from SNP_ABCB1_RS1045642 (c.3435C>T in exon 26), SNP_ABCB1_RS3842 (c.4036A>G in the 3’ UTR), and ABCB1_HAP_TTT (the phased rs1128503 / rs2032582 / rs1045642 haplotype); use this entry only when the source paper genotypes rs4148738 specifically. Follows the SNP_ABCG2_RS2231142_HET and SNP_IL10_RS1800896_HET precedents for a heterozygote-stratum indicator, and is an auto-approved member of the SNP_<GENE>_RS<rsid> family. No paired SNP_ABCB1_RS4148738_HOM is registered yet: Yang 2024 tested a TT dummy in the same forward-selection step and it did not enter, so no homozygote effect size exists to encode. A future paper that resolves a distinct TT effect should add the _HOM sibling and, at that point, re-express the reference category as the CC wild-type homozygote alone. The founding cohort was 99 genotyped healthy Chinese adults (Yang 2024 Table 2): CC 20 (20%), CT 48 (48%), TT 31 (31%), minor (C) allele frequency 44.44%. The modelled CT-only effect is non-monotonic in allele dose, which the source paper does not explain; treat the estimate as an empirical stratum contrast in a heterozygote-rich cohort rather than a mechanistic gene-dose effect (Yang 2024 Results 3.4 similarly reports the CT stratum as the one with a distinctly low median trough concentration). rs4148738 has been associated with dabigatran peak concentrations in the RE-LY genome-wide analysis (Pare 2013, cited in Yang 2024 Introduction).

SNP_ABCB1_RS1045642_GA (canonical for ABCB1 rs1045642 heterozygous GA genotype indicator)

• Description: Binary genotype indicator for the ABCB1 rs1045642 single-nucleotide polymorphism (conventionally written c.3435C>T; exon 26; synonymous Ile1145Ile; encodes P-glycoprotein / MDR1). 1 = subject’s reported genotype is GA (heterozygous); 0 = otherwise. Paired with SNP_ABCB1_RS1045642_GG; both indicators are 0 for the AA reference group. Time-fixed per subject (germline genotype). Named by the reported genotype letters rather than by wild-type / variant status, because the founding source does not state which genotype is the wild type (see Notes).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 in combination with SNP_ABCB1_RS1045642_GG = 0, i.e. the AA genotype group (n = 12 of 67 in Yao 2025).
• Source aliases:
  • ABCB1 (rs1045642) GA – the Yao 2025 Table 2 row label for the CL coefficient of the GA group.
• Example models: Yao_2025_flurbiprofen_s.R (exponential effect on apparent plasma clearance of S(+)-flurbiprofen: cl = exp(lcl + etalcl) * exp(e_snp_abcb1_rs1045642_ga_cl * SNP_ABCB1_RS1045642_GA + e_snp_abcb1_rs1045642_gg_cl * SNP_ABCB1_RS1045642_GG) with e_snp_abcb1_rs1045642_ga_cl = -1.52, i.e. GA subjects have about 78% lower apparent CL than the AA reference group).
• Notes: Genotype-level (three-category) encoding, distinct from the pooled carrier indicator SNP_ABCB1_RS1045642 (1 = any T allele, reference = c.3435CC wild type) and from the phased haplotype ABCB1_HAP_TTT. The two encodings are NOT composable: SNP_ABCB1_RS1045642 takes the wild-type homozygote as its reference, whereas this pair takes the source’s AA group as reference, and the wild-type / variant orientation of the reported A / G alleles is unresolved. Yao 2025 reports genotypes as AA/GA/GG (12/40/15) while writing the SNP as “3435C > T” in its Methods, and never states which of AA / GG is the wild type. Two readings are defensible: (i) ABCB1 is transcribed from the minus strand, so cDNA C = genomic G and cDNA T = genomic A, making GG the c.3435CC wild type and AA the variant homozygote – the observed A-allele frequency of 0.478 is consistent with the reported 3435T frequency in Han Chinese cohorts; (ii) the source intended A as the wild type and simply ordered genotypes alphabetically. Reading (i) is the better supported for this locus, which would mean the source used the variant homozygote as its reference category. Because the model’s numerical predictions are identical under either reading, the founding extraction encodes the covariates by reported genotype letter so that no unverifiable wild-type claim is baked into the register; an extraction that needs wild-type-anchored semantics must resolve the orientation against the source’s genotyping assay before pooling with SNP_ABCB1_RS1045642. The _HET / _HOM suffix pair pre-authorised in the SNP_ABCB1_RS1045642 Notes was not used for exactly this reason: _HOM presupposes which homozygote is the variant, and here the reference category is itself a homozygote.

SNP_ABCB1_RS1045642_GG (canonical for ABCB1 rs1045642 homozygous GG genotype indicator)

• Description: Binary genotype indicator for the ABCB1 rs1045642 single-nucleotide polymorphism (conventionally written c.3435C>T). 1 = subject’s reported genotype is GG (homozygous for the G allele); 0 = otherwise. Paired with SNP_ABCB1_RS1045642_GA; both indicators are 0 for the AA reference group. Time-fixed per subject (germline genotype). Named by the reported genotype letters rather than by wild-type / variant status.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 in combination with SNP_ABCB1_RS1045642_GA = 0, i.e. the AA genotype group (n = 12 of 67 in Yao 2025).
• Source aliases:
  • ABCB1 (rs1045642) GG – the Yao 2025 Table 2 row label for the CL coefficient of the GG group.
• Example models: Yao_2025_flurbiprofen_s.R (exponential effect on apparent plasma clearance of S(+)-flurbiprofen with e_snp_abcb1_rs1045642_gg_cl = 0.19, i.e. GG subjects have about 21% higher apparent CL than the AA reference group).
• Notes: Always used together with SNP_ABCB1_RS1045642_GA; a model that references one without the other has mis-specified the reference category. See the SNP_ABCB1_RS1045642_GA Notes for the unresolved wild-type / variant orientation and for why this pair must not be pooled with the carrier indicator SNP_ABCB1_RS1045642. In Yao 2025 the genotype effects are non-monotonic (GA markedly reduces apparent CL, GG slightly increases it, relative to AA); the source reports this without mechanistic explanation beyond noting that P-glycoprotein polymorphisms modulate transport, and the AA reference group contained only 12 subjects.

SNP_POR_RS1057868_GA (**canonical for POR rs1057868 (POR*28) heterozygous GA genotype indicator**)

• Description: Binary genotype indicator for the POR rs1057868 single-nucleotide polymorphism (POR*28; conventionally written c.1508C>T, p.Ala503Val; encodes cytochrome P450 oxidoreductase, the flavoprotein that supplies electrons to microsomal CYP enzymes). 1 = subject’s reported genotype is GA (heterozygous); 0 = otherwise. Paired with SNP_POR_RS1057868_GG; both indicators are 0 for the AA reference group. Time-fixed per subject (germline genotype). Named by the reported genotype letters rather than by wild-type / variant status (see Notes).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 in combination with SNP_POR_RS1057868_GG = 0, i.e. the AA genotype group (n = 17 of 67 in Yao 2025).
• Source aliases:
  • POR (rs1057868) GA – the Yao 2025 Table 3 row label for the CL coefficient of the GA group.
• Example models: Yao_2025_flurbiprofen_r.R (exponential effect on apparent plasma clearance of R(-)-flurbiprofen: cl = exp(lcl + etalcl) * exp(e_snp_por_rs1057868_ga_cl * SNP_POR_RS1057868_GA + e_snp_por_rs1057868_gg_cl * SNP_POR_RS1057868_GG) with e_snp_por_rs1057868_ga_cl = -0.29, i.e. GA subjects have about 25% lower apparent CL than the AA reference group).
• Notes: First POR entry in the register. Genotype-level (three-category) encoding following the SNP_ABCB1_RS1045642_GA / _GG pattern. As with that pair, the wild-type / variant orientation of the reported A / G alleles is unresolved: Yao 2025 reports AA/GA/GG = 17/41/9 and writes the locus as “POR28 (rs1057868)” but never states which genotype is the wild type. Two readings conflict. (i) POR* is transcribed from the minus strand, so cDNA C = genomic G and cDNA T = genomic A, making GG the c.1508CC wild type and AA the 28/28 variant homozygote. (ii) The source’s own mechanism narrative points the other way – the Discussion attributes the reduced clearance of the GA and GG groups to”compromised electron transfer from POR to CYP450 enzymes”, i.e. it treats G as the loss-of-function (28) allele and therefore AA as the wild type. The allele-frequency check cannot break the tie because this locus deviates from Hardy-Weinberg equilibrium in the source cohort (the observed A-allele frequency of 0.560 far exceeds the roughly 0.36-0.44 POR28 frequency reported for East Asian cohorts, and the authors attribute the deviation to their small sample size). Because the model’s numerical predictions are identical under either reading, the founding extraction encodes the covariates by reported genotype letter so that no unverifiable wild-type claim is baked into the register. An extraction that needs wild-type-anchored POR semantics must resolve the orientation against the source’s genotyping assay first, and should register a separate wild-type-anchored canonical rather than redefining this one.

SNP_POR_RS1057868_GG (**canonical for POR rs1057868 (POR*28) homozygous GG genotype indicator**)

• Description: Binary genotype indicator for the POR rs1057868 single-nucleotide polymorphism (POR*28; conventionally written c.1508C>T, p.Ala503Val). 1 = subject’s reported genotype is GG (homozygous for the G allele); 0 = otherwise. Paired with SNP_POR_RS1057868_GA; both indicators are 0 for the AA reference group. Time-fixed per subject (germline genotype). Named by the reported genotype letters rather than by wild-type / variant status.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 in combination with SNP_POR_RS1057868_GA = 0, i.e. the AA genotype group (n = 17 of 67 in Yao 2025).
• Source aliases:
  • POR (rs1057868) GG – the Yao 2025 Table 3 row label for the CL coefficient of the GG group.
• Example models: Yao_2025_flurbiprofen_r.R (exponential effect on apparent plasma clearance of R(-)-flurbiprofen with e_snp_por_rs1057868_gg_cl = -2.01, i.e. GG subjects have about 87% lower apparent CL than the AA reference group).
• Notes: Always used together with SNP_POR_RS1057868_GA; a model that references one without the other has mis-specified the reference category. See the SNP_POR_RS1057868_GA Notes for the unresolved wild-type / variant orientation and the Hardy-Weinberg deviation at this locus. The Yao 2025 genotype effects are monotonic in G-allele count (GA about 25% lower, GG about 87% lower apparent CL than AA), which is the pattern expected if G is the reduced-function allele; only 9 of 67 subjects were GG, so the GG coefficient is imprecise (bootstrap 95% CI -5.2 to -1.2).

SLCO1B1_HAP15_HET (**canonical for SLCO1B1*15 haplotype heterozygote indicator**)

• Description: Binary haplotype indicator for the SLCO1B1 *15 reduced-function haplotype (the cis combination of the 388A>G / rs2306283 and 521T>C / rs4149056 variants; encodes the OATP1B1 N130D + V174A double mutant). 1 = subject carries exactly one 15 allele (heterozygous: *1a/*15 or *1b/*15), 0 = otherwise (the union of 15-noncarriers and *15-homozygotes; the paired indicator SLCO1B1_HAP15_HOM flags the homozygous group). Time-fixed per subject (germline haplotype).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (homozygous 1a/1a, 1a/1b, or 1b/1b – i.e., no 15 allele). The reference group is the union of the three non-15 diplotypes; SLCO1B1_HAP15_HOM flags the 15/15 group.
• Source aliases:
  • HT – Ide 2009 (paper text Eq. for Frel = 1 * theta1^HT * theta2^HM where HT = 1 for heterozygotes *1a/*15 and *1b/*15).
  • OATP1B1 phenotype – Jeong 2022 (paper Section 3.2 + Table 1 phenotype-to-haplotype mapping: ET = 1a/1a + 1a/1b + 1b/1b, IT = 1a/15 + 1b/15, PT = 15/15; IT subjects map to SLCO1B1_HAP15_HET = 1).
• Example models: Ide_2009_pravastatin.R (multiplicative effect on relative bioavailability Frel: Frel = 1.50^SLCO1B1_HAP15_HET * 1.95^SLCO1B1_HAP15_HOM – 15 heterozygotes have 50% higher Frel than 15-noncarriers; dOFV = 32.2 in backward elimination, p < 0.001), Jeong_2022_torsemide.R (linear-deviation effect on apparent central volume V/F: V/F = tvV/F * (1 + (-0.410) * SLCO1B1_HAP15_HET + (-0.646) * SLCO1B1_HAP15_HOM) – 15 heterozygotes have 41% lower V/F than 15-noncarriers; Jeong 2022 Table 4).
• Notes: Paired with SLCO1B1_HAP15_HOM to encode a three-level haplotype categorical (noncarrier / heterozygote / homozygote) with *15-noncarrier as the implicit reference (both indicators = 0). Distinct from the SNP-level canonical SNP_SLCO1B1_RS11045819 (which encodes only the C>A variant at a different position; rs11045819 = P155T) and from the 15 component SNPs rs2306283 (388A>G) and rs4149056 (521T>C) individually: future Ide-style extractions that pool 5 (521T>C only) with *15 should still record their values under this canonical and document the pooling rule in covariateData[[SLCO1B1_HAP15_HET]]$notes. Distribution in the Ide 2009 cohort of 57 healthy Japanese male volunteers (Table I): 28 noncarriers, 23 heterozygotes, 6 homozygotes; in the Jeong 2022 cohort of 112 healthy Korean male volunteers (Table 1): 86 ET noncarriers (76.8%), 23 IT heterozygotes (20.5%), 3 PT homozygotes (2.7%).

SLCO1B1_HAP15_HOM (**canonical for SLCO1B1*15 haplotype homozygote indicator**)

• Description: Binary haplotype indicator for the SLCO1B1 *15 reduced-function haplotype. 1 = subject carries two 15 alleles (homozygous: *15/*15), 0 = otherwise (the union of 15-noncarriers and *15-heterozygotes; the paired indicator SLCO1B1_HAP15_HET flags the heterozygous group). Time-fixed per subject (germline haplotype).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (homozygous 1a/1a, 1a/1b, or 1b/1b – i.e., no 15 allele). The reference group is the union of the three non-15 diplotypes; SLCO1B1_HAP15_HET flags the *15-heterozygote group.
• Source aliases:
  • HM – Ide 2009 (paper text Eq. for Frel = 1 * theta1^HT * theta2^HM where HM = 1 for homozygotes *15/*15).
  • OATP1B1 phenotype – Jeong 2022 (paper Section 3.2 + Table 1 phenotype-to-haplotype mapping: PT = 15/15 corresponds to SLCO1B1_HAP15_HOM = 1).
• Example models: Ide_2009_pravastatin.R (multiplicative effect on relative bioavailability Frel: Frel = 1.50^SLCO1B1_HAP15_HET * 1.95^SLCO1B1_HAP15_HOM – 15 homozygotes have 95% higher Frel than 15-noncarriers; dOFV = 33.7 in backward elimination, p < 0.001), Jeong_2022_torsemide.R (linear-deviation effect on apparent central volume V/F: V/F = tvV/F * (1 + (-0.410) * SLCO1B1_HAP15_HET + (-0.646) * SLCO1B1_HAP15_HOM) – 15 homozygotes have 64.6% lower V/F than 15-noncarriers; Jeong 2022 Table 4).
• Notes: Paired with SLCO1B1_HAP15_HET to encode a three-level haplotype categorical (noncarrier / heterozygote / homozygote) with *15-noncarrier as the implicit reference (both indicators = 0). See SLCO1B1_HAP15_HET Notes for the broader context; population distribution in Ide 2009 was 6 of 57 (10.5%) homozygotes and in Jeong 2022 was 3 of 112 (2.7%) homozygotes.

CYP2C9_S1_COUNT (**canonical for CYP2C9*1 (wild-type) allele count**)

• Description: Continuous individual-level CYP2C91 allele count: 0 = no 1 allele, 1 = one 1 allele (heterozygous), 2 = two 1 alleles (homozygous wild-type). Time-invariant (germline genotype). Paired with CYP2C9_S2_COUNT and CYP2C9_S3_COUNT to encode the three loss-of-function-allele dosage form used by Hamberg-family warfarin models, where the subject’s CL is the sum of per-allele CL contributions across the two CYP2C9 alleles. The three count columns sum to 2 for each subject.
• Units: (count, 0/1/2 alleles per subject)
• Type: continuous
• Scope: general
• Reference category: n/a (continuous). Used directly as a multiplier on a fixed per-allele CL contribution (0.174 L/h per *1 allele in Hamberg / Xia 2024).
• Source aliases:
  • CYP2C9 (genotype string such as "*1/*1", "*1/*3"): derive CYP2C9_S1_COUNT = (length of *1 matches in the genotype string). The model Xia_2024_warfarin.R carries the source CYP2C9 genotype string mapped to the three count columns.
• Example models: Xia_2024_warfarin.R (per-allele CL contributions: cl = CYP2C9_S1_COUNT * 0.174 + CYP2C9_S2_COUNT * 0.0879 + CYP2C9_S3_COUNT * 0.0422, times an age effect), Huth_2019_siponimod.R (the three counts build six diplotype indicators selecting a relative CYP2C9 clearance activity running from 1 in 1/1 to 0.021 in 3/3).
• Notes: Hamberg’s warfarin K-PD model parameterises CL as a sum of two per-allele CL contributions (one per CYP2C9 allele on each chromosome), which is more flexible than a single genotype indicator because it naturally accommodates any combination of 1, 2, 3 alleles (six diplotypes: 1/1, 1/2, 1/3, 2/2, 2/3, 3/3). When a paper reports additional CYP2C9 alleles (e.g. 5, 6, 8, *11), register parallel canonicals (CYP2C9_S5_COUNT etc.) rather than overloading the existing three counts. Distinct from the categorical phenotype canonicals CYP3A5_EXPR (binary expresser) and from continuous-activity scores like CYP3A4 – the count form preserves loss-of-function-allele dosage exactly.

CYP2C9_S2_COUNT (**canonical for CYP2C9*2 reduced-function allele count**)

• Description: Continuous individual-level CYP2C92 allele count: 0 = no 2 allele, 1 = one 2 allele (heterozygous), 2 = two 2 alleles (homozygous). Time-invariant (germline genotype). Paired with CYP2C9_S1_COUNT and CYP2C9_S3_COUNT; the three counts sum to 2 for each subject. The *2 allele (rs1799853, R144C) encodes a reduced-function CYP2C9 isoform.
• Units: (count, 0/1/2 alleles per subject)
• Type: continuous
• Scope: general
• Reference category: n/a (continuous). Used directly as a multiplier on a fixed per-allele CL contribution (0.0879 L/h per *2 allele in Hamberg / Xia 2024).
• Source aliases:
  • CYP2C9 (genotype string such as "*1/*2", "*2/*2", "*2/*3"): derive CYP2C9_S2_COUNT = (length of *2 matches in the genotype string).
• Example models: Xia_2024_warfarin.R (per-allele CL contributions; the Xia 2024 Han cohort had no 2 carriers, but the model retains the term for general use across CYP2C9 papers), Huth_2019_siponimod.R (1/2 relative CYP2C9 activity 0.908, 2/2 0.648, 2/*3 0.391).
• Notes: See CYP2C9_S1_COUNT for the broader rationale.

CYP2C9_S3_COUNT (**canonical for CYP2C9*3 reduced-function allele count**)

• Description: Continuous individual-level CYP2C93 allele count: 0 = no 3 allele, 1 = one 3 allele (heterozygous), 2 = two 3 alleles (homozygous). Time-invariant (germline genotype). Paired with CYP2C9_S1_COUNT and CYP2C9_S2_COUNT; the three counts sum to 2 for each subject. The *3 allele (rs1057910, I359L) encodes a strongly reduced-function CYP2C9 isoform and is the dominant CYP2C9 pharmacogenomic risk variant in East-Asian populations (warfarin / phenytoin sensitivity).
• Units: (count, 0/1/2 alleles per subject)
• Type: continuous
• Scope: general
• Reference category: n/a (continuous). Used directly as a multiplier on a fixed per-allele CL contribution (0.0422 L/h per *3 allele in Hamberg / Xia 2024).
• Source aliases:
  • CYP2C9 (genotype string such as "*1/*3", "*3/*3"): derive CYP2C9_S3_COUNT = (length of *3 matches in the genotype string).
• Example models: Xia_2024_warfarin.R (per-allele CL contributions; 5.7% of the Han cohort were 1/3 heterozygous per Xia 2024 Table 1), Ohara_2014_warfarin_s.R (dichotomised to a carrier flag via CYP2C9_S3_COUNT > 0; CL(S) in carriers is 0.543x the 1/1 reference), Huth_2019_siponimod.R (1/3 relative CYP2C9 activity 0.515, 2/3 0.391, 3/3 0.021; siponimod is contraindicated in 3/3).
• Notes: See CYP2C9_S1_COUNT for the broader rationale.

CYP2C9_S5_COUNT (**canonical for CYP2C9*5 reduced-function allele count**)

• Description: Continuous individual-level CYP2C95 allele count: 0 = no 5 allele, 1 = one 5 allele (heterozygous), 2 = two 5 alleles (homozygous). Time-invariant (germline genotype). The *5 allele (rs28371686, D360E) encodes a markedly reduced-function CYP2C9 isoform found almost exclusively in individuals of African ancestry. Registered per the extension rule stated in CYP2C9_S1_COUNT Notes (“When a paper reports additional CYP2C9 alleles (e.g. 5, 6, 8, 11), register parallel canonicals”); with CYP2C9_S6_COUNT and CYP2C9_S8_COUNT it extends the CYP2C9_S{1,2,3}_COUNT family so the per-subject counts still sum to 2.
• Units: (count, 0/1/2 alleles per subject)
• Type: continuous
• Scope: general
• Reference category: n/a (continuous). In RodriguezFernandez_2024_warfarin.R this column is one of four summed into the pooled reduced-function allele class “n” (3, 5, 6, 8) that Reyes-Gonzalez 2020 uses to select the CYP2C9-diplotype elimination rate constant.
• Source aliases:
  • CYP2C9 (genotype string such as "*1/*5", "*2/*5"): derive CYP2C9_S5_COUNT = (number of *5 matches in the genotype string).
• Example models: RodriguezFernandez_2024_warfarin.R (Caribbean Hispanic cohort with 1/5 n=1 and 2/5 n=1; the 5 allele contributes to the pooled “n” class whose diplotypes carry kel 0.0132 (1/n), 0.009 (2/n) and 0.0075 (n/*n) 1/h).
• Notes: See CYP2C9_S1_COUNT for the broader rationale. A model that pools 3/5/6/8 must still declare the individual count columns rather than a single pooled column, so that a downstream user retains the resolved genotype and a future paper reporting allele-specific effects can use the same dataset.

CYP2C9_S6_COUNT (**canonical for CYP2C9*6 reduced-function allele count**)

• Description: Continuous individual-level CYP2C96 allele count: 0 = no 6 allele, 1 = one 6 allele, 2 = two 6 alleles. Time-invariant (germline genotype). The *6 allele (rs9332131, a single-base deletion producing a frameshift) is a null allele found predominantly in individuals of African ancestry. Registered per the extension rule in CYP2C9_S1_COUNT Notes.
• Units: (count, 0/1/2 alleles per subject)
• Type: continuous
• Scope: general
• Reference category: n/a (continuous). One of the four columns summed into the pooled “n” reduced-function class in RodriguezFernandez_2024_warfarin.R.
• Source aliases:
  • CYP2C9 (genotype string such as "*1/*6"): derive CYP2C9_S6_COUNT = (number of *6 matches in the genotype string).
• Example models: RodriguezFernandez_2024_warfarin.R (no 6 carriers in the Caribbean Hispanic cohort, but Reyes-Gonzalez 2020 explicitly names 6 as a member of the pooled “n” class, so the model is defined for *6 carriers and the column is referenced in model()).
• Notes: See CYP2C9_S1_COUNT for the broader rationale. Registering a count column that is identically zero in the source cohort is correct when the source model’s own allele-class definition names the allele – it makes the model usable on a cohort that does carry it, and keeps the “counts sum to 2” invariant checkable.

CYP2C9_S8_COUNT (**canonical for CYP2C9*8 reduced-function allele count**)

• Description: Continuous individual-level CYP2C98 allele count: 0 = no 8 allele, 1 = one 8 allele, 2 = two 8 alleles. Time-invariant (germline genotype). The 8 allele (rs7900194, R150H) encodes a reduced-function CYP2C9 isoform enriched in African-ancestry populations and is a major reason CYP2C9 genotyping panels restricted to 2 and *3 misclassify warfarin dose requirements in Black and admixed patients. Registered per the extension rule in CYP2C9_S1_COUNT Notes.
• Units: (count, 0/1/2 alleles per subject)
• Type: continuous
• Scope: general
• Reference category: n/a (continuous). One of the four columns summed into the pooled “n” reduced-function class in RodriguezFernandez_2024_warfarin.R.
• Source aliases:
  • CYP2C9 (genotype string such as "*1/*8"): derive CYP2C9_S8_COUNT = (number of *8 matches in the genotype string).
• Example models: RodriguezFernandez_2024_warfarin.R (1/8 n=2 in the Caribbean Hispanic cohort, Table 1).
• Notes: See CYP2C9_S1_COUNT for the broader rationale. The presence of 5 and 8 carriers is one of the features distinguishing admixed Caribbean Hispanic cohorts from the European cohorts underlying the Hamberg warfarin models, whose CYP2C9 covariate is limited to 1/2/*3.

CYP2C9_MISSING (canonical for CYP2C9 genotype-missing indicator)

• Description: Binary indicator for a subject whose CYP2C9 genotype was not measured / not available. 1 = CYP2C9 genotype missing; 0 = CYP2C9 genotype known (in which case CYP2C9_S1_COUNT + CYP2C9_S2_COUNT + CYP2C9_S3_COUNT = 2). Companion to the CYP2C9_S{1,2,3}_COUNT per-allele canonicals so popPK models that estimate a separate typical-value covariate effect for the missing-genotype subgroup (rather than imputing it as wild-type) can do so faithfully.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (CYP2C9 genotype known). When CYP2C9_MISSING == 1 the three CYP2C9_S{1,2,3}_COUNT columns must all be 0 so the diplotype-indicator products correctly evaluate to zero for the missing-genotype subject.
• Source aliases:
  • Missing (Lane 2011 Table 3 q_CYP2C9 Missing row – the paper treats unknown CYP2C9 genotype as a distinct categorical level with its own multiplicative CL effect, rather than imputing the missing subjects into the wild-type reference; 14 of 306 S-warfarin subjects were missing CYP2C9 genotype per Lane 2011 Table 1).
• Example models: Lane_2011_warfarin_s.R (multiplicative CL effect 0.782 for the missing-genotype subgroup vs the 1/1 wild-type reference; Lane 2011 Table 3).
• Notes: Follows the missing-data-indicator pattern established by ADA_MISSING (multiplicative CL effect for subjects with missing ADA results; Suri 2018 brentuximab) and HEPIMP_MOD_OR_MISSING (composite moderate-or-missing hepatic impairment indicator). Distinct from the strategy of imputing missing genotype as wild-type (which would set CYP2C9_S1_COUNT = 2 and not use a separate indicator); use CYP2C9_MISSING only when the source paper explicitly fits a separate coefficient for the missing-genotype subgroup.

VKORC1_1639G_COUNT (canonical for VKORC1 -1639G allele count)

• Description: Continuous individual-level count of VKORC1 -1639G alleles (rs9923231, also reported as VKORC1 -1639 G > A or 1173 C > T depending on numbering convention). 0 = AA homozygous (warfarin-sensitive), 1 = GA heterozygous, 2 = GG homozygous (warfarin-resistant). Time-invariant (germline genotype). The complementary -1639A count is 2 - VKORC1_1639G_COUNT.
• Units: (count, 0/1/2 alleles per subject)
• Type: continuous
• Scope: general
• Reference category: n/a (continuous). In the Hamberg / Xia 2024 EC50 model the typical EC50 is a per-allele sum: ec50_typ = VKORC1_1639G_COUNT * ec50_per_G + (2 - VKORC1_1639G_COUNT) * ec50_per_A. Distribution in the Xia 2024 Han cohort (Table 1): AA 80.3%, GA 18.7%, GG 0.9% (G allele frequency ~10%, consistent with East-Asian populations).
• Source aliases:
  • VKORC1 (genotype string such as "AA", "GA", "GG" or "1639AA" etc.): derive VKORC1_1639G_COUNT = (count of G in the two-letter genotype).
• Example models: Xia_2024_warfarin.R (per-allele EC50 contributions, re-estimated for the Han Chinese cohort: 4.3 mg/L per G allele, 1.14 mg/L per A allele), Ohara_2014_warfarin_s.R (dichotomised to a G-carrier flag via VKORC1_1639G_COUNT > 0, with -1639A/A as the reference and a 2.07-fold IC50 increase in G carriers).
• Notes: VKORC1 -1639G > A is the strongest single-SNP determinant of warfarin sensitivity (the A allele reduces VKORC1 expression via a promoter-region effect, requiring less warfarin to achieve target anticoagulation). The per-allele count form is preferred over a binary carrier indicator because the heterozygous and homozygous mutant subjects respond detectably differently to warfarin. The count column also reconstructs a carrier-flag parameterisation deterministically (count > 0), which is what papers reporting a single “VKORC12” indicator require – note that such papers usually take the A/A genotype as the reference and report an IC50/EC50 increase* per G carriage, the mirror image of the Hamberg / Xia G/G-reference orientation.

SNP_SCN1A_RS3812718_GA (canonical for SCN1A rs3812718 heterozygous G/A genotype indicator)

• Description: Binary genotype indicator for the SCN1A single-nucleotide polymorphism rs3812718 (conventionally written IVS5-91 G>A, also c.603-91G>A; intron 5 splice-donor region; SCN1A encodes the alpha-1 subunit Nav1.1 of the neuronal voltage-gated sodium channel). 1 = subject’s genotype is G/A (heterozygous), 0 = otherwise. Paired with [[SNP_SCN1A_RS3812718_AA]]; both indicators are 0 for the G/G reference group. Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 in combination with SNP_SCN1A_RS3812718_AA = 0, i.e. the G/G genotype group (9 of 77 patients, 11.7%, in Nakashima 2015).
• Source aliases:
  • SCN1A G/A genotype – the Nakashima 2015 Eq 6 / Eq 7 / Eq 8 term label and Table 3 row label.
• Example models: Nakashima_2015_valproic_acid.R (acts on BOTH the intercept and the exposure slope of a logistic exposure-response model for an over-50% reduction in seizure frequency: e_snp_scn1a_rs3812718_ga_logit = -5.87 and e_snp_scn1a_rs3812718_ga_slope = +10.1 per 100 ug/mL of predicted trough valproate; Nakashima 2015 Table 3), Nakashima_2015_valproic_acid_le18y.R (-4.75 and +7.62; Eq 7), Nakashima_2015_valproic_acid_ge19y.R (-9.88 on the intercept only; Eq 8).
• Notes: First SCN1A entry in the register. Genotype-level (three-category) encoding following the SNP_ABCB1_RS1045642_GA / _GG and SNP_POR_RS1057868_GA / _GG precedents: one indicator per non-reference genotype, named by the reported genotype letters. Unlike the ABCB1 rs1045642 case, the allele orientation here is not ambiguous – the founding source writes the polymorphism as “IVS5-91 G > A”, which names G as the reference allele and A as the variant, so the G/G reference group is the wild-type homozygote and the letter-based names coincide with wild-type-anchored semantics. rs3812718 lies in the splice-donor site of SCN1A exon 5 and modulates alternative splicing between the neonatal exon 5N and adult exon 5A isoforms; the A allele reduces exon-5N expression and alters Nav1.1 electrophysiology in vitro, which is the founding authors’ mechanistic reading of the reduced valproate efficacy they observe in A-allele carriers (Nakashima 2015 Discussion). Because the effect is estimated on both the intercept and the slope, the two coefficients partly offset and the net effect of genotype depends on the trough concentration – do not summarise this covariate by its intercept coefficient alone. Founding-cohort genotype frequencies: G/G 11.7%, G/A 53.2%, A/A 35.1%; A-allele frequency 61.7%; consistent with Hardy-Weinberg equilibrium (Nakashima 2015 Results and Table 1). Promote to general scope once a second model ratifies the name.

SNP_SCN1A_RS3812718_AA (canonical for SCN1A rs3812718 homozygous A/A genotype indicator)

• Description: Binary genotype indicator for the SCN1A single-nucleotide polymorphism rs3812718 (IVS5-91 G>A; intron 5 splice-donor region; SCN1A encodes the alpha-1 subunit Nav1.1 of the neuronal voltage-gated sodium channel). 1 = subject’s genotype is A/A (homozygous for the variant A allele), 0 = otherwise. Paired with [[SNP_SCN1A_RS3812718_GA]]; both indicators are 0 for the G/G reference group. Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 in combination with SNP_SCN1A_RS3812718_GA = 0, i.e. the G/G genotype group (9 of 77 patients, 11.7%, in Nakashima 2015).
• Source aliases:
  • SCN1A A/A genotype – the Nakashima 2015 Eq 6 / Eq 7 / Eq 8 term label and Table 3 row label.
• Example models: Nakashima_2015_valproic_acid.R (acts on BOTH the intercept and the exposure slope of a logistic exposure-response model for an over-50% reduction in seizure frequency: e_snp_scn1a_rs3812718_aa_logit = -4.88 and e_snp_scn1a_rs3812718_aa_slope = +9.48 per 100 ug/mL of predicted trough valproate; Nakashima 2015 Table 3), Nakashima_2015_valproic_acid_le18y.R (-4.30 and +7.60; Eq 7), Nakashima_2015_valproic_acid_ge19y.R (the only concentration-dependent term in that model, -14.3 on the slope; Eq 8).
• Notes: Paired with [[SNP_SCN1A_RS3812718_GA]]; see that entry for the locus biology, the unambiguous G-reference orientation, and the caution that the intercept and slope coefficients partly offset so genotype cannot be summarised by its intercept effect alone. The A allele is the variant and is the more common allele in the founding Japanese cohort (frequency 61.7%), so the G/G reference group is small (9 patients) and both coefficients carry wide bootstrap confidence intervals that span zero on the intercept (-13.61 to 1.09 for G/A, -12.70 to 1.96 for A/A) while remaining bounded away from zero on the slope (1.31 to 21.31 and 2.13 to 20.70) (Nakashima 2015 Table 3). The founding authors previously associated the A/A genotype with carbamazepine-resistant epilepsy in the same patient population. Promote to general scope once a second model ratifies the name.

SNP_CYP4F2_RS2108622_T_COUNT (**canonical for CYP4F2 1297C>T (rs2108622, CYP4F2*3) T-allele count**)

• Description: Continuous individual-level count of CYP4F2 c.1297C>T (rs2108622, p.V433M) T alleles, i.e. the CYP4F2*3 allele: 0 = CC homozygous wild-type (1/1), 1 = CT heterozygous (1/3), 2 = TT homozygous variant (3/3). Time-invariant (germline genotype). CYP4F2 is the hepatic vitamin-K1 oxidase; the 433M variant protein has reduced activity, so *3 carriers retain more hepatic vitamin K and require a higher warfarin concentration for the same inhibition of vitamin-K-dependent clotting-factor synthesis. It is the third pharmacogene (after VKORC1_1639G_COUNT and the CYP2C9_S{1,2,3}_COUNT family) included in the CPIC and IWPC warfarin dosing guidance.
• Units: (count, 0/1/2 alleles per subject)
• Type: continuous
• Scope: general
• Reference category: n/a (continuous). In Ohara_2014_warfarin_s.R the count is dichotomised in model to a carrier indicator (SNP_CYP4F2_RS2108622_T_COUNT > 0) because the source paper reports a single carriage effect rather than a per-allele dosage: IC50 is 1.30x higher in 1/3 and 3/3 subjects than in the 1/1 reference (Ohara 2014 Table 2).
• Source aliases:
  • CYP4F2*3 – Ohara 2014 / Shi 2024 (reported as a 0/1 carriage indicator; Shi 2024 Table 3 footnote c defines CYP4F2*3 = 0 for 1/1 and 1 otherwise, so the count column reconstructs it as SNP_CYP4F2_RS2108622_T_COUNT > 0).
  • CYP4F2 (genotype string such as "CC" / "CT" / "TT", or star-allele form "*1/*1" / "*1/*3" / "*3/*3"): derive the count as the number of T (equivalently *3) alleles in the diplotype.
  • V433M / rs2108622 / 1297C>T – alternative labels for the same variant.
• Example models: Ohara_2014_warfarin_s.R (multiplicative 1.30x effect on the S-warfarin IC50 for inhibition of normal-prothrombin synthesis, applied to carriers; Ohara 2014 Table 2, 95% CI 1.07-1.53).
• Notes: Distribution in the Ohara 2014 Taiwanese Chinese cohort (Table 1, n = 99): 50 CC, 43 CT, 6 TT (MAF 0.278), consistent with reported East-Asian frequencies. Shi 2024 assumed all simulated Chinese patients to be CYP4F21/1, so the column is 0 throughout their clinical-trial simulations. The count form (rather than a bare SNP_CYP4F2_RS2108622 indicator) follows the SNP_CYP2B6_RS3745274_T_COUNT and CYP2C9_S{1,2,3}_COUNT precedent: a single count column carries the full genotype without redundancy and lets model() derive either a per-allele effect (* count) or a carrier/HET/HOM decomposition depending on the source paper’s parameterisation.

SNP_CYP1A2_RS762551_C_CARRIER (canonical for CYP1A2 rs762551 (C163A) C-allele carrier indicator)

• Description: Binary indicator for carriage of the CYP1A2 rs762551 promoter-region single-nucleotide polymorphism (also known as CYP1A2 c.-163C>A; the paper’s “C163A” shorthand). 1 = subject carries at least one C allele (heterozygous AC or homozygous CC); 0 = homozygous AA (the CYP1A2*1F reference / ultra-inducible form, associated with enhanced CYP1A2 substrate metabolism in smokers). Time-fixed per subject (germline genotype). AC heterozygous and CC homozygous carriers are pooled in Hopkins 2015 because the CC homozygous frequency (9/105 = 8.6%) was too low for a separate phenotype layer; future extractions distinguishing heterozygous vs homozygous C-allele effects should register paired SNP_CYP1A2_RS762551_C_HET / SNP_CYP1A2_RS762551_C_HOM indicators following the SLCO1B1_HAP15_HET / SLCO1B1_HAP15_HOM precedent.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (homozygous AA, i.e., CYP1A21F/1F).
• Source aliases:
  • CYP1A2 C163A – Hopkins 2015 (paper Table 3 caption defines the covariate: “carriers of the C allele of CYP1A2 rs762551 (i.e. CC or CA genotype)”; same orientation as the canonical).
• Example models: Hopkins_2015_leflunomide.R (multiplicative effect on the cessation-hazard covariate model in a leflunomide time-to-event analysis: hazard <- h0(t) * (1 + e_cyp1a2_haz * SNP_CYP1A2_RS762551_C_CARRIER) with e_cyp1a2_haz = 1.29; C-allele carriers have a 2.29-fold (95% CI 2.24, 2.34) increase in instantaneous cessation hazard compared with AA homozygotes; Hopkins 2015 Table 3 covariate-model theta10 and Results paragraph on CYP1A2).
• Notes: rs762551 lies in the CYP1A2 promoter region. The A allele defines the CYP1A2*1F ultra-inducible haplotype associated with enhanced CYP1A2 substrate metabolism in cigarette smokers (Sachse et al. 1999, cited in Hopkins 2015 Discussion). Downstream mechanistic hypotheses for the observed leflunomide-toxicity association include (a) altered relative contributions of CYP1A2 vs CYP2C19 / CYP3A4 in converting leflunomide to its active metabolite teriflunomide, potentially producing higher concentrations of alternate toxic metabolites in C-allele carriers, and (b) direct leflunomide toxicity that is unmasked when the CYP1A2 clearance arm is slower (Hopkins 2015 Discussion). No association between smoking status and leflunomide cessation was detected in the Hopkins 2015 cohort, but the smoking rate (37.1%) was modest. Distinct from S1A2 (a specific-scope categorical CYP1A2-modifying co-medication indicator carried per subject in the DDMORE lidocaine bundle; not the same concept). The naming pattern SNP_<GENE>_<RSID>_<ALLELE>_CARRIER follows the existing SNP_ABCB1_RS1045642 / SNP_ABCB1_RS3842 precedent while making the binarised allele explicit.

SNP_CYP1A2_RS762551 (canonical for CYP1A2 rs762551 variant-allele carrier indicator, variant allele unspecified by the source)

• Description: Binary indicator for carriage of at least one VARIANT allele at the CYP1A2 rs762551 promoter single-nucleotide polymorphism (c.-163C>A, CYP1A2*1F), as the source paper labels its genotypes wt and v. 1 = wt/v or v/v; 0 = wt/wt. Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (wt/wt).
• Source aliases:
  • CYP1A2_v – Sima 2022 (Table 4 footnote: ‘CYP1A2_v is at least one variant allele in the CYP1A2 genotype’; Table 2 genotype rows wt/wt, wt/v, v/v).
• Example models: Sima_2022_ciprofloxacin.R (exponential effect on the desethylene ciprofloxacin elimination rate constant: Km = Km_pop * exp(-0.035 * AGE + 0.6 * SNP_CYP1A2_RS762551), i.e. carriers eliminate the metabolite 1.82-fold faster).
• Notes: Registered as the bare SNP_<GENE>_RS<rsid> variant-carrier form, without an allele letter, because Sima 2022 does not say which nucleotide it calls the variant allele, and its genotype frequencies (wt/wt 20.7%, wt/v 65.5%, v/v 13.8% in 29 Czech patients) do not settle it. Not interchangeable with SNP_CYP1A2_RS762551_C_CARRIER: a C-carrier indicator groups AC with CC, an A-carrier indicator groups AC with AA, and the two disagree for every homozygote. Pharmacogenetic nomenclature usually treats C as the reference (1A) and A as the variant (1F), but papers are not consistent, which is why the letter cannot be assumed; a source that names the allele letter should register a lettered sibling (SNP_CYP1A2_RS762551_A_CARRIER) rather than reuse this column.

SNP_CYP2B6_RS3745274_T_COUNT (canonical for CYP2B6 516G>T (rs3745274) T-allele count)

• Description: Continuous individual-level CYP2B6 c.516G>T (rs3745274, p.Q172H) T-allele count: 0 = GG homozygous wild-type, 1 = GT heterozygous, 2 = TT homozygous variant. Time-invariant (germline genotype). The 516G>T variant reduces CYP2B6 enzyme activity in a gene-dose-dependent but non-additive manner; in nevirapine and efavirenz populations, the TT-vs-GG decrease in apparent oral clearance is approximately 2-3x larger than the GT-vs-GG decrease, motivating a heterozygous-vs-homozygous indicator parameterization rather than a linear per-allele model.
• Units: (count, 0/1/2 alleles per subject)
• Type: continuous
• Scope: general
• Reference category: n/a (continuous). In the Schipani 2011 nevirapine model the count is decomposed in model into mutually-exclusive heterozygous (count == 1) and homozygous (count == 2) indicators, each multiplied by an independently estimated CL/F shift (-0.5 L/h for GT, -1.3 L/h for TT relative to the GG reference). In the Olagunju 2018 efavirenz model the count is summed together with SNP_CYP2B6_RS28399499_C_COUNT to derive a composite metaboliser status (n_variant == 0 fast, == 1 intermediate, >= 2 slow), and the per-group CL/F is encoded as log-ratio multiplicative shifts on the fast-metaboliser reference.
• Source aliases:
  • X_516GT / X_516TT – Schipani 2011 (paper Table 2 / final-model equation TVCL = theta0 + theta_BW * (BW - 72.5) + theta_516GT * X_516GT + theta_516TT * X_516TT + theta_983TC * X_983TC; the two indicators are mutually exclusive across the three 516 genotypes, so the canonical count column reconstructs them deterministically via X_516GT = as.integer(SNP_CYP2B6_RS3745274_T_COUNT == 1) and X_516TT = as.integer(SNP_CYP2B6_RS3745274_T_COUNT == 2)).
  • CYP2B6 516G>T (rs3745274) – Olagunju 2018 (paper Methods ‘Sample Collection…’ paragraph 1; genotype reported per-allele and used in combination with rs28399499 to define a composite CYP2B6 metaboliser status).
  • RSA – Vucicevic 2025 (paper Table 2 footnote: ‘binary parameter indicating whether the patient is a carrier of CYP2B6 516G>T (1) or not (0)’; a dominant-model carrier indicator that the canonical count column reconstructs via SNP_CYP2B6_RS3745274_T_COUNT >= 1. Results paragraph 2 reports the carriers as GT heterozygotes (n = 30) versus GG wild-type (n = 60) with no TT homozygotes observed, so within that cohort the carrier indicator and a heterozygote indicator coincide).
  • CYP2B6 516 G>T genotype – Siccardi 2012 (paper Methods ‘Population PK model at standard regimens’, final paragraph: CL = CL0 + theta1 * GT + theta2 * TT, where the paper describes GT and TT as ‘genotypic status being equal to 1 and 2’ but uses them as two mutually exclusive indicators over the three genotypes with 516GG as the reference; the canonical count column reconstructs them via (SNP_CYP2B6_RS3745274_T_COUNT == 1) and (SNP_CYP2B6_RS3745274_T_COUNT == 2)).
• Example models: Schipani_2011_nevirapine.R (additive linear shift on CL/F: cl = exp(lcl) + e_516gt_cl * (SNP_CYP2B6_RS3745274_T_COUNT == 1) + e_516tt_cl * (SNP_CYP2B6_RS3745274_T_COUNT == 2) + ...; 516TT homozygotes have approximately 37% lower CL/F than the GG reference; Schipani 2011 Table 2), Olagunju_2018_efavirenz.R (composite-metaboliser-status encoding: variant alleles from rs3745274 and rs28399499 are summed to classify subjects as fast / intermediate / slow, with per-group CL/F = 18.0 / 16.1 / 6.24 L/h reported in Olagunju 2018 Table 2), Vucicevic_2025_efavirenz.R (dominant-model carrier indicator raised as a power of a fractional-change factor on apparent oral clearance: (1 + e_cyp2b6_516t_cl)^(SNP_CYP2B6_RS3745274_T_COUNT >= 1) with e_cyp2b6_516t_cl = -0.364; carriers have CL/F 36.4% lower than GG wild-type; Vucicevic 2025 Table 2 theta_RSA), Siccardi_2012_efavirenz.R (additive linear shift on CL/F following the same form as Schipani 2011: tvcl = exp(lcl) + e_516gt_cl * (SNP_CYP2B6_RS3745274_T_COUNT == 1) + e_516tt_cl * (SNP_CYP2B6_RS3745274_T_COUNT == 2); the paper prints the three resulting typical values 13.3 / 9.7 / 2.6 L/h rather than the two shifts, so e_516gt_cl = -3.6 and e_516tt_cl = -10.7 are exact differences of printed values; Siccardi 2012 Table 2).
• Notes: Distribution in the Schipani 2011 European HIV-positive cohort (Table 1, n = 275): 516GG 47%, 516GT 46%, 516TT 7%. In the Olagunju 2018 Nigerian HIV-positive pregnant-women cohort (Table 1, n = 77): 516GG 32%, 516GT 54%, 516TT 14%. The CYP2B6 516G>T variant is one of the two most-extensively-studied CYP2B6 pharmacogenomic polymorphisms (together with 983T>C / rs28399499, registered as SNP_CYP2B6_RS28399499_C_COUNT) and is consistently associated with reduced metabolism of nevirapine, efavirenz, bupropion, and methadone. The count-form encoding is preferred over paired binary HET / HOM indicators because (a) it follows the established _COUNT precedent used for CYP2C9_S{1,2,3}_COUNT and VKORC1_1639G_COUNT, (b) a single count column captures the underlying genotype without redundancy, and (c) the model code can deterministically derive either a linear per-allele effect (* count) or a non-additive HET / HOM decomposition (* (count == 1), * (count == 2)) depending on the source paper’s parameterization.

SNP_CYP2B6_RS28399499_C_COUNT (canonical for CYP2B6 983T>C (rs28399499) C-allele count)

• Description: Continuous individual-level CYP2B6 c.983T>C (rs28399499, p.I328T) C-allele count: 0 = TT homozygous wild-type, 1 = TC heterozygous, 2 = CC homozygous variant. Time-invariant (germline genotype). The 983T>C variant lies within the CYP2B6 J-helix and changes a highly conserved hydrophobic Ile to polar Thr; recombinant expression in COS-1 cells shows essentially zero enzyme activity (Klein et al. 2005), and clinical cohorts consistently show large reductions in apparent oral clearance of CYP2B6 substrates in heterozygous carriers.
• Units: (count, 0/1/2 alleles per subject)
• Type: continuous
• Scope: general
• Reference category: n/a (continuous). In the Schipani 2011 nevirapine model the count is decomposed in model into a heterozygous indicator (count == 1) multiplied by an estimated CL/F shift (-1.4 L/h relative to the TT reference); the homozygous indicator (count == 2) is not estimated because no 983CC homozygotes have been reported in any published cohort. In the Olagunju 2018 efavirenz model the count is summed together with SNP_CYP2B6_RS3745274_T_COUNT to define a composite CYP2B6 metaboliser status (slow / intermediate / fast); a hypothetical 983CC subject is still classified as slow (each variant allele contributes to the composite count) although that substratum is not present in the fitted cohort.
• Source aliases:
  • X_983TC – Schipani 2011 (paper Table 2 / final-model equation; the heterozygous indicator is mutually exclusive with the TT-homozygous reference, so the canonical count column reconstructs it via X_983TC = as.integer(SNP_CYP2B6_RS28399499_C_COUNT == 1)).
  • CYP2B6 983T>C (rs28399499) – Olagunju 2018 (paper Methods ‘Sample Collection …’ paragraph 1; combined with rs3745274 to define a composite metaboliser status).
• Example models: Schipani_2011_nevirapine.R (additive linear shift on CL/F: cl = exp(lcl) + e_983tc_cl * (SNP_CYP2B6_RS28399499_C_COUNT == 1) + ...; 983TC heterozygotes have approximately 40% lower CL/F than the TT reference; Schipani 2011 Table 2), Olagunju_2018_efavirenz.R (composite-metaboliser-status encoding: variant alleles from rs3745274 and rs28399499 are summed to classify subjects as fast / intermediate / slow; Olagunju 2018 Table 2).
• Notes: Distribution in the Schipani 2011 European HIV-positive cohort (Table 1, n = 275): 983TT 97%, 983TC 3%, 983CC 0%. Distribution in the Olagunju 2018 Nigerian HIV-positive pregnant-women cohort (Table 1, n = 77): 983TT 75%, 983TC 25%, 983CC 0% – a much higher heterozygous-variant fraction than the Schipani 2011 cohort, consistent with the West-African ethnicity of the Olagunju 2018 study population. The 983T>C variant is also known as CYP2B6*18 and is found primarily in African and African-admixed populations (allele frequency ~6-8% in West Africans, ~0% in Europeans and East Asians); the Schipani 2011 cohort included 33% Black-ethnicity subjects, which explains the observed 3% heterozygote frequency, while the all-Nigerian Olagunju 2018 cohort shows the typical West-African 25% heterozygote frequency. No 983CC homozygotes have been described in the published literature as of the 2011 report (Schipani 2011 Results paragraph 4), so models that include the variant typically estimate only a heterozygous shift. Often co-tested with 516G>T (linkage-disequilibrium block: subjects homozygous for both 516TT and 983TC are CYP2B6 “slow metabolizers” of efavirenz / nevirapine). The count-form encoding follows the same rationale as SNP_CYP2B6_RS3745274_T_COUNT.

SNP_CYP2B6_RS4803419_T_COUNT (canonical for CYP2B6 c.485-18C>T (rs4803419) T-allele count)

• Description: Continuous individual-level CYP2B6 c.485-18C>T (rs4803419) T-allele count: 0 = CC homozygous wild-type, 1 = CT heterozygous, 2 = TT homozygous variant. Time-invariant (germline genotype). rs4803419 is an intron-3 splice-region variant (also reported as CYP2B6 15582C>T) that tags the CYP2B6*22-associated reduced-expression haplotype; it is in partial linkage disequilibrium with c.516G>T (rs3745274) but carries an independently detectable effect on efavirenz apparent oral clearance in some cohorts.
• Units: (count, 0/1/2 alleles per subject)
• Type: continuous
• Scope: specific
• Reference category: n/a (continuous). In the Vucicevic 2025 efavirenz model the count is decomposed in model into a single recessive (TT-versus-rest) indicator, SNP_CYP2B6_RS4803419_T_COUNT == 2, multiplied into a fractional-change factor on CL/F; heterozygotes are pooled with CC wild-type in the reference group.
• Source aliases:
  • RSB – Vucicevic 2025 (paper Table 2 final-model equation CL/F = CLp * (1 + theta_RSA)^RSA * (1 + theta_RSB)^RSB; the Table 2 footnote describes RSB as a ‘carrier’ indicator, but Results paragraph 2 identifies the flagged stratum as the TT genotype (n = 12) and Table 3 contrasts the simulated strata as ‘CYP2B6 c.485-18TT’ versus ‘CYP2B6 c.485-18CC’, so the canonical count column reconstructs RSB via SNP_CYP2B6_RS4803419_T_COUNT == 2).
• Example models: Vucicevic_2025_efavirenz.R (recessive multiplicative factor on apparent oral clearance: (1 + e_cyp2b6_485tt_cl)^(SNP_CYP2B6_RS4803419_T_COUNT == 2) with e_cyp2b6_485tt_cl = -0.268; TT homozygotes have CL/F 26.8% lower than the pooled CC / CT reference; Vucicevic 2025 Table 2 theta_RSB, RSE 16%).
• Notes: Cohort distribution in the Vucicevic 2025 Serbian Caucasian HIV-1 cohort (Table 1, n = 89): 12 (13.48%) TT homozygotes; the CC / CT split within the remaining 77 subjects is not reported. The recessive (TT-only) encoding is the published parameterization, so a _HET sibling is not identifiable from this source; a future paper resolving all three strata should add SNP_CYP2B6_RS4803419_T_HET following the SNP_ABCG2_RS2231142_HET / _HOM precedent, or use the count directly for a per-allele model. The literature on this variant is split: Bertrand 2014 (Cambodian adults on efavirenz plus antituberculosis treatment) reported a 30% CL decrease, consistent with the Vucicevic 2025 estimate, whereas Bienczak 2016 (Ugandan and Zambian children) found no effect; the variant is not included in the CPIC CYP2B6-efavirenz dosing guideline, which covers only c.516G>T and c.983T>C. Follows the _COUNT encoding used by the sibling CYP2B6 canonicals SNP_CYP2B6_RS3745274_T_COUNT, SNP_CYP2B6_RS28399499_C_COUNT, and SNP_CYP2B6_RS35303484_G_COUNT: a single count column captures the germline genotype without redundancy and lets model() derive a per-allele effect (* count), a HET / HOM decomposition (* (count == 1), * (count == 2)), or a dominant / recessive indicator (* (count >= 1), * (count == 2)) as the source paper’s parameterization requires. Auto-approved member of the SNP_<GENE>_RS<rsid> family.

SNP_CYP2C19_RS3814637_VAR_COUNT (canonical for CYP2C19 rs3814637 variant-allele count)

• Description: Continuous individual-level CYP2C19 rs3814637 variant-allele count: 0 = wild-type homozygote (the more common allele on both chromosomes), 1 = heterozygote, 2 = variant homozygote. Time-invariant (germline genotype). Companion missing-genotype indicator: SNP_CYP2C19_RS3814637_MISSING (when missing, all three genotype-indicator products evaluate to zero and the missing-multiplier is applied instead). The rs3814637 SNP is in the CYP2C19 gene region and is reported in Lane 2011 as significantly influencing R-warfarin clearance; the paper notes that this SNP is NOT in linkage disequilibrium with the canonical CYP2C19*2 loss-of-function allele (rs4244285) in the Lane 2011 patient population, so it is registered as a distinct canonical column and is not interchangeable with CYP2C19_S2_CARRIER.
• Units: (count, 0/1/2 alleles per subject)
• Type: continuous
• Scope: general
• Reference category: n/a (continuous). In the Lane 2011 R-warfarin model the count is decomposed in model into mutually exclusive wild-homozygote (count == 2 - VAR_COUNT == 2), heterozygote (VAR_COUNT == 1), and variant-homozygote (VAR_COUNT == 2) indicators, each multiplied by a distinct typical-value CL multiplier.
• Source aliases:
  • CYP2C19 genotype / rs3814637 – Lane 2011 (paper Methods ‘Covariate selection and models’; coded as wild-type / heterozygote / mutant-homozygote with a separate Missing category). Heterozygote vs wild-homozygote CL multiplier 0.761, mutant-homozygote multiplier 0.494 (Lane 2011 Table 3).
• Example models: Lane_2011_warfarin_r.R (categorical CL multiplier 1.00 / 0.761 / 0.494 for wild / heterozygote / variant homozygote; Lane 2011 Table 3 final R-warfarin model).
• Notes: Follows the SNP_<GENE>_RS<rsid>_<allele>_COUNT pattern established by SNP_CYP2B6_RS3745274_T_COUNT and SNP_CYP2B6_RS28399499_C_COUNT. The _VAR token (rather than a specific allele letter) is used because Lane 2011 does not specify which allele is the wild-type vs variant in the paper text, only reporting the categorical wild-type / heterozygote / mutant-homozygote distinction. Future extractions that report rs3814637 with explicit allele letters (e.g. T/C) should add an aliased canonical such as SNP_CYP2C19_RS3814637_T_COUNT and cross-reference. Distinct from CYP2C19_S2_CARRIER (rs4244285): Lane 2011 explicitly notes that rs3814637 is not in linkage disequilibrium with CYP2C19*2 in their cohort.

SNP_CYP2C19_RS3814637_MISSING (canonical for CYP2C19 rs3814637 genotype-missing indicator)

• Description: Binary indicator for a subject whose CYP2C19 rs3814637 genotype was not measured / not available. 1 = genotype missing; 0 = genotype known (in which case SNP_CYP2C19_RS3814637_VAR_COUNT carries the 0/1/2 variant-allele count). Companion to SNP_CYP2C19_RS3814637_VAR_COUNT so popPK models that estimate a separate typical-value covariate effect for the missing-genotype subgroup (rather than imputing it as wild-type) can do so faithfully.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (genotype known). When SNP_CYP2C19_RS3814637_MISSING == 1 the SNP_CYP2C19_RS3814637_VAR_COUNT column should be 0 so the genotype-indicator products correctly evaluate to zero for the missing-genotype subject.
• Source aliases:
  • Missing (Lane 2011 Table 3 q_CYP2C19 Missing row – the paper treats unknown CYP2C19 rs3814637 genotype as a distinct categorical level with its own multiplicative CL effect; 61 of 309 R-warfarin subjects were missing CYP2C19 genotype per Lane 2011 Table 1).
• Example models: Lane_2011_warfarin_r.R (multiplicative CL effect 0.804 for the missing-genotype subgroup vs the wild-homozygote reference; Lane 2011 Table 3 final R-warfarin model).
• Notes: Follows the same missing-data-indicator pattern as CYP2C9_MISSING, ADA_MISSING, and HEPIMP_MOD_OR_MISSING.

SNP_CYP2C19_RS12769205_GA (**canonical for CYP2C19*2 rs12769205 heterozygous GA genotype indicator**)

• Description: Binary genotype indicator for the CYP2C19 rs12769205 single-nucleotide polymorphism, the intronic tag SNP for the CYP2C19*2 loss-of-function allele (c.-98-273A>G; in strong linkage disequilibrium with the canonical *2 defining variant rs4244285, c.681G>A). 1 = subject’s reported genotype is GA (heterozygous); 0 = otherwise. Paired with SNP_CYP2C19_RS12769205_GG; both indicators are 0 for the AA reference group, which the founding source identifies explicitly as the wild type. Time-fixed per subject (germline genotype). Named by the reported genotype letters, following the SNP_ABCB1_RS1045642_GA / _GG and SNP_POR_RS1057868_GA / _GG pattern.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 in combination with SNP_CYP2C19_RS12769205_GG = 0, i.e. the AA genotype group (n = 56 of 133 in the Li 2025 model-development group).
• Source aliases:
  • CL-rs12769205*GA – the Li 2025 Table 2 row label for the CL/F fold-change of the GA group.
  • rs12769205 – the Li 2025 Table 1 genotype row, reported as AA/GA/GG counts.
• Example models: Li_2025_lacosamide_cyp2c19.R (multiplicative fold-change on apparent clearance of oral lacosamide in Chinese children with epilepsy: cl = exp(lcl + etalcl) * (WT/30)^e_wt_cl * genoCl with genoCl = 1 + (0.879 - 1) * SNP_CYP2C19_RS12769205_GA + (0.736 - 1) * SNP_CYP2C19_RS12769205_GG, i.e. GA subjects have 0.879-fold, about 12% lower, apparent CL/F than the AA reference group).
• Notes: Genotype-level (three-category) encoding. Unlike the ABCB1 rs1045642 and POR rs1057868 pairs, whose wild-type / variant orientation the register records as unresolved, the orientation here is resolved by the founding source: Li 2025 states “theta_rs12769205 = 1 for CYP2C19*2 AA” and the Discussion calls the AA group the “wild-type”, with the GG group showing “a 26.4% reduction in CL/F (i.e., 0.736 x wild-type)” consistent with the loss of function annotated for CYP2C19*2 in PharmGKB. So A is the wild-type allele and G the *2-tagging reduced-function allele, and the effects are monotonic in G-allele count (AA 1 > GA 0.879 > GG 0.736). Not composable with the phenotype-level canonicals CYP2C19_S2_CARRIER, CYP2C19_IM, CYP2C19_PM, CYP2C19_NON_EM or the activity proxy CYP2C19: those pool diplotypes across several loci (*2, *3, *17) into metabolizer classes, whereas this pair resolves a single locus by reported genotype and takes the wild-type homozygote as reference. A model needing metabolizer phenotype must derive it from the full diplotype, not from this indicator alone. Li 2025 genotyped nine SNPs across ABCB1, ABCC2, CYP2C9 and CYP2C19 and found rs12769205 to be the only one significantly affecting CL/F.

SNP_CYP2C19_RS12769205_GG (**canonical for CYP2C19*2 rs12769205 homozygous GG genotype indicator**)

• Description: Binary genotype indicator for the CYP2C19 rs12769205 single-nucleotide polymorphism, the intronic tag SNP for the CYP2C19*2 loss-of-function allele. 1 = subject’s reported genotype is GG (homozygous for the G, *2-tagging, allele); 0 = otherwise. Paired with SNP_CYP2C19_RS12769205_GA; both indicators are 0 for the AA wild-type reference group. Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 in combination with SNP_CYP2C19_RS12769205_GA = 0, i.e. the AA genotype group (n = 56 of 133 in the Li 2025 model-development group).
• Source aliases:
  • CL-rs12769205*GG – the Li 2025 Table 2 row label for the CL/F fold-change of the GG group.
  • rs12769205 – the Li 2025 Table 1 genotype row, reported as AA/GA/GG counts.
• Example models: Li_2025_lacosamide_cyp2c19.R (multiplicative fold-change 0.736 on apparent clearance, i.e. GG subjects have about 26% lower CL/F than the AA reference group and correspondingly higher steady-state troughs; Li 2025 reduces the recommended dose for this subgroup).
• Notes: Always used together with SNP_CYP2C19_RS12769205_GA; a model that references one without the other has mis-specified the reference category. See the SNP_CYP2C19_RS12769205_GA Notes for the resolved wild-type orientation and for why this pair must not be pooled with the CYP2C19 metabolizer-phenotype canonicals. Only 15 of 133 subjects in the Li 2025 model-development group were GG, so the coefficient rests on a small subgroup, though its reported precision is high (RSE 5.5%, bootstrap 95% CI 0.661-0.825). The Li 2025 Discussion notes the CYP2C19*2 allele frequency exceeds 30% in Asian populations, which amplifies the clinical significance of this indicator relative to European or African cohorts where *17 predominates.

SNP_CYP3A4_RS2242480_VAR_COUNT (canonical for CYP3A4 rs2242480 variant-allele count)

• Description: Continuous individual-level CYP3A4 rs2242480 variant-allele count: 0 = wild-type homozygote, 1 = heterozygote, 2 = variant homozygote. Time-invariant (germline genotype). Companion missing-genotype indicator: SNP_CYP3A4_RS2242480_MISSING. The rs2242480 SNP is one of the SNPs that defines the CYP3A4*1G haplotype (a 5’ UTR variant whose functional effect remains controversial – some studies report gain-of-function, others loss-of-function; Lane 2011 Discussion paragraph 5).
• Units: (count, 0/1/2 alleles per subject)
• Type: continuous
• Scope: general
• Reference category: n/a (continuous). In the Lane 2011 R-warfarin model the count is decomposed in model into mutually exclusive wild-homozygote, heterozygote, and variant-homozygote indicators, each multiplied by a distinct typical-value CL multiplier.
• Source aliases:
  • CYP3A4 genotype / rs2242480 – Lane 2011 (paper Methods ‘Covariate selection and models’; coded as wild-type / heterozygote / mutant-homozygote with a separate Missing category). Heterozygote vs wild-homozygote CL multiplier 1.32, variant-homozygote multiplier 1.06 (Lane 2011 Table 3); Lane 2011 explicitly notes the heterozygote vs wild contrast is the significant effect (“Those patients with the heterozygote genotype for CYP3A4 had a 32% increased clearance for R-warfarin”, Lane 2011 Results paragraph ‘R-Warfarin models’).
• Example models: Lane_2011_warfarin_r.R (categorical CL multiplier 1.00 / 1.32 / 1.06 for wild / heterozygote / variant homozygote; Lane 2011 Table 3 final R-warfarin model), Li_2019_cyclosporine.R (source column Gene coded 1 = CC, 2 = TT, 3 = CT; CL multiplier 0.984 for CC and 1.22 for CT and TT on intravenous cyclosporine in Chinese children after allo-HSCT; the T allele is the variant (minor) allele, frequency 23.2%).
• Notes: Follows the SNP_<GENE>_RS<rsid>_<allele>_COUNT pattern. The _VAR token (rather than a specific allele letter) is used because Lane 2011 does not specify allele letters in the genotype table. Lane 2011 Discussion paragraph 5 notes that the functional effect of rs2242480 / CYP3A4*1G has been subject to controversy in the literature (Gao 2008 and Du 2006 report gain-of-function; Wang 2011 and He 2011 report loss-of-function; a luciferase reporter study in Wang 2013 found higher transcriptional activity for the A allele than the G allele). The Lane 2011 finding of higher R-warfarin clearance in heterozygotes is consistent with a gain-of-function effect in heterozygotes, although the small variant-homozygote stratum (n=3) limits inference for that sub-group.

SNP_CYP3A4_RS2242480_MISSING (canonical for CYP3A4 rs2242480 genotype-missing indicator)

• Description: Binary indicator for a subject whose CYP3A4 rs2242480 genotype was not measured / not available. 1 = genotype missing; 0 = genotype known. Companion to SNP_CYP3A4_RS2242480_VAR_COUNT so popPK models that estimate a separate typical-value covariate effect for the missing-genotype subgroup can do so faithfully.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (genotype known). When SNP_CYP3A4_RS2242480_MISSING == 1 the SNP_CYP3A4_RS2242480_VAR_COUNT column should be 0 so the genotype-indicator products correctly evaluate to zero.
• Source aliases:
  • Missing (Lane 2011 Table 3 q_CYP3A4 Missing row; 38 of 309 R-warfarin subjects were missing CYP3A4 genotype per Lane 2011 Table 1).
• Example models: Lane_2011_warfarin_r.R (multiplicative CL effect 0.937 for the missing-genotype subgroup vs the wild-homozygote reference; Lane 2011 Table 3 final R-warfarin model).
• Notes: Follows the same missing-data-indicator pattern as SNP_CYP2C19_RS3814637_MISSING, CYP2C9_MISSING, ADA_MISSING, and HEPIMP_MOD_OR_MISSING.

SNP_CYP3A4_RS35599367 (**canonical for CYP3A4*22 (rs35599367) reduced-function allele carrier indicator**)

• Description: Binary genotype indicator for the CYP3A4*22 allele (rs35599367; c.522-191C>T; intron 6 T variant that lowers CYP3A4 mRNA splicing efficiency and hence hepatic CYP3A4 expression). 1 = subject carries at least one T (*22) allele (heterozygous *1/*22 or homozygous *22/*22); 0 = homozygous wild-type *1/*1. Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (CYP3A4 *1/*1 wild-type non-carrier).
• Source aliases:
  • CYP3A4 Genotype / CYP3A4*22 – used in MohammedAli_2023_tacrolimus.R (Mohammed Ali 2023 Methods section 2.3 CYP3A4*22 C > T (rs35599367) TaqMan genotyping; Table 1 reports *1/*1 86 (86.7%) vs *1/*22 12 (13.3%), so heterozygotes and the (unobserved) homozygotes are pooled into the carrier group exactly as the paper’s CYP3A4*22 carriers vs non-carriers contrast does).
• Example models: MohammedAli_2023_tacrolimus.R (one of the two binary inputs from which the three-level CYP3A4/CYP3A5 cluster metabolizer phenotype is reconstructed inside model(): is_hm <- (1 - SNP_CYP3A4_RS35599367) * CYP3A5_EXPR, is_pm <- SNP_CYP3A4_RS35599367 * (1 - CYP3A5_EXPR), is_im <- 1 - is_hm - is_pm; the three indicators then select one of three separately estimated typical CL/F values, 19.6 / 10.6 / 7.37 L/h).
• Notes: Distinct from the two already-registered CYP3A4 SNP canonicals: SNP_CYP3A4_RS2246709 (active-allele count) and SNP_CYP3A4_RS2242480_VAR_COUNT (variant-allele count for the *1G / c.1026+12G>A intronic variant) are different loci, and the continuous CYP3A4 canonical is a probe-substrate-derived activity score rather than a genotype. *22 is the principal reduced-function CYP3A4 allele in the tacrolimus / CYP3A pharmacogenetics literature and is nearly always analysed jointly with the CYP3A5 *3 expresser status (CYP3A5_EXPR), because the two genes together determine total CYP3A metabolic capacity. Follow the two-binary-input convention documented in the CYP3A5_EXPR_DONOR Notes: when a source paper fits a combined CYP3A4 + CYP3A5 “cluster” phenotype (high / intermediate / poor metabolizer), keep the underlying genotypes as separate canonical columns (CYP3A5_EXPR and SNP_CYP3A4_RS35599367) and reconstruct the cluster levels inside model(), rather than registering a single collapsed three-level cluster column – this keeps the underlying biology explicit in the dataset and lets a downstream user re-derive any other clustering rule.

SNP_NR1I2_RS2461817_HOM (canonical for NR1I2 (PXR) rs2461817 homozygous-variant indicator)

• Description: Binary genotype indicator for the NR1I2 (pregnane X receptor, PXR) rs2461817 single-nucleotide polymorphism. 1 = homozygous variant; 0 = otherwise (the union of homozygous wild-type and heterozygous carriers). This is a recessive-model encoding: heterozygotes are pooled with wild-type homozygotes because the founding paper resolved a distinct typical-value effect only for the homozygous-mutant stratum. Time-fixed per subject (germline genotype). NR1I2 encodes PXR, the nuclear receptor that transcriptionally regulates CYP3A4, CYP3A5 and ABCB1, so NR1I2 variants act on drug disposition indirectly by modulating downstream enzyme and transporter expression rather than by altering an enzyme’s own catalytic activity.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (homozygous wild-type or heterozygous). In a recessive (_HOM-only) encoding the 0-level pools both non-homozygous-variant strata. If a future paper resolves a distinct heterozygote effect, register a paired SNP_NR1I2_RS2461817_HET companion following the SNP_ABCG2_RS2231142_HET / _HOM precedent; the per-subject value of the _HOM column is unchanged by that pairing.
• Source aliases:
  • GENE – used in Xie_2025_midazolam.R. Xie 2025 names the column GENE in its Figure 6 legend and defines the recessive pooling verbatim: “GENE = 1 indicates that the NR1I2 rs2461817 genotype reflects a homozygous mutation. GENE = 0 indicates that the NR1I2 rs2461817 genotype is wild-type homozygous or mutant heterozygous.” The bare name GENE is far too generic to be canonical, hence the explicit gene + rsID + genetic-model canonical.
• Example models: Xie_2025_midazolam.R (fractional-shift effect on midazolam clearance in mechanically ventilated Chinese ICU patients: nr1i2_factor <- 1 + e_nr1i2_hom_cl * SNP_NR1I2_RS2461817_HOM with e_nr1i2_hom_cl = -0.405; homozygous mutants have midazolam CL 40.5% lower, 13.4 vs 22.6 L/h; paper Table 2 “rs2461817 on CL MDZ”, RSE 29%, bootstrap 95% CI -0.591 to -0.0659).
• Notes: Functional form is fractional, not exponential. Xie 2025 does not print the covariate equation, but the fractional form (1 + theta * GENE) is the only one consistent with its own reported simulation clearances: 22.6 * (1 - 0.405) = 13.45 matches the reported 13.4 L/h, whereas 22.6 * exp(-0.405) = 15.07 does not. Check this before reusing the coefficient in another parameterisation. Distinct from the CYP3A4 / CYP3A5 genotype canonicals (SNP_CYP3A4_RS2246709, SNP_CYP3A4_RS2242480_VAR_COUNT, SNP_CYP3A4_RS35599367, CYP3A5_EXPR): those are the metabolising enzymes themselves, whereas NR1I2 is their upstream transcriptional regulator. Xie 2025 genotyped 23 loci across four genes (CYP3A4 rs2242480, rs2246709; CYP3A5 rs776746, rs15524; ABCB1 rs2032582, rs1128503, rs1045642; and 16 NR1I2 loci) and found only rs2461817 significant on midazolam clearance – the Discussion attributes the null CYP3A4 / CYP3A5 result to ICU drug-drug interactions and inflammatory cytokines masking the genetic signal. Other NR1I2 loci screened in that cohort and available for future canonicals: rs1464603, rs1464602, rs3732357, rs6785049, rs2276707, rs10934498, rs3814055, rs2472677, rs3732359, rs3814058, rs3732360, rs4688040, rs2276706, rs1523130, rs1523127.

CYP2C19_S2_CARRIER (**canonical for CYP2C19*2 loss-of-function allele carrier indicator**)

• Description: Binary indicator for carriage of the CYP2C192 loss-of-function allele (rs4244285, c.681G>A; creates a cryptic splice site that abolishes CYP2C19 enzyme activity). 1 = subject carries at least one 2 allele (heterozygous 1/2 or homozygous 2/2); 0 = no 2 allele. Time-fixed per subject (germline genotype). Het and hom carriers are pooled in studies where the 2/2 (poor metabolizer) frequency is too low for a separate phenotype layer; future extractions that distinguish heterozygous-intermediate-metabolizer from homozygous-poor-metabolizer effects should register paired CYP2C19_S2_HET / CYP2C19_S2_HOM indicators (or, where the source uses the per-allele count form, a CYP2C19_S2_COUNT column following the CYP2C9_S2_COUNT precedent).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no 2 allele – 1/1, 1/17, or 17/*17).
• Source aliases:
  • CYP2C19*2 – Danielak 2017 (paper Methods ‘Determination of genetic polymorphisms’ and Table 2 final-model Effect of CYP2C19*2 on FM (COV) row; PCR-RFLP genotyping for rs4244285). The Danielak 2017 cohort had no 2/2 homozygous-poor-metabolizers, so heterozygous 1/2 carriers were pooled into the binary CYP2C19_S2_CARRIER = 1 group with no information loss.
• Example models: Danielak_2017_clopidogrel.R (linear-deviation effect on the fraction of clopidogrel metabolised to the active thiol H4: fm = TVFM * (1 + e_cyp2c19_s2_fm * CYP2C19_S2_CARRIER) with e_cyp2c19_s2_fm = -0.45; carriers convert 45% less of the absorbed clopidogrel to the active H4 metabolite, giving a 36.7% lower predicted AUC of H4 vs non-carriers; Danielak 2017 Table 2 final-model and Results page 1628), Yin_2021_mitotane.R (multiplicative effect on apparent clearance of mitotane; GA/AA carriers have CL/F 0.551x the GG reference, i.e. 44.9% lower, Yin 2021 Table 2 CL_SNP1 (GA/AA) = 0.551).
• Notes: CYP2C192 is the dominant loss-of-function* CYP2C19 allele in clopidogrel pharmacogenetics; it reduces clopidogrel’s metabolic activation to the antiplatelet-active H4 thiol and is associated with elevated rates of stent thrombosis and major adverse cardiovascular events on clopidogrel therapy (FDA boxed warning, 2010). The paired *17 ultra-rapid-metabolizer allele (rs12248560) is typically registered separately when present (a CYP2C19_S17_CARRIER indicator following the same pattern). The continuous-individual-activity-score consolidation TODO logged on CYP2D6 line 3321 also applies prospectively to CYP2C19, but the binary carrier indicator remains the standard discrete encoding used by most published clopidogrel popPK / PD models.

CYP2C19_IM (canonical for CYP2C19 intermediate-metabolizer phenotype indicator)

• Description: 1 = subject is a CYP2C19 intermediate metabolizer (one functional and one loss-of-function allele; e.g., *1/*2, *2/*17); 0 = any other CYP2C19 phenotype (EM, UM, PM, or RM). Time-fixed per subject (germline genotype-derived phenotype). Paired with CYP2C19_PM to encode the three-level EM/UM (reference, both indicators 0) / IM (CYP2C19_IM = 1) / PM (CYP2C19_PM = 1) phenotype with two binary indicators.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (EM, UM, PM, or RM phenotype). Reference category when paired with CYP2C19_PM = 0 is the extensive / ultrarapid-metabolizer phenotype pool used in Zhao 2018.
• Source aliases:
  • CYP2C19 IM / IM – Zhao 2018 (paper Table 2 reports F_CYP2C19 IM = 0.449 relative to the EM/UM reference; genotypes pooled into IM: *1/*2, *2/*17).
• Example models: Zhao_2018_omeprazole.R (power-of-binary-indicator multiplicative factor on CLOMZ-M1 formation clearance: e_cyp2c19_im_kmet_5oh ^ CYP2C19_IM with e_cyp2c19_im_kmet_5oh = 0.449; IM subjects have ~55% lower formation clearance of 5-hydroxy-omeprazole than the EM/UM reference; Zhao 2018 Table 2), Jung_2024_clopidogrel.R (additive shift on the LOGIT scale of two nested metabolized fractions: e_cyp2c19_im_logitfm1 = -0.450 and e_cyp2c19_im_logitfm2 = -1.428, taking the active-metabolite fraction fmH4 from 0.120 in EM to 0.071 in IM; Jung 2024 Tables 2 and 3), Xu_2020_clopidogrel_pbpk.R (multiplicative factor on the CYP2C19 Vmax of both clopidogrel oxidation steps: 0.5 in IM; the reference is EM only because UM is flagged separately by CYP2C19_UM; Xu 2020 Table 3).
• Notes: Follows the CYP2B6_IM / CYP2B6_SM / CYP2B6_USM three-binary precedent for multi-level metabolizer phenotypes. The Zhao 2018 cohort pooled extensive (EM, *1/*1) and ultrarapid (UM, *1/*17, *17/*17) metabolizers into a single reference because the typical-value clearance of 5-hydroxy-omeprazole formation was indistinguishable between the two strata in n = 38 EM/UM subjects (Zhao 2018 Methods ‘Population pharmacokinetic-pharmacogenetic modelling’). Future extractions that fit a separate UM coefficient should register a paired CYP2C19_UM companion indicator following this pattern. Distinct from CYP2C19_S2_CARRIER (binary *2-allele carrier indicator used by Danielak 2017 clopidogrel) – CYP2C19_S2_CARRIER pools heterozygous and homozygous *2 carriers into a single 0/1 contrast, while CYP2C19_IM resolves the heterozygous *2 (IM) stratum separately from the homozygous *2/*2 (PM) stratum that CYP2C19_PM flags.

CYP2C19_PM (canonical for CYP2C19 poor-metabolizer phenotype indicator)

• Description: 1 = subject is a CYP2C19 poor metabolizer (two loss-of-function alleles, e.g., *2/*2); 0 = any other CYP2C19 phenotype (EM, UM, IM, or RM). Time-fixed per subject (germline genotype-derived phenotype). Paired with CYP2C19_IM to encode the three-level EM/UM (reference) / IM / PM phenotype with two binary indicators.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (EM, UM, IM, or RM phenotype). Reference category when paired with CYP2C19_IM = 0 is the extensive / ultrarapid-metabolizer phenotype pool used in Zhao 2018.
• Source aliases:
  • CYP2C19 PM / PM – Zhao 2018 (paper Table 2 reports F_CYP2C19 PM = 0.125 relative to the EM/UM reference; genotype pooled into PM: *2/*2).
• Example models: Zhao_2018_omeprazole.R (power-of-binary-indicator multiplicative factor on CLOMZ-M1 formation clearance: e_cyp2c19_pm_kmet_5oh ^ CYP2C19_PM with e_cyp2c19_pm_kmet_5oh = 0.125; PM subjects have 87.5% lower formation clearance of 5-hydroxy-omeprazole than the EM/UM reference; Zhao 2018 Table 2), Marathe_2023_belzutifan.R (linear-deviation multiplicative factor on apparent clearance: cl_cyp2c19 = 1 + e_cyp2c19_pm_cl * CYP2C19_PM with e_cyp2c19_pm_cl = -0.36, i.e. -36.0% CL/F versus the non-poor-metabolizer reference (5.63 -> 3.60 L/h); Marathe 2023 Table 2 and Table 3. Here the reference pools ALL non-poor phenotypes (intermediate, extensive, rapid, and ultrarapid) because “only the PM category had a significantly different CL/F compared to all other categories”, so CYP2C19_IM is deliberately not used alongside it), Jung_2024_clopidogrel.R (additive shift on the LOGIT scale of two nested metabolized fractions: e_cyp2c19_pm_logitfm1 = -0.996 and e_cyp2c19_pm_logitfm2 = -2.432, taking the active-metabolite fraction fmH4 from 0.120 in EM to 0.034 in PM; Jung 2024 Tables 2 and 3), Xu_2020_clopidogrel_pbpk.R (multiplicative factor on the CYP2C19 Vmax of both clopidogrel oxidation steps: 0 in PM, which removes the CYP2C19 pathway; the reference is EM only because UM is flagged separately by CYP2C19_UM; Xu 2020 Table 3).
• Notes: Companion to CYP2C19_IM. See CYP2C19_IM Notes for the three-level decomposition rationale. The PM phenotype indicator carries a separate typical-value coefficient because the Zhao 2018 cohort observed *2/*2 poor metabolizers (n = 2) had substantially lower 5-OH-omeprazole formation clearance than the heterozygous *1/*2 and *2/*17 intermediate metabolizers (12.5% vs 44.9% of EM/UM reference).

CYP2C19_UM (canonical for CYP2C19 ultrarapid-metabolizer phenotype indicator)

• Description: 1 = subject is a CYP2C19 ultrarapid metabolizer (carrier of the gain-of-function *17 allele without a loss-of-function allele: *1/*17 or *17/*17); 0 = any other CYP2C19 phenotype (EM, IM or PM). Time-fixed per subject (germline genotype-derived phenotype). Paired with CYP2C19_IM and CYP2C19_PM to encode the four-level UM / EM (reference, all three indicators 0) / IM / PM phenotype with three mutually exclusive binary indicators.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (EM, IM or PM phenotype). With CYP2C19_IM = 0 and CYP2C19_PM = 0 the reference is the extensive-metabolizer (*1/*1) phenotype.
• Source aliases:
  • UM – Xu 2020 (Methods ‘In liver’ and Table 3 footnote b: UM = *1/*17 and *17/*17, CYP2C19 activity 1.58-fold of EM).
• Example models: Xu_2020_clopidogrel_pbpk.R (multiplicative factor on the CYP2C19 Vmax of both clopidogrel oxidation steps: 1 + (e_cyp2c19_um - 1) * CYP2C19_UM with e_cyp2c19_um = 1.58, so Vmax,CYP2C19 goes from 7.52 to 11.88 pmol/pmol P450/min for clopidogrel and from 9.06 to 14.31 for 2-oxo-clopidogrel; Xu 2020 Table 3).
• Notes: The paired UM companion that the CYP2C19_IM Notes anticipated for a paper that resolves UM separately. When a model uses CYP2C19_UM, the CYP2C19_IM / CYP2C19_PM reference narrows from the pooled EM/UM of Zhao 2018 to EM alone; record that in the model’s covariateData reference categories. Distinct from the rapid-metabolizer (*1/*17 only) grouping of CPIC 2022, which some papers split from *17/*17 ultrarapid metabolizers – a paper that separates the two should register a CYP2C19_RM companion rather than overload this flag.

CYP2C19_MISSING (canonical for CYP2C19 genotype-missing indicator)

• Description: Binary indicator for a subject whose CYP2C19 genotype / metabolizer phenotype was not determined. 1 = genotype missing; 0 = genotype known (in which case the phenotype indicators CYP2C19_IM, CYP2C19_PM, and any sibling phenotype columns carry the measured phenotype). Companion to the phenotype-indicator family so that popPK models estimating a separate typical value for the missing-genotype subgroup – rather than imputing those subjects into the extensive-metabolizer reference – can do so faithfully. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (genotype known). When CYP2C19_MISSING == 1 the paired phenotype indicators (CYP2C19_IM, CYP2C19_PM) must be 0, so the phenotype-indicator products evaluate to zero and only the missing-group multiplier applies.
• Source aliases:
  • missing = 3 – Akil 2016 (paper Methods ‘Final model’: “CYP2C19 genotype was regrouped to have three possible values (EM/RM = 1, IM/PM = 2 and missing = 3)”; Table 3 row CL Sp /F for Missing, L/h). 15 of 81 subjects (18.5 percent) were missing CYP2C19 genotype (Akil 2016 Table 1).
• Example models: Akil_2016_citalopram.R (multiplicative factor on S-citalopram apparent metabolic clearance: 1 + (16.8/22.1 - 1) * CYP2C19_MISSING, i.e. missing-genotype subjects have a typical CL/F of 16.8 L/h against 22.1 L/h in the EM/RM reference – close to, but estimated separately from, the 16.3 L/h IM/PM stratum).
• Notes: Member of the established <COV>_MISSING missing-indicator family (ADA_MISSING, AAG_MISSING, CYP2C9_MISSING, SNP_CYP2C19_RS3814637_MISSING, SNP_CYP3A4_RS2242480_MISSING): a flag that distinguishes “value unmeasured” from a measured value, so unmeasured subjects are not silently treated as exchangeable with the reference group. Phenotype-level, and therefore distinct from SNP_CYP2C19_RS3814637_MISSING, which flags a missing genotype at one specific locus (rs3814637) rather than a missing metabolizer-phenotype assignment derived from the full diplotype; a model that pools diplotypes into EM / IM / PM classes needs this column, not the locus-specific one. Whether the missing-group coefficient is interpretable is a per-paper question: it estimates the clearance of a mixture of unmeasured phenotypes weighted by their prevalence in that cohort, so it does not transfer across populations with different allele frequencies. Record in each model’s covariateData[[CYP2C19_MISSING]]$notes what the source’s missing fraction was.

UGT2B17_EM (canonical for UGT2B17 extensive-metabolizer phenotype indicator)

• Description: 1 = subject is a UGT2B17 extensive metabolizer (both alleles functional, corresponding to a wild-type diplotype at the highly polymorphic UGT2B17 copy-number / whole-gene-deletion locus); 0 = any other UGT2B17 phenotype (intermediate or poor). Time-fixed per subject (germline genotype-derived phenotype). Paired with UGT2B17_PM to encode the three-level EM (UGT2B17_EM = 1) / IM (both indicators 0, reference) / PM (UGT2B17_PM = 1) phenotype with two binary indicators.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (intermediate or poor metabolizer). When paired with UGT2B17_PM = 0 the reference collapses to the UGT2B17 intermediate-metabolizer stratum used in Marathe 2023.
• Source aliases:
  • UGT2B17P (three-level categorical, 1 = poor / 2 = intermediate / 3 = extensive, -99 = missing) – Marathe 2023 (supplement 2 NONMEM control stream branches IF(UGT2B17P.EQ.3) CLUGT2B17P = (1 + THETA(12)), with level 2 as “most common” reference and the -99 missing sentinel mapped onto the reference). Decomposed into the paired binary indicators UGT2B17_EM / UGT2B17_PM.
  • UGT2B17 EM / UGT2B17 (extensive) – Marathe 2023 Table 2 and Table 3 row labels.
• Example models: Marathe_2023_belzutifan.R (linear-deviation multiplicative factor on apparent clearance: cl_ugt2b17 = 1 + e_ugt2b17_em_cl * UGT2B17_EM + e_ugt2b17_pm_cl * UGT2B17_PM with e_ugt2b17_em_cl = 0.39, i.e. +39.1% CL/F for extensive metabolizers versus the intermediate-metabolizer reference (5.63 -> 7.83 L/h); Marathe 2023 Table 2 and Table 3).
• Notes: Follows the CYP2D6_PM / CYP2D6_EM and CYP2C19_IM / CYP2C19_PM paired-binary precedents for multi-level metabolizer phenotypes (two binary indicators encoding a three-level categorical with the intermediate stratum as the implicit reference). UGT2B17 is a phase-II glucuronidation enzyme whose poor-metabolizer phenotype arises from homozygous whole-gene deletion rather than from point mutations, which is why the phenotype – not a single SNP_UGT2B17_RS<rsid> genotype indicator – is the right canonical here. Note that Marathe 2023 retains all three UGT2B17 levels on CL/F but only two on bioavailability (EM and IM merged), so UGT2B17_EM has no bioavailability effect in that model while UGT2B17_PM does; a model needing the ultrarapid stratum should register a paired UGT2B17_UM on the same pattern. Distinct from the UGT2B7_* genotype canonicals above (a different UGT isoform, and genotype- rather than phenotype-based).

UGT2B17_PM (canonical for UGT2B17 poor-metabolizer phenotype indicator)

• Description: 1 = subject is a UGT2B17 poor metabolizer (homozygous UGT2B17 whole-gene deletion, i.e. del/del at the UGT2B17 copy-number locus, giving no enzyme activity); 0 = any other UGT2B17 phenotype (extensive or intermediate). Time-fixed per subject (germline genotype-derived phenotype). Paired with UGT2B17_EM to encode the three-level EM / IM (reference) / PM phenotype with two binary indicators.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (extensive or intermediate metabolizer). When paired with UGT2B17_EM = 0 the reference collapses to the UGT2B17 intermediate-metabolizer stratum. Note that a single model may use two different reference categories for two different effects of this indicator – in Marathe 2023 the CL/F effect is referenced to the intermediate stratum alone (all three levels retained) while the bioavailability effect is referenced to the pooled intermediate + extensive stratum (only two levels retainable on F).
• Source aliases:
  • UGT2B17P (three-level categorical, 1 = poor / 2 = intermediate / 3 = extensive, -99 = missing) – Marathe 2023 (supplement 2 NONMEM control stream branches IF(UGT2B17P.EQ.1) CLUGT2B17P = (1 + THETA(13)) for CL and IF(UGT2B17P.EQ.1) F1UGT2B17P = (1 + THETA(15)) for bioavailability). Decomposed into the paired binary indicators UGT2B17_EM / UGT2B17_PM.
  • UGT2B17 PM / UGT2B17 (poor) – Marathe 2023 Table 2 and Table 3 row labels.
• Example models: Marathe_2023_belzutifan.R (two distinct linear-deviation effects: on apparent clearance cl_ugt2b17 = 1 + e_ugt2b17_em_cl * UGT2B17_EM + e_ugt2b17_pm_cl * UGT2B17_PM with e_ugt2b17_pm_cl = -0.24, i.e. -24.2% CL/F versus the intermediate-metabolizer reference (5.63 -> 4.27 L/h); and on relative bioavailability f(depot) = exp(lfdepot) * (1 + e_ugt2b17_pm_fdepot * UGT2B17_PM) with e_ugt2b17_pm_fdepot = 0.11, i.e. +11.0% F versus the pooled intermediate + extensive reference; Marathe 2023 Table 2, Table 2 caption equation F = 1 * (1 + F-UGT2B17P), and Table 3).
• Notes: Companion to UGT2B17_EM; see that entry’s Notes for the three-level decomposition rationale and the phenotype-versus-genotype naming choice. Follows the CYP2C19_PM (Zhao 2018 omeprazole) and CYP2D6_PM (Knights 2015 aripiprazole) precedents. Marathe 2023 attributes the positive bioavailability effect to reduced first-pass glucuronidation in the gut when UGT2B17 activity is absent (Discussion), which is mechanistically consistent with the negative CL/F effect from the same deletion. The UGT2B17 deletion is strongly ancestry-stratified: Marathe 2023 reports the UGT2B17/CYP2C19 dual-PM phenotype frequency at ~0.5% in the overall US population but up to 15% in East Asians.

UGT2B15_STAR2_HET (**canonical for UGT2B15*2 heterozygote indicator**)

• Description: Binary germline genotype indicator for the UGT2B151/2 heterozygote group. 1 = subject carries exactly one UGT2B152 allele (diplotype 1/2); 0 = otherwise (the union of 1/1 wild-type homozygotes and 2/2 homozygous carriers; the paired indicator UGT2B15_STAR2_HOM flags the 2/2 group, and UGT2B15_MISSING flags subjects whose genotype was never collected). Time-fixed per subject. UGT2B152 is the reduced-glucuronidation variant allele, so *2 carriers clear UGT2B15 substrates more slowly and have higher exposure.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (UGT2B151/1 wild-type homozygote), which is realised when UGT2B15_STAR2_HET, UGT2B15_STAR2_HOM and UGT2B15_MISSING are all 0.
• Source aliases:
  • UGT2B15*1/*2 – Stringer 2014 Table 1 column header (UGT2B15*1/*2 (CL = 3.35 L/h)) and supplement Table S1 genotype-count row (UGT2B15 genotype (*1/*1:*1/*2:*2/*2) 149:357:194); Stringer 2013 Table 1 genotype-frequency column and Table 2 row label CL (*1/*2).
• Example models: Stringer_2014_sipoglitazar.R (selects the fixed per-genotype sipoglitazar clearance 3.35 L/h that converts total daily dose into the steady-state AUC0-24h driving the Emax stimulation of FPG elimination; Stringer 2014 Table 1), Stringer_2013_sipoglitazar.R (three stratum-specific typical clearances selected by the paired indicators: lcl_s1s1 = log(4.46), lcl_s1s2 = log(3.25), lcl_s2s2 = log(1.53) L/h; Stringer 2013 Table 2).
• Notes: Follows the CYP2D6_STAR10_HET / CYP2D6_STAR10_HOM, CYP3A5_STAR1_HET / CYP3A5_STAR1_HOM and SLCO1B1_HAP15_HET / SLCO1B1_HAP15_HOM precedents – paired binary indicators for a three-level germline diplotype with an implicit wild-type reference. The _STAR2_ token names the variant-allele orientation, matching the source paper’s star-allele nomenclature. Distinct from the UGT2B17_EM / UGT2B17_PM phenotype canonicals: those encode an inferred metabolizer PHENOTYPE from a whole-gene deletion locus, whereas these encode the assayed UGT2B15 DIPLOTYPE directly, which is what Stringer 2014 reports and stratifies clearance on. Use UGT2B15_MISSING alongside these two when the source paper handles ungenotyped subjects as their own stratum rather than imputing them as wild-type.

Cross-paper caution about which clearances belong to which genotype. Stringer 2014 Table 1 fixes the per-genotype sipoglitazar clearances at 5.04 / 3.35 / 1.53 L/h for *1/*1 / *1/*2 / *2/*2. Those are the mixture-model estimates from supplementary Table A1 of Stringer 2013, where they belong to the latent EM / IM / PM subpopulations, not to the genotype. The genotype-stratified estimates of Stringer 2013 Table 2 are 4.46 / 3.25 / 1.53 L/h. A model that needs “the published sipoglitazar clearance for genotype X” must therefore say which of the two Stringer 2013 models it means; Stringer_2013_sipoglitazar.R carries the Table 2 set and Stringer_2013_sipoglitazar_mixture.R carries the Table A1 set.

UGT2B15_STAR2_HOM (**canonical for UGT2B15*2 homozygote indicator**)

• Description: Binary germline genotype indicator for the UGT2B152/2 homozygous-variant group. 1 = subject carries two UGT2B152 alleles (diplotype 2/2); 0 = otherwise (the union of 1/1 wild-type homozygotes and 1/2 heterozygotes; the paired indicator UGT2B15_STAR2_HET flags the 1/*2 group). Time-fixed per subject. This is the lowest-clearance / highest-exposure UGT2B15 stratum.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (UGT2B151/1 wild-type homozygote), which is realised when UGT2B15_STAR2_HET, UGT2B15_STAR2_HOM and UGT2B15_MISSING are all 0.
• Source aliases:
  • UGT2B15*2/*2 – Stringer 2014 Table 1 column header (UGT2B15*2/*2 (CL = 1.53 L/h)) and supplement Table S1 genotype-count row; Stringer 2013 Table 1 genotype-frequency column and Table 2 row label CL (*2/*2).
• Example models: Stringer_2014_sipoglitazar.R (selects the fixed per-genotype sipoglitazar clearance 1.53 L/h – 3.3-fold below the 1/1 value of 5.04 L/h, matching the ‘3.3-fold higher’ exposure statement in the Stringer 2014 Introduction), Stringer_2013_sipoglitazar.R (selects lcl_s2s2 = log(1.53) L/h, i.e. 66% lower apparent clearance than the *1/*1 stratum; Stringer 2013 Table 2 and Covariate Analysis paragraph).
• Notes: Companion to UGT2B15_STAR2_HET; see that entry’s Notes for the paired-indicator rationale and the distinction from the UGT2B17_EM / UGT2B17_PM phenotype canonicals.

UGT2B15_MISSING (canonical for UGT2B15 genotype-missing indicator)

• Description: Binary indicator for a subject whose UGT2B15 genotype was not measured / not available. 1 = UGT2B15 genotype missing; 0 = UGT2B15 genotype known (in which case UGT2B15_STAR2_HET and UGT2B15_STAR2_HOM together identify the diplotype, and both being 0 means 1/1). When UGT2B15_MISSING == 1 both star-allele indicators must be 0 so the diplotype-indicator products correctly evaluate to zero for the missing-genotype subject.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (UGT2B15 genotype known)
• Source aliases:
  • genotype not collected – Stringer 2014 supplement Table S1 footnote (genotype information not collected in 80 subjects) and Methods ‘Intra-Individual Variability and Residual Error’ (Genotype information was not collected in 10% of the population, however, these subjects were included in the analysis using an average clearance value for the population).
• Example models: Stringer_2014_sipoglitazar.R (selects a population-average sipoglitazar clearance for the 80 ungenotyped subjects rather than imputing them into the 1/1 wild-type stratum).
• Notes: Follows the CYP2C9_MISSING (Lane 2011 warfarin) missing-genotype-stratum precedent, and behind it ADA_MISSING and HEPIMP_MOD_OR_MISSING. Use this canonical only when the source paper explicitly handles the ungenotyped subgroup as its own stratum – if the paper instead imputes missing genotype as wild-type, set the star-allele indicators to the wild-type pattern and do not carry a missing indicator. Note that Stringer 2014 states it used “an average clearance value for the population” for this stratum but does not print the number, so a consuming model must derive it (Stringer 2014 derives 3.21 L/h as the genotype-frequency-weighted mean of the Table 1 clearances) and record the derivation in its vignette.

ABCB1_C1236T_HET (canonical for ABCB1 C1236T heterozygote indicator)

• Description: Binary genotype indicator for the ABCB1 (P-glycoprotein, MDR1) C1236T heterozygote group at rs1128503 (exon 12, synonymous Gly412Gly). 1 = subject carries exactly one variant allele (genotype T/C); 0 = otherwise (the union of C/C homozygous wild-type and T/T homozygous variant strata). Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: Model-dependent, because the two strata this indicator separates are “heterozygous” and “homozygous (either allele)”, not “variant” and “wild-type”. Document the reference explicitly per model. In Liu_2025_avatrombopag.R the reference is ABCB1_C1236T_HET = 1 (the T/C heterozygote stratum, for which the absorption fraction Fa = 1 under fasting); the pooled C/C + T/T homozygote stratum carries the estimated effect. When a source paper instead fits distinct heterozygote and homozygous-variant effects against a C/C wild-type reference, register a paired ABCB1_C1236T_MUT indicator on the ABCB1_C3435T_HET / ABCB1_C3435T_MUT pattern and use the two jointly.
• Source aliases:
  • ABCB1 – Liu 2025 (Table 3 footnote: “If the genotype is heterozygote, ABCB1 is 0; homozygote is 1”). The paper’s column is the exact complement of this canonical, so Liu_2025_avatrombopag.R writes (1 - ABCB1_C1236T_HET) wherever the paper writes ABCB1.
  • ABCB1 (C1236T) (CC/TC/TT) – Liu 2025 Table 1 demographic row; the three-level genotype from which the binary heterozygote indicator is derived.
• Example models: Liu_2025_avatrombopag.R (linear-deviation food-by-genotype co-effect on the absorption fraction: Fa = 1 + e_fed_highfat_fdepot * FED_HIGHFAT + e_abcb1hom_fdepot * (1 - ABCB1_C1236T_HET) + e_fed_abcb1hom_fdepot * FED_HIGHFAT * (1 - ABCB1_C1236T_HET) with e_abcb1hom_fdepot = -0.272 and e_fed_abcb1hom_fdepot = 0.177; fasting C/C-or-T/T homozygotes have 27.2% lower Fa than fasting T/C heterozygotes, and the deficit narrows to 16.0% when dosed after a high-fat meal).
• Notes: Sibling of ABCB1_C3435T_HET / ABCB1_C3435T_MUT (rs1045642, exon 26) and of SNP_ABCB1_RS1045642 / SNP_ABCB1_RS3842; distinct from ABCB1_HAP_TTT, which is the phased cis haplotype across rs1128503 / rs2032582 / rs1045642 jointly rather than the single rs1128503 SNP. Use ABCB1_C1236T_HET only when the source paper models rs1128503 alone. The heterozygote-versus-pooled-homozygote (rather than variant-versus-wild-type) contrast is unusual but is the source paper’s own empirical finding, not an extraction convenience: Liu 2025 Results 2.2.2 reports that the C/C and T/T strata had statistically indistinguishable avatrombopag exposure (Cmax ratio 102.4% fasting / 113.7% fed; AUC0-t ratio 108.7% fasting / 108.4% fed) while both differed materially from T/C, so the paper pooled them. Genotype frequencies in the 92-subject Chinese cohort were 13.0% C/C, 46.7% T/C, 40.2% T/T, consistent with other Asian populations (Liu 2025 Discussion paragraph 4). Mechanistically rs1128503 is a synonymous exon-12 variant whose functional consequences for P-gp expression and efflux are poorly characterised and inconsistent across substrates (Liu 2025 Discussion paragraph 5); the association should be treated as empirical.

ABCB1_C3435T_HET (canonical for ABCB1 C3435T heterozygote indicator)

• Description: Binary genotype indicator for the ABCB1 (P-glycoprotein, MDR1) C3435T heterozygote group at rs1045642 (exon 26, synonymous Ile1145Ile). 1 = subject carries exactly one variant allele (genotype C/T); 0 = otherwise (the union of C/C homozygous wild-type and T/T homozygous variant strata; the paired indicator ABCB1_C3435T_MUT flags the homozygous variant group). Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (C/C homozygous wild-type, when paired with ABCB1_C3435T_MUT = 0). The reference group is the homozygous wild-type C/C stratum; ABCB1_C3435T_MUT flags the homozygous-variant T/T stratum.
• Source aliases:
  • ABCB1 C3435T C/T / C/T – Zhao 2018 (paper Table 2 reports a multiplicative scaling factor of 1.86 on Ka for the heterozygote stratum relative to the C/C reference).
• Example models: Zhao_2018_omeprazole.R (power-of-binary-indicator multiplicative factor on absorption rate constant Ka: e_abcb1_c3435t_het_ka ^ ABCB1_C3435T_HET with e_abcb1_c3435t_het_ka = 1.86; C/T heterozygotes have an absorption rate constant approximately 86% higher than the C/C wild-type reference; paired with ABCB1_C3435T_MUT and used jointly), Toshimoto_2017_irinotecan_pbpk.R (scales MDR1-mediated biliary excretion and enterocyte-to-lumen efflux to 66.6 percent of wild type, weighted by the MDR1 contribution: 50 percent for irinotecan, 33 percent for SN-38; Toshimoto 2017 Supplementary Table 2A).
• Notes: Follows the CYP3A5_STAR1_HET / CYP3A5_STAR1_HOM and SLCO1B1_HAP15_HET / SLCO1B1_HAP15_HOM paired-binary precedent for a three-level genotype where each stratum carries a distinct typical-value covariate effect. Distinct from the ABCB1_HAP_TTT haplotype canonical (which jointly tests the cis combination of rs1128503 / rs2032582 / rs1045642 SNPs as a single haplotype block; used in de Wit 2016 everolimus). Use ABCB1_C3435T_HET + ABCB1_C3435T_MUT when the source paper fits a distinct typical-value covariate effect to the single rs1045642 SNP without phasing it into a haplotype, and when both the heterozygous and homozygous-variant strata are large enough to identify independent effects (Zhao 2018 cohort: n = 22 heterozygotes, 43.1%, and n = 4 homozygous variant, 7.8%, with n = 25 wild-type, 49.0%, as the reference). Mechanistically the C3435T variant has been associated with altered P-gp expression and substrate efflux in some studies (often via linkage with functional variants in the same haplotype block), though directionality of the effect on substrate exposure varies across substrates and tissues.

ABCB1_C3435T_MUT (canonical for ABCB1 C3435T homozygous-variant indicator)

• Description: Binary genotype indicator for the ABCB1 (P-glycoprotein, MDR1) C3435T homozygous-variant group at rs1045642 (exon 26, synonymous Ile1145Ile). 1 = subject carries two variant alleles (genotype T/T); 0 = otherwise (the union of C/C homozygous wild-type and C/T heterozygote strata; the paired indicator ABCB1_C3435T_HET flags the heterozygous group). Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (C/C homozygous wild-type, when paired with ABCB1_C3435T_HET = 0). The reference group is the homozygous wild-type C/C stratum; ABCB1_C3435T_HET flags the heterozygous C/T stratum.
• Source aliases:
  • ABCB1 C3435T T/T / T/T – Zhao 2018 (paper Table 2 reports a multiplicative scaling factor of 6.93 on Ka for the homozygous-variant stratum relative to the C/C reference).
• Example models: Zhao_2018_omeprazole.R (power-of-binary-indicator multiplicative factor on absorption rate constant Ka: e_abcb1_c3435t_mut_ka ^ ABCB1_C3435T_MUT with e_abcb1_c3435t_mut_ka = 6.93; T/T homozygotes have an absorption rate constant approximately 6.93-fold higher than the C/C wild-type reference; paired with ABCB1_C3435T_HET and used jointly), Toshimoto_2017_irinotecan_pbpk.R (same processes scaled to 33.3 percent of wild type; Toshimoto 2017 Supplementary Table 2A).
• Notes: Companion to ABCB1_C3435T_HET. See ABCB1_C3435T_HET Notes for the three-level decomposition rationale, the distinction from the ABCB1_HAP_TTT haplotype canonical, and the Zhao 2018 cohort distribution.

ABCB1_HAP_TTT (canonical for ABCB1 TTT haplotype carrier indicator)

• Description: Binary haplotype indicator for the ABCB1 TTT haplotype across the rs1128503 (1236C>T, exon 12, synonymous Gly412Gly) / rs2032582 (2677G>T/A, exon 21, Ala893Ser/Thr) / rs1045642 (3435C>T, exon 26, synonymous Ile1145Ile) SNP block. 1 = subject carries at least one TTT haplotype (heterozygous or homozygous; pooled because the homozygote frequency was < 0.1 in the de Wit 2016 cohort); 0 = no TTT haplotype. Time-fixed per subject (germline haplotype).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no TTT haplotype).
• Source aliases:
  • ABCB1 TTT haplotype – de Wit 2016 (paper text Methods ‘Pharmacogenetic analysis’ and Table 2 final-model theta TTT on F row; haplotypes phased in gPLINK with certainty > 0.97).
• Example models: deWit_2016_everolimus.R (multiplicative effect on apparent bioavailability F: F = 1 * 0.792^ABCB1_HAP_TTT – carriers have 20.8% lower F than non-carriers; dOFV = 9.6 in backward elimination, P < 0.01).
• Notes: The TTT haplotype of ABCB1 (P-glycoprotein, MDR1 efflux transporter) is associated with enhanced P-gp efflux activity and reduced everolimus bioavailability (de Wit 2016 Discussion paragraph 5); de Wit cites prior evidence that the same TTT haplotype also reduces exposure / efficacy of other P-gp substrates [refs 20-22 in the paper], although directionally inconsistent results have been reported [refs 7, 23, 24]. Het and hom carriers were pooled in de Wit 2016 because the homozygote frequency was < 0.1 (Methods ‘Pharmacogenetic analysis’); future extractions that estimate separate het / hom effects should register paired ABCB1_HAP_TTT_HET and ABCB1_HAP_TTT_HOM indicators following the SLCO1B1_HAP15_HET / SLCO1B1_HAP15_HOM precedent above. Distinct from any individual ABCB1 SNP indicator (rs1128503, rs2032582, rs1045642 alone) because the haplotype is the cis combination tested jointly.

ALDH2_S2_CARRIER (**canonical for ALDH2*2 inactive-variant carrier indicator**)

• Description: Binary genotype indicator for the ALDH2 *2 (inactive) variant allele (rs671 G>A; Glu487Lys, aldehyde dehydrogenase 2 mitochondrial isoform). 1 = subject carries at least one ALDH22 allele (heterozygous 1/2 or homozygous 2/2); 0 = ALDH21/*1 wild-type. Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (ALDH21/1 wild-type, fully active aldehyde dehydrogenase 2).
• Source aliases:
  • ALDH2 (the source NONMEM indicator used in Nemoto_2017_ethanol.R; 1 for 1/2 carriers, 0 for 1/1 wild-type; the Nemoto 2017 cohort had no 2/2 homozygotes so the indicator is effectively heterozygote-vs-wild-type in that cohort).
• Example models: Nemoto_2017_ethanol.R (additive shift on Vd/F: -20.4 L when ALDH2_S2_CARRIER = 1 vs ALDH21/1 reference; Nemoto 2017 Table II final model).
• Notes: ALDH22 is the canonical East-Asian aldehyde-dehydrogenase deficiency allele (rs671 GAA -> AAA, Glu487Lys); homozygous 2/2 subjects are essentially ALDH2-null and experience severe acetaldehyde-flush after even small ethanol doses, so are typically excluded from alcohol-PK studies. The Nemoto 2017 Japanese cohort contained 21/34 (62%) 1/1 and 13/34 (38%) 1/2 subjects; no 2/2 homozygotes were enrolled. The single binary carrier indicator pools heterozygous and homozygous variant carriers; future ethanol-PK extractions that distinguish 1/2 from 2/*2 effects should register paired ALDH2_S2_HET and ALDH2_S2_HOM indicators following the SLCO1B1_HAP15_HET / SLCO1B1_HAP15_HOM precedent.

ADH1B_S2_HOM (**canonical for ADH1B*2 homozygote indicator**)

• Description: Binary genotype indicator for the ADH1B *2 (high-activity) variant allele (rs1229984 G>A; Arg47His, alcohol dehydrogenase 1B class I beta-subunit). 1 = subject is homozygous 2/2; 0 = otherwise (the default reference covers ADH1B2/1 heterozygotes; ADH1B1/1 wild-type subjects fall outside the published parameterization of Nemoto 2017 and are conventionally assigned the same reference value as heterozygotes). Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (ADH1B2/1 heterozygous; ADH1B1/1 wild-type by extension when the model is applied beyond the published parameter range).
• Source aliases:
  • ADH1B (the source NONMEM indicator used in Nemoto_2017_ethanol.R; 1 for 2/2 homozygotes, 0 for 2/1 heterozygotes; Nemoto 2017 Table II structural model encodes the Vmax conditional via two separate THETAs).
• Example models: Nemoto_2017_ethanol.R (additive shift on Vmax: +176 mg/h for 2/2 homozygotes relative to the 2/1 reference Vmax of 7790 mg/h; Nemoto 2017 Table II final model).
• Notes: ADH1B2 is the East-Asian high-activity alcohol-dehydrogenase variant (rs1229984 Arg47His); the 2 allele encodes a ~40x faster ethanol oxidation rate than the 1 wild-type. Allele frequency in Japanese populations is ~70%, so most subjects are 2/2 or 1/2 with a small 1/1 minority (~9% expected). Nemoto 2017 inherits the two-Vmax parameterization from Seng et al. 2014 (which estimated Vmax separately for 2/1 and 2/2 in a Chinese + Indian cohort) and does not report an ADH1B genotype distribution for its 34-subject Japanese cohort; the structural model is silent on ADH1B1/*1 subjects. Future ADH1B-aware ethanol-PK extractions that explicitly model all three genotypes should register ADH1B_S2_HET as a companion indicator following the SLCO1B1_HAP15_HET / SLCO1B1_HAP15_HOM precedent.

PFMDR1_86Y (canonical for P. falciparum pfmdr1 codon-86 tyrosine mutant indicator)

• Description: Plasmodium falciparum pfmdr1 codon-86 tyrosine mutant indicator (1 = single-copy pfmdr1 with the 86Y mutation, Simpson 2013 Genotype 2; 0 = otherwise). Time-fixed per parasite isolate. This is a parasite-genome indicator (not host pharmacogenetics): the “subject” in the NLME framework is a clinical P. falciparum isolate, not a human patient.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 = Simpson 2013 Genotype 1 single-copy wild-type 86N/1042N (when all four PFMDR1 indicators are 0). The four PFMDR1 indicators are mutually exclusive in the Simpson 2013 cohort (Thai pfmdr1 point mutations occur almost exclusively on single-copy parasites; amplifications occur exclusively on WT 86N/1042N parasites).
• Source aliases:
  • X1 – used in Simpson_2013_artesunate.R, Simpson_2013_chloroquine.R, Simpson_2013_lumefantrine.R, Simpson_2013_mefloquine.R (per-isolate Genotype-2 indicator).
• Example models: Simpson_2013_artesunate.R, Simpson_2013_chloroquine.R, Simpson_2013_lumefantrine.R, Simpson_2013_mefloquine.R (multiplicative proportional effect on the sigmoid-Emax EC50, 1 + e_pfmdr1_86y_ec50 * PFMDR1_86Y, drug-specific magnitude per Simpson 2013 Table 3 Genotype-2 percent-change column).
• Notes: Specific scope because the canonical name is a parasite-genome covariate tied to the Simpson 2013 in-vitro antimalarial study design. Explicitly DISTINCT from the human host-pharmacogenetics ABCB1_* canonicals (ABCB1 / MDR1 is the human P-glycoprotein efflux-transporter gene; pfmdr1 is the orthologous P. falciparum multidrug-resistance transporter). Member of the four-member mutually-exclusive PFMDR1_* Simpson 2013 genotype set (PFMDR1_86Y, PFMDR1_1042D, PFMDR1_CN2, PFMDR1_CN3PLUS), with single-copy WT 86N/1042N as the all-zero reference. Future pfmdr1-genotype antimalarial extractions should reuse this set or register sibling parasite-genome canonicals.

PFMDR1_1042D (canonical for P. falciparum pfmdr1 codon-1042 aspartate mutant indicator)

• Description: Plasmodium falciparum pfmdr1 codon-1042 aspartate mutant indicator (1 = single-copy pfmdr1 with the 1042D mutation, Simpson 2013 Genotype 3; 0 = otherwise). Time-fixed per parasite isolate. Parasite-genome indicator (not host pharmacogenetics).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 = Simpson 2013 Genotype 1 single-copy wild-type 86N/1042N (when all four PFMDR1 indicators are 0). Mutually exclusive with the other three PFMDR1 indicators.
• Source aliases:
  • X2 – used in Simpson_2013_artesunate.R, Simpson_2013_chloroquine.R, Simpson_2013_lumefantrine.R, Simpson_2013_mefloquine.R (per-isolate Genotype-3 indicator).
• Example models: Simpson_2013_artesunate.R, Simpson_2013_chloroquine.R, Simpson_2013_lumefantrine.R, Simpson_2013_mefloquine.R (multiplicative proportional effect on the sigmoid-Emax EC50, 1 + e_pfmdr1_1042d_ec50 * PFMDR1_1042D, drug-specific magnitude per Simpson 2013 Table 3 Genotype-3 percent-change column).
• Notes: Specific scope. Distinct from the human ABCB1_* canonicals (see PFMDR1_86Y notes). Member of the four-member mutually-exclusive PFMDR1_* Simpson 2013 genotype set with single-copy WT as the all-zero reference.

PFMDR1_CN2 (canonical for P. falciparum pfmdr1 double-copy amplification indicator)

• Description: Plasmodium falciparum pfmdr1 double-copy amplification indicator (1 = two copies of pfmdr1, all wild-type 86N/1042N, Simpson 2013 Genotype 4; 0 = otherwise). Time-fixed per parasite isolate. Parasite-genome indicator (not host pharmacogenetics).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 = Simpson 2013 Genotype 1 single-copy wild-type 86N/1042N (when all four PFMDR1 indicators are 0). Mutually exclusive with the other three PFMDR1 indicators.
• Source aliases:
  • X3 – used in Simpson_2013_artesunate.R, Simpson_2013_chloroquine.R, Simpson_2013_lumefantrine.R, Simpson_2013_mefloquine.R (per-isolate Genotype-4 indicator).
• Example models: Simpson_2013_artesunate.R, Simpson_2013_chloroquine.R, Simpson_2013_lumefantrine.R, Simpson_2013_mefloquine.R (multiplicative proportional effect on the sigmoid-Emax EC50, 1 + e_pfmdr1_cn2_ec50 * PFMDR1_CN2, drug-specific magnitude per Simpson 2013 Table 3 Genotype-4 percent-change column).
• Notes: Specific scope. Distinct from the human ABCB1_* canonicals (see PFMDR1_86Y notes). Member of the four-member mutually-exclusive PFMDR1_* Simpson 2013 genotype set with single-copy WT as the all-zero reference; PFMDR1_CN2 (two copies) and PFMDR1_CN3PLUS (three-or-more copies) jointly encode the copy-number amplification axis.

PFMDR1_CN3PLUS (canonical for P. falciparum pfmdr1 triple-or-more-copy amplification indicator)

• Description: Plasmodium falciparum pfmdr1 triple-or-more-copy amplification indicator (1 = three or more copies of pfmdr1, all wild-type 86N/1042N, Simpson 2013 Genotype 5; 0 = otherwise). Time-fixed per parasite isolate. Parasite-genome indicator (not host pharmacogenetics).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 = Simpson 2013 Genotype 1 single-copy wild-type 86N/1042N (when all four PFMDR1 indicators are 0). Mutually exclusive with the other three PFMDR1 indicators.
• Source aliases:
  • X4 – used in Simpson_2013_artesunate.R, Simpson_2013_chloroquine.R, Simpson_2013_lumefantrine.R, Simpson_2013_mefloquine.R (per-isolate Genotype-5 indicator).
• Example models: Simpson_2013_artesunate.R, Simpson_2013_chloroquine.R, Simpson_2013_lumefantrine.R, Simpson_2013_mefloquine.R (multiplicative proportional effect on the sigmoid-Emax EC50, 1 + e_pfmdr1_cn3plus_ec50 * PFMDR1_CN3PLUS, drug-specific magnitude per Simpson 2013 Table 3 Genotype-5 percent-change column; for artesunate and mefloquine the triple+ copy group carries the largest EC50 shift, +127% and +188% respectively).
• Notes: Specific scope. Distinct from the human ABCB1_* canonicals (see PFMDR1_86Y notes). Member of the four-member mutually-exclusive PFMDR1_* Simpson 2013 genotype set with single-copy WT as the all-zero reference; paired with PFMDR1_CN2 on the copy-number amplification axis.

CYP2C19 (canonical for CYP2C19 individual metabolic-activity proxy)

• Description: Continuous individual-level CYP2C19 metabolic-activity proxy. The intent is a single canonical column for any CYP2C19 phenotype proxy that the source paper reports as a continuous number (probe-substrate model-based individual clearance, urinary or serum S/R-mephenytoin ratio, copy-number-corrected expression score, activity-score sum from *allele genotypes, etc.); the per-model covariateData[[CYP2C19]]$units, description, and notes document which proxy is in force, the orientation (higher value = higher or lower activity, which depends on the probe), and the population-mean reference value used inside the model. Time-invariant in all known examples (germline-genotype-determined enzyme expression measured once via probe-substrate assay).
• Units: Paper-specific – document per-model (e.g., unitless urinary S/R-mephenytoin ratio in Areberg 2006).
• Type: continuous
• Scope: general
• Reference category: n/a (continuous). Models center on a population mean (e.g., 0.769 in Areberg 2006); document the reference value per-model in covariateData[[CYP2C19]]$notes.
• Source aliases:
  • CYP2C19 – used directly in Areberg_2006_escitalopram.R (Areberg 2006 Equation 2 variable name; the value is the urinary S/R-mephenytoin ratio defined in the Materials and Methods ‘CYP2C19 Phenotyping’ section).
• Example models: Areberg_2006_escitalopram.R (linear-deviation effect on apparent oral clearance: CL/F = 19 - 17 * (CYP2C19 - 0.769); CYP2C19 = urinary S/R-mephenytoin ratio, population mean = 0.769; one unit of additional ratio = 17 L/h lower clearance, consistent with the biology that higher S/R ratio = lower CYP2C19 activity = lower escitalopram metabolic clearance; Areberg 2006 Equations 1-2 and Table 3).
• Notes: Companion to the continuous CYP2D6 (line 4055) and CYP3A4 (line 4089) canonicals, ratified here per the prospective registration TODO on the CYP3A4 line. Orientation depends on the probe. Two patterns coexist in the CYP2C19-aware popPK literature:
  • Probe-derived clearance form (analogous to the dextromethorphan-probe CYP2D6 / CYP3A4 columns): higher value = higher CYP2C19 activity. Use this orientation when the source reports a probe-substrate-derived individual clearance number.
  • Probe-derived metabolic ratio form (e.g., urinary or serum S/R-mephenytoin ratio): higher value = lower CYP2C19 activity, because high CYP2C19 activity selectively metabolizes S-mephenytoin and lowers the residual S enantiomer. Per-model covariateData[[CYP2C19]]$notes must document which orientation is in force alongside the population-mean reference value and the column’s units, mirroring the per-model documentation already used for the CYP2D6 and CYP3A4 continuous canonicals. Distinct from the binary genotype-derived canonicals CYP2C19_S2_CARRIER (loss-of-function *2 allele carrier), CYP2C19_IM (intermediate-metabolizer phenotype), and CYP2C19_PM (poor-metabolizer phenotype); use the binary indicators when the source paper reports only a discrete phenotype label, and use CYP2C19 (continuous) when the source reports a probe-derived continuous number.

CYP2C19_NON_EM (canonical for composite CYP2C19 non-homozygous-extensive-metabolizer indicator)

• Description: 1 = subject is genotyped as a CYP2C19 poor metabolizer (homozygous loss-of-function, e.g. 2/2, 2/3, 3/3) OR a heterozygous-extensive metabolizer (one functional allele and one loss-of-function allele, e.g. 1/2, 1/3); 0 = subject is genotyped as a homozygous extensive metabolizer (both alleles functional, 1/1). Distinct from CYP2C19_PM (strict homozygous-PM only) and from CYP2C19_IM (intermediate-metabolizer-only); CYP2C19_NON_EM is a paper-defined composite of PM + IM that the source paper used as a single multiplicative covariate on clearance.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (homozygous extensive metabolizer; both *1 alleles).
• Source aliases:
  • PM (Wahlby 2004 / Walsh 2004 voriconazole source-column convention) – used in Wahlby_2004_voriconazole.R. Note that the column name PM in this paper denotes the composite non-EM group, not the strict CYP2C19_PM phenotype; users converting from the source paper’s data should not collapse PM to canonical CYP2C19_PM without re-checking the genotype-to-indicator mapping.
  • CYP2C19 (Guo 2020 valproic acid, Figure 4 legend) – coded 1 = *1/*1, 2 = *1/*2, *1/*3, *2/*2, *2/*3 or *3/*3; CYP2C19_NON_EM = CYP2C19 - 1. *17 was genotyped but not used in the grouping.
• Example models: Wahlby_2004_voriconazole.R (multiplicative effect on CL: (1 - 0.46 * CYP2C19_NON_EM), so a non-EM subject has 46 percent lower CL than a homozygous-EM subject), Guo_2020_valproic_acid.R (exponential effect on CL: exp(-0.45 * CYP2C19_NON_EM)).
• Notes: Specific scope because the PM+IM composite grouping is paper-defined; future papers that report PM and IM separately should use the strict CYP2C19_PM and CYP2C19_IM canonicals instead. The Wahlby 2004 / Walsh 2004 voriconazole encoding has been retained because the source paper does not provide separate per-genotype coefficient estimates.

SNP_ABCB1_RS3842 (canonical for ABCB1 rs3842 (c.4036A>G, 3’ UTR) mutant allele carrier indicator)

• Description: Binary genotype indicator for the ABCB1 rs3842 single-nucleotide polymorphism (c.4036A>G; in the 3’ untranslated region of the ABCB1 / MDR1 / P-glycoprotein gene). 1 = subject carries at least one G (mutant) allele (heterozygous AG or homozygous GG; pooled because the source paper reports the same effect for heterozygotes and homozygotes); 0 = homozygous wild-type (AA). Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (homozygous wild-type AA).
• Source aliases:
  • ABCB1 (rs3842) – Mukonzo 2009 (paper Table 1 ‘ABCB1’ section: position c.4036 A/G, rs number rs3842, allele New SNP, protein 3' UTR, observed SNP frequency 16.8%; Table 3 final-model ‘Effect of ABCB1 (rs 3842)’ row).
• Example models: Mukonzo_2009_efavirenz.R (linear-deviation effect on relative bioavailability Frel: f(depot) = (1 + e_rs3842_fdepot * SNP_ABCB1_RS3842) * exp(etalfdepot) with e_rs3842_fdepot = 0.257; mutant carriers (any G allele) have 25.7% higher Frel than AA wild-type carriers; Mukonzo 2009 Table 3 and Results paragraph ‘Pharmacokinetic modelling’).
• Notes: Distinct from SNP_ABCB1_RS1045642 (a different SNP at c.3435C>T in exon 26 that affects P-glycoprotein expression / function) and from ABCB1_HAP_TTT (the multi-SNP haplotype across rs1128503 / rs2032582 / rs1045642 jointly); use SNP_ABCB1_RS3842 only when the source paper genotypes the c.4036A>G variant in the 3’ UTR. Mukonzo 2009 selected rs3842 from a panel of 13 ABCB1 SNPs as a “previously uncharacterised polymorphism” predicted to have potential regulatory function by bioinformatics tools (Mukonzo 2009 Table 1 ‘Relevance’ = ‘Undetermined’ at extraction time); the modeled +25.7% effect on relative bioavailability was the strongest single-SNP signal among the 13 ABCB1 candidates in the Mukonzo 2009 backward-elimination step. Heterozygote and homozygote G-carriers are pooled in Mukonzo 2009 because the paper explicitly groups them (Results: “Mutant homozygote and heterozygote individuals for ABCB1 rs3842 exhibited 26% greater efavirenz bioavailability than wild-type carriers”); future extractions that estimate separate het / hom effects should register paired SNP_ABCB1_RS3842_HET and SNP_ABCB1_RS3842_HOM indicators following the SLCO1B1_HAP15_HET / SLCO1B1_HAP15_HOM precedent. The Mukonzo 2009 Discussion notes that the rs3842 minor-allele frequency is >20% in Asians, Europeans, and Sub-Saharan Africans according to dbSNP, so the canonical may have broad applicability once additional papers ratify the same encoding.

SNP_ABCB1_RS3842_A_CARRIER (canonical for ABCB1 rs3842 (c.4036A>G, 3’ UTR) A-allele carrier indicator)

• Description: Binary genotype indicator for the ABCB1 rs3842 single-nucleotide polymorphism (c.4036A>G; 3’ untranslated region of the ABCB1 / MDR1 / P-glycoprotein gene), pooled in the A-allele-carrier direction. 1 = subject carries at least one A allele (heterozygous G/A or homozygous A/A); 0 = homozygous G/G. Time-fixed per subject (germline genotype). This is the mirror-image pooling of SNP_ABCB1_RS3842, which is a G-allele-carrier indicator with A/A as its reference category; register both because neither column can be derived from the other – the heterozygote falls in the 1 group under both poolings, so the mapping G/G, G/A, A/A -> 1, 1, 0 (SNP_ABCB1_RS3842) and -> 0, 1, 1 (this column) is not a sign flip or a value inversion. Use whichever column matches the contrast the source paper actually estimated.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (homozygous G/G).
• Source aliases:
  • ABRS3842 – Chala 2023 (paper Table 1 row ABCB1c.4036A > G(rs3842) with levels G/G and G/A or A/A; Table 2 row ‘Fold of typical CL for subjects with ABCB1.rs3842 G/A or A/A genotype’; supplement Appendix S1 control stream IF(ABRS3842.EQ.1) ABCB1RS3842 = 1).
• Example models: Chala_2023_efavirenz.R (multiplicative fold-factor on apparent oral clearance raised to the 0/1 indicator power, ... * e_snp_abcb1_rs3842_a_cl^SNP_ABCB1_RS3842_A_CARRIER with e_snp_abcb1_rs3842_a_cl = 1.452; A-allele carriers have 45% higher efavirenz CL/F than G/G homozygotes, Chala 2023 Table 2 / Equation 1 factor 1.45(f3)).
• Notes: The two rs3842 columns are directionally consistent about the biology even though they pool opposite ways. Mukonzo 2009 (Ugandan adults, SNP_ABCB1_RS3842) found G-allele carriers had 25.7% higher relative bioavailability, i.e. higher efavirenz exposure; Chala 2023 (Ethiopian children, this column) found G/G homozygotes had the lower apparent clearance, i.e. also higher exposure. Both therefore associate the G allele with higher exposure. The cohort genotype distributions differ markedly between the two papers – Mukonzo 2009 reports a 16.8% observed rs3842 SNP frequency, while Chala 2023 Table 1 reports G/G in 17% and G/A-or-A/A in 80% of subjects – so do not assume a shared reference allele when transcribing a third rs3842 paper; read the paper’s own genotype-level table and pick the column whose reference category matches. Distinct from SNP_ABCB1_RS1045642 (c.3435C>T in exon 26) and ABCB1_HAP_TTT (multi-SNP haplotype).

SNP_SLC22A1_RS683369 (canonical for SLC22A1 (OCT1) rs683369 c.480C>G L160F variant carrier indicator)

• Description: Binary germline-genotype indicator for the SLC22A1 (OCT1, organic cation transporter 1) c.480C>G polymorphism, dbSNP rs683369, encoding the p.Leu160Phe substitution. 1 = carrier of at least one G (variant) allele, i.e. genotype CG or GG; 0 = CC wild-type homozygote. Time-fixed per subject. OCT1 is the principal hepatic uptake transporter for imatinib and several other cationic drugs, so a reduced-function variant lowers hepatocellular uptake and therefore apparent oral clearance.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (CC wild-type homozygote), whose typical-value multiplier is 1 by construction.
• Source aliases:
  • hOCT1 (Di Paolo 2014, which names the gene by its protein alias in the covariate column and titles the paper ‘The c.480C>G polymorphism of hOCT1 influences imatinib clearance’).
  • theta_SLC22A1 (Yang 2025 Table 1 symbol for the multiplicative coefficient, resolved by Table 1 footnote f).
  • rs683369 / c.480C>G / L160F – interchangeable identifiers for the same variant.
• Example models: DiPaolo_2014_imatinib.R (dominant genetic model: CL/F is multiplied by 0.882 for CG and GG carriers alike and by 1 for CC, per Yang 2025 Table 1 footnote f, ‘When SLC22A1 CG or GG theta_SLC22A1 = 0.882, else theta_SLC22A1 = 1 when SLC22A1 CC’).
• Notes: Encoded as a single CARRIER indicator rather than the paired _HET / _HOM indicators used by SNP_ABCG2_RS2231142_HET / _HOM in Jiang_2023_imatinib.R, because the founding model fits a DOMINANT genetic model: heterozygotes and variant homozygotes share one coefficient, so two indicators would be unidentifiable. A future paper that resolves CG and GG separately should add paired SNP_SLC22A1_RS683369_HET / _HOM entries rather than redefining this one. Shape follows the existing SNP_SLC22A2_808GT carrier-indicator precedent. Note that Jiang_2023_imatinib.R records SNP_SLC22A1_RS683369 under covariatesDataExcluded (screened, not retained) using the paper’s own three-level genotype description; that documentation-only entry predates this canonical and does not conflict with it.

SNP_SLC22A2_808GT (canonical for SLC22A2 (OCT2) c.808G>T A270S variant carrier indicator)

• Description: Binary genotype indicator for the SLC22A2 (OCT2) c.808G>T single-nucleotide polymorphism (protein change A270S / p.Ala270Ser; widely reported in the pharmacogenomic literature as rs316019). 1 = at least one variant (T) allele present (heterozygous 808GT or homozygous 808TT carrier); 0 = homozygous wild-type 808GG. Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (homozygous wild-type 808GG).
• Source aliases:
  • OCT2-808 G>T – Yoon 2013 (paper Table I ‘OCT2 808 G>T’ row: 96 subjects genotyped as 81 GG + 15 GT + 0 TT, allele frequency 7.8%; Methods ‘Statistical Analysis for the Influence of Genetic Polymorphisms’ pooled heterozygotes and homozygous variants because the Korean cohort contained no homozygous variants).
• Example models: Yoon_2013_metformin.R (multiplicative fractional shift on CL/F: cl = exp(lcl + etalcl) * (1 + e_snp_oct2_cl * SNP_SLC22A2_808GT) * (1 + e_snp_octn1_cl * SNP_SLC22A4_917CT) with e_snp_oct2_cl = -0.248; 808G>T variant carriers have 24.8% lower apparent oral clearance than 808GG wild-type at the OCTN1 reference; Yoon 2013 Table III theta_OCT2 with RSE 15.8%).
• Notes: Position-based canonical name because the source paper reports the SNP by nucleotide position and protein change without an accompanying rsID; the SLC22A2 A270S variant is nevertheless documented across the pharmacogenomic literature as rs316019 and downstream extractions that report the rsID directly may register a paired canonical SNP_SLC22A2_RS316019 with an alias link back to this entry. OCT2 (organic cation transporter 2) is the basolateral renal-tubule uptake transporter for organic cations, including metformin; the A270S substitution reduces cellular metformin uptake in HEK293 in-vitro transport assays. In the Yoon 2013 Korean-male healthy-volunteer cohort the 808G>T variant allele frequency was 7.8% (Table I), comparable to previously reported frequencies of 16.8% in Japanese-Asian and 16% in European-American populations (Yoon 2013 Table I ‘Asian’ and ‘European’ columns).

SNP_SLC22A4_917CT (canonical for SLC22A4 (OCTN1) c.917C>T T306I variant carrier indicator)

• Description: Binary genotype indicator for the SLC22A4 (OCTN1) c.917C>T single-nucleotide polymorphism (protein change T306I / p.Thr306Ile). 1 = at least one variant (T) allele present (heterozygous 917CT or homozygous 917TT carrier); 0 = homozygous wild-type 917CC. Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (homozygous wild-type 917CC).
• Source aliases:
  • OCTN1-917C>T – Yoon 2013 (paper Table I ‘OCTN1 917C>T’ row: 96 subjects genotyped as 10 CC + 52 CT + 34 TT, allele frequency 62.5%; Methods ‘Statistical Analysis for the Influence of Genetic Polymorphisms’ pooled heterozygotes and homozygous variants).
• Example models: Yoon_2013_metformin.R (multiplicative fractional shift on CL/F: cl = exp(lcl + etalcl) * (1 + e_snp_oct2_cl * SNP_SLC22A2_808GT) * (1 + e_snp_octn1_cl * SNP_SLC22A4_917CT) with e_snp_octn1_cl = -0.234; 917C>T variant carriers have 23.4% lower apparent oral clearance than 917CC wild-type at the OCT2 reference; Yoon 2013 Table III theta_OCTN1 with RSE 61.1%).
• Notes: Position-based canonical name because the source paper reports the SNP by nucleotide position and protein change without an accompanying rsID. Distinct from the well-characterised SLC22A4 c.1507C>T (L503F, rs1050152) OCTN1 variant which was also genotyped by Yoon 2013 but occurred with zero variant-allele frequency in the Korean cohort (Table I) and was not carried forward as a model covariate. OCTN1 (organic cation transporter, novel, type 1) is expressed at the apical membrane of renal proximal-tubule cells and contributes to the renal secretion of metformin; the T306I substitution is a candidate function-altering variant whose direct in-vitro effect on metformin transport has not been characterised (Yoon 2013 Discussion notes that in-vitro confirmation is still needed). Yoon 2013 Discussion further notes that the 917C>T variant-allele carrier rate is ~90% in Koreans, ~60% in other Asians, and ~33% in Caucasians, so the impact of OCTN1 917C>T on metformin PK is expected to be larger in Asian populations.

SNP_SOD2_RS4880_TT (canonical for SOD2 Val16Ala (rs4880) homozygous wild-type Val/Val indicator)

• Description: Binary genotype indicator for the SOD2 (manganese superoxide dismutase, MnSOD) rs4880 single-nucleotide polymorphism in the mitochondrial targeting sequence (c.47T>C; p.Val16Ala; also written Ala16Val or -9T>C in the older literature). 1 = homozygous wild-type Val/Val (genotype T/T on the reference orientation); 0 = otherwise (the union of Val/Ala heterozygotes and Ala/Ala homozygous-variant carriers). Time-fixed per subject (germline genotype). Note the inverted polarity relative to most SNP_* entries in this register: the indicator flags the wild-type homozygote, not the variant carrier, because the founding paper found the risk to be carried by the Val/Val stratum.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (Val/Ala heterozygotes pooled with Ala/Ala homozygous-variant carriers). The pooling is a source-paper decision driven by cell counts, not a general genetic-model claim: Ogusu 2014 had only 2 Ala/Ala subjects among the 169 in the PK-PD analysis (1.7%), too few to estimate a separate stratum, so Ala/Ala and Val/Ala were combined (Ogusu 2014 Results, “Patient Demographics”).
• Source aliases:
  • SOD2 Val/Val genotype – used in Ogusu_2014_valproic_acid.R (Ogusu 2014 Eq. 9 superscript SOD2 Val/Val genotype, defined in the text below the equation as “SOD2 Val/Val genotype = 1, SOD2 Val/Ala or Ala/Ala genotype = 0”; Table 3 row “SOD2 genotype on BASE”).
• Example models: Ogusu_2014_valproic_acid.R (additive shift on the intercept of a logistic regression for the probability of gamma-glutamyltransferase elevation: logit_ggt_elevation <- logit_ref + ... + e_sod2_vv_logit * SNP_SOD2_RS4880_TT + ... with e_sod2_vv_logit = 1.96 log-odds; Ogusu 2014 Table 3, RSE 44.1%, bootstrap 95% CI 0.33-4.42).
• Notes: Strand and nomenclature. rs4880 is c.47T>C in the reference orientation, so the T allele codes valine (Val, wild-type) and the C allele codes alanine (Ala, variant); papers variously label the polymorphism Val16Ala, Ala16Val, or -9T>C, and some report it on the opposite strand as A>G. Always map the paper’s genotype labels onto the Val / Ala amino-acid strata before assigning indicator values – the letter in this canonical’s _TT suffix refers to the reference-orientation nucleotide genotype (T/T = Val/Val), consistent with the SNP_ABCG2_RS2231142_HOM entry’s strand-orientation guidance. Biology. SOD2 is the mitochondrial first-line defence against superoxide; the Ala variant is imported into mitochondria more efficiently and yields higher enzyme activity, which is the mechanistic rationale for the Val/Val stratum carrying the higher oxidative-stress risk in Ogusu 2014’s valproic-acid hepatotoxicity setting. Encoding orientation. Because the polarity is inverted relative to the usual variant-carrier convention, do NOT reuse this canonical for a paper that codes Ala-carriers = 1; register a paired companion (e.g. SNP_SOD2_RS4880_ALA_CARRIER) instead, following the SNP_ABCG2_RS2231142_HOM / _HET / hypothetical _CARRIER precedent, so the two genetic-model orientations remain separate canonicals. Scope is specific because the 0-level pools two genotype strata for a paper-specific cell-count reason; promote to general when a second independent paper adopts the same Val/Val-versus-rest dichotomy. Ratified 2026-09-02 (PMC9833552, after a register search) alongside the Ogusu 2014 valproic acid extraction.

SNP_CYP2B6_RS35303484_G_COUNT (canonical for CYP2B6 136A>G (rs35303484) G-allele count)

• Description: Continuous individual-level CYP2B6 c.136A>G (rs35303484, p.M46V) G-allele count: 0 = AA homozygous wild-type, 1 = AG heterozygous, 2 = GG homozygous variant. Time-invariant (germline genotype). The 136A>G variant defines the CYP2B6*11 haplotype (a phenotypic null allele in the CYP allele nomenclature) and is most commonly observed in Sub-Saharan African populations.
• Units: (count, 0/1/2 alleles per subject)
• Type: continuous
• Scope: specific
• Reference category: n/a (continuous). In the Mukonzo 2009 efavirenz model the count is decomposed in model into a homozygous-mutant-only indicator (count == 2) multiplied by a -0.199 multiplicative shift on CL/F. Heterozygotes are pooled with wild-type and receive no shift, consistent with the source paper’s reported effect for “homozygous CYP2B6*11” only (Mukonzo 2009 Results paragraph ‘Pharmacokinetic modelling’).
• Source aliases:
  • CYP2B6*11 – Mukonzo 2009 (paper Table 1 SNP-table column ‘Allele’ and Table 3 ’Effect of CYP2B6*11’ row; rs35303484 is the defining SNP at c.136 A->G with p.M46V protein change).
• Example models: Mukonzo_2009_efavirenz.R (homozygous-mutant-only multiplicative shift on CL/F: cl = exp(lcl + etalcl) * (1 + e_2b6_6_cl * (SNP_CYP2B6_RS3745274_T_COUNT == 2)) * (1 + e_2b6_11_cl * (SNP_CYP2B6_RS35303484_G_COUNT == 2)) with e_2b6_11_cl = -0.199; homozygous *11 carriers have 19.9% lower CL/F than the wild-type / heterozygous reference; Mukonzo 2009 Table 3).
• Notes: Distribution in the Mukonzo 2009 Ugandan healthy-volunteer cohort (Table 1, n = 121): 136A>G allele frequency 13.6%, expected genotype frequencies under Hardy-Weinberg AA 74.6%, AG 23.5%, GG 1.9%. CYP2B611 is found primarily in African populations (allele frequency ~7-15% in Sub-Saharan Africans, ~0-1% in Europeans and East Asians); the M46V substitution lies in the substrate-binding pocket and abolishes detectable enzyme activity in recombinant expression studies. Often co-tested with the more common 516G>T (CYP2B66) variant; in the Mukonzo 2009 Ugandan cohort the two variants segregated independently. The count-form encoding follows the same rationale as SNP_CYP2B6_RS3745274_T_COUNT and SNP_CYP2B6_RS28399499_C_COUNT: a single count column captures the underlying genotype without redundancy and lets the model code derive either a linear per-allele effect (* count) or a non-additive HET / HOM decomposition (* (count == 1), * (count == 2)) depending on the source paper’s parameterization.

UGT2B15_IM (canonical for UGT2B15 intermediate-metabolizer phenotype indicator)

• Description: 1 = subject is assigned to the UGT2B15 intermediate-metabolizer (IM) subpopulation; 0 = any other UGT2B15 phenotype (extensive or poor). Time-fixed per subject. Paired with UGT2B15_PM to encode the three-level EM (both indicators 0, reference) / IM / PM phenotype with two binary indicators.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 with UGT2B15_PM = 0 (the UGT2B15 extensive-metabolizer stratum).
• Source aliases:
  • POP2 (CL2 IM) – Stringer 2013 supplementary mixture-model analysis, where the latent subpopulation assignment is estimated from apparent clearance alone via NONMEM $MIX without using the genotype.
• Example models: Stringer_2013_sipoglitazar_mixture.R (selects lcl_im = log(3.35) L/h; Stringer 2013 supplementary Table A1. The mixture prior weights estimated by the model are EM 0.18, IM 0.522, PM 0.30).
• Notes: Follows the UGT2B17_EM / UGT2B17_PM, CYP2D6_EM / CYP2D6_IM / CYP2D6_PM and CYP2C19_IM / CYP2C19_PM paired-binary precedents for multi-level metabolizer phenotypes. This canonical names the LATENT subpopulation membership a mixture model assigns from observed clearance, which is deliberately distinct from the observed germline genotype indicators UGT2B15_STAR2_HET / UGT2B15_STAR2_HOM above: Stringer 2013 fits both and reports that 61 of 744 subjects (8%) genotyped *1/*1 or *1/*2 were assigned to the PM subpopulation, i.e. phenotype and genotype disagree in a small but clinically relevant fraction. A model that assigns subpopulations from a genotype should use the _STAR2_ genotype indicators, not these.

UGT2B15_PM (canonical for UGT2B15 poor-metabolizer phenotype indicator)

• Description: 1 = subject is assigned to the UGT2B15 poor-metabolizer (PM) subpopulation; 0 = any other UGT2B15 phenotype (extensive or intermediate). Time-fixed per subject. Paired with UGT2B15_IM to encode the three-level EM (reference) / IM / PM phenotype with two binary indicators.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 with UGT2B15_IM = 0 (the UGT2B15 extensive-metabolizer stratum).
• Source aliases:
  • POP3 (CL3 PM) – Stringer 2013 supplementary mixture-model analysis.
• Example models: Stringer_2013_sipoglitazar_mixture.R (selects lcl_pm = log(1.53) L/h; Stringer 2013 supplementary Table A1).
• Notes: Companion to UGT2B15_IM; see that entry’s Notes for the latent-phenotype-versus-observed-genotype boundary.

SNP_ABCB1_RS2032582_TG (canonical for ABCB1 rs2032582 TG genotype indicator)

• Description: Binary genotype indicator for the ABCB1 rs2032582 single-nucleotide polymorphism (conventionally written c.2677G>T/A, p.Ala893Ser/Thr; exon 21; encodes P-glycoprotein / MDR1). 1 = the subject’s reported genotype string is TG; 0 = otherwise. Time-fixed per subject (germline genotype). One of five indicators (_TG, _GG, _TT, _AG, _GT) that together encode the founding source’s six-level genotype covariate, with the AT genotype as the common reference (all five indicators 0). Named by the reported genotype letters, following the SNP_ABCB1_RS1045642_GA / _GG precedent, because the founding source does not state which allele is wild type.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 in combination with SNP_ABCB1_RS2032582_GG, _TT, _AG and _GT all 0, i.e. the AT genotype group (Z3 = 0 in Zhang 2018).
• Source aliases:
  • Z3 = 1 – the Zhang 2018 “final equation” genotype code for the TG stratum (coefficient dClmdZ31, Table 6).
• Example models: Zhang_2018_sunitinib.R (linear proportional effect on the apparent clearance of the metabolite SU12662: cl_su12662 = exp(lcl_su12662 + etalcl_su12662) * (WT / 68.3)^0.538 * (1 - e_snp_abcb1_rs2032582_tg_cl_su12662 * SNP_ABCB1_RS2032582_TG) * ... with e_snp_abcb1_rs2032582_tg_cl_su12662 = 0.314, i.e. TG subjects have about 31% lower apparent SU12662 clearance than the AT reference group).
• Notes: rs2032582 is tri-allelic (G / T / A), so the genotype space is larger than the usual three levels; Zhang 2018 observed six genotype strings (AT, TG, GG, TT, AG, GT) in 53 Chinese renal-cell-carcinoma patients and fitted a separate coefficient for each non-reference level. The source distinguishes TG from GT and assigns them materially different coefficients (0.314 versus 0.0456) even though the two strings denote the same unordered heterozygous genotype; the allele order in the source’s genotype strings is therefore load-bearing and is reproduced rather than collapsed, since collapsing would require choosing one of two published estimates. An extraction whose source reports rs2032582 as unordered genotypes should register its own members rather than reusing this pair. Distinct from the pooled-stratum encoding in Zhang 2018 Table 2 (GG / “GT/A” / “AA/TT/TA”), which is the association analysis’s grouping and does not match the popPK model’s six-level coding, so the Table 2 counts cannot be used to recover per-level frequencies for this family. Both indicators in a pair must always be used together with the other three: a model that references a subset has mis-specified the reference category.

SNP_ABCB1_RS2032582_GG (canonical for ABCB1 rs2032582 GG genotype indicator)

• Description: Binary genotype indicator for the ABCB1 rs2032582 single-nucleotide polymorphism (c.2677G>T/A). 1 = the subject’s reported genotype string is GG (homozygous for the G allele); 0 = otherwise. Time-fixed per subject. Member of the five-indicator AT-referenced family described under SNP_ABCB1_RS2032582_TG.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 in combination with SNP_ABCB1_RS2032582_TG, _TT, _AG and _GT all 0, i.e. the AT genotype group (Z3 = 0 in Zhang 2018).
• Source aliases:
  • Z3 = 2 – the Zhang 2018 “final equation” genotype code for the GG stratum (coefficient dClmdZ32, Table 6).
• Example models: Zhang_2018_sunitinib.R (e_snp_abcb1_rs2032582_gg_cl_su12662 = 0.269, i.e. GG subjects have about 27% lower apparent SU12662 clearance than the AT reference group).
• Notes: Zhang 2018 Table 2 reports 10 of 53 subjects (18.87%) as GG, the only one of the six model strata whose count is directly recoverable from the publication; the other five are pooled in that table. Always used together with the other four members of the family – see SNP_ABCB1_RS2032582_TG for the reference-category and tri-allelic notes.

SNP_ABCB1_RS2032582_TT (canonical for ABCB1 rs2032582 TT genotype indicator)

• Description: Binary genotype indicator for the ABCB1 rs2032582 single-nucleotide polymorphism (c.2677G>T/A). 1 = the subject’s reported genotype string is TT (homozygous for the T allele); 0 = otherwise. Time-fixed per subject. Member of the five-indicator AT-referenced family described under SNP_ABCB1_RS2032582_TG.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 in combination with SNP_ABCB1_RS2032582_TG, _GG, _AG and _GT all 0, i.e. the AT genotype group (Z3 = 0 in Zhang 2018).
• Source aliases:
  • Z3 = 3 – the Zhang 2018 “final equation” genotype code for the TT stratum (coefficient dClmdZ33, Table 6).
• Example models: Zhang_2018_sunitinib.R (e_snp_abcb1_rs2032582_tt_cl_su12662 = 0.308, i.e. TT subjects have about 31% lower apparent SU12662 clearance than the AT reference group).
• Notes: Zhang 2018 Table 2 pools TT with AA and TA into a single 17-subject (32.08%) row, so the TT-only count is not recoverable from the publication. Always used together with the other four members of the family.

SNP_ABCB1_RS2032582_AG (canonical for ABCB1 rs2032582 AG genotype indicator)

• Description: Binary genotype indicator for the ABCB1 rs2032582 single-nucleotide polymorphism (c.2677G>T/A). 1 = the subject’s reported genotype string is AG; 0 = otherwise. Time-fixed per subject. Member of the five-indicator AT-referenced family described under SNP_ABCB1_RS2032582_TG.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 in combination with SNP_ABCB1_RS2032582_TG, _GG, _TT and _GT all 0, i.e. the AT genotype group (Z3 = 0 in Zhang 2018).
• Source aliases:
  • Z3 = 4 – the Zhang 2018 “final equation” genotype code for the AG stratum (coefficient dClmdZ34, Table 6).
• Example models: Zhang_2018_sunitinib.R (e_snp_abcb1_rs2032582_ag_cl_su12662 = 0.0368, i.e. AG subjects have about 4% lower apparent SU12662 clearance than the AT reference group).
• Notes: Genotypes containing an A allele exist at this locus because rs2032582 is tri-allelic; Zhang 2018 Table 2 reports an A-allele frequency of 0.12 in its cohort and notes that only G and T alleles were listed for this SNP in dbSNP at the time, attributing the discrepancy to the cohort not being exclusively Han. Always used together with the other four members of the family.

SNP_ABCB1_RS2032582_GT (canonical for ABCB1 rs2032582 GT genotype indicator)

• Description: Binary genotype indicator for the ABCB1 rs2032582 single-nucleotide polymorphism (c.2677G>T/A). 1 = the subject’s reported genotype string is GT; 0 = otherwise. Time-fixed per subject. Member of the five-indicator AT-referenced family described under SNP_ABCB1_RS2032582_TG.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 in combination with SNP_ABCB1_RS2032582_TG, _GG, _TT and _AG all 0, i.e. the AT genotype group (Z3 = 0 in Zhang 2018).
• Source aliases:
  • Z3 = 5 – the Zhang 2018 “final equation” genotype code for the GT stratum (coefficient dClmdZ35, Table 6).
• Example models: Zhang_2018_sunitinib.R (e_snp_abcb1_rs2032582_gt_cl_su12662 = 0.0456, i.e. GT subjects have about 5% lower apparent SU12662 clearance than the AT reference group).
• Notes: Not interchangeable with SNP_ABCB1_RS2032582_TG, despite denoting the same unordered heterozygous genotype – Zhang 2018 fitted the two strings as separate strata with coefficients an order of magnitude apart (0.0456 here versus 0.314 for TG). Reproducing the source’s allele ordering is the only way to apply its published equation; see SNP_ABCB1_RS2032582_TG Notes. Always used together with the other four members of the family.

SNP_NT5E_RS2295890 (canonical for NT5E (CD73) 5’-UTR rs2295890 G>C variant-allele carrier indicator)

• Description: Binary genotype indicator for the NT5E rs2295890 single-nucleotide polymorphism (G>C, 5’-untranslated / promoter region; NT5E encodes ecto-5’-nucleotidase, CD73, the GPI-anchored ectoenzyme that dephosphorylates extracellular nucleoside monophosphates in the purine salvage pathway). 1 = subject carries at least one variant C allele (heterozygous GC or homozygous CC); 0 = homozygous wild-type GG. Time-fixed per subject (germline genotype). This is the carrier / dominant-model encoding that is the default of the SNP_<GENE>_<RSID> family.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (homozygous wild-type GG).
• Source aliases:
  • rs2295890 (coded 0 = WT, 1 = heterozygous variant or mutant) – Mohanan 2017 Table 2 footnote a and the Results covariate-definition caption; same orientation as the canonical, no value transformation.
• Example models: Mohanan_2017_fludarabine.R (log-scale multiplicative effect on the BSA-normalised clearance of the fludarabine nucleoside F-ara-A, exp(e_snp_nt5e_rs2295890_cl * SNP_NT5E_RS2295890) with e_snp_nt5e_rs2295890_cl = log(5.03 / 7.12) = -0.347, i.e. variant carriers clear F-ara-A 29% more slowly than GG wild-type homozygotes, P = 0.038; Mohanan 2017 Abstract and Table 2).
• Notes: Complete linkage disequilibrium caveat. Mohanan 2017 reports that rs2295890 was in complete linkage disequilibrium with four other NT5E 5’-region SNPs in its cohort – rs9450278, rs9450279, rs4599602 and rs4458647 – so in that dataset the column may equivalently be derived from any of the five markers, and the effect cannot be attributed to rs2295890 specifically. A paper that genotypes one of the other four and reports it separately should record the alias here rather than registering a sibling canonical, unless it also breaks the LD block. Variant-carrier rate in the founding cohort was 15 of 48 successfully genotyped patients (31%), with a variant-allele frequency of 0.18 matching the 1000 Genomes expectation. Mohanan 2017 states this is the first report of an NT5E polymorphism affecting fludarabine pharmacokinetics and that the functional mechanism was still under investigation, so the coefficient should be treated as an empirical cohort association rather than an established transporter- or enzyme-mediated effect; scope is specific for that reason as well as because the affected parameter is paper-defined. Distinct from the nucleoside-transporter SNPs screened but not retained in the same analysis (SLC28A3/CNT3 rs7853758, SLC29A1/hENT1 rs747199) and from NT5C2 rs4917996, which is a different 5’-nucleotidase gene (cytosolic NT5C2 rather than ecto-5’-nucleotidase NT5E); those three are carried in that model’s covariatesDataExcluded because no coefficient was reported for any of them. Promote to general if a second paper ratifies the same carrier-versus-GG semantics.

Lifestyle / medical history

SMOKE (canonical for current-smoker binary indicator)

• Description: 1 = current smoker at baseline, 0 = non-smoker.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (non-smoker).
• Source aliases:
  • Smoking (case-insensitive) – used in Ma_2020_sarilumab_anc.R.
  • Smoker_active – used in Rini_2013_axitinib.R (Rini 2013 Table 2 row “Smoking status on CL” and the typical-value clearance equation in Results “PK model”, where Smoker_active “is 1 if applicable and 0 otherwise” and the reference group pools ex-smokers with non-smokers).
• Example models: Ma_2020_sarilumab_anc.R (power-form on baseline ANC: BASE * 1.15^SMOKE), Rini_2013_axitinib.R (linear-proportional effect on CL: (1 + 1.02 * SMOKE), i.e. active smokers have 102% greater clearance than the pooled ex-smoker / non-smoker reference; the estimate is imprecise – 44% RSE, 95% CI 0.144 to 1.90 – because only 19 of 590 subjects were active smokers, so Rini 2013 excluded the smoking effect from its covariate-impact simulations and called the clinical significance unclear).
• Notes: Baseline-only indicator; does not track within-study smoking-cessation changes. Use this two-level (current vs non-smoker) encoding when the source paper does not split former and never smokers. When the source uses a 3-level smoking-status categorical (never / former / current), use the paired SMOKE_CURRENT + SMOKE_NEVER indicators below instead – the 3-level encoding cannot be reduced to a single SMOKE column without losing information.

SMOKE_CURRENT (canonical for current-smoker indicator (paired with SMOKE_NEVER))

• Description: 1 = current smoker at baseline, 0 otherwise (former or never smoker). Paired with SMOKE_NEVER to encode a 3-level smoking-status categorical with former smoker as the implicit reference (both indicators = 0).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (former smoker, when paired with SMOKE_NEVER = 0). The pairing follows the RACE_<GROUP> convention for paired indicators.
• Source aliases:
  • Smoking status = Current / SMOK = 2 (case-insensitive) – derived from a 3-level smoking-status column.
• Example models: Hwang_2023_monalizumab.R (proportional-shift effect on V1: (1 + 0.0484)^SMOKE_CURRENT; reference category former smoker).
• Notes: Baseline-only indicator. See also SMOKE_NEVER (paired indicator) and SMOKE (binary current-vs-non-smoker encoding when the source paper does not split former vs never).

SMOKE_NEVER (canonical for never-smoker indicator (paired with SMOKE_CURRENT))

• Description: 1 = never smoker at baseline, 0 otherwise (former or current smoker). Paired with SMOKE_CURRENT to encode a 3-level smoking-status categorical with former smoker as the implicit reference (both indicators = 0).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (former smoker, when paired with SMOKE_CURRENT = 0). The pairing follows the RACE_<GROUP> convention for paired indicators.
• Source aliases:
  • Smoking status = Never / SMOK = 0 (case-insensitive) – derived from a 3-level smoking-status column.
• Example models: Hwang_2023_monalizumab.R (proportional-shift effect on V1: (1 - 0.141)^SMOKE_NEVER; reference category former smoker), Claret_2014_motesanib_tumorsize_OS.R (unpaired two-level use: the source dichotomises smoking history as “former and current smokers vs. those who had never smoked”, so the paper’s ever-smoker indicator is formed inside model() as 1 - SMOKE_NEVER and the published coefficient -0.264 is applied to it unchanged).
• Notes: Baseline-only indicator. See also SMOKE_CURRENT (paired indicator) and SMOKE (binary current-vs-non-smoker encoding when the source paper does not split former vs never). SMOKE_NEVER may also be used unpaired, without SMOKE_CURRENT, when the source paper collapses smoking history to the two-level contrast never vs. ever (former or current) – the description above already covers that encoding exactly (1 = never, 0 = former or current), and the implicit reference category is then the pooled ever-smoker group rather than the former-smoker group. When the source’s published coefficient is stated on the ever-smoker side of that contrast, form the ever-smoker indicator in model() as 1 - SMOKE_NEVER and keep the published coefficient and intercept unchanged, rather than re-parameterising onto SMOKE_NEVER and shifting the intercept – that preserves the correspondence with any published parameter covariance matrix.

SMOKE_TTFC_SCORE (canonical for the Fagerstrom time-to-first-cigarette nicotine-dependence score)

• Description: Fagerstrom Test for Nicotine Dependence (FTND) item 1 – “How soon after you wake up do you smoke your first cigarette?” – carried as the instrument’s own published 0-3 ordinal score. Canonical scoring, following the FTND and Ravva 2010 Methods: 0 = more than 60 min after waking; 1 = 31-60 min; 2 = 6-30 min; 3 = within 5 min. Higher score means greater nicotine dependence. Baseline-only, time-fixed per subject.
• Units: (ordinal score 0-3)
• Type: categorical
• Scope: general
• Reference category: 0 (first cigarette more than 60 min after waking) is the conventional model reference level. Beware a distinction that recurs in this literature: the score-0 model reference category is frequently NOT the same as the reporting reference population a paper uses to quote probabilities (Ravva 2010 quotes everything against a score-2 “first cigarette within 6-30 min” typical smoker). Record both in covariateData[[SMOKE_TTFC_SCORE]]$notes.
• Source aliases:
  • FSQ1 – “Fagerstrom question 1”; the printed name in Ravva 2010 Table 1, Table 3 and Equation 2. Same 0-3 orientation as the canonical, so no value transformation.
  • TTFC, FTND_Q1 – common alternate abbreviations in the smoking-cessation literature.
• Example models: Ravva_2010_varenicline_car_w9_12.R, Ravva_2010_varenicline_nausea.R (decomposed inside model() into the three non-reference indicators, each with its own estimated exponent multiplying the baseline logit: e_smoke_ttfc_31_60_base_logit^(SMOKE_TTFC_SCORE == 1) * e_smoke_ttfc_6_30_base_logit^(SMOKE_TTFC_SCORE == 2) * e_smoke_ttfc_le5_base_logit^(SMOKE_TTFC_SCORE == 3); efficacy exponents 1.54 / 1.92 / 2.59, nausea exponents 0.894 / 0.867 / 0.961), Fediuk_2021_varenicline_nausea.R (same decomposition on an adolescent nausea-or-vomiting logit; exponents 0.0750 / 0.0773 / 0.0779 against an intercept of -30.9, because the score-0 reference level held only 5 of 238 subjects).
• Notes: A single ordinal column, decomposed in model() – deliberately not three pre-binned indicator columns. This follows the SNP_CYP2B6_RS35303484_G_COUNT precedent, whose own Notes make the general argument: one column captures the underlying measurement without redundancy and lets the model code derive either a linear per-level effect or a non-additive level-by-level decomposition, depending on the source paper’s parameterisation. That matters here because papers bin time-to-first-cigarette inconsistently (the FTND’s own 4-level split, a 2-level <=5 min vs rest split, and heaviness-of-smoking-index variants all occur), and a dataset carrying the raw 0-3 score can serve any of them, whereas pre-binned indicator columns can only serve the paper that created them. Encoding the score does not imply a linear dose-response: the founding models estimate a separate exponent per level, and the Ravva 2010 nausea estimates are visibly non-monotone in the score (0.894 at level 1, 0.867 at level 2, 0.961 at level 3), which a linear-in-score encoding could not represent. Genuinely distinct from the existing smoking canonicals, which is why it needed its own entry rather than an alias: SMOKE, SMOKE_CURRENT and SMOKE_NEVER all encode smoking status (current / former / never) and are uninformative in a smoking-cessation cohort where every subject is a current smoker by enrolment criterion; SMOKE_TTFC_SCORE encodes dependence severity among current smokers. Sibling SMOKE_CPD_SCORE (FTND item 4, cigarettes per day) is the other commonly-collected FTND item; the two are strongly collinear and a source paper will typically retain only one.

SMOKE_CPD_SCORE (canonical for the Fagerstrom cigarettes-per-day score)

• Description: Fagerstrom Test for Nicotine Dependence (FTND) item 4 – “How many cigarettes per day do you smoke?” – carried as the instrument’s own published 0-3 ordinal score. Canonical scoring, following the FTND and Ravva 2010 Methods: 0 = 10 or fewer per day; 1 = 11-20; 2 = 21-30; 3 = 31 or more. Higher score means greater nicotine dependence. Baseline-only, time-fixed per subject.
• Units: (ordinal score 0-3)
• Type: categorical
• Scope: general
• Reference category: 0 (10 or fewer cigarettes per day).
• Source aliases:
  • FSQ4 – “Fagerstrom question 4”; the printed name in Ravva 2010 Table 1. Same 0-3 orientation as the canonical, so no value transformation.
  • CPD, FTND_Q4 – common alternate abbreviations.
• Example models: Ravva_2010_varenicline_car_w9_12.R, Ravva_2010_varenicline_nausea.R (both list it in covariatesDataExcluded: screened as a candidate nicotine-dependence predictor but dropped for collinearity with SMOKE_TTFC_SCORE).
• Notes: Distinct from a raw cigarettes-per-day count: this is the FTND’s banded 0-3 score, not the underlying integer. A source paper reporting the untransformed count should register a separate continuous canonical rather than coercing it into these bands, since the band edges are not recoverable from the count in reverse. Strongly collinear with SMOKE_TTFC_SCORE; papers that screen both typically retain only one (Ravva 2010 Results: “Simultaneous inclusion of the correlated predictors FSQ1, FSQ4, and CO in the model was avoided by selecting the FSQ1 variable as the most relevant predictor of nicotine dependence”, and the Discussion adds that time to the first cigarette “was found to be more predictive of smoking cessation than the number of cigarettes smoked per day”). First registered on 2026-08-31 alongside the Ravva 2010 varenicline exposure-response extraction, as a documented-but-excluded covariate.

CO_EXHALED_PPM (canonical for exhaled carbon monoxide concentration)

• Description: Carbon monoxide concentration in exhaled breath, in parts per million, measured with a handheld CO monitor. A short-half-life biochemical marker of recent smoke inhalation, used both as a continuous index of smoking intensity and, against a threshold, as the standard objective confirmation of self-reported smoking abstinence.
• Units: ppm
• Type: continuous
• Scope: general
• Reference category: n/a (continuous). Typical smoking-cessation-trial abstinence threshold is <= 10 ppm; document the threshold a given paper uses in covariateData[[CO_EXHALED_PPM]]$notes, since it varies (6, 8 and 10 ppm all appear in the literature).
• Source aliases:
  • CO – the bare two-letter form used in Ravva 2010 Table 1, Methods and Figure 2. The full-word canonical is preferred over the bare abbreviation per the standing canonical-naming rule and the CARDIAC_OUTPUT precedent (where the same bare CO abbreviation denotes an unrelated hemodynamic quantity); the alias is preserved so source datasets translate without value transformation.
  • ECO, eCO, CO_BREATH – alternate abbreviations in the smoking-cessation literature.
• Example models: Ravva_2010_varenicline_car_w9_12.R, Ravva_2010_varenicline_nausea.R (both list it in covariatesDataExcluded: screened as a candidate baseline-smoking-status predictor but dropped for collinearity with SMOKE_TTFC_SCORE; retained in the source analysis only to confirm self-reported abstinence at <= 10 ppm and to impute eight missing baseline values by multivariate regression on FSQ1 and gender).
• Notes: Name collision warning: the bare source abbreviation CO is used in the literature for both exhaled carbon monoxide and cardiac output. The register resolves this by giving each concept an explicit full-word canonical – CO_EXHALED_PPM here and CARDIAC_OUTPUT for the hemodynamic quantity – and by keeping CO only as a per-concept source alias. Always disambiguate from the source paper’s context before mapping a CO column. The _PPM suffix is retained in the canonical name because the abstinence threshold that gives this covariate most of its operational meaning is quoted in ppm, and a percent-carboxyhemoglobin (COHb) measurement is a different quantity on a different scale that should register its own canonical rather than reuse this one. General scope: exhaled CO is an assay-independent measurement usable by any smoking-related PK or PD analysis (it also appears as a CYP1A2-induction surrogate, since polycyclic aromatic hydrocarbons in smoke induce CYP1A2 and exhaled CO tracks recent smoke exposure). First registered on 2026-08-31 alongside the Ravva 2010 varenicline exposure-response extraction, as a documented-but-excluded covariate.

ALCOHOL_ABUSE (canonical for chronic alcohol abuse clinical indicator)

• Description: 1 = subject has a documented history of chronic alcohol abuse (defined per paper; commonly >= 6 standard drinks / units per day for a sustained period, per Swart 2004 which follows the Vrije Universiteit Medical Center clinical criteria); 0 = no such history. Subject-level baseline behavioural / clinical-history flag captured at study or ICU admission from patient / next-of-kin interview, not a laboratory measurement and NOT a pharmacogenomic marker. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no chronic alcohol abuse).
• Source aliases:
  • alcohol abuse – printed prose label in Swart 2004 Results / Table 4 / Table 5 CL formulas (“CL no alcohol abuse” vs “CL alcohol abuse”); used in Swart_2004_lorazepam.R and Swart_2004_midazolam.R.
• Example models: Swart_2004_lorazepam.R (selector effect on CL: alcohol-abuse patients take a flat cl = 0.74 L/h; non-alcohol-abuse patients take the PEEP-adjusted cl = 4.13 - (PEEP - 5) * 0.417 L/h; contrast is a ~5.6-fold reduction in CL), Swart_2004_midazolam.R (selector effect on CL: alcohol-abuse patients take cl = 7.3 - (AGE - 57) * 0.145 L/h; non-alcohol-abuse patients take cl = 11.3 - (AGE - 57) * 0.145 L/h; same age slope in both strata, ~4 L/h lower baseline in the alcohol-abuse stratum at age 57).
• Notes: General scope because chronic alcohol abuse is a clinically-defined behavioural indicator applicable across drug classes and populations, and appears frequently as a hepatic-metabolism modifier (predominantly via CYP2E1 induction and cirrhosis-linked reductions in phase-I and phase-II capacity). Critically distinct from the pharmacogenomic ADH1B_S2_HOM / ADH1B_S2_CARRIER canonicals (Nemoto 2017 ethanol PK): those encode a germline high-activity aldehyde-dehydrogenase-1B allele (rs1229984) that changes intrinsic ethanol clearance; ALCOHOL_ABUSE encodes patient behaviour / history and would remain 1 regardless of ADH1B genotype. Per-model covariateData[[ALCOHOL_ABUSE]]$notes must document the paper’s exact definition (units/day, duration, ascertainment source); Swart 2004 uses “> 6 units per day” as the threshold. Future popPK papers using a different definition (e.g., DSM alcohol use disorder criteria, positive CAGE screen, ICD codes) can extend Example models provided they document the definition in per-model notes; if the paper uses a numerically-graded alcohol-use covariate rather than a binary flag, prefer a separate ordinal canonical rather than reusing this binary.

ALCOHOL_USE (canonical for any-history-of-alcohol-use indicator)

• Description: 1 = subject has any history of alcohol use (any regular or occasional consumption, per the source paper’s own ascertainment); 0 = no history of alcohol use. Subject-level baseline behavioural-history flag taken from patient interview / medical history, not a laboratory measurement. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no history of alcohol use).
• Source aliases:
  • ALC – Riggs 2013 Table 2 row “CL/F x ALC (theta_10)”; the matching Figure 3A panel label is “Consumes alcohol” and the Figure 3 caption defines the reference individual as having “no history of alcohol use”. Used in Riggs_2013_empagliflozin.R.
• Example models: Riggs_2013_empagliflozin.R (multiplicative effect on CL/F: 1.02^ALCOHOL_USE, i.e. 2% higher apparent oral clearance in patients with any alcohol-use history; Riggs 2013 Table 2 theta_10 = 1.02, 95% CI 0.984-1.06).
• Notes: Critically distinct from ALCOHOL_ABUSE, and the two must never be pooled into one column. ALCOHOL_ABUSE encodes a narrow clinical diagnosis (chronic abuse, commonly >= 6 units/day sustained, per the Swart 2004 / VUMC criteria); ALCOHOL_USE encodes the far broader “drinks at all” contrast, so a subject who has one drink a week is ALCOHOL_USE = 1 and ALCOHOL_ABUSE = 0. Merging them would silently make a shared column mean “alcohol abuser” in one dataset and “has ever consumed alcohol” in another. General scope because any-alcohol-use is a lifestyle / medical-history flag with a stable paper-independent meaning, matching how the sibling behavioural flags SMOKE, ALCOHOL_ABUSE and CANNABIS_DAILY are scoped. Per-model covariateData[[ALCOHOL_USE]]$notes should record whatever ascertainment detail the source paper gives, because “alcohol use” is not standardised across the literature; Riggs 2013 states only that “smoking and alcohol histories” were evaluated as covariates and does not publish a threshold. A future paper that grades alcohol consumption numerically (units/week, AUDIT score) should register a separate continuous or ordinal canonical rather than overload this binary.

CANNABIS_DAILY (canonical for daily-versus-occasional cannabis use indicator)

• Description: 1 = habitual daily cannabis user, 0 = occasional cannabis user. Subject-level baseline behavioural-history flag describing the participant’s habitual pattern of cannabis consumption before the study visit, ascertained by self-reported use-frequency interview rather than by a laboratory measurement. Time-fixed per subject and independent of the product consumed on the study day (flower versus concentrate).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (occasional user).
• Source aliases:
  • pattern of use (daily versus occasional) – prose label in Henthorn 2024 Methods (“a modeled analysis was conducted, in which the pattern of use was dichotomized (daily use versus occasional use) without regard to whether flower or concentrate was consumed in the dosing protocol”) and Table 3 footnote (“Covariate_daily user, value of the covariate for the estimated inhaled THC dose for both concentrate and flower users where the default is Covariate_occasional user = 0”); used in Henthorn_2024_tetrahydrocannabinol.R.
• Example models: Henthorn_2024_tetrahydrocannabinol.R (exponential effect on the estimated inhaled-dose fraction Fi: fdepot = exp(lfdepot + e_cannabis_daily_fdepot * CANNABIS_DAILY + etalfdepot) with e_cannabis_daily_fdepot = 1.79; daily users inhale exp(1.79) = 5.99-fold more bioavailable THC than the occasional-user reference, i.e. 10.78 mg versus 1.80 mg of a nominal 15 mg fully bioavailable dose).
• Notes: General scope because habitual cannabis use is a behavioural-history indicator applicable beyond a single drug: it plausibly modifies the disposition of co-administered drugs (CYP1A2 induction from combusted-material exposure, CYP2C9 / CYP3A interactions) and characterises co-use populations, matching how the sibling per-substance flags SMOKE (tobacco) and ALCOHOL_ABUSE (alcohol) are scoped. Distinct from SMOKE: a daily cannabis user may be a tobacco non-smoker and vice versa, and concentrate users (vape pens, dabs) do not combust plant material at all, so the two columns must be carried separately rather than collapsed. Distinct from the model-specific OPIOID_PATIENT_TYPE selector, which is scoped specific and names a parameter-set selector rather than a behavioural-history flag. Per-model covariateData[[CANNABIS_DAILY]]$notes must document the paper’s exact frequency definitions, because “occasional” is not standardised across the literature; Henthorn 2024 Methods defines occasional use as “an average of at least 2 days per month and no more than 3 days per week in the 90 days before enrollment” and daily use as daily consumption of flower or concentrate over the same window. A future paper that resolves usage pattern more finely (e.g. occasional / daily-flower / daily-concentrate, which Henthorn 2024 tested and rejected as non-significant, or a graded days-per-month count) should register a separate ordinal or paired-indicator canonical rather than overload this binary.

ATHLETE (canonical for athlete / endurance-trained-individual indicator)

• Description: 1 = competitive athlete or healthy endurance-trained individual, 0 = the study’s non-athlete reference group. Subject-level baseline training-status flag describing habitual endurance training rather than a measured physiological quantity. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (non-athlete). The exact composition of the reference group is paper-defined and must be recorded in covariateData[[ATHLETE]]$notes; in the founding example it is healthy participants pooled with chronic asthmatics, because the paper’s rich model found asthmatics not significantly different from healthy participants on either affected parameter and therefore regrouped them.
• Source aliases:
  • TYPE (category 3 of healthy / chronic asthmatic / athlete-endurance-trained) – Thoueille 2026 Data S1 $PK derives ATHLETES = 1 when TYPE == 3.
  • athletes/endurance-trained individuals – Thoueille 2026 Table 2 row label and Results prose.
• Example models: Thoueille_2026_salmeterol.R (multiplicative-power effect on two parameters of the joint plasma-and-urine inhaled-salmeterol model: cl = exp(lcl + etalcl) * 1.63^ATHLETE and kurine = exp(lkurine + etalkurine) * 2.91^ATHLETE, i.e. athletes have 63% higher apparent salmeterol plasma clearance and a 191% higher salmeterol urinary excretion rate constant, together producing substantially higher urine concentrations at the same inhaled dose).
• Notes: Record the study’s actual training modality in covariateData[[ATHLETE]]$notes so the specific cohort is not lost – the founding example’s athletes are endurance-trained (cyclists and comparable endurance disciplines), and a future paper contrasting strength-trained against endurance-trained athletes within an athlete-only cohort should register a separate within-cohort modality indicator rather than widening or re-pointing this entry. General scope because training status is a lifestyle / medical-history flag with a stable paper-independent meaning, matching how the sibling behavioural flags SMOKE, ALCOHOL_ABUSE and CANNABIS_DAILY are scoped, and because sports-pharmacology and anti-doping models are a recurring class in this library (see also Heuberger_2018_salbutamol.R, whose whole simulated cohort is elite cyclists but which carries no covariate because it does not contrast them against a non-athlete group). Distinct from any measured surrogate of fitness (VO2max, lean body mass, haematocrit): this column records group membership as the source paper assigned it, not a physiological measurement.

Formulation / assay / study

ROUTE_* family – section-header policy

All ROUTE_<TARGET> canonicals follow the same shape: a binary indicator where 1 = subject received the <TARGET> administration route and 0 = subject received the reference (non-target) route. The reference category is route-pair-specific and is documented per-canonical in the Reference category: field (e.g., ROUTE_IV references SC; ROUTE_IP references non-IP; ROUTE_NGT references oral). The 2026-06-19 canonical-register standardization audit reviewed renaming this family to a more explicit ROUTE_<TARGET>_VS_<REFERENCE> shape but the maintainers declined the rename and instead set the following documentation discipline:

Policy: every ROUTE_* canonical MUST document its reference category explicitly in the Reference category: field of its H3 block. The reference is what ROUTE_<TARGET> = 0 means in the source dataset and is paper-specific. New ROUTE_* canonicals registered after 2026-06-19 must follow this rule; existing entries already comply. The discipline eliminates the silent-mis-encoding risk (where a model file might assume the wrong reference category) by making the reference explicit in the register rather than relying on the column name alone.

ROUTE_IV (canonical for IV-vs-SC administration route indicator)

• Description: 1 = subject received intravenous (IV) administration, 0 = subcutaneous (SC) administration. Per-subject (study-fixed) covariate flagging the dosing route when a population analysis pools cohorts that differ by route, with covariate effects on PK parameters that capture route-specific disposition behaviour.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (SC).
• Source aliases:
  • “Admin route = IV” (categorical effect column in Yu 2022 covariate equations).
  • IV – used in Zierhut_2008_osteoprotegerin.R (DDMODEL00000233 dataObj column flagging IV vs SC cohort, switching the PK observation residual SD between CcpropSdIV and CcpropSdSC).
  • “IV” / “SC arm” – used in Wang_2021_pertuzumab.R (per-subject FeDeriCa arm indicator P+H IV vs PH FDC SC, switching the proportional residual SD between CcpropSdIv and CcpropSdSc).
  • “Route of administration: IV / IM” (supplementary Table S1 row in Clegg 2024) – used in Clegg_2024_tixagevimab_cilgavimab.R, Clegg_2024_tixagevimab.R, Clegg_2024_cilgavimab.R. The comparator non-IV route here is intramuscular, not SC.
• Example models:
  • Yu_2022_ofatumumab.R (exponential effect on R0, CL, Q, ksyninf).
  • Zierhut_2008_osteoprotegerin.R (per-subject indicator switching the PK observation residual SD between the IV cohort (CcpropSdIV) and the SC cohort (CcpropSdSC)).
  • Wang_2021_pertuzumab.R (per-subject indicator switching the proportional residual SD between the IV (CcpropSdIv = 0.175) and SC (CcpropSdSc = 0.155) cohorts of the FeDeriCa popPK).
  • FiedlerKelly_2019_fremanezumab.R (per-subject indicator switching both the central volume of distribution (Vc,IV = 2.98 L FIXED vs Vc,SC = 1.88 L) and the residual-error structure (IV: proportional-only with SD = sqrt(0.0467) = 0.21610; SC: combined additive sqrt(0.204) + proportional sqrt(0.0531)) in the pooled phase 1/2b/3 fremanezumab popPK).
  • Fanta_2007_ciclosporin.R (per-occasion indicator carrying a 23% lower typical CL on oral occasions than on IV occasions (exp(e_route_iv_cl * ROUTE_IV) with e_route_iv_cl = log(1/0.77) = 0.2614) and switching the residual-error structure between IV (propSdIv = 0.089 + addSdIv = 1.5 ug/L) and oral (propSdPo = 0.20) arms; the comparator non-IV route is oral microemulsion ciclosporin, not SC).
  • Koolen_2010_docetaxel.R (per-dose-record indicator switching the docetaxel central volume of distribution between the IV value (V_central_iv = 9.8 L, polysorbate-80-micelle-sequestered) and the oral value (V_central_po = 44.0 L) per Koolen 2010 Results ‘Volume of distribution’ section; the reference category is oral here, not SC, because the source paper pools IV and oral docetaxel cohorts).
  • Roepcke_2018_tak_079.R (per-subject / per-dose-record indicator switching the TAK-079 central volume of distribution between the IV reference (Vc = 0.141 L, monkey) and SC (Vc = 0.043 L, ca. 70% smaller) via vc = exp(lvc + etalvc) * (1 - e_route_iv_vc * (1 - ROUTE_IV)) with e_route_iv_vc = 0.697 per Roepcke 2018 Table 2 ‘ROUT on V_C’; SC data come from the four lower single-dose groups <= 1 mg/kg in studies 7 and 8).
  • Toutain_2025_doxycycline_pig.R (per-dose-record indicator separating the single-dose IV catheter arm of the VetCAST pig doxycycline meta-analysis from the four oral modalities; when ROUTE_IV = 1 both the absorption rate constant and the bioavailability collapse to 0 so the depot equation contributes nothing and the dose must be placed in the central compartment via cmt = 'central', and the IV residual-error pair (proportional 13.9%, additive 0.0128 ug/mL) is selected. Toutain 2025 Appendix S3 Phoenix script block A. The comparator non-IV route is oral, not SC).
  • Kuroda_2024_quinidine_horse.R (per-dose-record indicator selecting BOTH the proportional and the additive residual-error term of the Thoroughbred-horse quinidine popPK between the intravenous dataset (propSdIv = 0.0594 with addSdIv = 0.0338 ug/mL) and the oral dataset (propSdOral = 0.1571 with addSdOral = 0.0479 ug/mL) per Kuroda 2024 Table 1 rows CMultStdev0/Stdev0 and CMultStdev1/Stdev1. Structural disposition is shared between the routes and no route effect was retained on any structural parameter, so this is the same residual-only role as Zierhut_2008_osteoprotegerin.R, Wang_2021_pertuzumab.R and Clegg_2024_tixagevimab_cilgavimab.R, but switching the additive term as well as the proportional one. The reference category here is oral quinidine sulfate dihydrate given by nasogastric tube, not SC. Dose into depot when ROUTE_IV = 0 and into central when ROUTE_IV = 1).
  • vandenBerg_2021_uprifosbuvir_pbpk.R (per-dose-record indicator switching between IV and oral uprifosbuvir; when ROUTE_IV = 1 the oral-absorption fast/slow fractions F1/F2 are set to 0 so the depot compartments contribute no mass and the dose must be placed directly into the central compartment via cmt = 'central' on the event record; when ROUTE_IV = 0 the standard oral absorption applies per FORM_CAPSULE and CONMED_ITRACONAZOLE. Corresponds to NONMEM FORM = 10 for the IV cohort in van den Berg 2021 supplement Table 1).
  • Clegg_2024_tixagevimab_cilgavimab.R, Clegg_2024_tixagevimab.R, Clegg_2024_cilgavimab.R (per-dose-record indicator switching the log-scale additive residual SD of the AZD7442 popPK between the intravenous value (expSdIv, 0.104 for the summed AZD7442 analyte) and the intramuscular value (expSdIm, 0.24) per Clegg 2024 Table 1 rows error_ADD2 / error_ADD1. The reference category here is intramuscular, not SC – Clegg 2024 pools an IM cohort (4,598 participants, 93.1%) with an IV cohort (342, 6.9%) and no SC arm exists. Structural disposition is shared between the routes; only the residual-error magnitude differs, so this is the same role as Zierhut_2008_osteoprotegerin.R and Wang_2021_pertuzumab.R. Dose into depot when ROUTE_IV = 0 and into central when ROUTE_IV = 1).
  • Pelligand_2016_robenacoxib_cat.R (per-dose-record indicator selecting which bioavailability applies to a dose placed in the central compartment of the feline robenacoxib popPK. The MLXTRAN listing in the paper’s supplement declares all three inputs on compartment 1 – iv(cmt=1,adm=2) with no bioavailability term beside the two simultaneous subcutaneous inputs oral(cmt=1,adm=1,Tk0=Tk0,p=Ftot*F0) and oral(cmt=1,adm=1,ka=ka,p=Ftot*(1-F0)) – but rxode2 applies one f() per compartment rather than per administration type, so f(central) <- ROUTE_IV + (1 - ROUTE_IV) * fdepot * (1 - ffo) collapses to 1 for an IV bolus and to Ftot*F0 for the zero-order half of an SC dose. A single SC administration therefore needs TWO dose records at the same time: a bolus into depot and a rate = -2 record into central so the modelled dur(central) is applied; an IV administration is one plain bolus into central, and because that record carries no rate rxode2 ignores dur(central) and keeps it a bolus. 14 of the 83 cats received both routes at least a week apart, so the column varies within subject. The reference category here is subcutaneous, as in the register’s base description).
  • Wang_2020_vitacoxib_cat.R (per-dose-record indicator selecting which bioavailability applies to a dose placed in the central compartment of the feline vitacoxib popPK, the same parallel zero- + first-order input construction as Pelligand_2016_robenacoxib_cat.R: f(central) <- ROUTE_IV + (1 - ROUTE_IV) * fdepot * (1 - ffo) collapses to 1 for an i.v. bolus and to F*(1-Fr) for the zero-order part of an oral dose. An oral administration needs TWO records at the same time (a bolus into depot and a rate = -2 record into central); an i.v. administration is one plain bolus into central. The same eight cats received both routes (Studies 1-3 and 6), so the column varies within subject. The reference category here is oral, not SC).
  • Han_2025_midazolam_pbpk.R, Han_2025_fentanyl_pbpk.R, Han_2025_alfentanil_pbpk.R, Han_2025_sufentanil_pbpk.R (per-dose-record indicator gating the gut-wall unbound fraction of the whole-body age-dependent CYP3A4 PBPK. Han 2025 Supplementary Eq S5 states that fu,gut “was defaulted to 1 for oral administration and equaled to free fraction of drug (fu,b) in blood for intravenous administration”, so fu_gut = ROUTE_IV * fu_b + (1 - ROUTE_IV) * 1: an oral dose sees full first-pass extraction of the absorbed flux by duodenal, jejunal and ileal CYP3A4 while an intravenous dose sees only the protein-binding-limited rate. Same gating role as vandenBerg_2021_uprifosbuvir_pbpk.R. The reference category here is oral, not SC, because Han 2025 pool oral and intravenous midazolam; the three opioid siblings are intravenous throughout. Dose into stomach when ROUTE_IV = 0 and into venous when ROUTE_IV = 1).
  • Checchio_2017_psoriasis_pasi_landmark_mbma.R (arm-level indicator selecting the alefacept ED50 between the intravenous branch (3.42 mg, ROUTE_IV = 1) and the reference intramuscular branch (15.4 mg) via exp(led50_alefacept + e_route_iv_led50_alefacept * ROUTE_IV) with e_route_iv_led50_alefacept = -1.502, per Checchio 2017 Supplementary Table 2 row alefacetp.I(bio.route = 'IV'). The reference category here is intramuscular, not SC, as in the Clegg 2024 entries above. The indicator affects ALEFACEPT ONLY and must be left at 0 for every other agent in that model, and the source’s own Table 2 alefacept prediction is reproduced by the IV branch, not the IM one).
  • Duflot_2022_morphine.R (per-subject indicator of the parallel-group intravenous-bolus versus nebulized morphine trial. The reference category here is nebulized (inhaled) morphine, not SC. The source codes the route the other way round – 0 = i.v. (reference), 1 = NEB, applied as log(theta) = log(theta_pop) + beta – so the model writes neb <- 1 - ROUTE_IV and carries the paper’s fixed NEB effects on the central-to-peripheral2 transfer rate (e_neb_k13 = 1.39) and on M3G elimination (e_neb_kel_m3g = -1.33) unchanged, per Duflot 2022 Table 3. The route also selects the dose target: central for an intravenous bolus, depot for a nebulized dose, whose Savic transit input carries the nebulized bioavailability F = 0.035).
• Notes: This is the per-subject covariate-equation indicator, distinct from the dosing-event cmt column that names the target compartment. When simulating, set ROUTE_IV = 1 for IV cohorts and dose into the central compartment; set ROUTE_IV = 0 for SC cohorts and dose into the depot. Scope: specific because the set of parameters that differ by route is paper-specific (Yu 2022 carries route-specific exponential effects on disposition parameters; Zierhut 2008, Wang 2021, and Fiedler-Kelly 2019 carry route-specific PK observation residual SDs; Fiedler-Kelly 2019 and Roepcke 2018 additionally carry a route-specific central volume of distribution; van den Berg 2021 gates the entire oral-absorption cascade including gut wall extraction, four-step transit chain, and pseudo-M4 gut pool so that IV doses bypass the first-pass metabolism to gut UMP-UTP-CTP-CMP cycle).

ROUTE_IP (canonical for intraperitoneal-vs-non-IP administration route indicator)

• Description: 1 = subject (or dose record) received intraperitoneal (IP) administration; 0 = subcutaneous (SC), intravenous (IV), or any other non-IP route. Per-dose-record covariate flagging the IP route when a preclinical popPK pools multiple routes with route-specific bioavailability and the IP arm is the only route that carries a non-unity F.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-IP, typically SC or IV with bioavailability fixed at 1).
• Source aliases:
  • “IP” (route label in Johnson 2011 Table I, per-study route designation across 12 rat studies – IP studies 1-6b, SC studies 7-11, IV study 12).
• Example models:
  • Johnson_2011_olanzapine_rat.R (per-dose-record indicator selecting the IP bioavailability FIP = 0.636 with 87% CV log-normal IIV; ROUTE_IP = 0 selects F = 1 for SC and IV. The encoding f(central) <- exp(ROUTE_IP * (lfip + etalfip)) collapses to 1 when ROUTE_IP = 0 because exp(0) = 1, so subjects dosed via SC or IV inherit complete bioavailability without IIV on F).
• Notes: This is the per-dose-record covariate-equation indicator, distinct from the dosing-event cmt column that names the target compartment (Johnson 2011 doses all routes directly into central because the absorption rate constant was not estimable from the available data). When simulating IP doses, set ROUTE_IP = 1 on the dose record(s); set ROUTE_IP = 0 for SC and IV dose records. Scope: specific because the IP-vs-other contrast and which parameter it modifies (here, bioavailability) is paper-specific; complementary to ROUTE_IV (IV-vs-SC indicator) – a future tri-route study could use both indicators jointly.

ROUTE_NGT (canonical for nasogastric-tube-vs-oral administration route indicator)

• Description: 1 = dose record administered by nasogastric tube (NGT), 0 = oral administration. Per-dose-record covariate flagging NGT delivery when an oral popPK pools whole-tablet / crushed-tablet oral dosing with crushed-tablet-via-NGT dosing, with the route effect captured on an absorption parameter.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (oral administration; the structural reference for the Denti 2018 typical absorption lag time T_lag of 0.242 h).
• Source aliases:
  • NGT – used in Denti_2018_levofloxacin.R (per-dose-record nasogastric-tube indicator).
  • SOL (ST) – used in Toutain_2025_doxycycline_pig.R (Toutain 2025 Table 3 route column: an oral doxycycline solution delivered by oro-gastric “stomach tubing” rather than taken spontaneously in drinking water).
• Example models: Denti_2018_levofloxacin.R (multiplicative effect on the absorption lag time: (1 + e_route_ngt_tlag * ROUTE_NGT) with e_route_ngt_tlag = -0.856, so NGT delivery shortens T_lag to ~14.4% of its oral value; in the cohort 90/109 (82.6%) children were dosed by crushed tablet via NGT, Denti 2018 Table 2).
  • Denti_2018_levofloxacin.R (multiplicative effect on the absorption lag time: (1 + e_route_ngt_tlag * ROUTE_NGT) with e_route_ngt_tlag = -0.856, so NGT delivery shortens T_lag to ~14.4% of its oral value; in the cohort 90/109 (82.6%) children were dosed by crushed tablet via NGT, Denti 2018 Table 2).
  • Toutain_2025_doxycycline_pig.R (selects the stomach-tube sub-model of the pig doxycycline meta-analysis – Ka = 0.725 1/h, F = 0.258, proportional residual 27.5% plus additive 0.0026 ug/mL – against the spontaneous drinking-water sub-model at ROUTE_NGT = 0 – Ka = 0.689 1/h, F = 0.307, proportional residual 29.3% plus additive 0.0057 ug/mL; Toutain 2025 Table 6. Only meaningful when ROUTE_IV = 0 and FORM_DOX_FEED = 0).
• Notes: This is the per-dose-record covariate-equation indicator, distinct from the dosing-event cmt column that names the target compartment. When simulating NGT doses, set ROUTE_NGT = 1 on the dose record(s); set ROUTE_NGT = 0 for oral dose records. Scope: specific because the NGT-vs-oral contrast and which absorption parameter it modifies are paper-specific. Distinct from the ROUTE_IV (IV-vs-SC) and ROUTE_IP (IP-vs-non-IP) parenteral-route indicators – ROUTE_NGT is an enteral-delivery-method indicator within oral administration. The canonical covers tube-delivered vs voluntarily-ingested enteral dosing generally, so a veterinary oro-gastric tube (Toutain 2025) uses the same column as a human nasogastric tube (Denti 2018); the reference category is spontaneous oral intake in both.

ROUTE_ORAL (canonical for oral-vs-non-oral administration route indicator)

• Description: 1 = dose record administered orally (capsule / tablet / oral solution), 0 = any non-oral route pooled in the same analysis (intranasal, intravenous, or other). Per-dose-record covariate flagging oral delivery when a population analysis pools an oral arm with one or more non-enteral arms and estimates a separate absorption parameter set – bioavailability, absorption rate constant, zero-order input duration, lag time – for each route, and/or restricts a systemic covariate effect to the oral route only.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-oral; the pooled reference is paper-specific and must be documented per model). In Comisar_2025_zavegepant.R the reference pools the intranasal nasal-spray arm and the intravenous-infusion arm: IV doses are placed directly into central by the dose record’s cmt and therefore never read the depot parameters at all, so ROUTE_ORAL = 0 selects the intranasal absorption set on the records that actually enter the depot. In Thoueille_2024_rilpivirine.R the reference is the long-acting intramuscular gluteal nanosuspension and there is no third route.
• Source aliases:
  • “Administration route” – Comisar 2025 Table 1 per-study route column (Nasal spray / Oral / Intravenous infusion).
  • ROUT – Willmann 2019 supplementary NONMEM $INPUT column, an integer route code branched in $ERROR as IF(ROUT.EQ.1), with ROUT = 1 oral and ROUT = 2 intravenous. Derive ROUTE_ORAL = as.integer(ROUT == 1).
  • LAI – Thoueille 2024 supplementary NONMEM $INPUT column, carried with the opposite polarity (LAI = 1 marks the long-acting intramuscular records, LAI = 0 the oral records), so ROUTE_ORAL = 1 - LAI.
• Example models:
  • Thoueille_2024_rilpivirine.R (row-level indicator on both dose and observation records, selecting the residual-error structure of the pooled oral-plus-long-acting-intramuscular rilpivirine popPK: the source $ERROR block computes Y0 = IPRED0 + ERR(1) when LAI = 0 and Y1 = IPRED1 * (1 + ERR(2)) when LAI = 1, i.e. an additive SD of sqrt(317) = 17.8 ng/mL on oral records and a proportional SD of sqrt(0.031) = 0.176 on intramuscular records. The model reproduces both branches with one combined error model whose two magnitudes are gated by the indicator, ruvAdd = addSdOral * ROUTE_ORAL and ruvProp = propSdIm * (1 - ROUTE_ORAL). Same residual-magnitude-switching role as Zierhut_2008_osteoprotegerin.R, Wang_2021_pertuzumab.R and Clegg_2024_tixagevimab_cilgavimab.R under ROUTE_IV, but with an oral-vs-non-oral contrast. The route’s structural branch – the source $PK block selects the oral zero-order duration D3 when LAI = 0 and the two intramuscular first-order rate constants otherwise – is carried by the dose record’s target compartment instead (central with rate = -2 for oral, depot plus depot2 for intramuscular), so the covariate is only read by the error model).
  • Comisar_2025_zavegepant.R (per-dose-record indicator selecting between the intranasal and oral absorption parameter sets of the three-compartment zavegepant popPK – F 5.1% vs 0.65%, ka 5.8 vs 0.81 1/h, D1 8.6 vs 57.2 min, and an oral-only 12.2 min absorption lag that collapses to zero via tlag = ROUTE_ORAL * exp(ltlag_oral); each switch is written as a log-scale convex combination exp((1 - ROUTE_ORAL) * l<param>_intranasal + ROUTE_ORAL * l<param>_oral + eta<param>) so the shared per-quantity eta of Comisar 2025 Table 3 rides on both routes. It additionally gates two effects that Comisar 2025 restricts to oral dosing: the itraconazole clearance effect, applied as (1 + e_conmed_itraconazole_cl * CONMED_ITRACONAZOLE * ROUTE_ORAL) = -27.9% on CL, and the fed-state bioavailability effect (1 + e_fed_fdepot_oral * FED * ROUTE_ORAL) = -50.9% on F.)
  • Willmann_2019_moxifloxacin.R (per-observation-record indicator selecting the residual-error magnitude of the pooled intravenous-plus-oral pediatric moxifloxacin popPK, with intravenous infusion as the 0 reference. The source $ERROR block writes three proportional magnitudes and orders the oral branch last so it overrides the study branch, Y = IPRED + W*EPS(1) / IF(STUD.EQ.11826) ... EPS(2) / IF(ROUT.EQ.1) ... EPS(3); the model reproduces the precedence by nesting the STUDY_PHASE3 selection inside (1 - ROUTE_ORAL). Same residual-magnitude-switching role as Thoueille_2024_rilpivirine.R, and the route’s structural branch is likewise carried by the dose record’s target compartment – depot for oral, central for the intravenous infusion – so the covariate is read by the error model alone.)
  • Li_2021_tedizolid.R (per-observation-record indicator scaling the log-scale residual SD by 2.01-fold after oral dosing – Li 2021 Table 1 ‘RV for oral data (fold)’ – multiplicatively with the STUDY_PHASE3 4.92-fold factor; intravenous infusion is the 0 reference and the route’s structural branch is carried by the dose record’s target compartment, so the covariate is read by the error model alone.)
  • Jayachandran_2021_tenofovir.R (per-dose-record indicator with rectal 1% TFV gel as the 0 reference, selecting the full route-specific parameter set of a plasma-plus-biophase tenofovir PK model: oral vs rectal first-order absorption rate constant, oral (F = 1) vs rectal (F = 0.102) bioavailability, the plasma residual-error structure (proportional for oral, combined additive-plus-proportional for rectal), and for each tissue / cellular biophase matrix the route-specific equilibration rate and plasma-to-matrix ratio, each written as exp(l<param>_oral * ROUTE_ORAL + l<param>_rectal * (1 - ROUTE_ORAL)). The PBMC ratio is additionally gated to zero for rectal dosing because all rectal-arm PBMC concentrations were below the limit of quantification. Source Supplementary Model Code 1 branches the same information as IF(ROUTE.EQ.2) with ROUTE = 1 oral, ROUTE = 2 rectal, so ROUTE_ORAL = as.integer(ROUTE == 1).)
• Notes: This is the per-dose-record covariate-equation indicator, distinct from the dosing-event cmt column that names the target compartment. Complementary to ROUTE_IV (IV-vs-SC), ROUTE_IP (IP-vs-non-IP), ROUTE_SC (SC-vs-IM), ROUTE_NGT (NGT-vs-oral) and ROUTE_VAGINAL (vaginal-vs-buccal) – none of those carries an oral-vs-non-oral contrast, and ROUTE_NGT is its near-inverse (it distinguishes delivery methods within oral administration, with oral as its reference). In a three-or-more-route design a single ROUTE_ORAL column suffices whenever the remaining routes are separated by the target compartment rather than by a covariate, as in Comisar 2025; register an additional ROUTE_* indicator only when two non-oral routes must be told apart inside the same compartment’s parameter equations. Scope: specific because which parameters switch by route, and what the 0 level pools, are paper-specific.

ROUTE_VAGINAL (canonical for vaginal-vs-buccal administration route indicator)

• Description: 1 = subject received vaginal misoprostol administration, 0 = buccal administration. Per-subject (study-fixed) covariate flagging the dosing route when a population analysis pools vaginal and buccal cohorts of a mucosally absorbed drug, with the buccal cohort as the structural reference (F = 1) and vaginal-specific effects encoded as a relative bioavailability multiplier and separate absorption rate constants per route x dose combination.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (buccal administration; the structural reference in Vorontsova 2022, whose relative bioavailability parameter F_v/b = 2.3 is the ratio of vaginal to buccal bioavailability with the buccal value fixed at 1 per the paper’s model-construction statement ‘we included a relative bioavailability term (F_v/b) for the vaginal route relative to buccal (F_b, assumed to be one)’).
• Source aliases:
  • ROUTE – used in Vorontsova_2022_misoprostol.R (IMPROVE-trial per-subject randomization indicator flagging vaginal vs buccal misoprostol for full-term labor induction).
• Example models: Vorontsova_2022_misoprostol.R (per-subject indicator selecting between four absorption rate constants (ka_buccal_25, ka_buccal_50, ka_vaginal_25, ka_vaginal_50 indexed by ROUTE_VAGINAL x DOSE) and applying the relative bioavailability multiplier F_v/b = 2.3 on the depot dose for vaginal subjects only via fdepot = 1 - ROUTE_VAGINAL + ROUTE_VAGINAL * fvb; inter-occasion variability on F_v/b is estimated across the two sampled dose events).
• Notes: This is the per-subject covariate-equation indicator, distinct from the dosing-event cmt column that names the target compartment (Vorontsova 2022 doses both routes into the same depot absorption compartment; only the ka value and bioavailability switch by route). Complementary to ROUTE_IV (IV-vs-SC), ROUTE_IP (IP-vs-non-IP), ROUTE_NGT (NGT-vs-oral), and ROUTE_SC (SC-vs-IM) – ROUTE_VAGINAL fills the vaginal-vs-buccal contrast for mucosally absorbed drugs (misoprostol, obstetric prostaglandins, hormonal contraceptives) that neither of those cover. Scope: specific because the vaginal-vs-buccal contrast and which parameters switch by route are paper-specific (Vorontsova 2022 carries route-specific ka values crossed with dose level and a single relative bioavailability multiplier; a future paper could carry different route effects on CL, V, or Tlag).

ROUTE_SC (canonical for subcutaneous-vs-intramuscular administration route indicator)

• Description: 1 = dose record (or subject occasion) administered by subcutaneous (SC) injection, 0 = intramuscular (IM) injection. Per-dose-record / per-occasion covariate flagging SC delivery when a population analysis pools IM and SC injection cohorts of an injectable long-acting formulation, with the IM cohort as the structural reference and SC-specific effects encoded as multiplicative factors on absorption half-lives, dose-split ratios, and relative bioavailability.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (IM injection; the structural reference in Kado 2020, whose “Relative structural model parameters for SC administration” are all ratios of SC-value / IM-value with IM’s F fixed at 1).
• Source aliases:
  • SC / route = SC – Kado 2020 Table 1 SC-relative parameter block (SC vs IM crossover of benzathine benzylpenicillin G).
• Example models: Kado_2020_benzathine_benzylpenicillin_g.R (per-dose-record indicator switching the three absorption half-lives (t1/2,abs-1/-2/-3), the transit half-life (t1/2,tr), the two dose-split ratios (RAT-transit, RAT-slowfast), and the relative bioavailability (F_SC = 0.957) between the IM reference and the SC structural values; all IIVs on absorption parameters are also route-specific per Kado 2020 Methods ‘PK modelling’), Tan_2025_cabotegravir.R (per-dose-record linear-effect (1 + e_route_sc_ka * ROUTE_SC) on the typical absorption rate constant of long-acting intramuscular cabotegravir, with theta = -0.563 -> a 56.3% slower ka for a subcutaneous depot per Tan 2025 Table 2 ‘Effect of subcutaneous location on k a (%)’. Because the drug is flip-flop the slower absorption RAISES the 8-weekly trough. Unlike Kado 2020 the SC value is not an assigned route: every injection was an intended ventrogluteal IM injection and ROUTE_SC records where post-injection ultrasound actually found the injectate – see notes).
• Notes: This is the per-dose-record covariate-equation indicator, distinct from the dosing-event cmt column that names the target compartment (Kado 2020 doses all routes into the same triple-depot absorption chain; only the parameter values switch by route). Complementary to ROUTE_IV (IV-vs-SC, where SC is the reference) and ROUTE_IP (IP-vs-non-IP, where non-IP including SC is the reference) – ROUTE_SC fills the SC-vs-IM contrast that neither of those cover. Scope: specific because the IM-vs-SC contrast and the set of parameters that differ by route are paper-specific. The canonical covers two distinct study designs, and the per-model covariateData entry states which one applies. In Kado 2020 the route is assigned – a deliberate SC-versus-IM crossover. In Tan 2025 it is observed: all injections were intended to be IM, and the column records the depot location found by a post-injection ultrasound, so a value of 1 flags an unintentionally ectopic deposition rather than a chosen route. The observed form can also absorb covariates that would otherwise appear in their own right: Tan 2025 reports that sex and BMI both lost significance once ROUTE_SC entered the model, because gluteal subcutaneous fat depth mediates both (see the covariatesDataExcluded block of Tan_2025_cabotegravir.R). A source that grades deposition on three levels (intramuscular / mixed / subcutaneous) but reports a single coefficient collapses to this binary with the mixed depots in the reference category; record that grouping in the model’s covariateData notes.

ROUTE_SUBLINGUAL (canonical for sublingual-vs-buccal oromucosal administration indicator)

• Description: 1 = the oromucosal dose was directed under the tongue (sublingual), 0 = into the buccal cavity. Per-dose-record indicator for oromucosal products whose absorption rate depends on where in the mouth the dose is placed.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (buccal administration, the standard use of the Nicorette QuickMist mouth spray in Olsson Gisleskog 2021).
• Source aliases:
  • TMT (TMT.EQ.9003, control-stream flag FLAG_KA) – Olsson Gisleskog 2021 ESM mouth-spray control stream.
• Example models: OlssonGisleskog_2021_nicotine_mouthspray.R (selects the oromucosal absorption rate constant: ka = 86.0 1/h sublingual vs 15.9 1/h buccal; Table 5 footnote c).
• Notes: Follows the ROUTE_* policy of documenting the reference route explicitly. Complementary to ROUTE_VAGINAL (vaginal vs buccal), which also takes buccal as its reference but contrasts a different site. Only the absorption rate switches in the founding model; the fraction swallowed and the bioavailabilities do not.

DEVICE_AI (canonical for autoinjector-vs-prefilled-syringe SC device indicator)

• Description: 1 = subject’s SC dose delivered via autoinjector (AI), 0 = prefilled syringe (PFS). Per-subject (study-fixed) covariate flagging the SC delivery device when a model carries device-specific PK effects.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (PFS).
• Source aliases: “Formulation = AI” (categorical effect column in Yu 2022 covariate equations).
• Example models: Yu_2022_ofatumumab.R (exponential effect on k_e(P) and R0), Diep_2026_donidalorsen.R (Phoenix linear-effect (1 + e_device_ai_ka * DEVICE_AI) on the typical SC absorption rate constant with theta = +0.262 -> multiplier 1.262 for autoinjector vs vial-and-syringe reference; characterized in the ISIS 721744-CS9 single-dose bioequivalence cohort).
• Notes: Set to 0 (PFS / vial reference) for IV subjects, since the device is undefined for IV; the IV-specific effects are captured by ROUTE_IV instead. Scope: specific because the AI / vial / PFS contrast and which parameters it affects depend on the study’s device-comparison design.

INJSITE_ARM (canonical for SC injection-site = arm indicator)

• Description: 1 = subject’s SC dose injected into the arm, 0 = abdomen (the universal SC reference site across the popPK literature). Per-dose-record covariate flagging the SC injection site when a population analysis estimates site-specific absorption parameters.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (abdomen).
• Source aliases: paper narrative “arm” / “abdomen” subgroup labels driving site-specific ka in Diep 2022.
• Example models: Diep_2022_eplontersen.R (additive log-shift e_injsite_arm_ka = log(ka_arm / ka_ab) on the typical absorption rate constant: ka_arm = 0.217 1/h vs ka_ab = 0.282 1/h, ~30% higher ka for abdomen; INJSITE_ARM = 1 selects the arm typical value), Diep_2026_donidalorsen.R (Phoenix linear-effect (1 + e_injsite_arm_ka * INJSITE_ARM) on the typical absorption rate constant with theta = -0.338 -> multiplier 0.662 for arm; the paper’s reference category is “abdomen or thigh” rather than “abdomen” alone, but is consistent with the canonical reference because abdomen is the universal SC reference site and the thigh effect is pooled into the reference category by the Diep 2026 model), CarlssonPetri_2018_semaglutide.R (multiplicative CL/F ratio 1.08 for upper-arm vs abdomen injection site, encoded as e_site_arm_cl^INJSITE_ARM per Table S3; per-subject dominant injection-site indicator since the paper assigns each subject their most frequently used site), Perlstein_2025_risperidone_tv46000.R (categorical multiplicative effect (1 + e_injsite_arm_ka1)^INJSITE_ARM on the fast direct-release rate constant of the TV-46000 subcutaneous risperidone depot, with theta = 0.331 -> 33% higher ka1 for the upper arm; per-dose-record, since the paper’s conclusion is that a patient may alternate sites between injections without an exposure consequence), Marier_2010_teduglutide.R (read together with INJSITE_THIGH as a POOLED non-abdomen indicator injsite_nonabd = INJSITE_ARM + INJSITE_THIGH that selects the second of two parallel stratum-suffixed typical absorption rate constants, lka_armthigh = log(0.206 1/h) in place of lka_abdomen = log(0.299 1/h); Marier 2010 Table III mod 16 collapsed its three-category injection-site model to abdomen-versus-other at a cost of only dMOF = +1.552, so arm and thigh share one estimated value and one shared BSV), Marier_2021_teduglutide.R (pooled with INJSITE_THIGH into one non-abdomen stratum, injsite_nonabd <- INJSITE_ARM + INJSITE_THIGH, which multiplies Ka by 0.766, the absorption lag time by 1.458 and the relative bioavailability by 0.936 per Marier 2021 Table S8; supplement control stream EXLOC1 = 0).
• Notes: Specific scope because the arm-vs-abdomen contrast is paper-specific. Sister canonical to INJSITE_THIGH (thigh-vs-abdomen indicator; founded 2026-07-09 alongside the CarlssonPetri_2018_semaglutide extraction). When a source paper pools arm and thigh into a single non-abdomen absorption stratum, encode the pool as the sum of the two indicators inside model rather than registering a separate INJSITE_OTHER canonical: the family’s members are mutually exclusive with all-zero denoting abdomen, so the sum is already the 0/1 non-abdomen indicator, and keeping the two source-level columns preserves which anatomical sites the cohort actually used (founding example Marier_2010_teduglutide.R). Per-administration rather than per-subject in the Diep 2022 / 2026 usage (a subject in a multi-dose simulation can switch SC injection sites between doses; supply the indicator on each dose record), but per-subject in the Carlsson Petri 2018 usage (each subject’s dominant / most frequently used injection site was used as the covariate value). Distinct from ROUTE_IV (IV vs SC route, not within-SC site) and from DEVICE_AI (autoinjector vs prefilled syringe, device rather than anatomical site).

INJSITE_THIGH (canonical for injection-site = thigh indicator)

• Description: 1 = subject’s dose injected into the (antero-lateral) thigh, 0 = abdomen (the universal SC reference site across the popPK literature) or a different non-thigh injection site. Per-dose-record OR per-subject covariate flagging the injection site when a population analysis estimates site-specific absorption, bioavailability or clearance effects. Most members of the family come from subcutaneous popPK, but the canonical also covers intramuscular injection-site contrasts, where the reference complement is typically the gluteal region rather than the abdomen (founding IM example Clegg_2024_tixagevimab_cilgavimab.R).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (abdomen for SC models; gluteal region for IM models – read the per-model covariateData entry, which states which complement the source paper used).
• Source aliases: paper narrative “thigh” / “abdomen” subgroup labels driving site-specific PK parameters; ACTIV2 (a study-name column that Clegg 2024 supplementary Table S2 states is “used to distinguish the site of IM administrations”, the ACTIV-2 study being the only one that injected into the thigh) – used in Clegg_2024_tixagevimab_cilgavimab.R, Clegg_2024_tixagevimab.R, Clegg_2024_cilgavimab.R.
• Example models: Clegg_2024_tixagevimab_cilgavimab.R, Clegg_2024_tixagevimab.R, Clegg_2024_cilgavimab.R (log-additive effect on absolute intramuscular bioavailability, f(depot) <- exp(lfdepot + e_injsite_thigh_fdepot * INJSITE_THIGH) with e_injsite_thigh_fdepot = 0.416 / 0.378 / 0.414 per Clegg 2024 Table 1 row beta_FIM(ACTIV2_1), i.e. 51.6% / 45.9% / 51.3% higher bioavailability for a thigh injection than for the gluteal reference; the source column is the ACTIV-2 study indicator, which supplementary Table S2 documents as a proxy for injection site. Because the effect is log-additive on an already-high gluteal FIM, the implied thigh bioavailability of the summed AZD7442 analyte is 0.671 * exp(0.416) = 1.017, marginally above unity – see the model’s vignette Errata), CarlssonPetri_2018_semaglutide.R (multiplicative CL/F ratio 1.04 for thigh vs abdomen injection site, encoded as e_site_thigh_cl^INJSITE_THIGH per Table S3; per-subject dominant injection-site indicator since the paper assigns each subject their most frequently used site), Bjornsson_2023_buprenorphine.R (multiplicative gate (1 - INJSITE_THIGH) on the CAM2038 Q1W fast-pathway dose fraction Fq1w1, i.e. a thigh injection routes the entire weekly buprenorphine depot dose through the slow absorption pathway; Bjornsson 2023 Results 3.1.1 estimated Fq1w1 for thigh injections “close to 0” and subsequently fixed it to 0, so the effect is encoded as a structural gate rather than an estimated coefficient; per-dose-record covariate, and the reference complement here is buttock / abdomen / upper arm, which Discussion 4.1 reports as indistinguishable from one another), Glatard_2025_octreotide.R (fractional change -0.351 on the fast-release mean absorption time MAT_fast of the CAM2029 octreotide depot relative to abdominal injection, per Table 3 / Eq 11; supplied per dose record alongside INJSITE_BUTTOCK, both 0 meaning abdomen), Riggs_2012_albinterferon.R (the site acts in TWO places at once for self-administered subcutaneous albinterferon alfa-2b: a multiplicative 1.09 on the absorption rate constant AND a relative bioavailability of 1.07 versus the abdomen, encoded as f(depot) <- exp(lfdepot) * e_injsite_thigh_fdepot^INJSITE_THIGH with lfdepot anchored at log(1) because CL/F and V/F already carry the abdomen-site bioavailability; per Table I rows “Thigh as injection site” and “Frel for thigh as injection site”), Marier_2010_teduglutide.R (pooled with INJSITE_ARM into a single non-abdomen absorption stratum selecting the stratum-suffixed lka_armthigh = log(0.206 1/h) in place of lka_abdomen = log(0.299 1/h); see the INJSITE_ARM entry for the pooling rule. In Marier 2010 the thigh arm of the pool carries most of the patient rather than healthy-volunteer data, because the phase II Crohn’s study CL0600-008 and the phase III short-bowel-syndrome study CL0600-004 both permitted self-administration into the abdomen or the thigh), Marier_2021_teduglutide.R (pooled with INJSITE_ARM into the single non-abdomen stratum of Marier 2021 Table S8 – Ka x 0.766, lag time x 1.458, relative bioavailability x 0.936 – see the INJSITE_ARM entry).
• Notes: Specific scope because the thigh-vs-abdomen contrast is paper-specific. Sister canonical to INJSITE_ARM (arm-vs-abdomen). Per-administration or per-subject depending on the paper’s dosing granularity – in Carlsson Petri 2018 the covariate is per-subject because ‘the most frequently used injection site for an individual patient was used as the covariate value’ (Methods); Bjornsson 2023 supplies it per dose record because a single subject’s weekly depot injections can rotate between sites. The reference complement is paper-defined: strictly abdomen in Carlsson Petri 2018, the pooled buttock / abdomen / upper arm group in Bjornsson 2023 (where those three sites showed no absorption differences), and the gluteal region in Clegg 2024 (the only intramuscular member of the family so far). A thigh effect may be a structural gate rather than an estimated coefficient when the source paper fixes the thigh parameter at a boundary value. Distinct from ROUTE_IV (IV vs SC route, not within-SC site) and from DEVICE_AI (autoinjector vs prefilled syringe, device rather than anatomical site). Founded alongside the CarlssonPetri_2018_semaglutide extraction.

STUDY_93NNBT007 (canonical for the 93NNBT007 single-dose oral nicotine microtablet study indicator)

• Description: 1 = record from study 93NNBT007, in which nicotine microtablets were swallowed whole as a single dose; 0 = study 92NNBT005, in which the microtablets were chewed before swallowing and dosed repeatedly. Subject-level (time-fixed).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (study 92NNBT005).
• Source aliases:
  • STUDY (STUDY.EQ.6; STUDY.EQ.5 is 92NNBT005) – Olsson Gisleskog 2021 ESM oral control stream.
• Example models: OlssonGisleskog_2021_nicotine_oral.R (switches the absorption structure from direct first-order absorption to a three-compartment transit chain with ktr 3.60 1/h, the bioavailability from 39.5% to 22.3%, and turns off the time-dependent clearance increase).
• Notes: Well-formed member of the auto-approved STUDY_<id> family. The paper attributes the difference to how the tablets were taken (whole vs chewed), but the model keys it on the study, so the study label is kept.

STUDY_97NNIN024 (canonical for the 97NNIN024 nicotine-inhaler study indicator)

• Description: 1 = record from nicotine-inhaler study 97NNIN024; 0 = either of the other two inhaler studies pooled in Olsson Gisleskog 2021. Subject-level (time-fixed).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the other inhaler studies).
• Source aliases:
  • TMT (codes 28001 and 28002, control-stream flag FL_STUDY28) – Olsson Gisleskog 2021 ESM inhaler control stream; Table 5 row ‘F increase, Study 97NNIN024’.
• Example models: OlssonGisleskog_2021_nicotine_inhaler.R (multiplies both the oromucosal and the oral bioavailability by 1 + 0.660).
• Notes: Well-formed member of the auto-approved STUDY_<id> family.

STUDY_INNOVATV202 (canonical for the innovaTV 202 tisotumab vedotin phase I/II study indicator)

• Description: 1 = record from innovaTV 202 (NCT02552121, sponsor code GEN702), an open-label phase I/II dose-escalation / expansion trial of tisotumab vedotin in locally advanced or metastatic solid tumors; 0 = any other study in the same pooled analysis. Subject-level (time-fixed).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (innovaTV 201, 204 and 207 pooled).
• Source aliases:
  • innovaTV 202 study – Gibiansky 2022 Table S1 (‘[theta36 if innovaTV 202 study]’); Study 702 / sigma_702 in the FDA BLA 761208 review Tables 47-49.
• Example models: Gibiansky_2022_tisotumab.R (multiplies the ADC residual-error SD by 1.42; no effect on any PK parameter).
• Notes: Well-formed member of the auto-approved STUDY_<id> family. Acts on the residual error only, so simulations of a new population set it to 0.

STUDY_NC003 (canonical for the NC-003 pretomanid combination-regimen study cohort indicator)

• Description: 1 = subject enrolled in NC-003, a phase 2a study of pretomanid in combination regimens containing bedaquiline, clofazimine and/or pyrazinamide; 0 = subject from any other study in the pooled analysis. Subject-level (time-fixed).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (any other pooled study).
• Source aliases:
  • NC3 – Salinger 2019 supplemental Table S3 $INPUT column; study number 103 in the supplemental figure labelling.
• Example models: Salinger_2019_pretomanid.R (scales the MAGNITUDE of the mean-transit-time random effect rather than any typical value: sdScaleMtt <- exp(e_study_nc003_etalmtt * STUDY_NC003) with e_study_nc003_etalmtt = -0.645, so the MTT eta is multiplied by exp(-0.645) = 0.525 in NC-003, i.e. absorption variability is about half that of the other studies).
• Notes: Well-formed member of the auto-approved STUDY_<id> canonical family. Unusual among registered STUDY_ members in acting purely on a random-effect magnitude, with no fixed-effect counterpart – Salinger 2019 introduced per-study absorption terms because fed/fasted conditions were not well controlled in NC-002, NC-003, NC-005, NC-006 and Nix-TB, and for NC-003 only the variability term was retained. Sibling to STUDY_NC005 and STUDY_NIXTB from the same pooled analysis. STUDY_NC002 and STUDY_NC006 were tested in the same modelling step but retained no effect and are therefore not registered.

STUDY_NC005 (canonical for the NC-005 pretomanid combination-regimen study cohort indicator)

• Description: 1 = subject enrolled in NC-005, a phase 2b study in which drug-susceptible-TB patients received bedaquiline + pretomanid + pyrazinamide and MDR-TB patients received bedaquiline + pretomanid + moxifloxacin + pyrazinamide (BPaMZ); 0 = subject from any other study in the pooled analysis. Subject-level (time-fixed).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (any other pooled study).
• Source aliases:
  • NC5 – Salinger 2019 supplemental Table S3 $INPUT column; study number 105 in the supplemental figure labelling.
• Example models: Salinger_2019_pretomanid.R (two absorption effects: e_study_nc005_ka^STUDY_NC005 with e_study_nc005_ka = 0.186, a five-fold slowing of the first-order absorption rate constant, and e_study_nc005_mtt^STUDY_NC005 with e_study_nc005_mtt = 6.95e-07, a boundary estimate that collapses the transit-chain mean transit time to about 1e-06 h. The two together reproduce the study’s anomalously late median steady-state Tmax of 6.5 h, which the slow absorption rate supplies on its own).
• Notes: Well-formed member of the auto-approved STUDY_<id> canonical family. The MTT coefficient is a genuine boundary estimate, not a transcription error: Salinger 2019 Table 1 reports it as 6.95E-07 with a confidence interval of -0.0350 to 0.0350 that cannot bracket a positive parameter entering the model through a logarithm, and the supplemental Table S4 summary confirms the consequence – the NC-005 median MTT + 1/KA of 3.8 h equals 1/KA = 3.9 h on its own, leaving no room for a transit delay. A data set carrying STUDY_NC005 = 1 therefore drives the transit rate constant to roughly 3e+06 1/h; this solves, but a user tightening solver tolerances should expect the transit chain to behave as an instantaneous pass-through. Sibling to STUDY_NC003 and STUDY_NIXTB.

STUDY_NIXTB (canonical for the Nix-TB study cohort indicator)

• Description: 1 = subject enrolled in Nix-TB (NCT02333799), the phase 3 open-label study of bedaquiline + pretomanid + linezolid (BPaL) in XDR-TB or treatment-intolerant / non-responsive MDR-TB; 0 = subject from any other study in the pooled analysis. Subject-level (time-fixed).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (any other pooled study).
• Source aliases:
  • NIX – Salinger 2019 supplemental Table S3 $INPUT column; study number 201 in the supplemental figure labelling.
• Example models: Salinger_2019_pretomanid.R (the most structurally load-bearing STUDY_ member in the registry: it supplies a 1.54-fold relative-bioavailability effect, inflates the bioavailability random effect via exp(0.919 * STUDY_NIXTB), selects between two pairs of Box-Cox shape parameters for the clearance and central-volume random effects, and gates a SECOND piecewise-constant clearance breakpoint at week 6 through clStep2 <- STUDY_NIXTB * (tafdNow >= tclchange2)).
• Notes: Well-formed member of the auto-approved STUDY_<id> canonical family. Nix-TB was added at the final stage of Salinger 2019’s staged model build, after a visual predictive check showed the pre-Nix model could not describe it, so the study indicator carries structure rather than only a nuisance offset – in particular it is the only condition under which the model’s lcl_late2 / ltclchange2 second clearance step is active, and it changes the SHAPE of two random-effect distributions rather than just their location. The step-up in clearance from week 6 was retained after the alternative explanation (rising body weight as patients recovered) was tested and rejected. Sibling to STUDY_NC003 and STUDY_NC005.

DUR_SL_CONTACT (canonical for contact time of a sublingual dosage form under the tongue)

• Description: Time, in minutes, that a sublingual dosage form (film or tablet) remained in contact with the sublingual mucosa under the tongue before it had dissolved or was removed, as recorded at each sublingual administration. Per-dose-record covariate.
• Units: min
• Type: continuous
• Scope: general
• Reference category: n/a – continuous; the founding model uses power scaling normalised to 2 min.
• Source aliases: “contact time under the tongue”, “contact time for sublingual administration” (Agbo 2021 Table 1 and Table 3 row “Contact time under the tongue for sublingual film on ka for sublingual administration”).
• Example models: Agbo_2021_apomorphine.R (power effect -0.194 on the sublingual absorption rate constant, reference 2 min; Figure 2’s typical patient uses 3 min).
• Notes: A dosing-procedure covariate rather than a formulation indicator: distinct from FORM_ODT and the other FORM_* columns, which name the dosage form, and from NTIME / TAD, which are times since dose. General scope because sublingual residence time is a recurring determinant of sublingual absorption across drugs. Named in the DUR_<event> duration family (sibling of DUR_MAP_BELOW_PRESURG); the units live in the Units field, not the name, so the column is DUR_SL_CONTACT rather than a _MIN-suffixed form.

STUDY_APLIOS (canonical for APLIOS bioequivalence study indicator)

• Description: 1 = subject enrolled in the APLIOS bioequivalence study (NCT03560739; phase 2; ofatumumab AI vs PFS in RMS), 0 = other study in the Yu 2022 pooled analysis.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-APLIOS studies: OMS115102, MIRROR, ASCLEPIOS I, ASCLEPIOS II).
• Source aliases: “Study = APLIOS” (categorical effect column in Yu 2022 covariate equations).
• Example models: Yu_2022_ofatumumab.R (exponential effect on Emax of B cell lysis).
• Notes: Captures a between-study shift in the maximum B-cell lysis stimulatory effect not explained by the other covariates in the final model.

STUDY_CTH103 (canonical for apomorphine sublingual film study CTH-103 indicator)

• Description: 1 = subject enrolled in study CTH-103 (phase I, three-dose active-comparator, placebo-controlled randomised crossover study in healthy subjects comparing apomorphine sublingual film 10, 15 and 25 mg, formulation K, with subcutaneous apomorphine 2, 3 and 4 mg), 0 = any other study in the Agbo 2021 pooled analysis.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the other eight pooled studies: CTH-104, -105, -106, -107, -200, -201, -203, -301).
• Source aliases: “Study CTH-103” (Agbo 2021 Table 3 row “Study CTH-103 on V2/F”).
• Example models: Agbo_2021_apomorphine.R (proportional shift -0.555 on the apomorphine central volume V2/F).
• Notes: Added by the source authors at the base-model stage to absorb the very high peak concentrations observed in CTH-103, particularly after subcutaneous dosing. It is a between-study artefact rather than a patient characteristic, so simulations of new subjects set it to 0. Follows the STUDY_<id> auto-approve family.

STUDY_NAPOLI1 (canonical for NAPOLI-1 study / NAPOLI-1 drug-product manufacturing-site indicator)

• Description: 1 = subject enrolled in NAPOLI-1 (NCT01494506; phase 3; nanoliposomal irinotecan with or without 5-FU/LV in metastatic pancreatic cancer previously treated with gemcitabine), 0 = subject enrolled in one of the five PharmaEngine (PEI) phase I-II nal-IRI studies pooled in the Adiwijaya 2017 analysis (PEP0201, PEP0203, PEP0206, and two others listed in that paper’s Table S1).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the pooled PharmaEngine phase I-II studies / PEI manufacturing site).
• Source aliases:
  • mfg == NAPOLI (vs mfg == PEI) – Adiwijaya 2017 Tables S6 and S7 theta row labels. The source names the covariate manufacturing site, not study: nal-IRI used in the phase III study was manufactured at a different site from the material used in the phase I-II studies, and Table S3 gives “Potential difference in manufacturing of nal-IRI in phase 2 and phase 3 studies” as the rationale. The two levels are perfectly collinear with the study (258 of 353 subjects are NAPOLI-1), so the canonical study indicator carries the manufacturing site rather than minting a parallel MFGSITE_* column that could never take a different value.
• Example models: Adiwijaya_2017_irinotecan_liposomal.R (exponential effects on four parameters at once: total-irinotecan V1 -0.172 and CL -0.189, the SN-38 formation rate constant +3.79e-05, and the encapsulated-SN-38 mass fraction -0.615 – the last being by far the largest, a 46% lower contaminant fraction in the phase III material, which is the effect the manufacturing rationale most directly predicts).
• Notes: Scope is specific because the indicator names one trial. A model that needs the manufacturing site to vary independently of the study – e.g. a later analysis pooling several trials that each drew on both sites – should propose a separate covariate rather than reusing this one; see the Source aliases note for why the two are indistinguishable in the founding analysis.

STUDY_MRX4002 (canonical for the MRX4-002 intravenous contezolid acefosamil phase 1 study indicator)

• Description: 1 = healthy volunteer from MRX4-002, the single- and multiple-ascending-dose phase 1 study of intravenous contezolid acefosamil (150-2400 mg), 0 = subject from the oral contezolid studies MRX-I-02 (phase 1) or MRX-I-03 (phase 2 ABSSSI) pooled in Bulitta 2024. Subject-level (time-fixed).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (oral contezolid studies).
• Source aliases: none; Bulitta 2024 Table 2 footnote h distinguishes the residual errors “after IV dosing of CZA in healthy volunteers” and “after oral dosing of contezolid in healthy volunteers and phase 2 patients”.
• Example models: Bulitta_2024_contezolid.R (selects the IV-study additive and proportional residual errors for contezolid, 0.0392 mg/L and 0.192, and for MRX-1320, 0.00636 mg/L and 0.117, instead of the oral-study values; acts on no structural parameter).
• Notes: Well-formed member of the auto-approved STUDY_<id> canonical family. The study and the dosing route coincide in the source (MRX4-002 is the only IV study), so the indicator can equally be read as “subject received IV contezolid acefosamil”.

STUDY_MIRROR (canonical for MIRROR dose-finding study indicator)

• Description: 1 = subject enrolled in the MIRROR dose-finding study (NCT01457924; phase 2; SC ofatumumab dose-ranging in RRMS), 0 = other study in the Yu 2022 pooled analysis.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-MIRROR studies: OMS115102, APLIOS, ASCLEPIOS I, ASCLEPIOS II).
• Source aliases: “Study = MIRROR” (categorical effect column in Yu 2022 covariate equations).
• Example models: Yu_2022_ofatumumab.R (exponential effect on B cell elimination rate kout).
• Notes: Captures a between-study shift in the B cell elimination rate not explained by the other covariates in the final model.

STUDY_MICP220 (canonical for benralizumab phase II study MI-CP220 indicator)

• Description: 1 = subject enrolled in the phase II benralizumab study MI-CP220 (NCT01238861), 0 = subject enrolled in any of the other studies of the pooled analysis. MI-CP220 is one of the two phase II studies that administered benralizumab 2-200 mg subcutaneously Q4W/Q8W, and in both benralizumab pooled analyses that use this indicator it is the single study carrying both its own absolute subcutaneous bioavailability and its own residual-error magnitude.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (any of the other pooled studies – the other 11 in the Jin 2025 analysis, the other 8 in the Yan 2019 analysis).
• Source aliases:
  • STUDYN – the numeric study identifier column (Jin 2025 Resource 1); the indicator is derived from it.
  • Fa1S220, study 220 – the parameter-name and footnote spellings Jin 2025 Resource 10 uses for the study-specific bioavailability.
  • CP220, study CP220 – the short form Yan 2019 Table 5 uses in the row label “Change in F with study CP220, fraction”.
• Example models: Jin_2025_benralizumab.R (selects an absolute subcutaneous bioavailability of 0.457, with its own lognormal IIV of omega 0.373, in place of the reference 0.539; the same study also has its own log-scale additive residual error of 0.549 versus 0.175 for the other early studies, encoded as the separate Cc_micp220 endpoint).
  • Jin_2025_benralizumab.R (selects an absolute subcutaneous bioavailability of 0.457, with its own lognormal IIV of omega 0.373, in place of the reference 0.539; the same study also has its own log-scale additive residual error of 0.549 versus 0.175 for the other early studies, encoded as the separate Cc_micp220 endpoint).
  • Yan_2019_benralizumab.R (the legacy nine-study analysis that Jin 2025 updates: selects an absolute subcutaneous bioavailability of 0.490, with its own lognormal IIV of omega 0.350, in place of the reference 0.589; the same study likewise has its own residual error of 0.545 versus 0.250 for the other early studies, encoded as the separate Cc_micp220 endpoint).
• Notes: Sibling of STUDY_AMES, the other study-specific bioavailability stratum in the Jin 2025 model; the two are mutually exclusive per subject and a subject in neither takes the reference bioavailability. Distinct from a bare between-study shift on a disposition parameter – here the study effect is on bioavailability, which is why the strata are estimated as separate absolute fractions rather than as relative multipliers. That reading is what lets the two analyses be compared directly: Yan 2019 Table 5 labels the row “Change in F with study CP220” but tabulates an absolute fraction, and the resulting MI-CP220-to-reference bioavailability ratio is 0.569 / 0.671 = 0.848 in the Yan 2019 base model and 0.457 / 0.539 = 0.848 in the Jin 2025 final model – an agreement the relative-multiplier reading of the Yan 2019 row could not produce.

STUDY_MICP200 (canonical for inebilizumab phase I systemic-sclerosis study MI-CP200 indicator)

• Description: 1 = subject enrolled in the phase I single-ascending-dose inebilizumab study MI-CP200 (NCT00946699; single IV doses of 0.1-10 mg/kg in adults with systemic sclerosis), 0 = subject enrolled in either of the other two studies of the pooled analysis (CD-IA-MEDI-551-1102 in relapsing MS, CD-IA-MEDI-551-1155 in NMOSD). Subject-level (time-fixed).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (studies 1102 and 1155; the most common realisation per Yan 2022 Eq. 4).
• Source aliases:
  • Study CP200, CP200 – Yan 2022 Table 6 row label ‘Study CP200 on Vmax (%)’ and Table 4 model 5 ‘SSc Study on Vmax’.
• Example models: Yan_2022_inebilizumab.R (fractional-change effect on the baseline Michaelis-Menten Vmax, vmax * (1 + 2.10 * STUDY_MICP200), so the typical MI-CP200 subject has a 3.1-fold higher Vmax than subjects in the other two studies).
• Notes: Well-formed member of the auto-approved STUDY_<id> family, spelled like its STUDY_MICP220 sibling. The indicator is a STUDY effect, and all MI-CP200 subjects had systemic sclerosis, so study and disease are confounded (Yan 2022 Section 3.3). Do not reinterpret it as a systemic-sclerosis disease indicator. The Discussion’s “2.1-fold higher” paraphrases the tabulated 210% fractional change. Eq. 4 (P = theta1 * (1 + theta2 * Factor)) with theta2 = 2.10 gives a 3.1-fold ratio.

STUDY_AMES (canonical for the benralizumab AMES autoinjector-versus-prefilled-syringe study indicator)

• Description: 1 = subject enrolled in the phase I AMES study (NCT02968914), which compared benralizumab PK exposure after administration by accessorized prefilled syringe versus autoinjector in 180 healthy volunteers given a single 30 mg subcutaneous dose, 0 = subject enrolled in any of the other studies of the pooled analysis.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (any of the other 11 studies in the Jin 2025 pooled analysis).
• Source aliases:
  • STUDYN – the numeric study identifier column (Jin 2025 Resource 1); the indicator is derived from it.
  • Fa1S30, FaslS30, AMES – the parameter-name and footnote spellings Jin 2025 Resource 10 uses for the study-specific bioavailability (the FaslS30 spelling in the Resource 10 row label is a typo for Fa1S30, which the same table’s footnote uses).
• Example models: Jin_2025_benralizumab.R (selects an absolute subcutaneous bioavailability of 0.688, with its own lognormal IIV of omega 0.144, in place of the reference 0.539).
• Notes: Sibling of STUDY_MICP220. This stratum exists for a diagnostic reason worth preserving: external validation of the legacy model under-predicted the AMES concentrations, and Jin 2025 Results 3.2 tested injection site (ILOC), healthy status and an AMES study effect on Fa1, CL and V2 before concluding that ‘the model featuring an effect of AMES on Fa1 performed best’. The injection-site and healthy-status alternatives were rejected – the healthy-status effect improved AMES but degraded the other healthy-volunteer study – so the study indicator is deliberately narrower than the physiological covariates it displaced, and must not be re-interpreted as a healthy-volunteer or injection-device effect.

STUDY_IM018003 (canonical for the IM018003 BMS-986166 multiple-ascending dose study indicator)

• Description: 1 = subject from Study IM018003, the multiple-ascending dose study of BMS-986166 (0.25, 0.75 or 1.5 mg once daily for 28 days); 0 = Study IM018001, the single-ascending dose study (0.75, 2 or 5 mg). Subject-level (time-fixed).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (Study IM018001).
• Source aliases:
  • MAD / SAD subscripts on F1 and F2 in Bihorel 2021 Table 1 (F1 MAD, F2 MAD,LOW, …) – STUDY_IM018003 = 1 for the MAD rows.
• Example models: Bihorel_2021_BMS986166.R (selects the study-specific relative bioavailability of BMS-986166, 1 FIXED vs 0.764, the fraction of the molar dose absorbed as BMS-986166-P, 0.0771 vs 0.0797, and the study-specific low-exposure fold shifts on both).
• Notes: Well-formed member of the auto-approved STUDY_<id> family. Bihorel 2021 Discussion warns that the F1 difference may compensate for time-dependent PK changes after repeated dosing rather than reflect a true bioavailability difference, which is why the model keys it on the study and not on a single- vs multiple-dose regimen flag.

STUDY_IRIDIUM (canonical for the IRIDIUM Phase III asthma study indicator)

• Description: 1 = subject enrolled in IRIDIUM (NCT02571777), the 52-week Phase III study of the indacaterol/glycopyrronium/mometasone furoate (IND/GLY/MF) and indacaterol/mometasone furoate (IND/MF) Breezhaler fixed-dose combinations in inadequately controlled asthma; 0 = subject enrolled in any other study of the pooled analysis (PALLADIUM, NCT02554786, or the E2201 device-bridging study, NCT01555151).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (PALLADIUM or E2201).
• Source aliases: Study effect on Vc/F in IRIDIUM – the Bartels 2021 Table 5 row label; the analysis dataset column itself is not named.
• Example models: Bartels_2021_indacaterol.R (vc *= exp(0.25 * STUDY_IRIDIUM)), Bartels_2021_mometasoneFuroate.R (vc *= exp(0.19 * STUDY_IRIDIUM)).
• Notes: An empirical study effect on the apparent central volume that absorbs the higher maximal concentrations seen in PALLADIUM (Bartels 2021 Sect. 3.1 and 3.6); the authors attribute it to no mechanism. The IND/GLY/MF formulation was studied only in IRIDIUM, so in the source data FORM_MF_INDGLYMF = 1 implies STUDY_IRIDIUM = 1; the two effects were nonetheless estimated separately on the MF Vc/F.

STUDY_ARROW_PART2 (canonical for ARROW PK Substudy Part 2 study indicator)

• Description: 1 = record from the ARROW PK Substudy Part 2 (the antiretroviral-dosing substudy of the ARROW trial in African children living with HIV-1), 0 = record from any other study pooled into the analysis. Per-record (study-fixed) binary indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (all other pooled studies; relative bioavailability fixed at 1).
• Source aliases:
  • F1 study term – used in Chandasana_2024b_abacavir.R; Chandasana 2024 Table 1 names the two estimates “F, tablet ARROW PK Substudy Part 2” and “F, solution ARROW PK Substudy Part 2”.
• Example models: Chandasana_2024b_abacavir.R (study-specific relative bioavailability on the abacavir depot, applied jointly with FORM_SOLUTION so that the ARROW Part 2 tablet gets 1.62 and the ARROW Part 2 oral solution gets 1.75, both relative to the all-other-studies reference of 1; encoded on the log scale as exp(e_arrow_tab_fdepot * STUDY_ARROW_PART2 * (1 - FORM_SOLUTION) + e_arrow_sol_fdepot * STUDY_ARROW_PART2 * FORM_SOLUTION)).
• Notes: The source model included this term because abacavir exposure observed in ARROW PK Substudy Part 2 was substantially higher than in the other pooled studies despite similar doses and formulations (Chandasana 2024 “ABC Pediatric PopPK Model”), i.e. it absorbs an unexplained between-study bioavailability shift rather than a mechanistic covariate effect. The effect is formulation-specific within the substudy, so it is always paired with a formulation indicator. Sibling of the other STUDY_<id> members; scoped specific because the reference category is the paper’s own pooled-study set.

STUDY_PHASE3 (canonical for pooled phase III study-stratum indicator)

• Description: 1 = record from a phase III study in the pooled analysis, 0 = record from an earlier-phase (phase I / II) study. Per-record (study-fixed) binary indicator. Use when a source paper pools early-phase and phase III data and stratifies a model component by development phase as a whole rather than by individual trial – the usual reason being that the phase III trials share a sparse (trough-only) sampling design while the early-phase studies contributed the rich profiles, so the two strata need different residual-error models.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (the pooled phase I / II studies).
• Source aliases:
  • Study phase – Rich 2026 Table 1 residual-error row labels “Proportional error phase III studies” and “Proportional error phase I/II studies” (Rich_2026_momelotinib.R); the source dataset column name is not published.
  • STUD – Willmann 2019 supplementary NONMEM $INPUT column, an integer Bayer study number branched in $ERROR as IF(STUD.EQ.11826), with 11826 the phase I study and 11643 the phase III study. Derive STUDY_PHASE3 = as.integer(STUD == 11643).
• Example models: Rich_2026_momelotinib.R (selects the residual-error model only, for both the momelotinib and the M21 output: proportional-only in the phase III trials SIMPLIFY-1 / SIMPLIFY-2 / MOMENTUM, proportional-plus-additive in the phase I clinical-pharmacology studies and the phase II study; it touches no structural or covariate parameter), Vaddady_2020_doravirine.R (residual-error stratum only, together with STUDY_PHASE2; see that entry).
  • Rich_2026_momelotinib.R (selects the residual-error model only, for both the momelotinib and the M21 output: proportional-only in the phase III trials SIMPLIFY-1 / SIMPLIFY-2 / MOMENTUM, proportional-plus-additive in the phase I clinical-pharmacology studies and the phase II study; it touches no structural or covariate parameter).
  • Ide_2022_elotuzumab.R (the saturable log-scale residual SD is multiplied by SDphase1,2 = 0.843 when STUDY_PHASE3 = 0; ELOQUENT-2 is the only phase III study. Source flag STOTHER = as.integer(STUDY != 204004) = 1 - STUDY_PHASE3.)
  • Willmann_2019_moxifloxacin.R (selects the proportional residual-error magnitude of the pediatric moxifloxacin popPK: 15.3% CV in the phase I single-dose intravenous study 11826 versus 33.8% CV in the phase III complicated intra-abdominal infection study 11643, with a third 47.5% CV magnitude for the phase III oral records gated by ROUTE_ORAL, which the source $ERROR block orders last so that it overrides this indicator. It touches no structural or covariate parameter.)
  • Li_2021_tedizolid.R (4.92-fold increase of the log-scale residual SD in ‘study 104 and phase 3 trials’ relative to the ‘non-phase 3 trials’ – Li 2021 Table 1. The source pools the phase 2 ABSSSI study 104 into the phase 3 stratum, so that study is coded 1 too; the stratum is effectively the sparse-sampled patient trials. It touches no structural or covariate parameter.)
  • Beck_2022_elagolix.R (selects the combined proportional-plus-additive residual-error magnitudes of the elagolix popPK: proportional variance 0.145 and additive variance 5.26e-05 (ng/mL)^2 in the six intensively sampled phase I studies versus 0.284 and 0.266 in the sparse-sampled phase III endometriosis and uterine-fibroid studies – Beck 2022 Table 3. It touches no structural or covariate parameter.)
• Notes: Sibling of the trial-named STUDY_<id> members, but the id names a development phase stratum rather than one trial, which is what makes the column reusable across papers (hence general rather than specific scope). Prefer the trial-named indicators when the source estimates a genuinely per-trial effect; use this one only when the source itself pools by phase. Distinct from STUDY_PHASE2A / STUDY_PHASE2B, which despite their names are indicators for two specific vupanorsen trials rather than phase strata. A model using this column should state in covariateData[[STUDY_PHASE3]]$notes which value a user should set when simulating, since the choice changes only the residual error and not the typical-value prediction.

STUDY_PHASE2 (canonical for pooled phase II study-stratum indicator)

• Description: 1 = record from a phase II study in the pooled analysis, 0 = record from a study of another phase. Per-record (study-fixed) binary indicator. Sibling of STUDY_PHASE3, completing the three-way development-phase stratification that a pooled phase I / II / III analysis needs: phase I is the reference when STUDY_PHASE2 and STUDY_PHASE3 are both 0.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0. Together with STUDY_PHASE3 = 0 this selects the pooled phase I studies.
• Source aliases:
  • PHASEN – the NONMEM $INPUT column in Ooi 2026 Supplementary Datafiles S1 and S2, an integer study-phase code branched in $ERROR as IF(PHASEN.EQ.2) / IF(PHASEN.EQ.3). Derive STUDY_PHASE2 = as.integer(PHASEN == 2).
• Example models: Ooi_2026_elafibranor.R, Ooi_2026_elafibranor_gft1007.R (selects the residual-error magnitudes only; Table S3 tabulates the residual-error rows per study phase), Vaddady_2020_doravirine.R (residual-error stratum only: phase 2b and phase 3 doravirine records share one log-scale SD of 0.504, while phase 1 records use separate SDs for <= 0.5 h and > 0.5 h postdose), Takada_2022_roxadustat.R (a structural effect, not a residual stratum: F1 is multiplied by 0.935 when STUDY_PHASE2 = 1 and at least one phosphate binder is coadministered, because the single phase II study did not require the 1-hour separation of roxadustat from phosphate binders that the phase III studies did; there is no phase I stratum, so STUDY_PHASE2 = 0 selects the pooled phase III studies).
• Notes: Introduced because STUDY_PHASE3 alone can only express a two-way phase III versus everything-else split, which is what its founding example Rich_2026_momelotinib.R needed. STUDY_PHASE2A / STUDY_PHASE2B are vupanorsen-specific single-study indicators and are not substitutes for this pooled-stratum column.

STUDY_PHASE4 (canonical for pooled phase IV study-stratum indicator)

• Description: 1 = record from a phase IV (post-approval) study in the pooled analysis, 0 = record from an earlier-phase study. Per-record (study-fixed) binary indicator. Sibling of STUDY_PHASE2 and STUDY_PHASE3 for analyses that pool intensively sampled phase I PK studies with a sparsely sampled post-approval efficacy/safety study and stratify a model component by that split.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (the pooled earlier-phase studies; in the founding model, the two phase I PK studies).
• Source aliases:
  • IND / PROT – Fediuk 2021 Text S1 $ERROR, IND = 0; IF(PROT.EQ.1073) IND = 1, where PROT is the Pfizer protocol number and 1073 is the phase 4 study NCT01312909. Derive STUDY_PHASE4 = as.integer(PROT == 1073).
• Example models: Fediuk_2021_varenicline.R (selects the residual-error magnitudes only: proportional 29.1% / additive 0.240 ng/mL in the phase 1 studies versus 43.2% / 0.580 ng/mL in the phase 4 study; it touches no structural or covariate parameter).
• Notes: Completes the STUDY_PHASE<n> development-phase stratum family with STUDY_PHASE2 and STUDY_PHASE3. As with its siblings, prefer a trial-named STUDY_<id> indicator when the source estimates a genuinely per-trial effect, and use this one only when the source stratifies by phase. A model using this column should state in covariateData[[STUDY_PHASE4]]$notes which value to set when simulating, since in the founding model the choice changes only the residual error and not the typical-value prediction.

STUDY_NOEARLYPK (canonical for the no-early-sampling study-stratum indicator that switches off central-volume IIV)

• Description: 1 = subject comes from a study whose first post-dose concentration was generally drawn more than about 2 minutes after a bolus or an infusion-rate change, so the study carries no information on the central volume; 0 = subject comes from a study with early (first-minutes) sampling. Per-subject (study-fixed) binary indicator. In the founding model it switches the inter-individual random effect on the central volume off, so subjects with the flag set take the typical V1: the random effect was not estimable from late-sampled data, and leaving it on produced unphysiologically high and low individual V1 estimates.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (study with early sampling; the full random effect applies). For simulation of new subjects set it to 0 – the switch is an estimation device for sparse data, not a claim that V1 does not vary in those populations.
• Source aliases:
  • STDY – integer study code in the Zhou 2021 Figure S3 control stream, branched in $PK as IF (STDY.GT.11.AND.STDY.NE.17.OR.STDY.EQ.3) THEN V1=TVV1 ELSE V1=TVV1*EXP(ETA(4)). The code-to-study key is not published; the paper’s Methods name the studies instead (see Example models).
• Example models: Zhou_2021_remimazolam.R (1 for the procedural-sedation studies CNS7056-002, -004, -006, -008 and -015 and the general-anaesthesia study ONO-2745-03; 0 for CNS7056-001, CNS7056-017, ONO-2745-01, ONO-2745-02 and ONO-2745-IVU007; the random effect on V1, 61.7% CV, is applied only when the indicator is 0).
• Notes: A study-STRATUM member of the STUDY_<id> family, like STUDY_PHASE3: the id names a sampling-design property shared by several trials rather than one trial, which is what makes it reusable. Distinct from SAMPLE_INTENSIVE, which is a per-OBSERVATION rich-versus-sparse switch on residual error; this indicator is per-SUBJECT and gates a random effect on a structural parameter.

STUDY_GFT505B_319_1 (canonical for the ELATIVE phase III elafibranor study indicator)

• Description: 1 = record from study GFT505B-319-1 (ELATIVE, NCT04526665), the phase III efficacy and safety trial of elafibranor in primary biliary cholangitis; 0 = record from any other pooled study. Per-record (study-fixed) binary indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (any other of the 17 pooled studies listed in Table S1).
• Source aliases:
  • STUDYIDN – Ooi 2026 Supplementary Datafiles S1 and S2 $INPUT; the study carries code 20, branched as IF(STUDYIDN.EQ.20). Derive STUDY_GFT505B_319_1 = as.integer(STUDYIDN == 20).
• Example models: Ooi_2026_elafibranor.R, Ooi_2026_elafibranor_gft1007.R (selects a study-specific absorption lag time, fixed to 0 h, for both analytes; Table S3 row ‘Lag-time for Study GFT505B-319-1 (h)’).
• Notes: Member of the STUDY_<id> family. Distinct from STUDY_PHASE3, which is 1 for every phase III study in the pool; the elafibranor analysis included two phase III trials (GFT505B-319-1 in PBC and GFT505-315-1 in MASH) and only the former carries the lag-time override.

STUDY_ODYSSEY (canonical for ODYSSEY trial study indicator)

• Description: 1 = record from the ODYSSEY trial (a multicentre randomised trial of antiretroviral regimens in children and adolescents living with HIV-1), 0 = record from the comparator study pooled into the analysis. Per-record (study-fixed) binary indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the per-paper comparator pooled study; IMPAACT P1093 in Chandasana 2024).
• Source aliases:
  • STDY – used in Chandasana_2023_dolutegravir.R; the source NONMEM control stream (Chandasana 2023 ESM Supplementary Text 1) codes STDY = 1093 for IMPAACT P1093 and STDY = 201296 for ODYSSEY, and collapses them with IF(STDY.EQ.201296) STUD=1.
  • study – used in Chandasana_2024b_dolutegravir.R; Chandasana 2024 Table 2 labels the residual-error rows “study P1093” and “ODYSSEY”.
• Example models: Chandasana_2023_dolutegravir.R (founding example, and the primary source of the fit; selects between the two study-specific residual-error magnitudes: proportional 28.6% plus additive 0.0405 ug/mL for the IMPAACT P1093 reference, and proportional 11.1% plus additive 0.300 ug/mL for ODYSSEY, applied inside model() as propSd_i <- propSd * (1 - STUDY_ODYSSEY) + propSd_odyssey * STUDY_ODYSSEY and likewise for the additive term), Chandasana_2024b_dolutegravir.R (selects between the two study-specific residual-error magnitudes of the pooled dolutegravir pediatric model: proportional 28.6% plus additive 0.00164 ug/mL for the IMPAACT P1093 reference, and proportional 11.1% plus additive 0.090 ug/mL for ODYSSEY; applied inside model() as propSd_i <- propSd * (1 - STUDY_ODYSSEY) + propSd_odyssey * STUDY_ODYSSEY and likewise for the additive term).
• Notes: Distinguishes the two pooled source studies of the pediatric dolutegravir popPK model. Separate residual errors were estimated because the trials used different bioanalytical methods, with lower limits of quantification of 5 ng/mL (P1093) and 9.7 ng/mL (ODYSSEY). The between-study difference is confined to residual error, not to any structural or covariate parameter. Records that belong to neither pooled study – for example the IMPAACT 2019 external-validation cohort – take 0 so the P1093 residual error applies, that being the pediatric dolutegravir single-entity study most comparable to IMPAACT 2019. The residual-error rows of Chandasana 2023 Table 2 print NONMEM $SIGMA VARIANCES in the point-estimate column and their square roots in the SD / CV% column; nlmixr2’s add() / prop() take the standard-deviation scale, so the additive terms are 0.0405 = sqrt(0.00164) and 0.300 = sqrt(0.0900). Reading the variance column as an SD understates the P1093 additive error 25-fold and was a live defect in Chandasana_2024b_dolutegravir.R until the primary paper was extracted. Sibling of the other STUDY_<id> members.

STUDY_PHASE2A (canonical for the vupanorsen phase IIa dose-finding study indicator)

• Description: 1 = participant enrolled in the phase IIa dose-finding study of vupanorsen in patients with hypertriglyceridemia, type 2 diabetes and nonalcoholic fatty liver disease (NCT03371355; N = 105), 0 = participant from any other study pooled into the analysis. Per-subject (study-fixed) binary indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the phase I studies NCT02709850 and NCT04459767, which are the PD reference population, and the phase IIb study; mutually exclusive with STUDY_PHASE2B).
• Source aliases:
  • Phase IIa – Ahn 2023 Table 1 column heading and Table 3 parameter rows (“Phase IIa factor in ANGPTL3 baseline”, etc.).
• Example models: Ahn_2023_vupanorsen.R (selects the phase IIa baseline factor for each of the three indirect-response end points – ANGPTL3 0.978, triglycerides 1.46, non-HDL-cholesterol 0.840 – applied inside model() as e_study_phase2a_rbase_<endpoint>^STUDY_PHASE2A; together with STUDY_PHASE2B it also forms the patient indicator that selects the single phase II potency factor on IC50, and the phase II residual-error magnitudes for the vupanorsen concentration).
• Notes: Sibling of the other STUDY_<id> members, named by the study’s development phase because the source paper identifies it that way throughout and reports no trial acronym for it (unlike the phase IIb study, TRANSLATE-TIMI 70). The between-study differences the indicator carries are confined to the PD layer (baseline and potency) and to the PK residual-error magnitude; no structural PK parameter depends on it. Scoped specific because the reference category is the paper’s own pooled-study set.

STUDY_PHASE2B (canonical for the TRANSLATE-TIMI 70 phase IIb study indicator)

• Description: 1 = participant enrolled in TRANSLATE-TIMI 70, the phase IIb dose-ranging study of vupanorsen in statin-treated patients with dyslipidemia (NCT04516291; N = 286), 0 = participant from any other study pooled into the analysis. Per-subject (study-fixed) binary indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the phase I studies NCT02709850 and NCT04459767, which are the PD reference population, and the phase IIa study; mutually exclusive with STUDY_PHASE2A).
• Source aliases:
  • Phase IIb – Ahn 2023 Table 1 column heading (“TRANSLATE TIMI 70: Phase IIb”) and Table 3 parameter rows (“Phase IIb factor in ANGPTL3 baseline”, etc.).
• Example models: Ahn_2023_vupanorsen.R (selects the phase IIb baseline factor for each of the three indirect-response end points – ANGPTL3 0.897, triglycerides 1.17, non-HDL-cholesterol 0.768 – applied inside model() as e_study_phase2b_rbase_<endpoint>^STUDY_PHASE2B; together with STUDY_PHASE2A it also forms the patient indicator that selects the single phase II potency factor on IC50, and the phase II residual-error magnitudes for the vupanorsen concentration).
• Notes: Sibling of the other STUDY_<id> members. Named by development phase rather than by the TRANSLATE-TIMI 70 acronym so that it pairs symmetrically with STUDY_PHASE2A, whose study has no acronym; the acronym is recorded here and in the model’s covariateData notes. All 286 participants were statin-treated, so within this pooled analysis the indicator is confounded with baseline statin use. Scoped specific because the reference category is the paper’s own pooled-study set.

STUDY_TLS (canonical for the TLS on-farm field trial cohort indicator)

• Description: 1 = record from the TLS trial, the 215-pig on-farm field trial of doxycycline administered in medicated feed (del Castillo 2006); 0 = record from one of the laboratory-condition in-feed trials pooled into the same analysis (AFSSA, BIOEQ, PARADOX, Company 9203, Company 9204). Per-dose-record (trial-fixed) binary indicator separating field-condition from laboratory-condition in-feed medication.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the laboratory-condition in-feed trials).
• Source aliases:
  • Trial ID = TLS – Toutain 2025 Table 1 trial identifier; Table 6 labels the two blocks “Trial TLS (feed, field conditions)” and “Trials AFSSA, BIOEQ, PARADOX, Company 9203 and Company 9204 (feed, laboratory conditions)”.
• Example models: Toutain_2025_doxycycline_pig.R (selects between the field-condition feed sub-model (Ka = 0.072 1/h, F = 0.501 with 84.8% between-subject variability, proportional residual 22.8% plus additive 0.113 ug/mL) and the laboratory-condition feed sub-model (Ka = 0.144 1/h, F = 0.340 with 36.4% between-subject variability, proportional residual 18.4% plus additive 0.019 ug/mL); Toutain 2025 Table 6).
• Notes: Well-formed member of the auto-approved STUDY_<id> canonical family; the mechanistic reading is field vs laboratory husbandry, but the paper’s own sub-model boundary is exactly the TLS trial, so the trial identifier is the honest column. Only meaningful when FORM_DOX_FEED = 1; set to 0 for every non-feed record. Sibling of STUDY_ODYSSEY (which likewise carries a purely between-study contrast) except that here the contrast spans absorption rate and bioavailability as well as residual error, because the field trial’s group-fed dosing process introduces competition between animals for access to the medicated meal.

STUDY_DORZA_EARLY (canonical for early-phase dorzagliatin study indicator (HMM0102 / HMM0103 / HMM0110))

• Description: 1 = subject enrolled in one of the three early-phase dorzagliatin studies pooled by Wang 2023 – HMM0102 (NCT02077452, multiple-ascending-dose), HMM0103 (NCT02386982, 28-day beta-cell-function study) or HMM0110 (NCT04324424, renal-impairment study with matched healthy volunteers); 0 = one of the three later-phase studies HMM0201 (NCT02561338, phase II dose-ranging), HMM0301 (NCT03173391, phase III monotherapy) or HMM0302 (NCT03141073, phase III metformin add-on). Used to switch apparent clearance between the two study groups.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (later-phase studies HMM0201 / HMM0301 / HMM0302, which used near-commercial preparations).
• Source aliases:
  • Study – used in Wang_2023_dorzagliatin.R (Wang 2023 Table 3 footnote: “Study: 1 for study 102/103/110, 0 for others”).
• Example models: Wang_2023_dorzagliatin.R (multiplicative exponential effect on CL/F: exp(0.203 * STUDY_DORZA_EARLY) = a 1.23-fold, i.e. 22.5% higher, apparent clearance in the early-phase studies; Wang 2023 Table 3 CL_STUDY = 1.23).
• Notes: Specific scope; tied to the six-trial Hua Medicine dorzagliatin development programme. Subject-level (time-fixed); set from the trial identifier. Wang 2023 Discussion paragraph 2 documents the mechanism and why the effect lands on CL/F rather than on F: the early studies showed systematically lower exposure, plausibly a formulation difference, but because the later studies were mainly sparsely sampled “the model was unable to correct for prediction bias by bioavailability”, so the between-trial difference was absorbed into apparent clearance. This is why the indicator is registered under the STUDY_<id> family rather than as a FORM_<drug>_<formulation> entry – the formulation attribution is the paper’s hypothesis, not its parameterisation. Directly analogous to STUDY_LBSL (a binary indicator over a group of early-phase studies switching CL and V magnitudes) and to STUDY_FARLETUZUMAB_PHASE2 / STUDY_POSA_PHASE3 / STUDY_NIPOCALIMAB_PHASE1. Follows the auto-approved STUDY_<id> canonical family.

STUDY_NOVA (canonical for the NOVA niraparib trial cohort indicator)

• Description: 1 = record from NOVA (NCT01847274), the phase 3 randomised maintenance trial of niraparib in platinum-sensitive recurrent ovarian cancer; 0 = record from one of the other five studies pooled into the same analysis. Per-record (study-fixed) binary indicator, one of the five non-reference members of the six-study niraparib pool.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 with STUDY_QUADRA, STUDY_PRIMA, STUDY_TABLET and STUDY_HEPATIC also 0, which selects PN001 (NCT00749502), the phase 1 dose-escalation study and the reference stratum of the set.
• Source aliases:
  • study – Gaffney 2026 Table 1 trial identifier; Table 3 labels the residual-error rows “Res variability CV PN001 / NOVA / QUADRA / PRIMA / TABLET / HEPATIC”.
• Example models: Gaffney_2026_niraparib.R (selects the NOVA log-scale residual SD 0.350; applied inside model() as a five-indicator switch over the six study-specific magnitudes, with the bare expSd carrying the PN001 reference).
• Notes: Well-formed member of the auto-approved STUDY_<id> canonical family, named for the published trial rather than a phase number because all six niraparib studies are individually named and two of them share a phase. The between-study difference is confined to residual error (assay and sampling-design differences: NOVA and the other phase 2/3 studies are sparsely sampled, whereas TABLET and HEPATIC used intensive serial sampling), not to any structural or covariate parameter – the same role STUDY_ODYSSEY plays in Chandasana_2024b_dolutegravir.R. Subject-level (time-fixed); set once from the trial identifier on each subject record. Exactly one of the five indicators is 1 per record, or all five are 0 for a PN001 record.

STUDY_QUADRA (canonical for the QUADRA niraparib trial cohort indicator)

• Description: 1 = record from QUADRA (NCT02354586), the phase 2 single-arm trial of niraparib in relapsed high-grade serous advanced ovarian cancer; 0 = record from one of the other five studies pooled into the same analysis.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 with the other four niraparib study indicators also 0 (selects PN001).
• Source aliases:
  • study – Gaffney 2026 Table 1 trial identifier.
• Example models: Gaffney_2026_niraparib.R (selects the QUADRA log-scale residual SD 0.381).
• Notes: See STUDY_NOVA for the family rationale and the five-indicator / PN001-reference convention.

STUDY_PRIMA (canonical for the PRIMA niraparib trial cohort indicator)

• Description: 1 = record from PRIMA (NCT02655016), the phase 3 randomised trial of niraparib versus placebo as first-line maintenance after platinum-based chemotherapy in advanced ovarian cancer; 0 = record from one of the other five studies pooled into the same analysis.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 with the other four niraparib study indicators also 0 (selects PN001).
• Source aliases:
  • study – Gaffney 2026 Table 1 trial identifier.
• Example models: Gaffney_2026_niraparib.R (selects the PRIMA log-scale residual SD 0.451, the largest of the six).
• Notes: See STUDY_NOVA for the family rationale and the five-indicator / PN001-reference convention.

STUDY_TABLET (canonical for the TABLET niraparib trial cohort indicator)

• Description: 1 = record from TABLET (NCT03329001), the phase 1 open-label multicentre study in advanced solid tumours that evaluated tablet-versus-capsule bioequivalence and the effect of a high-fat meal on a single 300 mg niraparib dose; 0 = record from one of the other five studies pooled into the same analysis.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 with the other four niraparib study indicators also 0 (selects PN001).
• Source aliases:
  • study – Gaffney 2026 Table 1 trial identifier.
• Example models: Gaffney_2026_niraparib.R (selects the TABLET log-scale residual SD 0.324).
• Notes: One of the two intensively sampled phase 1 studies added by Gaffney 2026 to the earlier niraparib pool, contributing 6487 of the 14,106 included observations (46%) and essentially all of the fed records that identify the FED effects on relative bioavailability and mean transit time. See STUDY_NOVA for the family rationale and the five-indicator / PN001-reference convention.

STUDY_ALK3831A305 (canonical for the olanzapine/samidorphan phase 3 study ALK3831-A305 record indicator)

• Description: 1 = record from ALK3831-A305, the phase 3 randomised double-blind placebo-controlled 4-week study of once-daily olanzapine/samidorphan (OLZ/SAM 10/10 or 20/20 mg) or olanzapine in adults with schizophrenia, with sparse PK sampling (predose on day 1, then weeks 2 and 4); 0 = record from one of the other studies pooled into the same analysis.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (any of the other pooled olanzapine/samidorphan studies).
• Source aliases:
  • STUDY – Sun 2021 Table S1 study identifier.
• Example models: Sun_2021_samidorphan.R (multiplicative 10.1-fold factor on the absorption lag time, Sun 2021 Table 4 ‘Change in ALAG’).
• Notes: A nuisance indicator rather than a physiological covariate: dose times were not recorded in ALK3831-A305, so Sun 2021 imputed dose timing and estimated a study-specific change in the lag time to absorb the timing error (Table 4 footnote c). Set to 0 when simulating a patient with known dose times. Auto-approved member of the STUDY_<id> family.

STUDY_HEPATIC (canonical for the HEPATIC niraparib trial cohort indicator)

• Description: 1 = record from HEPATIC (NCT03359850), the phase 1 open-label multicentre study of a single 300 mg niraparib dose in patients with advanced solid tumours and either normal hepatic function or moderate hepatic impairment; 0 = record from one of the other five studies pooled into the same analysis.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 with the other four niraparib study indicators also 0 (selects PN001).
• Source aliases:
  • study – Gaffney 2026 Table 1 trial identifier.
• Example models: Gaffney_2026_niraparib.R (selects the HEPATIC log-scale residual SD 0.146, the smallest of the six).
• Notes: The study name describes the trial’s hepatic-impairment design, but the indicator is a study-membership flag and not a hepatic-function covariate – do not substitute it for HEPIMP_MOD or any other impairment column. Gaffney 2026 explicitly did not retain hepatic impairment as a model effect (only 8 of the 17 HEPATIC patients had moderate impairment, which the paper names as a limitation), so the study enters the model solely through its residual-error magnitude. A model that needs the impairment contrast itself must carry a separate impairment column. See STUDY_NOVA for the family rationale and the five-indicator / PN001-reference convention.

STUDY_RADIO (canonical for the RADIO dolutegravir-rifampicin healthy-volunteer study cohort indicator)

• Description: 1 = record from RADIO, the phase II open-label sequential healthy-volunteer pharmacokinetic study of dolutegravir 50 mg and 100 mg once daily with and without rifampicin 600 mg once daily conducted at the St Stephen’s Centre, London; 0 = record from NCT01231542, the phase I two-arm fixed-sequence healthy-volunteer study conducted in Baltimore that is the other study pooled into the same analysis. Per-record (study-fixed) binary indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (NCT01231542).
• Source aliases:
  • Study (RADIO vs NCT01231542) – Kawuma 2023 Table 1 residual-error row labels Additive error RADIO study and Additive error NCT01231542 study; used in Kawuma_2023_dolutegravir.R.
• Example models: Kawuma_2023_dolutegravir.R (selects the study-specific additive residual standard deviation, 0.0485 mg/L for RADIO against 0.036 mg/L for NCT01231542, via addSd * (1 - STUDY_RADIO) + addSd_radio * STUDY_RADIO).
• Notes: In the founding model the indicator carries a purely bioanalytical contrast: RADIO quantified dolutegravir by UHPLC-UV validated over 0.050-10 mg/L and NCT01231542 by LC-MS/MS validated over 0.020-20 mg/L, and the additive residual error was constrained to be at least 20% of each study’s LLOQ (Kawuma 2022 Methods and ‘Analytical assay’), so the two estimates differ mainly through their 0.2 x LLOQ floors. The two studies are also perfectly confounded with the fed / fasted prandial contrast that drives the absorption lag time in the same model, but that effect is carried by the general [[FED]] indicator rather than by this one, because the authors themselves index the lag time by prandial state when simulating and attribute it mechanistically to food. Keep the two separate: a downstream user simulating a fed dose analysed on the LC-MS/MS assay is a coherent combination that a single study indicator could not express. Sibling of the other STUDY_<id> members; scoped specific because the reference category is this analysis’s own two-study set.

STUDY_B7451005 (canonical for the abrocitinib phase II psoriasis study B7451005 (NCT02201524) indicator)

• Description: 1 = record from the abrocitinib phase II psoriasis study B7451005 (NCT02201524), 0 = record from any other study in the pooled analysis. Per-record (study-fixed) binary indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the reference studies B7451001, B7451004, B7451013, B7451017, B7451019, B7451020 and B7451027).
• Source aliases:
  • PROT = 1005 – Wojciechowski 2022 Online Resource 2 protocol column (IF (PROT.EQ.1005) COVPROTRUV = ...).
• Example models: Wojciechowski_2022_abrocitinib.R (founding example; moderate-variability study, proportional residual SD multiplied by 1 + 0.495).
• Notes: Auto-approved member of the STUDY_<id> family; affects only the residual error.

STUDY_B7451006 (canonical for the abrocitinib phase IIb atopic-dermatitis dose-ranging study B7451006 (NCT02780167) indicator)

• Description: 1 = record from the abrocitinib phase IIb atopic-dermatitis dose-ranging study B7451006 (NCT02780167), 0 = record from any other study in the pooled analysis. Per-record (study-fixed) binary indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the reference studies B7451001, B7451004, B7451013, B7451017, B7451019, B7451020 and B7451027).
• Source aliases:
  • PROT = 1006 – Wojciechowski 2022 Online Resource 2 protocol column (IF (PROT.EQ.1006) COVPROTRUV = ...).
• Example models: Wojciechowski_2022_abrocitinib.R (founding example; high-variability study, proportional residual SD multiplied by 1 + 1.16).
• Notes: Auto-approved member of the STUDY_<id> family; affects only the residual error.

STUDY_B7451012 (canonical for the abrocitinib phase III atopic-dermatitis study B7451012 (JADE MONO-1, NCT03349060) indicator)

• Description: 1 = record from the abrocitinib phase III atopic-dermatitis study B7451012 (JADE MONO-1, NCT03349060), 0 = record from any other study in the pooled analysis. Per-record (study-fixed) binary indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the reference studies B7451001, B7451004, B7451013, B7451017, B7451019, B7451020 and B7451027).
• Source aliases:
  • PROT = 1012 – Wojciechowski 2022 Online Resource 2 protocol column (IF (PROT.EQ.1012) COVPROTRUV = ...).
• Example models: Wojciechowski_2022_abrocitinib.R (founding example; moderate-variability study, proportional residual SD multiplied by 1 + 0.495).
• Notes: Auto-approved member of the STUDY_<id> family; affects only the residual error.

STUDY_B7451043 (canonical for the abrocitinib phase I probenecid drug-interaction study B7451043 (NCT03937258) indicator)

• Description: 1 = record from the abrocitinib phase I probenecid drug-interaction study B7451043 (NCT03937258), 0 = record from any other study in the pooled analysis. Per-record (study-fixed) binary indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the reference studies B7451001, B7451004, B7451013, B7451017, B7451019, B7451020 and B7451027).
• Source aliases:
  • PROT = 1043 – Wojciechowski 2022 Online Resource 2 protocol column (IF (PROT.EQ.1043) COVPROTRUV = ...).
• Example models: Wojciechowski_2022_abrocitinib.R (founding example; moderate-variability study, proportional residual SD multiplied by 1 + 0.495).
• Notes: Auto-approved member of the STUDY_<id> family; affects only the residual error.

STUDY_DAILY_RPE (canonical for the Daily RPE rifapentine trial cohort indicator)

• Description: 1 = record from the Daily RPE study (NCT00814671), the open-label two-stage activity-safety trial in which 450 or 600 mg rifapentine daily replaced 600 mg rifampin during the intensive phase of standard tuberculosis therapy in the Western Cape, South Africa; 0 = record from RIFAQUIN (ISRCTN44153044), the phase III trial whose continuation-phase arms gave 1,200 mg rifapentine once weekly or 900 mg twice weekly at sites in the Western Cape, Gauteng and Harare, and which is the other trial pooled into the same analysis. Per-record (study-fixed) binary indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (RIFAQUIN).
• Source aliases:
  • study (Daily RPE vs RIFAQUIN) – Francis 2019 Table 2 row Effect of Daily RPE study on F (%); used in Francis_2019_rifapentine.R.
• Example models: Francis_2019_rifapentine.R (multiplicative fractional shift on bioavailability, (1 + e_study_daily_rpe_fdepot * STUDY_DAILY_RPE) with the coefficient -0.233, i.e. 23.3 percent lower bioavailability in Daily RPE than in RIFAQUIN; Francis 2019 Table 2, delta-OFV 59, P < 0.001).
• Notes: The indicator is a proxy for concomitant food intake, not an intrinsic trial effect, and the source says so directly: RIFAQUIN gave every dose with 240 ml of water 15 min after a standardised light meal of two hard-boiled eggs with bread, whereas Daily RPE advised patients to dose with food but neither standardised the meal nor recorded what was eaten. Rifapentine absorption is strongly food-dependent (33 to 86 percent increase in extent with a meal), so the 23.3 percent shift is the largest covariate effect in the model and the Discussion treats it as evidence for the importance of food-intake recommendations in programmatic use. A downstream user simulating a fed cohort on a standardised meal should set this column to 0 regardless of which trial the rest of the covariates came from. The column is partially confounded with dose level and dosing frequency by design, since only RIFAQUIN contributed 900 and 1,200 mg arms and only Daily RPE contributed 450 and 600 mg arms; the separate DOSE_HIGH indicator in the same model carries the 1,200-mg-once-weekly clearance contrast. Sibling of the other STUDY_<id> members; scoped specific because the reference category is this analysis’s own two-trial set.

STUDY_TACTCV (canonical for the TACT-CV antimalarial trial cohort indicator)

• Description: 1 = record from TACT-CV (NCT03355664, “Triple ACTs in Cambodia and Vietnam”), the multicentre open-label randomised trial of artemether-lumefantrine plus amodiaquine versus artemether-lumefantrine alone conducted at three sites in western and eastern Cambodia and in Vietnam; 0 = record from TRACII (NCT02453308, “Tracking Resistance to Artemisinin Collaboration II”), the seven-site trial in Bangladesh, India, Myanmar, the Democratic Republic of Congo and Lao PDR that is the other trial pooled into the same analysis. Per-record (study-fixed) binary indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (TRACII, NCT02453308).
• Source aliases:
  • TACT – the short form used in the Ding 2026 Table 4 covariate row Study effect (TACT) on VC/F LF (%) and in its footnote, Study effect (TACT) was implemented on VC/F LF as a proportional equation [1 + theta x TACT]; used in Ding_2026_lumefantrine.R (same orientation as the canonical, no value transformation).
• Example models: Ding_2026_lumefantrine.R (proportional effect on the apparent lumefantrine central volume of distribution: vc <- exp(lvc + etalvc) * (WT/45)^e_wt_vc * (1 + e_study_tactcv_vc * STUDY_TACTCV) with e_study_tactcv_vc = -0.281, i.e. Vc/F is 28.1% lower in TACT-CV than in TRACII).
• Notes: The indicator distinguishes the two trials, not the two treatment arms – coadministration of amodiaquine was tested separately as a drug-drug-interaction covariate on every parameter and found not to affect lumefantrine PK, so STUDY_TACTCV and a hypothetical triple-ACT-arm indicator are different columns and must not be conflated (the abbreviation TACT in the source table is the trial acronym, and Ding 2026 Results describes the term in words as “a study effect on the central volume of distribution”). Ding 2026 attributes the contrast to the different fatty food given with each dose to promote lumefantrine absorption – a fatty snack in TRACII against 80 mL of milk in TACT-CV – and judges it not clinically meaningful, since it moves the terminal half-life from 191 h to 182 h and the Day 7 concentration from 454 to 444 ng/mL with no change in total exposure. The same two-trial pooling underlies the sibling Ding_2026_artemether.R and Ding_2026_amodiaquine.R, neither of which retains a study term; for amodiaquine a trial difference in inter-compartmental clearance was detected but discarded as biologically implausible, and that model records the column in covariatesDataExcluded instead. Sibling of the other STUDY_<id> members; scoped specific because the reference category is this analysis’s own two-trial set.

STUDY_BDQ_C208_C209 (canonical for the bedaquiline phase IIb MDR-TB studies TMC207-TiDP13-C208 / C209 indicator)

• Description: 1 = subject from the bedaquiline phase IIb registration studies in patients with multidrug-resistant tuberculosis, TMC207-TiDP13-C208 (stages 1 and 2) or TMC207-TiDP13-C209; 0 = subject from any other study of the McLeay 2014 pooled bedaquiline analysis, or from a later study. Subject-level (time-fixed).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (any other study). Note that in the source parameterisation C208 / C209 is the REFERENCE group for relative bioavailability (F = 1), so the indicator selects F = 1 rather than applying a shift to a typical value.
• Source aliases:
  • study – McLeay 2014 Table 3 rows ‘Study R207910-CDE102 or TiDP13-C104 on F’ / ‘Other studies on F’ (reference C208 / C209) and ‘Residual unexplained variability on TiDP13-C208 or TiDP13-C209’; reprinted in Kurosawa 2021 Table S1.
• Example models: Kurosawa_2021_bedaquiline.R (relative bioavailability 1 for C208 / C209 vs 2.03 for ‘other studies’ and 1.51 for STUDY_BDQ_CDE102_C104; log-scale residual SD 0.277 vs 0.206 elsewhere).
• Notes: Well-formed member of the auto-approved STUDY_<id> family, prefixed BDQ because the C-numbers are bedaquiline (TMC207) protocol codes. Mutually exclusive with STUDY_BDQ_CDE102_C104; a subject with both 0 is in McLeay 2014’s ‘other studies’ group (C109, C110, C111, C202, TBC1003 and any later study, including the Kurosawa 2021 healthy volunteers). In McLeay 2014 every MDR-TB patient came from C208 / C209, so this indicator coincides with DIS_TB_MDR = 1 in that dataset; they are kept separate because Kurosawa 2021 decoupled them in its nontuberculous-mycobacteria simulations (MDR-TB-like clearance with the ‘other studies’ bioavailability).

STUDY_BDQ_CDE102_C104 (canonical for the bedaquiline phase 1 oral-solution studies R207910-CDE102 / TMC207-TiDP13-C104 indicator)

• Description: 1 = subject from the bedaquiline phase 1 healthy-volunteer multiple-dose oral-solution studies R207910-CDE102 or TMC207-TiDP13-C104; 0 = subject from any other study. Subject-level (time-fixed).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (any other study).
• Source aliases:
  • study – McLeay 2014 Table 3 row ‘Study R207910-CDE102 or TiDP13-C104 on F’; reprinted in Kurosawa 2021 Table S1.
• Example models: Kurosawa_2021_bedaquiline.R (relative bioavailability 1.51 versus the C208 / C209 reference).
• Notes: Well-formed member of the auto-approved STUDY_<id> family. Sibling of STUDY_BDQ_C208_C209 from the same McLeay 2014 pooled analysis; the two are mutually exclusive.

STUDY_DNDI_SHINE (canonical for the DNDi super-boosted lopinavir / SHINE paediatric TB study indicator)

• Description: 1 = record from either (i) the Drugs for Neglected Diseases initiative (DNDi) study of lopinavir super-boosted with additional ritonavir (4:4) in infants and young children co-infected with HIV and TB, or (ii) the SHINE trial (Shorter treatment for minimal TB in children; 4 vs 6 months of first-line treatment with WHO fixed-dose combinations); 0 = record from the DATiC study (Optimal Dosing of 1st Line Antituberculosis and Antiretroviral Drugs in Children; South Africa and Malawi). Subject-level (study-fixed) binary indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (DATiC).
• Source aliases:
  • STUDY > 1 – the Tikiso 2022 supplementary control stream codes STUDY = 1 DATiC, 2 DNDi, 3 SHINE and applies IF(STUDY>1) STUDY_ABS = THETA(13). Derive STUDY_DNDI_SHINE = as.integer(STUDY > 1).
• Example models: Tikiso_2022_ethambutol.R (23.6% slower absorption in DNDi and SHINE: ka multiplied and MTT divided by 1 + e_study_dndi_shine_ka * STUDY_DNDI_SHINE with e_study_dndi_shine_ka = -0.236).
• Notes: Well-formed member of the auto-approved STUDY_<id> family. The two studies share one indicator because the source model estimates a single shared effect for both relative to DATiC; the paper attributes it to differences in assay, sampling schedule and formulation between studies rather than to a population difference. A future model that separates DNDi from SHINE should register STUDY_DNDI and STUDY_SHINE as siblings rather than re-orient this indicator.

STUDY_DINO (canonical for the DINO neonatal dosing study cohort indicator)

• Description: 1 = record from the DINO study (Drug dosage Improvement in NeOnates; NCT02421068, MEC-2014-067, NL47409.078.14), the prospective multicentre study of preterm neonates in four Dutch level III NICUs (Erasmus MC-Sophia, Maxima Medical Center, Maastricht UMC, Radboud UMC) with opportunistic scavenge sampling and LC-MS/MS bioanalysis; 0 = record from the other dataset pooled into the same analysis. Per-record (study-fixed) binary indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the non-DINO dataset of the pooled analysis; for Wu_2022_fentanyl.R the Helsinki continuous-infusion dataset of Saarenmaa 2000).
• Source aliases:
  • ASY – the Wu 2022 ESM control stream column “Dataset identification. 0 for dataset 1 and 1 for dataset 2”, where dataset 2 is DINO. Same orientation as the canonical, no value transformation.
• Example models: Wu_2022_fentanyl.R (selects the residual-error magnitudes only: additive 0.0297 ug/L and proportional 0.361 for DINO against 0.246 ug/L and 0.23 for the Helsinki dataset).
• Notes: The indicator identifies the study, not the assay: in Wu_2022_fentanyl.R the two datasets also differ in bioanalysis (DINO LC-MS/MS vs Helsinki radioimmunoassay), sampling route and BLOQ handling, and the paper attributes the residual split to the dataset. Use RIA_ASSAY / IMMUNOASSAY instead when a paper attributes a residual split specifically to the analytical method. The DINO cohort also underlies Voller_2019_midazolam.R, which retains no study term.

REGI_BID (canonical for twice-daily dosing-regimen indicator)

• Description: 1 = subject’s dosing regimen is BID (twice daily), 0 = QD (once daily) or other non-BID regimen. Per-subject (regimen-fixed) categorical indicator for population analyses that pool QD and BID arms and test regimen as a covariate.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-BID regimen – typically QD).
• Source aliases:
  • BID – used in Girard_2012_pimasertib.R.
• Example models: Girard_2012_pimasertib.R (additive shift on the cumulative-logit AE-score model: theta_bid * REGI_BID; -0.399 logit units for BID vs QD), Kengo_2025_atazanavir.R, Kengo_2025_ritonavir.R (interaction-only use: the indicator carries no main effect and acts solely in combination with CONMED_RIF, because the DERIVE trial escalated ritonavir-boosted atazanavir from once to twice daily only after rifampicin had been added and so never sampled twice-daily ATV/r without rifampicin. Doubling the dosing frequency lowers rifampicin’s induction of atazanavir clearance from 3.05-fold to 2.03-fold and fully restores atazanavir bioavailability, and partially restores ritonavir bioavailability from 31% to 67% of reference. Setting REGI_BID = 1 with CONMED_RIF = 0 is extrapolation beyond the studied design and correctly selects the unmodified reference).
• Notes: Specific scope because the QD-vs-BID contrast is study-specific; future regimen-comparison models that contrast different schedules should either extend this entry’s example list (when QD is the reference) or register a sibling indicator (REGI_TID, REGI_QW) following the same pattern. Distinct from DOSE (dose level in mg) and from total-daily-dose aggregates: a 60 mg/day cohort can include either a 60 mg QD subgroup or a 30 mg BID subgroup, and both share the same DOSE = 60 while differing in REGI_BID. Sibling REGI_QD carries the same QD-vs-BID contrast with the polarity reversed, for papers whose reference arm is the twice-daily one.

REGI_QD (canonical for once-daily dosing-regimen indicator)

• Description: 1 = subject’s dosing regimen is QD (once daily, every 24 h), 0 = the per-paper comparator regimen. Per-subject (regimen-fixed) categorical indicator for population analyses that pool a once-daily arm with a more frequent arm and test regimen as a covariate, in the case where the more frequent arm is the reference.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the paper’s comparator regimen; twice-daily / every-12-hour dosing in Gosselin_2015_motesanib.R). Document the per-paper comparator in covariateData[[REGI_QD]]$notes.
• Source aliases:
  • FREQ – “dosing frequency”, the Gosselin 2015 notation in Supplement Table 1 (“+FREQ on V M4 /F M4”, “+ FREQ on CL M4 /F M4”) and in the Table 3 rows “Dosing interval on CLM4/FM4” and “Dosing interval on VM4/FM4”, both of which report their multiplicative factor explicitly “for once-daily dose”.
• Example models: Gosselin_2015_motesanib.R (multiplicative factors on both metabolite disposition parameters: e_regi_qd_cl_m4 = 1.52 on M4 apparent clearance and e_regi_qd_vc_m4 = 0.45 on M4 apparent volume, with twice-daily dosing as the reference arm; the reference orientation is confirmed arithmetically by the paper’s Table 3 footnote, which quotes the M4 half-lives 25.7 h and 14.7 h “for twice-daily administration” and which the unmodified typical values reproduce), Huang_2021_cholecalciferol_pbpk.R (comparator a SINGLE oral dose: REGI_QD switches the hepatic vitamin D3 clearance between mclh = 0.21 L/h under repeated daily dosing and sclh = 0.32 L/h after a single dose, Huang 2021 Table 1 and supplement Final Model section, ‘For single dose: CLH = SCLH. For repeated dose: CLH = MCLH.’).
• Notes: Member of the REGI_* regimen-indicator family alongside REGI_BID, REGI_QM and REGI_Q2M. Mutually exclusive with REGI_BID in any analysis that pools only q24h and q12h arms (REGI_QD = 1 - REGI_BID there); carry whichever polarity matches the source paper’s reference arm rather than translating the coefficient, so the sign and magnitude stay literally traceable to the published table. When a model pools three or more schedules, use one indicator per active arm and let all-zero select the reference arm, as the REGI_Q2M entry documents. Distinct from DOSE (dose level in mg) and from total-daily-dose aggregates. Note that a regimen indicator on a disposition parameter (as in Gosselin 2015, where once-daily dosing raises M4 apparent clearance 1.52-fold and lowers M4 apparent volume to 0.45-fold) is an empirical, regimen-level correction – Gosselin 2015 attributes it to “saturation in the formation/elimination of M4 under the twice-daily regimen” – and models carrying it must not be extrapolated to dosing intervals the paper did not study.

REGI_QM (canonical for once-monthly dosing-regimen indicator)

• Description: 1 = subject’s dosing regimen is once-monthly (QM, typically every 4 weeks for a long-half-life mAb), 0 = the per-paper comparator regimen (typically once-every-2-weeks, Q2W, for evolocumab-class mAbs). Per-subject (regimen-fixed) categorical indicator used by static Emax-on-AUC exposure-response models that pool QM and Q2W arms and need to distinguish them because the Emax-on-AUC formulation cannot represent the difference in target-saturation time courses between the two regimens. This is purely an exposure-response model device; dynamic ODE-based PK / PD layers do not need this indicator because they resolve the concentration time course explicitly.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (non-QM regimen – paper-defined; Q2W for Kuchimanchi_2018_evolocumab_ldlc.R, and sham / every-other-month for Crass_2025_pegcetacoplan_ga_doseresponse.R). Document the per-paper comparator in covariateData[[REGI_QM]]$notes.
• Source aliases:
  • REG – Kuchimanchi 2018 notation for the binary indicator i in the exposure-response formula Eff = Emax * AUC / (EC50 * REG^i + AUC). The linear-scale multiplier paired with this indicator (paper estimate 2.30) maps to the canonical parameter reg_qm.
• Example models: Kuchimanchi_2018_evolocumab_ldlc.R (founding example; exposure-response Emax on LDL-C distinguishes 420 mg SC QM from 140 mg SC Q2W via ec50_eff = ec50 * reg_qm^REGI_QM), Crass_2025_pegcetacoplan_ga_doseresponse.R (proportional step-function shift on the study-eye geographic-atrophy lesion progression rate, (1 + e_regi_qm_slope_study * REGI_QM + e_regi_q2m_slope_study * REGI_Q2M) with e_regi_qm_slope_study = -0.204, i.e. a 0.80-fold lower rate of lesion growth on intravitreal pegcetacoplan 15 mg monthly vs sham).
• Notes: Promoted from Scope: specific to Scope: general alongside the Crass 2025 pegcetacoplan geographic-atrophy extraction: “once-monthly dosing regimen” is a regimen-interval fact that is independent of the comparator, and a second unrelated analysis (a different drug, a different route, and a sham rather than Q2W comparator) now uses the same column. The comparator is paper-specific and is recorded per-model in covariateData[[REGI_QM]]$notes, not in the column name. Sibling of REGI_BID (within-QD-vs-BID schedules) and REGI_Q2M (every other month) under the REGI_* family pattern; future regimen-comparison models that contrast different intervals (e.g., Q4W vs Q8W) should register a sibling indicator (REGI_Q4W, REGI_Q8W) following the same pattern. Distinct from DOSE (dose level in mg) and from regimen-implied total-AUC aggregates: a steady-state exposure-matched 140 mg Q2W and 420 mg QM pair share the same AUC over 4 weeks but differ in REGI_QM. Pairs with the canonical parameter lreg_qm / reg_qm (the EC50 multiplier exponentiated by this indicator). Use only in static Emax-on-AUC exposure-response models; dynamic ODE-based PK / PD layers do not need this indicator.

REGI_Q2M (canonical for once-every-other-month dosing-regimen indicator)

• Description: 1 = subject’s dosing regimen is every other month (Q2M, i.e. once every 8 weeks or once every second monthly visit), 0 = the per-paper comparator regimen. Per-subject (regimen-fixed) categorical indicator used by exposure-response and dose-response models that pool arms differing only in dosing interval and need to distinguish them because the model’s drug-effect term is a per-regimen step function rather than a resolved concentration time course.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (not on the Q2M regimen – paper-defined). Document the per-paper comparator in covariateData[[REGI_Q2M]]$notes. Reference values observed: in Crass_2025_pegcetacoplan_ga_doseresponse.R the reference state is REGI_QM = REGI_Q2M = 0, i.e. sham treatment; 498 of 1501 patients were on the Q2M (every-other-month) pegcetacoplan arm and 505 on the monthly arm.
• Source aliases:
  • EOM – “every other month”, the Crass 2025 notation for the arm and for the corresponding drug-effect theta (“Every-other-month drug effect, proportion”).
• Example models: Crass_2025_pegcetacoplan_ga_doseresponse.R (proportional step-function shift on the study-eye GA lesion progression rate: (1 + e_regi_qm_slope_study * REGI_QM + e_regi_q2m_slope_study * REGI_Q2M) with e_regi_q2m_slope_study = -0.172, i.e. a 0.83-fold lower rate of lesion growth vs sham).
• Notes: Sibling of REGI_QM (once-monthly) and REGI_BID (twice-daily) under the REGI_* regimen-indicator family; the REGI_QM entry’s Notes explicitly anticipate registering siblings for additional dosing intervals. When a model pools three or more regimens, use one indicator per active arm and let all-zero select the reference arm – the indicators are then mutually exclusive and must be constructed so at most one is 1 per subject. Distinct from DOSE (dose level in mg): the Crass 2025 monthly and every-other-month arms received the identical 15 mg intravitreal dose and differ only in interval. Use only where the model cannot resolve the concentration time course; a dynamic ODE-based PK/PD layer expresses the same information through the dosing records.

REGI_Q4W (canonical for once-every-4-weeks dosing-regimen indicator)

• Description: 1 = the subject’s or study arm’s dosing regimen is every 4 weeks (Q4W), 0 = the per-paper comparator interval. Regimen-fixed categorical indicator for dose-response and exposure-response models that pool arms differing in dosing interval and need to distinguish them because the drug-effect term is a per-regimen step function rather than a resolved concentration time course.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (not on the Q4W regimen – paper-defined; every 2 weeks in Checchio_2017_psoriasis_pasi_landmark_mbma.R). Document the per-paper comparator in covariateData[[REGI_Q4W]]$notes.
• Source aliases:
  • bio.freq = 'Q4W' – Checchio 2017 Supplementary Table 2 row label brodalumab.I(bio.freq = 'Q4W').
• Example models: Checchio_2017_psoriasis_pasi_landmark_mbma.R (additive shift on the brodalumab LOG ED50: exp(led50_brodalumab + e_regi_q4w_led50_brodalumab * REGI_Q4W) with e_regi_q4w_led50_brodalumab = 0.151, i.e. an ED50 of 104.1 mg per administration on Q4W against 89.5 mg on Q2W, so a Q4W arm needs about 16% more drug per administration for the same Week-12 PASI response. The 90% CI (-0.033, 0.334) includes zero, so the interval effect is weak, and the source’s own published brodalumab prediction is the Q2W arm).
• Notes: Registered as the sibling the REGI_QM entry explicitly anticipates (“future regimen-comparison models that contrast different intervals (e.g., Q4W vs Q8W) should register a sibling indicator (REGI_Q4W, REGI_Q8W) following the same pattern”). Scope general for the same reason REGI_QM and REGI_Q2M are: “dosed every 4 weeks” is a regimen-interval fact independent of what it is being contrasted against, and the comparator belongs in the per-model notes rather than in the column name. Distinct from REGI_QM, which names a once-MONTHLY regimen: the two coincide for a long-half-life monoclonal antibody dosed on a 4-week cycle but not for a calendar-month regimen, and a model should carry whichever the source’s own row label names. Also distinct from DOSE (dose level in mg): a Q2W and a Q4W arm can share an identical per-administration dose and differ only in this column, which is exactly the founding case. Drug-specific in a multi-drug meta-analysis. Where a source estimates an interval term for one agent only, the indicator must be left at 0 for every other agent even when that agent’s clinical regimen is also Q4W – in the founding model ixekizumab is dosed Q4W but has a single ED50 with no interval term, so setting the indicator on an ixekizumab arm would silently apply brodalumab’s offset. Use only where the model cannot resolve the concentration time course; a dynamic ODE-based PK/PD layer expresses the same information through its dosing records.

MIL_REGIMEN (canonical for miltefosine monotherapy vs combination-with-LAmB regimen indicator)

• Description: Per-subject (time-fixed) binary indicator carrying the visceral-leishmaniasis miltefosine treatment-arm assignment in the Dorlo 2017 Eastern African LEAP-0208 study. 1 = monotherapy arm (28 days oral miltefosine 2.5 mg/kg/day, maximum 150 mg/day); 0 = combination arm (single IV liposomal amphotericin B 10 mg/kg on day 1 plus 10 days oral miltefosine 2.5 mg/kg/day, maximum 150 mg/day). Used to select the duration of the initial reduced-bioavailability window: 7 days for monotherapy, 1 day for the combination arm (Dorlo 2017 Table 2 footnote d).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (combination arm: single LAmB IV on day 1 + 10-day oral miltefosine).
• Source aliases: derived per subject from the trial-arm assignment in NCT01067443; the source paper does not name an NMTRAN column directly. Dorlo 2017 narrative labels the arms ‘MIL’ (monotherapy) and ‘combination therapy arm’.
• Example models: Dorlo_2017_miltefosine.R (sets tlowf <- 7 * MIL_REGIMEN + 1 * (1 - MIL_REGIMEN) inside model(); tlowf is then used in the (t > 0) * (t <= tlowf) predicate that gates the typical 74.3% reduction in relative bioavailability during the initial absorption window).
• Notes: Specific scope because the miltefosine-vs-miltefosine+LAmB head-to-head and the per-arm absorption-window durations are tied to the Dorlo 2017 design. The combination-arm reduction window is shorter than the monotherapy window because the LAmB infusion is hypothesised to accelerate the patients’ overall physiological recovery (Discussion paragraph ‘Pharmacokinetics’); future Eastern African VL combination-with-LAmB miltefosine popPK extractions that share the same regimen pair can extend this entry’s example list, while a contrast against a different miltefosine combination partner (e.g., paromomycin, fexinidazole) should register a sibling indicator. Distinct from REGI_BID (within-monotherapy QD vs BID schedule) and from the CONMED_* family (the LAmB co-administration in the combination arm is captured implicitly via the regimen indicator rather than via an explicit concomitant-medication binary, because the LAmB-driven effect modelled here is on the absorption-window duration rather than on miltefosine clearance or distribution).

MEAL_FLAG (canonical for intra-day meal-window indicator (time-varying))

• Description: 1 = the current observation time falls within a meal window (typically lunch or dinner, ~1 hour duration each); 0 = no meal-driven physiological perturbation active. Distinct from FED (per-dose-record fed-vs-fasted indicator): MEAL_FLAG is a time-varying intra-day flag that switches on at meal onset, stays on through the meal duration, and switches off afterward. Used by enterohepatic-recirculation / gallbladder-contraction models that need to scale post-prandial transport rate constants for the duration of a meal.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no meal effect active at this time).
• Source aliases: paper narrative “meal effect on K_BI” / “post-prandial gallbladder contraction” in Zuo 2016.
• Example models: Zuo_2016_UDCA.R (multiplies the biliary-to-intestine rate constants K_BI,0/1/2 by E_meal = 35.33 during meal windows to simulate gallbladder contraction; two meals modelled per day, lunch at +4 h and dinner at +10 h after the morning dose, each 1 hour long), Edwards_2016_obeticholicAcid_pbpk.R (gates the gallbladder-to-gut output rate constant f23 = 1.2 /h entirely on the flag rather than scaling it: outside a meal window the gallbladder only fills via f22 and does not empty, so gb_out = f23 * MEAL_FLAG; the window is 90 min from the start of each standardized meal per Edwards 2016 “Source data”).
• Notes: Specific scope because the operational definition (meal duration, schedule, magnitude of the rate-constant scaling) is paper-defined. Two materially different encodings are in use and must not be conflated: a multiplicative scaling of an always-active biliary-to-intestinal rate constant (Zuo_2016_UDCA.R, x35.33 during the window) versus an on/off gate where the emptying flux is exactly zero outside the window (Edwards_2016_obeticholicAcid_pbpk.R). Read the source’s gallbladder mechanism before choosing; the gated form makes the gallbladder a true fill-and-discharge reservoir, the scaled form does not. Meal duration is likewise paper-defined (1 h in Zuo 2016, 90 min in Edwards 2016). Time-varying: must be supplied at every observation row in the event dataset. Pair with SNACK_FLAG for studies that distinguish meal and snack effects. Distinct from FED and FED_HIGHFAT (per-dose-record meal-state indicators tied to a single dosing event); MEAL_FLAG is decoupled from any specific dose record and instead drives ongoing physiology over a multi-hour window.

SNACK_FLAG (canonical for intra-day snack-window indicator (time-varying))

• Description: 1 = the current observation time falls within a snack window (typically 0.5 hour duration, a smaller perturbation than a meal); 0 = no snack-driven physiological perturbation active. Used in tandem with MEAL_FLAG by enterohepatic-recirculation / gallbladder-contraction models that scale post-prandial transport rate constants with a smaller magnitude for snacks than for full meals.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no snack effect active at this time).
• Source aliases: paper narrative “snack effect on K_BI” in Zuo 2016.
• Example models: Zuo_2016_UDCA.R (multiplies the biliary-to-intestine rate constants K_BI,0/1/2 by E_snack = 9.53 during snack windows; one snack modelled per day at +7 h after the morning dose in the Xiang 2011 single-dose study, 0.5 hour long).
• Notes: Specific scope; same per-paper-defined operational definition as MEAL_FLAG. Time-varying.

FRACABS (canonical for dose-record fractional absorption supplied as a data column)

• Description: Per-dose-record fractional absorption (0-1) supplied as a covariate when the source paper reports F as a dose-dependent function rather than as a single estimable bioavailability parameter. The model wires the covariate into a bioavailability hook (f(<depot>) <- FRACABS); the user supplies the per-dose F value derived from the paper’s regression / lookup table.
• Units: fraction
• Type: continuous
• Scope: specific
• Reference category: n/a (continuous covariate).
• Source aliases: paper narrative “F = 0.66 at 150 mg” / “F = 0.31 at 1000 mg” / Walker 1992 / Crosignani 1991 / Dilger 2012 reported absorption values in Zuo 2016.
• Example models: Zuo_2016_UDCA.R (dose-dependent fractional absorption derived from a log-dose linear regression: F = 0.66 at 150 mg and F = 0.31 at 1000 mg, with R^2 = 0.99 over 200-2000 mg per the paper; combined with the UDCA molecular-weight conversion in the bioavailability hook f(stomach_udca) <- FRACABS / mw_udca).
• Notes: Specific scope because the operational definition (which doses get which F value; whether F is treated as a known input or as a parameter to be re-estimated) is paper-defined. Distinct from lfdepot / f(depot) in models where bioavailability is an estimable PK parameter rather than a supplied covariate. When the paper’s F regression is itself a function of DOSE, the user can derive FRACABS from the DOSE column upstream of rxSolve rather than carrying the regression inside model.

FASTED_STRICT (canonical for strict-fasting-vs-relaxed-fasting dose-record indicator)

• Description: 1 = the dose was taken under a strictly enforced fast, in which the protocol prohibits food for a stated interval both BEFORE and AFTER the dose; 0 = any less strict prandial state, i.e. a relaxed / partial fast in which food is permitted on one side of the dose, or a fed state. Encodes fast strictness, the axis that the binary FED indicator cannot express: FED = 0 pools every non-fed state into one category, so a study that contrasts two different non-fed protocols needs this second indicator to separate them. FASTED_STRICT composes with FED to span a three-level prandial factor: strict fast is FASTED_STRICT = 1, FED = 0; relaxed fast is FASTED_STRICT = 0, FED = 0; fed is FASTED_STRICT = 0, FED = 1.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (relaxed / partial fast, or fed). Document the two protocols verbatim in covariateData[[FASTED_STRICT]]$notes, since what counts as “strict” is protocol-defined and the pre-dose and post-dose intervals differ between papers.
• Source aliases:
  • prandial-state label (Table 2 row labels fast / mod fast / fed) – used in OlssonGisleskog_2025_ibrutinib.R, where the three-level prandial factor of Olsson Gisleskog 2025 Table 2 is decomposed into FASTED_STRICT (acting on F1) plus FED (acting on the zero-order input duration D1).
  • FASTED – used in FernandezTeruel_2024_capivasertib.R (Fernandez-Teruel 2024 equation (c)); same orientation as the canonical, so no value transformation is needed.
• Example models: FernandezTeruel_2024_capivasertib.R (the paper’s own three prandial protocols – overnight fast, semi-fasted with no food from 2 h before to 1 h after the dose, and fed – collapse onto exactly this canonical’s axis, with the semi-fasted and fed states POOLED at the reference level; the indicator gates the whole absorption lag term rather than scaling it, tlag = (Lag1_tab*(1-FORM_CAPSULE) + Lag1_cap*FORM_CAPSULE) * (1 - FASTED_STRICT) * exp(eta), so an overnight-fasted dose has no lag at all and a non-overnight-fasted one has 0.212 h (tablet) or 0.46 h (capsule). Per dose record, not per subject: the OAK study dosed the same patients under both prandial states in a fixed-sequence crossover. Fernandez-Teruel 2024 Table 3 and Sect. 3.2), OlssonGisleskog_2025_ibrutinib.R (multiplicative ratio effect on relative bioavailability: 0.666^FASTED_STRICT, i.e. 33.4% lower F1 under the strict fast than under the paper’s relaxed-fast or fed reference, per Olsson Gisleskog 2025 Table 2 rows F1 mod fast/fed = 1 FIX and F1 fast = 0.666 FIX. The D1 rows of the same table pool the two non-fed states at a single 2.45 h, so FED rather than FASTED_STRICT is the indicator that acts on the absorption duration – the two indicators land on different parameters).
• Notes: General scope: an enforced fasting window is a protocol-level design variable, and pre-dose fasting instructions of the form “nothing by mouth from midnight” appear in the approved labelling of more than one oral oncology product, so a study contrasting two fast strictnesses is expected to recur. Do NOT identify this indicator with the bare phrase “modified fasting.” That phrase is not portable between papers and must be read from each paper’s own definition, because two papers in the same extraction wave used it for opposite protocols: Olsson Gisleskog 2025 ibrutinib defines it as dose first, food later (Methods 2.1: ibrutinib “taken at least 30 min before or at least 2 h after a meal”), and that state is this entry’s reference category (FASTED_STRICT = 0), with the paper’s stricter PK-sampling-day instruction to “fast from midnight prior (or at a minimum, 2 h prior) to dosing and continue fasting until approximately 30 min after capsule intake” being FASTED_STRICT = 1. Mauro 2025 nilotinib uses the same phrase for food first, dose about 2 hours later, which is MEAL_PREDOSE_2H = 1 and belongs to a different family entirely. Always encode the protocol, never the phrase. Distinct from the rest of the meal family on the axis each one carries: FED / FED_HIGHFAT / FED_LOWFAT carry meal presence and composition; MEAL_DELAY_<n> carries a dose-then-meal interval and MEAL_PREDOSE_<n> a meal-then-dose interval, both of which describe when a meal happened relative to the dose; FASTED_STRICT carries how tightly food was excluded on BOTH sides of the dose, so both of its levels may be food-free. MEAL_FLAG / SNACK_FLAG are time-varying intra-day meal windows driving ongoing physiology and are a different construct again. Per dose record: in a study where routine dosing and PK-sampling-day dosing follow different fasting instructions (as in SHINE) the indicator varies within subject across days, which is also why the covariate is best set from the dosing-diary record rather than assigned once per subject. Register further strictness contrasts as siblings rather than redefining this entry’s levels.

FED (canonical for fed-vs-fasted dose-record indicator)

• Description: 1 = fed state at dosing, 0 = fasted.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (fasted).
• Example models: Kyhl_2016_nalmefene.R, Goel_2016_Sonidegib.R (the Goel 2016 healthy-fasted F effect is applied via e_healthy_fast_f ^ (DIS_HEALTHY * (1 - FED)); cancer-patient records have FED = 1, healthy-fasted-arm records have FED = 0, high-fat-meal arm records have FED = 1 and the additional FED_HIGHFAT = 1 indicator), Niebecker_2015_edoxaban.R (the Niebecker 2015 “study 6” indicator is encoded via FED – study 6 was the only Fed-state phase 1 study in the pooled analysis; all 12 other phase 1 studies and the Hokusai-VTE phase 3 study were overnight-fast. Two multiplicative effects: ka * (1 + (-0.690) * FED) for the -69% absorption slowdown and CLnr/F * (1 + 0.204 * FED) for the +20.4% non-renal CL with food), Chen_2023_nemonoxacin.R (three multiplicative power-form effects from Chen 2023 Eq. 8: ka * 0.44^FED, T_lag * 1.6^FED, and F1 * 0.88^FED; the paper’s operational definition is any food taken within 2 h before or 30 min after the dose, i.e. a general fed-vs-fasted flag rather than a high-fat-meal challenge, so FED applies rather than FED_HIGHFAT), Marathe_2023_belzutifan.R (linear-deviation effect on the absorption rate constant: ka * (1 + (-0.88) * FED), an 87.6% reduction in ka when dosed fed (2.40 -> 0.30 1/h) per Marathe 2023 Table 2 KA-FED and Table 3; assessed in the crossover food-effect Study 2, so a subject contributes both FED levels. Food had no effect on AUCss, and the food effect on absorption lag time reached SCM significance but could not be retained in the final model), Comisar_2025_rimegepant.R (two multiplicative fractional effects per Comisar 2025 Table 3: -0.315 on relative bioavailability F1 and -0.706 on the transit rate constant ktr. The only fed data come from the crossover food-effect study BHV3000-112, whose fed arm received a HIGH-FAT meal, so the effect size is calibrated on a high-fat challenge – but FED is used rather than FED_HIGHFAT because the paper labels the covariate generically as ‘Food effect (fasting/fed)’ in Table 2 and applies it as a general fed flag across all 11 pooled studies. Neither effect was judged clinically meaningful, partly because migraine attacks are often accompanied by nausea so a food-intake recommendation would not reflect clinical practice), Rini_2013_axitinib.R (two linear-proportional effects per Rini 2013 Table 2: ka * (1 + 1.97 * (1 - FED)) regardless of crystal polymorph, and F * (1 + 0.330 * (1 - FED) * (1 - FORM_AXI_XLI)) for crystal polymorph Form IV only, since Table 2 footnote g reports no observed food effect with Form XLI – the same indicator therefore acts on two parameters with different gating), Hong_2016_pregabalin.R (the founding example of FED selecting between two INDEPENDENTLY ESTIMATED absorption parameter sets rather than scaling a single reference value: Table 5 reports separate MTT and ka for the fasted and fed states, so the model carries four absorption thetas and switches between them on the dose record. Two further features are worth carrying forward. The between-subject variability is asymmetric across the levels – MTT has a SEPARATE eta per prandial state (43.7% vs 81.9% CV) while ka has ONE eta shared by both – so the covariate partitions the random-effects structure and not just the fixed effects. And the source deliberately collapses finer prandial detail into this single flag: it gave regular and high-fat meals at 30 min and 4 h before dosing, and reports that neither meal type nor dose timing was ‘significantly different or discernible’, so FED_HIGHFAT is correctly NOT used despite a high-fat arm existing. Bioavailability is held constant across levels because the paper’s trapezoidal AUC comparison found no effect of food on extent, so FED alters rate only), Assmus_2022_emodepside.R (two fractional effects per Assmus 2022 Table 3: +114% on the absorption mean transit time and -24.4% on relative bioavailability, (1 + e * FED); the S1 Code FOOD column is coded 1 = fasted, 2 = fed, so FED = FOOD - 1), Chan_2020_fenebrutinib.R (time-varying; multiplies the transit MTT by 1.43 and the number of transit compartments by 0.864, and interacts with CONMED_PPI on MTT and F1).

FED_HIGHFAT (canonical for high-fat-meal-at-dosing indicator)

• Description: 1 = oral dose administered after a high-fat meal, 0 = oral dose administered under any other meal condition (typically fasted or light meal). Refines the more general FED indicator for studies that specifically test the high-fat-meal food effect on bioavailability or absorption rate (a common solubility-limited absorption phenotype).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (non-high-fat meal condition; most often “fasted” or “2 h post light meal” as defined per study protocol). Document the operational reference per model.
• Source aliases:
  • Fatmeal – used in Goel_2016_Sonidegib.R (covariate on F).
  • FATM – used in Goel_2016_Sonidegib.R (covariate on Ka; same indicator as Fatmeal in that paper).
  • YS – used in Yang_2024_dabigatran.R (Yang 2024 Supplementary Material $INPUT column list and the KAYS / CLYS / ALAG1YS covariate blocks; 1 = dosed within 30 min after a standard high-fat meal, 0 = fasted, so the same orientation as the canonical).
• Example models: Goel_2016_Sonidegib.R (multiplicative effect on F: 5.74^FED_HIGHFAT – ~5.7-fold higher F under high-fat meal vs 2 h post-light-meal reference; multiplicative effect on Ka: 1.01^FED_HIGHFAT – no meaningful effect), Yang_2024_dabigatran.R (linear proportional deviations on three absorption / disposition parameters simultaneously: ka * (1 - 0.24 * FED_HIGHFAT), tlag * (1 + 2.65 * FED_HIGHFAT) and cl * (1 + 0.51 * FED_HIGHFAT) – the high-fat meal slows absorption, more than triples the absorption lag time, and raises apparent clearance, reproducing the observed 2.05 h Tmax delay and 23% AUC reduction; Yang 2024 Table 4), Hu_2026_utreloxastat.R (linear additive effect on relative bioavailability: Frel *= (1 + 0.57 * FED_HIGHFAT), Hu 2026 Table 3 theta8, a +57% increase (90% CI 42% to 72%) relative to the fasted reference; the paired low-fat arm carries FED_LOWFAT at +36%. The two indicators are mutually exclusive per dose record and both are 0 under fasting), Gosselin_2015_motesanib.R (multiplicative factors on both absorption parameters: e_fed_highfat_ka = 0.0245 and e_fed_highfat_tlag = 1.079, i.e. a high-fat breakfast cuts the motesanib absorption rate constant roughly 40-fold, from 9.84 to 0.241 1/h, and lengthens the lag time by 7.9%; the indicator is time-varying because the food-effect substudy dosed the same 10 patients fed on one occasion and fasted on others), Bihorel_2023_molnupiravir.R (linear proportional shift on the Savic transit-chain mean transit time only: mtt * (1 + 4.22 * FED_HIGHFAT), a 422% increase per Bihorel 2023 Table 2, with relative bioavailability F1 fixed at 1 – so the high-fat meal delays and blunts the NHC peak (Tmax 1.5 h to 3.6 h, Cmax 11,455 to 9,133 nmol/L in the packaged model) while leaving AUC exactly unchanged, which is why molnupiravir may be dosed without regard to food. Time-varying per dose record: the food-effect part of the phase I trial MK-4482-004 dosed the same participants fasted and after a high-fat breakfast in a balanced crossover. Note the unusual provenance in the three COVID-19 trials, where food status was never collected (Table 1) – Bihorel 2023 assigned it by mixture modelling and fixed the high-fat fraction at 25%, then hard-coded that assignment into the analysis dataset, so for those 1107 participants the indicator is a modelling construct rather than observed data and simulations should draw FED_HIGHFAT ~ Bernoulli(0.25)), Kuchimanchi_2022_diroximelFumarate.R (source column FAT = 3; fractional changes -0.666 and -0.621 on the monomethyl fumarate and 2-hydroxyethyl succinimide absorption rate constants and -0.131 on monomethyl fumarate bioavailability, fixed at phase I estimates; sibling of FED_LOWFAT and FED_MEDFAT in a four-level meal factor), Lalovic_2020_lemborexant.R (standard high-fat meal: multiplicative factors 0.695 on ka and 1.21 on relative bioavailability, both fixed).
• Notes: Distinct from FED (binary fed-vs-fasted): FED_HIGHFAT carries the specific “high-fat meal” semantic (typically >= 800 kcal, >= 50% calories from fat per FDA guidance). Document the per-protocol meal definition in covariateData[[FED_HIGHFAT]]$notes. When a paper reports a high-fat-meal arm and a separate fasted-healthy arm (e.g., Goel 2016), use FED_HIGHFAT for the high-fat semantic and the existing FED + DIS_HEALTHY indicators for the composite healthy-fasted effect (Goel 2016 applies the e_healthy_fast_f effect via (DIS_HEALTHY * (1 - FED))); the retired HV_FAST composite indicator was deleted on 2026-05-11. Sibling of FED_LOWFAT for the low-fat-meal end of the meal-type spectrum.

FED_LOWFAT (canonical for low-fat-meal-at-dosing indicator)

• Description: 1 = oral dose administered after a low-fat meal, 0 = oral dose administered under any other meal condition (typically fasted, moderate-fat, or high-fat). Refines the more general FED indicator for studies that specifically test a low-fat-meal food effect on bioavailability or absorption rate, most commonly for drugs whose apparent bioavailability decreases with a light / low-fat meal relative to fasted (e.g., HIV integrase inhibitors, some antiviral tablets whose dissolution is optimised under fasted or moderate-fat conditions).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (non-low-fat meal condition; most often “fasted” or “moderate-fat / high-fat meal” as defined per study protocol). Document the operational reference per model.
• Source aliases:
  • LOWFAT – used in Bukkems_2021_raltegravir.R (NONMEM covariate flag; source paper Bukkems 2021 Methods ‘Development population pharmacokinetic model’ subsection ‘A low-fat meal (389 kcal, 6.9% fat)’; same orientation as the canonical, 1 = low-fat meal).
• Example models: Bukkems_2021_raltegravir.R (linear additive effect on bioavailability: F *= (1 + e_lowfat_fdepot * FED_LOWFAT) with e_lowfat_fdepot = -0.459; low-fat-meal-associated raltegravir dose records have -46% relative bioavailability vs the fasted / moderate-fat reference, Bukkems 2021 Table 2 ‘Factor change in F low-fat meal’), Hu_2026_utreloxastat.R (linear additive effect on relative bioavailability: Frel *= (1 + 0.36 * FED_LOWFAT), Hu 2026 Table 3 theta7, a +36% increase (90% CI 25% to 48%) relative to the fasted reference – note the sign is opposite to the raltegravir case. Hu 2026 pooled three meal conditions into this one indicator: the protocol low-fat low-calorie meal of the food-effect part, the site-standardized meal used throughout the SAD and MAD parts (fat providing at least 25% of calories, “very close to the standard low-fat meal as defined by FDA guidance”), and a customized medium-fat meal whose effect was not statistically distinguishable from low-fat. Document that pooling per model, because it makes FED_LOWFAT = 1 the usual state in this study rather than a special arm), Kuchimanchi_2022_diroximelFumarate.R (source column FAT = 1; one of four mutually exclusive meal levels alongside FED_MEDFAT, FED_HIGHFAT and FED_MISSING, with fasted as the reference. Fractional changes -0.368 on the monomethyl fumarate absorption rate constant, -0.335 on the 2-hydroxyethyl succinimide one and -0.296 on monomethyl fumarate bioavailability, plus a 0.421 h 2-hydroxyethyl succinimide absorption lag found only with the low-fat meal).
• Notes: Distinct from FED (binary fed-vs-fasted, generic) and from FED_HIGHFAT (high-fat-meal semantic). The low-fat semantic is typically defined per protocol as <= ~400 kcal with <= ~10% calories from fat (Bukkems 2021 references 389 kcal / 6.9% fat as the low-fat definition from the Rizk 2012 raltegravir food-effect study, and 650-844 kcal / 48% fat as the moderate-fat reference). Document the per-protocol meal definition in covariateData[[FED_LOWFAT]]$notes. When a paper reports discrete meal-type strata (fasted / low-fat / moderate-fat / high-fat), the recommended encoding is FED for any-food-vs-fasted plus one or both of FED_LOWFAT / FED_HIGHFAT for the specific low-/high-fat semantics; the moderate-fat stratum is the typical fed-reference unless the paper estimates it separately, in which case use the FED_MEDFAT sibling. Sibling of FED_HIGHFAT for the high-fat-meal end of the meal-type spectrum. General scope because the low-fat semantic is well-defined across studies (rather than tied to a single paper’s protocol).

FED_MEDFAT (canonical for medium-fat-meal-at-dosing indicator)

• Description: 1 = oral dose administered after a medium-fat (moderate-fat) meal, 0 = oral dose administered under any other meal condition. Completes the fat-content axis of the FED_* family between FED_LOWFAT and FED_HIGHFAT, for papers that estimate a separate effect for each of three standardised test meals rather than using the moderate-fat meal as the fed reference.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (not a medium-fat meal; usually the fasted state, as in the founding model). Mutually exclusive with FED_LOWFAT, FED_HIGHFAT, FED_PATIENTSELECTED, FED_LIQUIDSUPP and FED_MISSING per dose record. Document the operational reference per model.
• Source aliases:
  • MED, BFAT2, FAT = 2 – used in Kuchimanchi_2022_diroximelFumarate.R (Kuchimanchi 2022 Table 2 ‘MED administration with medium-fat meal’; the ESM control stream decodes the four-level FAT column into BFAT1 to BFAT4, and ESM Table S2 labels the level ‘Moderate’).
  • FED_MODERATE – the placeholder name that the FED_LOWFAT notes used for this sibling before it was ratified.
• Example models: Kuchimanchi_2022_diroximelFumarate.R (founding example; fractional changes -0.512 on the monomethyl fumarate absorption rate constant, -0.492 on the 2-hydroxyethyl succinimide absorption rate constant and -0.301 on monomethyl fumarate bioavailability, all fixed at the phase I base-model estimates from the study A109 food-effect crossover, whose medium-fat meal was consumed 30 min before the dose).
• Notes: General scope for the same reason as its low- and high-fat siblings: the medium-fat test meal is a standard food-effect protocol level (FDA guidance describes a moderate-fat meal of roughly 500-600 kcal with about 30% of calories from fat) rather than a single protocol’s construct. Record the per-protocol meal definition in covariateData[[FED_MEDFAT]]$notes. When a paper treats the moderate-fat meal as its fed reference and estimates only low- and high-fat deviations, this indicator is not used; the moderate-fat records then take all meal indicators 0 and FED = 1.

FED_LIQUIDSUPP (canonical for liquid-nutritional-supplement meal at dosing indicator)

• Description: 1 = the oral dose was taken with a ready-to-drink liquid nutritional supplement (PediaSure, Ensure, Boost and equivalents) rather than a solid test meal; 0 = any other prandial state. Extends the FED_* meal-type family along the meal form and lipid load axis rather than the fat-percentage axis: a paediatric or enteral nutritional drink delivers a far smaller absolute quantity of lipid than a standardised solid meal even when its fat percentage is comparable. That distinction matters for drugs whose absorption depends on bile-salt micellar solubilisation or on chylomicron-mediated lymphatic uptake, where the absolute lipid load, not the fat fraction, sets bioavailability.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (not a liquid nutritional supplement). As with FED_HIGHFAT, FED_LOWFAT and FED_PATIENTSELECTED, the meal-type indicators are mutually exclusive per dose record. Note that the reference is NOT necessarily the fasted state: in Hu_2025_vatiquinone.R the prandial reference is the medium-fat solid meal (FED = 1, FED_LIQUIDSUPP = 0) and the fasted state is carried separately on the complement (1 - FED). Document the per-model reference explicitly.
• Source aliases:
  • LQD – used in Hu_2025_vatiquinone.R (Hu 2025 Section 2.3 covariate equations, “LQDi assumes a value of 1 when a liquid PediaSure meal is ingested”; same orientation as the canonical).
• Example models:
  • Hu_2025_vatiquinone.R (multiplicative log-scale effect on relative bioavailability: Frel *= exp(-2.671 * FED_LIQUIDSUPP), Hu 2025 Table 2 theta12, i.e. exposure falls to 6.9% of the medium-fat-meal reference – reproduced as 0.07 for Cmax,ss, Cmin,ss and AUC0-24h,ss alike in Hu 2025 Table 3. Vatiquinone has cLogP 7.8 and is thought to reach the systemic circulation partly by chylomicron-mediated lymphatic transport, which is why the lipid load of the meal dominates its bioavailability).
• Notes: Distinct from FED (generic fed-vs-fasted), from FED_LOWFAT / FED_HIGHFAT (protocol-supplied solid test meals classified by fat percentage) and from FED_PATIENTSELECTED (who chose the meal). It is also distinct from the FORM_* family: FORM_SUSPENSION, FORM_SYRUP and siblings describe the DRUG PRODUCT’s physical form, whereas this column describes the MEAL taken with it. The two co-occur and are independent – Hu_2025_vatiquinone.R pools capsule (82.8%) and oral-solution (17.2%) records and models neither formulation effect, while modelling the liquid-supplement meal effect. Scope is specific rather than general because the effect size depends on the particular supplement’s volume and composition and on what the study’s reference meal was; promote to general when a second paper ratifies the concept with a comparable definition.

FED_PATIENTSELECTED (canonical for patient-selected-meal-at-dosing indicator)

• Description: 1 = the oral dose was taken with a meal the participant chose themselves under protocol dietary instruction (rather than a protocol-supplied standardised test meal); 0 = any other prandial state. Completes the FED_* meal-type family for the outpatient / real-world stratum that later-phase trials use, where the protocol prescribes a dietary pattern – typically a therapeutic-lifestyle-change or equivalent low-fat diet – but does not standardise the individual meal. The distinguishing feature is that the fat and calorie content is only nominally controlled, so the stratum’s bioavailability effect is an average over whatever participants actually ate and is not interchangeable with a standardised low-fat-meal effect.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (not a patient-selected meal). As with FED_HIGHFAT and FED_LOWFAT, the indicators are mutually exclusive per dose record and all are 0 under fasting, so the fasted state is the model’s prandial reference and is derived inside model() as the complement rather than carried as a fourth column. Document the protocol’s dietary instruction verbatim in covariateData[[FED_PATIENTSELECTED]]$notes, since what participants were told to eat is protocol-defined.
• Source aliases:
  • I_PB – Krishna 2011 Eq. 5 indicator variable (“patient-selected meal”); same orientation as the canonical.
• Example models: Krishna_2011_anacetrapib.R (multiplicative prandial term on bioavailability, e_fed_patientselected_fdepot = 3.3, selected by this indicator in Krishna 2011 Eq. 5. The phase Ib and phase IIb trials instructed participants to choose a meal conforming to the American Heart Association’s TLC diet, described as “similar in fat and caloric content to a low-fat meal”; the estimate is nonetheless the LARGEST of the three prandial multipliers – 3.3 against 2.4 for the standardised low-fat meal and 2.7 for the standardised high-fat meal – so it must not be collapsed into FED_LOWFAT, which would misreproduce every patient-selected prediction including the paper’s 100 mg phase III dose-selection case).
• Notes: Distinct from FED (generic fed-vs-fasted), FED_LOWFAT (a protocol-supplied standardised low-fat test meal) and FED_HIGHFAT (a protocol-supplied standardised high-fat test meal). The register’s FED_LOWFAT notes anticipate this family growing and flag that no FED_MODERATE sibling has been ratified; FED_PATIENTSELECTED is orthogonal to a fat-content axis, because it names who chose the meal, not how much fat it contained. That is why it is specific rather than general scope: the effect size is a property of one trial’s dietary-counselling protocol and its participants’ adherence, not of a reproducible meal definition. A paper that both prescribes a diet AND standardises the meal should use the fat-content indicator, not this one.

MEAL_DELAY_1H (canonical for delayed-meal indicator: food consumed >= 1 h after dosing rather than ~0.5 h after dosing)

• Description: 1 = the subject started eating at least 1 hour after taking the oral dose; 0 = the subject started eating approximately 0.5 hour after taking the dose. Encodes the dose-to-meal interval in studies where every subject is fed but the delay between dosing and the start of the meal is protocol-controlled or recorded. The concept it captures is how long the drug has to dissolve and empty from a relatively unfed stomach before the meal arrives, which shifts gastric-emptying-limited absorption timing (typically a zero-order absorption duration D1, a lag time, or ka) without necessarily changing overall bioavailability.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (food consumed ~0.5 h after dosing – the shorter dose-to-meal interval). Document the per-protocol interval definition in covariateData[[MEAL_DELAY_1H]]$notes, since the two levels a given paper contrasts may not be exactly 0.5 h vs 1 h.
• Source aliases:
  • FOOD – used in Wang_2023_dorzagliatin.R (Wang 2023 Table 3 footnote: “FOOD: 0 and 1 for the time of food consumption after 0.5 h and >= 1 h of drug administration, respectively”; same orientation as the canonical, 1 = the longer delay).
• Example models: Wang_2023_dorzagliatin.R (multiplicative exponential effect on the zero-order absorption duration D1: exp(0.816 * MEAL_DELAY_1H) = a 2.26-fold longer zero-order absorption window, 0.418 h -> 0.945 h, when the meal is delayed by at least 1 h; Wang 2023 Table 3 D1,FOOD = 2.26).
• Notes: Specific scope. The concept – the dose-to-meal interval – is genuinely new to the register, but the particular 1-hour threshold is Wang 2023’s own protocol design choice and has not yet been shown to be reusable, so the entry is scoped to that paper pending a second ratification. Promote to general when a second paper adopts the same 0.5 h vs >= 1 h dichotomy. Distinct from the whole FED family: FED is fed-vs-fasted (meal presence), FED_HIGHFAT / FED_LOWFAT / MEAL_A-MEAL_D are meal composition, and MEAL_FLAG / SNACK_FLAG are time-varying intra-day meal windows driving ongoing physiology. MEAL_DELAY_1H is orthogonal to all of these – both of its levels are fed, with identical meal composition, differing only in timing relative to the dose. A study that varies both composition and delay should carry FED_HIGHFAT (or FED_LOWFAT) and this indicator. Per-dose-record in principle; in a fixed-protocol study it is effectively subject-level. Because the effect typically lands on an absorption-timing parameter rather than on F or CL, the downstream exposure consequences are usually small: Wang 2023 Section 3.3 reports no change in steady-state AUCtau and only -1.26% / +2.71% shifts in Cmax,ss / Cmin,ss. Future models contrasting a different pair of intervals (e.g. immediately-with-food vs 2 h post-dose) should register a sibling indicator (MEAL_DELAY_2H) following the same pattern rather than redefining this entry’s levels. See MEAL_PREDOSE_2H for the mirror-image concept, in which the meal PRECEDES the dose.

MEAL_PREDOSE_2H (canonical for pre-dose-meal interval indicator: dose taken ~2 h after the start of the meal rather than ~0.5 h after it)

• Description: 1 = the oral dose was taken about 2 hours after the subject started eating; 0 = the oral dose was taken about 0.5 hour after the subject started eating, or the subject was fasted. Encodes the meal-to-dose interval for protocols in which the meal PRECEDES the dose, so the drug arrives in a stomach whose emptying and bile-salt state has already been perturbed for two hours rather than for thirty minutes. Composes with the meal-composition indicators (FED, FED_HIGHFAT, FED_LOWFAT), which continue to carry what was eaten; this entry carries only when it was eaten relative to the dose. The concept it captures is a bioavailability and absorption-timing difference between two fed protocols of identical meal composition.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (dose taken about 0.5 h after the start of the meal, or fasted). Document the per-protocol intervals in covariateData[[MEAL_PREDOSE_2H]]$notes, since the pair of intervals a given paper contrasts may not be exactly 0.5 h vs 2 h.
• Source aliases:
  • FOOD / prandial-state label – used in Mauro_2025_nilotinib.R and Mauro_2025_nilotinib_lowfat.R, where the four-level prandial-state factor of Mauro 2025 Methods, “Prandial state definitions”, is decomposed into FED + FED_HIGHFAT / FED_LOWFAT + this indicator. The paper’s level names are “fed” (MEAL_PREDOSE_2H = 0) and “modified fasting” (MEAL_PREDOSE_2H = 1).
• Example models: Mauro_2025_nilotinib.R (separates the paper’s “fed” state – high-fat meal started 0.5 h before the dose – from its “modified fasting with a high-fat meal” state – the same high-fat meal started 2 h before the dose. The two states carry different proportional effects on nilotinib tablet relative bioavailability: at 142 mg +48.4% at the 0.5 h interval vs +58.5% at the 2 h interval, and at 190 mg +61.7% vs +60.3%, per Mauro 2025 Supplemental Table 2. Applied as hf05h <- FED_HIGHFAT * (1 - MEAL_PREDOSE_2H) and hf2h <- FED_HIGHFAT * MEAL_PREDOSE_2H), Mauro_2025_nilotinib_lowfat.R (same decomposition with the low-fat 2 h state added as FED_LOWFAT * MEAL_PREDOSE_2H).
• Notes: General scope: the meal-to-dose interval is a protocol-level design variable, and a 2 h pre-dose interval is an approved-label dosing instruction for more than one marketed oral oncology product, so the concept is expected to recur. Direction matters and is the whole point of the name. MEAL_DELAY_<n> is the mirror concept – the meal comes AFTER the dose (MEAL_DELAY_1H = food consumed >= 1 h after dosing) – and MEAL_PREDOSE_<n> is the meal BEFORE the dose. Registering this as MEAL_DELAY_2H would have been the easy and wrong choice: it would have put a meal-precedes-dose concept into the meal-follows-dose family and inverted the physiology. Register further pre-dose intervals as siblings (MEAL_PREDOSE_4H,…) rather than redefining this entry’s levels. Do not identify this indicator with the bare phrase “modified fasting.” That phrase is not portable between papers and must be read from each paper’s own definition: Mauro 2025 nilotinib uses it for food first, dose about 2 hours later, whereas Olsson Gisleskog 2025 ibrutinib uses the same phrase for the opposite protocol – no food from about 30 min before to about 2 h after dosing, i.e. dose first, food later – which belongs to FASTED_STRICT, not to this entry. Always encode the protocol, never the phrase. Distinct from FED / FED_HIGHFAT / FED_LOWFAT (meal composition and presence) and from MEAL_FLAG / SNACK_FLAG (time-varying intra-day meal windows driving ongoing physiology). Per dose record in principle; in a fixed-protocol crossover it varies within subject across periods and is naturally paired with OCC.

MEAL_PREDOSE_4H (canonical for pre-dose-meal interval indicator: dose taken at least 4 h after the start of the preceding meal)

• Description: 1 = the oral dose was taken at least about 4 hours after the subject’s preceding meal (e.g. an evening dose taken >= 4 h after dinner); 0 = the reference prandial state of the paper (typically an overnight fast of >= 10 h before a morning dose) or a fed state carried by the meal-composition indicators. Sibling of MEAL_PREDOSE_2H on the same meal-precedes-dose axis, for a longer meal-to-dose interval: the stomach may still be emptying the earlier meal, so absorption rate and extent can fall short of the overnight-fasted reference even though no food is eaten near the dose.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (overnight fast >= 10 h, or any fed state carried by FED / FED_HIGHFAT / FED_LOWFAT). Document the per-protocol intervals in covariateData[[MEAL_PREDOSE_4H]]$notes.
• Source aliases:
  • Food = 1 of a three-level food status – used in Guan_2021_yimitasvir.R, where Guan 2021’s food status (0 = fasted >= 10 h, 1 = evening dose >= 4 h after dinner in the phase 1b study, 2 = high-fat meal) is decomposed into this indicator plus FED_HIGHFAT. The paper labels level 1 “Fasted at least 4 h” (Table 1, Figure 5).
• Example models: Guan_2021_yimitasvir.R (exponential effects on ka, exp(-1.71 * MEAL_PREDOSE_4H), and on relative bioavailability, exp(-0.341 * MEAL_PREDOSE_4H); steady-state AUC 28.9% and Cmax 46.1% lower than after an overnight fast, Guan 2021 Figure 5).
• Notes: Encode the protocol, not the label: Guan 2021 calls level 1 “fasted at least 4 h”, but the design contrast is a dose after an evening meal versus a dose after an overnight fast, i.e. a meal-then-dose interval, which is this family. In Guan 2021 the level-1 subjects also dosed in the evening rather than the morning, so the indicator absorbs any time-of-day effect as well; the paper attributes the difference to slower gastric emptying after dinner. Distinct from FASTED_STRICT (how tightly food is excluded on both sides of the dose) and from DOSETIME_EVENING (clock-time of dosing in a chronopharmacokinetic model).

MULTI_DOSE_PT (canonical for multiple-dose-phase-in-patients indicator)

• Description: 1 = dose record from the multiple-dose phase of a clinical-pharmacology study in patients, 0 = otherwise (single-dose run-in records, healthy-volunteer records, or first-dose records in patient studies). Captures any systematic shift in apparent bioavailability between the controlled run-in and the longer multiple-dose phase, typically driven by variable food-restriction compliance over many dosing days.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (single-dose / run-in / healthy-volunteer dose records).
• Source aliases:
  • FMDD – used in Goel_2016_Sonidegib.R (Goel 2016 covariate on F).
• Example models: Goel_2016_Sonidegib.R (multiplicative effect on F: 1.16^MULTI_DOSE_PT – ~16% higher apparent F during the multiple-dose phase relative to first dose, attributed in the paper to occasional non-fasting compliance), Fang_2010_etanercept.R (multiplicative effect on F: 0.674^MULTI_DOSE_PT – ~33% lower apparent F during the multiple-dose phase relative to the single-dose reference; Fang 2010 attributes the reduction to partitioning of the rhTNFR-Fc fusion protein into local subcutaneous adipose tissue with repeated injection. In Fang 2010 the multi-dose cohort is the AS-patient arm and the single-dose cohort is the healthy-volunteer arm, so MULTI_DOSE_PT is effectively subject-level: source column M), vanIersel_2018_posaconazole.R (multiplicative effect on apparent clearance: cl *= (1 + 0.750 * MULTI_DOSE_PT) – 75% higher CL in multiple-dose records relative to the single-dose reference; van Iersel 2018 Table 2 final-model ‘Dosing regimen on CL’ = 0.750), Han_2024_ainuovirine.R (multiplicative effect on apparent clearance: cl *= (1 + 1.47 * MULTI_DOSE_PT) – 147% higher CL/F at steady state than after the first dose, i.e. the 2.47-fold multiplier Han 2024 states in its Table 3 footnote, from the row labelled ‘Drugno on CL’. The source column name is Drugno (dose number). Han 2024 attributes the step chiefly to auto-induction of CYP3A4 by ainuovirine itself and estimates no time course for it, because the phase 1 design samples only two occasions – after the first dose on day 1 and after the last dose on day 10 – so records between those landmarks are uninformed. Record-level: a phase 1 subject contributes both levels, and every phase 3 sample (weeks 12-48) is a steady-state record), Jiang_2021_ivosidenib.R (per-record step between the first ivosidenib dose and repeated dosing, on BOTH relative bioavailability (0.50^(MULTI_DOSE_PT - 1)) and clearance (1.66^(MULTI_DOSE_PT - 1)); the steady-state state is the reference, and the indicator switches at the second dose, so the escalation cohort’s day -3 single dose and the expansion cohort’s cycle 1 day 1 dose both take 0).
• Notes: Specific scope because the indicator’s exact definition (dose-record level vs subject level, run-in inclusion, occasion boundary) is paper-specific. In Goel 2016, the dataset distinguishes the run-in single dose from the daily multiple-dose phase; the indicator switches at the start of the multiple-dose phase for cancer patients. In Fang 2010, the indicator is subject-level (all dose records of a multi-dose AS subject carry MULTI_DOSE_PT = 1; all dose records of a healthy-volunteer subject carry MULTI_DOSE_PT = 0). Distinct from FED and FED_HIGHFAT (which are per-record meal-state indicators) and from REGI_BID (regimen indicator). Future models that need a generic “occasion boundary” effect should consider the existing ooc<n> IOV pattern instead.

MULTI_DOSE (canonical for repeated-dosing indicator)

• Description: 1 = record within a repeated-dose regimen after its first dose, 0 = record from a single-dose administration or from the first dose of a regimen. Applies whatever the population (healthy volunteers in multiple-ascending-dose studies as well as patients). Per-record indicator.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (single or first dose).
• Source aliases:
  • MULTI – Wojciechowski 2022 Online Resource 2 (Single or multiple dosing (0 = single dose, 1 = repeated dose)).
• Example models: Wojciechowski_2022_abrocitinib.R (founding example; additive shift of +0.241 on logit absolute bioavailability, which also raises clearance through the effective-daily-dose term, so a single-dose subject keeps MULTI_DOSE = 0 after the dose).
• Notes: The population-independent form of [[MULTI_DOSE_PT]], whose 0 side explicitly includes healthy-volunteer records. Use MULTI_DOSE_PT only when a source restricts the contrast to the multiple-dose phase in patients. Carry it as a data column rather than deriving it from time when the source does: a time-derived flag cannot tell a single-dose subject’s post-dose records from a repeated-dose subject’s.

FORM_TABLET (canonical for tablet vs non-tablet oral liquid formulation indicator)

• Description: 1 = tablet formulation, 0 = the per-paper non-tablet oral liquid comparator (solution or suspension). Document the reference oral liquid form per-model in covariateData[[FORM_TABLET]]$notes.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (per-paper non-tablet oral liquid comparator; per-model documentation identifies whether the reference is a solution or a suspension).
• Source aliases:
  • TABLET – earlier name used by Kyhl_2016_nalmefene.R and Tikiso_2021_abacavir.R; renamed to FORM_TABLET for consistency with the FORM_* family (FORM_CAPSULE, future FORM_SUSPENSION, etc.).
• Example models: Kyhl_2016_nalmefene.R (additive shift on residual error: tablet vs solution), Tikiso_2021_abacavir.R (multiplicative effect on the absorption mean transit time MTT: tablet (abacavir + lamivudine fixed-dose-combination tablet) vs liquid solution; mtt *= (1 + 0.249 * FORM_TABLET), i.e. 24.9% slower absorption for the FDC tablet relative to the abacavir liquid reference), Kleideiter_2017_cebranopadol.R (tablet is the typical-value reference, FORM_TABLET = 1 leaves the formulation effects on ka, klag, and bioavailability at zero; paired with FORM_CAPSULE and the derived is_solution = (1 - FORM_TABLET) * (1 - FORM_CAPSULE) to encode a three-level formulation stratification), Fisher_2008_fosamprenavir.R (relative bioavailability of tablet vs fed-suspension reference: F_tab = 1.09; encoded as log(F) <- log(F_tab) * FORM_TABLET + log(F_food,sus) * (1 - FORM_TABLET) * (1 - FED) so tablet bioavailability does not additionally depend on food, while fasted suspension carries the F_food,sus reduction), Aruldhas_2021_R_methadone.R, Aruldhas_2021_S_methadone.R (per-oral-dose covariate that selects between two independently-estimated first-order absorption rate constants: suspension is the canonical reference Ka and tablet is a log-additive multiplicative shift; ka <- exp(lka + e_form_tablet_ka * FORM_TABLET)), Wada_2023_sparsentan.R (three-level sparsentan formulation stratification {100 mg capsule reference, whole tablet, crushed tablet} encoded into FORM_TABLET plus the sibling new canonical FORM_CRUSHED_TABLET, with both = 0 selecting the capsule; the whole tablet is absorbed more slowly than the capsule via ka * exp(-0.306) and has a shorter lag via tlag * exp(-0.269). NOTE: here the comparator is a CAPSULE, not the non-tablet oral liquid named in this entry’s default reference category – document the comparator per-model), Knebel_2011_pantoprazole.R (three-level {intravenous, delayed-release tablet, delayed-release granule} stratification encoded as FORM_TABLET plus FORM_GRANULE with is_iv = (1 - FORM_TABLET) * (1 - FORM_GRANULE); the tablet is the oral bioavailability anchor with F1 fixed to 1 and supplies the default absorption rate constant that the granule branch overrides. NOTE: here the comparator selected by both indicators = 0 is the INTRAVENOUS route, not an oral liquid. The pair also selects the residual-error model – proportional-only intravenously, proportional plus a shared additive term for either oral formulation), vanEijk_2022_paclitaxel.R (paclitaxel ModraPac spray-dried amorphous-solid-dispersion tablet vs the paclitaxel drinking solution, the oral liquid reference with rF = 1; paired with FORM_CAPSULE for a three-level formulation factor. The tablet carries relative gut bioavailability 0.97 and shares the Weibull absorption shape 3.57 with the capsule, against 2.53 for the drinking solution; the ritonavir booster has its own formulation indicator FORM_RTV_TABLET).
  • Zhu_2018_asunaprevir.R (Phase II tablet versus the Phase III / commercial SOFT-GEL CAPSULE reference – a second instance, alongside Wada 2023 sparsentan, where the comparator is a capsule rather than this entry’s default non-tablet oral liquid. Unusually the single indicator acts on three absorption quantities at once: ka * exp(-0.503), the zero-order release duration D1 * exp(0.864), and relative bioavailability exp(-0.215), so the tablet is absorbed more slowly, over a longer zero-order release, and less completely. Zhu 2018 notes the fitted 1.24-fold soft-gel-over-tablet bioavailability ratio is smaller than the roughly 2-fold seen in a dedicated relative-bioavailability study, because the tablet arm is confounded with the 600 mg dose level that carries its own DOSE_600MG bioavailability increment).
  • Lalovic_2020_lemborexant.R (tablet versus the early-phase-1 CAPSULE reference, as for Zhu 2018; multiplicative factors 0.254 on the zero-order duration D1 and 1.12 on ka, both fixed).
  • Liao_2022_lucitanib.R (film-coated tablet versus the hard gelatin CAPSULE reference; the single formulation effect is a multiplicative factor 0.299 on the zero-order release duration D1, 0.814 h for the capsule to 0.243 h for the tablet, estimated; source column FORM with FORM = 1 the tablet per the control stream TVD1 = THETA(8)*THETA(10)**FORM).
• Notes: Scope promoted from specific to general on 2026-07-24 alongside the Aruldhas 2021 methadone extraction (the sixth example model), where the non-tablet comparator is an oral suspension rather than an oral solution. Per-model documentation should identify which non-tablet oral liquid formulation is the reference (solution or suspension) in covariateData[[FORM_TABLET]]$notes. Distinct from the FORM_FDC canonical (Wilkins 2008 antitubercular fixed-dose-combination of multiple drugs, contrasted against single-drug tablets) because here the comparator is a non-tablet oral liquid rather than a separate tablet product. Future formulation-comparison models should extend this entry’s example list when the comparator is any non-tablet oral liquid (solution or suspension); register a sibling canonical when contrasting two tablet products.

FORM_FENEBRUTINIB_TABLET (canonical for fenebrutinib tablet vs powder-in-capsule formulation indicator)

• Description: 1 = fenebrutinib tablet, 0 = fenebrutinib powder-in-capsule. Time-varying across crossover periods.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (powder-in-capsule).
• Source aliases:
  • FORM – Chan 2020 Model S1 column, FORM = 2 = tablet; transform FORM_FENEBRUTINIB_TABLET = (FORM == 2).
• Example models: Chan_2020_fenebrutinib.R (multiplier 0.049 on the number of Savic transit compartments: 14.9 for the capsule, 0.73 for the fasted tablet).
• Notes: Member of the FORM_<drug>_<formulation> family. Distinct from the general FORM_TABLET, whose comparator is a non-tablet oral LIQUID; here the comparator is a powder-in-capsule. The capsule was given in phase 1 study GA29347 and one arm of GP29832; all other arms and the phase 2 ANDES trial used the tablet.

FORM_ALISERTIB_ECT (canonical for alisertib enteric-coated tablet vs powder-in-capsule formulation indicator)

• Description: 1 = alisertib enteric-coated tablet, 0 = alisertib powder-in-capsule. Subject-level, time-fixed.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (powder-in-capsule).
• Source aliases:
  • Formulation – Zhou 2022 Table 1 (Capsule / Enteric-coated tablet); Zhou 2018 Alisertib formulation.
• Example models: Zhou_2022_alisertib.R (relative bioavailability of the tablet 0.671, applied as f(transit1) <- e_form_ect_fdepot^FORM_ALISERTIB_ECT – founding example), Zhou_2018_alisertib.R (screened on KTR and F and not retained; documented in covariatesDataExcluded).
• Notes: Member of the FORM_<drug>_<formulation> family. In the paediatric analysis each formulation was used in only one study (capsule in ADVL0812, tablet in ADVL0921), so the effect is confounded with study and sampling design; the adult analysis found no formulation difference.

FORM_RTV_TABLET (canonical for ritonavir tablet vs soft-gel capsule formulation indicator)

• Description: 1 = ritonavir (Norvir) film-coated tablet, 0 = ritonavir (Norvir) soft-gel capsule. A per-dose-record property of the ritonavir booster dose, for models that embed ritonavir PK to drive a CYP3A4-inhibition term on a victim drug.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (soft-gel capsule; relative bioavailability 1).
• Source aliases:
  • Ftablet/capsule – Yu 2020 (J Clin Pharmacol 60:340) Table 2 parameter name for the relative ritonavir bioavailability of tablet versus capsule; the underlying data item is the switch from the capsule to the tablet formulation during the oral docetaxel studies.
• Example models: vanEijk_2022_paclitaxel.R (multiplies the ritonavir relative bioavailability by exp(lftab_rtv + etalftab_rtv) with lftab_rtv = log(1.06) and a 30% CV between-subject variability, both carried from the embedded Yu 2020 ritonavir model; van Eijk 2022 does not state which ritonavir formulation its studies used).
• Notes: Member of the FORM_<drug>_<formulation> family, with RTV the ritonavir abbreviation already used by CONMED_RTV and CONMED_RTV_AUC. Distinct from the general FORM_TABLET, whose comparator is a non-tablet oral liquid and which, in a booster model, already describes the victim drug’s formulation: carrying the booster’s formulation in its own column keeps the two drugs’ formulations independent. Distinct from CONMED_RTV, which records whether ritonavir is given at all.

FORM_CLO_GENERIC (canonical for generic-vs-Plavix clopidogrel product indicator)

• Description: 1 = the generic 75 mg film-coated clopidogrel test tablet evaluated in a bioequivalence study, 0 = the Plavix 75 mg film-coated reference tablet (Sanofi Winthrop Industrie, Ambares, France). Both products are film-coated tablets of the same strength, so this indicator distinguishes two products rather than two dosage forms; the contrast it carries is relative bioavailability, not a change in absorption mechanism.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (Plavix film-coated tablet – the reference medicine, whose bioavailability F is the fixed 100% anchor that makes the generic’s relative bioavailability identifiable).
• Source aliases: formulation, treatment (the crossover treatment-arm column of a 2-treatment bioequivalence dataset; FORM_CLO_GENERIC = as.integer(treatment == "test")).
• Example models: Pejcic_2024_clopidogrel.R (selects between the fixed reference F = 1 and the estimated study-specific generic relative bioavailability Fgen_st1 = 1.08 / Fgen_st2 = 0.960; frel <- exp(lfdepot) * (1 - FORM_CLO_GENERIC) + exp(lfgenTv) * FORM_CLO_GENERIC).
• Notes: Follows the auto-approved FORM_<drug>_<formulation> canonical family. Time-varying within subject in a crossover design – each subject receives both products, one per period – so it must be set per dose record and is naturally paired with OCC. Distinct from FORM_TABLET / FORM_CAPSULE, which contrast dosage forms; here both arms are the same dosage form and the covariate identifies the manufacturer’s product. Register a sibling FORM_<drug>_GENERIC canonical for other generic-vs-reference bioequivalence extractions rather than reusing this drug-specific entry.

FORM_TEDUGLUTIDE_GE10MGVIAL (canonical for teduglutide vial strength of 10 mg/vial or higher)

• Description: 1 = teduglutide administered from a vial strength of 10 mg/vial or higher (10, 20 or 65 mg/vial in the founding analysis), 0 = a lower vial strength (1.25, 2.5 or 5 mg/vial). A formulation-strength indicator, not a dose indicator: the same dose can be drawn from different vial strengths.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (vial strength below 10 mg/vial).
• Source aliases:
  • STRENGTH2 – Marier 2021 supplement control stream; a four-level category whose levels 1, 2 and 3 all take the single factor (1 + THETA(12)) (‘Combined Strength cat’), so FORM_TEDUGLUTIDE_GE10MGVIAL = as.integer(STRENGTH2 >= 1). Table S8 row label ‘x 0.476 for formulation strength >= 10 mg/vial’.
• Example models: Marier_2021_teduglutide.R (absorption lag time x 0.476, e_form_tlag^FORM_TEDUGLUTIDE_GE10MGVIAL, Marier 2021 Table S8; the largest single OFV drop of the forward-inclusion covariate search, Table S7 step 1).
• Notes: Follows the auto-approved FORM_<drug>_<formulation> canonical family. In Marier 2021 the higher strengths were used almost only in phase 1 studies (SBS patients received 1.25-10 mg/vials, non-SBS subjects 5-65 mg/vials; Table S4), so the indicator is partly confounded with study population.

FORM_CZP_LUVERINA (canonical for the Luverina-vs-Leponex clozapine product indicator)

• Description: 1 = Luverina (Celsius Laboratories), the “similar” clozapine product registered in Uruguay; 0 = Leponex (Novartis Laboratories), the brand-name reference product. Both are 100 mg oral clozapine tablets, so the indicator distinguishes two manufacturers’ products rather than two dosage forms, and the contrast it carries is relative bioavailability.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (Leponex, the brand-name reference product, whose oral bioavailability F is the fixed 1 anchor that makes the Luverina relative bioavailability identifiable).
• Source aliases: CZP formulation, brand (the Leponex / Luverina treatment-period column; FORM_CZP_LUVERINA = as.integer(brand == "Luverina")).
• Example models: Olmos_2019_clozapine.R (selects between the fixed reference F = 1 and the estimated relative bioavailability 0.892, which alone carries the bioavailability random effect; frel <- exp(lfdepot_leponex) * (1 - FORM_CZP_LUVERINA) + exp(lfdepot_luverina + etalfdepot_luverina) * FORM_CZP_LUVERINA).
• Notes: Auto-approved member of the FORM_<drug>_<formulation> canonical family, registered as a sibling of FORM_CLO_GENERIC per that entry’s own instruction to register a new canonical for each generic-vs-reference bioequivalence extraction rather than reuse a drug-specific one. The CZP token is the source paper’s own abbreviation for clozapine and is deliberately distinct from CLO, which FORM_CLO_GENERIC already uses for clopidogrel. Time-varying within subject, but by a sequential switch rather than a randomised crossover: every patient was on Leponex for more than a year, the hospital’s purchasing then changed, and sampling followed two months on Luverina. Period and product are therefore confounded, which matters because the founding paper’s design cannot separate a product effect from a period effect; record that limitation in covariateData[[FORM_CZP_LUVERINA]]$notes for any model reusing it. Must be set per dose record. Distinct from FORM_TABLET / FORM_CAPSULE, which contrast dosage forms.

FORM_DCN_SD (canonical for the diacerein PEG 8000 solid-dispersion formulation indicator)

• Description: 1 = the dose is the optimised diacerein solid dispersion (PEG 8000, drug:polymer 1:4 w/w, prepared by solvent evaporation), which is amorphous and releases about 95% of the dose within 15 minutes in vitro; 0 = plain crystalline diacerein powder, which releases about 48% within 60 minutes and no more.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (plain crystalline diacerein).
• Source aliases: optimized SD vs plain DCN (Fouad 2021 Table 6 row labels).
• Example models: Fouad_2021_diacerein.R (selects which in-vitro dissolution profile drives release from the depot of undissolved drug into gut_lumen: rdis <- ((1 - FORM_DCN_SD) * rdis_plain + FORM_DCN_SD * rdis_sd) / 100, where each rdis_* is the piecewise-constant slope of that formulation’s linearly-interpolated profile over 0-1 h and is zero afterwards).
• Notes: Auto-approved member of the FORM_<drug>_<formulation> canonical family. DCN is the source paper’s own abbreviation for diacerein. The indicator selects a release law, not a multiplier on ka or F. The bioavailability difference between the two formulations follows from where each profile plateaus (48.3% vs 99.1% dissolved), so do not add a separate relative-bioavailability term on top. Must be set per dose record.

FORM_RUTIN_EXTRACT (canonical for rutin-within-a-botanical-extract-matrix versus pure-compound indicator)

• Description: 1 = the administered rutin (quercetin-3-O-rutinoside) was delivered inside a botanical extract matrix; 0 = the same flavonoid was administered as the isolated pure compound. The contrast is a whole-matrix effect: the extract carries an uncharacterised mixture of co-constituents whose proposed mechanisms are displacement of rutin from plasma protein binding (raising free fraction, hence distribution volume and clearance) and P-glycoprotein inhibition at the enterocyte (raising conjugate absorption and enteric recirculation). It is therefore a property of the administered product, not of the animal. Name the specific extract in covariateData[[FORM_RUTIN_EXTRACT]]$notes; the founding models use the hydroethanolic extract of Physalis peruviana calyces standardised to 14.80 ug rutin per mg.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (pure rutin, the isolated analytical-standard compound).
• Source aliases: source of rutin – Dominguez More 2024 Section 2.3.2 (“using the source of rutin as categorical covariate (pure compound or extract)”); RUT / EXT – the arm labels used throughout that paper’s tables and figures.
• Example models: DominguezMore_2024_rutin_rabbit.R (shifts logit(V) by +0.678, log(k) by +0.625 and log(k12) by -0.634 after intravenous dosing), DominguezMore_2024_quercetin_rabbit.R (shifts log(ka1) by -0.949, log(ka2) by +3.528, log(V) by -1.910, log(k12) by +1.076 and log(k21) by +2.067 for the rutin conjugates after oral dosing), DominguezMore_2024_rutin_human.R (selects the pure-rutin or extract row of the Table 5 allometric projection, switching both the 70 kg typical value and the weight exponent).
• Notes: Auto-approved member of the FORM_<drug>_<formulation> canonical family, alongside FORM_CLO_GENERIC and the FORM_LNG_IUS* device selectors. The <formulation> token is EXTRACT rather than a named product because what is contrasted is “inside a plant extract” against “pure compound”, the standard design of a herbal-medicine matrix-effect study; a second botanical extract of rutin would reuse this column and record its identity in the per-model notes rather than register a sibling. Distinct from FORM_TABLET / FORM_CAPSULE / FORM_SOLUTION, which contrast dosage forms of one drug product, and from DOSE_GHI_MLKG, which is the volume dose of a herbal injection whose rutin content is one of eight quantified constituents (Chen_2024_rutin_rat.R) – that column is a continuous dose, not a matrix indicator. Confounding warning: in both founding rabbit models the extract arm and the pure-compound arm did not receive equal rutin doses (1.48 vs 0.37 mg/kg intravenously; 7.4 vs 100 mg/kg orally), because the extract dose was set by the extract’s own reference dose and rutin content. The authors fitted no dose effect, so any dose-nonlinearity is absorbed into this covariate’s coefficients; record the per-arm doses in covariateData[[FORM_RUTIN_EXTRACT]]$notes for every model that uses it.

FORM_MTX_NANO (canonical for the methotrexate nanoformulation versus free-solution indicator)

• Description: 1 = methotrexate (MTX) was administered inside a nanoformulation (drug-loaded polymeric nanoparticles or nanoemulsions, pooled); 0 = administered as the free methotrexate solution. Per-dose-record indicator. The contrast is between an encapsulated drug product and the unencapsulated drug, not between patients.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (free methotrexate solution).
• Source aliases: formulation type (Free methotrexate = 0; Nanoformulation = 1) – Jeong 2021 Equations (1)-(5); dPdFormulation – the coefficient names in Jeong 2021 Table 3.
• Example models: Jeong_2021_methotrexate_nanoformulation_rat.R (linear-deviation effects on V +0.429, CL -0.355, ka +10.883 and F +4.246).
• Notes: Auto-approved member of the FORM_<drug>_<formulation> canonical family. The value 1 pools two physically different products (PLGA nanoparticles and olive-oil/Labrasol nanoemulsions); a model that separates them uses FORM_MTX_NANOEMULSION within the nanoformulation arms instead, and the two columns are not complements of one another. Confounding warning: in the founding model the nanoemulsion arms received 83- to 208-fold lower doses (0.06 mg/kg oral, 0.024 mg/kg IV) than the free-solution and nanoparticle arms (5 mg/kg), and no dose effect was fitted, so dose-nonlinearity is absorbed into the coefficients.

FORM_MTX_NANOEMULSION (canonical for the methotrexate nanoemulsion versus polymeric-nanoparticle indicator)

• Description: 1 = methotrexate (MTX) was administered in a nanoemulsion (a soft-type, lipid-based nanoformulation); 0 = administered in polymeric (PLGA) nanoparticles (a hard-type nanoformulation). Per-dose-record indicator, defined only within nanoformulated doses; free methotrexate solution is outside its scope.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (methotrexate-loaded PLGA nanoparticles).
• Source aliases: formulation type (Nanoparticles = 0; Nanoemulsions = 1) – Jeong 2021 Equations (6)-(10); dPdFormulation – the coefficient names in Jeong 2021 Table 9.
• Example models: Jeong_2021_methotrexate_nanoemulsion_rat.R (linear-deviation effects on V -0.986, CL -0.990, ka +1.552 and F +0.193).
• Notes: Auto-approved member of the FORM_<drug>_<formulation> canonical family, a sibling of FORM_MTX_NANO (which contrasts any nanoformulation against the free solution). Name the specific nanoemulsion and nanoparticle compositions in the per-model notes; the founding model uses olive-oil/Labrasol nanoemulsions (173.77 nm) and PLGA nanoparticles (163.7 nm). The same dose confounding as FORM_MTX_NANO applies: 0.06 / 0.024 mg/kg nanoemulsion versus 5 mg/kg nanoparticle doses, with no dose effect fitted.

FORM_LNG_IUS20 (canonical for the levonorgestrel-releasing intrauterine system 20 (Mirena) device indicator)

• Description: 1 = the modelled device is the levonorgestrel-releasing intrauterine system 20 (LNG-IUS 20, Mirena), 0 = one of the sibling intrauterine systems. Mutually exclusive with FORM_LNG_IUS12 and FORM_LNG_IUS8; exactly one of the three is 1 in any record. The three devices differ in reservoir loading, release kinetics and indicated duration of use, so the indicator selects a whole parameter set rather than scaling a single effect.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: none of the three indicators is a reference level in the usual sense – the set is a mutually-exclusive selector, not a contrast against a baseline formulation. FORM_LNG_IUS20 = 1 additionally selects a structurally different release law (see Notes).
• Source aliases: not a data column in the source publication; Reinecke 2018 coded the device as a NONMEM branch on the treatment identifier (If LNG-IUS 20:... If LNG-IUS 12, LNG-IUS 8, implant:..., Supplemental Table S3a). Product-name aliases: Mirena, LNG-IUS 52 mg, LNG-IUS 20.
• Example models: Reinecke_2018_levonorgestrel_iusRelease.R (selects the LNG-IUS 20 release parameters and, uniquely among the three devices, switches release process 1 from zero-order to first-order in the remaining reservoir content and sets the shared correlation factor fc2 to zero: input1 <- FORM_LNG_IUS20 * c12 * depot + (1 - FORM_LNG_IUS20) * c12 and input2 <- c13 * (FORM_LNG_IUS20 + depot / (t1 + t))).
• Notes: Follows the auto-approved FORM_<drug>_<formulation> canonical family, alongside FORM_CLO_GENERIC. This indicator is load-bearing on model STRUCTURE and not only on parameter values: LNG-IUS 20 is the one device whose first release process is first-order and whose fc2 is zero. A single-device model for LNG-IUS 20 (for example Jensen_2023_lngIus52mg.R or Reinecke_2018_levonorgestrel_lngIus20.R) needs no indicator at all and should not declare one – the indicators exist only for models that carry more than one device in one file.

FORM_LNG_IUS12 (canonical for the levonorgestrel-releasing intrauterine system 12 (Kyleena) device indicator)

• Description: 1 = the modelled device is the levonorgestrel-releasing intrauterine system 12 (LNG-IUS 12, Kyleena), 0 = one of the sibling intrauterine systems. Mutually exclusive with FORM_LNG_IUS20 and FORM_LNG_IUS8.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: not a reference level; one of a mutually-exclusive selector set (see FORM_LNG_IUS20).
• Source aliases: not a data column in the source publication (see FORM_LNG_IUS20). Product-name aliases: Kyleena, LNG-IUS 19.5 mg, LNG-IUS 12.
• Example models: Reinecke_2018_levonorgestrel_iusRelease.R (selects the LNG-IUS 12 release parameters; shares the zero-order first release process and the fitted correlation factor fc2 with LNG-IUS 8).
• Notes: Follows the auto-approved FORM_<drug>_<formulation> canonical family. LNG-IUS 12 and LNG-IUS 8 share both the release law and the fc2 estimate in Reinecke 2018, differing only in reservoir loading and in their own c12 / c13 values, so a model carrying only those two devices needs a single indicator rather than two. See FORM_LNG_IUS20 for the family rationale.

FORM_LNG_IUS8 (canonical for the levonorgestrel-releasing intrauterine system 8 (Jaydess/Skyla) device indicator)

• Description: 1 = the modelled device is the levonorgestrel-releasing intrauterine system 8 (LNG-IUS 8, marketed as Jaydess and as Skyla), 0 = one of the sibling intrauterine systems. Mutually exclusive with FORM_LNG_IUS20 and FORM_LNG_IUS12.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: not a reference level; one of a mutually-exclusive selector set (see FORM_LNG_IUS20).
• Source aliases: not a data column in the source publication (see FORM_LNG_IUS20). Product-name aliases: Jaydess, Skyla, LNG-IUS 13.5 mg, LNG-IUS 8.
• Example models: Reinecke_2018_levonorgestrel_iusRelease.R (selects the LNG-IUS 8 release parameters; shares the zero-order first release process and the fitted correlation factor fc2 with LNG-IUS 12).
• Notes: Follows the auto-approved FORM_<drug>_<formulation> canonical family. The device numeral in the product name is the nominal in-vivo release rate in ug/day, not the reservoir loading (LNG-IUS 8 is loaded with 13.5 mg), so do not read a dose from the canonical name. See FORM_LNG_IUS20 for the family rationale.

FORM_DPV_FILM (canonical for the quick-dissolve dapivirine vaginal film indicator)

• Description: 1 = the dose record is a quick-dissolve dapivirine (DPV) vaginal film, 0 = the dose record is a Ring-004 dapivirine silicone intravaginal ring. Both are vaginally administered microbicide dosage forms for HIV pre-exposure prophylaxis, but they release drug on completely different timescales – the ring over weeks from a 25 mg silicone-matrix reservoir, the film over minutes from a 1.25 mg dissolving solid – so the indicator selects a release law rather than scaling a single effect.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (Ring-004 intravaginal ring; the formulation the source model was built on before the film was added).
• Source aliases: not a data column in the source publication. Kay 2018 implemented the ring and the film as two separate MATLAB SimBiology models rather than one model with a formulation covariate; the indicator exists so both live in a single nlmixr2 file, as FORM_LNG_IUS20 does for the Reinecke 2018 intrauterine systems.
• Example models: Kay_2018_dapivirine_pbpk.R (selects the release law – krel <- (1 - FORM_DPV_FILM) * k_depot_lumen_ring * exp(-k_ring_decay * tad(depot)) + FORM_DPV_FILM * k_depot_lumen_film – the luminal-fluid-to-epithelium transfer rate constant, 0.058 1/h for the ring versus 2.773 1/h for the film, and the bioavailability term, f(depot) <- 1 - FORM_DPV_FILM * (1 - exp(lfdepot))).
• Notes: Follows the auto-approved FORM_<drug>_<formulation> canonical family, alongside FORM_CLO_GENERIC and the FORM_LNG_IUS* device set. Load-bearing on model STRUCTURE, like FORM_LNG_IUS20: the ring’s release rate constant decays exponentially from the insertion of that particular ring and the film’s does not. It also gates bioavailability asymmetrically, which is easy to misread – Kay 2018 applies the same 30% bioavailability assumption to both formulations but enters it differently, back-solved into the ring’s release rate constant (so the ring’s F term is 1 and the dose is the full 25 mg load) and as a dose reduction for the film (1.25 mg x 0.30). A single-formulation vaginal model needs no indicator and should not declare one. A future vaginal microbicide extraction covering a gel arm should register a sibling FORM_<drug>_GEL rather than overload this entry.

FORM_BPN_FILM (canonical for the sublingual buprenorphine/naloxone film vs buprenorphine sublingual tablet indicator)

• Description: 1 = the sublingual (SL) buprenorphine dose record is a buprenorphine/naloxone SL film, 0 = the dose record is a buprenorphine SL tablet. Both are transmucosal SL products used for induction and maintenance in opioid use disorder; the film has a higher relative bioavailability and a slower apparent absorption than the tablet.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (buprenorphine SL tablet; the formulation of the phase IIa run-in on which the SL absorption parameters were estimated).
• Source aliases: not a data column in the source publication. Jones 2021 estimates the film-vs-tablet contrast as two multiplicative parameters, FRK14 (relative change in the SL absorption rate constant k14) and FRF1 (relative change in the SL bioavailability F1), applied to the phase III (film) run-in records.
• Example models: Jones_2021_buprenorphine.R (ka <- exp(lka + etalka) * e_form_film_ka^FORM_BPN_FILM with FRK14 = 0.636, and fdepot multiplied by e_form_film_fdepot^FORM_BPN_FILM with FRF1 = 1.47; scales only the SL depot, not the BUP-XR subcutaneous depots).
• Notes: Follows the auto-approved FORM_<drug>_<formulation> canonical family; BPN is the buprenorphine abbreviation already used by DOSE_BPN_SL_MG. Per-dose-record covariate: a subject switching between tablet and film carries the value of the product on each SL dose record. The naloxone component is poorly absorbed sublingually and has no effect on buprenorphine PK (Jones 2021 Methods 2.1), so the indicator describes the buprenorphine product only. Distinct from FORM_ODT (an orally disintegrating tablet that is swallowed and absorbed enterally) and from FORM_DPV_FILM (a vaginal film).

FORM_CAPSULE (canonical for capsule formulation indicator)

• Description: 1 = capsule formulation, 0 = the per-paper comparator non-capsule formulation. The complement formulation is paper-defined: solution for Hennig 2006 / 2007 itraconazole, tablet for Gupta 2016 lenvatinib, Doryx delayed-release tablet (with FORM_DOX_DORYX_MPC = 0) for Hopkins 2017 doxycycline. Document the comparator and the reference-category bioavailability per-model in covariateData[[FORM_CAPSULE]]$notes.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: Either FORM_CAPSULE = 0 or FORM_CAPSULE = 1, paper-defined. Hennig 2006 and Gupta 2016 anchor F = 1 on the non-capsule comparator (FORM_CAPSULE = 0): solution in Hennig 2006 with fdepot fixed to 1 for the solution arm; tablet in Gupta 2016 with F fixed to 1 for the tablet arm. Salem 2014 anchors F = 1 on the capsule arm (FORM_CAPSULE = 1): the capsule is the structural F = 1 reference and the liquid (suspension or solution) arm carries the estimated age-dependent relative bioavailability. The IIV (when present) is gated to the arm that carries the estimated F; document the orientation in covariateData[[FORM_CAPSULE]]$notes.
• Source aliases:
  • PREP – used in Hennig_2006_itraconazole.R (Clin Pharmacokinet 2006;45(11):1099-1114; PREP = 1 = capsule, PREP = 0 = oral solution) and in Hennig_2007_itraconazole.R (Br J Clin Pharmacol 2007;63(4):438-450; DOI 10.1111/j.1365-2125.2006.02778.x; same orientation, capsule typical absorption parameters as the published reference).
  • CAPSULE – earlier name used in the Hennig_2006_itraconazole.R and Hennig_2007_itraconazole.R model files before the FORM_* rename.
• Example models: Hennig_2006_itraconazole.R (Hennig 2006 Table II final estimates: capsule ka 0.09 h^-1 vs solution ka 0.96 h^-1 and capsule relative bioavailability 0.55 vs solution 1; the etalfdepot IIV applies only to the capsule arm), Hennig_2007_itraconazole.R (selects between lka_cap and lka_sol typical-value absorption rate constants and applies f(depot) <- (1 - FORM_CAPSULE) + FORM_CAPSULE * fdepot so the relative bioavailability F_rel = 0.817 is applied only to the capsule arm), Gupta_2016_lenvatinib.R (relative bioavailability of capsule vs tablet is 0.896; F1 fixed to 1 for the tablet reference; the etalfcap IIV (30.2% CV) applies only to the capsule arm), Lacy_2018_cabozantinib.R (multiplicative fractional effect of capsule (vs tablet reference) on Ka = -0.579 (57.9% slower absorption for capsule) and on overall bioavailability F = -0.144 (14.4% lower exposure for capsule); tablet F fixed at 1 as the reference; comparator capsule = Cometriq 140 mg approved for MTC), Kleideiter_2017_cebranopadol.R (paired with FORM_TABLET to encode the three-level cebranopadol formulation stratification: tablet reference, oral solution, liquid-filled capsule; multiplicative effects on ka (log shift log(2.09 / 0.864) = 0.883), klag (log shift log(0.077 / 0.087) = -0.122), and bioavailability (factor 1.174) for capsules relative to the tablet reference), Kleideiter_2018_cebranopadol.R (three-level formulation factor {tablet, oral solution, liquid-filled capsule} encoded into two binary indicators FORM_CAPSULE and the sibling new canonical FORM_SOLUTION; tablet is the reference when both indicators are 0; capsule multiplicative effects per Kleideiter 2018 Table 13 are 2.09 / 0.864 = 2.419 on Ka, 0.077 / 0.087 = 0.885 on klag, and 1.174 on bioavailability F), Hopkins_2017_doxycycline.R (three-level Doryx formulation factor {Doryx tablet (delayed-release), Doryx MPC (modified-acid-resistance delayed-release), Doryx capsule (conventional-release)} encoded into two binary indicators FORM_CAPSULE and the sibling canonical FORM_DOX_DORYX_MPC; Doryx tablet is the reference when both indicators are 0; Doryx-capsule relative bioavailability vs the Doryx-tablet reference is 0.978 per Hopkins 2017 Table 3 F1CAP, shared 0.115 h absorption lag with FORM_DOX_DORYX_MPC per Table 3 ALAG1, and the Doryx-capsule food effect on KTR matches the Doryx-tablet -20.9% reduction per Table 3 COVFED rather than the Doryx-MPC -54.9% reduction), Salem_2014_efavirenz.R (capsule is the structural F = 1 reference; the oral suspension and oral solution arms are pooled as ‘liquid’ since Salem 2014 found no difference in F between them, and the liquid arm carries an Emax age-dependent relative bioavailability with mature asymptote 0.79 (Salem 2014 Table 2 TVF, RSE 12.5%) and TM50,F = 10.6 months (Salem 2014 Table 2 TM50,F, RSE 38.7%); IIV on the mature liquid F is 39.9% CV (etaltvf_liq) and is gated to the liquid arm via fdepot <- FORM_CAPSULE * 1 + (1 - FORM_CAPSULE) * f_liquid), Sano_2023_fesoterodine.R (fesoterodine beads-in-capsule (BIC) vs tablet in pediatric patients with neurogenic detrusor overactivity; same orientation as Gupta 2016 lenvatinib, with the tablet arm fixed at F = 1 and the BIC arm carrying the estimated relative bioavailability 0.648 per Sano 2023 Table 2, applied as f(depot) <- (1 - FORM_CAPSULE) + FORM_CAPSULE * exp(lfdepot); note that in the source studies BIC was given only to the 25-kg-or-less cohort, so formulation is confounded with body weight and dose), vandenBerg_2021_uprifosbuvir_pbpk.R (uprifosbuvir capsule vs tablet formulation stratification per van den Berg 2021 Table 3: capsule multiplicative factors are 2.02 on KA1, 3.23 on the pre-logit for the fast-absorption fraction FDOS1 (corresponding to F1 = 0.91 for capsules vs 0.67 for tablets), 1.17 on bioavailability, and 1.65 on the fast-absorption lag time ALAG1; the capsule dose-slope on KA1 is -0.00189 (linear on log-KA1 per mg above the 150 mg reference); the gut-M6 concentration-dependent elimination rate uses a separate capsule-specific baseline KelM6g = 5.76 with a capsule dose slope of -0.00420), Majid_2024_lenvatinib.R (relative bioavailability of the lenvatinib capsule versus the tablet reference: f(depot) <- 0.882^FORM_CAPSULE, Majid 2024 Table 1 footnote F1 = 1 * 0.882^FORM; unlike the Gupta 2016 predecessor model no inter-individual variability is estimated on F1), Koh_2025_aspirin.R (enteric-coated aspirin capsule (Astrix, Boryungbio) vs enteric-coated tablet (Aspirin Protect, Bayer); the ONLY entry in this list where the indicator does not touch bioavailability at all – every Koh 2025 parameter is apparent (/F) and F is common to both arms, so the formulation acts solely on the first-order absorption rate constant, ka = 0.22 1/h for the capsule (estimated, RSE 21.8%) vs 0.053 1/h for the tablet (fixed during covariate analysis); the tablet is the reference category because it carries the fixed value, and Koh 2025 tested and rejected the same formulation split on Tk0, Lag0 and fr), Xu_2025_aficamten.R (formulation affects only the absorption lag time: two separately estimated lag times, 0.229 h for the pooled phase 2 / phase 3 / commercial tablet reference and 0.248 h for the phase 1 capsule, selected as tlag <- exp(ltlag_tab) * (1 - FORM_CAPSULE) + exp(ltlag_cap) * FORM_CAPSULE per Xu 2025 Table 1 and the Data S1 control stream line TVALAG1 = (THETA(6)*(1-CAPSULE) + CAPSULE*THETA(7)). Unlike the other members of this family no relative-bioavailability and no Ka effect of formulation was retained in the final model, so F1 is anchored at 1 on both arms and the two arms differ by 8% in lag time alone; source column FORM, where FORM <= 1 is capsule and FORM 2 / 3 / 4 are the phase 2, phase 3, and commercial tablets), Comisar_2025_rimegepant.R (three-level formulation factor {immediate-release tablet, capsule, orally disintegrating tablet} encoded into two binary indicators FORM_CAPSULE and the sibling new canonical FORM_ODT, with the tablet as the reference when both indicators are 0; the capsule effect is on the transit rate constant only, ktr * (1 + 2.03) per Comisar 2025 Table 3 ‘Capsule formulation on ktr’, i.e. 3.03-fold faster transit, with formulation effects on relative bioavailability and on CL/F both screened and then fixed to 0 in the final model – so unlike the other capsule-vs-tablet examples here this one is bioequivalent to its tablet reference in extent of absorption and differs only in rate. Only 17 subjects (4.9%) received the capsule, consistent with the wide interval on the effect), FernandezTeruel_2024_capivasertib.R (second member of the narrow lag-time-only subgroup alongside Xu_2025_aficamten.R: two separately estimated absorption lag times selected as (exp(ltlag_tab)*(1 - FORM_CAPSULE) + exp(ltlag_cap)*FORM_CAPSULE), 0.212 h for the tablet reference and 0.46 h for the capsule, per Fernandez-Teruel 2024 Table 3 Lag1_tab / Lag1_cap. Every parameter of that model is apparent (/F) and F is common to both arms – the paper states explicitly that “no other differences were detected between tablets and capsules” – so neither bioavailability nor Ka carries a formulation effect. Unlike Xu 2025, the lag term is additionally gated OFF entirely by FASTED_STRICT, so the formulation contrast exists only on non-overnight-fasted dose records), FernandezTeruel_2025_capivasertib.R (the successor fit of that same model by the same first author, adding the Phase III CAPItello-291 and China PK studies; identical lag-time-only encoding and identical FASTED_STRICT gating, with both lag times re-estimated – 0.287 h for the tablet reference and 0.468 h for the capsule per Fernandez Teruel 2025 Table 2. The formulation contrast narrows from 2.2-fold in the 2024 fit to 1.6-fold here. CAPItello-291, the Phase III trial that motivates the update, dosed the tablet exclusively, so the indicator is 0 throughout any Phase III simulation), Tamai_2017_lenvatinib.R (relative bioavailability of capsule vs tablet 0.867, the tablet again anchored at F = 1; unlike the Gupta 2016 fit this one estimates NO IIV on the formulation factor, the Table 2 interindividual-variability column being blank on that row), Nguyen_2019_cabozantinib.R, Nguyen_2019_cabozantinib_liver_dysfunction.R (the updated refit of the Lacy 2018 integrated cabozantinib model, and like it a member of the subgroup where capsule affects BOTH absorption rate and extent; Nguyen 2019 Table 3 reports both as direct MULTIPLICATIVE factors rather than the fractional changes its predecessor used, giving 0.402^FORM_CAPSULE on Ka and 0.847^FORM_CAPSULE on overall relative oral bioavailability in the final model, and 0.528 / 0.841 in the liver-dysfunction variant; tablet remains the F = 1 reference and the comparator capsule is Cometriq. Note the Ka factor is the least stable parameter across the two columns of Table 3, moving from 0.402 to 0.528 when the liver-dysfunction covariates are added, with both 90% CIs wide and overlapping), Thakkar_2018_tafenoquine.R (tablet anchored at F = 1 per Table 3 footnote a; the capsule carries BOTH an estimated relative bioavailability of 0.863 and a separate multiplicative effect of 0.924 on the absorption rate constant, the only registered member so far to act on Ka as well as on F – the Ka term survived into the final model on Cmax-prediction grounds even though backward elimination favoured dropping it), Liva_2021_rec2282.R (lag-time-only, like Xu_2025_aficamten.R, but with the CAPSULE as the reference: Liva 2021 Table 3 FORM = 0.74 is applied in the linear categorical form of Equation 2, ALAG = TV * (1 + 0.74 * (1 - FORM_CAPSULE)). The paper defines FORM only as ‘formulation (capsule vs. tablet)’ and does not state which level is coded 1 or how many patients took each, so the orientation is the maintainers’ reading of the first-named level as the reference; the effect is clinically negligible because the typical lag is under 10 minutes either way), vanEijk_2022_paclitaxel.R (paclitaxel ModraPac freeze-dried amorphous-solid-dispersion capsule vs the paclitaxel drinking solution reference with rF = 1, the orientation of Hennig 2006: relative gut bioavailability 0.46 for the capsule, Weibull absorption shape 3.57 shared with the FORM_TABLET arm; both indicators 0 selects the drinking solution), Tomita_2022_imeglimin.R (capsule vs tablet, with the optimised tablet as the reference and the conventional tablet carried by FORM_IMEGLIMIN_CONVENTIONAL_TABLET; fractional effects +2.45 on lag time and, shared with the conventional tablet, +0.298 on ka), Taneja_2019_glpg1690.R (capsule vs the oral-suspension comparator of the first-in-human study; the capsule acts only on the absorption rate constant, ka * (1 + 0.162 * FORM_CAPSULE) per Taneja 2019 Table 3 ‘KA (capsule)’, with F1 anchored at 1 for both formulations – only rifampicin coadministration moves relative bioavailability in that model).
• Notes: Scoped specific because the complement reference category is paper-defined (solution for Hennig itraconazole, tablet for Gupta lenvatinib). Sibling to FORM_TABLET (Kyhl 2016 / Tikiso 2021 tablet vs solution) under the FORM_* family. Future formulation-comparison models that need a capsule indicator should reuse this canonical, extending the example list and documenting the comparator in per-model notes.

FORM_ODT (canonical for orally disintegrating tablet formulation indicator)

• Description: 1 = the dose was administered as an orally disintegrating tablet (ODT), a fast-dispersing solid oral dosage form placed sublingually or on the tongue that disintegrates in saliva without water; 0 = the per-paper comparator formulation (usually a conventional swallowed immediate-release tablet). Per-dose-record indicator. Use this canonical when a source paper carries an ODT arm as a distinct level of a formulation factor and estimates separate absorption parameters (ka, transit rate constant, lag time, or relative bioavailability) for it. An ODT is pharmaceutically distinct from a conventional tablet because disintegration is not gastric-transit-limited, which typically manifests as a faster absorption rate rather than a change in extent of absorption.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (the per-paper comparator formulation; conventional immediate-release tablet in Comisar_2025_rimegepant.R). Document the comparator per-model in covariateData[[FORM_ODT]]$notes.
• Source aliases:
  • FORM (value 3) – used in Comisar_2025_rimegepant.R (Comisar 2025 Supplementary Listing 1 three-level formulation factor IF(FORM.EQ.1)...; Most common = tablet reference, FORM.EQ.2 = capsule, FORM.EQ.3 = ODT).
• Example models: Comisar_2025_rimegepant.R (three-level formulation factor {immediate-release tablet, capsule, orally disintegrating tablet} encoded into two binary indicators FORM_ODT and the sibling canonical FORM_CAPSULE, with the tablet as the reference when both indicators are 0; multiplicative fractional effect on the transit rate constant only, ktr * (1 + 0.355) per Comisar 2025 Table 3 ‘Oral disintegrating tablet on ktr’, a 35.5% faster transit versus the tablet. Formulation was also screened on relative bioavailability and on CL/F but both were fixed to 0 in the final model, so the ODT is bioequivalent to the tablet in extent of absorption and differs only in rate – the expected ODT signature. The rimegepant ODT is the marketed Nurtec / Vydura presentation).
• Notes: Auto-approved member of the FORM_* family, registered as a general dosage-form type alongside FORM_TABLET, FORM_CAPSULE, FORM_SOLUTION, FORM_SUSPENSION, FORM_SYRUP, FORM_GRANULE, and FORM_POWDER rather than as a drug-specific FORM_<drug>_<formulation> entry, because orally-disintegrating tablets are a generic presentation used across many drugs (rimegepant, ondansetron, olanzapine, rizatriptan, donepezil,…). Distinct from FORM_TABLET (which contrasts a conventional tablet against a non-tablet oral liquid) – an ODT arm compared against a conventional tablet needs its own indicator, since both arms are tablets. Distinct from FORM_SUBLINGUAL_FILM or similar buccal / transmucosal-absorption forms should one be needed later: an ODT disintegrates in saliva but is then swallowed and absorbed enterally, so it does not bypass first-pass metabolism, whereas a genuine sublingual / buccal product does. Where a paper administers the ODT both sublingually and on top of the tongue and finds no difference (as Comisar 2025 study BHV3000-110 part 2 did), a single FORM_ODT indicator covers both placements.

FORM_ODMT (canonical for orodispersible-minitablet formulation indicator)

• Description: 1 = the dose was administered as orodispersible minitablets (ODMTs), i.e. a multiple-unit presentation in which the full dose is made up of several small (typically 1-3 mm) fast-disintegrating tablets taken together; 0 = the per-paper comparator formulation (usually a single conventional swallowed immediate-release tablet of the same total strength). Per-dose-record indicator. Use this canonical when a source paper carries an ODMT arm as a distinct level of a formulation factor and estimates separate absorption parameters (ka, transit rate constant or mean transit time, lag time, or relative bioavailability) for it. ODMTs are the standard child-appropriate solid oral presentation in paediatric formulation development, because the dose can be titrated by counting units and the small size is swallowable from infancy.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (the per-paper comparator formulation; the market-authorised Renitec 2 x 5 mg conventional tablet in Faisal_2019_enalapril.R). Document the comparator per-model in covariateData[[FORM_ODMT]]$notes.
• Source aliases:
  • FORM (value 1 = ODMT, 0 = reference tablet) – used in Faisal_2019_enalapril.R (Faisal 2019 Equation 7, TV = theta_X * theta_12^FORM * exp(eta)).
• Example models: Faisal_2019_enalapril.R (ten 1 mg enalapril ODMTs versus two 5 mg Renitec tablets in a two-period crossover in 24 healthy adults; multiplicative effect on the enalapril mean absorption transit time only, mtt * 0.730^FORM_ODMT per Faisal 2019 Table 1 ‘THETA (X)’, i.e. a 27% shorter MTT1 and roughly 5 min earlier appearance of enalapril in serum. Formulation was screened on every other model parameter and retained on none, so the two presentations are bioequivalent in extent of absorption and differ only in the onset of absorption – and, because the prodrug’s active metabolite enalaprilat showed no formulation difference at all, with no expected pharmacodynamic consequence).
• Notes: Auto-approved member of the FORM_* family, registered as a general dosage-form type alongside FORM_TABLET, FORM_CAPSULE, FORM_ODT, FORM_SOLUTION, FORM_SUSPENSION, FORM_SYRUP, FORM_GRANULE and FORM_POWDER rather than as a drug-specific FORM_<drug>_<formulation> entry, because orodispersible minitablets are a generic paediatric presentation used across many drugs (enalapril, furosemide, lisinopril, spironolactone, and the wider EU FP7 LENA and CloSed programmes). Distinct from FORM_ODT, which is a SINGLE orally-disintegrating tablet placed on the tongue or sublingually and allowed to disperse in saliva WITHOUT water: the defining feature of an ODMT arm is the multiple-unit, small-particle presentation, and in the founding example the minitablets were deliberately swallowed with 240 mL of water in both arms so that in-mouth disintegration was NOT the contrast being estimated. The mechanism the founding paper invokes for the faster absorption onset is accelerated gastric emptying of small units rather than avoided disintegration time, which is a multiparticulate property and not an ODT property. A paper that administers minitablets sublingually without water, or that pools ODT and ODMT arms it did not separate, should carry FORM_ODT instead; a paper resolving both needs both columns.

FORM_CRUSHED_TABLET (canonical for crushed-tablet (dry, not resuspended) formulation indicator)

• Description: 1 = the dose was administered as a tablet crushed immediately before dosing and swallowed dry (or sprinkled on food) without being suspended in a liquid vehicle; 0 = the per-paper comparator intact formulation. Per-dose-record indicator. Use this canonical when a source paper carries a crushed-tablet arm as a distinct level of a formulation factor and estimates separate absorption parameters (Ka, lag time, bioavailability, or transit rate) for it.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (per-paper intact comparator; document which formulation is the reference per-model). For Wada 2023 sparsentan the reference is the 100 mg capsule, selected when FORM_CRUSHED_TABLET and the sibling FORM_TABLET are both 0.
• Source aliases:
  • Formulation three-level categorical column (capsule / tablet / crushed tablet) – the crushed-tablet level maps to FORM_CRUSHED_TABLET = 1. Used in Wada_2023_sparsentan.R.
• Example models: Wada_2023_sparsentan.R (third member of the three-level sparsentan formulation stratification {100 mg capsule reference, whole tablet, crushed tablet}, encoded into the two binary indicators FORM_TABLET and FORM_CRUSHED_TABLET; the crushed tablet is absorbed faster than either intact form via ka * exp(0.080) (Wada 2023 Table 2, RSE 159.1%) and has a markedly shorter absorption lag via tlag * exp(-1.175) (0.099 h vs the 0.32 h capsule reference); formulation had no effect on relative bioavailability, so AUC is unchanged and Cmax differs by <10%).
• Notes: Distinct from FORM_SUSPENSION (Svensson 2018: tablets crushed AND suspended in a liquid vehicle, i.e. an extemporaneously-prepared liquid dosage form) – the defining feature here is that no liquid vehicle is involved, so the dissolution and gastric-emptying behaviour differ. Also distinct from ROUTE_NGT (Denti 2018: crushed tablet delivered through a nasogastric tube), which is a route-of-administration indicator rather than a formulation state; a model pooling dry-crushed oral dosing with crushed-via-NGT dosing needs both columns. Sibling to FORM_TABLET / FORM_CAPSULE / FORM_SOLUTION / FORM_POWDER / FORM_GRANULE under the generic FORM_* family. Promote to general scope when a second paper ratifies the same encoding.

FORM_SOLUTION (canonical for oral-solution formulation indicator)

• Description: 1 = subject received the modelled drug as an oral solution, 0 = the per-paper comparator non-solution formulation (tablet or capsule). The complement reference is paper-defined; document the comparator and the reference-category absorption / bioavailability per-model in covariateData[[FORM_SOLUTION]]$notes.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (per-paper non-solution comparator; for Kleideiter 2018 cebranopadol the film-coated tablet is the typical-value reference, with the sibling FORM_CAPSULE also at 0 in the reference state).
• Source aliases:
  • FORM (three-level formulation column {tablet, oral solution, liquid-filled capsule}) – used in Kleideiter_2018_cebranopadol.R; the oral-solution level maps to FORM_SOLUTION = 1 and is paired with FORM_CAPSULE so that both = 0 selects the tablet reference.
• Example models: Kleideiter_2018_cebranopadol.R (third member of the three-level cebranopadol formulation stratification {tablet reference, oral solution, liquid-filled capsule}, encoded into the two binary indicators FORM_SOLUTION and FORM_CAPSULE; oral-solution multiplicative effects relative to the tablet reference per Kleideiter 2018 Table 13 are 2.43 / 0.864 = 2.813 on Ka, 0.077 / 0.087 = 0.885 on klag, and 1.045 on bioavailability F), Othman_2013_ABT_102.R (two-level oral-solution vs solid-dispersion comparator for the TRPV1 antagonist ABT-102; solution-vs-solid-dispersion-reference effects per Othman 2013 PK/PD-model Results paragraph 1 are 0.40 multiplicative on bioavailability F and 0.5 multiplicative on absorption lag time), PerezRuixo_2006_tipifarnib.R (oral-solution-vs-solid (capsule or tablet) absorption-effect indicator; ratio-form multiplicative effects on D1 (0.348), Ka (2.07), and tlag (0.183) applied as ratio^FORM_SOLUTION; reference category is the solid oral form – Perez-Ruixo 2006 found the absorption profile statistically indistinguishable between capsule and tablet, so the two solid forms are pooled as the reference, and the solution shows ~3-fold faster zero-order release, 2-fold higher Ka, and a 5.5-fold shorter lag time relative to the solid reference), Petric_2023_vinpocetine.R (third member of the three-level Petric 2023 vinpocetine formulation stratification {Ultra Vinca sustained-release beta-cyclodextrin tablet reference, Cavinton immediate-release tablet, 10 mg / 5 mL oral solution}, encoded into the two binary indicators FORM_SOLUTION and the sibling drug-specific FORM_VINP_IR; oral-solution log-additive effects relative to the SR-tablet reference per Petric 2023 Table 1 are exp(beta = -0.68) on Tk0 (49% shorter zero-order absorption duration) and exp(beta = -1.24) on V1/F (71% lower apparent central volume, i.e. higher metabolite exposure); metabolite AVA is the observed compound, so V/F folds in both the vinpocetine oral bioavailability and the vinpocetine-to-AVA conversion fraction), Chandasana_2024b_lamivudine.R (two-level oral-solution vs solid-oral-dosage-form contrast for absolute bioavailability F1; f(depot) <- exp(lfdepot * (1 - FORM_SOLUTION) + lfdepot_sol * FORM_SOLUTION) with F1 = 0.609 for the solid reference (tablet, capsule, and the ABC/DTG/3TC dispersible tablet) and F1 = 0.496 for the oral solution per Chandasana 2024 Table 3; the dispersible tablet is assigned to the solid reference because that assignment reproduces Chandasana 2024 Table 4 to within 2.4% in every weight band while the solution value under-predicts by ~26%), Chandasana_2024b_abacavir.R (used only as the formulation half of the ARROW PK Substudy Part 2 relative-bioavailability term, paired with STUDY_ARROW_PART2: the substudy tablet gets 1.62 and the substudy oral solution 1.75; outside that substudy the abacavir model applies no formulation effect on bioavailability, so this indicator is inert when STUDY_ARROW_PART2 = 0), Nguyen_2025_valbenazine.R (two-level oral-solution vs capsule contrast on the transit absorption rate constant KTR; per Nguyen 2025 Supplemental Table S2 equation (3) the coefficient is a PROPORTIONAL shift, KTR * (1 + 1.51 * FORM_SOLUTION), i.e. a 2.51-fold higher absorption rate constant for the oral solution than for the capsule reference. The oral solution appeared only in the six phase 1 studies; both phase 3 studies – KINECT-HD and KINECT 3 – used capsules, so the indicator is 0 throughout the phase 3 simulations), Bihorel_2023_molnupiravir.R (two-level oral-solution vs capsule contrast on the Savic transit-chain mean transit time MTT; per Bihorel 2023 Table 2 the coefficient is a PROPORTIONAL shift, MTT * (1 - 0.616 * FORM_SOLUTION), i.e. a 61.6% shorter mean transit time for the oral solution than for the capsule reference, with no accompanying change in the extent of absorption – relative bioavailability F1 is fixed at 1 and the Discussion states that neither food status nor formulation influenced bioavailability. The oral solution was given to only 36 of the 100 phase I MK-4482-004 participants; every participant in the three COVID-19 trials received capsules, so the indicator is 0 throughout any phase II/III simulation. Note the Table 2 row label reads “capsule or suspension”, i.e. an extemporaneous suspension shares the capsule reference category rather than the solution level), He_2022_paclitaxel.R (two-level contrast between paclitaxel intravenous solution given orally and the oral capsule, both co-administered with the P-glycoprotein inhibitor encequidar; per He 2022 Table S6 TVF1 = THETA(6)*(1 + THETA(9)*(2-FORM)) the coefficient is a PROPORTIONAL shift on bioavailability, F1 = 0.119 * (1 + 0.895 * FORM_SOLUTION), i.e. F1 = 0.226 for the solution against the capsule reference fixed at 0.119. Source column FORM is 1 = solution, 2 = capsule, so FORM_SOLUTION = 2 - FORM. The solution was used only in the first study, 15 of 197 patients).
• Notes: Scoped specific because the complement reference category is paper-defined. Third member of the oral solid-dosage FORM_* family alongside FORM_TABLET (tablet vs solution) and FORM_CAPSULE (capsule vs per-paper comparator); register prose for those entries already anticipates this name. Where a model carries all three formulations, FORM_CAPSULE + FORM_SOLUTION form the two-indicator encoding with tablet as the all-zero reference. Distinct from FORM_ASV_LIQUID (a drug-specific suspension/solution-vs-capsule/tablet indicator for asunaprevir).

FORM_ASV_LIQUID (canonical for asunaprevir liquid (suspension/solution) formulation indicator)

• Description: Asunaprevir (ASV) formulation indicator. 1 = ASV given as a suspension or oral solution (the higher zero-order absorption-fraction route); 0 = ASV given as a capsule or tablet (the reference formulation). The covariate has no effect when no ASV dose is administered.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (capsule or tablet; the reference formulation in the Wang 2018 ASV PK fit).
• Source aliases: none – the source Formulation column (“Suspension” / “Solution” -> 1; “Capsule” / “Tablet” -> 0, Wang 2018 Table 1) maps directly onto the canonical orientation; the model column is the canonical FORM_ASV_LIQUID.
• Example models: Wang_2018_daclatasvir_asunaprevir.R (switches the structural zero-order absorption fraction FK between fk_cap_asv = 0.184 (capsule/tablet) and fk_sol_asv = 0.334 (suspension/solution); both estimated with a shared 65.0% inter-arm-variability CV encoded as a single eta on the logit of FK, Wang 2018 Table 3).
• Notes: Specific scope because the liquid-vs-solid ASV-formulation contrast and the FK absorption-fraction switch are paper-specific to the Wang 2018 daclatasvir + asunaprevir analysis. Drug-specific member of the formulation indicator family; distinct from the generic oral solid-dosage FORM_TABLET / FORM_CAPSULE / FORM_SOLUTION indicators because the contrast here lumps both liquid forms (suspension and solution) against both solid forms (capsule and tablet) for a single named drug (ASV). Per-dose-occasion in principle (a participant could receive both formulations across occasions), but in the Wang 2018 trials each subject received a single ASV formulation. Not the complement of FORM_ASV_SOFTGEL, the other registered ASV formulation indicator: that one contrasts the phase-3 soft-gel capsule against the phase-2 film-coated tablet, a contrast that lives entirely inside this entry’s 0 level.

FORM_ASV_SOFTGEL (canonical for the asunaprevir soft-gel capsule formulation indicator)

• Description: Asunaprevir (ASV) formulation indicator. 1 = the dose was given as the phase-3 soft-gel capsule (100 mg twice daily), the food-effect-mitigating formulation selected for commercial use; 0 = given as the phase-2 film-coated tablet (200 or 600 mg twice daily), the reference formulation.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (phase-2 film-coated tablet; FORMref is stated as the phase-2 tablet formulation in Osawa 2018).
• Source aliases:
  • FORM – Osawa 2018 source-column convention, defined in the Table 2 footnote as “FORM, formulation (phase 2 film-coated tablet vs phase 3 soft-gel capsule)”.
• Example models: Osawa_2018_asunaprevir.R (single estimated coefficient -0.314 applied to BOTH CL/F and V/F via the shared-exponent parameter e_form_asv_softgel_cl_vc, giving 1 / exp(-0.314) = 1.369-fold higher relative bioavailability for the soft-gel capsule, matching the paper’s stated 1.37-fold).
• Notes: The paper tabulates this as an effect on F, not on CL/F and V/F separately, and the two readings are the same parameter. Bioavailability is not identifiable from oral-only data, so an F effect manifests identically in the apparent quantities CL/F and V/F; Osawa 2018 Table 2 therefore prints one coefficient labelled “F ~ FORM” while its displayed equations apply that coefficient to each of CL/F and V/F. Encode it once and multiply both – the elimination rate constant CL/V is then formulation-independent and only the concentration level moves, which is the intended meaning. The Osawa 2018 Figure 3 caption states the reading explicitly (“Relative bioavailability was computed from CL/F (phase 3 formulation, soft-gel capsule)/CL/F (phase 2 formulation, film-coated tablet), or V/F (phase 3 formulation)/V/F (phase 2 formulation)”). Distinct from FORM_ASV_LIQUID, the other registered asunaprevir formulation indicator: that one contrasts liquid forms (suspension, solution) against solid forms (capsule, tablet) in the Wang 2018 meta-analysis, whereas both arms here are solid oral forms and the contrast is soft-gel capsule versus film-coated tablet. A dataset pooling both analyses would need both columns, and they are not complements of one another. Also distinct from the generic FORM_CAPSULE / FORM_TABLET indicators, which do not name a drug. Specific scope, as for every drug-specific member of the FORM_<drug>_<formulation> family.

FORM_TRI_TAB200 (canonical for the tribendimidine 200-mg-tablet strength indicator)

• Description: Tribendimidine (TRI) formulation indicator. 1 = the dose was given as 200-mg enteric-coated tablets; 0 = given as 50-mg enteric-coated tablets (the reference formulation). Both arms are enteric-coated tablets from the same manufacturer (Shandong Xinghua Pharmaceutical Corporation), so the contrast is between two strengths of the same dosage form rather than between two dosage forms.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (50-mg enteric-coated tablets).
• Source aliases: STUDY – the trial index in the Vanobberghen 2016 final NONMEM control stream (supplemental File S1), which writes each effect as IF(STUDY.EQ.2) X = 1 (reference) and IF(STUDY.EQ.1) X = (1 + THETA), so FORM_TRI_TAB200 = as.integer(STUDY == 1). Study 1 used 200-mg tablets and study 2 used 50-mg tablets, so over this pooled data set the study index and the tablet strength are the same column; the canonical is named for the formulation because that is the mechanism the authors attribute the effect to.
• Example models: Meister_2019_tribendimidine.R (single multiplicative effect on absorption only, mtt *= (1 + 0.429 * FORM_TRI_TAB200); Meister 2019 Table 2 row “Formulation on MTT (%)” = 42.9. The larger pooled data set retained no formulation effect on either central volume, and pairs this indicator with the new FORM_TRI_SPLIT50 sibling), Vanobberghen_2016_tribendimidine.R (three simultaneous multiplicative effects relative to the 50-mg reference: mtt *= (1 + 0.401 * FORM_TRI_TAB200), vc *= (1 + 1.13 * FORM_TRI_TAB200) and vc_adadt *= (1 + 3.64 * FORM_TRI_TAB200), i.e. a 40.1% longer absorption mean transit time and 113% / 364% larger dADT / adADT apparent central volumes; Vanobberghen 2016 Table 1).
• Notes: Specific scope because the contrast is between two named tablet strengths of one drug. Drug-specific member of the FORM_<drug>_<formulation> branch of the FORM_* family, alongside FORM_ASV_LIQUID, FORM_COL_LIT and FORM_UPA_ER; it does not interoperate with the generic dosage-form members FORM_TABLET / FORM_CAPSULE / FORM_SOLUTION, and the FORM_TABLET entry’s own Notes anticipate exactly this case (“register a sibling canonical when contrasting two tablet products”). The mechanism the authors propose for the delayed absorption is that the 200-mg tablets were observed in vitro to float, prolonging gastric residence; they offer no mechanism for the volume effects and state the reason “is not clear”, so treat the two volume coefficients as empirical. Note the level this indicator does not resolve: the 25-mg dose in study 2 was given as split 50-mg tablets, destroying the enteric coating and producing visibly faster and less variable absorption, but Vanobberghen 2016’s model for that interaction did not converge, so split tablets are pooled into the FORM_TRI_TAB200 = 0 reference. A later extraction that needs to distinguish them should register a sibling rather than overload this indicator.

FORM_MF_INDGLYMF (canonical for mometasone furoate delivered as the indacaterol/glycopyrronium/mometasone furoate Breezhaler fixed-dose combination)

• Description: Mometasone furoate (MF) product indicator. 1 = the MF dose was given as the indacaterol/glycopyrronium/mometasone furoate (IND/GLY/MF) fixed-dose combination via the Breezhaler device (150/50/80 or 150/50/160 ug); 0 = MF given as the indacaterol/mometasone furoate (IND/MF) Breezhaler combination or as MF alone via Breezhaler.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (IND/MF Breezhaler FDC, or MF monotherapy via Breezhaler).
• Source aliases: IND/GLY/MF on Vc/F – the Bartels 2021 Table 5 row label for the estimated effect; the dataset column is not named.
• Example models: Bartels_2021_mometasoneFuroate.R (two effects: an estimated vc *= exp(-0.32 * FORM_MF_INDGLYMF) from Table 5, and a fixed relative bioavailability of 2 for the nominal MF dose, frel = 2^FORM_MF_INDGLYMF, from the Sect. 2.3 fixed dose-normalisation factor of 0.5).
• Notes: The fixed factor exists because the IND/GLY/MF product delivers a higher fine-particle mass of MF, so its 80 and 160 ug strengths were developed to match the 160 and 320 ug strengths of IND/MF (Bartels 2021 Table 2). The paper calls 0.5 “a fixed multiplicative factor on the bioavailability”, but read as a multiplier it would give IND/GLY/MF one quarter of the IND/MF exposure at matched ICS strength, contradicting the paper’s own report of equal simulated AUC0-24h; it is applied as a divisor of the nominal dose. Distinct from [[FORM_MF_MEDIUM]], the dose-strength indicator that applies to both FDC products. The MF Twisthaler monotherapy comparators (dose-normalisation factor 2.5) are a separate formulation whose effects Bartels 2021 does not report; do not encode them through this indicator.

FORM_MF_MEDIUM (canonical for the medium-dose mometasone furoate strength of the Breezhaler fixed-dose combinations)

• Description: Mometasone furoate (MF) dose-strength indicator. 1 = the MF dose was the medium-dose inhaled-corticosteroid strength of a Breezhaler fixed-dose combination (IND/MF 150/160 ug or IND/GLY/MF 150/50/80 ug); 0 = the high-dose strengths (IND/MF 150/320 ug or IND/GLY/MF 150/50/160 ug), or MF monotherapy via Breezhaler.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (high-dose MF FDC strengths).
• Source aliases: IND/MF and IND/GLY/MF medium-dose MF on F – the Bartels 2021 Table 5 row label; the dataset column is not named.
• Example models: Bartels_2021_mometasoneFuroate.R (relative bioavailability frel *= exp(0.18 * FORM_MF_MEDIUM), i.e. about 20% higher F for the medium strengths; Bartels 2021 Table 5 and Sect. 3.2).
• Notes: A strength-as-formulation member of the FORM_<drug>_<formulation> family, like [[FORM_TRI_TAB200]] and [[FORM_BXM_TAB10]]: the medium and high strengths are separate products, and the effect was estimated only for the medium FDC strengths. The E2201 MF Breezhaler monotherapy arms (80 and 320 ug) carry no strength effect and take the reference value. Medium and high follow the ICS dose-level labels of Bartels 2021 Table 2. Pair with [[FORM_MF_INDGLYMF]] to select the FDC product.

FORM_TRI_SPLIT50 (canonical for the split tribendimidine 50-mg-tablet indicator)

• Description: Tribendimidine (TRI) tablet-integrity indicator. 1 = the administered 50-mg enteric-coated tablet was broken in half before dosing, destroying the enteric coating; 0 = the tablet was swallowed whole (the reference). The contrast is the physical integrity of one tablet product, not a different product: both arms are the same 50-mg enteric-coated tablet from the same manufacturer.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (whole tablet).
• Source aliases: derived per dose from the trial’s dose level – in the pooled Meister 2019 data set only the 25-mg level of phase 2a trial 2 was administered as a split 50-mg tablet, so FORM_TRI_SPLIT50 = as.integer(dose_mg == 25) over that data set.
• Example models: Meister_2019_tribendimidine.R (founding example; mtt *= (1 + (-0.794) * FORM_TRI_SPLIT50), i.e. a 79.4% shorter mean absorption transit time than a whole 50-mg tablet; Meister 2019 Table 2 row “Split 50-mg tablets on MTT (%)” and the matching direction statement in Results, “a 79.4% faster mean absorption transit time for broken tablets than for whole 50-mg tablets”).
• Notes: The sibling canonical that the FORM_TRI_TAB200 entry’s Notes anticipated. FORM_TRI_TAB200 contrasts two tablet strengths (200 mg vs 50 mg) and explicitly does not resolve tablet integrity, because Vanobberghen 2016’s split-tablet interaction model did not converge and that earlier extraction therefore pooled split tablets into its FORM_TRI_TAB200 = 0 reference; Meister 2019’s larger pooled data set did identify the effect, so the two indicators are now carried side by side and a model using this column should set both. The two are not independent in the source data: split tablets were only ever 50-mg tablets, so FORM_TRI_SPLIT50 = 1 implies FORM_TRI_TAB200 = 0, and the combination FORM_TRI_TAB200 = 1, FORM_TRI_SPLIT50 = 1 is an extrapolation with no supporting observations. Distinct from FORM_CRUSHED_TABLET, which records a tablet reduced to powder (typically for administration through a feeding tube) rather than halved; the mechanism here is specifically loss of the enteric coating, which the authors expect to cause immediate release.

FORM_COL_LIT (canonical for the literature-extracted colchicine tablet formulation indicator)

• Description: Colchicine (COL) formulation indicator. 1 = the dose was one of the colchicine tablet products used in the published studies from which concentration-time data were digitised into the analysis dataset; 0 = the colchicine tablets used in the New Zealand Gout Study (the reference formulation). Per-dose-occasion in principle, per-study in the source data.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the NZ Gout Study colchicine tablets).
• Source aliases: STUDY – the analysis-dataset study index in the Wright 2025 final NONMEM control stream (IF(STUDY.EQ.1) FORM=1 for the NZ Gout Study; IF(STUDY.GT.1) FORM=1+THETA(11) for the three literature-extracted studies), so FORM_COL_LIT = as.integer(STUDY > 1). The paper’s Table 2 names the coefficient theta_FORM and glosses it as “bioavailability differences between data extracted from published studies and the present dataset”.
• Example models: Wright_2025_colchicine.R (multiplies the fixed oral bioavailability anchor: f1 <- exp(lfcentral) * (1 + e_form_col_lit_fcentral * FORM_COL_LIT), so F1 = 0.469 for the NZ Gout Study tablets and 0.469 * 1.55 = 0.727 for the literature formulations; e_form_col_lit_fcentral = 0.55, Wright 2025 Table 2 and footnote d).
• Notes: Specific scope because the contrast is between one named study’s colchicine tablets and the pooled tablet products of the studies Wright 2025 digitised (Thomas 1989, Ferron 1996, Rochdi 1994); it is not a generic dosage-form contrast and does not interoperate with FORM_TABLET / FORM_CAPSULE / FORM_SOLUTION, which distinguish dosage forms rather than products. Drug-specific member of the FORM_<drug>_<formulation> branch of the FORM_* family, alongside FORM_ASV_LIQUID and FORM_UPA_ER. Set to 0 when simulating the gout-patient population the model was built to describe – the literature arm exists only to stabilise the structural model against intensively sampled profiles, and its coefficient is imprecise (RSE 54.8%, SIR 95% CI 0.14-1.17, i.e. the 1.55 multiplier is not distinguishable from 1.14 or 2.17).

FORM_EMODEPSIDE_ASDA, FORM_EMODEPSIDE_ASDB (canonical for the emodepside amorphous-solid-dispersion tablet A / B indicators)

• Description: Emodepside formulation indicators against the oral liquid service formulation (LSF) solution reference. FORM_EMODEPSIDE_ASDA = 1 for the amorphous-solid-dispersion (ASD) tablet A (hypromellose acetate succinate polymer); FORM_EMODEPSIDE_ASDB = 1 for ASD-tablet B (copovidone polymer). Mutually exclusive; both 0 selects the 1 mg/mL LSF solution. Per-dose-record indicators.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 for both (LSF solution).
• Source aliases:
  • FORM2 – the Assmus 2022 S1 Code NONMEM column, coded 1 = LSF solution, 3 = ASD-tablet A, 4 = ASD-tablet B (code 2, the crystalline tablet X, was excluded from the analysis). FORM_EMODEPSIDE_ASDA = (FORM2 == 3), FORM_EMODEPSIDE_ASDB = (FORM2 == 4).
• Example models: Assmus_2022_emodepside.R (tablet A: +243% mean transit time, -31.4% relative bioavailability; tablet B: +124% mean transit time, -20.0% relative bioavailability; fractional multipliers (1 + e * FORM) on each parameter).
• Notes: Members of the FORM_<drug>_<formulation> family. Specific scope because the contrasts are particular to the emodepside formulations developed by Bayer / DNDi. ASD-tablet B is the formulation carried into phase II. Distinct from the generic FORM_TABLET / FORM_SOLUTION, which would not separate the two tablet variants.

FORM_UPA_ER (canonical for the upadacitinib extended-release tablet indicator)

• Description: Upadacitinib (UPA) formulation indicator. 1 = the dose was given as the extended-release tablet; 0 = given as the immediate-release capsule (the reference formulation). Per-dose-occasion indicator. The contrast is between two dosage forms of the same drug with materially different absorption behaviour, so it belongs to the FORM_<drug>_<formulation> drug-specific branch of the FORM_* family rather than to the generic oral solid-dosage members.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (immediate-release capsule). The immediate-release form is the bioavailability reference: Bhatnagar 2024 Table S3 footnote a states that the tabulated CL/F, Vc/F, Q/F and Vp/F “are for the immediate-release formulation (based on immediate-release bioavailability)”, and the extended-release arm carries a 76.2% relative-bioavailability factor on top of it.
• Source aliases:
  • FORM – source column in the Bhatnagar 2024 Appendix S1 NONMEM control streams, coded 1 = immediate release and 2 = extended release, i.e. FORM_UPA_ER = FORM - 1. Note the source coding is the OPPOSITE polarity to the canonical orientation in the sense that its low level is the extended-release-absent case only after the shift; always apply the subtraction rather than using FORM directly.
• Example models: Bhatnagar_2024_upadacitinib.R (switches the entire absorption branch: extended release uses Ka = 0.0523 1/h carrying the only absorption random effect, a 0.154 h lag, 74.5% of the absorbed dose entering the central compartment by a 3.29 h zero-order input, and 76.2% relative bioavailability; immediate release uses Ka = 2.77 1/h with no random effect, a 0.200 h lag, no zero-order arm and unit relative bioavailability – Bhatnagar 2024 Table S3), Klunder_2019_upadacitinib.R (the PARENT analysis from which Bhatnagar 2024 fixed those same absorption parameters; identical branch switch, with every value estimated rather than fixed – Klunder 2019 Table 3. This is the paper that actually pooled the immediate-release and extended-release arms, so both branches of the switch are exercised by real data here, whereas in Bhatnagar 2024 every subject received the extended-release tablet).
• Notes: Specific scope because the contrast and the absorption-branch switch are particular to the upadacitinib formulation pair. Unlike most FORM_* members, this indicator does not modify a single parameter: it selects between two complete absorption sub-models, including whether the zero-order arm exists at all and whether the absorption rate constant carries interindividual variability. Every subject in the Bhatnagar 2024 axial-spondyloarthritis analysis received the extended-release tablet, so FORM_UPA_ER = 1 reproduces that paper; the immediate-release branch is retained because it is fully parameterized in the source and defines the bioavailability reference. Distinct from FORM_TABLET / FORM_CAPSULE, which contrast dosage forms generically without implying a release-rate difference.

FORM_SAM_TAB (canonical for the samidorphan-alone tablet vs olanzapine/samidorphan bilayer tablet indicator)

• Description: Samidorphan (SAM) formulation indicator. 1 = the samidorphan dose was given as the samidorphan-alone immediate-release tablet; 0 = given as the olanzapine/samidorphan (OLZ/SAM) coformulated bilayer tablet (one olanzapine layer, one samidorphan layer). Per-dose indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (OLZ/SAM bilayer tablet, the marketed combination product).
• Source aliases:
  • FORM – Sun 2021 formulation covariate (‘OLZ/SAM bilayer tablet or tablet of olanzapine or samidorphan’).
• Example models: Sun_2021_samidorphan.R (multiplicative 1.41-fold factor on the absorption lag time, Sun 2021 Table 4 ‘Formulation (samidorphan tablet vs OLZ/SAM bilayer tablet) on ALAG’).
• Notes: Drug-specific member of the FORM_<drug>_<formulation> family. The contrast is a coformulated combination tablet versus the single-agent tablet of the same drug, so it is not the generic FORM_TABLET (tablet vs oral liquid). Sun 2021 screened the parallel olanzapine contrast and did not retain it.

FORM_VOSO_SOLN02 (canonical for the 0.2 mg/mL vosoritide dosing-solution indicator)

• Description: Vosoritide dosing-solution strength indicator. 1 = the subcutaneous dose was prepared from the 0.2 mg/mL vosoritide solution; 0 = prepared from the 0.8 mg/mL or 2 mg/mL solution. Per-dose-record indicator. The contrast is between reconstituted-solution concentrations of the same lyophilised drug product, not between dosage forms, so the covariate belongs to the FORM_<drug>_<formulation> drug-specific branch of the FORM_* family rather than to the oral solid-dosage members.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the 0.8 mg/mL and 2 mg/mL solutions – the strengths used in the phase III studies and in the commercial stock keeping units; Qi 2024 Discussion states these “did not have any effect on F”, so no adjustment is needed between commercial products).
• Source aliases:
  • SOLNC (“solution concentration”, a three-level column {0.2, 0.8, 2 mg/mL}) – used in Qi_2024_vosoritide.R; only the 0.2 mg/mL level carries an effect, so the three-level column collapses to this single binary indicator.
• Example models: Qi_2024_vosoritide.R (multiplies relative bioavailability by 1.56, Qi 2024 Table 5 “Effect of SOLNC (0.2 mg/mL)”; encoded on the log scale as e_form_voso_soln02_fdepot = log(1.56) and confirmed by Table 3, where the 0.2 mg/mL column is a constant 56% above the time-only reference at every tabulated time).
• Notes: Specific scope because the effect is tied to vosoritide’s reconstituted-solution strengths. The 0.2 mg/mL solution was used only in study 111-202 of the pooled five-trial analysis, so the indicator is 0 for every phase III and commercial record. Qi 2024 draws the analogy to insulin, where dilution of the dosing solution is also known to change subcutaneous bioavailability. A future paper reporting a dosing-solution-strength effect for a different drug should register a parallel FORM_<drug>_SOLN<strength> canonical rather than overloading this name.

FORM_ROPI_SOLN05 (canonical for the 0.5 % w/v ropivacaine injectate-concentration indicator)

• Description: Ropivacaine injectate-concentration indicator. 1 = the regional block was performed with the 0.5 % w/v (5 mg/mL) ropivacaine solution; 0 = a different solution strength was used. Per-administration indicator, but time-fixed per subject in single-shot block studies. As with FORM_VOSO_SOLN02, the contrast is between concentrations of the same injectable drug product rather than between dosage forms, so the covariate belongs to the FORM_<drug>_<formulation> drug-specific branch of the FORM_* family. The administered mass is held constant across strengths (3 mg/kg in the founding study), so only the injected volume, and hence the geometry and surface area of the fascial-plane drug depot, changes.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the 0.25 % w/v solution, when FORM_ROPI_SOLN075 is also 0). The two indicators are mutually exclusive and both zero selects the 0.25 % stratum.
• Source aliases:
  • The concentration of ropivacaine – the three-level randomisation column {0.25 %, 0.5 %, 0.75 %} of Ling 2025 Table 1 and Table 3, decomposed here into two binary indicators against the 0.25 % reference.
• Example models: Ling_2025_ropivacaine.R (selects the first-order absorption rate constant estimated in the 0.5 % stratum, ka = 19.4 1/h, against the 0.25 % reference stratum ka = 32.0 1/h; Ling 2025 Table 3 splits its “ka” row into one sub-row per solution strength, so the three estimates are parallel stratum values rather than a reference plus offsets).
• Notes: Specific scope until a second regional-anaesthesia paper ratifies a solution-strength effect for ropivacaine. The numeric token drops the decimal point exactly as FORM_VOSO_SOLN02 does for 0.2 mg/mL, so SOLN05 is 0.5 % w/v and SOLN075 is 0.75 % w/v; a future model that needs an explicit 0.25 % indicator (rather than using it as the reference) should register FORM_ROPI_SOLN025. Ling 2025 also enrolled two patients given a 0.375 % solution, but they were held out for external validation and are not part of any fitted stratum, so no indicator is defined for that strength. Distinct from DOSE_* covariates: the administered dose in mg/kg is identical across the strata and it is the concentration of the injectate, not the amount of drug, that carries the effect.

FORM_ROPI_SOLN075 (canonical for the 0.75 % w/v ropivacaine injectate-concentration indicator)

• Description: Ropivacaine injectate-concentration indicator. 1 = the regional block was performed with the 0.75 % w/v (7.5 mg/mL) ropivacaine solution; 0 = a different solution strength was used. The direct partner of FORM_ROPI_SOLN05; see that entry for the family placement and for why the contrast is a solution-strength rather than a dosage-form one.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the 0.25 % w/v solution, when FORM_ROPI_SOLN05 is also 0).
• Source aliases:
  • The concentration of ropivacaine – the same three-level Ling 2025 randomisation column decomposed against the 0.25 % reference.
• Example models: Ling_2025_ropivacaine.R (selects the 0.75 % stratum first-order absorption rate constant, ka = 14.4 1/h; Ling 2025 Discussion reads the monotone decline in ka across 0.25 %, 0.5 % and 0.75 % as poorer diffusion of the more concentrated, lower-volume injectate away from the serratus anterior fascial plane).
• Notes: Specific scope, registered together with FORM_ROPI_SOLN05. Ling 2025 additionally fits a post-hoc log-linear regression through the three stratum estimates in order to interpolate ka for the held-out 0.375 % patients; that regression is an external-validation device rather than part of the final model and is not encoded by these indicators.

FORM_UNDIL_SUSP (canonical for undiluted oral suspension formulation indicator)

• Description: 1 = subject received the modelled drug as an undiluted oral suspension, 0 = the per-paper comparator non-undiluted-suspension formulation (tablet or diluted oral suspension in Willmann 2018; documents the comparator per-model in covariateData[[FORM_UNDIL_SUSP]]$notes). The undiluted oral suspension was distinguished from the diluted oral suspension because pre-clinical in-vitro dissolution work showed the rivaroxaban suspension excipients limit dissolution of drug particles at low pH unless diluted before administration (Willmann 2018 Discussion paragraph 5).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (per-paper comparator non-undiluted-suspension; for Willmann 2018 this lumps the tablet and the diluted oral suspension as a single reference because the paper found their ka statistically indistinguishable).
• Source aliases: derived per dose record from the EINSTEIN-Jr phase I formulation column (Tablet / Undiluted suspension / Diluted suspension); FORM_UNDIL_SUSP = 1 iff the formulation column is “Undiluted suspension”.
• Example models: Willmann_2018b_rivaroxaban.R (multiplicative log-shift on the absorption rate constant: ka <- exp(lka + e_undilsusp_ka * FORM_UNDIL_SUSP) with lka = log(0.717) h^-1 and e_undilsusp_ka = log(0.208 / 0.717) = -1.238 per Willmann 2018 Table 1, RSE 21.3% and 15.4% respectively; the undiluted-suspension ka is ~3.4-fold slower than the tablet / diluted-suspension reference, consistent with the in-vitro dissolution result reported in the Discussion).
• Notes: Specific scope because the “undiluted suspension vs tablet-or-diluted-suspension” lumping is tied to the Willmann 2018 EINSTEIN-Jr phase I formulation design (which deliberately tested both undiluted and diluted variants of the same oral suspension after observing delayed absorption with the undiluted formulation). Sibling to FORM_TABLET / FORM_CAPSULE / FORM_SOLUTION under the FORM_* family. Future paediatric or low-pH-sensitive-drug models that distinguish an undiluted suspension from a diluted one with the same composition may extend this entry’s example list; models that contrast a suspension against a tablet without the diluted-vs-undiluted distinction should reuse FORM_TABLET (= 0 for the suspension comparator) instead.

FORM_ABA_PHASE2 (canonical for the abatacept SC phase-2 formulation indicator)

• Description: 1 = subject received the abatacept Phase-2 SC formulation (lower-pH excipient blend, ~56% absolute bioavailability), 0 = subject received the Phase-3 / commercial SC formulation (the ~81% bioavailability reference). Per-dose-occasion indicator: a single subject can carry both indicator values across phase-2-to-phase-3 transition arms.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (Phase-3 / commercial 125 mg SC formulation; the typical-value bioavailability reference).
• Source aliases: none – the canonical name matches the source-paper column.
• Example models: Li_2019_abatacept.R (additive shift on the logit-scale bioavailability: logit_F = logit_F_TV + FORM_ABA_PHASE2 * (-1.16); Li 2019 Table 2 reports the Phase-2 formulation effect as -1.16 logit-scale units, mapping to an absolute F of ~0.56 vs the Phase-3 reference ~0.81).
• Notes: Scoped specific because the Phase-2 vs Phase-3 formulation contrast is tied to the abatacept clinical-development timeline and does not generalise to other drugs. Future Phase-2-vs-commercial bioavailability comparisons for unrelated drug programmes should register their own FORM_<drug>_PHASE2 canonical rather than reuse this entry. Distinct from the generic FORM_TABLET / FORM_CAPSULE / FORM_SUSPENSION family (oral solid-dosage manipulations) because the contrast here is two SC injectable formulations with different excipient pH.

FORM_PANO_CSF (canonical for the panobinostat clinical service formulation indicator)

• Description: 1 = subject received the panobinostat clinical service formulation (CSF), the early-development oral formulation used in studies B2101, B2102 and B1101; 0 = subject received the final market image (FMI), the formulation intended for commercialization and used in every other oral study of the pooled analysis. Per-subject (study-fixed) indicator. In Savelieva 2015, 106 of the 494 orally dosed patients received the CSF and 388 the FMI. The indicator is irrelevant to intravenously dosed subjects, who bypass the depot entirely; the source control stream handles them with a separate IF (IV.EQ.1) branch that zeroes the absorption rate constant and the lag.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (final market image, FMI; the commercial formulation and the typical-value absorption reference).
• Source aliases:
  • FORM – used in Savelieva_2015_panobinostat.R and Savelieva_2015_panobinostat_allometric.R (Savelieva 2015 Supplementary Tables S2a and S3a, where the absorption rate constant is written as the linear switch KA = THETA(7)*(1-FORM) + THETA(8)*FORM, with THETA(7) labelled KA.FMI and THETA(8) labelled KA.CSF in the matching parameter tables). The canonical orientation matches the source coding, so no value flip is needed.
• Example models: Savelieva_2015_panobinostat.R (first final model: selects the absorption rate constant only, 0.544/h for the CSF against 0.321/h for the FMI, i.e. absorption roughly 70 percent faster from the early formulation), Savelieva_2015_panobinostat_allometric.R (second final model: selects BOTH the absorption rate constant, 0.631/h against 0.420/h, AND the absorption lag time, 0.296 h against 0.162 h).
• Notes: Formulation affects absorption but not bioavailability. Savelieva 2015 states in its Covariate analysis that the formulation “had a significant impact on the Ka (but not on the bioavailability) of oral panobinostat,” and both control streams apply a single bioavailability THETA to the two oral formulations. Do not reach for this entry to encode a formulation effect on F; that would contradict the founding source. Scoped specific because the CSF-vs-FMI contrast is tied to the panobinostat development programme. Distinct from the generic FORM_TABLET / FORM_CAPSULE / FORM_SUSPENSION family, which name a dosage form rather than a development-stage formulation, and analogous in role to FORM_PEX_PHASE1 and FORM_ABA_PHASE2; a future early-vs-commercial formulation contrast for an unrelated drug should register its own FORM_<drug>_<formulation> canonical rather than reuse this one.

FORM_PEX_PHASE1 (canonical for the pexidartinib Phase 1 clinical formulation indicator)

• Description: 1 = subject received the pexidartinib Phase 1 clinical formulation (used in Phase 1 dose-ranging studies PLX108-01 and the relative-bioavailability study U114; relative bioavailability 0.855 vs the Phase 3 / commercial formulation reference); 0 = subject received the Phase 3 / commercial formulation (used in healthy-volunteer studies U116-U121 and in the Phase 3 ENLIVEN trial; the typical-value bioavailability reference at F1 = 1.0). Per-dose-occasion indicator: within a single study the formulation is fixed, but a subject can in principle carry both values across studies.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (Phase 3 / commercial formulation; the typical-value bioavailability reference).
• Source aliases: none – Yin 2020 codes the formulation as a categorical “Phase 1” vs “Phase 3” column; the canonical FORM_PEX_PHASE1 name matches the indicator direction (1 = Phase 1).
• Example models: Yin_2020_pexidartinib.R (multiplicative effect on the depot bioavailability: F1 = 1.0 when FORM_PEX_PHASE1 = 0 (Phase 3 reference) and F1 = 0.855 * exp(etalfdepot) when FORM_PEX_PHASE1 = 1, where the 0.855 typical-value anchor is held fixed in the source model and the IIV etalfdepot ~ N(0, 0.101) is gated by FORM_PEX_PHASE1 so it has no effect on Phase 3 subjects).
• Notes: Scoped specific because the Phase-1-vs-Phase-3 formulation contrast is tied to the pexidartinib clinical-development timeline and does not generalise to other drugs. Future drug-specific Phase-1-vs-commercial bioavailability comparisons should register their own FORM_<drug>_PHASE1 canonical rather than reuse this entry (sibling to FORM_ABA_PHASE2 for abatacept). Distinct from the generic FORM_TABLET / FORM_CAPSULE / FORM_SUSPENSION family (oral solid-dosage manipulations) because the contrast here is two oral formulations differentiated by clinical-development phase rather than by dosage form. For typical simulations of the approved 800 mg/day TGCT regimen used in ENLIVEN, set FORM_PEX_PHASE1 = 0 for every subject.

FORM_GEF_F02 (canonical for the gefapixant F02 wet-granulation citric-acid tablet formulation indicator)

• Description: 1 = dose record administered as gefapixant F02, the earlier wet-granulation film-coated immediate-release tablet (7.5, 20, and 50 mg strengths) that contains citric acid as an acidulant; 0 = dose record administered as F04 (gefapixant citrate active ingredient with a 20A film coating, used in two phase I studies) or F04A (gefapixant citrate with a 03K film coating, used in the phase III COUGH-1 and COUGH-2 trials). Per-dose-occasion indicator: within a single study the formulation is fixed, but a subject in a relative-bioavailability crossover study can carry both values across periods.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (F04 / F04A gefapixant-citrate formulations).
• Source aliases:
  • FORM – Chawla 2023 codes the formulation as a categorical F02 / F04 / F04A column (Table S2); the canonical name matches the indicator direction (1 = F02).
• Example models: Chawla_2023_gefapixant.R (carries NO main effect on any PK parameter – no F02-vs-F04 relative-bioavailability difference was retained in the final model – and exists solely to gate the FED effect on the absorption rate constant to F02 records: ka = exp(lka + etalka) * (1 + e_fed_ka * FED * FORM_GEF_F02) with e_fed_ka = -0.594, a 59.4% reduction in Ka when an F02 dose is taken fed. Dedicated phase I relative-bioavailability studies had already established that fed status and concomitant proton-pump-inhibitor use affect gefapixant exposure for F02 but not for F04, so food and PPI effects were tested for F02 only.).
• Notes: Scoped specific because the F02-vs-F04 contrast is tied to the gefapixant clinical-development formulation timeline and does not generalise to other drugs. This is a gating indicator rather than an effect-carrying covariate – the distinguishing pattern in the FORM_* family, where siblings such as FORM_PEX_PHASE1 and FORM_SUSPENSION carry their own multiplicative effect on bioavailability or absorption. When a formulation indicator exists only to restrict another covariate’s effect to a formulation subset, encode it in covariateData (it IS referenced in model()) and state the gating role explicitly in covariateData[[FORM_GEF_F02]]$notes. The marketed F04B formulation is compositionally identical to F04A except that it omits citric acid, and was shown bioequivalent to F04A, so F04B records take FORM_GEF_F02 = 0. The excluded F01 formulation predates F02 and does not appear in the analysis data set; a future extraction that needs it should register a separate FORM_GEF_F01. For typical simulations of the approved 45 mg b.i.d. chronic-cough regimen, set FORM_GEF_F02 = 0 for every subject (COUGH-1 and COUGH-2 used F04A), which makes the food effect inoperative.

FORM_SUSPENSION (canonical for extemporaneously-prepared liquid suspension formulation indicator)

• Description: 1 = subject received the modelled drug as an extemporaneously-prepared liquid suspension (the solid drug substance – typically a tablet, but also possibly powder or another solid form – compounded into a liquid suspension at bedside or in the clinical-pharmacy immediately before administration, rather than dispensed as a formally manufactured suspension product); 0 = subject received the comparator solid oral formulation (typically the same tablet swallowed whole or a sugar-coated tablet). Per-dose-occasion (not per-subject) indicator because a single participant can receive both formulations across study occasions in a crossover bioequivalence design.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the per-paper solid oral comparator: tablets swallowed whole in Svensson 2018; sugar-coated tablet swallowed whole in Valle 2005). Document the per-model comparator in covariateData[[FORM_SUSPENSION]]$notes.
• Source aliases:
  • FORM – used in Svensson_2018_bedaquiline.R (paper’s narrative “whole vs suspended” with the suspended formulation as the 1 level in the analytical control stream).
  • Treatment-arm column “Treatment 2 (suspension after fasting)” vs “Treatment 1 (SCT after fasting)” – used in Valle_2005_exemestane.R (3x3 Latin-square crossover with the extemporaneous suspension as the 1 level).
• Example models: Svensson_2018_bedaquiline.R (multiplicative effect on the typical mean absorption time MAT: mat_typ = exp(lmat) * (1 + e_susp_mat * FORM_SUSPENSION) with e_susp_mat = +0.23 (suspended-tablet MAT is 23% longer than whole-tablet MAT; Svensson 2018 Table 2, 95% CI 2.1-48%, P = 0.03); relative bioavailability F is held identical between formulations because the paper found no statistically significant difference (95% nonparametric CI 94-108% within the 80-125% bioequivalence criteria)), Valle_2005_exemestane.R (multiplicative effects on absorption rate ka (suspension/SCT-fasting ratio 7.6/2.35 = 3.234x; suspension absorbs ~3.2x faster than the SCT swallowed whole) and on apparent bioavailability F (suspension/SCT-fasting ratio = 1.2x); intrinsic V/F is shared across formulations; the paper’s per-treatment V/F values 1360 vs 1120 L collapse to a constant intrinsic V at V/F_SCT_fasting / F_suspension = 1133 ~= 1120 L).
• Notes: Specific scope because the per-paper comparator solid-oral form is paper-defined (tablets swallowed whole in Svensson 2018 paediatric-tuberculosis; sugar-coated tablet swallowed whole in Valle 2005 healthy-postmenopausal-women bioequivalence). Both papers contrast the same drug substance prepared two different ways (extemporaneous liquid suspension vs swallowed-whole solid) rather than two different manufactured drug products – distinct from FORM_TABLET (Kyhl 2016 / Tikiso 2021 tablet vs liquid solution), FORM_CAPSULE, and FORM_POWDER. Future extemporaneous-suspension bioequivalence extractions should reuse this canonical and extend the example list, documenting the per-paper comparator solid-oral form.

FORM_UNDILUTED_SUSP (canonical for undiluted ready-to-use oral suspension preparation-state indicator)

• Description: 1 = subject received a manufactured ready-to-use oral suspension administered undiluted (squirted directly into the mouth without pre-administration mixing with another liquid); 0 = subject received any of the rapidly-absorbed comparator formulations (e.g. tablet, granules for oral suspension after reconstitution, or the same ready-to-use suspension after dilution with a defined volume of non-sparkling liquid). Per-dose-occasion indicator because a single pediatric participant can switch preparation state across study phases. Distinct from FORM_SUSPENSION (Svensson 2018 / Valle 2005 extemporaneously-compounded suspension contrasted with a whole-swallowed solid) and from FORM_GRANULE (Ayyoub 2016 reconstitution-from-granules contrasted with a tablet) because here all subjects receive the same manufactured suspension product and the contrast is whether the suspension was diluted or not before swallowing.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (rapidly-absorbed comparator pool: tablet, granules for oral suspension after reconstitution, or diluted ready-to-use suspension; these share the same ka in Willmann 2021).
• Source aliases:
  • FORM,DILU (composite: FORM = 2 for ready-to-use suspension AND DILU = 1 for undiluted) – used in Willmann_2021_rivaroxaban.R (supplement S2 NONMEM $PK block IF(FORM.EQ.2.AND.DILU.EQ.1) FD = 2 ; IF(FD.EQ.2) KA = THETA(6)).
• Example models: Willmann_2021_rivaroxaban.R (piecewise selection of the typical first-order absorption rate constant Ka: ka = exp(lka) * (1 - FORM_UNDILUTED_SUSP) + exp(lka_susp_undiluted) * FORM_UNDILUTED_SUSP; the undiluted suspension Ka = 0.226 1/h (95% CI 0.154-0.297) is ~3.5x slower than the comparator pool Ka = 0.799 1/h (95% CI 0.655-0.944); the relative bioavailability F is unaffected by preparation state because absorption rate does not enter the area under the curve at the dose levels studied per Willmann 2021 Discussion p. 1201 ‘A lower rate of absorption of the undiluted ready-to-use suspension was found in this model … but the extent of absorption is not affected by ka and, thus, AUC(0-24h),ss is independent of the formulation’).
• Notes: Specific scope because the contrast pairs a specific preparation step (mix the suspension with another liquid before swallowing vs swallow undiluted) against itself for one specific drug product. The dilution step was introduced in EINSTEIN-Jr for the ready-to-use suspension after the delayed oral absorption became obvious when administered undiluted (Willmann 2021 Discussion p. 1201). Sits in the FORM_* family alongside FORM_SUSPENSION (extemporaneous suspension), FORM_GRANULE (granules for reconstitution), and the tablet/capsule/solution siblings, capturing a third type of liquid-administration contrast (preparation state of the same manufactured product). Future ready-to-use-suspension dilution-preparation extractions should reuse this canonical with the per-paper comparator pool documented in covariateData[[FORM_UNDILUTED_SUSP]]$notes. Promote to general scope when a second paper ratifies the same encoding.

FORM_APAPIBU_SUSP (canonical for the acetaminophen + ibuprofen ready-to-use oral suspension indicator)

• Description: 1 = subject received the fixed-dose acetaminophen + ibuprofen combination as the manufactured ready-to-use oral suspension (Maxigesic Oral Suspension, 160 mg acetaminophen + 48 mg ibuprofen per 5 mL); 0 = any other formulation of the same combination (film-coated tablet or sachet). Per-dose-occasion indicator because the source studies are crossovers in which each participant receives several formulations.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the film-coated tablet when the sibling FORM_POWDER is also 0).
• Source aliases: none (the source reports a formulation column only).
• Example models: Morse_2022_acetaminophen.R and Morse_2022_ibuprofen.R (multiplicative factors on the absorption half-life and the absorption lag time, with separate fasted and fed factor sets that are all relative to the fasted tablet: fasted 0.394 / 0.743 for acetaminophen and 0.719 / 0.984 for ibuprofen, fed 2.52 / 2.93 and 2.45 / 2.52; Morse 2022 Tables 2 and 3. Bioavailability is shared across the oral formulations).
• Notes: Auto-approved member of the FORM_<drug>_<formulation> family, with the drug token APAPIBU naming the acetaminophen + ibuprofen combination product because both drugs’ models carry the same indicator. Not FORM_SUSPENSION, which is defined for an extemporaneously compounded suspension, nor FORM_UNDILUTED_SUSP / FORM_UNDIL_SUSP, which encode a dilution contrast for a single product. Promote to a general manufactured-oral-suspension canonical if a second drug needs the same contrast.

FORM_GRANULE (canonical for granule-for-oral-suspension pediatric formulation indicator)

• Description: 1 = subject received the modelled drug as a pediatric granule (sachet) formulation intended for reconstitution with water or food immediately before administration to children (i.e. a regulated “granules for oral suspension” dosage form); 0 = subject received the comparator solid oral formulation (typically a tablet, the adult / older-child reference; document the per-paper comparator in covariateData[[FORM_GRANULE]]$notes). Per-subject (regimen-fixed) categorical indicator typical of pediatric-development popPK pooled analyses.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (the per-paper comparator formulation; the tablet in Ayyoub 2016 Table 2, the intravenous route in Knebel 2011 where FORM_GRANULE is paired with FORM_TABLET). Document the comparator per-model in covariateData[[FORM_GRANULE]]$notes.
• Source aliases:
  • FORM / FORMULATION – used in Ayyoub_2016_pyronaridine.R (paper’s formulation indicator with the granule formulation as the 1 level; matches the canonical encoding directly without value transformation).
  • spheroid – Knebel 2011’s name for its delayed-release pediatric granule formulation throughout the equations and Table IV; same orientation, no transformation. Used in Knebel_2011_pantoprazole.R.
• Example models: Ayyoub_2016_pyronaridine.R (additive-multiplier effect on the typical absorption rate constant Ka: ka_typ = exp(lka) * (1 + e_form_granule_ka * FORM_GRANULE) with e_form_granule_ka = +1.63 (granule Ka is 2.63x the tablet Ka, i.e. 47.1 1/day vs 17.9 1/day; Ayyoub 2016 Table 2, %RSE 37.8); CL/F, V2/F, V3/F, and Q/F are held identical between formulations because backward elimination (P < 0.001) retained formulation only on Ka), Knebel_2011_pantoprazole.R (three-level {intravenous, tablet, granule} stratification encoded as FORM_TABLET plus FORM_GRANULE with both = 0 selecting the intravenous route; the granule branch overrides the tablet defaults for BOTH relative bioavailability (0.295 vs the fixed anchor of 1) and absorption rate constant (0.613 vs 1.32 1/h), carries the model’s only inter-occasion variability (56.7% CV on F1, first dose vs all others), and selects its own proportional residual SD).
• Notes: The “pediatric granule” dosage form is a regulated granules- or sachets-for-oral-suspension product reconstituted with water or food, distinct from the bedside-improvised FORM_SUSPENSION (Svensson 2018: tablets crushed-and-suspended) and from FORM_TABLET / FORM_SOLUTION / FORM_POWDER (different sibling formulation contrasts). Future pediatric-granule popPK extractions (a common pattern in antimalarial, antiretroviral, and antitubercular pediatric trials – e.g. Pyramax granules, Coartem dispersible, lopinavir-ritonavir pellets) should reuse this canonical and extend the example list. Scope promoted from specific to general on 2026-08-26 alongside the Knebel 2011 pantoprazole extraction, which ratifies the same encoding in a second therapeutic area; note that Knebel 2011’s comparator is the intravenous route rather than Ayyoub 2016’s tablet, so the reference category is per-model rather than fixed.

FORM_BXM_TAB10 (canonical for the baloxavir marboxil 10 mg tablet formulation indicator)

• Description: 1 = baloxavir marboxil administered as the 10 mg tablet; 0 = any of the other formulations pooled in the source analysis (20 mg tablet, 2% granules, and granules for oral suspension 2 mg/mL). The indicator carries a relative-bioavailability contrast between two strengths of the same solid oral product, not a change in dosage form or in absorption mechanism, which is why it is a product-specific canonical rather than a member of the FORM_TABLET / FORM_GRANULE dosage-form family.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the 20 mg tablet, 2% granules and 2 mg/mL granules for oral suspension, all of which the source analysis assigned F_rel = 1.00).
• Source aliases: none formally; Retout 2026 identifies the level in prose as “the 10 mg tablet formulation used in T0822”.
• Example models: Retout_2026_baloxavir.R (relative bioavailability on the depot compartment: f(depot) = exp(lfdepot) * (1 + (e_form_bxm_tab10_fdepot - 1) * FORM_BXM_TAB10) with lfdepot fixed at log(1) and e_form_bxm_tab10_fdepot fixed at 0.88; Retout 2026 Methods, PopPK analysis).
• Notes: Specific scope because the contrast is tied to one product strength in one legacy Japanese study (T0822). The F_rel value was set rather than estimated, from biopharmaceutical clinical studies cited as “Roche data on file”, so it is encoded with fixed. Retout 2026 Table S1 records that study T0821 also used a 10 mg tablet alongside the 20 mg tablet, but the paper’s Methods text applies the 0.88 relative bioavailability only to T0822; the model encodes the printed statement. Downstream users simulating the currently marketed formulations should leave this indicator at 0.

FORM_DELAMANID_DT (canonical for the delamanid pediatric dispersible-tablet formulation indicator)

• Description: 1 = delamanid given as the pediatric dispersible tablet (5 mg or 25 mg); 0 = delamanid given as the adult 50-mg film-coated tablet.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (film-coated tablet).
• Source aliases:
  • Formulation – Sasaki 2022 Table 1 (‘Film-coated tablet’ / ‘Dispersible tablet’) and Table 2 rows ‘Formulation (dispersible tablet) on MAT’ and ‘Formulation (dispersible tablet) on F1’.
• Example models: Sasaki_2022_delamanid.R (founding example; mtt <- ... * (1 + 0.495 * FORM_DELAMANID_DT) and f_form <- 1 + (-0.158) * FORM_DELAMANID_DT, both coefficients fixed from a model that included the adult bioequivalence trial 245).
• Notes: Follows the auto-approved FORM_<drug>_<formulation> family and mirrors FORM_DTG_DT. In the founding trials the formulation is confounded with age group (0-5 years dispersible, 6-17 years film-coated), which is why the effects were fixed from adult bioequivalence data rather than estimated.

FORM_PRETOMANID_DT (canonical for the pretomanid pediatric dispersible-tablet formulation indicator)

• Description: 1 = pretomanid given as the pediatric dispersible tablet (10 mg or 50 mg strengths, dispersed in water before administration); 0 = pretomanid given as the marketed adult 200 mg tablet.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 1 (dispersible tablet) is the typical-value reference for KA in the founding model; the 0 level carries the multiplicative marketed-formulation effect.
• Source aliases:
  • DTF – Zou 2022 ESM Section C $INPUT column (0/1), with MF = 1 - DTF formed in $PK.
• Example models: Zou_2022_pretomanid.R (founding example; ka <- exp(lka + iov_ka) * e_form_mf_ka^(1 - FORM_PRETOMANID_DT) with e_form_mf_ka = 1.65, so the marketed tablet absorbs 1.65-fold faster than the dispersible tablet. Formulation was tested on relative bioavailability (ratio 1.00, 90% CI 0.87-1.14) and not retained).
• Notes: Follows the auto-approved FORM_<drug>_<formulation> family and mirrors FORM_DELAMANID_DT and FORM_DTG_DT. The polarity matches the source column (1 = dispersible tablet) even though the paper’s reference level for KA is the dispersible tablet; the model converts to the marketed-formulation indicator internally. In the founding study the 200 mg dispersible-tablet dose was given as 4 x 50 mg tablets, all with a high-fat meal.

FORM_DTG_DT (canonical for the dolutegravir dispersible-tablet / granule formulation indicator)

• Description: 1 = dolutegravir given as the pediatric dispersible tablet or as granules for oral suspension; 0 = dolutegravir given as the film-coated tablet (FCT). The dispersible tablet and the granules share a single set of absorption and bioavailability estimates in the source model, so one indicator covers both.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (film-coated tablet; the typical-value Ka and relative-bioavailability reference in Chandasana 2023 Table 2).
• Source aliases:
  • FORM – used in Chandasana_2023_dolutegravir.R; the source NONMEM control stream (Chandasana 2023 ESM Supplementary Text 1) codes FORM = 1 for the film-coated tablet and FORM = 2 / FORM = 3 for the dispersible tablet and granules, and collapses them with IF(FORM.GE.2) FORMK=1.
  • formulation – used in Chandasana_2024b_dolutegravir.R; Chandasana 2024 Table 2 names the levels “FCT” and “DT and granules”.
• Example models: Chandasana_2023_dolutegravir.R (founding example, and the primary source of the fit; two simultaneous multiplicative effects, ka <- exp(lka + e_form_dt_ka * FORM_DTG_DT + ...) with exp(e_form_dt_ka) = 2.04 and exp(lka) = 0.854 1/h, and frel <- exp(e_food_fdepot * FED + e_form_dt_fdepot * FORM_DTG_DT) with exp(e_form_dt_fdepot) = 1.53), Chandasana_2024b_dolutegravir.R (two simultaneous effects: a blend selecting the typical absorption rate constant, lka <- lka_fct * (1 - FORM_DTG_DT) + lka_dt * FORM_DTG_DT with Ka = 0.854 1/h for the FCT and 2.04 1/h for the DT / granules, and a multiplicative relative-bioavailability effect exp(e_form_dt_fdepot * FORM_DTG_DT) with exp(e_form_dt_fdepot) = 1.53 for the fasted DT / granules relative to the fasted FCT; the sibling FED food effect is gated on (1 - FORM_DTG_DT) because Chandasana 2024 Table 2 reports a fed estimate for the FCT only).
• Both reported effects are RATIOS to the film-coated tablet, not absolute values. Chandasana 2023 Table 2 prints KA~DT and granules = 2.04, which is the multiplier THETA(17) of the control stream’s KAFORM = THETA(17)**FORMK, not the dispersible-tablet Ka. The dispersible-tablet Ka is the product 0.854 x 2.04 = 1.74 1/h, as the paper’s own covariate-relationship block and Table S1 both spell out. Encoding 2.04 as an absolute Ka overstates dispersible-tablet absorption by a factor of 2.4 and was a live defect in Chandasana_2024b_dolutegravir.R until the primary paper was extracted.
• The sibling FED effect is NOT gated on this indicator. The control stream forms F1 = 1 * THETA(4)**SFLAG * THETA(9)**FFLAG, so the 1.10 food ratio multiplies the 1.53 formulation ratio; Chandasana 2023 Table S1 tabulates the product as F, without regards to food, DT/Granules = 1.68. Chandasana 2023 Table 2 lists a fed estimate on the FCT row only because the FCT is the reference level, not because the effect is FCT-specific.
• Notes: A dispersible tablet is a solid dosage form dispersed in water immediately before administration, so it is pharmaceutically distinct from the bedside-improvised FORM_SUSPENSION (crushed tablets), from FORM_SOLUTION (a manufactured oral liquid), and from FORM_GRANULE (sachet granules alone) – but the source model pools the dispersible tablet with the granules against the film-coated tablet, which is why this is a single drug-scoped indicator rather than a combination of the general siblings. Dispersible tablets are ubiquitous in pediatric antiretroviral, antimalarial and antitubercular development, so a general FORM_DISPERSIBLE_TABLET canonical is the natural promotion target: promote when a second paper ratifies a dispersible-tablet-vs-solid contrast that is not pooled with granules.

FORM_ELA_F1 (canonical for the elafibranor clinical formulation 1 indicator)

• Description: 1 = the dose was given as elafibranor clinical formulation 1, 0 = any other formulation. Per-dose-record indicator (formulation can change within a subject across study periods).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0. All six FORM_ELA_* indicators equal to 0 selects formulation 5, the phase II/III formulation and the model reference; there is deliberately no FORM_ELA_F5 column (see Notes).
• Source aliases:
  • FORMN / FORM1..FORM6 – Ooi 2026 Supplementary Datafiles S1 and S2 $INPUT. The binarised FORM<k> columns of the control streams are NOT in the paper’s formulation numbering: stream FORM1..FORM6 map to Table S1 / Table S3 formulations 5, 6, 1, 2, 3, 4 respectively. The canonical columns use the PAPER’s numbering, which is what Table S1 and Table S3 print.
• Example models: Ooi_2026_elafibranor.R, Ooi_2026_elafibranor_gft1007.R (formulation-specific relative bioavailability, zero-order input duration D1, mean absorption time and absorption lag time, for both the parent and the metabolite).
• Notes: Member of the FORM_<drug>_<formulation> family. Used in the phase I studies GFT505-106-1, GFT505-106-2, GFT505-108-4 and GFT505-108-3 (Table S1).

FORM_ELA_F2 (canonical for the elafibranor clinical formulation 2 indicator)

• Description: 1 = the dose was given as elafibranor clinical formulation 2, 0 = any other formulation. Per-dose-record indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (formulation 5 when all six FORM_ELA_* indicators are 0).
• Source aliases: FORMN – see FORM_ELA_F1 for the stream-versus-paper numbering caveat.
• Example models: Ooi_2026_elafibranor.R, Ooi_2026_elafibranor_gft1007.R.
• Notes: Member of the FORM_<drug>_<formulation> family.

FORM_ELA_F3 (canonical for the elafibranor clinical formulation 3 indicator)

• Description: 1 = the dose was given as elafibranor clinical formulation 3, 0 = any other formulation. Per-dose-record indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (formulation 5 when all six FORM_ELA_* indicators are 0).
• Source aliases: FORMN – see FORM_ELA_F1 for the stream-versus-paper numbering caveat.
• Example models: Ooi_2026_elafibranor.R, Ooi_2026_elafibranor_gft1007.R.
• Notes: Member of the FORM_<drug>_<formulation> family. In the GFT1007 model the relative bioavailability of formulation 3 is also fixed to 1.00, so formulations 3 and 5 share a bioavailability but differ in D1, MAT and lag time (Table S3).

FORM_ELA_F4 (canonical for the elafibranor clinical formulation 4 indicator)

• Description: 1 = the dose was given as elafibranor clinical formulation 4, 0 = any other formulation. Per-dose-record indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (formulation 5 when all six FORM_ELA_* indicators are 0).
• Source aliases: FORMN – see FORM_ELA_F1 for the stream-versus-paper numbering caveat.
• Example models: Ooi_2026_elafibranor.R, Ooi_2026_elafibranor_gft1007.R.
• Notes: Member of the FORM_<drug>_<formulation> family. The phase IIb MASH formulation (study GFT505-212-7, Table S1).

FORM_ELA_F6 (canonical for the elafibranor clinical formulation 6 indicator)

• Description: 1 = the dose was given as elafibranor clinical formulation 6, 0 = any other formulation. Per-dose-record indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (formulation 5 when all six FORM_ELA_* indicators are 0).
• Source aliases: FORMN – see FORM_ELA_F1 for the stream-versus-paper numbering caveat.
• Example models: Ooi_2026_elafibranor.R, Ooi_2026_elafibranor_gft1007.R.
• Notes: Member of the FORM_<drug>_<formulation> family. Used in the formulation-comparison study GFT505-119-17 (Table S1).

FORM_POWDER (canonical for oral powder formulation indicator)

• Description: 1 = subject received the oral powder formulation of the modelled drug; 0 = subject received the comparator solid oral formulation (typically a tablet, but document the per-paper comparator in covariateData[[FORM_POWDER]]$notes).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (tablet; F = 1 fixed in Yukawa 1990 Model 2).
• Source aliases:
  • FORM_POWDER – used in Yukawa_1990_phenytoin.R (paper’s BA indicator inverted: source BA = 1 if tablet, 0 if powder; canonical FORM_POWDER = 1 - BA_indicator so 0 is the tablet reference).
• Example models: Yukawa_1990_phenytoin.R (Yukawa 1990 Model 2 dose-dependent powder bioavailability F_powder = 1 - exp(-9.92 / DOSE_PHT_MGKGD); tablet F fixed at 1), Retlich_2015_linagliptin.R (multiplicative shift on the linagliptin first-order absorption rate constant Ka: powder-in-bottle Ka = 0.933 1/h vs tablet formulation 2 reference Ka = 0.441 1/h; the tablet formulation 1 comparator is captured by the sibling canonical FORM_LINAG_TAB1), Hong_2011_atazanavir.R (multiplicative linear-deviation effect on relative bioavailability: fdepot *= (1 + e_form_powder_frel * FORM_POWDER) with e_form_powder_frel = -0.355, i.e. the pediatric atazanavir oral powder has 35.5% lower bioavailability than the capsule reference; Hong 2011 Table 4), Morse_2022_acetaminophen.R and Morse_2022_ibuprofen.R (the Maxigesic sachet – an acetaminophen + ibuprofen powder dissolved in 200 mL water – against the film-coated tablet, paired with the sibling FORM_APAPIBU_SUSP; both indicators 0 selects the tablet. The indicator selects multiplicative factors on the absorption half-life and the absorption lag time, with separate fasted and fed factor sets that are all relative to the fasted tablet: fasted 0.462 / 0.845 for acetaminophen and 0.235 / 0.539 for ibuprofen, fed 2.30 / 2.10 and 3.79 / 0.178; Morse 2022 Tables 2 and 3. Bioavailability is shared across the oral formulations).
• Notes: Specific scope because the “powder vs tablet” contrast is tied to a particular drug-product manufacturing comparison (Yukawa 1990 contrasts Aleviatin brand phenytoin powder with Aleviatin tablets, both from Dainippon Pharmaceutical Co.; Retlich 2015 contrasts an early-phase linagliptin powder-in-bottle formulation against the marketed linagliptin tablet). Mirrors the sibling FORM_TABLET (Kyhl 2016 / Tikiso 2021 tablet vs solution) and FORM_CAPSULE (Hennig 2006 / Hennig 2007 capsule vs solution) under the FORM_* family. Future powder-formulation models should reuse this canonical, extending the example list and documenting the per-paper comparator.

FORM_CPX351_FROZEN (canonical for the frozen vs lyophilized CPX-351 liposome formulation indicator)

• Description: 1 = the intravenous CPX-351 dose was prepared from the frozen liposomal suspension (the early-development presentation used throughout the phase 1 study 101), 0 = the dose was prepared from the lyophilized powder reconstituted before infusion (the commercial presentation used in the phase 2 study 206 and the phase 3 study 301; the typical-value reference). CPX-351 is a dual-drug liposome co-encapsulating cytarabine and daunorubicin at a fixed 5:1 molar ratio, so the covariate describes the drug product, not either analyte separately. Study-fixed per subject in the founding source: all 38 study-101 patients received the frozen formulation and all 157 study-206/301 patients the lyophilized, which makes formulation perfectly confounded with study and therefore inseparable from a study effect in that dataset. Set it per dose record if a future dataset changes formulation within a subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (lyophilized formulation; the commercial presentation and the typical-value reference for every parameter in Wang 2019 Table 2, per its footnote b).
• Source aliases:
  • Formulation – Wang 2019 Table 1 row ‘Formulation, n (%)’ with levels ‘Frozen’ and ‘Lyophilized’; the canonical orientation takes Lyophilized as the reference, matching the paper’s typical-value convention, so no value flip is needed relative to the tabulated multipliers.
• Example models: Wang_2019_daunorubicin_cpx351.R (log-scale multiplicative shifts on all four disposition parameters of total daunorubicin: Form_CL -0.275, Form_Vc -0.221, Form_Q -0.551, Form_Vp -0.686, i.e. multipliers 0.760, 0.802, 0.576 and 0.504 relative to the lyophilized reference; Wang 2019 Supplemental Table S3, with Table 2 printing the already-exponentiated form), Wang_2019_cytarabine_cpx351.R.
• Notes: Well-formed member of the auto-approved FORM_<drug>_<formulation> family, alongside FORM_PEGCET_LYOPHILIZED, FORM_ASV_LIQUID, FORM_VOSO_SOLN02, FORM_ABA_PHASE2, FORM_PEX_PHASE1, FORM_GEF_F02 and FORM_DTG_DT. Deliberately NOT folded into FORM_PEGCET_LYOPHILIZED even though both contrast a lyophilisate with a non-lyophilised presentation: there the lyophilisate is the indicator level and a ready-to-use solution the reference, here the lyophilisate is the reference and the frozen suspension the indicator, so reusing that name would invert the reference category. Also distinct from the generic FORM_POWDER family, which is scoped to oral powder-versus-tablet dissolution comparisons; the mechanism at stake here is liposome integrity through a freeze-thaw versus a freeze-drying process for an intravenous product. Because the effect is confounded with study in the founding source, treat a fitted FORM_CPX351_FROZEN coefficient as a combined formulation-plus-study term rather than as a pure formulation effect; Wang 2019’s Discussion separately reports that ‘study phase had no measurable effect on the pharmacokinetics of CPX-351’, but that assessment could not be made independently of formulation. Scope: specific because the frozen-versus-lyophilized contrast is defined by the CPX-351 development programme.

FORM_PEGCET_LYOPHILIZED (canonical for the pegcetacoplan lyophilized-powder formulation indicator)

• Description: 1 = the subcutaneous pegcetacoplan dose was given as the lyophilized powder reconstituted before administration; 0 = the dose was given as one of the ready-to-use solution formulations. Per-regimen categorical indicator (a subject may change formulation between study periods, so set it per dose record rather than per subject when the source data do).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (ready-to-use solution formulations; the source FORM column levels 1 = sorbitol vehicle, 2 = dextrose vehicle, 3 = mannitol vehicle, which the Crass 2024 control stream pools as the reference because only the powder level carries an effect).
• Source aliases:
  • FORM (level 4 = POWDER) – Crass 2024 ESM Table 1 PK control stream, which expands the four-level FORM column into indicators FORM1..FORM4 and retains only FORM4 in the final model.
  • LYO, FORM_LYO – plausible alternative NONMEM $INPUT forms.
• Example models: Crass_2024_pegcetacoplan.R (fractional linear effect on subcutaneous bioavailability, F1 = THETA(4)*(1+FORM4*THETA(8)) with THETA(8) = 0.220, i.e. the lyophilized powder gives about 22% higher relative bioavailability than the solution reference, raising F from 0.758 to 0.925; Crass 2024 ESM Table 3 theta 8, 95% CI 0.145-0.296), Crass_2024_pegcetacoplan_hemoglobin.R, Crass_2024_pegcetacoplan_ldh.R.
• Notes: Well-formed member of the auto-approved FORM_<drug>_<formulation> family, alongside FORM_ASV_LIQUID, FORM_VOSO_SOLN02, FORM_ABA_PHASE2, FORM_PEX_PHASE1, FORM_GEF_F02, and FORM_DTG_DT. Deliberately NOT folded into the generic FORM_POWDER canonical: that entry is scoped to oral powder-versus-tablet product comparisons (Yukawa 1990 phenytoin, Retlich 2015 linagliptin, Hong 2011 atazanavir), whereas this contrast is a reconstituted lyophilisate versus a ready-to-use solution for subcutaneous injection, where the mechanism at stake is injection-site depot behaviour rather than oral dissolution. A future generic FORM_LYOPHILIZED could be promoted if a second parenteral lyophilisate-versus-solution model appears; until then the drug-specific form keeps the paper-defined solution reference unambiguous. Scope: specific because the reference category pools three named vehicle formulations defined by the pegcetacoplan development programme.

FORM_THEO_APNECUT (canonical for Apnecut vs theophylline-alcohol in-house preparation oral-theophylline product indicator)

• Description: 1 = subject received the commercial Apnecut (APC) oral theophylline product (4 mg/mL aqueous internal-use solution; Kowa Co., Ltd., Japan; launched August 2006), 0 = subject received the theophylline-alcohol (TA) in-house oral preparation (5 mg/mL theophylline dissolved in ethanol then diluted with sterile purified water to give 10 percent final ethanol concentration; compounded at the National Center for Child Health and Development pharmacy). Per-subject (regimen-fixed) categorical indicator in the Suda 2008 retrospective neonatal-apnea cohort.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (TA in-house preparation; the typical-value CL/F reference in Suda 2008 final model).
• Source aliases:
  • FORM_APNECUT – prior canonical name (pre-2026-06-19 standardization audit).
  • AP – used in Suda_2008_theophylline.R (Suda 2008 NONMEM AP indicator; same orientation, no transformation; AP = 1 if Apnecut, 0 if theophylline alcohol).
• Example models: Suda_2008_theophylline.R (multiplicative effect on CL/F per Suda 2008 final model page 638: cl <- exp(lcl + etalcl) * (WT / 1)^1.08 * (1 + e_form_apnecut_cl * FORM_THEO_APNECUT) with e_form_apnecut_cl = -0.282; the APC formulation has approximately 0.71x the CL/F of the TA reference, equivalent to approximately 1.41x higher dose-normalised exposure consistent with the higher trough concentrations the authors observed clinically that motivated the analysis).
• Notes: Specific scope because the Apnecut-vs-TA contrast is a Japan-specific paediatric-theophylline drug-product comparison local to the Suda 2008 cohort: the TA in-house preparation is unique to the National Center for Child Health and Development pharmacy (compounded ad hoc from bulk theophylline + ethanol + sterile water), and Apnecut is a Kowa Co. commercial Japanese product not marketed elsewhere. Both are oral liquid formulations - distinct from FORM_POWDER/FORM_SYRUP/FORM_TABLET/FORM_CAPSULE because the contrast here is between two specific liquid drug products rather than between dosage-form categories. Suda 2008 attributes the CL/F difference primarily to absorption (per Discussion, page 641: HPLC content analysis confirmed both products were within label, so the formulation effect on apparent oral clearance is interpreted as a bioavailability difference rather than an actual clearance difference) but the parameter is encoded as an effect on CL/F to match the published equation. Mirrors the sibling drug-product-version FORM_* entries (FORM_SAR_DP2 sarilumab, FORM_ISA_P2F2 isatuximab, FORM_LINAG_TAB1 linagliptin, FORM_VISMO_PHASEI vismodegib, FORM_TAC_IR tacrolimus) under the FORM_* family. Set to 0 to simulate the TA reference; in the absence of TA-specific dosing in a downstream cohort, set FORM_THEO_APNECUT = 1 to represent Apnecut.

FORM_SYRUP (canonical for oral syrup / liquid-suspension formulation indicator)

• Description: 1 = subject received the modelled drug as an oral syrup or liquid-suspension formulation (paediatric or extemporaneously-compounded oral suspension); 0 = subject received the comparator solid oral formulation (typically a capsule, but document the per-paper comparator in covariateData[[FORM_SYRUP]]$notes). Per-dose-occasion indicator in principle; in paediatric cohorts often time-fixed per subject by clinical convention.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (per-paper non-syrup comparator: capsule in Nanga 2019 tacrolimus; tablet in Sarashina 2005 epinastine).
• Source aliases:
  • syrup formulation – used in Nanga_2019_tacrolimus_metaanalysis.R (Table 3 covariate-effect label; relative bioavailability of syrup vs capsule = 0.53).
  • FORM (dry syrup vs tablet) – used in Sarashina_2005_epinastine.R (Table 4 theta_9 multiplier on CL/F: dry-syrup / tablet ratio 1.06; paediatric arm received dry syrup only, adult arm received tablet or dry syrup).
  • formulation type (capsule or suspension) – used in Isla_2024_fosfomycinCalcium.R (Methods ‘Population PK modelling’ covariate-screening list; the manufactured Fosfocina 250 mg/5 mL oral suspension is the 1 level and the manufactured Fosfocina 500 mg hard capsule the 0 reference, both commercial products of the same manufacturer).
• Example models: Nanga_2019_tacrolimus_metaanalysis.R (multiplicative effect on bioavailability per Eq. 4: f(depot) <- 0.53^FORM_SYRUP, so capsule users have F = 1 and syrup users have F = 0.53; Nanga 2019 Table 3 ‘Bioavailability for syrup formulation’ = 0.53 with 95% bootstrap CI 0.31 - 0.75), Sarashina_2005_epinastine.R (multiplicative ratio 1.06 on CL/F for dry syrup vs tablet; paediatric atopic-dermatitis patients all received dry syrup, healthy adults received either tablet or dry syrup; tablet is the reference (FORM_SYRUP = 0) and dry syrup is FORM_SYRUP = 1), Isla_2024_fosfomycinCalcium.R (the founding example in which the column acts on ABSORPTION KINETICS rather than on bioavailability or clearance: two multiplicative ratios, ka = exp(lka + etalka + log(1.17) * FORM_SYRUP) and tlag = exp(ltlag + etaltlag + log(0.84) * FORM_SYRUP), so the suspension absorbs 1.17x faster and starts 0.84x sooner than the capsule; Isla 2024 Table 2 theta_KA2 = 1.17 and theta_TLAG2 = 0.84, whose products with the capsule reference values reproduce the paper’s own printed ‘KA (suspension) = 0.18’ and ‘TLAG (suspension) = 0.70’ sub-rows. No relative-bioavailability term was estimated between the two products because the trial had no intravenous arm and all disposition parameters are apparent, and the paper’s Discussion states explicitly that disposition is unaffected by formulation – so this is the one registered instance where the syrup/capsule contrast is purely an absorption-rate contrast).
• Notes: Specific scope because the comparator solid-oral reference is paper-defined (capsule in Nanga 2019 and Isla 2024, tablet in Sarashina 2005) and because the affected parameter differs by paper (bioavailability in Nanga 2019, apparent clearance in Sarashina 2005, absorption rate and lag time in Isla 2024). Distinct from FORM_SUSPENSION (Svensson 2018: tablets extemporaneously suspended in water at bedside immediately before swallowing – same tablet swallowed two different ways, where the manipulation affects MAT rather than F) because here the contrast is between two distinct drug products (commercial capsules vs paediatric syrup / suspension). Distinct from FORM_CAPSULE, FORM_TABLET, and FORM_POWDER, which compare those solid-oral forms against a liquid solution or against each other. Future paediatric-syrup / oral-suspension formulation comparisons should reuse this canonical; if a future model contrasts syrup against tablet (rather than capsule), extend the per-model notes rather than registering a sibling canonical.

FORM_CACO3 (canonical for calcium-carbonate-tablet vs calcium-containing mineral-water formulation indicator)

• Description: 1 = subject received calcium as a calcium carbonate (CaCO3) tablet; 0 = subject received calcium as calcium-containing thermal mineral water (Geumjin thermal spring water, the Ahn 2014 reference comparator). Per-subject (treatment-arm) categorical indicator in the Ahn 2014 parallel-arm calcium-absorption design.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (Geumjin thermal spring water; the typical-value bioavailability reference in Ahn 2014, F = 1 fixed for the reference arm).
• Source aliases: none known; Ahn 2014 narratives the treatment factor as “thermal spring water versus calcium carbonate tablet” without a single NMTRAN column name.
• Example models: Ahn_2014_parathyroidHormone.R (multiplicative effect on the depot bioavailability per Ahn 2014 Table 2 ‘Relative F1’ = 1.98: fdepot = exp(lfdepot + e_form_caco3_fdepot * FORM_CACO3) with lfdepot fixed at log(1) for the thermal-water reference and e_form_caco3_fdepot = log(1.98) for the CaCO3 arm; 95% CI on relative F1 1.06-2.90).
• Notes: Specific scope because the contrast pairs a specific calcium product (calcium carbonate tablet) against a specific natural mineral-water source (Geumjin thermal spring water) rather than a generic tablet-vs-solution contrast (FORM_TABLET) – the 0-level here is calcium-containing spring water naturally rich in dissolved calcium, not a manufactured aqueous solution of the modelled drug. Future calcium-bioavailability studies that compare CaCO3 against a different calcium source (citrate, gluconate, dietary-calcium baseline) should reuse this canonical with the reference clearly documented in per-model notes, or register a sibling canonical if the comparator is meaningfully different. Sits in the FORM_* family alongside FORM_FDC (Wilkins 2008 rifampicin co-formulation), FORM_SYRUP (Nanga 2019 tacrolimus paediatric syrup), and FORM_TABLET (Kyhl 2016 nalmefene / Tikiso 2021 abacavir tablet vs solution).

FORM_FDC (canonical for fixed-dose-combination tablet formulation indicator)

• Description: 1 = subject received a fixed-dose-combination (FDC) tablet co-formulating the modelled drug with one or more other drugs in a single tablet; 0 = subject received the modelled drug as a single-drug tablet (or, in the antitubercular context, as a separate-drug-combination “SDC” of single-drug tablets). Per-subject (regimen-fixed) categorical covariate flagging the formulation when a population analysis pools FDC and single-drug-tablet arms and tests formulation as a covariate on absorption / disposition parameters.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: Paper-defined (the typical-value reference is not fixed across the register because different popPK papers anchor their typical-value parameters to different formulation arms). Wilkins 2008 set FDC = 1 as the typical-value reference (SDC = 0 enters as a deviation on MTT and CL); Choi 2018 set the single-agent tablet (FORM_FDC = 0) as the typical-value reference (FDC = 1 enters as a deviation on Ka and F). The canonical column value semantics (1 = FDC, 0 = single-drug tablet) are preserved across both papers; the choice of which arm is the structural reference is documented per-model in covariateData[[FORM_FDC]]$notes. Reference values observed: 1 = FDC (Wilkins 2008 rifampicin + isoniazid + pyrazinamide +/- ethambutol antitubercular FDC); 0 = single-agent metformin tablet (Choi 2018 metformin vs metformin-containing antidiabetic FDC).
• Source aliases:
  • FDC – used in Wilkins_2008_rifampicin.R (DDMODEL00000280 NMTRAN $INPUT column; values 0 / 1 with the same orientation as the canonical, 1 = FDC).
  • formulation (lower-case as printed in the paper) – used in Choi_2018_metformin.R (Choi 2018 Methods ‘Covariate analysis’ equation: formulation = 0 single-agent reference, formulation = 1 FDC test).
• Example models:
  • Wilkins_2008_rifampicin.R (multiplicative (1 + e_fdc0_mtt * (1 - FORM_FDC)) shift on MTT and (1 + e_fdc0_cl * (1 - FORM_FDC)) shift on CL; SDC subjects (FDC = 0) had 104% longer MTT and 23.6% higher CL than the FDC = 1 reference per Wilkins 2008 final estimates. Antitubercular co-formulation: rifampicin + isoniazid + pyrazinamide +/- ethambutol).
  • Choi_2018_metformin.R (multiplicative effects on first-order absorption rate Ka and on relative bioavailability F: ka *= 0.83^FORM_FDC (Ka shrinks to 83.0% of the single-agent value for FDC) and f_rel *= 0.94^FORM_FDC (relative bioavailability shrinks to 94.0% of the single-agent value for FDC) per Choi 2018 Table 3 final estimates. Antidiabetic co-formulation: the FDC drug-product is metformin co-formulated with an unspecified second antidiabetic agent – typical Korean metformin FDCs co-formulate with sitagliptin / glimepiride / vildagliptin / dapagliflozin; the paper does not name the specific co-formulant. Reference category 0 = single-agent metformin tablet).
• Notes: General scope: the FDC vs single-drug-tablet contrast applies to any drug class where a popPK study compares the modelled drug as a single-drug tablet against an FDC tablet co-formulating it with one or more other drugs. Promoted from specific to general scope on 2026-05-30 when Choi 2018 metformin (antidiabetic FDC) ratified the canonical alongside the pre-existing Wilkins 2008 rifampicin (antitubercular FDC) example. The drug-class context (antitubercular / antidiabetic / antihypertensive / antiretroviral / etc.) and the specific co-formulant identity (isoniazid + pyrazinamide / sitagliptin / etc.) are documented per-model in covariateData[[FORM_FDC]]$notes. The mechanism is co-formulation-driven perturbation of absorption (excipient interactions, dissolution rate, gastric residence) rather than drug-product manufacturing of the single drug – distinct from FORM_TABLET (Kyhl 2016 nalmefene tablet vs solution) and the rest of the FORM_* (drug-product-version) family because the FDC-vs-single-drug-tablet contrast compares two tablet products that both contain the modelled drug, not a tablet vs a non-tablet. When the source paper’s typical-value parameters are anchored to the FDC arm, set the per-paper note accordingly; when anchored to the single-drug-tablet arm, set the per-paper note accordingly. Both orientations are valid and the canonical column semantics (1 = FDC, 0 = single-drug tablet) preserve a consistent input column across simulations.

RIA_ASSAY (canonical for radioimmunoassay vs LC-MS/MS bioanalytical method indicator)

• Description: 1 = radioimmunoassay; 0 = LC-MS/MS.
• Units: (binary)
• Type: binary
• Scope: specific
• Example models: Kyhl_2016_nalmefene.R.
• Notes: Switches the additive residual-error magnitude. Use this canonical only when the source paper specifically identifies the immunoassay as radioimmunoassay. For non-radioactive immunoassay methods (microparticle enzyme immunoassay MEIA, chemiluminescence microparticle immunoassay CMIA, EMIT, or generic “immunoassay”), use the sibling canonical IMMUNOASSAY below.

IMMUNOASSAY (canonical for non-radioactive immunoassay vs LC-MS/MS bioanalytical method indicator)

• Description: Binary indicator selecting between a non-radioactive immunoassay bioanalytical method (microparticle enzyme immunoassay MEIA, chemiluminescence microparticle immunoassay CMIA, EMIT, ELISA-based, or generic “immunoassay”) and an LC-MS/MS reference method. 1 = immunoassay; 0 = LC-MS/MS. Time-varying per sample (per-row) in pooled datasets that span the historical introduction of LC-MS/MS at the analytical lab. Used to switch the additive and/or proportional residual-error magnitudes between the two analytical methods, which typically have different precision, accuracy, and cross-reactivity profiles.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (LC-MS/MS – the modern reference bioanalytical method with lower LLOQ, higher specificity, and less cross-reactivity with parent drug metabolites; selected as the reference because most current popPK datasets are LC-MS/MS based and the indicator captures the residual-error shift for the legacy-immunoassay subset).
• Source aliases:
  • IMMUNOASSAY / IA / ASSAY – common NONMEM $INPUT forms.
• Example models: Andrews_2017_tacrolimus.R (per-sample binary indicator switching the additive + proportional residual error between immunoassay (9% of samples; pre-2013 microparticle / chemiluminescence immunoassay) and LC-MS/MS (91% of samples; lower LLOQ 1.0 ng/mL vs 1.5 ng/mL) – both magnitudes estimated jointly in the final model per Andrews 2017 Section 2.3 and Table 2).
• Notes: Distinct from the radioimmunoassay-specific RIA_ASSAY canonical (use that one when the paper identifies the immunoassay as radioimmunoassay; use this IMMUNOASSAY for MEIA / CMIA / EMIT / ELISA / generic immunoassay). The two siblings could in principle be unified into a single IMMUNOASSAY canonical with per-model notes documenting the immunoassay subtype, but the existing RIA_ASSAY registration is preserved for backwards compatibility with Kyhl_2016_nalmefene.R. Per-row time-varying when the dataset spans the historical introduction of LC-MS/MS at the analytical lab (e.g., Andrews 2017’s 2009-2016 sampling window crosses the lab’s switch from immunoassay to LC-MS/MS; the per-sample assay method is recorded). When the dataset is fully LC-MS/MS, set IMMUNOASSAY = 0 for every row and the immunoassay residual-error parameters become non-identifiable – consider dropping them from the model.

ASSAY_LCMSMS (canonical for LC-MS/MS vs other-chromatographic bioanalytical method indicator)

• Description: Binary indicator selecting between an LC-MS/MS (liquid chromatography with tandem mass spectrometry) bioanalytical method and a NON-mass-spectrometric chromatographic reference method, typically HPLC with UV or fluorescence detection. 1 = LC-MS/MS; 0 = the other chromatographic method. Per-sample (per-row) in pooled datasets that span several source studies analysed at different laboratories or in different eras. Used to switch the additive and/or proportional residual-error magnitudes, LC-MS/MS generally giving the lower LLOQ and the better precision of the two.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the non-LC-MS/MS chromatographic assay). Note the orientation is the MIRROR of the IMMUNOASSAY / RIA_ASSAY siblings, which take LC-MS/MS as their 0 reference and flag the legacy method with 1. The reference here is the older method in both families; what changes is which method the canonical is NAMED for. Record the direction explicitly in covariateData[[ASSAY_LCMSMS]]$notes so a data assembler cannot invert it.
• Source aliases:
  • ASSY / ASSY1 / ASSY0 – Xu 2017 supplementary Table S2, whose two additive residual variances are footnoted “Residual error parameter for the study that used LC/MS/MS assay” (ASSY1) and “Residual error parameter for studies that did not use the LC/MS/MS assay” (ASSY0). Same orientation as the canonical.
• Example models: Xu_2017_daptomycin.R (per-sample switch between two ADDITIVE residual SDs on plasma daptomycin, 1.44 mg/L for the LC-MS/MS study and 2.30 mg/L for the non-LC-MS/MS studies, derived as the square roots of the printed variances 2.07 and 5.31; a pooled database spanning the Chaves 2014 registrational studies plus two later CRRT studies), Kloosterboer_2020_pipamperone.R (per-sample switch adding an EXTRA additive residual error of 26.6 ug/L, in variance, to the common 0.21 ug/L additive error for the German HPLC-UV samples (ASSAY_LCMSMS = 0); the Dutch UHPLC-MS plasma and DBS samples take ASSAY_LCMSMS = 1; Kloosterboer 2020 Table 2).
• Notes: Use this canonical when BOTH arms of the contrast are chromatographic and the paper names LC-MS/MS as the distinguishing method. When the comparator is an immunoassay use IMMUNOASSAY (MEIA / CMIA / EMIT / ELISA / generic) or RIA_ASSAY (radioimmunoassay) instead – those two already encode LC-MS/MS as their reference level and adding this column alongside them would double-code the same contrast. When the comparator is a study identity rather than a named method, the STUDY_<id> family is the better fit; use this one only when the paper attributes the residual-error split to the ANALYTICAL METHOD. Where the non-LC-MS/MS method is named precisely enough to matter (HPLC-UV vs HPLC-fluorescence, and its validated range) record it per model. In a prospective all-LC-MS/MS dataset set ASSAY_LCMSMS = 1 for every row and the second residual parameter becomes non-identifiable.

ASSAY_OSA (canonical for one-stage clotting (OSA) vs chromogenic substrate assay (CSA) bioanalytical method indicator for factor VIII activity)

• Description: 1 = factor VIII activity measured by the one-stage activated partial thromboplastin time (APTT) clotting assay (OSA); 0 = measured by the chromogenic substrate assay (CSA). Per-observation (per-row) binary indicator. Used to switch (a) a multiplier capturing the systematic activity bias between the two assay platforms – applied either to bioavailability (Abrantes 2017) or directly to the predicted FVIII activity (Valke 2024) – and (b) the residual-error magnitudes, which differ between the two platforms.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (CSA). The column’s coding is fixed (1 = OSA, 0 = CSA), but which level a given model treats as its structural reference scale is a per-model property and must be recorded in that model’s covariateData notes. Abrantes_2017_moroctocog.R estimates its typical parameters on the CSA scale (F = 1 anchor at CSA, the assay used to calibrate the European-Union potency labels for Refacto AF) and applies its effects to the OSA rows; Valke_2024_factorviii.R and its PD siblings estimate CL / V1 on the OSA scale and therefore apply their effects to the CSA (ASSAY_OSA = 0) rows.
• Source aliases:
  • METH1 – used in Abrantes_2017_moroctocog.R (Abrantes 2017 Table 2 footnote g: OSA-central indicator on the bioavailability multiplier and on the proportional residual-error magnitude).
  • Assay method – used in Valke_2024_factorviii.R, Valke_2024_factorviii_thrombinPeak.R, Valke_2024_factorviii_thrombinPotential.R, Valke_2024_factorviii_plasminPeak.R (Valke 2024 Results 3.4: every sample was assayed by both methods and both were fitted simultaneously).
• Example models: Abrantes_2017_moroctocog.R (multiplicative effects: F multiplier (1 - 0.390 * ASSAY_OSA) so OSA-assayed observations have F reduced by 39.0% relative to the CSA reference; proportional residual SD multiplier (1 + 0.403 * ASSAY_OSA) so OSA-assayed observations have 40.3% larger residual error magnitude), Valke_2024_factorviii.R, Valke_2024_factorviii_thrombinPeak.R, Valke_2024_factorviii_thrombinPotential.R, Valke_2024_factorviii_plasminPeak.R (OSA-referenced: Cc <- fviii * 0.939^(1 - ASSAY_OSA) so CSA-assayed observations read 0.939x the one-stage value, with the proportional and additive residual SDs likewise switched by 0.840^(1 - ASSAY_OSA) and 5.012^(1 - ASSAY_OSA); the theta^flag form follows Valke 2024 Supplementary Methods Eq. 6).
• Notes: Specific scope because OSA-vs-CSA is a FVIII-domain bioanalytical contrast (the OSA is a functional clotting assay that uses patient APTT-based clot formation; the CSA is an amidolytic chromogenic substrate assay). The direction and magnitude of the OSA-CSA discrepancy are product-dependent: B-domain-deleted recombinant FVIII products read lower by OSA than CSA (Abrantes 2017 Discussion; Lippi 2008; Hubbard 2013), whereas for the plasma-derived VWF/FVIII concentrate of Valke 2024 the CSA read 6.1% lower than the OSA (correction factor 0.939). Do not carry a magnitude across models; source it per product. Per-observation (per-row) when a dataset mixes assays – Valke 2024 assayed every sample by both methods, so the same nominal sampling time appears twice with different ASSAY_OSA values. Paired with ASSAY_OSA_LOCAL to capture an additional inter-laboratory shift for OSA performed at non-central laboratories.

ASSAY_OSA_LOCAL (canonical for OSA performed at a local (vs central) laboratory indicator for factor VIII activity)

• Description: 1 = factor VIII activity measured by the OSA at a local (study-site or regional) laboratory rather than the centralised analytical laboratory; 0 = central-laboratory OSA, or any non-OSA-local observation. Per-observation (per-row) binary indicator. Sub-variant of ASSAY_OSA: when ASSAY_OSA_LOCAL = 1, ASSAY_OSA is also 1 (the local-lab samples are a subset of OSA-assayed samples). Used to capture an additional inter-laboratory variability term on the bioavailability multiplier.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (central laboratory OSA, or any non-OSA-local observation – the dominant subset in pooled FVIII PK datasets).
• Source aliases:
  • METH2 – used in Abrantes_2017_moroctocog.R (Abrantes 2017 Table 2 footnote g: OSA-local indicator on the bioavailability multiplier; only one study in the Abrantes 2017 13-study pool, B1831003, used a local laboratory).
• Example models: Abrantes_2017_moroctocog.R (multiplicative effect on F: (1 - 0.146 * ASSAY_OSA_LOCAL) so OSA-local-laboratory observations have an additional 14.6% F reduction beyond the central-laboratory OSA effect captured by ASSAY_OSA).
• Notes: Specific scope because OSA inter-laboratory variability is a FVIII-domain phenomenon. Paired with ASSAY_OSA. When the pooled dataset contains no local-laboratory observations, set ASSAY_OSA_LOCAL = 0 for every row and the corresponding F multiplier collapses to 1.

ASSAY_SEPIP (canonical for the separation-of-interfering-peak bioanalytical-method indicator)

• Description: 1 = the concentration was measured by a liquid-chromatography method in which an interfering chromatographic peak was separated from the analyte peak, WITHOUT addition of formic acid to the mobile phase; 0 = neither modification was applied. Per-observation indicator used to shift the predicted concentration and to select the residual-error magnitudes between bioanalytical method versions. Mutually exclusive with ASSAY_SEPIP_FA, which carries the peak separation AND the formic-acid modification; both 0 selects the unmodified reference method.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (the unmodified method: no peak separation and no formic acid).
• Source aliases:
  • BIOANN – the NONMEM $INPUT column in Ooi 2026 Supplementary Datafile S1, a four-level code read in $PK / $ERROR as 1 = neither modification, 2 = separation of the interfering peak only, 3 and 4 = separation plus formic acid. Derive ASSAY_SEPIP = as.integer(BIOANN == 2).
• Example models: Ooi_2026_elafibranor.R (founding example; multiplies the predicted elafibranor concentration by exp(-0.182) and selects four of the twelve residual-error magnitudes).
• Notes: Distinct from ASSAY_CMIA / ASSAY_OSA / ELISA, which contrast two named assay PLATFORMS. This pair names two chromatographic SAMPLE-PREPARATION / mobile-phase modifications of one LC-MS/MS method, which is why the indicator is named after the modification rather than after a platform. A paper that reports only one modification uses ASSAY_SEPIP alone.

ASSAY_SEPIP_FA (canonical for the separation-of-interfering-peak-plus-formic-acid bioanalytical-method indicator)

• Description: 1 = the concentration was measured by a liquid-chromatography method in which an interfering chromatographic peak was separated from the analyte peak AND formic acid was added; 0 otherwise. Per-observation indicator. Mutually exclusive with ASSAY_SEPIP; both 0 selects the unmodified reference method.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (the unmodified method: no peak separation and no formic acid).
• Source aliases:
  • BIOANN – Ooi 2026 Supplementary Datafile S1 $INPUT; levels 3 and 4 both select this stratum, so ASSAY_SEPIP_FA = as.integer(BIOANN %in% c(3, 4)).
• Example models: Ooi_2026_elafibranor.R (founding example; multiplies the predicted elafibranor concentration by exp(-0.760) and selects two of the twelve residual-error magnitudes).
• Notes: Sibling of ASSAY_SEPIP; see that entry’s Notes for why this pair is named after the sample-preparation modification rather than after an assay platform.

ASSAY_IFX (canonical for the commercial infliximab ELISA identity stratum)

• Description: Integer code naming which commercial infliximab immunoassay measured a serum infliximab concentration, used to select a per-assay residual-error magnitude. Canonical code mapping: 1 = Sanquin (Amsterdam), 2 = Immundiagnostik AG (Germany), 3 = Caltag Laboratories, 4 = Matrix Biotek, 5 = the in-house assay developed at Sheba Medical Center and referred to in the literature by its developer’s name, Shomron Ben-Horin, 6 = Promonitor (Grifols / Progenika). Per-observation (per-row) code. A dataset that uses a vendor not yet in the mapping extends it with the next unused integer and records the addition here; 7 is the next free code (the obvious first claimant being apDIA, which appears in Zhao 2026 Supplementary Table 1 but contributed no fitted observations).
• Units: (integer 1-6)
• Type: categorical
• Reference category: None – the stratification is symmetric, with no reference level: every observation belongs to exactly one assay and each assay selects its own estimated residual SD. Code 2 (Immundiagnostik) is the recommended default when a user does not know which kit produced a measurement: it is the assay Zhao 2026 pivots on in its Table 5 pairwise comparisons, it is one of the two commercial kits the Zhao 2026 Discussion identifies as mutually harmonised (“no significant differences were found between the commercially developed Immundiagnostik and Caltag assays, likely reflecting greater methodological similarity and harmonization in their commercial manufacturing processes”), and its estimate sits mid-range among the six.
• Source aliases:
  • IFX assay – used in Zhao_2026_infliximab.R (Zhao 2026 Table 1 tabulates the six assays by patient count and Table 2 gives each one a residual-error row).
• Example models: Zhao_2026_infliximab.R (six proportional residual SDs, selected inline as propSdSanquin * (ASSAY_IFX == 1) + ... + propSdPromonitor * (ASSAY_IFX == 6) and passed to prop()).
• Notes: An integer stratum column rather than a family of paired binary indicators, following the guidance already recorded in the ASSAY_CMIA entry (“A pooled dataset that mixes CMIA, EMIT and LC-MS/MS needs more than one binary indicator, or an integer stratum column on the STUDY_<id> pattern of STUDY_TACRO_FRANCKE”) and the STUDY_TACRO_FRANCKE precedent itself; a model file derives the (ASSAY_IFX == n) indicators inline. Auto-approved member of the ASSAY_<method> canonical family. General scope, and deliberately keyed on vendor identity rather than on a single paper’s coding, because inter-assay discordance in infliximab therapeutic drug monitoring is a standing problem in the IBD literature with these same kits recurring across studies – a later infliximab popPK that stratifies its residual error by kit reuses the same codes, and a study using only a subset simply never emits the unused ones. Distinct from ELISA and ASSAY_CMIA, which each contrast exactly two named immunoassay platforms and cannot express a six-level stratification. Distinct from IMMUNOASSAY and RIA_ASSAY, which contrast an immunoassay against an LC-MS/MS reference; there is no chromatographic arm here (Zhao 2026 Supplementary Table 1 does label the Utrecht centre “Sanquin (LC-MS/MS)”, but that centre’s observations fall in the Sanquin stratum and the paper models no separate LC-MS/MS magnitude). Distinct also from a STUDY_<id> cohort indicator: assay and contributing centre are close to confounded in Zhao 2026 (Supplementary Table 1 maps 14 centres onto 7 vendor labels, with several centres per vendor), but the estimand is the assay’s measurement precision, and a centre that switches kits mid-course – or a sample split across two kits – would break the confounding with no way to recover the distinction from a centre label. The assay must enter only the residual-error model, never a structural parameter and never a multiplicative bias on the predicted concentration. Zhao 2026 found no assay effect on exposure (Table 5: all 15 pairwise comparisons of AUC week 6-14 gave P > 0.05) and none on post-hoc clearance except a single underpowered Sanquin-versus-Immundiagnostik contrast (P = 0.033, from n = 8 Sanquin measurements, which the paper itself discounts as likely to “lack robustness and clinical significance”). A future paper that does estimate a concentration-scale bias between kits should apply it as a multiplier on Cc in the manner of ASSAY_OSA in Valke_2024_factorviii.R, and must source the magnitude per kit pair rather than carrying one across models.

TRACER_TC99M_DTPA (canonical for the 99mTc-DTPA-vs-51Cr-EDTA exogenous GFR-tracer identity indicator)

• Description: 1 = the record’s administered exogenous glomerular-filtration tracer is technetium-99m diethylenetriaminepentaacetate (99mTc-DTPA); 0 = chromium-51 ethylenediaminetetraacetate (51Cr-EDTA). Per-record (per-occasion) binary covariate identifying which tracer compound a plasma-concentration record and its PK parameters belong to, when a population analysis fits two exogenous GFR tracers simultaneously in order to estimate the systematic bias between their measured clearances.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (51Cr-EDTA, the historical GFR reference tracer).
• Source aliases:
  • MES (“GFR measurement”) – used in Gracia_2025_cr51edta_tc99mdtpa.R (Gracia 2025 Methods “Pharmacokinetic analysis” and Table 2 footnote: “MES = 0 or = 1, respectively, for 51Cr-EDTA or 99mTc-DTPA data”).
• Example models: Gracia_2025_cr51edta_tc99mdtpa.R (multiplicative (1 - theta * TRACER_TC99M_DTPA) on both the typical CL and the typical V, with theta6 = -0.009 on CL and theta8 = +0.017 on V, and switching the proportional residual SD between 4.99% CV for 51Cr-EDTA and 7.18% CV for 99mTc-DTPA).
• Notes: Distinct from CRCL, which carries the measured GFR value regardless of which tracer produced it (the CRCL entry already lists iohexol, inulin, 99mTc-DTPA and 51Cr-EDTA as interchangeable measurement methods for that value); TRACER_TC99M_DTPA carries the tracer identity for a record in a joint two-tracer fit. Distinct from the ASSAY_* / RIA_ASSAY / IMMUNOASSAY bioanalytical-method family, which distinguishes how a single fixed analyte was measured – here the two tracers are different administered molecules with their own disposition, and the covariate names which compound the record belongs to. Also distinct from a STUDY_* cohort indicator: in Gracia 2025 tracer and source study happen to be perfectly confounded (Cysped NCT02822404 supplied all 51Cr-EDTA data, CysPedVal RnIPH 2020-101 all 99mTc-DTPA data, and “No patient successively received both radioisotopic tracers”), but the estimand is the tracer bias, and a future study administering both tracers to the same patients would break the confounding with no way to recover the distinction from a study label. Founds a TRACER_<agent> family; a future iohexol-vs-inulin or 99mTc-DTPA-vs-iohexol comparison should register a sibling (e.g. TRACER_IOHEXOL) with the same shape.

FORM_LEB_NS0 (canonical for lebrikizumab NS0 cell-line formulation indicator)

• Description: 1 = NS0 cell-line formulation (lebrikizumab), 0 = other.
• Units: (binary)
• Type: binary
• Scope: specific
• Source aliases:
  • FORM_NS0 – prior canonical name (pre-2026-06-19 FORM__ standardization).
• Example models: Zhu_2017_lebrikizumab.R.

FORM_LEC_PROCESSB (canonical for lecanemab manufacturing Process B drug-product indicator)

• Description: 1 = drug product manufactured by Process B, 0 = drug product manufactured by Process A (the earlier process). A per-DOSE-RECORD indicator rather than a per-subject flag: the lecanemab Study 201 open-label extension switched material mid-study (United States, Canada and South Korea from May 2020; Japan from July 2020), so one subject can contribute records under both processes. Studies 101, 104 and 201 Core used Process A exclusively; Study 301 Core (Clarity AD) used Process B exclusively.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (Process A; relative bioavailability fixed at 1).
• Source aliases:
  • FORM – used in Majid_2024_lecanemab.R (Majid 2024 supplement Text S1 $INPUT; the footnote under the printed covariate equations defines it as “FORM, 0 (Process A) or 1 (Process B)”).
• Example models: Majid_2024_lecanemab.R (relative bioavailability ratio 0.904 on the intravenous dose, i.e. 9.6% lower exposure under Process B, applied as ratio^FORM_LEC_PROCESSB).
• Notes: The LEC drug abbreviation matches the existing CSS_LEC and AUC_LEC entries. This is a manufacturing-process comparability indicator in the sense of ICH Q5E, so it sits in the same family as FORM_LEB_NS0 / FORM_LEB_CHO_PHASE2 (lebrikizumab cell line) and FORM_SAR_DP2 (sarilumab drug-product version) rather than describing a dosage form. One structural caution for anyone re-using this column: in the founding model the between-subject variability on relative bioavailability applies to Process B records only – Majid 2024 supplement Text S1 $PK codes F1=1; IF (FORM.EQ.1) F1=THETA(5)*EXP(ETA(4)), so Process A bioavailability is exactly 1 with no variability, and putting the eta on a shared anchor instead would change the model. Set to 0 to simulate the earlier drug product and to 1 for the commercial material used in Clarity AD.

FORM_NALIRI_PREVSITE (canonical for the liposomal irinotecan previous-manufacturing-site drug-product indicator)

• Description: 1 = liposomal irinotecan (nal-IRI, irinotecan liposome injection) manufactured at the PREVIOUS manufacturing site; 0 = product from the current commercial site. Per-subject in the founding analysis: previous-site material was used in phase I studies conducted before 2012, current-site material in all pivotal studies (NAPOLI-1, NCT02551991).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (current commercial site; 81.4% of the Brendel 2021 cohort).
• Source aliases:
  • MFG – Brendel 2021 final NONMEM control stream. Opposite orientation: MFG = 1 is the current site (the control stream’s ‘most common’ category) and MFG = 0 the previous site, so FORM_NALIRI_PREVSITE = 1 - MFG.
  • Manufacturing site – Brendel 2021 Table 2 row label (‘previous site vs. current site’ in the Results).
• Example models: Brendel_2021_irinotecan_liposomal.R (previous-site factors 1.515 on total-irinotecan clearance, 0.872 on the irinotecan central volume and 1.376 on the delayed SN-38 formation-pathway ratio).
• Notes: A manufacturing-comparability indicator in the same family as FORM_LEC_PROCESSB and FORM_SAR_DP2. Distinct from STUDY_NAPOLI1, which carries the manufacturing site in Adiwijaya_2017_irinotecan_liposomal.R because there the site was perfectly collinear with the NAPOLI-1 study. In Brendel 2021 it is not: only 18.6% of patients, a subset of the 124 patients from the five non-pivotal studies, received previous-site material, so a separate column is needed. The NALIRI token follows the source’s ‘nal-IRI’ abbreviation.

FORM_LEB_CHO_PHASE2 (canonical for lebrikizumab CHO Phase 2 formulation indicator)

• Description: 1 = CHO Phase 2 formulation (lebrikizumab), 0 = other.
• Units: (binary)
• Type: binary
• Scope: specific
• Source aliases:
  • FORM_CHO_PHASE2 – prior canonical name (pre-2026-06-19 FORM__ standardization).
• Example models: Zhu_2017_lebrikizumab.R.

FORM_GCSF_PEG (canonical for pegylated vs non-pegylated recombinant human G-CSF formulation indicator)

• Description: 1 = pegfilgrastim (recombinant human G-CSF conjugated to a 20 kDa monomethoxy-polyethylene-glycol moiety, apparent MW approximately 39 kDa), 0 = filgrastim (non-pegylated recombinant human G-CSF, MW 18.8 kDa). Categorical drug-type covariate on the SC absorption / distribution / clearance / binding-affinity parameters (FSC, KSC, VD, CLD, KD). Subject-level (each subject received one drug per study cohort in the pooled analyses).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (filgrastim; the structurally-simpler non-pegylated parent molecule and the SC bioavailability anchor FSC = 1).
• Source aliases:
  • DRUG – generic paper-text label for the two-level filgrastim / pegfilgrastim covariate in Melhem_2018_g_csf.R (Melhem 2018 Methods ‘Covariate analysis’; the paper reports separate FIL and PEG parameter estimates in Table 2 rather than encoding a base + shift form directly).
• Example models: Melhem_2018_g_csf.R.
• Notes: Follows the FORM_<DRUG>_<FEATURE> naming pattern used by FORM_LEB_NS0, FORM_ITR_SUBA, FORM_DOX_DORYX_MPC, FORM_TAC_IR, FORM_LINAG_TAB1, FORM_SAR_DP2, FORM_ISA_P2F2, FORM_VISMO_PHASEI, FORM_XYNTHA, and FORM_AXI_XLI for drug-specific product / molecular-form indicators. Distinct from CONMED_IFNALPHA (pegylated interferon alpha coadministration flag) because FORM_GCSF_PEG is the primary G-CSF drug being modelled, not a coadministered agent. Distinct from CONMED_PLDH (PEGylated liposomal doxorubicin coadministration flag) for the same reason. Auto-approved sibling of the FORM_<DRUG>_<FEATURE> family.

FORM_EPO_ALFA (canonical for epoetin alfa vs epoetin delta recombinant-erythropoietin product indicator)

• Description: 1 = epoetin alfa (recombinant human erythropoietin expressed in Chinese hamster ovary (CHO) cells by insertion of the human EPO cDNA), 0 = epoetin delta (recombinant human erythropoietin produced in a human cell line by activation of the endogenous EPO gene, giving a human-cell glycosylation and carbohydrate profile). Categorical drug-product covariate for population PK models fit jointly to both products, whose PK differs in the rate and extent of subcutaneous absorption rather than in disposition. Subject-level and time-fixed (each subject receives one product for the whole study). Both products are quantified by the same erythropoietin ELISA, so the indicator distinguishes the administered product and not the measured analyte.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (epoetin delta; the product under study in the ratifying paper, whose Ka and F1 are the base estimates the epoetin alfa factors multiply).
• Source aliases:
  • TRT – treatment-arm column in Knebel_2008_epoetinDelta.R (Knebel 2008 Table III rows *theta6^TRT[epoetin alfa] on Ka and *theta7^TRT[epoetin alfa] on F1; the epoetin alfa level is the non-reference category, so the mapping onto FORM_EPO_ALFA needs no value transformation).
• Example models: Knebel_2008_epoetinDelta.R (exponentiated-switch factors on the subcutaneous absorption rate constant, ka = theta3 * 1.23^FORM_EPO_ALFA, and on subcutaneous bioavailability, F1 = theta4 * 0.544^FORM_EPO_ALFA, in a joint pediatric CKD fit of 47 epoetin delta and 13 epoetin alfa subjects).
• Notes: Follows the FORM_<DRUG>_<FEATURE> naming pattern; the closest precedent is FORM_GCSF_PEG, which likewise encodes two distinct marketed recombinant-protein products of the same parent cytokine as one binary covariate inside a single joint fit, acting on the SC absorption and bioavailability parameters. Distinct from ESAD (prior erythropoiesis-stimulating-agent dose at baseline, a continuous surrogate for residual hematopoietic stimulation in ESA-switch models) – FORM_EPO_ALFA names which ESA is being administered and modelled, not how much of a previous one the patient had. A future paper adding a third recombinant-erythropoietin product (epoetin beta, darbepoetin alfa, methoxy-polyethylene-glycol epoetin beta) to a joint fit should register its own sibling indicator (FORM_EPO_BETA,…) against the same epoetin delta or epoetin alfa reference rather than widening this column to a multi-level code. Auto-approved sibling of the FORM_<DRUG>_<FEATURE> family; ratified alongside the Knebel 2008 epoetin delta extraction.

CONMED_EOX (canonical for concomitant EOX (epirubicin + oxaliplatin + capecitabine) chemotherapy backbone indicator)

• Description: 1 = concomitant epirubicin + oxaliplatin + capecitabine (EOX) chemotherapy backbone, 0 = other backbone (e.g., mFOLFOX6, CAPOX, or single-agent).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-EOX backbone).
• Source aliases: COMB (used by Yamada 2025 Table 1 with the EOX level coded as the non-reference category; renamed to CONMED_EOX to preserve the semantic meaning of the 1-level).
• Example models: Yamada_2025_zolbetuximab.R (fractional effect on V1).
• Notes: Disease-backbone indicator. If a future model needs more backbone categories, encode each as its own indicator (COMB_CAPOX, COMB_FOLFOX,…) with a single reference group.

CONMED_CCB (canonical for concomitant calcium-channel blocker coadministration indicator)

• Description: 1 = subject was receiving a calcium-channel blocker (CCB) as a co-medication during the study, 0 = no CCB. Captures the documented CYP3A4-inhibition CCB-tacrolimus interaction (CCBs reduce tacrolimus apparent oral clearance because amlodipine, diltiazem, and verapamil inhibit CYP3A4 in the gut wall and liver where tacrolimus is metabolised).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no CCB).
• Source aliases: none.
• Example models: Passey_2011_tacrolimus.R (multiplicative effect on apparent oral CL: cl_typ * e_ccb_cl^CONMED_CCB with e_ccb_cl = 0.812, i.e. CCB coadministration reduces tacrolimus CL/F by ~19% in adult kidney transplant recipients; Passey 2011 final model Table 4 row “CCB”).
• Notes: Specific scope because the CCB-tacrolimus interaction is documented for CYP3A4 substrates and the magnitude is drug-pair specific. Future popPK models for CYP3A4 substrates that need a CCB conmed indicator should reuse this canonical.

CONMED_OMEPRAZOLE (canonical for concomitant omeprazole indicator)

• Description: Binary indicator of concomitant omeprazole; 1 = omeprazole co-administered, 0 = not. Drug-specific member of the CONMED_<INN> family, used when a paper estimates an omeprazole effect separately rather than pooling proton-pump inhibitors.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no omeprazole).
• Source aliases: omeprazole (Wang 2019 covariate name; same orientation).
• Example models: Wang_2019b_tacrolimus.R (fractional effect on CL/F, (1 - 0.362 * CONMED_OMEPRAZOLE); attributed to CYP3A inhibition), Gao_2021_methotrexate.R, Zhang_2026_linezolid.R.
• Notes: Distinct from the class-level CONMED_PPI; use CONMED_PPI when the paper pools PPIs. Record whether the indicator is time-varying in the per-model notes. Auto-approved as a well-formed CONMED_<INN> member during the Wang 2019 tacrolimus extraction (PMC6861867).

CONMED_LORATADINE (canonical for concomitant loratadine indicator)

• Description: Binary indicator of concomitant loratadine; 1 = loratadine co-administered, 0 = not. Member of the CONMED_<INN> family.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no loratadine).
• Source aliases: loratadine (Wang 2019 covariate name; same orientation).
• Example models: Wang_2019b_tacrolimus.R (fractional effect on CL/F, (1 - 0.322 * CONMED_LORATADINE); attributed to competition for CYP3A).
• Notes: Auto-approved as a well-formed CONMED_<INN> member during the Wang 2019 tacrolimus extraction (PMC6861867).

CONMED_DILTIAZEM (canonical for concomitant diltiazem indicator)

• Description: Binary indicator of concomitant diltiazem; 1 = diltiazem co-administered, 0 = not. Drug-specific member of the CONMED_<INN> family.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no diltiazem).
• Source aliases: diltiazem (Wang 2019 covariate name; same orientation).
• Example models: Wang_2019b_tacrolimus.R (fractional effect on CL/F, (1 - 0.307 * CONMED_DILTIAZEM); attributed to CYP3A inhibition).
• Notes: Distinct from the class-level CONMED_CCB, which pools calcium-channel blockers of differing CYP3A-inhibition potency (diltiazem, amlodipine, verapamil). Use CONMED_DILTIAZEM when the paper estimates the diltiazem effect separately. Auto-approved as a well-formed CONMED_<INN> member during the Wang 2019 tacrolimus extraction (PMC6861867).

CONMED_ABATACEPT_DOSE, CONMED_ADALIMUMAB_DOSE, CONMED_ANAKINRA_DOSE, CONMED_CERTOLIZUMAB_DOSE, CONMED_ETANERCEPT_DOSE, CONMED_GOLIMUMAB_DOSE, CONMED_INFLIXIMAB_DOSE, CONMED_RITUXIMAB_DOSE, CONMED_TOCILIZUMAB_DOSE, CONMED_USTEKINUMAB_DOSE, CONMED_MTX_DOSE (canonical for the dose of a named biologic DMARD or immunomodulatory monoclonal antibody, or of randomized methotrexate, in an immune-mediated-inflammatory-disease dose-response meta-analysis)

• Description: Dose of the named agent assigned to the study arm (per-arm MBMA) or to the individual subject (subject-level trial simulation), expressed on the dose metric the source paper’s ED50 is written against. 0 = the agent was not given as a randomized treatment, so it contributes nothing to the dose-response; every column of the family 0 = a placebo (or background-therapy-only) arm or subject. Suffix is the INN, except MTX for methotrexate, matching the abbreviation already used by CONMED_MTX and DOSE_MTX_MGM2. Members of the CONMED_<drug>_DOSE family; each column drives its own drug’s dose-response term in a model-based meta-analysis, exactly as the statin and anticoagulant members of that family do. The family spans indications: rheumatoid arthritis (Mandema_2011_biologicDMARDs_mbma.R), plaque psoriasis (Dodds_2013_psoriasis_biologics_mbma.R) and Crohn’s disease (the six Yu_2022_crohns_*_mbma.R models).
• Units: Per drug AND per model, and NOT interchangeable. In the rheumatoid-arthritis maintenance-regimen setting: mg per administration for adalimumab (q2w), certolizumab pegol (q2w), etanercept (twice weekly), golimumab (q4w) and rituximab (weeks 0 and 2); mg/day for anakinra; mg/kg per administration for abatacept (q4w), infliximab (weeks 0, 2, 6 then q8w) and tocilizumab (q4w); mg/week for methotrexate. In the psoriasis single-dose setting: flat mg for a SINGLE administration of adalimumab, golimumab or ustekinumab. The unit of each column must match the unit of the matching ED50 parameter in the model. Record the per-column unit in covariateData[[<name>]]$units.
• Type: continuous
• Scope: specific
• Reference category: 0 (this agent was not given as a randomized treatment in this arm, so it contributes nothing to the dose-response). All columns 0 = placebo / background-therapy-only arm.
• Source aliases: “Dose” / “Dose range” in per-drug meta-analysis summary tables; the x-axis label of a per-drug dose-response panel.
• Example models: Mandema_2011_biologicDMARDs_mbma.R (each biologic column drives its drug’s Emax term through the drug’s own ED50 and its mechanism-of-action class Emax; per-drug dose ranges are in Mandema 2011 Table 1), Dodds_2013_psoriasis_biologics_mbma.R (CONMED_ADALIMUMAB_DOSE, CONMED_GOLIMUMAB_DOSE and CONMED_USTEKINUMAB_DOSE carry a SINGLE-dose milligram amount read against per-drug ED50s of 16.9, 45.5 and 13.9 mg, driving one shared Emax on the log PASI ratio; the simulated dose levels are 0, 21, 70, 210 and 700 mg), Checchio_2017_psoriasis_pasi75_longitudinal_mbma.R, Checchio_2017_psoriasis_pasi_landmark_mbma.R, He_2021_psoriasis_pasi75_mbma.R, He_2021_psoriasis_pasi90_mbma.R (adalimumab, certolizumab, etanercept, infliximab, ustekinumab and methotrexate columns as per-administration maintenance doses; CONMED_USTEKINUMAB_DOSE is a step in the PASI90 model only), Yu_2022_crohns_cdai150_mbma.R, Yu_2022_crohns_cdai100_mbma.R, Yu_2022_crohns_cdai70_mbma.R, Yu_2022_crohns_dcdai_mbma.R, Yu_2022_crohns_dcrp_mbma.R, Yu_2022_crohns_dibdq_mbma.R (Crohn’s disease: abatacept, adalimumab, certolizumab pegol, etanercept, infliximab and ustekinumab columns carry the DOSE values of the paper’s deposited analysis dataset – weight-based regimens at 70 kg in mg, adalimumab as the week-0 induction dose; each column is a STEP except where the endpoint’s model estimates a dose-response: adalimumab graded in the CDAI150, CDAI-100, Delta CDAI and Delta IBDQ models, certolizumab pegol graded in the Delta CRP model).
• Notes: The same column name carries DIFFERENT dose metrics in different models, which is why the units field is per-paper and must always be read from covariateData[[<name>]]$units. CONMED_ADALIMUMAB_DOSE is 40 mg per q2w administration in Mandema_2011_biologicDMARDs_mbma.R (ED50 17.8 mg q2w) but a single-dose milligram amount in Dodds_2013_psoriasis_biologics_mbma.R (ED50 16.9 mg); the same holds for CONMED_GOLIMUMAB_DOSE (mg q4w, ED50 45.7, versus single-dose mg, ED50 45.5). A value must never be carried across the two models. CONMED_MTX_DOSE is the per-arm dose of methotrexate given as a RANDOMIZED treatment and is distinct from the binary CONMED_MTX, which records background concomitant methotrexate. The distinction is load-bearing in Mandema_2011_biologicDMARDs_mbma.R: background methotrexate is absorbed into the trial-specific placebo response and carries no parameter, while randomized methotrexate carries an estimated mean effect. It is also distinct from DOSE_MTX_MGM2, a body-surface-area-normalized oncology methotrexate dose. Where a source estimates only a mean effect for an agent rather than a dose-response (as Mandema 2011 does for methotrexate, which was titrated to a similar 15-20 mg/week range in every trial), the column still carries the dose so the arm is self-describing, and the model keys the effect on the column being non-zero; record that in the per-model notes so a downstream user does not read dose-proportionality into it. Because a biologic dose column and a weight-normalized dose column can carry the same INN in different papers, always check the units field of the per-model covariateData entry rather than assuming milligrams.

CONMED_ACITRETIN_DOSE, CONMED_ALEFACEPT_DOSE, CONMED_APREMILAST_DOSE, CONMED_BARICITINIB_DOSE, CONMED_BRIAKINUMAB_DOSE, CONMED_BRODALUMAB_DOSE, CONMED_GUSELKUMAB_DOSE, CONMED_IXEKIZUMAB_DOSE, CONMED_RISANKIZUMAB_DOSE, CONMED_SECUKINUMAB_DOSE, CONMED_TILDRAKIZUMAB_DOSE, CONMED_TOFACITINIB_DOSE (canonical for the dose of a named systemic antipsoriatic agent in a psoriasis dose-response meta-analysis)

• Description: One column per drug carrying the dose of that agent in a study arm, and 0 for every arm not randomised to it. Arm-level, not subject-level. A model-based meta-analysis of a therapeutic area has no single DOSE column, because each arm’s dose is only interpretable against that drug’s own ED50; giving each agent its own column lets a zero collapse that drug’s Emax term to exactly zero while the arm’s actual dose drives the one agent it received. Well-formed siblings of the biologic-DMARD block CONMED_ABATACEPT_DOSE, CONMED_ADALIMUMAB_DOSE, ..., registered separately because these agents are psoriasis-specific and do not appear in the rheumatoid-arthritis analyses that founded that block. The three IL-23 inhibitors (guselkumab, tildrakizumab, risankizumab) were added with He 2021.
• Units: mg (per-paper; read covariateData[[<name>]]$units, which is authoritative – the family deliberately does NOT fix a dose metric, and the sibling CONMED_INFLIXIMAB_DOSE is mg/kg in the same models)
• Type: continuous
• Reference category: n/a – continuous. 0 means the arm did not receive that agent, which is the value every column takes on a placebo arm.
• Source aliases:
  • Dose / DOSE_d – Checchio 2017 Equations 4 and 7; the per-arm value is the Clinical dose column of that paper’s Tables 1 and 2.
  • dose – He 2021 Equation 5; the per-arm regimen is the Route (regimen) column of that paper’s Table 1.
• Example models: Checchio_2017_psoriasis_pasi75_longitudinal_mbma.R (eight of the original nine: acitretin and baricitinib are absent because Checchio 2017 Results ‘Available data’ excluded them, with ciclosporin, from the longitudinal analysis for insufficient longitudinal data), Checchio_2017_psoriasis_pasi_landmark_mbma.R (the original nine), He_2021_psoriasis_pasi75_mbma.R (every member except acitretin), He_2021_psoriasis_pasi90_mbma.R (every member except acitretin and alefacept; He 2021 dropped alefacept from the PASI90 model for insufficient data), Yu_2022_crohns_cdai150_mbma.R, Yu_2022_crohns_cdai100_mbma.R, Yu_2022_crohns_cdai70_mbma.R, Yu_2022_crohns_dcdai_mbma.R, Yu_2022_crohns_dcrp_mbma.R, Yu_2022_crohns_dibdq_mbma.R (Crohn’s disease; CONMED_RISANKIZUMAB_DOSE and CONMED_TOFACITINIB_DOSE only. The two columns are reused outside psoriasis because the agents are the same; the dose metric is that of Yu 2022’s deposited dataset – risankizumab mg per q4w infusion, tofacitinib mg per administration of a twice-daily regimen – and the step/graded status is per model: risankizumab is graded in the CDAI150 and Delta IBDQ models and a step elsewhere, tofacitinib is a step in every model).
• Notes: Dose metric. In both founding models the value is the dose PER ADMINISTRATION, not a daily or weekly total. This is settled arithmetically rather than by assumption: the two tofacitinib arms (5 and 10 mg b.i.d.) and the two etanercept arms (25 and 50 mg twice weekly) each share a single ED50 and differ only in this column, and only the per-administration reading reproduces the published responder rates for both members of each pair. Apremilast is the clearest single case – 30 mg reproduces its Table 1 and Table 2 values and 60 mg does not. He 2021 uses the same per-administration maintenance-dose metric, settled the same way: only the per-administration reading reproduces its Tables 3 and 5 (apremilast 30 mg, tofacitinib 5 and 10 mg, ixekizumab 160 vs 80 mg). Baricitinib is once daily, so per administration equals per day. In He 2021 a column is a STEP wherever that paper fixed the drug’s ED50 to 0 – CONMED_RISANKIZUMAB_DOSE and CONMED_ALEFACEPT_DOSE in the PASI75 model (with CONMED_MTX_DOSE of the sibling block), and CONMED_USTEKINUMAB_DOSE (sibling block) in the PASI90 model, where risankizumab is graded – so the step/graded status is per MODEL, not per drug. Three of these columns are STEPS, not graded doses (in the Checchio 2017 models). CONMED_ACITRETIN_DOSE and CONMED_ALEFACEPT_DOSE (longitudinal model only) trigger a fixed offset at any positive value, because the source could not estimate an ED50 for them – for the traditional oral agents because per-patient titration on safety and efficacy grounds left insufficient literature dose-response information, and for alefacept because the dose-response data were too limited. Only the comparison > 0 is read, so the magnitude records the arm’s clinical dose for provenance and must NOT be read as a prediction for any other dose of those agents. CONMED_ALEFACEPT_DOSE is graded in the landmark model, which did estimate two alefacept ED50 values (one per route; see ROUTE_IV). Two columns pair with an indicator. CONMED_ALEFACEPT_DOSE pairs with ROUTE_IV and CONMED_BRODALUMAB_DOSE with REGI_Q4W, each of which shifts only that drug’s log ED50 and must be left at 0 for every other agent. In particular, ixekizumab’s clinical regimen is also every 4 weeks but ixekizumab has a single ED50 with no interval term, so setting REGI_Q4W = 1 on an ixekizumab arm would be wrong. At most one column positive per arm. Setting two positive simultaneously is outside the founding source’s design and makes the model ADD the two drug effects, which it was never fitted to do. Briakinumab (withdrawn from development) and alefacept (withdrawn from market) are retained because the source models and predicts them; a column for a withdrawn agent is a historical record, not a recommendation.

CONMED_ABRILUMAB_DOSE, CONMED_ANDECALIXIMAB_DOSE, CONMED_APILIMOD_DOSE, CONMED_BRAZIKUMAB_DOSE, CONMED_CDP571_DOSE, CONMED_FILGOTINIB_DOSE, CONMED_FONTOLIZUMAB_DOSE, CONMED_LAQUINIMOD_DOSE, CONMED_NATALIZUMAB_DOSE, CONMED_ONERCEPT_DOSE, CONMED_ONTAMALIMAB_DOSE, CONMED_PF04236921_DOSE, CONMED_SEMAPIMOD_DOSE, CONMED_UPADACITINIB_DOSE, CONMED_VEDOLIZUMAB_DOSE, CONMED_VERCIRNON_DOSE (canonical for the dose of a named biologic or small targeted molecule in a Crohn’s disease dose-response meta-analysis)

• Description: One column per drug carrying the dose of that agent in a study arm, and 0 for every arm not randomised to it. Arm-level, not subject-level. Well-formed siblings of the biologic-DMARD block (CONMED_ABATACEPT_DOSE, CONMED_ADALIMUMAB_DOSE, ...) and of the antipsoriatic block (CONMED_ACITRETIN_DOSE, ...), registered separately because these agents first appear in the Crohn’s disease analysis of Yu 2022. Drug classes (Yu 2022 Table 1): TNF-alpha inhibitors CDP571, onercept and semapimod; integrin inhibitors natalizumab (alpha4), vedolizumab and abrilumab (alpha4beta7); IL-12/23 inhibitor apilimod; IL-23 inhibitor brazikumab; IL-6 inhibitor PF-04236921; MMP-9 inhibitor andecaliximab; JAK inhibitors filgotinib and upadacitinib; T-cell activation inhibitor laquinimod; MAdCAM inhibitor ontamalimab (PF-00547659); IFN-gamma inhibitor fontolizumab; CCR9 antagonist vercirnon. CDP571 and PF04236921 carry the development code because neither agent has an INN; the hyphen of PF-04236921 is dropped so the column is a syntactic name.
• Units: mg (per-paper; read covariateData[[<name>]]$units and notes, which are authoritative – the family deliberately does NOT fix a dose metric)
• Type: continuous
• Scope: specific
• Reference category: n/a – continuous. 0 means the arm did not receive that agent, which is the value every column takes on a placebo arm.
• Source aliases:
  • DOSE with the drug’s FLAG value – Yu 2022 deposited dataset (Supplementary DataSheet 1), one sheet per endpoint; the model code selects each drug’s term with (FLAG == k).
• Example models: Yu_2022_crohns_cdai150_mbma.R, Yu_2022_crohns_cdai100_mbma.R, Yu_2022_crohns_cdai70_mbma.R, Yu_2022_crohns_dcdai_mbma.R, Yu_2022_crohns_dcrp_mbma.R, Yu_2022_crohns_dibdq_mbma.R (Yu 2022 fits six independent endpoint models; not every agent enters every model, and the covariateData of each file lists the columns it reads).
• Notes: Dose metric. The value is the dataset’s DOSE column, which is NOT uniformly the daily dose that Yu 2022 Methods ‘Data Development’ describes: weight-based regimens are mg at 70 kg (natalizumab 3 and 6 mg/kg = 210 and 420 mg; vedolizumab 0.5 and 2 mg/kg = 35 and 140 mg; fontolizumab 1, 4 and 10 mg/kg = 70, 280 and 700 mg; CDP571 10 mg/kg = 700 mg), upadacitinib is mg per administration of a twice-daily regimen with the 24 mg once-daily arm recorded as 12, PF-04236921 is mg per subcutaneous administration, and only vercirnon (250 mg bid = 500) and andecaliximab (150 mg q2w = 75) are daily or weekly totals. Only upadacitinib (graded in the CDAI-100, CDAI-70 and Delta CRP models) and PF-04236921 (graded in the CDAI150 and Delta CRP models) read the magnitude; every other column is a STEP in which only > 0 is read and the magnitude records the arm’s regimen for provenance. The step/graded status is per MODEL, not per drug. At most one column positive per arm; setting two makes the model add the two drug effects, which it was never fitted to do. ### FORM_FLV_BID_XR (canonical for fluvastatin twice-daily / extended-release formulation indicator)
• Description: 1 = fluvastatin arm used twice-daily (BID) dosing or an extended-release (XR) formulation; 0 = once-daily immediate-release fluvastatin (the model reference).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (q.d. IR fluvastatin).
• Source aliases: BID_XR_FLV legacy form (used in Vargo 2014 statins / ezetimibe MBMA pre-rename); alias of the canonical FORM_FLV_BID_XR form rename.
• Example models: Vargo_2014_statins_ezetimibe_mbma.R (multiplies the fluvastatin ED50 by 0.645 when set to 1; Vargo 2014 Table 3 row “ED50,fluvastatin (b.i.d.|XR) / ED50,fluvastatin”; same ratio used for either regimen because the paper found b.i.d. and XR ED50 estimates were similar).
• Notes: Specific scope because the b.i.d.-vs-q.d. / IR-vs-XR formulation comparison is paper-specific to fluvastatin meta-analyses. Sibling of FORM_LOV_BID_XR (lovastatin) under the FORM_<drug>_BID_XR family pattern.

FORM_LOV_BID_XR (canonical for lovastatin twice-daily / extended-release formulation indicator)

• Description: 1 = lovastatin arm used twice-daily (BID) dosing or an extended-release (XR) formulation; 0 = once-daily immediate-release lovastatin (the model reference).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (q.d. IR lovastatin).
• Source aliases: BID_XR_LOV legacy form (used in Vargo 2014 statins / ezetimibe MBMA pre-rename); alias of the canonical FORM_LOV_BID_XR form rename.
• Example models: Vargo_2014_statins_ezetimibe_mbma.R (multiplies the lovastatin ED50 by 0.59 when set to 1; Vargo 2014 Table 3 row “ED50,lovastatin (b.i.d.|XR) / ED50,lovastatin”).
• Notes: Specific scope because the b.i.d.-vs-q.d. / IR-vs-XR formulation comparison is paper-specific to lovastatin meta-analyses. Sibling of FORM_FLV_BID_XR (fluvastatin) under the FORM_<drug>_BID_XR family pattern.

DIS_ACS (canonical for acute coronary syndrome cohort indicator)

• Description: 1 = study arm enrolled patients with acute coronary syndrome (ACS), 0 = non-ACS cohort (the model reference).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (non-ACS arm).
• Source aliases: ACS legacy form (used in Vargo 2014 statins / ezetimibe MBMA pre-rename); alias of the canonical DIS_ACS form rename.
• Example models: Vargo_2014_statins_ezetimibe_mbma.R (additive shift of -0.117 on the statin Emax in ACS arms; greater statin LDL-C lowering in ACS patients than in the non-ACS reference), Cirincione_2018_apixaban.R, Cirincione_2018_apixaban_asian_subgroups.R (fractional-change effects on TWO structural parameters at once: (1 + e_dis_acs_cl * DIS_ACS) on total apparent apixaban clearance, -21.5 percent in the stage 2 final model, and (1 + e_dis_acs_vc * DIS_ACS) on the apparent central volume, -18.0 percent. Reference category here is the healthy volunteer, NOT the non-ACS patient: the indicator is paired with DIS_NVAF and both zero selects the 270 phase I healthy subjects. The ACS records come from the APPRAISE-1 global phase II study and the Japan ACS phase II study, which the authors included specifically to supply a non-NVAF patient population of similar demographics so that race and patient-status effects could be separated), Willmann_2018a_rivaroxaban.R (multiplicative study / indication effect on rivaroxaban CL/F: 1.14^DIS_ACS, i.e. 14% higher apparent clearance than the VTE-treatment reference; n = 2,251 patients, the largest single cohort in the pooled analysis).
• Notes: Specific scope: the two ratifying papers use the name with the same 1 = ACS-arm orientation but different reference complements (Vargo 2014 pools against non-ACS statin-trial arms; Cirincione 2018 against healthy volunteers, with NVAF carried in a separate indicator), so the coefficient is not portable between them and the per-model covariateData[[DIS_ACS]]$reference_category is load-bearing. Sibling of DIS_NVAF and DIS_HEFH (HeFH cohort indicator) and the broader DIS_<indication> family.

DIS_NVAF (canonical for nonvalvular atrial fibrillation cohort indicator)

• Description: 1 = subject enrolled with nonvalvular atrial fibrillation (NVAF), 0 = non-NVAF cohort. “Nonvalvular” is the therapeutically load-bearing qualifier for this indication, not a stylistic one: the direct oral anticoagulant trials that define the population (ARISTOTLE, AVERROES, ROCKET-AF, RE-LY, ENGAGE AF-TIMI 48) all excluded atrial fibrillation secondary to moderate-to-severe mitral stenosis or a prosthetic heart valve, because a DOAC is not indicated in those patients. A model carrying this indicator is describing the trial population that qualifier defines.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-NVAF cohort). In a pooled analysis that also carries DIS_ACS, the two indicators are mutually exclusive and both zero selects the healthy-volunteer reference; document the complement explicitly per model, because a paper pooling NVAF against a patient rather than a healthy reference gives the same coefficient a different meaning.
• Source aliases:
  • AF – the Cirincione 2018 NONMEM control-stream variable, derived from the subject-type column as AF=0; IF (STYP.EQ.4) AF=1. Note that the source symbol drops the “nonvalvular” qualifier the canonical keeps; the analysis dataset contains only nonvalvular subjects, so the two are the same column.
  • NVAF patients – the Cirincione 2018 Table 3 row label.
• Example models: Cirincione_2018_apixaban.R, Cirincione_2018_apixaban_asian_subgroups.R (fractional-change effects on TWO structural parameters at once: (1 + e_dis_nvaf_cl * DIS_NVAF) on total apparent clearance, -13.9 percent in the stage 2 final model, and (1 + e_dis_nvaf_vc * DIS_NVAF) on the apparent central volume, -4.05 percent. The volume effect is the weaker of the two and is not distinguishable from zero on its own standard error, -0.0405 +/- 0.0491, but the authors retained it because the paired DIS_ACS volume effect is well estimated and the two indicators share a reference category – dropping one alone would silently re-point the other).
• Notes: Specific scope until a second independent paper ratifies the name. Sibling of DIS_ACS and the broader DIS_<indication> family. Distinct from an “any atrial fibrillation” indicator: do not use DIS_NVAF for a cohort that admits valvular AF, and do not treat a real-world-database AF flag as an alias without checking the exclusion criteria, because registry cohorts routinely include the valvular patients the trials excluded. When every subject in an analysis has NVAF (e.g. Lai_2026_rivaroxaban.R, a single-indication real-world cohort) the indicator is degenerate and no column is needed – this canonical exists for analyses that pool NVAF subjects with a different population.

DIS_HEFH (canonical for heterozygous familial hypercholesterolemia cohort indicator)

• Description: 1 = study arm enrolled patients with heterozygous familial hypercholesterolemia (HeFH), 0 = non-HeFH cohort (the model reference).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-HeFH arm).
• Source aliases: HEFH legacy form (used in Vargo 2014 statins / ezetimibe MBMA pre-rename); alias of the canonical DIS_HEFH form rename.
• Example models: Vargo_2014_statins_ezetimibe_mbma.R (additive shift of +0.127 on the statin Emax in HeFH arms; smaller statin LDL-C lowering in HeFH patients than in the non-HeFH reference; biologically consistent with the LDLR-pathway disruption in HeFH), Kuchimanchi_2018_evolocumab_ldlc.R (multiplicative exponent 1.28 on baseline LDL-C in the evolocumab Emax-on-AUC exposure-response layer, Table 4; HeFH patients have higher baseline LDL-C than the non-HeFH reference), Jadhav_2023_bempedoicAcid_ldlc.R (proportional shift of +0.0671 on the indirect-response baseline LDL-C, Jadhav 2023 Table 3; two of the four pivotal phase 3 studies enrolled patients with prior ASCVD and/or HeFH on maximally tolerated statin therapy).
• Notes: Specific scope because the HeFH-cohort effect on statin response is paper-specific. Sibling of DIS_HOFH (homozygous form, Pu_2021_evinacumab) and the broader DIS_<indication> family.

CONMED_RIF_CC, CONMED_INH_CC, CONMED_EMB_CC, CONMED_STR_CC, CONMED_CAB_CC, CONMED_COL_CC, CONMED_MER_CC, CONMED_GEN_CC, CONMED_CIP_CC (canonical for time-varying plasma / in-vitro concentration of a co-administered named drug)

• Description: Time-varying plasma (or in-vitro experiment) concentration of the named drug (suffix = INN lowercase abbreviation: rif rifampicin, inh isoniazid, emb ethambutol, str streptomycin, cab capreomycin / a second antibiotic, col colistin, mer meropenem, gen gentamicin, cip ciprofloxacin). Used as the driving exposure for the corresponding drug-effect term in combination PK/PD or in vitro time-kill models. 0 = the named drug is not part of the regimen for this subject / arm.
• Units: mg/L (or ug/mL, equivalent) for in vitro time-kill experiments; ug/mL for clinical plasma-PK-driven PD subsystems. Per-paper unit is documented in covariateData[[CONMED_<drug>_CC]]$units.
• Type: continuous
• Scope: specific
• Reference category: 0 (drug not in regimen).
• Source aliases: Ccol / Cmer / Cgen / Ccip (Mohamed 2016, Sadouki 2025) and bare EMB / INH / RIF / STR / CAB (Clewe 2018, Khan 2015) – legacy forms used before the 2026-05-28 naming-audit rename. All map to the canonical CONMED_<drug>_CC form.
• Example models: Clewe_2018_rifampicin.R (in vitro time-kill of rifampicin + isoniazid + ethambutol against M. tuberculosis; each drug exposure is a fixed concentration driving the Hill-Emax effect on the F / S / N bacterial states), Khan_2015_ciprofloxacin.R (in vitro streptomycin + capreomycin time-kill), Mohamed_2016_colistin_meropenem.R (in vitro colistin + meropenem time-kill against WT and meropenem-resistant P. aeruginosa), Sadouki_2025_meropenem_gentamicin_ciprofloxacin.R (combination PD with meropenem / gentamicin / ciprofloxacin time-varying exposures).
• Notes: Companion to the dose covariate CONMED_<drug>_DOSE (registered above). The _CC suffix distinguishes the dynamic concentration covariate from the dose-level covariate; both share the <drug> INN abbreviation.

CONMED_MER, CONMED_GEN, CONMED_CIP (canonical for concomitant coadministration indicators of meropenem / gentamicin / ciprofloxacin)

• Description: Binary indicator: 1 = the named antibiotic is coadministered with the index drug for this subject / record / study arm, 0 = the antibiotic is not coadministered. The indicator covers two related use cases that share the same data semantic: (1) in combination-antibiotic PD models, the indicator gates the corresponding drug-effect term (and any pairwise-interaction term that requires both partners present); (2) in popPK models of the index drug, the indicator enters as a covariate on a PK parameter (typically clearance) to capture a multiplicative shift attributable to (or correlated with) the coadministered antibiotic.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (drug not coadministered; the corresponding CONMED_<drug>_CC concentration covariate is then unused / 0).
• Source aliases: MER_PRESENT, GEN_PRESENT, CIP_PRESENT (Sadouki 2025) – the legacy _PRESENT suffix is dropped because binary semantics are conveyed by the type field; renamed 2026-05-28 per the naming audit. GENT (Cohen-Wolkowiez 2014) maps to CONMED_GEN (1 if gentamicin was given concurrently with piperacillin-tazobactam, 0 otherwise).
• Example models: Sadouki_2025_meropenem_gentamicin_ciprofloxacin.R (used together to compute a combination-of-three indicator that triggers a categorical -1 shift on the BETA parameter when all three antibiotics are co-administered), CohenWolkowiez_2014_tazobactam.R (CONMED_GEN enters the tazobactam clearance covariate equation as a multiplicative power-form factor 1.52^CONMED_GEN; in the source paper the gentamicin coadministration covariate was found to be confounded by postmenstrual / postnatal age differences between infants who received gentamicin and those who did not, so the effect should be interpreted as a marker of an older / sicker subset rather than a mechanistic drug-drug interaction).
• Notes: Companion to the concentration covariate CONMED_<drug>_CC (above). For a given drug, CONMED_<drug> is the binary “coadministered?” indicator and CONMED_<drug>_CC is the time-varying plasma concentration. Scope promoted to general on 2026-06-01 alongside the Cohen-Wolkowiez 2014 piperacillin-tazobactam extraction; the underlying semantic (binary coadministration of a named antibiotic) is the same whether the indicator gates a PD interaction term or scales a PK covariate equation, so a single canonical name is preferred over splitting into PD-only and PK-only variants.

BACT (canonical for bacterial-strain indicator in in vitro time-kill models)

• Description: Integer-coded indicator of the bacterial strain used in an in vitro time-kill experiment. Each level represents a structurally distinct strain (e.g., wild-type ATCC reference vs a clinical isolate with a specific resistance pattern); models switch between strain-specific structural parameters at the indicator value. Paper-specific level coding is documented per-file in covariateData[[BACT]]$notes.
• Units: (categorical)
• Type: categorical
• Scope: specific
• Reference category: paper-specific (typically the wild-type ATCC reference strain; e.g., BACT = 2 in Mohamed 2016).
• Source aliases: none.
• Example models: Mohamed_2016_colistin_meropenem.R (1 = ARU552 meropenem-resistant clinical isolate, 2 = ATCC 27853 wild-type; hard switch on 25+ strain-specific structural parameters), Zhao_2024_ciprofloxacin_colistin_invitro.R, Zhao_2024_ciprofloxacin_colistin_plasma.R, Zhao_2024_ciprofloxacin_colistin_kidney.R (47 = C47 clinical urinary isolate, 347 = MG1655 wild type LM347, 378 = LM378 gyrA1 S83L, 421 = LM421 gyrA1 S83L plus marR knockout; levels are the source FSTRAIN codes and switch the colistin Emax, EC50, Hill and interaction parameters plus the pre-existing-resting inoculum fraction), HernandezLozano_2025_apramycin_invitro.R, HernandezLozano_2025_apramycin_mouse.R, HernandezLozano_2025_apramycin_human.R (591 = Escherichia coli EN591, an MDR rmtB clinical isolate; 700336 = Escherichia coli ATCC 700336 / EN1085, a trimethoprim/sulfamethoxazole-resistant urinary isolate. In the in vitro file the level selects the growth rate constant, and jointly with PH_MEDIUM the three drug-effect parameters SlopeS / SlopeR / kada and the MIC; in the two in vivo files it selects the per-organ growth and death rate constants and the strain’s MICs at pH 7.4 and pH 6, while the drug-effect parameters are shared across strains).
• Notes: Specific scope because the per-level strain identity is defined by the source experiment. New in vitro time-kill extractions should reuse this canonical and document the per-level strain mapping in covariate notes. Levels are integer codes with no shared meaning across papers, and there need be no reference level: Zhao 2024 gives every level its own parameter set rather than contrasting against a baseline, whereas Mohamed 2016 treats the wild-type ATCC strain as the reference. A level codes the strain and nothing else: when the source parameter table is indexed by strain and a separately-varied experimental condition, register (or reuse) a canonical for that condition and switch on the pair, rather than multiplying the two factors into composite BACT levels. HernandezLozano_2025_apramycin_invitro.R is the founding example – Table 1 of that paper indexes the in vitro drug-effect parameters by strain and by growth-medium pH, so the file carries BACT (591 / 700336) alongside PH_MEDIUM (6 / 7.4); composite levels would have made a single canonical carry two independent factors, so a model varying strain at fixed pH could not be expressed.

PH_MEDIUM (canonical for the pH of the incubation medium or site of action in an in vitro or site-specific model)

• Description: pH of the medium in which an in vitro experiment was conducted – a bacterial growth medium, or an incubation / uptake buffer – or of the tissue / fluid at the site of action to which a site-referenced constant applies (in vivo). Carried as a covariate whenever pH is a factor the experiment deliberately varied, because pH-sensitive quantities change with it: the MIC of a pH-sensitive antibacterial and, for some strains, its drug-effect parameters; or the affinity and capacity of a proton-coupled membrane transporter. Enters the model as a selector on the affected parameter set, not as a continuous slope, so a model that has only been characterised at the pH levels the source studied should switch on those levels rather than interpolating between them.
• Units: pH units
• Type: continuous
• Scope: general
• Reference category: n/a – not a contrast against a baseline. pH 7.4 is the conventional neutral condition used in standard broth time-kill work and is the natural default for a model that has to pick one; the founding example’s other level is pH 6, chosen because urine is acidic.
• Source aliases: none.
• Example models: HernandezLozano_2025_apramycin_invitro.R (levels 6 and 7.4; apramycin’s MIC is 4-fold higher at pH 6 for both strains – EN591 32 vs 8 mg/L, ATCC 700336 16 vs 4 mg/L – and for EN591 the pH levels additionally take separate SlopeS / SlopeR / kada estimates, whereas MIC-normalization alone lets ATCC 700336 share one drug-effect set across both), Morse_2012_ghb_rbc_invitro.R (levels 6.5 and 7.4; the whole Michaelis-Menten uptake parameter set of the proton-coupled transporter MCT1 switches on it, most sharply the Michaelis constant, 2.2 mM at pH 6.5 against 17.0 mM at pH 7.4 – the transporter binds GHB nearly 8-fold more avidly at the lower pH).
• Notes: General scope because medium / site pH is an experimental condition with the same meaning in any in vitro experiment, unlike the paper-specific integer coding of BACT. Scope broadened 2026-09-02 from antibacterial PD to in vitro work generally, when the Morse 2012 extraction needed the identical column for the pH of an erythrocyte uptake buffer: the covariate’s role there – deliberately varied, selecting a whole parameter set rather than entering as a slope, characterised at only two levels – is the same in every respect, and forking a second pH canonical would have split one concept across two columns. Proton-coupled transporters (the MCT / SLC16 and SMCT / SLC5 families) are the expected recurring non-antibacterial use, because their affinity is pH-dependent by mechanism and the pH difference between blood and the renal tubule is often the scientific point of the paper. Recorded on the natural pH scale (a numeric such as 6 or 7.4), not as an integer level code, so the value is self-describing across papers. Pairs with BACT, which codes the strain and only the strain: the two factors are switched on jointly when a source parameter table is indexed by both. Do not use this column for a model in which pH is fixed rather than varied – the in vivo siblings HernandezLozano_2025_apramycin_mouse.R and HernandezLozano_2025_apramycin_human.R deliberately omit it, because there pH is a property of the organ (kidney assumed 7.4, bladder assumed 6) that is baked into which per-site MIC applies, and the MICs are fixed() ini() parameters a user overrides for a different isolate. The motivating methodological point of the founding paper is that time-kill studies are conventionally run at neutral pH while the target site in a urinary tract infection is acidic, so this covariate is expected to recur in future urinary and other site-specific antibacterial PD extractions.

PERFUSION_FETAL_TO_MATERNAL (canonical for the dosed-circulation (perfusion-direction) indicator of an ex vivo dual-side placenta perfusion)

• Description: Binary indicator of which circulation of an ex vivo dual-side placental cotyledon perfusion the drug was added to: 1 = drug added to the fetal circulation (a fetal-to-maternal, FTM, transfer experiment); 0 = drug added to the maternal circulation (a maternal-to-fetal, MTF, experiment). The structural transfer model is the same in both directions. The indicator selects whatever the source fitted per direction, such as a direction-specific residual error, and in a closed-open configuration it also selects which circulation is the recirculating (closed) one and which is single-pass (open).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 = maternal-to-fetal (drug added to the maternal circulation), the conventional primary direction of placenta perfusion work.
• Source aliases:
  • dosing compartment – Bukkems 2022 Online Resource 16 wording (‘the relevant compartments were turned on based on the dosing compartment covariate’). Encode the maternal-reservoir dosing compartment as 0 and the fetal-reservoir dosing compartment as 1.
  • MTF / FTM – the usual direction labels in perfusion papers (MTF -> 0, FTM -> 1).
• Example models: Bukkems_2022_doravirine_placentaperfusion.R (closed-closed configuration; selects the proportional residual error, 7.5% CV after maternal dosing and 12.6% after fetal dosing), Bukkems_2022_doravirine_placentaperfusion_closedopen.R (closed-open configuration; also selects the open side, a 3 mL collecting reservoir in place of the 200 mL recirculating one, and both the additive and the proportional residual error).
• Notes: General scope because every dual-side placenta perfusion has a dosed circulation, and the same column applies to any drug. The dose record itself still goes to the dosed reservoir state, so the indicator must agree with the dose compartment; it does not redirect the dose. The name is spelled out rather than built on the ROUTE_* family, because the dosed circulation is an experimental configuration of an isolated organ and not a route of administration to a subject. Do not use it for a single-side (open-circuit) perfusion, where there is only one direction.

WT_COTYLEDON (canonical for the wet weight of a perfused placental cotyledon)

• Description: Wet weight of the single placental cotyledon perfused in an ex vivo dual-side placenta perfusion experiment, measured after the experiment. A per-experiment physiologic input used to scale the cotyledon’s tissue sub-volumes (maternal intervillous space, placental barrier, fetal vascular space), and the transfer clearances when the source estimates them per unit of tissue, from a typical cotyledon to the one actually perfused.
• Units: g
• Type: continuous
• Scope: general
• Reference category: n/a – enters as the ratio WT_COTYLEDON / <reference weight>. Reference value observed: 42 g, taken as 44.02 mL (Bukkems 2022 Table 1 and Table 2 footnote a).
• Source aliases:
  • cotyledon weight – Bukkems 2022 Section 2.2 and Online Resources 4 and 14; same orientation and units, no transformation.
• Example models: Bukkems_2022_doravirine_placentaperfusion.R (closed-closed; cohort 18.9-64.4 g, median 32.55 g), Bukkems_2022_doravirine_placentaperfusion_closedopen.R (closed-open; cohort 22.5-41.4 g, median 33.15 g). In both, the three placental sub-volumes and both passive-diffusion clearances scale linearly with WT_COTYLEDON / 42.
• Notes: General scope: cotyledon size varies roughly threefold between perfusions and is the natural normaliser for any mechanistic cotyledon model. Kept separate from WT_BIRTH (neonate birth weight), with which a name lookup can confuse it; the two are different quantities and are often both tabulated in the same placenta-characteristics table. Not a placental weight: a whole term placenta weighs roughly 500 g, and a whole-placenta weight would need its own canonical.

LOWINOC (canonical for low-inoculum experiment indicator)

• Description: Binary indicator: 1 = study arm / replicate used the low-inoculum experimental design (typically 10^5 - 10^6 CFU/mL starting bacterial count), 0 = standard or high inoculum (typically 10^7 - 10^9 CFU/mL). Captures the inoculum-effect deviation in bacterial-kill kinetics across initial-density groups.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (standard / high inoculum).
• Source aliases: none.
• Example models: Sadouki_2025_meropenem_gentamicin_ciprofloxacin.R (gates a low-inoculum-specific shift on the bacterial-kill amplitude; the paper found the low-inoculum arm had a measurably different kill rate from the standard-inoculum arm).
• Notes: Specific scope because the per-paper definition of “low” varies by experiment design.

FLARE (canonical for flare-design study-arm indicator)

• Description: Binary study-arm indicator: 1 = the trial used a flare design (subjects washed out of pain medication and required a predefined pain flare-up before randomisation), 0 = non-flare design. A property of the trial design in a model-based meta-analysis (MBMA), not of an individual patient.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-flare design).
• Source aliases:
  • If – used in Boucher_2018_naproxen_mbma.R (Boucher 2018 Eqs 2-3).
  • Flare / Fi – used in Boucher_2016_naproxen_mbma.R (Boucher 2016 Supplementary Materials Table 1 column header, and Fi in the “Meta-regression model” equation).
• Example models: Boucher_2018_naproxen_mbma.R (shifts both baseline WOMAC pain e0 and emax; 12 of 18 trials were flare designs), Boucher_2016_naproxen_mbma.R (the earlier landmark analysis by the same authors; 9 of 13 trials were flare designs).
• Notes: MBMA study-arm-level covariate (a property of the trial design). Specific scope because “flare design” is defined per the osteoarthritis-pain trial-design literature. The effect target depends on what the meta-analysis pools. When the modelled data point is a study-ARM mean (Boucher_2018_naproxen_mbma.R), flare shifts the arm’s baseline and maximal effect, because a flare design raises the pain level at randomisation in the placebo and active arms alike. When the modelled data point is an already-differenced within-trial CONTRAST (Boucher_2016_naproxen_mbma.R), the trial main effect has been eliminated, so the single flare coefficient is a treatment-by-design interaction – whether the drug-placebo separation itself is larger in flare trials – and is a much smaller quantity. Do not carry a flare coefficient across the two framings. In the founding landmark analysis that interaction was NOT significant (-0.14, 95% CI -0.47 to 0.18).

NAPROXEN (canonical for naproxen treatment-arm indicator)

• Description: Binary study-arm treatment indicator: 1 = the arm received naproxen, 0 = the arm received placebo. A property of the trial arm in a model-based meta-analysis (MBMA), not of an individual patient.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (placebo arm).
• Source aliases:
  • In – used in Boucher_2018_naproxen_mbma.R (Boucher 2018 Eqs 3-4).
• Example models: Boucher_2018_naproxen_mbma.R (shifts emax and shortens et50; all 18 included trials had both a naproxen and a placebo arm).
• Notes: MBMA study-arm-level treatment indicator. Specific scope (a per-drug arm indicator).

TRAMADOL (canonical for tramadol treatment-arm indicator)

• Description: Binary study-arm treatment indicator: 1 = the arm received tramadol, 0 = the arm did not receive tramadol (placebo, tapentadol, or another active-control arm). A property of the trial arm in a model-based meta-analysis (MBMA), not of an individual patient.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (arm did not receive tramadol).
• Source aliases: none (the source paper uses the drug name directly to select the extent-of-reduction term).
• Example models: Mercier_2014_tramadol_tapentadol_mbma.R (selects the tramadol Emax-in-dose term e_tramadol_emax * CONMED_TRAMADOL_DOSE / (CONMED_TRAMADOL_DOSE + ED50) * TRAMADOL in the extent-of-reduction R; 43 of 81 arms in the pooled 45-trial database were tramadol arms).
• Notes: MBMA study-arm-level treatment indicator, sibling of NAPROXEN in the per-drug MBMA arm-indicator family. Specific scope (a per-drug arm indicator). Pairs with TAPENTADOL and CONMED_TRAMADOL_DOSE in the Mercier 2014 chronic-non-malignant-pain extraction; the two treatment indicators are mutually exclusive (a single arm can be tramadol OR tapentadol, never both).

TAPENTADOL (canonical for tapentadol treatment-arm indicator)

• Description: Binary study-arm treatment indicator: 1 = the arm received tapentadol, 0 = the arm did not receive tapentadol (placebo, tramadol, or another active-control arm). A property of the trial arm in a model-based meta-analysis (MBMA), not of an individual patient.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (arm did not receive tapentadol).
• Source aliases: none (the source paper uses the drug name directly).
• Example models: Mercier_2014_tramadol_tapentadol_mbma.R (enters the extent-of-reduction R additively via e_tapentadol_emax * TAPENTADOL; no dose-response is fit because tapentadol was studied only over the narrow 100-250 mg bid range in the Mercier 2014 database; 8 of 81 arms were tapentadol arms).
• Notes: MBMA study-arm-level treatment indicator, sibling of NAPROXEN and TRAMADOL in the per-drug MBMA arm-indicator family. Specific scope. Pairs with TRAMADOL in the Mercier 2014 extraction; the two are mutually exclusive per arm.

ACETAMINOPHEN (canonical for acetaminophen (paracetamol) treatment-arm indicator)

• Description: Binary study-arm treatment indicator: 1 = the arm received systemic acetaminophen (paracetamol), 0 = it did not. A property of the trial arm in a model-based meta-analysis (MBMA), not of an individual patient.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (arm did not receive acetaminophen).
• Source aliases:
  • ace – used in Sethi_2024_acetaminophen_diclofenac_mbma.R (Sethi 2024 Table 1 Treatments column; the parameter is e.ace in Table 3).
• Example models: Sethi_2024_acetaminophen_diclofenac_mbma.R (additive -18.93 mg shift on arm-mean opioid PCA consumption via e_acetaminophen_e0 * ACETAMINOPHEN; set to 1 on the combination arms as well as the monotherapy arms, so the arm indicator and DICLOFENAC are NOT mutually exclusive in this extraction).
• Notes: MBMA study-arm-level treatment indicator, sibling of NAPROXEN, TRAMADOL and TAPENTADOL in the per-drug MBMA arm-indicator family. Specific scope. Distinct from CONMED_PARA, which flags concomitant paracetamol taken alongside the drug under study at an individual-patient observation; use ACETAMINOPHEN when acetaminophen is itself a randomised study treatment at the trial-arm level. Dose-independent in the founding model: Sethi 2024 pooled arms at 1000-2400 mg because five trials could not identify a dose-response, so a future extraction that does fit an acetaminophen dose-response should pair this indicator with a CONMED_ACETAMINOPHEN_DOSE-style dose column rather than overload the indicator.

DICLOFENAC (canonical for diclofenac treatment-arm indicator)

• Description: Binary study-arm treatment indicator: 1 = the arm received diclofenac, 0 = it did not. A property of the trial arm in a model-based meta-analysis (MBMA), not of an individual patient.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (arm did not receive diclofenac).
• Source aliases:
  • dic – used in Sethi_2024_acetaminophen_diclofenac_mbma.R (Sethi 2024 Table 1 Treatments column; the parameter is e.dic in Table 3).
• Example models: Sethi_2024_acetaminophen_diclofenac_mbma.R (additive -28.41 mg shift on arm-mean opioid PCA consumption via e_diclofenac_e0 * DICLOFENAC; set to 1 on the combination arms as well as the monotherapy arms, so the arm indicator and ACETAMINOPHEN are NOT mutually exclusive in this extraction).
• Notes: MBMA study-arm-level treatment indicator, sibling of NAPROXEN, TRAMADOL, TAPENTADOL and ACETAMINOPHEN in the per-drug MBMA arm-indicator family. Specific scope. The route is not encoded in the indicator and must be documented per model. In the founding model every diclofenac arm was systemic (oral, rectal or intravenous, 75-100 mg) even though the source paper’s title concerns topical diclofenac – Sethi 2024 Supplementary Table S5 labels its rows “Systemic diclofenac”, and the paper’s move from systemic to topical is a qualitative Discussion extrapolation, not a fitted effect. A future extraction that distinguishes topical from systemic diclofenac arms should carry a separate FORM_DICLOFENAC_TOPICAL-style formulation indicator alongside this one. Dose-independent in the founding model (75 mg and 100 mg arms share one effect).

NINTEDANIB (canonical for nintedanib treatment-arm indicator)

• Description: Binary study-arm treatment indicator: 1 = the arm received nintedanib, 0 = the arm received placebo or a different active treatment. A property of the trial arm in a model-based meta-analysis (MBMA), not of an individual patient. Drug class: a tyrosine kinase inhibitor and one of the two FDA-approved IPF therapies.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the arm received placebo or a different active treatment).
• Source aliases:
  • Nintedanib treatment – the Chan 2017 Table 2 row label. Free-text label, not a column header; downstream data sets should use the canonical column name NINTEDANIB.
• Example models: Chan_2017_ipf_fvc_mbma.R (Chan 2017 Table 1: 3 arms, 21 timepoints, 723 subjects, 300 mg/day, from TOMORROW and INPULSIS-1/2. Selects emax_nintedanib = 3.31 %predicted FVC, one of only two effects whose 95% CI excluded zero. The Figure 4 forest plot labels the two INPULSIS arms 281.7 and 280.5 mg/day – exposure-weighted mean daily doses after interruptions – while Table 1 and the model use the nominal 300 mg/day; the dose does not enter the model. Note that nintedanib also appears in this register inside STUDY_LUMELUNG2 and DIS_SSC_ILD, neither of which is an arm indicator.).
• Notes: MBMA study-arm-level treatment indicator, sibling of NAPROXEN, TRAMADOL, TAPENTADOL, ACETAMINOPHEN and DICLOFENAC in the per-drug MBMA arm-indicator family. Specific scope. Named under the register’s explicit written rule, recorded in the TRT_BENRALIZUMAB entry Notes: use TRT_<INN> when the drug is the study treatment under investigation, CONMED_<INN> when it is background co-medication, and a bare INN only for MBMA trial-arm indicators. The register carries a competing precedent that upstream has not reconciled – the Sandra 2024 anti-HBV siRNA batch (Sandra_2024_siRNA_hbsag_mouse_mbma.R) registered 13 MBMA arm indicators as TRT_<drug>, citing the bare-INN family and then departing from it. A ruling in favour of TRT_<INN> should migrate the whole bare-INN family, now 19 members, rather than leave a one-off. In the founding model all 14 Chan 2017 indicators are mutually exclusive and all are 0 on a placebo arm, which is what makes the placebo-corrected output zero there by construction.

PIRFENIDONE (canonical for pirfenidone treatment-arm indicator)

• Description: Binary study-arm treatment indicator: 1 = the arm received pirfenidone, 0 = the arm received placebo or a different active treatment. A property of the trial arm in a model-based meta-analysis (MBMA), not of an individual patient. Drug class: an antifibrotic and the other FDA-approved IPF therapy.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the arm received placebo or a different active treatment).
• Source aliases:
  • Pirfenidone treatment – the Chan 2017 Table 2 row label. Free-text label, not a column header; downstream data sets should use the canonical column name PIRFENIDONE.
• Example models: Chan_2017_ipf_fvc_mbma.R (Chan 2017 Table 1: 4 arms, 22 timepoints, 710 subjects, from CAPACITY-1/2 and ASCEND. The only member of this family that must be paired with a dose column: pirfenidone is the one treatment for which Chan 2017 estimated a dose-response, as a step function, so this indicator selects the pirfenidone branch and DOSE_HIGH picks 2.42 (1,197 mg/day) or 3.87 (2,403 mg/day) within it. Largest effect among the approved drugs at its recommended dose, and one of only two whose 95% CI excluded zero.).
• Notes: MBMA study-arm-level treatment indicator, sibling of NAPROXEN, TRAMADOL, TAPENTADOL, ACETAMINOPHEN and DICLOFENAC in the per-drug MBMA arm-indicator family. Specific scope. Named under the register’s explicit written rule, recorded in the TRT_BENRALIZUMAB entry Notes: use TRT_<INN> when the drug is the study treatment under investigation, CONMED_<INN> when it is background co-medication, and a bare INN only for MBMA trial-arm indicators. The register carries a competing precedent that upstream has not reconciled – the Sandra 2024 anti-HBV siRNA batch (Sandra_2024_siRNA_hbsag_mouse_mbma.R) registered 13 MBMA arm indicators as TRT_<drug>, citing the bare-INN family and then departing from it. A ruling in favour of TRT_<INN> should migrate the whole bare-INN family, now 19 members, rather than leave a one-off. In the founding model all 14 Chan 2017 indicators are mutually exclusive and all are 0 on a placebo arm, which is what makes the placebo-corrected output zero there by construction.

PRM151 (canonical for PRM-151 treatment-arm indicator)

• Description: Binary study-arm treatment indicator: 1 = the arm received PRM-151 (recombinant human pentraxin-2 / serum amyloid P; later zinpentraxin alfa), 0 = the arm received placebo or a different active treatment. A property of the trial arm in a model-based meta-analysis (MBMA), not of an individual patient. Drug class: an antifibrotic investigational agent.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the arm received placebo or a different active treatment).
• Source aliases:
  • Prm-151 treatment – the Chan 2017 Table 2 row label. Free-text label, not a column header; downstream data sets should use the canonical column name PRM151.
• Example models: Chan_2017_ipf_fvc_mbma.R (Named by the development code the source uses throughout, following the CONMED_AZD9668_DOSE / CONMED_GSK233705_DOSE precedent for agents a source identifies only by code. Chan 2017 Table 1: 3 arms, 3 timepoints, 15 subjects, at a single 8-week timepoint (trial PRM151F-12GL). Selects emax_prm151 = 12.5 %predicted FVC – numerically the largest effect in the analysis and the least reliable, 95% CI -5.13 to 30.1; the Discussion states the estimate ‘was not precise or reliable’. Three dose strengths (1, 5, 10 mg/kg) are in the dataset but a dose-response would not minimize, so all three arms share this one indicator and no dose column applies.).
• Notes: MBMA study-arm-level treatment indicator, sibling of NAPROXEN, TRAMADOL, TAPENTADOL, ACETAMINOPHEN and DICLOFENAC in the per-drug MBMA arm-indicator family. Specific scope. Named under the register’s explicit written rule, recorded in the TRT_BENRALIZUMAB entry Notes: use TRT_<INN> when the drug is the study treatment under investigation, CONMED_<INN> when it is background co-medication, and a bare INN only for MBMA trial-arm indicators. The register carries a competing precedent that upstream has not reconciled – the Sandra 2024 anti-HBV siRNA batch (Sandra_2024_siRNA_hbsag_mouse_mbma.R) registered 13 MBMA arm indicators as TRT_<drug>, citing the bare-INN family and then departing from it. A ruling in favour of TRT_<INN> should migrate the whole bare-INN family, now 19 members, rather than leave a one-off. In the founding model all 14 Chan 2017 indicators are mutually exclusive and all are 0 on a placebo arm, which is what makes the placebo-corrected output zero there by construction.

INTERFERON_GAMMA (canonical for interferon gamma-1b treatment-arm indicator)

• Description: Binary study-arm treatment indicator: 1 = the arm received interferon gamma-1b, 0 = the arm received placebo or a different active treatment. A property of the trial arm in a model-based meta-analysis (MBMA), not of an individual patient. Drug class: an interferon.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the arm received placebo or a different active treatment).
• Source aliases:
  • Interferon gamma-1b treatment – the Chan 2017 Table 2 row label. Free-text label, not a column header; downstream data sets should use the canonical column name INTERFERON_GAMMA.
• Example models: Chan_2017_ipf_fvc_mbma.R (Chan 2017 Table 1: 4 arms, 8 timepoints, 607 subjects, 200 ug – the most-studied active treatment in the dataset by trial count. Selects emax_ifngamma = -0.618 %predicted FVC. Load-bearing well beyond its own arms: colchicine and prednisone had no placebo-controlled IPF data and reach placebo only INDIRECTLY through small active-control trials against interferon gamma, so their estimates inherit this one’s. Chan 2017 Discussion flags it as ‘an obvious outlier’ – two trials reported a large positive effect and two none, with the large INSPIRE trial dominating. The underscore in the canonical name separates the two words of a compound biological name and does not indicate a sub-scope.).
• Notes: MBMA study-arm-level treatment indicator, sibling of NAPROXEN, TRAMADOL, TAPENTADOL, ACETAMINOPHEN and DICLOFENAC in the per-drug MBMA arm-indicator family. Specific scope. Named under the register’s explicit written rule, recorded in the TRT_BENRALIZUMAB entry Notes: use TRT_<INN> when the drug is the study treatment under investigation, CONMED_<INN> when it is background co-medication, and a bare INN only for MBMA trial-arm indicators. The register carries a competing precedent that upstream has not reconciled – the Sandra 2024 anti-HBV siRNA batch (Sandra_2024_siRNA_hbsag_mouse_mbma.R) registered 13 MBMA arm indicators as TRT_<drug>, citing the bare-INN family and then departing from it. A ruling in favour of TRT_<INN> should migrate the whole bare-INN family, now 19 members, rather than leave a one-off. In the founding model all 14 Chan 2017 indicators are mutually exclusive and all are 0 on a placebo arm, which is what makes the placebo-corrected output zero there by construction.

COLCHICINE (canonical for colchicine treatment-arm indicator)

• Description: Binary study-arm treatment indicator: 1 = the arm received colchicine, 0 = the arm received placebo or a different active treatment. A property of the trial arm in a model-based meta-analysis (MBMA), not of an individual patient. Drug class: an antigout agent.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the arm received placebo or a different active treatment).
• Source aliases:
  • Colchicine treatment – the Chan 2017 Table 2 row label. Free-text label, not a column header; downstream data sets should use the canonical column name COLCHICINE.
• Example models: Chan_2017_ipf_fvc_mbma.R (Chan 2017 Table 1: 3 arms, 7 timepoints, 39 subjects, 1 mg (the Figure 4 forest plot gives the Douglas 1998 arm as 0.9 mg). Selects emax_colchicine = -7.63 %predicted FVC, the second most negative effect in the analysis. Indirectly estimated: the Discussion records that colchicine ‘lacked placebo-controlled trial data in subjects with IPF’ and reached placebo only through small active-control trials against interferon gamma, which itself showed no effect in INSPIRE – hence the large negative point estimate on 39 subjects and the wide CI.).
• Notes: MBMA study-arm-level treatment indicator, sibling of NAPROXEN, TRAMADOL, TAPENTADOL, ACETAMINOPHEN and DICLOFENAC in the per-drug MBMA arm-indicator family. Specific scope. Named under the register’s explicit written rule, recorded in the TRT_BENRALIZUMAB entry Notes: use TRT_<INN> when the drug is the study treatment under investigation, CONMED_<INN> when it is background co-medication, and a bare INN only for MBMA trial-arm indicators. The register carries a competing precedent that upstream has not reconciled – the Sandra 2024 anti-HBV siRNA batch (Sandra_2024_siRNA_hbsag_mouse_mbma.R) registered 13 MBMA arm indicators as TRT_<drug>, citing the bare-INN family and then departing from it. A ruling in favour of TRT_<INN> should migrate the whole bare-INN family, now 19 members, rather than leave a one-off. In the founding model all 14 Chan 2017 indicators are mutually exclusive and all are 0 on a placebo arm, which is what makes the placebo-corrected output zero there by construction.

PREDNISONE (canonical for prednisone treatment-arm indicator)

• Description: Binary study-arm treatment indicator: 1 = the arm received prednisone, 0 = the arm received placebo or a different active treatment. A property of the trial arm in a model-based meta-analysis (MBMA), not of an individual patient. Drug class: a corticosteroid.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the arm received placebo or a different active treatment).
• Source aliases:
  • Prednisone treatment – the Chan 2017 Table 2 row label. Free-text label, not a column header; downstream data sets should use the canonical column name PREDNISONE.
• Example models: Chan_2017_ipf_fvc_mbma.R (Chan 2017 Table 1: 1 arm, 1 timepoint, 12 subjects, 40 mg. Selects emax_prednisone = -11.9 %predicted FVC, the most negative effect in the analysis, with a CI spanning 28 %predicted FVC on 12 subjects. Indirectly estimated through the same interferon-gamma active-control chain as COLCHICINE. Distinct from the registered DRUG_PRED, a prednisone-versus-fosdagrocorat comparator indicator, and from PRED_DOSE, a concomitant oral prednisolone daily dose – this entry is prednisone as the randomized study treatment of a meta-analysed trial arm.).
• Notes: MBMA study-arm-level treatment indicator, sibling of NAPROXEN, TRAMADOL, TAPENTADOL, ACETAMINOPHEN and DICLOFENAC in the per-drug MBMA arm-indicator family. Specific scope. Named under the register’s explicit written rule, recorded in the TRT_BENRALIZUMAB entry Notes: use TRT_<INN> when the drug is the study treatment under investigation, CONMED_<INN> when it is background co-medication, and a bare INN only for MBMA trial-arm indicators. The register carries a competing precedent that upstream has not reconciled – the Sandra 2024 anti-HBV siRNA batch (Sandra_2024_siRNA_hbsag_mouse_mbma.R) registered 13 MBMA arm indicators as TRT_<drug>, citing the bare-INN family and then departing from it. A ruling in favour of TRT_<INN> should migrate the whole bare-INN family, now 19 members, rather than leave a one-off. In the founding model all 14 Chan 2017 indicators are mutually exclusive and all are 0 on a placebo arm, which is what makes the placebo-corrected output zero there by construction.

SILDENAFIL (canonical for sildenafil treatment-arm indicator)

• Description: Binary study-arm treatment indicator: 1 = the arm received sildenafil, 0 = the arm received placebo or a different active treatment. A property of the trial arm in a model-based meta-analysis (MBMA), not of an individual patient. Drug class: a phosphodiesterase inhibitor.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the arm received placebo or a different active treatment).
• Source aliases:
  • Sildenafil treatment – the Chan 2017 Table 2 row label. Free-text label, not a column header; downstream data sets should use the canonical column name SILDENAFIL.
• Example models: Chan_2017_ipf_fvc_mbma.R (Chan 2017 Table 1: 1 arm, 1 timepoint, 14 subjects, 20 mg (trial CLIN-009-05F). Selects emax_sildenafil = 2.67 %predicted FVC, 95% CI -7.15 to 12.5. Distinct from the registered CONMED_SILDENAFIL, which flags sildenafil as background co-medication at the individual-patient level; the two encode opposite roles and the bare form is the arm-indicator one.).
• Notes: MBMA study-arm-level treatment indicator, sibling of NAPROXEN, TRAMADOL, TAPENTADOL, ACETAMINOPHEN and DICLOFENAC in the per-drug MBMA arm-indicator family. Specific scope. Named under the register’s explicit written rule, recorded in the TRT_BENRALIZUMAB entry Notes: use TRT_<INN> when the drug is the study treatment under investigation, CONMED_<INN> when it is background co-medication, and a bare INN only for MBMA trial-arm indicators. The register carries a competing precedent that upstream has not reconciled – the Sandra 2024 anti-HBV siRNA batch (Sandra_2024_siRNA_hbsag_mouse_mbma.R) registered 13 MBMA arm indicators as TRT_<drug>, citing the bare-INN family and then departing from it. A ruling in favour of TRT_<INN> should migrate the whole bare-INN family, now 19 members, rather than leave a one-off. In the founding model all 14 Chan 2017 indicators are mutually exclusive and all are 0 on a placebo arm, which is what makes the placebo-corrected output zero there by construction.

BOSENTAN (canonical for bosentan treatment-arm indicator)

• Description: Binary study-arm treatment indicator: 1 = the arm received bosentan, 0 = the arm received placebo or a different active treatment. A property of the trial arm in a model-based meta-analysis (MBMA), not of an individual patient. Drug class: an endothelin receptor antagonist.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the arm received placebo or a different active treatment).
• Source aliases:
  • Bosentan treatment – the Chan 2017 Table 2 row label. Free-text label, not a column header; downstream data sets should use the canonical column name BOSENTAN.
• Example models: Chan_2017_ipf_fvc_mbma.R (Chan 2017 Table 1: 1 arm, 1 timepoint, 74 subjects, 125 mg (trial BUILD-1). Selects emax_bosentan = 1.75 %predicted FVC. Chan 2017 Results records that the endothelin-receptor-antagonist class – ambrisentan, bosentan and macitentan – had a HIGHER predicted residual SD than other classes in the separate SD-imputation model, which affects the inverse-variance study weights but is not a parameter of the published model. Distinct from the registered CONMED_BOSENTAN (background co-medication).).
• Notes: MBMA study-arm-level treatment indicator, sibling of NAPROXEN, TRAMADOL, TAPENTADOL, ACETAMINOPHEN and DICLOFENAC in the per-drug MBMA arm-indicator family. Specific scope. Named under the register’s explicit written rule, recorded in the TRT_BENRALIZUMAB entry Notes: use TRT_<INN> when the drug is the study treatment under investigation, CONMED_<INN> when it is background co-medication, and a bare INN only for MBMA trial-arm indicators. The register carries a competing precedent that upstream has not reconciled – the Sandra 2024 anti-HBV siRNA batch (Sandra_2024_siRNA_hbsag_mouse_mbma.R) registered 13 MBMA arm indicators as TRT_<drug>, citing the bare-INN family and then departing from it. A ruling in favour of TRT_<INN> should migrate the whole bare-INN family, now 19 members, rather than leave a one-off. In the founding model all 14 Chan 2017 indicators are mutually exclusive and all are 0 on a placebo arm, which is what makes the placebo-corrected output zero there by construction.

AMBRISENTAN (canonical for ambrisentan treatment-arm indicator)

• Description: Binary study-arm treatment indicator: 1 = the arm received ambrisentan, 0 = the arm received placebo or a different active treatment. A property of the trial arm in a model-based meta-analysis (MBMA), not of an individual patient. Drug class: an endothelin receptor antagonist.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the arm received placebo or a different active treatment).
• Source aliases:
  • Ambrisentan treatment – the Chan 2017 Table 2 row label. Free-text label, not a column header; downstream data sets should use the canonical column name AMBRISENTAN.
• Example models: Chan_2017_ipf_fvc_mbma.R (Chan 2017 Table 1: 1 arm, 7 timepoints, 330 subjects, 10 mg (trial ARTEMIS-IPF) – one of the eight treatments contributing longitudinal data. Selects emax_ambrisentan = -2.03 %predicted FVC, whose 95% CI (-4.18 to 0.126) is the narrowest of any non-significant treatment. See BOSENTAN on the class-level residual-SD finding.).
• Notes: MBMA study-arm-level treatment indicator, sibling of NAPROXEN, TRAMADOL, TAPENTADOL, ACETAMINOPHEN and DICLOFENAC in the per-drug MBMA arm-indicator family. Specific scope. Named under the register’s explicit written rule, recorded in the TRT_BENRALIZUMAB entry Notes: use TRT_<INN> when the drug is the study treatment under investigation, CONMED_<INN> when it is background co-medication, and a bare INN only for MBMA trial-arm indicators. The register carries a competing precedent that upstream has not reconciled – the Sandra 2024 anti-HBV siRNA batch (Sandra_2024_siRNA_hbsag_mouse_mbma.R) registered 13 MBMA arm indicators as TRT_<drug>, citing the bare-INN family and then departing from it. A ruling in favour of TRT_<INN> should migrate the whole bare-INN family, now 19 members, rather than leave a one-off. In the founding model all 14 Chan 2017 indicators are mutually exclusive and all are 0 on a placebo arm, which is what makes the placebo-corrected output zero there by construction.

ACETYLCYSTEINE (canonical for acetylcysteine treatment-arm indicator)

• Description: Binary study-arm treatment indicator: 1 = the arm received acetylcysteine (the source writes N-acetylcysteine), 0 = the arm received placebo or a different active treatment. A property of the trial arm in a model-based meta-analysis (MBMA), not of an individual patient. Drug class: a mucokinetic.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the arm received placebo or a different active treatment).
• Source aliases:
  • Acetylcysteine treatment – the Chan 2017 Table 2 row label. Free-text label, not a column header; downstream data sets should use the canonical column name ACETYLCYSTEINE.
• Example models: Chan_2017_ipf_fvc_mbma.R (Registered under the INN acetylcysteine; sources commonly write N-acetylcysteine or NAC, which translate to this canonical. Chan 2017 Table 1: 1 arm, 4 timepoints, 133 subjects, 1,800 mg/day (trial PANTHER-IPF) – one of the eight treatments contributing longitudinal data. Selects emax_acetylcysteine = -0.0568 %predicted FVC, the effect closest to exactly zero in the analysis.).
• Notes: MBMA study-arm-level treatment indicator, sibling of NAPROXEN, TRAMADOL, TAPENTADOL, ACETAMINOPHEN and DICLOFENAC in the per-drug MBMA arm-indicator family. Specific scope. Named under the register’s explicit written rule, recorded in the TRT_BENRALIZUMAB entry Notes: use TRT_<INN> when the drug is the study treatment under investigation, CONMED_<INN> when it is background co-medication, and a bare INN only for MBMA trial-arm indicators. The register carries a competing precedent that upstream has not reconciled – the Sandra 2024 anti-HBV siRNA batch (Sandra_2024_siRNA_hbsag_mouse_mbma.R) registered 13 MBMA arm indicators as TRT_<drug>, citing the bare-INN family and then departing from it. A ruling in favour of TRT_<INN> should migrate the whole bare-INN family, now 19 members, rather than leave a one-off. In the founding model all 14 Chan 2017 indicators are mutually exclusive and all are 0 on a placebo arm, which is what makes the placebo-corrected output zero there by construction.

WARFARIN (canonical for warfarin treatment-arm indicator)

• Description: Binary study-arm treatment indicator: 1 = the arm received warfarin, 0 = the arm received placebo or a different active treatment. A property of the trial arm in a model-based meta-analysis (MBMA), not of an individual patient. Drug class: an anticoagulant.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the arm received placebo or a different active treatment).
• Source aliases:
  • Warfarin treatment – the Chan 2017 Table 2 row label. Free-text label, not a column header; downstream data sets should use the canonical column name WARFARIN.
• Example models: Chan_2017_ipf_fvc_mbma.R (Chan 2017 Table 1: 1 arm, 3 timepoints, 72 subjects, 1.8 mg (trial ACE-IPF) – one of the eight treatments contributing longitudinal data. Selects emax_warfarin = -0.276 %predicted FVC. The 1.8 mg is the trial’s mean ACHIEVED daily dose under INR-guided titration rather than a fixed assigned dose, so it should not be read as a regimen; the dose does not enter the model.).
• Notes: MBMA study-arm-level treatment indicator, sibling of NAPROXEN, TRAMADOL, TAPENTADOL, ACETAMINOPHEN and DICLOFENAC in the per-drug MBMA arm-indicator family. Specific scope. Named under the register’s explicit written rule, recorded in the TRT_BENRALIZUMAB entry Notes: use TRT_<INN> when the drug is the study treatment under investigation, CONMED_<INN> when it is background co-medication, and a bare INN only for MBMA trial-arm indicators. The register carries a competing precedent that upstream has not reconciled – the Sandra 2024 anti-HBV siRNA batch (Sandra_2024_siRNA_hbsag_mouse_mbma.R) registered 13 MBMA arm indicators as TRT_<drug>, citing the bare-INN family and then departing from it. A ruling in favour of TRT_<INN> should migrate the whole bare-INN family, now 19 members, rather than leave a one-off. In the founding model all 14 Chan 2017 indicators are mutually exclusive and all are 0 on a placebo arm, which is what makes the placebo-corrected output zero there by construction.

ETANERCEPT (canonical for etanercept treatment-arm indicator)

• Description: Binary study-arm treatment indicator: 1 = the arm received etanercept, 0 = the arm received placebo or a different active treatment. A property of the trial arm in a model-based meta-analysis (MBMA), not of an individual patient. Drug class: a tumor necrosis factor inhibitor.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the arm received placebo or a different active treatment).
• Source aliases:
  • Etanercept treatment – the Chan 2017 Table 2 row label. Free-text label, not a column header; downstream data sets should use the canonical column name ETANERCEPT.
• Example models: Chan_2017_ipf_fvc_mbma.R (Chan 2017 Table 1: 1 arm, 5 timepoints, 46 subjects, 25 mg (trial 0881A4-203) – one of the eight treatments contributing longitudinal data, and the only TNF inhibitor in the dataset. Selects emax_etanercept = 1.47 %predicted FVC.), Leil_2021_rheumatoidArthritis_das28_mbma.R (Leil 2021: 8 trials, 1,747 patients on 50 mg SC qw; selects lemax_etanercept / let50_etanercept of the DAS28 change-from-baseline Emax-in-time treatment effect).
• Notes: MBMA study-arm-level treatment indicator, sibling of NAPROXEN, TRAMADOL, TAPENTADOL, ACETAMINOPHEN and DICLOFENAC in the per-drug MBMA arm-indicator family. Specific scope. Named under the register’s explicit written rule, recorded in the TRT_BENRALIZUMAB entry Notes: use TRT_<INN> when the drug is the study treatment under investigation, CONMED_<INN> when it is background co-medication, and a bare INN only for MBMA trial-arm indicators. The register carries a competing precedent that upstream has not reconciled – the Sandra 2024 anti-HBV siRNA batch (Sandra_2024_siRNA_hbsag_mouse_mbma.R) registered 13 MBMA arm indicators as TRT_<drug>, citing the bare-INN family and then departing from it. A ruling in favour of TRT_<INN> should migrate the whole bare-INN family, now 19 members, rather than leave a one-off. In the founding model all 14 Chan 2017 indicators are mutually exclusive and all are 0 on a placebo arm, which is what makes the placebo-corrected output zero there by construction.

ABATACEPT (canonical for abatacept treatment-arm indicator)

• Description: Binary study-arm treatment indicator: 1 = the arm received abatacept, 0 = the arm received placebo or a different active treatment. A property of the trial arm in a model-based meta-analysis (MBMA), not of an individual patient. Drug class: a T-cell costimulation modulator (CTLA-4-Ig).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the arm received placebo or a different active treatment).
• Source aliases: none (Leil 2021 identifies the arm’s drug by name in Table 2 and Supplementary Table S1).
• Example models: Leil_2021_rheumatoidArthritis_das28_mbma.R (Leil 2021 Table 2: 10 mg/kg IV q4w or 125 mg SC qw; 16 trials, 3,478 patients; selects lemax_abatacept / let50_abatacept of the DAS28 change-from-baseline Emax-in-time treatment effect).
• Notes: MBMA study-arm-level treatment indicator in the bare-INN per-drug arm-indicator family (siblings ETANERCEPT, NAPROXEN, TRAMADOL, …); see the ETANERCEPT entry Notes for the naming rule. Mutually exclusive with the other Leil 2021 drug indicators; all zero marks a background-therapy (placebo) arm.

ADALIMUMAB (canonical for adalimumab treatment-arm indicator)

• Description: Binary study-arm treatment indicator: 1 = the arm received adalimumab, 0 = the arm received placebo or a different active treatment. A property of the trial arm in a model-based meta-analysis (MBMA), not of an individual patient. Drug class: a tumor necrosis factor inhibitor.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the arm received placebo or a different active treatment).
• Source aliases: none (Leil 2021 identifies the arm’s drug by name in Table 2 and Supplementary Table S1).
• Example models: Leil_2021_rheumatoidArthritis_das28_mbma.R (Leil 2021 Table 2: 40 mg SC q2w; 15 trials, 2,802 patients; selects lemax_adalimumab / let50_adalimumab of the DAS28 change-from-baseline Emax-in-time treatment effect).
• Notes: MBMA study-arm-level treatment indicator in the bare-INN per-drug arm-indicator family (siblings ETANERCEPT, NAPROXEN, TRAMADOL, …); see the ETANERCEPT entry Notes for the naming rule. Mutually exclusive with the other Leil 2021 drug indicators; all zero marks a background-therapy (placebo) arm.

CERTOLIZUMAB (canonical for certolizumab pegol treatment-arm indicator)

• Description: Binary study-arm treatment indicator: 1 = the arm received certolizumab pegol, 0 = the arm received placebo or a different active treatment. A property of the trial arm in a model-based meta-analysis (MBMA), not of an individual patient. Drug class: a tumor necrosis factor inhibitor.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the arm received placebo or a different active treatment).
• Source aliases: none (Leil 2021 identifies the arm’s drug by name in Table 2 and Supplementary Table S1).
• Example models: Leil_2021_rheumatoidArthritis_das28_mbma.R (Leil 2021 Table 2: 200 mg SC q2w or 400 mg SC q4w; 10 trials, 2,654 patients; selects lemax_certolizumab / let50_certolizumab of the DAS28 change-from-baseline Emax-in-time treatment effect).
• Notes: MBMA study-arm-level treatment indicator in the bare-INN per-drug arm-indicator family (siblings ETANERCEPT, NAPROXEN, TRAMADOL, …); see the ETANERCEPT entry Notes for the naming rule. Mutually exclusive with the other Leil 2021 drug indicators; all zero marks a background-therapy (placebo) arm.

RITUXIMAB (canonical for rituximab treatment-arm indicator)

• Description: Binary study-arm treatment indicator: 1 = the arm received rituximab, 0 = the arm received placebo or a different active treatment. A property of the trial arm in a model-based meta-analysis (MBMA), not of an individual patient. Drug class: an anti-CD20 B-cell-depleting antibody.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the arm received placebo or a different active treatment).
• Source aliases: none (Leil 2021 identifies the arm’s drug by name in Table 2 and Supplementary Table S1).
• Example models: Leil_2021_rheumatoidArthritis_das28_mbma.R (Leil 2021 Table 2: 1,000 mg IV infusion x 2 per year; 14 trials, 1,336 patients; selects lemax_rituximab / let50_rituximab of the DAS28 change-from-baseline Emax-in-time treatment effect).
• Notes: MBMA study-arm-level treatment indicator in the bare-INN per-drug arm-indicator family (siblings ETANERCEPT, NAPROXEN, TRAMADOL, …); see the ETANERCEPT entry Notes for the naming rule. Mutually exclusive with the other Leil 2021 drug indicators; all zero marks a background-therapy (placebo) arm.

TOCILIZUMAB (canonical for tocilizumab treatment-arm indicator)

• Description: Binary study-arm treatment indicator: 1 = the arm received tocilizumab, 0 = the arm received placebo or a different active treatment. A property of the trial arm in a model-based meta-analysis (MBMA), not of an individual patient. Drug class: an interleukin-6 receptor antagonist.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the arm received placebo or a different active treatment).
• Source aliases: none (Leil 2021 identifies the arm’s drug by name in Table 2 and Supplementary Table S1).
• Example models: Leil_2021_rheumatoidArthritis_das28_mbma.R (Leil 2021 Table 2: 8 mg/kg IV q4w; 18 trials, 3,724 patients; selects lemax_tocilizumab / let50_tocilizumab of the DAS28 change-from-baseline Emax-in-time treatment effect).
• Notes: MBMA study-arm-level treatment indicator in the bare-INN per-drug arm-indicator family (siblings ETANERCEPT, NAPROXEN, TRAMADOL, …); see the ETANERCEPT entry Notes for the naming rule. Mutually exclusive with the other Leil 2021 drug indicators; all zero marks a background-therapy (placebo) arm.

TOFACITINIB (canonical for tofacitinib treatment-arm indicator)

• Description: Binary study-arm treatment indicator: 1 = the arm received tofacitinib, 0 = the arm received placebo or a different active treatment. A property of the trial arm in a model-based meta-analysis (MBMA), not of an individual patient. Drug class: a Janus kinase inhibitor.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the arm received placebo or a different active treatment).
• Source aliases: none (Leil 2021 identifies the arm’s drug by name in Table 2 and Supplementary Table S1).
• Example models: Leil_2021_rheumatoidArthritis_das28_mbma.R (Leil 2021 Table 2: 5 mg PO bid; 10 trials, 1,611 patients; selects lemax_tofacitinib / let50_tofacitinib of the DAS28 change-from-baseline Emax-in-time treatment effect).
• Notes: MBMA study-arm-level treatment indicator in the bare-INN per-drug arm-indicator family (siblings ETANERCEPT, NAPROXEN, TRAMADOL, …); see the ETANERCEPT entry Notes for the naming rule. Mutually exclusive with the other Leil 2021 drug indicators; all zero marks a background-therapy (placebo) arm.

AZATHIOPRINE (canonical for azathioprine treatment-arm indicator)

• Description: Binary study-arm treatment indicator: 1 = the arm received azathioprine, 0 = the arm received placebo or a different active treatment. A property of the trial arm in a model-based meta-analysis (MBMA), not of an individual patient. Drug class: an immunosuppressant.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the arm received placebo or a different active treatment).
• Source aliases:
  • Azathioprine treatment – the Chan 2017 Table 2 row label. Free-text label, not a column header; downstream data sets should use the canonical column name AZATHIOPRINE.
• Example models: Chan_2017_ipf_fvc_mbma.R (Chan 2017 Table 1: 1 arm, 1 timepoint, 14 subjects, recorded as ‘3 mg’ (Raghu 1991); the Figure 4 forest plot gives the same arm as 3 mg/kg, which is the clinically standard azathioprine dosing and is almost certainly what the unqualified Table 1 entry abbreviates. Selects emax_azathioprine = 5.89 %predicted FVC, the second largest point estimate in the analysis but on 14 subjects with a CI of -4.14 to 15.9. The dose does not enter the model, so the discrepancy does not affect the prediction.).
• Notes: MBMA study-arm-level treatment indicator, sibling of NAPROXEN, TRAMADOL, TAPENTADOL, ACETAMINOPHEN and DICLOFENAC in the per-drug MBMA arm-indicator family. Specific scope. Named under the register’s explicit written rule, recorded in the TRT_BENRALIZUMAB entry Notes: use TRT_<INN> when the drug is the study treatment under investigation, CONMED_<INN> when it is background co-medication, and a bare INN only for MBMA trial-arm indicators. The register carries a competing precedent that upstream has not reconciled – the Sandra 2024 anti-HBV siRNA batch (Sandra_2024_siRNA_hbsag_mouse_mbma.R) registered 13 MBMA arm indicators as TRT_<drug>, citing the bare-INN family and then departing from it. A ruling in favour of TRT_<INN> should migrate the whole bare-INN family, now 19 members, rather than leave a one-off. In the founding model all 14 Chan 2017 indicators are mutually exclusive and all are 0 on a placebo arm, which is what makes the placebo-corrected output zero there by construction.

COTRIMOXAZOLE (canonical for co-trimoxazole treatment-arm indicator)

• Description: Binary study-arm treatment indicator: 1 = the arm received co-trimoxazole (the fixed combination of sulfamethoxazole and trimethoprim), 0 = the arm received placebo or a different active treatment. A property of the trial arm in a model-based meta-analysis (MBMA), not of an individual patient. Drug class: an antibiotic.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the arm received placebo or a different active treatment).
• Source aliases:
  • Co-trimoxazole treatment – the Chan 2017 Table 2 row label. Free-text label, not a column header; downstream data sets should use the canonical column name COTRIMOXAZOLE.
• Example models: Chan_2017_ipf_fvc_mbma.R (Chan 2017 Table 1: 1 arm, 1 timepoint, 95 subjects, 960 mg (Shulgina 2013). Selects emax_cotrimoxazole = 0.172 %predicted FVC. The 960 mg is the standard double-strength tablet (800 mg sulfamethoxazole plus 160 mg trimethoprim), so one indicator covers the fixed combination and no per-component split is needed; a future extraction that models the two components separately would need separate indicators.).
• Notes: MBMA study-arm-level treatment indicator, sibling of NAPROXEN, TRAMADOL, TAPENTADOL, ACETAMINOPHEN and DICLOFENAC in the per-drug MBMA arm-indicator family. Specific scope. Named under the register’s explicit written rule, recorded in the TRT_BENRALIZUMAB entry Notes: use TRT_<INN> when the drug is the study treatment under investigation, CONMED_<INN> when it is background co-medication, and a bare INN only for MBMA trial-arm indicators. The register carries a competing precedent that upstream has not reconciled – the Sandra 2024 anti-HBV siRNA batch (Sandra_2024_siRNA_hbsag_mouse_mbma.R) registered 13 MBMA arm indicators as TRT_<drug>, citing the bare-INN family and then departing from it. A ruling in favour of TRT_<INN> should migrate the whole bare-INN family, now 19 members, rather than leave a one-off. In the founding model all 14 Chan 2017 indicators are mutually exclusive and all are 0 on a placebo arm, which is what makes the placebo-corrected output zero there by construction.

TRT_EPHEDRINE (canonical for ephedrine treatment-arm indicator)

• Description: Binary treatment-arm indicator: 1 = the subject received ephedrine, 0 = the subject did not. Unlike the NAPROXEN / TRAMADOL / TAPENTADOL MBMA arm indicators, this is a property of an individual patient in a two-armed observational study, not of a trial arm in a meta-analysis.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0. The reference differs by model and must be documented per model – in Dings_2026_cafedrine_theodrenaline_ephedrine.R the reference is the cafedrine/theodrenaline arm (every subject received one drug or the other), whereas in Dings_2026_neonatal_acidosis.R the reference is untreated and TRT_CAFEDRINE_THEODRENALINE is carried as a separate indicator alongside it.
• Source aliases:
  • Treatment with E / Treatment with ephedrine – Dings 2026 Tables A1, A3 and A4 row labels.
• Example models: Dings_2026_cafedrine_theodrenaline_ephedrine.R (raises the maximum attainable heart rate by a fraction of 0.15 and lowers all three EC50 values by a fraction of 0.711, both via the paper’s 1 + COV * theta categorical form), Dings_2026_neonatal_acidosis.R (+2.64 mmol/L base excess and +5.46 mg/dL lactate relative to untreated).
• Notes: Specific scope (a per-drug arm indicator). The TRT_ prefix distinguishes a patient-level randomised / assigned treatment indicator from the bare-INN MBMA arm indicators (NAPROXEN, TRAMADOL, TAPENTADOL) and from the CONMED_<INN> concomitant-medication family; use TRT_<INN> when the drug is the study treatment under investigation. Pairs with TRT_CAFEDRINE_THEODRENALINE; the two are mutually exclusive per subject.

TRT_CAFEDRINE_THEODRENALINE (canonical for cafedrine/theodrenaline treatment-arm indicator)

• Description: Binary treatment-arm indicator: 1 = the subject received the 20:1 fixed combination of cafedrine and theodrenaline, 0 = the subject did not. A patient-level indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (subject did not receive cafedrine/theodrenaline). See the reference-category caveat on TRT_EPHEDRINE.
• Source aliases:
  • Treatment with C/T – Dings 2026 Table A3 row label.
• Example models: Dings_2026_neonatal_acidosis.R (+3.26 mmol/L neonatal base excess relative to untreated newborns – the larger of the two treatment effects, which is the paper’s central comparative finding).
• Notes: Specific scope. A single canonical for the fixed-dose combination rather than two separate indicators, because the two components are never given apart – cafedrine and theodrenaline are co-formulated at a fixed 20:1 ratio and doses are expressed as cafedrine equivalents. Do not split into TRT_CAFEDRINE and TRT_THEODRENALINE: no source reports them independently and the model cannot identify separate effects. Sibling of TRT_EPHEDRINE; mutually exclusive with it.

TRT_TAMSULOSIN, TRT_DUTASTERIDE, TRT_TAMSULOSIN_DUTASTERIDE, TRT_WATCHFUL_WAITING (canonical for BPH treatment-arm indicators)

• Description: Four mutually exclusive binary patient-level treatment-arm indicators for benign prostatic hyperplasia trials. TRT_TAMSULOSIN = 1 when the patient was randomised to tamsulosin (alpha-1 blocker) monotherapy; TRT_DUTASTERIDE = 1 for dutasteride (5-alpha-reductase inhibitor) monotherapy; TRT_TAMSULOSIN_DUTASTERIDE = 1 for tamsulosin + dutasteride combination therapy, free or fixed-dose; TRT_WATCHFUL_WAITING = 1 for watchful waiting (conservative management, with or without protocol-defined later initiation of drug therapy). All four 0 = placebo.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0; all four 0 = the placebo arm. In DAgate_2020_bph_ipss_mbma.R each indicator selects its own estimated treatment effect, its own IIV and its own residual SD, and the placebo arm additionally carries IIV on the disease-progression rate and the placebo magnitude. In DAgate_2021_bph_aurs_mbma.R each drug indicator selects its own log hazard ratio on the placebo hazard; watchful waiting has no estimate there, so a watchful-waiting patient carries the placebo hazard.
• Source aliases:
  • ARM – the D’Agate 2020 control stream column: 0 = placebo, 1 or 2 = watchful waiting (pooled), 3 = tamsulosin, 4 = dutasteride, 5 = combination (‘FDC’).
• Example models: DAgate_2020_bph_ipss_mbma.R (founding example; constant first-order effects on IPSS of 0.015 /month for tamsulosin, 0.016 /month for dutasteride, 0.032 /month for the combination and 0.018 /month for watchful waiting, each added to the decaying placebo effect), DAgate_2021_bph_aurs_mbma.R (log hazard ratios on the time to first acute urinary retention or BPH-related surgery: dutasteride 0.432, combination 0.336, tamsulosin fixed to 1; watchful waiting had no estimate and is listed under covariatesDataExcluded).
• Notes: Follows the register’s written TRT_<INN> rule (TRT_EPHEDRINE, TRT_BENRALIZUMAB Notes): the drug is the study treatment under investigation and the indicator is a property of an individual patient (individual-patient-data meta-analysis), not a published trial arm, so the bare-INN MBMA arm family does not apply. The combination is ONE indicator rather than TRT_TAMSULOSIN = TRT_DUTASTERIDE = 1, on the TRT_CAFEDRINE_THEODRENALINE precedent, because the source estimates a separate combination effect that is not the sum of the monotherapy effects. TRT_WATCHFUL_WAITING names a management strategy rather than an INN, on the TRT_PBT precedent.

TRT_IPSOS_CONTROL (canonical for IPSOS-trial control-arm indicator)

• Description: Binary trial-arm indicator: 1 = the study arm is the CONTROL arm of the phase III IPSOS trial (NCT03191786, reported as Lee 2023), in which investigators chose single-agent gemcitabine or vinorelbine for treatment-naive advanced NSCLC patients unsuitable for platinum-doublet chemotherapy; 0 = the arm is one of the historical single-agent-chemotherapy control arms drawn from the published literature.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (a historical published single-agent-chemotherapy control arm). This is the reference treatment against which the IPSOS control arm’s hazard ratio is expressed.
• Source aliases:
  • IPSOS control – Chen 2026 Tables 3, 4 and S7 Covariate column, and the IPSOSHR results column.
  • ID == 15 – the trial selector in the Chen 2026 NONMEM dataset psp470197-sup-0001-Supinfo1.csv and the IF (ID.EQ.15) branch of the distributed $PRED block, evaluated with the atezolizumab arm excluded.
• Example models: Chen_2026_nsclc_os_mbma.R (coefficient -0.633 on the overall-survival log hazard ratio, i.e. HR 0.531 [0.424, 0.664] for the IPSOS control arm relative to historical controls after adjusting for ECOG performance status).
• Notes: MBMA trial-arm indicator, not a patient characteristic – it identifies which published cohort an arm came from, in the same spirit as the STUDY_<id> family, but is filed under TRT_ because the quantity it modifies is a treatment-effect hazard ratio rather than a PK parameter of a study cohort. Member of the TRT_<arm> treatment-arm-indicator family (TRT_EPHEDRINE, TRT_PBT, TRT_PDFVIII_VWF, …). Do not set this flag on an IPSOS atezolizumab arm. Chen 2026’s final model (Model 010) is fitted with the atezolizumab arm excluded, so within that model the IPSOS trial identifier and the IPSOS control arm coincide; the atezolizumab treatment effect belongs to the paper’s Model 041, whose reference survival curve is not published. A future extraction that needs the atezolizumab arm should add a separate TRT_IPSOS_ATEZOLIZUMAB indicator rather than overloading this one. Founding example: Chen 2026.

TRT_PBT (canonical for plasma-based-therapy treatment-arm indicator)

• Description: Binary treatment-arm indicator: 1 = the infusion is a plasma-based therapy (PBT) – fresh frozen plasma, solvent/detergent-treated plasma, or an equivalent whole-plasma infusion – 0 = it is not. Used in ADAMTS13 / coagulation-factor replacement models where plasma infusion is the comparator against a recombinant or plasma-derived concentrate, and the products differ in how much of the active factor they deliver per administered unit.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0. The reference differs by model and must be documented per model – in Patel_2025_radamts13.R the reference is the recombinant product (rADAMTS13), whose relative activity is fixed to 1, and TRT_PDFVIII_VWF is carried as a separate mutually-exclusive indicator alongside this one.
• Source aliases:
  • PBT – Patel 2025 Table 1 row label “x (1-0.390) if PBT”; also the abbreviation used throughout that paper’s text and figures.
• Example models: Patel_2025_radamts13.R (multiplies the delivered dose by a relative-ADAMTS13-activity factor of 1 - 0.390, i.e. plasma infusion delivers 39.0% less ADAMTS13 activity per administered IU than rADAMTS13; enters as relative bioavailability f(central), not as an effect on clearance or volume).
• Notes: Specific scope. Unlike TRT_EPHEDRINE / TRT_CAFEDRINE_THEODRENALINE, which are patient-level indicators, this is a per-dose-record covariate: the cTTP pivotal trial is a crossover in which each patient receives the recombinant product in one period and PBT in another, so the indicator changes within a subject. Deliberately kept separate from TRT_PDFVIII_VWF rather than collapsed into one “non-recombinant” indicator, because the two product classes differ by more than a factor of nine in delivered activity (39.0% vs 93.3% lower); a source that pools all plasma-derived products into a single arm should use this canonical and say so in its notes. Distinct from FORM_* (a formulation of the same drug) – these are different products with different active-factor content, which is why the TRT_ prefix applies. Founding example: Patel 2025.

TRT_PDFVIII_VWF (canonical for plasma-derived FVIII/VWF concentrate treatment-arm indicator)

• Description: Binary treatment-arm indicator: 1 = the infusion is a plasma-derived factor VIII / von Willebrand factor concentrate (pdFVIII:VWF), 0 = it is not. These concentrates are given in congenital thrombotic thrombocytopenic purpura (cTTP) and in von Willebrand disease as a source of VWF and, incidentally, of the ADAMTS13 that co-purifies with it; the ADAMTS13 content is far lower per administered unit than in either plasma or a recombinant product.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0. Reference differs by model and must be documented per model – in Patel_2025_radamts13.R the reference is rADAMTS13 and this indicator is mutually exclusive with TRT_PBT.
• Source aliases:
  • FVIII:VWF concentrates – Patel 2025 Table 1 row label “x (1-0.933) if FVIII:VWF concentrates”.
  • pdFVIII:VWF – the abbreviation used in that paper’s text and in Figure 3.
• Example models: Patel_2025_radamts13.R (multiplies the delivered dose by a relative-ADAMTS13-activity factor of 1 - 0.933, i.e. pdFVIII:VWF concentrates deliver 93.3% less ADAMTS13 activity per administered IU than rADAMTS13).
• Notes: Specific scope. Per-dose-record, for the same crossover reason given under TRT_PBT; mutually exclusive with it. Orthogonal to VWF (a measured plasma von Willebrand factor concentration used as a covariate on FVIII clearance) and to FVIIIRECENT (a measured FVIII:C activity) – this canonical identifies which product was administered, not a measured analyte concentration. Founding example: Patel 2025.

TRT_ANTIPCSK9_SIRNA (canonical for anti-PCSK9 siRNA treatment-modality indicator)

• Description: Binary treatment-modality indicator in a model that carries both anti-PCSK9 modalities in one ODE system: 1 = the arm received an anti-PCSK9 short interfering RNA (inclisiran or ALN-PCS), 0 = the arm received an anti-PCSK9 monoclonal antibody (alirocumab, evolocumab or RG-7652). It selects which of two separately fitted values of a shared structural parameter applies to the arm, rather than switching a drug-specific potency on or off.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (monoclonal-antibody arm).
• Source aliases:
  • mAbs (alirocumab and evolocumab) vs siRNA (inclisiran and ALN-PCS) – Sokolov 2019 Methods, “Summary of model development”, step 2.
  • two previously identified sets of parameters – Sokolov 2019 Results, “PCSK9-LDL-C relationship”.
• Example models: Sokolov_2019_antipcsk9_qsp.R (selects the PCSK9-on-LDL-receptor power n1 between the mAb value 0.14 and the siRNA value 0.26; every other one of the model’s 51 parameters is shared across the two modalities).
• Notes: Specific scope. A class-level member of the TRT_<arm> treatment-arm-indicator family, following TRT_PBT (plasma-based therapy) and TRT_PDFVIII_VWF (plasma-derived concentrate class) rather than the per-INN members – the contrast Sokolov 2019 fitted is between pharmacological modalities, not between named drugs, and all three mAbs share one n1 while both siRNAs share the other. Do not decompose this into five per-drug TRT_<INN> indicators: the source reports only two fitted values and a per-drug encoding would imply five identifiable ones. Distinct from a dose or exposure column – it carries no potency information, and the drug actually administered is resolved by which depot compartment the dose event targets. A future anti-PCSK9 model that needs to separate the three mAbs (for example, to give RG-7652 its own LDL-receptor parameters) should register per-drug siblings alongside this entry rather than overload it. Founding example: Sokolov 2019.

TRT_OLX703A, TRT_AB729, TRT_ALG125755, TRT_ALG125819, TRT_VIR2218, TRT_ARB1740, TRT_ARB1467, TRT_ARC520, TRT_SIHBV74, TRT_SIHBV75, TRT_SIHBV76, TRT_SIHBV77, TRT_SIHBV74_77 (canonical for anti-HBV short-interfering-RNA treatment-arm indicators)

• Description: Mutually exclusive binary treatment-arm indicators naming which anti-hepatitis-B short interfering RNA (siRNA) a study arm received. Exactly one of the thirteen is 1 on any arm; all thirteen are 0 only on a placebo / vehicle arm. Members of the auto-approved TRT_<INN-or-development-code> family. Grouped in one entry because they are a single mutually exclusive set drawn from one source paper, and because the model that consumes them selects one drug-specific potency per arm from exactly this list. The thirteen split across three delivery classes, which is what makes the set worth keeping intact: GalNAc-conjugated – TRT_OLX703A (OLX703A), TRT_AB729 (AB-729 / imdusiran), TRT_ALG125755 (ALG-125755), TRT_ALG125819 (ALG-125819), TRT_VIR2218 (VIR-2218 / elebsiran); lipid-nanoparticle formulated – TRT_ARB1740 (ARB-1740), TRT_ARB1467 (ARB-1467); cholesterol-conjugated, all co-administered with the GalNAc-conjugated melittin-like endolytic peptide NAG-MLP – TRT_ARC520 (ARC-520), TRT_SIHBV74 / TRT_SIHBV75 / TRT_SIHBV76 / TRT_SIHBV77 (NAG-MLP chol-siHBV-74 / -75 / -76 / -77) and TRT_SIHBV74_77 (the fixed 74 + 77 co-formulation).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (this arm received a different siRNA, or placebo / vehicle). Because the set is mutually exclusive, the model reference is “no siRNA”, not any one of the thirteen.
• Source aliases:
  • OLX703A, AB-729, ALG-125755, ALG-125819, VIR-2218, ARB-1740, ARB-1467, ARC-520, NAG-MLP chol-siHBV-74, NAG-MLP chol-siHBV-75, NAG-MLP chol-siHBV-76, NAG-MLP chol-siHBV-77, NAG-MLP chol-siHBV-74 + 77 – Sandra 2024 Tables 1 and 3 row labels.
• Example models: Sandra_2024_siRNA_hbsag_mouse_mbma.R (each indicator selects that siRNA’s lid50, and – via the delivery class derived from it inside model() – the class-specific lkel; three of them, TRT_AB729, TRT_ARB1740 and TRT_SIHBV77, additionally select an estimated Hill exponent instead of the value fixed to 1).
• Notes: MBMA study-arm-level treatment indicators (properties of a published trial arm, not of an individual animal), following the FLARE / NAPROXEN / TRAMADOL MBMA-arm precedent; the TRT_ prefix marks the drug as the study treatment under investigation rather than concomitant medication (CONMED_<INN>). Specific scope: these are per-drug arm indicators tied to the anti-HBV siRNA development programmes. Development codes are used verbatim because most of these molecules had no INN at the time of the source studies; where an INN exists it is recorded above so a future extraction can find the column by either name. Non-alphanumeric characters in the source labels are dropped (AB-729 becomes TRT_AB729); the combination arm keeps an underscore between its two components (TRT_SIHBV74_77). Do not encode the 74 + 77 combination by setting both TRT_SIHBV74 and TRT_SIHBV77 to 1 – Sandra 2024 fits the co-formulation as a single entity with its own ID50 (41.46 ng, the most potent in the analysis), not as a combination of two separately parameterised siRNAs. A future paper reporting one of these siRNAs in a clinical population should reuse the same column rather than adding a species-qualified variant.

TRT_PINATUZUMAB_VEDOTIN, TRT_POLATUZUMAB_VEDOTIN (canonical for MMAE antibody-drug-conjugate treatment-arm indicators)

• Description: Mutually exclusive binary indicators naming which monomethyl-auristatin-E (MMAE) antibody-drug conjugate a subject received: TRT_PINATUZUMAB_VEDOTIN = 1 for pinatuzumab vedotin (anti-CD22, DCDT2980S) and TRT_POLATUZUMAB_VEDOTIN = 1 for polatuzumab vedotin (anti-CD79b, DCDS4501A). Exactly one of the two is 1 on any record. Patient-level, not trial-arm-level: both ADCs were studied in single-agent and rituximab-combination cohorts of separate phase I/II trials whose data were pooled into one joint fit. Members of the auto-approved TRT_<INN> family.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the subject received the other ADC). Because the pair is mutually exclusive and the source model carries no drug-independent reference, neither indicator has a meaningful “untreated” reference; a model consuming them selects one drug’s parameter block rather than applying an offset to a reference drug.
• Source aliases:
  • MOL – Lu 2016 supplement S1 $INPUT column, an integer taking 1 (pinatuzumab vedotin) or 2 (polatuzumab vedotin). The control stream selects parameters with the arithmetic switch THETA(i)*(2-MOL) + THETA(j)*(MOL-1), so TRT_PINATUZUMAB_VEDOTIN = 2 - MOL and TRT_POLATUZUMAB_VEDOTIN = MOL - 1.
  • molecule indicator – Lu 2016 Methods prose.
• Example models: Lu_2016_pinatuzumab_polatuzumab.R (each indicator selects that ADC’s clearance, central volume, deconjugation rate and acMMAE assay cross-calibration slope, while the intercompartmental clearance, peripheral volume, all five IIV variances and both residual errors are shared between the two drugs).
• Notes: Kept as two symmetric indicators rather than one, following the TRT_OLX703A … TRT_SIHBV74_77 precedent for a mutually exclusive drug set, because the source estimates a parallel parameter block per drug rather than a reference plus a deviation; the model file’s parameters carry matching _pina / _pola stratum suffixes. The TRT_ prefix marks the ADC as the study treatment under investigation, distinguishing these from CONMED_RITUXIMAB-style concomitant medication – rituximab was co-administered in some cohorts of both programmes, and Lu 2016 reports it does not affect ADC PK, so it is not carried as a covariate here. Do not reuse these columns for a different MMAE ADC: the two names are drug-specific, and a third MMAE conjugate needs its own TRT_<INN> column even though Lu 2016 argues the structural model generalises across the MMAE ADC platform.

DOSE_BUPIVACAINE_MG (canonical for administered intrathecal bupivacaine dose)

• Description: Total dose of bupivacaine administered for the neuraxial block, in mg. Determines the extent and duration of the sympathetic block and hence the depth of the resulting hypotension. Member of the auto-approved DOSE_<drug>_<units> family.
• Units: mg
• Type: continuous
• Scope: general
• Reference category: n/a – linear regression term. Reference values observed: cohort means 10.5 mg (cafedrine/theodrenaline) and 10.2 mg (ephedrine) in Dings 2026 Table 2.
• Source aliases:
  • Bupivacaine amount – Dings 2026 Tables A3 / A4 row label.
• Example models: Dings_2026_neonatal_acidosis.R (-0.186 mmol/L base excess and +1.05 mg/dL lactate per mg of maternal bupivacaine).
• Notes: General scope – bupivacaine is used for neuraxial and peripheral blocks across many surgical populations.

MAXHR_KPD (canonical for individual maximum attainable heart rate from an upstream K/PD fit)

• Description: A subject’s individual (empirical-Bayes) estimate of the maximum attainable heart rate parameter from an upstream haemodynamic K/PD model, supplied to a downstream model as a covariate. A model-derived quantity, not a measurement.
• Units: beats/min
• Type: continuous
• Scope: specific
• Reference category: n/a – linear regression term. Reference values observed: population typical values 77.8 beats/min (cafedrine/theodrenaline) and 89.5 beats/min (ephedrine) in Dings 2026 Table A1.
• Source aliases:
  • Model parameter MAXHR – Dings 2026 Table A3 row label.
• Example models: Dings_2026_neonatal_acidosis.R (-0.0256 mmol/L neonatal base excess per beat/min – the quantitative link behind the paper’s conclusion that a lower maternal peak heart rate benefits the newborn).
• Notes: Specific scope – the value is only meaningful as the rmax_hr parameter of the paired Dings_2026_cafedrine_theodrenaline_ephedrine.R model, and a downstream model consuming it must obtain it from that model rather than from an unrelated fit. Same upstream-posthoc-input pattern as BAS_SVEGFR3, IBASE and MBL_END.

KEL_KPD (canonical for individual K/PD elimination rate constant from an upstream fit)

• Description: A subject’s individual (empirical-Bayes) estimate of the elimination rate constant of an upstream kinetic/pharmacodynamic (K/PD) model, supplied to a downstream model as a covariate. In a K/PD framework this is a functional descriptor of how quickly the drug effect dissipates, not a systemic pharmacokinetic elimination rate – no drug concentrations were measured.
• Units: 1/min
• Type: continuous
• Scope: specific
• Reference category: n/a – linear regression term. Reference values observed: population typical value 0.0971/min (Dings 2026 Table A1, CL = 0.0971 L/min with V fixed to 1 L), corresponding to a 7.14 min effect half-life.
• Source aliases:
  • Model parameter kel – Dings 2026 Table A3 row label.
• Example models: Dings_2026_neonatal_acidosis.R (-1.74 mmol/L neonatal base excess per 1/min; a slower effect decay – a smaller kel – predicts better neonatal acid-base status).
• Notes: Specific scope, for the same reason as MAXHR_KPD. The founding paper’s Discussion cautions explicitly against a causal reading of this term, since within a K/PD framework the parameter may absorb unmeasured determinants of response duration; a model consuming this covariate should carry that caveat. The units in the founding source’s Table A3 are printed as mmol/L x min, which is the reciprocal of the per-(1/min) coefficient unit the regression actually implies – see the Errata section of the paired vignette.

QBL, QEFF (canonical for baseline / effective renal Q in HD/CRRT renal-replacement models)

• Description: Renal-replacement-therapy clearance terms used in hemodialysis (HD) / continuous renal replacement therapy (CRRT) popPK models for renally cleared drugs.
  • QBL = baseline (off-HD / off-CRRT) renal Q (L/h); the typical patient’s residual renal clearance.
  • QEFF = effective on-HD/CRRT renal Q (L/h); the apparent dialysis-augmented clearance during a session (including the effective renal Q plus the dialyser/filter clearance, depending on the paper’s parameterisation).
• Units: L/h
• Type: continuous
• Scope: specific
• Reference category: none (continuous).
• Source aliases: none.
• Example models: LeuppiTaegtmeyer_2019_colistin.R (colistin / colistimethate sodium popPK in CRRT recipients; QBL drives the off-CRRT typical-value CL, QEFF drives the on-CRRT augmented CL during the session window), Zhang_2025_fluconazole.R (QEFF is the per-record fluconazole CRRT clearance CLcrrt, supplied as a data column rather than estimated, and removes drug from the paper-specific crrt compartment at QEFF * RRT_CRRT_ACTIVE / vcrrt; computed from CRRT mode and circuit flows as (Q_uf + Q_d) * S, equivalently dose [mL/kg/h] * WT [kg] * 0.67 / 1000 at the cohort sieving/saturation coefficient – Zhang 2025 Supplementary Material Section 1. This model uses QEFF without a companion QBL: residual body clearance is a separately estimated structural parameter stratified by DIS_ARF, not a covariate), Peng_2025_meropenem.R (the same QEFF-without-QBL shape as Zhang 2025, in the simplest possible form: Peng 2025 Eq. 1 is CL_total = CL_body * exp(theta_CLCR * (CLCR - 13.6)) * exp(eta) + CL_CRRT, so QEFF is added directly to the single estimated body-clearance arm with no compartment of its own and no on/off gate – every subject in the cohort was on CVVH continuously. Computed as Q_uf [mL/h] * Sc / 1000 from the per-subject ultrafiltrate flow and the measured effluent-to-plasma meropenem sieving coefficient; cohort medians Q_uf 2477.5 mL/h and Sc 0.75. The paper’s own dosing simulations standardise this to 25 mL/h/kg * 65 kg * 0.75 / 1000 = 1.219 L/h, which reproduces all 24 cells of its Table 3 toxicity-probability grid to within 1.3 percentage points RMSE and is therefore the value to use when replicating them), He_2022_gentamicin.R (the same QEFF-without-QBL shape from the same Changsha group: He 2022 writes the individual clearance as CLbody*EXP(ETA(1)) + CLCRRT, a one-compartment gentamicin model in 14 CVVH / CVVHDF patients with no covariate on the 1.20 L/h body arm. Computed from the circuit settings by Equations 2-8; every Table 1 value is reproduced by the post-dilution form CRRT dose [mL/kg/h] * WT [kg] * Sc-or-Sd / 1000. The paper’s simulations fix Sd at 0.77 and set the CRRT dose to 30 / 40 / 50 mL/kg/h, i.e. QEFF = 0.77 * dose * WT / 1000 per virtual patient), Por_2021_imipenem.R (the same added-after-the-eta, ungated shape as Peng 2025, cl <- cl_body + QEFF per Por 2021 Equations 5 and 11, in a cohort where only 12 of 23 burn patients were on CVVH: QEFF is 0 for the others. It is deliberately not gated by RRT_CRRT_STATUS, which selects the body-clearance branch, because the paper’s Figure 5 adds CVVH clearance to the CrCl-driven no-CVVH branch. The per-patient haemofilter clearance is computed as Qf * Sc * CF from the ultrafiltrate flow, the measured sieving coefficient C_filter / ((C_pre + C_post) / 2) and the prefilter-replacement correction Qb / (Qb + Qrep) (Equations 1-3); cohort mean 1.56 +/- 0.7 L/h (Table 1). The paper’s simulations use 3 and 5 L/h (Figure 5) and 3.27 L/h (Figure 3)), Soraluce_2020_linezolid.R (two-compartment linezolid popPK in critically ill adults, 23 of 40 on CVVHDF or CVVHD. The QEFF-without-QBL shape again, but here QEFF sits INSIDE the random effect: Table 3 footnote b is CL = (CLNR + CLR * (Clcr / 44) + Sc * Qef) * exp(eta), so the per-subject extracorporeal clearance is summed with the estimated non-renal arm and the measured-creatinine-clearance renal arm before the exponential IIV scales the total. QEFF = Sc * Qef is set to 0 for subjects not on CRRT, with no separate on/off indicator; Table 1 median 2.51 L/h, range 0.79-3.09), Selig_2022_piperacillin_mbma.R, Selig_2022_tazobactam_mbma.R (aggregate-data one-compartment models fitted to arm-mean curves digitised from published CKRT studies plus individual CVVH patients, from the same group as Por 2021 and with the same Qf * Sc * CF derivation (Selig 2022 Equations 1-3). The Por 2021 shape, cl <- exp(lcl + eta_study_lcl) + QEFF, ungated: QEFF is the per-arm CKRT clearance (the CLCKRT column of the supplementary dataset; piperacillin arms 0.288-3 L/h, tazobactam 0.56-1.4 L/h), taken from each study or assumed to be 70% of the reported total CKRT dose. The additive placement is not printed as an equation; refitting the posted dataset with and without it, and the closed-form reproduction of the continuous-infusion panel of Figure 3 at QEFF = 0-3 L/h, both select it).
• Notes: Specific scope because the on-/off- HD/CRRT splitting is paper-specific. Future renal-replacement popPK extractions should reuse the QBL/QEFF pair. Note that the pair need not be used together: QEFF alone is correct when the paper supplies the extracorporeal clearance per record and estimates residual body clearance as a structural parameter (Zhang 2025), whereas the QBL/QEFF pair is used when both the off- and on-therapy renal Q are covariate-supplied (Leuppi-Taegtmeyer 2019).

DIS_CHD_PERCENT (canonical for coronary-heart-disease cohort prevalence percentage)

• Description: Study-arm-level percentage (0-100) of the enrolled cohort who carry a coronary heart disease (CHD) diagnosis. Continuous covariate scaled in percent (not fraction).
• Units: %
• Type: continuous
• Scope: specific
• Reference category: 0% (healthy non-CHD cohort).
• Source aliases: CHD_PCT legacy form (used in Vargo 2014 statins / ezetimibe MBMA pre-rename); alias of the canonical DIS_CHD_PERCENT form rename.
• Example models: Vargo_2014_statins_ezetimibe_mbma.R (linear coefficient e_chd_emax_statin = -0.000649 per percentage point on Emax_statin, i.e. a 24% CHD arm reduces Emax_statin by 0.000649 * 24 = 0.016; the paper’s typical-patient definition uses 24% CHD).
• Notes: MBMA study-arm-level covariate; the canonical register’s individual-level pop-PK covariates do not directly fit aggregate-percentage columns, so this canonical is specific-scope and explicitly carries a study-arm aggregate meaning. Future MBMA models should reuse for the CHD-cohort prevalence column.

DIS_HYPERT_PERCENT (canonical for hypertension cohort prevalence percentage)

• Description: Study- or trial-level percentage (0-100) of the enrolled cohort carrying a hypertension diagnosis at baseline. Continuous covariate scaled in percent (not fraction). The aggregate counterpart of the individual-level binary DIS_HYPERT.
• Units: %
• Type: continuous
• Scope: specific
• Reference category: 0% (no hypertensive patients enrolled).
• Source aliases:
  • Hypertension – Volkova 2023 Table 1 covariate label.
  • Hypertension (%) – Volkova 2023 Supplementary Table 3 row label.
• Example models: Volkova_2023_lipidLowering_mace_mbma.R (linear coefficient +0.011 per percentage point on the log risk ratio of ischemic stroke, so a 75%-hypertensive trial population sees exp(0.011 * 50) = 1.73-fold less ischemic-stroke benefit from lipid-lowering therapy than a 25%-hypertensive one).
• Notes: MBMA study-arm / trial-level covariate. Family precedent: DIS_CHD_PERCENT (Vargo 2014). Distinct from the per-subject binary DIS_HYPERT (an individual medical-history flag) and from continuous SBP / DBP blood-pressure measurements. The percent-vs-fraction scale is load-bearing because the coefficient is a per-percentage-point slope; confirm it against a published contrast before use.

DIS_CHB_PERCENT (canonical for hepatitis-B-positive cohort prevalence percentage)

• Description: Study-arm- or trial-level percentage (0-100) of the enrolled cohort who were hepatitis B virus positive at baseline. Continuous covariate scaled in percent, not as a fraction. The aggregate counterpart of the individual-level binary DIS_CHB.
• Units: %
• Type: continuous
• Scope: specific
• Reference category: 0% (no hepatitis-B-positive patients enrolled). Note that consuming models usually enter this covariate MEDIAN-CENTRED, in which case the model’s own centring constant – not 0% – is the value at which the intercept is quoted.
• Source aliases:
  • HBV – Zierhut 2016 Results “Final model” equation and Table 3 (Table 3’s footnote defines it as “median-centered percent of hepatitis B-positive patients”).
  • Percent with HBV – Zierhut 2016 Table 2 row label.
• Example models: Zierhut_2016_hcc_antiangiogenic_os_mbma.R (linear coefficient -0.00418 per percentage point on ln(median overall survival), entered centred at the analysis-set median of 34.0%, so each additional percentage point of hepatitis-B positivity costs exp(-0.00418) - 1 = -0.417% of median overall survival in advanced hepatocellular carcinoma).
• Notes: MBMA study-arm / trial-level covariate. Family precedent: DIS_CHD_PERCENT (Vargo 2014) and DIS_HYPERT_PERCENT (Volkova 2023), of which this is the direct analogue – an aggregate prevalence column whose individual-level sibling is already registered. The percent-vs-fraction scale is load-bearing because the coefficient is a per-percentage-point slope; confirm it against a published contrast before use (in the founding model the paper’s own clinical-trial-simulation medians reproduce to three digits only on the percent scale with 34.0 subtracted). Distinct from the per-subject binary DIS_CHB, and orthogonal to HBSAG_BL_LOG10 / HBEAG_POS, which quantify serologic state within an infected cohort rather than counting how much of an arm is infected. In hepatocellular-carcinoma meta-analyses this column is a disease-aetiology descriptor (hepatitis B as the cause of the cancer) rather than a co-infection flag, and is frequently IMPUTED where a source publication did not report it – record the imputation provenance in the model’s covariateData notes.

TUMTP_SQUAM_PCT (canonical for squamous-tumor-histology cohort prevalence percentage)

• Description: Study-arm-level percentage (0-100) of the enrolled cohort whose tumor histology is squamous (versus non-squamous). Continuous covariate scaled in percent (not fraction). In non-small cell lung cancer (NSCLC) MBMA papers the arm’s squamous fraction interacts with the chemotherapy treatment class (chemotherapy tends to be more effective on squamous NSCLC than on non-squamous NSCLC).
• Units: %
• Type: continuous
• Scope: specific
• Reference category: 0% (all-non-squamous arm).
• Source aliases: %squamous histology (Franzese 2026 Table 1 / Table S1 covariate label).
• Example models: Franzese_2026_pdl1_nsclc_mbma.R (linear coefficients on ORR chemotherapy intercept (+0.282), OS chemotherapy hazard (+0.213), and PFS chemotherapy hazard (+0.186) in a mNSCLC MBMA of PD-(L)1 immunotherapy).
• Notes: MBMA study-arm-level covariate. Family precedent: DIS_CHD_PERCENT (Vargo 2014). Distinct from a per-subject binary squamous-histology indicator; here the arm’s squamous fraction is a continuous proportion of participants.

TUMTP_FL_PCT (canonical for follicular-lymphoma cohort prevalence percentage)

• Description: Study-arm-level percentage (0-100) of the enrolled cohort whose non-Hodgkin-lymphoma (NHL) histology is follicular lymphoma (FL). Continuous covariate scaled in percent (not fraction). The aggregate counterpart of the individual-level binary TUMTP_FL.
• Units: %
• Type: continuous
• Scope: specific
• Reference category: 0% (no FL patients). In a model carrying the whole NHL-histology _PCT set, mantle cell lymphoma is the natural reference level and is represented by leaving TUMTP_FL_PCT, TUMTP_DLBCL_PCT and TUMTP_OTHER_NHL_PCT all at 0.
• Source aliases:
  • Percentage of patients with each NHL subtype / FL – Li 2017 Methods and Table 1 NHL subtype block; the Table S1 coefficient row is Follicular lymphoma on lambda.
• Example models: Li_2017_nhl_pfs_mbma.R (log-linear coefficient +0.46 on the Weibull scale lambda of progression-free survival, i.e. an all-FL cohort has exp(0.46) = 1.58-fold the median PFS time of an all-MCL cohort).
• Notes: MBMA study-arm-level covariate; sibling of TUMTP_SQUAM_PCT, from which the naming pattern is taken unchanged. Family precedent: DIS_CHD_PERCENT (Vargo 2014). A continuous prevalence rather than a per-cohort histology label because meta-analytic databases contain MIXED-histology arms – 25 of the 155 cohorts in Li 2017 (16.1%) are classified Mixed and take intermediate values in every member of this set. The percent-versus-fraction scale is load-bearing: sources routinely tabulate these columns as fractions in 0-1 (Li 2017 Table 1 prints its sibling aggregate covariates that way) while the canonical is in percent, and consuming models fit on the fraction scale must divide by 100. Supplying a fraction where a percent is expected shrinks the effect by a factor of 100 and is invisible in any summary statistic of the column itself, so verify against a published contrast (for Li 2017, the median-PFS-time ratios in Figure 3B) before use.

TUMTP_DLBCL_PCT (canonical for diffuse-large-B-cell-lymphoma cohort prevalence percentage)

• Description: Study-arm-level percentage (0-100) of the enrolled cohort whose NHL histology is diffuse large B-cell lymphoma (DLBCL). Continuous covariate scaled in percent (not fraction). The aggregate counterpart of the individual-level binary TUMTP_DLBCL.
• Units: %
• Type: continuous
• Scope: specific
• Reference category: 0% (no DLBCL patients); see TUMTP_FL_PCT for the shared MCL reference level.
• Source aliases:
  • Percentage of patients with each NHL subtype / DLBCL – Li 2017 Methods and Table 1 NHL subtype block; the Table S1 coefficient row is DLBCL on lambda.
• Example models: Li_2017_nhl_pfs_mbma.R (log-linear coefficient -1.41 on the Weibull scale lambda of progression-free survival, i.e. an all-DLBCL cohort has exp(-1.41) = 0.24 of the median PFS time of an all-MCL cohort – the largest histology effect in that model, consistent with DLBCL being the aggressive high-grade subtype).
• Notes: MBMA study-arm-level covariate; sibling of TUMTP_FL_PCT and TUMTP_OTHER_NHL_PCT. See TUMTP_FL_PCT for the percent-versus-fraction warning and the mixed-histology rationale. Distinct from the per-subject binary TUMTP_DLBCL (an individual-level indicator).

TUMTP_OTHER_NHL_PCT (canonical for ‘other non-FL non-DLBCL non-MCL NHL’ cohort prevalence percentage)

• Description: Study-arm-level percentage (0-100) of the enrolled cohort whose NHL histology is none of follicular lymphoma, DLBCL or mantle cell lymphoma (e.g. peripheral T-cell lymphoma, marginal zone lymphoma). Continuous covariate scaled in percent (not fraction).
• Units: %
• Type: continuous
• Scope: specific
• Reference category: 0% (no other-NHL patients); see TUMTP_FL_PCT for the shared MCL reference level.
• Source aliases:
  • Percentage of patients with each NHL subtype / Other – Li 2017 Methods and Table 1 NHL subtype block; the Table S1 coefficient row is Other lymphomas on lambda.
• Example models: Li_2017_nhl_pfs_mbma.R (log-linear coefficient -0.17 on the Weibull scale lambda of progression-free survival, with a 113% RSE and a 95% CI of -0.54 to 0.21 that brackets zero; the paper reports other subtypes as similar to the MCL patient population).
• Notes: MBMA study-arm-level covariate; sibling of TUMTP_FL_PCT and TUMTP_DLBCL_PCT. Its membership is narrower than that of the individual-level TUMTP_OTHER_NHL canonical, which pools MCL into the residual bucket: here MCL is EXCLUDED because MCL is the reference histology of the founding model and carries its own zero-by-construction level. A model that uses a different reference histology must document its own bucket composition in covariateData[[TUMTP_OTHER_NHL_PCT]]$notes rather than overloading this canonical. See TUMTP_FL_PCT for the percent-versus-fraction warning. Typically a small column – only 4 of the 155 Li 2017 cohorts (2.5%) are predominantly other-NHL, which is why that model cannot distinguish its effect from the reference.

LINE_1L_PCT (canonical for treatment-naive (first-line) cohort prevalence percentage)

• Description: Study-arm-level percentage (0-100) of the enrolled cohort receiving the trial regimen as FIRST-LINE therapy, i.e. treatment-naive with no prior anticancer therapy for the indication. Continuous covariate scaled in percent (not fraction). The aggregate counterpart of the individual-level binary LINE_1L, whose 1 = first-line / treatment-naive polarity it keeps.
• Units: %
• Type: continuous
• Scope: specific
• Reference category: paper-specific and frequently 100%, not 0%. Li 2017 defines its reference population as an all-treatment-naive cohort, so the reference for that model is LINE_1L_PCT = 100. Record the model’s own reference level in covariateData[[LINE_1L_PCT]]$reference_category.
• Source aliases:
  • Percentage of patients with different numbers of prior treatments (naive, 1, 2+) – Li 2017 Methods.
  • % experienced patients – Li 2017 Table S1 coefficient row. Opposite polarity: the source coefficient multiplies the treatment-EXPERIENCED fraction, so a consuming model forms (100 - LINE_1L_PCT) / 100. Do not register a reverse-coded LINE_2L_PLUS_PCT companion; the two would have to sum to 100 and could silently disagree.
• Example models: Li_2017_nhl_pfs_mbma.R (log-linear coefficient -1.85 applied to the experienced fraction (100 - LINE_1L_PCT) / 100 on the Weibull scale lambda of progression-free survival, i.e. an all-experienced cohort has exp(-1.85) = 0.16 of the median PFS time of an all-naive cohort – a 6.4-fold swing and the largest covariate effect in that model).
• Notes: MBMA study-arm-level covariate; follows the _PCT aggregate-prevalence pattern founded by DIS_CHD_PERCENT (Vargo 2014) and applied to an oncology histology column by TUMTP_SQUAM_PCT. Distinct from the per-subject binary LINE_1L. A single linear column approximates what is often a threshold effect: Li 2017’s own subanalysis found treatment-naive and one-prior-line cohorts statistically indistinguishable while two-or-more-prior-line cohorts had under a tenth the median PFS time, so the linear prevalence column compresses a step at two prior lines. When a source resolves prior-therapy count into more than two levels at the arm level, register parallel LINE_3L_PCT / LINE_4L_PLUS_PCT members rather than overloading this one – the same guidance the individual-level LINE_1L entry gives, and note the N prior regimens -> line N + 1 off-by-one documented under LINE_3L. See TUMTP_FL_PCT for the percent-versus-fraction warning.

PS_ECOG_0_PCT (canonical for ECOG-performance-status-0 cohort prevalence percentage)

• Description: Study-arm-level percentage (0-100) of the enrolled cohort with an Eastern Cooperative Oncology Group (ECOG) Performance Status score of 0 at baseline (fully active / asymptomatic). Continuous covariate scaled in percent (not fraction). Higher arm-level ECOG-0 fraction typically indicates a healthier / more-active enrolled cohort and is associated with better survival outcomes.
• Units: %
• Type: continuous
• Scope: specific
• Reference category: 0% (no ECOG-0 patients; all ECOG >= 1).
• Source aliases: %ECOG PS score of 0 / ps.0 (Franzese 2026 Table 1 / Table S1 covariate label).
• Example models: Franzese_2026_pdl1_nsclc_mbma.R (linear coefficient on OS non-chemotherapy hazard (-0.400) and on PFS global hazard (-0.293) in a mNSCLC MBMA of PD-(L)1 immunotherapy).
• Notes: MBMA study-arm-level covariate. Distinct from the per-subject ECOG_GE1 and ECOG_GE2 binaries (individual-level indicators). Family precedent: DIS_CHD_PERCENT (Vargo 2014). Sibling members of the same family: PS_ECOG_1_PCT, PS_ECOG_2_PCT, PS_ECOG_3_PCT.

PS_ECOG_1_PCT (canonical for ECOG-performance-status-1 cohort prevalence percentage)

• Description: Study-arm-level percentage (0-100) of the enrolled cohort with an Eastern Cooperative Oncology Group (ECOG) Performance Status score of 1 at baseline (restricted in strenuous activity but ambulatory and able to carry out light work). Continuous covariate scaled in percent (not fraction).
• Units: %
• Type: continuous
• Scope: specific
• Reference category: 0% (no ECOG-1 patients). In a model that carries the whole PS_ECOG_*_PCT set, ECOG 0 is the natural reference level and is represented by leaving all of PS_ECOG_1_PCT / PS_ECOG_2_PCT / PS_ECOG_3_PCT at 0.
• Source aliases: PS1 (Chen 2026 NONMEM $INPUT); PS.1 (Chen 2026 Tables S5 / S6); ECOG PS 1 (%) (Chen 2026 Table 2).
• Example models: Chen_2026_nsclc_os_mbma.R (linear coefficient 0.004 on the overall-survival log hazard ratio, i.e. HR 1.004 for an all-ECOG-1 arm relative to an all-ECOG-0 arm, in an MBMA of platinum-ineligible advanced NSCLC).
• Notes: MBMA study-arm-level covariate; sibling of PS_ECOG_0_PCT (founding example Franzese 2026), from which the naming pattern is taken unchanged. The four PS_ECOG_<n>_PCT members sum to 100 within an arm, so a model that uses ECOG 0 as its reference level consumes only levels 1, 2 and 3. Distinct from the per-subject ECOG_GE1 / ECOG_GE2 binaries (individual-level indicators). Arm-level ECOG percentages are frequently IMPUTED in meta-analytic databases – Chen 2026 mapped Karnofsky Performance Scores where available and otherwise split composite ECOG categories with a logistic regression calibrated on studies with similar entry criteria – so record the imputation provenance in the model’s covariateData notes.

PS_ECOG_2_PCT (canonical for ECOG-performance-status-2 cohort prevalence percentage)

• Description: Study-arm-level percentage (0-100) of the enrolled cohort with an Eastern Cooperative Oncology Group (ECOG) Performance Status score of 2 at baseline (ambulatory and capable of all self-care but unable to work; up and about more than half of waking hours). Continuous covariate scaled in percent (not fraction).
• Units: %
• Type: continuous
• Scope: specific
• Reference category: 0% (no ECOG-2 patients); ECOG 0 is the usual reference performance-status level.
• Source aliases: PS2 (Chen 2026 NONMEM $INPUT); PS.2 (Chen 2026 Tables S5 / S6).
• Example models: Chen_2026_nsclc_os_mbma.R (pooled with PS_ECOG_3_PCT under a single coefficient 0.769, i.e. HR 2.158 relative to an all-ECOG-0 arm; the dominant covariate effect in that paper).
• Notes: MBMA study-arm-level covariate; sibling of PS_ECOG_0_PCT and PS_ECOG_1_PCT. Sources very often POOL ECOG 2 and ECOG 3 into one composite category, either in reporting or in the model itself. Keep the two as separate columns even then: the pooling is a property of the model equation (sum the two columns inside model(), as Chen_2026_nsclc_os_mbma does) rather than of the data, and source datasets routinely carry them apart even when the main-text table shows them together. Do not introduce a pooled PS_ECOG_23_PCT column.

PS_ECOG_3_PCT (canonical for ECOG-performance-status-3 cohort prevalence percentage)

• Description: Study-arm-level percentage (0-100) of the enrolled cohort with an Eastern Cooperative Oncology Group (ECOG) Performance Status score of 3 at baseline (capable of only limited self-care; confined to bed or chair more than half of waking hours). Continuous covariate scaled in percent (not fraction).
• Units: %
• Type: continuous
• Scope: specific
• Reference category: 0% (no ECOG-3 patients); ECOG 0 is the usual reference performance-status level.
• Source aliases: PS3 (Chen 2026 NONMEM $INPUT); PS.3 (Chen 2026 Tables S5 / S6).
• Example models: Chen_2026_nsclc_os_mbma.R (pooled with PS_ECOG_2_PCT under the shared coefficient 0.769).
• Notes: MBMA study-arm-level covariate; sibling of PS_ECOG_0_PCT, PS_ECOG_1_PCT and PS_ECOG_2_PCT. Usually zero or near-zero, because most oncology trials exclude ECOG 3 patients; it is non-zero in only four of the 41 arms of the Chen 2026 database. See PS_ECOG_2_PCT for why the 2-and-3 pooling stays in the model equation rather than in the column set.

RACE_ASIAN_PCT (canonical for Asian-race cohort prevalence percentage)

• Description: Study-arm-level percentage (0-100) of the enrolled cohort who are Asian (any Asian subgroup, matching the individual-level RACE_ASIAN canonical). Continuous covariate scaled in percent (not fraction).
• Units: %
• Type: continuous
• Scope: specific
• Reference category: 0% (all-non-Asian arm).
• Source aliases: %Asian race / Race.Asian (Franzese 2026 Table 1 / Table S1 covariate label).
• Example models: Franzese_2026_pdl1_nsclc_mbma.R (enters the OS ORR-slope as an interaction term (eta_orr_os - 0.595) * (ORR/100) * (RACE_ASIAN_PCT/100); the paper’s Discussion attributes the effect to regional trial-conduct differences rather than an inherent race effect).
• Notes: MBMA study-arm-level covariate. Distinct from the per-subject binary RACE_ASIAN canonical (individual-level 0 or 1); here the arm’s Asian fraction is a continuous proportion of participants. Also distinct from Yang 2010’s use of RACE_ASIAN as a binary at the arm level (whole-arm-Asian vs whole-arm-Western), which is coarser than the continuous-fraction form. Family precedent: DIS_CHD_PERCENT (Vargo 2014).

SEXF_PCT (canonical for female cohort prevalence percentage)

• Description: Study-arm-level percentage (0-100) of the enrolled cohort who are female, matching the individual-level SEXF canonical. Continuous covariate scaled in percent (not fraction).
• Units: %
• Type: continuous
• Scope: general
• Reference category: n/a – continuous; a paper that reports the percentage MALE maps as SEXF_PCT = 100 - percent male.
• Source aliases:
  • % male – Leil 2021 Table 2 (value transformation SEXF_PCT = 100 - % male).
  • MaleP – Zou 2026 Monolix column (percentage male; SEXF_PCT = 100 - MaleP).
• Example models: Leil_2021_rheumatoidArthritis_das28_mbma.R (the placebo Hill coefficient is shifted by exp(-0.176) in arms with fewer than 18.5 percent male patients, derived in model() as (100 - SEXF_PCT) < 18.5). Also documented (screened, not retained) in Zou_2026_pembrolizumab_qol_mbma.R and Li_2017_nhl_pfs_mbma.R.
• Notes: MBMA study-arm-level covariate in the <CONCEPT>_PCT family (percent 0-100, divided by 100 inside model() when used as a fraction). Distinct from the per-subject binary SEXF. When the source dichotomises the percentage, keep the column continuous and derive the indicator in model() so the cut point stays visible.

N_ARM (canonical for the number of participants contributing to a study-arm-level aggregate observation)

• Description: The arm sample size behind one aggregate (arm-mean or arm-proportion) data point in a model-based meta-analysis: how many participants were summarised into that value. A study-design quantity, not a patient characteristic, and supplied per observation row rather than estimated. It is the meta-analytic weight: an arm mean computed from N participants has a standard error proportional to 1 / sqrt(N), so MBMA models divide the residual SD – and sometimes a between-arm random effect – by sqrt(N_ARM). Time-varying in principle, because an arm’s evaluable count can fall between visits as participants drop out.
• Units: participants
• Type: count
• Scope: general
• Reference category: n/a – continuous count. Must be strictly positive; consuming models divide by sqrt(N_ARM).
• Source aliases:
  • NOC – Zou 2026 Monolix regressor (NOC = {use=regressor}), the number of individuals behind each arm-level QoL mean.
  • N_ijk – Zou 2026 Equations 2 and 3 symbol for the same quantity.
  • N, N_arm, n, SS, NSUB – common column-name variants in MBMA datasets; all map directly with no value transformation.
• Example models: Zou_2026_pembrolizumab_qol_mbma.R (used twice: the between-treatment-arm random effect on logit baseline QoL enters as eta_arm_e0 / sqrt(N_ARM), and the residual SD is addSd / sqrt(N_ARM)), Yang_2026_copd_fev1_adipd_mbma.R, Checchio_2017_psoriasis_pasi75_longitudinal_mbma.R (both residual terms of Checchio 2017 Equation 13 are multiplied by the binomial standard error W = sqrt(Pr*(1-Pr)/N_ARM) of Equation 12: the independent term becomes the record’s residual SD, sdArm <- wArm * addSd_prob_pasi75, and the arm-level correlated term is added on the probability scale as wArm * eta_study_corr. It is the second term that forces the column – a per-arm RANDOM-EFFECT SD cannot be reproduced by rescaling the output of a single rxSolve()), He_2021_psoriasis_pasi75_mbma.R, He_2021_psoriasis_pasi90_mbma.R (the residual SD of He 2021 Equations 7-8 is sigma * sqrt(P*(1-P)/N_ARM), formed in model() as in Checchio 2017; He 2021 has no arm-level correlated term), Yu_2022_crohns_cdai150_mbma.R, Yu_2022_crohns_cdai100_mbma.R, Yu_2022_crohns_cdai70_mbma.R (residual SD addSd * sqrt(P*(1-P)/N_ARM), the gnls varPower weight of the source), Yu_2022_crohns_dcdai_mbma.R, Yu_2022_crohns_dcrp_mbma.R, Yu_2022_crohns_dibdq_mbma.R (residual SD addSd / sqrt(N_ARM), the gnls varFixed SD/sqrt(N) weight of the source), Leil_2021_rheumatoidArthritis_das28_mbma.R (Leil 2021 Equations 7, 8 and 12 scale every between-trial and between-arm random effect and the residual by (N / 100)^-0.5, encoded as wN <- sqrt(100 / N_ARM)).
• Notes: MBMA / trial-design covariate; family precedent K_CTRL and LOGRR_TARGET (Volkova 2023), which are likewise trial-design inputs rather than patient characteristics. Registered as an explicit column because a growing set of MBMA models needs the arm size INSIDE model(), not merely as a post-hoc scaling. Most of the library’s earlier MBMA extractions (Mercier_2014_tramadol_tapentadol_mbma, Chen_2025_methotrexate_acr20_mbma / _acr50_ / _das28_, Asiimwe_2025_trastuzumab*_mbma, Hanan_2026_peginterferon_alfa_*_mbma) store the residual SD UNWEIGHTED and state in ini() that downstream code must apply 1 / sqrt(N); that remains valid where the arm size touches only the residual. Use N_ARM instead when the arm size enters a structural or random-effect expression, where no downstream rescaling can reproduce it – a per-arm random-effect SD cannot be varied across arms within a single rxSolve() without a data column. Scope general because arm size carries the same meaning in every MBMA, with no paper-specific semantics. Do NOT use for an individual-level covariate: a row carrying N_ARM is an aggregate, and a model consuming it is simulating arm means, not patients.

YEAR_PUB (canonical for the calendar year of a trial in a model-based meta-analysis – publication or conduct year, stated per model)

• Description: Calendar year of a trial (a four-digit year), a study-level characteristic in a model-based meta-analysis used as a proxy for secular change in background care. Continuous; one value per trial. Sources use either the publication year or the year the trial was conducted; each model’s covariateData notes state which one its value is.
• Units: year
• Type: continuous
• Scope: general
• Reference category: n/a – continuous. Leil 2021 centres it (log-transformed) at 2013, the median publication year of its Supplementary Table S1 trials, because the data-set median is not printed.
• Source aliases:
  • trial year / year of conduct of the trial – Leil 2021 Methods and Table S3 (conduct year).
  • YEAR_PUBLICATION – Chen 2026 spelling (screened, not retained, in Chen_2026_nsclc_os_mbma.R; publication year).
  • YEAR, PUBYEAR, YEAR_TRIAL – common dataset column variants; map directly, after confirming which year the source recorded.
• Example models: Leil_2021_rheumatoidArthritis_das28_mbma.R (conduct year; exponent -397 of YEAR_PUB / 2013 on the DAS28 progression slope; earlier trials show faster background worsening), Zierhut_2016_hcc_antiangiogenic_os_mbma.R (publication year; screened and not retained, in covariatesDataExcluded).
• Notes: MBMA / trial-design covariate, sibling of N_ARM. Publication year lags conduct year by a few years, so do not pool values of the two kinds without checking; a model with a steep year effect is sensitive to the difference. Ratified 2026-09-29 alongside the Leil 2021 RA DAS28 extraction: the operator chose the existing Zierhut 2016 spelling over a separate YEAR_TRIAL canonical, with the year kind recorded per model.

K_CTRL (canonical for control-arm event proportion over the planned follow-up)

• Description: The proportion of control-arm participants expected to experience the endpoint of interest over a planned trial’s follow-up period; a fraction in (0, 1). A trial-design input, not a measured patient characteristic: it is what a study-level model needs, alongside an effect size, to turn a predicted treatment effect into a required enrolment. Time-dependent in principle – the proportion accrues with follow-up duration – so the value supplied is always tied to a stated follow-up length.
• Units: (fraction)
• Type: continuous
• Scope: specific
• Reference category: n/a – continuous. Must be strictly greater than 0; the sample-size formula divides by it.
• Source aliases:
  • k_i(t) – Volkova 2023 Section 2.4 / Supplementary File 2 symbol.
  • control-arm event rate, baseline risk, p_control – common prose forms.
• Example models: Volkova_2023_lipidLowering_mace_mbma.R (input to the n_required output; for a composite endpoint the paper’s Supplementary File 3 shows the same formula applies with k_sum = sum(k_i) over the component events, assuming overlap between components is negligible).
• Notes: MBMA / trial-design covariate; a fraction, never a percentage, so it does not belong to the *_PERCENT cohort-prevalence family. Pairs with LOGRR_TARGET; a sample-size calculation needs both. Supplied externally because the source that provides the effect size often does not publish the control-arm rate model in usable form – Volkova 2023 fitted a control-arm event-rate meta-regression but reported it only as fitted lines in Supplementary Figure 1, with no coefficients printed anywhere. Where the source publishes worked sample-size results, a value can be recovered by inverting the published formula (the Volkova 2023 vignette does this for nonfatal myocardial infarction, recovering k ~ 0.043 at 4 years consistently from two independent Figure 3 points).

LOGRR_TARGET (canonical for the target log risk ratio in a trial sample-size calculation)

• Description: The log risk ratio a planned trial must demonstrate for the treatment benefit to reach statistical significance; the effect-size input to a sample-size calculation. A trial-design input, not a patient characteristic. Negative values favour the active treatment.
• Units: (unitless log ratio)
• Type: continuous
• Scope: specific
• Reference category: n/a – continuous. Must be strictly non-zero for a finite required sample size; as it approaches 0 the required enrolment diverges.
• Source aliases:
  • theta_i, u_i – Volkova 2023 Supplementary File 2 and Section 2.4 symbols for the same quantity.
• Example models: Volkova_2023_lipidLowering_mace_mbma.R (input to the n_required output; supply one of the model’s own logrr_* predictions to size a single endpoint, or the Supplementary File 3 aggregate theta_sum = log(sum(k_i * exp(theta_i)) / sum(k_i)) to size a composite endpoint).
• Notes: MBMA / trial-design covariate; pairs with K_CTRL. Deliberately an input rather than an internally-computed quantity, so that one calculator serves both the single-endpoint and the composite-endpoint case. When the effect size is itself a prediction of the same model, reproducing a published sample size is a two-pass exercise (solve for the effect size, then re-solve supplying it here) – that mirrors the two-stage structure of the source method rather than working around a limitation. Record in covariateData[[LOGRR_TARGET]]$notes whether a given value is a point estimate or a draw from the effect-size uncertainty distribution, since the published percentile ranges around a sample size come from sampling this input.

FORM_ISA_P2F2 (canonical for isatuximab P2F2 drug-material indicator)

• Description: 1 = isatuximab P2F2 drug material (intended commercial / phase III material, used in the EFC14335 / ICARIA-MM study), 0 = P1F1 drug material (early-phase material used in TED10893 / TED14154 / TCD14079).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (P1F1).
• Source aliases:
  • FORM_P2F2 – prior canonical name (pre-2026-06-19 standardization audit).
  • Drug_mat – used in Fau_2020_isatuximab.R. Values 0 / 1 with the same orientation as the canonical (1 = P2F2 / commercial-bound material).
• Example models: Fau_2020_isatuximab.R (exponential effect on Vc with coefficient -0.137; P2F2 patients had ~13% lower Vc than P1F1).
• Notes: Phase III / commercial-bound formulation indicator for isatuximab; the FORM_* family stays scope-specific per nlmixr2lib policy that drug-product-version indicators are kept model-specific unless they generalize across multiple drugs. Set to 1 to simulate the marketed material.

FORM_LINAG_TAB1 (canonical for linagliptin tablet formulation 1 indicator)

• Description: 1 = subject received the linagliptin “tablet formulation 1” (used in Retlich 2015 Study 2), 0 = subject received tablet formulation 2 (the marketed linagliptin tablet, used in Studies 3 and 4) OR the powder-in-bottle formulation (used in Study 1). The powder-vs-tablet contrast is captured by the sibling canonical FORM_POWDER; this indicator switches between the two tablet formulations.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (tablet formulation 2 = marketed linagliptin tablet, the typical-value Ka reference; or FORM_POWDER = 1 for the powder).
• Source aliases: none known.
• Example models: Retlich_2015_linagliptin.R (multiplicative shift on the linagliptin first-order absorption rate constant Ka; typical Ka = 0.441 1/h for tablet formulation 2 (reference), 0.795 1/h for tablet formulation 1, 0.933 1/h for the powder formulation).
• Notes: Specific scope because the linagliptin “tablet 1 vs tablet 2” distinction is a drug-product-version comparison local to the Retlich 2015 popPK dataset; tablet formulation 1 was a development formulation that is not marketed. Mirrors the FORM_SAR_DP2 (sarilumab) and FORM_ISA_P2F2 (isatuximab) entries under the FORM_* family. Set to 0 for routine marketed-formulation simulation.

FORM_TEPOTINIB_CF1 (canonical for tepotinib capsule formulation 1 (non-micronised) indicator)

• Description: 1 = dose record given as tepotinib capsule formulation 1, the only formulation made with non-micronised drug substance, 0 = any other formulation (reference tablet TF2; capsule CF2 and the oral solution carry no separate effect).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (tablet formulation TF2).
• Source aliases:
  • CF1 – Xiong 2022 Tables 1-3.
• Example models: Xiong_2022_tepotinib.R (fractional effects on relative bioavailability, -0.656, and on ka, -0.442; also selects a CF1-specific IIV variance on F, 0.713 versus 0.283 for other doses in patients).
• Notes: Time-varying per dose record. Used only for the initial doses (up to 230 mg) of the first-in-human study 001. Development formulation; set to 0 for simulation of the marketed product (TF3).

FORM_TEPOTINIB_TF1 (canonical for tepotinib tablet formulation 1 indicator)

• Description: 1 = dose record given as tepotinib tablet formulation 1 made with micronised drug substance, 0 = any other formulation (reference tablet TF2).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (tablet formulation TF2).
• Source aliases:
  • TF1 – Xiong 2022 Tables 1-3.
• Example models: Xiong_2022_tepotinib.R (fractional effect on ka, +0.305).
• Notes: Time-varying per dose record. Mutually exclusive with FORM_TEPOTINIB_TF1FINE, the version of TF1 made with finely micronised drug substance. Development formulation.

FORM_TEPOTINIB_TF1FINE (**canonical for tepotinib tablet formulation 1 with finely micronised drug substance (TF1*) indicator**)

• Description: 1 = dose record given as TF1*, the version of tepotinib tablet formulation 1 that contains finely micronised (rather than just-micronised) drug substance, 0 = any other formulation (reference tablet TF2).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (tablet formulation TF2).
• Source aliases:
  • TF1* – Xiong 2022 Tables 2-3 (the asterisk is not a valid column-name character, hence the FINE suffix).
• Example models: Xiong_2022_tepotinib.R (fractional effect on ka, +0.674).
• Notes: Time-varying per dose record. Mutually exclusive with FORM_TEPOTINIB_TF1. Development formulation.

FORM_TEPOTINIB_TF3 (canonical for tepotinib tablet formulation 3 (marketed tablet) indicator)

• Description: 1 = dose record given as tepotinib tablet formulation 3, 0 = any other formulation (reference tablet TF2).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (tablet formulation TF2).
• Source aliases:
  • TF3 – Xiong 2022 Tables 1-3.
• Example models: Xiong_2022_tepotinib.R (fractional effect on relative bioavailability, +0.154; Figure 3 simulates TF3 500 mg QD with food).
• Notes: Time-varying per dose record. TF3 is the formulation used in VISION and in the bioequivalence study 0044; set to 1 to simulate the marketed product.

FORM_SAR_DP2 (canonical for sarilumab drug-product version 2 indicator)

• Description: 1 = sarilumab drug product 2 formulation (used in some phase I studies and the dose-ranging phase II study), 0 = other drug product (DP1 or DP3; DP3 is the commercial formulation).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (DP1 or DP3).
• Source aliases:
  • FORM_DP2 – prior canonical name (pre-2026-06-19 standardization audit).
  • DP2 – used in Xu_2019_sarilumab.R.
• Example models: Xu_2019_sarilumab.R.
• Notes: Affects both CLO/F (1.30x multiplier) and Ka (0.663x multiplier) in Xu 2019. Set to 0 for routine commercial-formulation simulation.

FORM_POSA_AB (canonical for posaconazole prototype tablet (A or B) vs later (C or D) tablet formulation indicator)

• Description: 1 = subject received the posaconazole delayed-release solid tablet formulation A or B (the two early prototype formulations used in some phase 1 studies of the van Iersel 2018 pooled analysis); 0 = subject received tablet formulation C or D (the later formulations, with D being the marketed / commercial formulation and the predominant formulation across the pooled data set). Per-dose-record categorical indicator switching relative bioavailability F1 between the prototype and later tablet formulations.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (tablet C or D; the predominant formulation across the pooled data set and the typical-value F1 = 1 reference in van Iersel 2018 final model per Results ‘Posaconazole exposure… and impact of covariates’: ‘tablet formulation D is the marketed image and the predominant formulation used across the data set and was taken as the reference formulation for the relative bioavailability estimate for tablet formulations A and B’).
• Source aliases: none – the paper narrates the contrast as “tablet formulation A/B versus tablet formulation C/D” without a single NMTRAN column name.
• Example models: vanIersel_2018_posaconazole.R (multiplicative effect on relative bioavailability F1: fdepot *= (1 + e_formposaab_fdepot * FORM_POSA_AB) with e_formposaab_fdepot = +0.247, i.e. the prototype A/B formulations have 24.7% higher F1 than the later C/D reference; van Iersel 2018 Table 2 final-model ‘Tablet formulation A/B on F1’ = 0.247, RSE 21.5%).
• Notes: Specific scope because the prototype-A-or-B vs later-C-or-D contrast is paper-specific to the van Iersel 2018 posaconazole solid-tablet clinical-development pooled analysis. Drug-specific member of the FORM_<drug>_<contrast> family alongside FORM_ASV_LIQUID (asunaprevir), FORM_SAR_DP2 (sarilumab), FORM_VISMO_PHASEI (vismodegib), FORM_LINAG_TAB1 (linagliptin), FORM_ABA_PHASE2 (abatacept). Set to 0 for routine commercial-formulation simulation (tablet D).

FORM_VISMO_PHASEI (canonical for vismodegib Phase I (dry-blend capsule) formulation indicator)

• Description: 1 = subject received the Phase I clinical-development vismodegib formulation (dry-blend capsules used in early-phase studies SHH3925g and the Phase I cohort of SHH4610g); 0 = subject received the Phase II / commercial-bound formulation (wet-granulation capsules used in SHH4476g, SHH4433g, and most of SHH4683g / SHH4871g). Per-subject (regimen-fixed) categorical indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (Phase II / commercial wet-granulation capsule; F = 1 fixed as the reference in Lu 2015 Eq. 4 and the typical-value ka reference in Eq. 4).
• Source aliases: none known; Lu 2015 reports formulation as a paper-defined “Phase I vs Phase II formulation” categorical without committing to a single column name.
• Example models: Lu_2015_vismodegib.R (multiplicative shifts on Ka and relative bioavailability F: typical Ka = 9.025 1/day for the Phase II reference in patients, with exp(-0.602) = 0.55x for the Phase I formulation and exp(0.671) = 1.96x for healthy volunteers; F = 1 for the Phase II reference, F = 0.346 for the Phase I formulation in patients and 0.836 in HV).
• Notes: Specific scope because the “Phase I (dry blend) vs Phase II (wet granulation) capsule” contrast is tied to the vismodegib drug-product-version comparison in Lu 2015. Mirrors the FORM_SAR_DP2 (sarilumab), FORM_ISA_P2F2 (isatuximab), and FORM_LINAG_TAB1 (linagliptin) entries under the FORM_* family of drug-product-version indicators. Set to 0 for routine commercial-formulation simulation.

FORM_DOX_DORYX_MPC (canonical for Doryx MPC delayed-release tablet doxycycline formulation indicator)

• Description: 1 = subject received the Doryx MPC delayed-release tablet (Mayne Pharma International; modified-acid-resistance delayed-release formulation, 120 mg doxycycline hyclate per tablet, formulation code MP336); 0 = subject received any other Doryx doxycycline product (the Doryx delayed-release tablet reference at 75 / 100 / 150 / 200 mg, or the conventional-release Doryx capsule at 100 mg, flagged separately by FORM_CAPSULE). Per-subject (regimen-fixed) categorical indicator used in popPK analyses that pool the three Doryx formulations and test formulation as a covariate on relative bioavailability, absorption lag, and the transit absorption rate’s food effect.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (Doryx delayed-release tablet; the typical-value reference for relative bioavailability F = 1 and the typical-value reference for the food effect on KTR in Hopkins 2017 Table 3).
• Source aliases:
  • FORM_DORYX_MPC – prior canonical name (pre-2026-06-19 standardization audit).
  • FMPC – used in Hopkins_2017_doxycycline.R (Hopkins 2017 Methods ‘Base model’ paragraph: “Structural covariates… a formulation effect (for Doryx MPC, FMPC; for Doryx capsule, FCAP) on relative bioavailability (RELF)”).
• Example models: Hopkins_2017_doxycycline.R (multiplicative effects per Hopkins 2017 Table 3 final model: relative bioavailability F1MPC = 0.863 vs the Doryx-tablet reference (Theta 8); shared absorption lag ALAG1 = 0.115 h with the Doryx capsule (Theta 12); strengthened food effect on the transit rate constant KTR: fed state reduces KTR by 54.9 % for Doryx MPC (COVFED2 = -0.549, Theta 11) vs 20.9 % for Doryx tablet / capsule (COVFED = -0.209, Theta 10)).
• Notes: Specific scope because the Doryx MPC vs Doryx tablet vs Doryx capsule contrast is tied to the doxycycline drug-product-version comparison in Hopkins 2017 (Mayne Pharma International’s modified-acid-resistance MP336 formulation vs the marketed Doryx delayed-release tablet vs the conventional-release Doryx capsule). Paired with FORM_CAPSULE (the Doryx capsule indicator) so that both indicators = 0 selects the Doryx-tablet reference; the two indicators are mutually exclusive at any given dose record. Mirrors the existing FORM_SAR_DP2 (sarilumab), FORM_ISA_P2F2 (isatuximab), FORM_LINAG_TAB1 (linagliptin), FORM_VISMO_PHASEI (vismodegib), FORM_TAC_IR (tacrolimus IR vs PR), and FORM_THEO_APNECUT (theophylline) entries under the FORM_* family of drug-product-version indicators. Set to 0 for routine Doryx-tablet simulation, 1 for Doryx MPC simulation.

FORM_DOX_FEED (canonical for doxycycline-in-medicated-feed vs doxycycline-in-aqueous-solution indicator)

• Description: 1 = the oral doxycycline dose was administered mixed into medicated feed (meal or pelleted ration); 0 = the oral dose was administered as an aqueous solution, either taken spontaneously in drinking water or delivered by gastric tube. Per-dose-record indicator used in veterinary oral-doxycycline population analyses that pool in-feed and in-water medication, which differ substantially in both absorption rate and absolute bioavailability.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (doxycycline given as an aqueous solution; in Toutain 2025 this reference splits further into drinking water vs stomach tube via ROUTE_NGT).
• Source aliases:
  • “Route of administration = oral feed” – Toutain 2025 Table 1 route column, distinguishing the six in-feed trials (AFSSA, BIOEQ, PARADOX, TLS, Company 9203, Company 9204) from the five aqueous-solution trials (104NL, 3205NL, GHENT, KING_NL, Bea).
• Example models: Toutain_2025_doxycycline_pig.R (selects between the two in-feed absorption sub-models and the two aqueous-solution sub-models of the pig meta-analysis; combined with STUDY_TLS it picks Ka = 0.072 / F = 0.501 for feed under field conditions or Ka = 0.144 / F = 0.340 for feed under laboratory conditions, and combined with ROUTE_NGT it picks Ka = 0.689 / F = 0.307 for drinking water or Ka = 0.725 / F = 0.258 for stomach tube, each stratum also carrying its own additive-plus-proportional residual error; Toutain 2025 Table 6).
• Notes: Well-formed member of the auto-approved FORM_<drug>_<formulation> family, using the same DOX drug token as the existing FORM_DOX_DORYX_MPC entry. Distinct from a food-effect covariate: this is not “drug taken with a meal” but “drug incorporated into the ration”, which in production animals also changes the dosing process (group access to a medicated meal, competition between animals) and is the reason Toutain 2025 reports a between-subject variability on bioavailability of 84.8% in feed versus 34.3% in drinking water. Ignored when ROUTE_IV = 1.

FORM_RAL_600 (canonical for raltegravir 600 mg vs 400 mg tablet formulation indicator)

• Description: 1 = subject received the raltegravir 600 mg film-coated tablet formulation (Isentress HD; the 1200 mg once-daily reformulation approved 2017); 0 = subject received the raltegravir 400 mg film-coated tablet formulation (Isentress; the original 400 mg twice-daily formulation approved 2007). Per-dose-record indicator used in raltegravir popPK analyses that pool the two tablet strengths to test formulation-specific effects on bioavailability, inter-occasion variability magnitude, or food-effect susceptibility. The 600 mg tablet is a “reformulated” polymer-based product with faster disintegration / dissolution than the original 400 mg tablet at physiological gastric pH.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (400 mg tablet reference; typical-value F, IOV F, and food-effect coefficient reported for the original 400 mg formulation in Bukkems 2021 Table 2).
• Source aliases:
  • NEW – used in Bukkems_2021_raltegravir.R (NONMEM covariate flag; source paper Bukkems 2021 Supporting Information S4 $PK section: IF (NEW.EQ.1) selects the 600 mg formulation multipliers on F and IOV F).
• Example models: Bukkems_2021_raltegravir.R (two multiplicative effects per Bukkems 2021 Table 2: linear additive effect on typical bioavailability F *= (1 + e_ral_600_fdepot * FORM_RAL_600) with e_ral_600_fdepot = 0.209 – 600 mg tablet has +21% bioavailability vs 400 mg reference; and multiplicative effect on the IOV F magnitude SD_iov_f *= (1 + e_ral_600_iov_fdepot * FORM_RAL_600) with e_ral_600_iov_fdepot = -0.718 – 600 mg tablet has 72% smaller inter-occasion variability magnitude on F than the 400 mg reference).
• Notes: Specific scope because the raltegravir 600 mg vs 400 mg contrast is tied to the two-tablet-strength Merck reformulation program (Isentress HD / SWITCHmrk regulatory package). Mirrors the existing FORM_DOX_DORYX_MPC (doxycycline Doryx MPC vs Doryx tablet), FORM_ITR_SUBA (itraconazole SUBA vs Sporanox), FORM_TAC_IR (tacrolimus Prograf IR vs Advagraf PR), FORM_SAR_DP2 (sarilumab DP2), FORM_ISA_P2F2 (isatuximab P2F2), FORM_LINAG_TAB1 (linagliptin), and FORM_VISMO_PHASEI (vismodegib Phase I) entries under the FORM_<drug>_<formulation> family of drug-product-version indicators. Set to 0 for routine 400 mg BID simulation, 1 for 600 mg (typically dosed as two 600 mg tablets = 1200 mg QD) simulation.

FORM_NMV_TAB150 (canonical for nirmatrelvir 150-mg tablet formulation indicator)

• Description: 1 = the nirmatrelvir dose was given as the 150-mg film-coated tablet (the commercial PAXLOVID tablet strength used for the bulk of the phase II/III EPIC-HR population); 0 = the nirmatrelvir dose was given as the oral suspension or as the 100-mg tablet. Per-dose-record indicator used in nirmatrelvir popPK analyses that pool the three clinical formulations and estimate a relative-bioavailability offset for the 150-mg tablet only.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (oral suspension or 100-mg tablet; both share the F1 = 1 structural anchor because the source estimates an effect only for the 150-mg level of the three-level formulation column).
• Source aliases:
  • FFORM (three-level formulation column; 0 = suspension, 1 = 100-mg tablet, 2 = 150-mg tablet) – used in Chan_2023_nirmatrelvir.R; the 150-mg-tablet level maps to FORM_NMV_TAB150 = 1 (Chan 2023 Data S1 $PK: F1FFORM = 1; IF(FFORM.EQ.2) F1FFORM = 1+THETA(15)).
• Example models: Chan_2023_nirmatrelvir.R (linear fractional effect on relative bioavailability F1 *= (1 + e_form_nmv_tab150_fdepot * FORM_NMV_TAB150) with e_form_nmv_tab150_fdepot = -0.379, i.e. the 150-mg tablet carries 37.9% lower relative bioavailability than the suspension / 100-mg-tablet reference, applied on top of the dose power function (DOSE/300)^-0.409).
• Notes: Specific scope because the reference category pools two distinct products (suspension and 100-mg tablet) rather than contrasting a single pair, which is a consequence of Chan 2023 estimating an effect only for the FFORM = 2 level. Sibling of FORM_RAL_600 (raltegravir 600 mg vs 400 mg tablet), FORM_LINAG_TAB1, FORM_TAC_IR, FORM_ITR_SUBA and the rest of the FORM_<drug>_<formulation> family of drug-product-version indicators. Chan 2023 Discussion notes that the 150-mg tablet was evaluated only among healthy participants in a single single-dose study, so the estimated formulation effect on F1 is likely partially confounded with the COVID-19 effect on CL (DIS_COVID19); set both to their reference values when simulating the phase I suspension / 100-mg-tablet condition. A separate phase I study (NCT05263895) evaluated relative bioequivalence of the two tablet strengths but did not contribute to this analysis.

FORM_ANACETRAPIB_LFC (canonical for the anacetrapib Imwitor/Tween liquid-filled-capsule formulation indicator)

• Description: 1 = the dose was given as the Imwitor/Tween liquid-filled capsule (LFC) used through phase I and phase IIb; 0 = the dose was given as the hot-melt-extruded (HME) tablet that became the intended commercial and phase III formulation.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (hot-melt-extruded tablet). Six of the eight studies pooled in Krishna 2011 and 474 of the 576 subjects used the LFC; two trials used the HME tablet – a 78-subject bridging study and a 24-subject food-effect study.
• Source aliases:
  • formulation type (HME tablets vs. LFC) – Krishna 2011 Population PK Model prose; the canonical’s orientation is the reverse of the paper’s phrasing (1 = LFC here).
• Example models: Krishna_2011_anacetrapib.R (gates both NDOSEUNITS terms on bioavailability – the Eq. 7 Emax term and the fasted Eq. 5 exponential – so that FORM_ANACETRAPIB_LFC = 0 collapses each to 1 for tablet records).
• Notes: The formulation effect in this model is the dosage-unit-count effect, and the indicator carries no coefficient of its own. Krishna 2011 Table I contains no formulation theta, even though the text reports that adding the phase IIb data brought in “formulation effects, capsule effects”; the two are the same thing, because unit count is only defined for the LFC. The gating is not printed in Krishna 2011 Eqs. 4-7 but is fixed by two independent pieces of the paper: the Methods state the model assessed “number of capsules per dose (for LFC only)”, and the arithmetic agrees – the paper reports F1 = 0.35 for a fasted 100 mg TABLET, and only a collapsed count term reproduces it (1 - 100/155 = 0.3548), whereas carrying the Eq. 7 term for the tablet gives 0.21 at one unit and 0.32 at the 5.5-unit fasted median. The same gating then reproduces the paper’s three printed relative bioavailabilities (0.85 low-fat, 1.17 patient-selected, 2.07 high-fat) to the printed precision. Krishna 2011’s Discussion reports the two formulations were “generally similar” in AUC with the HME running about 25% higher exposure at 200 mg with a high-fat meal and about 40% lower Cmax fasted; those differences are what the count terms absorb.

FORM_IG_HYALURONIDASE (canonical for hyaluronidase-facilitated subcutaneous immunoglobulin (fSCIg) product indicator)

• Description: 1 = the subcutaneous immunoglobulin dose was given as hyaluronidase-facilitated subcutaneous immunoglobulin (fSCIg, e.g. HyQvia), in which recombinant human hyaluronidase is co-administered to depolymerise subcutaneous hyaluronan, increase dispersion and permit infusion volumes up to about 600 mL at a single site; 0 = the dose was given as conventional unfacilitated subcutaneous immunoglobulin (SCIg). Per-occasion, since a single patient may receive both products across a study. Applies to subcutaneous doses only and is undefined for intravenous doses, whose bioavailability is 1 by definition.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (unfacilitated SCIg).
• Source aliases:
  • hyaluronidase product yes/no and IgG product – used in Li_2022_immunoglobulin.R (van der Zeeuw 2026 Table 3 lists both among the covariates tested).
• Example models: Li_2022_immunoglobulin.R (multiplicative ratio 1.1262 on subcutaneous bioavailability, reference category unfacilitated SCIg; the two printed absolute bioavailabilities are 70.5% without hyaluronidase and 79.4% with it).
• Notes: Member of the auto-approved FORM_<drug>_<formulation> drug-product-version family. The effect is specifically on bioavailability, not on the absorption rate constant: in the only reviewed model that carries fSCIg data, product type was the sole covariate any model retained on F (van der Zeeuw 2026 section 3.2.1.2). Expressing the fSCIg bioavailability relative to IVIg gives approximately 92%, consistent with the roughly 93% AUC-based figure reported elsewhere in the literature, whereas unfacilitated SCIg sits in the 65-74% band. Use FORM_IG_HYALURONIDASE rather than a route column: the intravenous-versus-subcutaneous distinction is already carried by which compartment receives the dose (central versus depot), so this column encodes only the within-subcutaneous product contrast.

FORM_ITR_SUBA (canonical for SUBA-itraconazole vs Sporanox capsule formulation indicator)

• Description: 1 = subject received the SUBA-itraconazole formulation (a solid dispersion of itraconazole in a pH-dependent polymeric matrix that enhances dissolution and targets drug release in the proximal small intestine; marketed as Lozanoc in Australia, Itragerm in Spain); 0 = subject received the Sporanox capsule (Janssen Pharmaceuticals innovator product). Per-dose-record categorical indicator used in itraconazole bioequivalence-comparison popPK analyses.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (Sporanox capsule; relative bioavailability F = 1 and FVAR-variability scaling ETASCALE = 1 as the structural references in Abuhelwa 2015 Table 3).
• Source aliases:
  • FORM_SUBA – prior canonical name (pre-2026-06-19 standardization audit).
  • DRUG – used in Abuhelwa_2015_itraconazole.R (Appendix S1 NONMEM control stream: IF (DRUG.EQ.0) THEN DRUGF = 1 (Sporanox) ELSE DRUGF = (1 + THETA(8)) (SUBA-itraconazole)).
• Example models: Abuhelwa_2015_itraconazole.R (multiplicative effects per Abuhelwa 2015 Table 3: SUBA-itraconazole has +73% relative bioavailability vs Sporanox (F = 1 + 0.729 x FORM_ITR_SUBA) and 21.3% less F variability (ETASCALE = 1 + (-0.213) x FORM_ITR_SUBA = 0.787 scales the shared FVAR random effect)).
• Notes: Specific scope because the SUBA-itraconazole vs Sporanox capsule contrast is tied to the Mayne Pharma bioequivalence-development programme. Mirrors the FORM_SAR_DP2 (sarilumab), FORM_ISA_P2F2 (isatuximab), FORM_LINAG_TAB1 (linagliptin), and FORM_VISMO_PHASEI (vismodegib) entries under the FORM_* family of drug-product-version / formulation-comparison indicators. Distinct from FORM_CAPSULE (Hennig 2006 / 2007 capsule-vs-solution contrast for the same itraconazole molecule) because here both arms are capsule formulations and the contrast is between two capsule drug products.

FORM_TAC_IR (canonical for tacrolimus immediate-release vs prolonged-release formulation indicator)

• Description: 1 = subject received the twice-daily immediate-release tacrolimus formulation (Prograf, Astellas) administered every 12 hours; 0 = subject received the once-daily prolonged-release tacrolimus formulation (Advagraf in Europe, Astagraf XL in the US; both Astellas) administered every 24 hours. Per-subject (regimen-fixed; per-occasion in cross-over conversion studies) categorical indicator used in popPK analyses that pool the two oral tacrolimus formulations and test formulation as a covariate on absorption and disposition parameters.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (Advagraf / Astagraf XL prolonged-release; the typical-value reference for Ktr and Vc/F in Woillard 2011 Table 4).
• Source aliases:
  • study – used in Woillard_2011_tacrolimus.R (Woillard 2011 Methods: “study factor (assumed to be similar to drug formulation) … study = 1 for the Prograf cohort, study = 0 for the Advagraf cohort”). The source paper’s study and formulation factors are aliased; the canonical preserves the source paper’s orientation (Prograf = 1 = IR, Advagraf = 0 = PR), which also matches the standard convention of treating the older immediate-release formulation as the non-reference level.
  • FORMULATION (with 1 = PR-T, opposite polarity) – used in Lu_2019_tacrolimus_industry_meta.R. Lu 2019 writes the Ka covariate equation as Ka = 0.375 * (1 - (1 - theta_form) * FORMULATION) with FORMULATION = 1 for prolonged-release; the model derives form_pr = 1 - FORM_TAC_IR inside model() to match the paper’s published equation while keeping the canonical column oriented Prograf = 1.
• Example models: Woillard_2011_tacrolimus.R (multiplicative effects per Woillard 2011 Table 4: Ktr = theta1 * theta2^FORM_TAC_IR with theta1 = 3.34/h and theta2 = 1.53 (Prograf ~53 % faster absorption than Advagraf); Vc/F = theta6 * theta7^FORM_TAC_IR with theta6 = 486 L and theta7 = 0.29 (Prograf Vc/F is 29 % of the Advagraf reference)), Lu_2019_tacrolimus_industry_meta.R (linear effect on Ka per Lu 2019 Table 3 ‘Prolonged-release tacrolimus on Ka’ = 0.499, encoded as Ka(PR-T) = Ka(IR-T) * 0.499 via form_pr = 1 - FORM_TAC_IR; 50 % slower absorption for PR-T vs IR-T).
• Notes: Scope promoted to general after Lu 2019 corroborated the Prograf-vs-Advagraf contrast in a second, larger-cohort population (Woillard 2011 n = 173 + 174, Lu 2019 n = 408 across 8 Astellas Phase II studies). Mirrors the existing FORM_SAR_DP2 (sarilumab), FORM_ISA_P2F2 (isatuximab), FORM_LINAG_TAB1 (linagliptin), and FORM_VISMO_PHASEI (vismodegib) entries under the FORM_* family of drug-product-version indicators. The Woillard 2011 paper notes the formulation effect partially confounds with time-post-transplant (Prograf cohort sampled within the first 6 months post-transplant, Advagraf cohort > 12 months post-transplant); Lu 2019 has no such confounding because most studies are within-subject IR-T-to-PR-T conversions. Future tacrolimus models that include Envarsus XR (modified-release once-daily granules) or LCP-Tacro (life-cycle-pharma melt-extrusion tablets) should register a sibling canonical (e.g., FORM_TAC_ENVARSUS) rather than overloading FORM_TAC_IR.

FORM_TAC_ENVARSUS (canonical for the LCP-Tac / Envarsus MeltDose extended-release tacrolimus formulation indicator)

• Description: 1 = subject (or occasion) received once-daily LCP-Tac, the life-cycle-pharma MeltDose melt-extrusion extended-release tacrolimus tablet marketed as Envarsus / Envarsus XR (Chiesi Farmaceutici in Europe, Veloxis in the US); 0 = subject received a non-LCP-Tac oral tacrolimus product, in practice the twice-daily immediate-release formulation IR-Tac (Prograf, Astellas). Per-subject in parallel-group designs and per-occasion in the within-patient conversion designs that dominate this literature. MeltDose technology raises oral bioavailability and spreads release along the whole gastrointestinal tract to the colon, which is why LCP-Tac needs a lower daily dose than IR-Tac for equal exposure.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (IR-Tac immediate-release Prograf; the F = 1 reference arm is selected jointly with CYP3A5_EXPR in Mohammed Ali 2025, where the LCP-Tac nonexpresser group is the anchor).
• Source aliases:
  • formulation – used in MohammedAli_2025_tacrolimus.R (Mohammed Ali 2025 Methods section 2.2; patients were converted from Prograf to Envarsus, giving one IR-Tac and one LCP-Tac 24 h profile each).
• Example models: MohammedAli_2025_tacrolimus.R (switches three quantities at once: the first-order absorption rate constant (ka 0.111/h on LCP-Tac vs a 2.04/h mesor on IR-Tac), the absorption lag time (1.4 h vs 0.465 h), and the relative bioavailability F crossed with CYP3A5_EXPR (Mohammed Ali 2025 Table 3: F = 1 FIX for LCP-Tac nonexpressers, 0.745 for IR-Tac nonexpressers, 0.693 for LCP-Tac expressers, 0.427 for IR-Tac expressers). The circadian rhythm on the absorption rate constant is IR-Tac-specific and is gated by 1 - FORM_TAC_ENVARSUS).
• Notes: The distinction is load-bearing and not cosmetic: FORM_TAC_IR = 0 means the Advagraf / Astagraf XL prolonged-release capsule, which is a different once-daily product from LCP-Tac with different absorption characteristics and a different conversion ratio from IR-Tac (1:1 for Advagraf versus 1:0.7 for Envarsus in the European guidelines), so a dataset that pooled Advagraf and Envarsus into one “prolonged-release” level would confound two distinct products. A study covering all three oral products encodes them with the two columns jointly: FORM_TAC_IR = 1 selects Prograf, FORM_TAC_ENVARSUS = 1 selects Envarsus, and both = 0 selects Advagraf / Astagraf XL. Mirrors the FORM_* family of drug-product-version indicators (FORM_TAC_IR, FORM_VINP_IR, FORM_SAR_DP2, FORM_ISA_P2F2, FORM_LINAG_TAB1, FORM_VISMO_PHASEI). Scoped general because the IR-Tac-to-LCP-Tac conversion is a routine clinical manoeuvre in solid-organ transplantation and the contrast recurs across the tacrolimus popPK literature (ASERTAA, and this group’s own earlier LCP-Tac analyses). Frequently interacts with CYP3A5_EXPR: Mohammed Ali 2025 found the bioavailability gain on conversion to be genotype-dependent, which is the basis for its genotype-specific conversion ratios (1:0.6 in CYP3A5*1 expressers, 1:0.7 in nonexpressers).

FORM_VINP_IR (canonical for vinpocetine immediate-release Cavinton tablet formulation indicator)

• Description: 1 = subject received the Cavinton immediate-release 5 mg vinpocetine tablet (Organon / Gedeon Richter Ltd., Budapest, Hungary; the reference immediate-release solid dosage form of vinpocetine); 0 = subject received the Ultra Vinca sustained-release beta-cyclodextrin-complex 10 mg vinpocetine tablet (Tecnimede, Portugal; the Petric 2023 reference formulation). Per-dose-occasion indicator (used in the Petric 2023 crossover design where each subject received all three formulations across separate occasions with a 7-day washout).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (Ultra Vinca sustained-release beta-cyclodextrin complex tablet; the typical-value Tk0 and V1/F reference in Petric 2023 Table 1).
• Source aliases:
  • Formulation#2 – the second level of the three-level Formulation categorical covariate in Petric 2023 Table 1; maps directly to FORM_VINP_IR = 1 with Formulation#1 (Ultra Vinca SR) as the reference.
• Example models: Petric_2023_vinpocetine.R (log-additive effect on Tk0 (exp(beta = -0.4) = 33% shorter zero-order absorption duration relative to the Ultra Vinca SR reference) and on V1/F (exp(beta = -1.26) = 72% lower apparent central volume relative to the SR reference, consistent with a higher vinpocetine oral bioavailability / metabolite yield for the Cavinton IR arm; Petric 2023 Table 1 beta_Tk0_Formulation#2 = -0.4 (RSE 29.0%) and beta_V1/F_Formulation#2 = -1.26 (RSE 5.44%))…………….
• Notes: Scoped specific because the Cavinton-IR-vs-Ultra-Vinca-SR contrast is tied to the Petric 2023 relative-bioavailability crossover design (a three-level formulation stratification of vinpocetine as its main active metabolite apovincaminic acid). Paired with the general FORM_SOLUTION canonical (both = 0 selects the Ultra Vinca SR reference; FORM_SOLUTION = 1 selects the 10 mg / 5 mL oral solution) to encode the full three-level factor. Sibling to the other drug-specific FORM_*_IR / FORM_*_<formulation> entries (FORM_TAC_IR, FORM_LINAG_TAB1, FORM_SAR_DP2, FORM_ISA_P2F2, FORM_VISMO_PHASEI) under the FORM_* family of drug-product-version indicators. Future vinpocetine-formulation extractions comparing other vinpocetine products against Ultra Vinca SR (or Cavinton IR) should reuse this canonical and extend the example list; future SR-vs-IR contrasts for unrelated drugs should register their own drug-specific sibling canonical rather than overloading this entry.

FORM_XYNTHA (canonical for Xyntha vs Refacto / Refacto AF moroctocog-alfa product indicator)

• Description: 1 = Xyntha (Wyeth Pharmaceuticals / Pfizer; OSA-calibrated potency label, marketed in the United States, Canada, and other regions); 0 = Refacto or Refacto AF (CSA-calibrated potency label, marketed in the European Union and other regions). Per-observation (per-row) binary indicator. Moroctocog alfa is the same B-domain-deleted recombinant factor VIII active moiety in all three products; Xyntha and Refacto AF are the same drug product with different potency-assay calibrations (OSA vs CSA) for label dosing. Used as a multiplicative effect on bioavailability F to harmonise the model-predicted FVIII activity across products.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (Refacto or Refacto AF; CSA-calibrated potency label).
• Source aliases:
  • PROD – used in Abrantes_2017_moroctocog.R (Abrantes 2017 Table 2 footnote g and Methods: F multiplier 1.38^PROD so Xyntha effectively delivers 1.38x the labelled dose in CSA-reference units).
• Example models: Abrantes_2017_moroctocog.R (multiplicative effect on F: (1 + 0.38 * FORM_XYNTHA) – equivalent to 1.38^FORM_XYNTHA when binary; the model anchors F = 1 for the CSA-reference Refacto/Refacto AF products and applies the +38.0% multiplier when the product administered is Xyntha).
• Notes: Specific scope because the contrast is tied to the Pfizer / Wyeth moroctocog-alfa product line (Refacto, Refacto AF, and Xyntha). Mirrors the FORM_<drug>_<variant> family pattern (FORM_LEB_NS0 lebrikizumab cell-line, FORM_SAR_DP2 sarilumab drug-product version, FORM_DOX_DORYX_MPC doxycycline Doryx-MPC). Per-observation (per-row) when a subject could switch between products across trials (e.g., Abrantes 2017 study B1831066 enrolled both Refacto and Refacto AF in the same protocol); subject-level otherwise. The 1.38 OSA-vs-CSA potency ratio for B-domain-deleted FVIII is documented in Mikaelsson 1998 and Hubbard 2013 (cited in Abrantes 2017 reference 13).

FORM_FVIII_BDD (canonical for B-domain-deleted (vs full-length) recombinant factor VIII product indicator)

• Description: 1 = the administered factor VIII (FVIII) concentrate is a B-domain-deleted recombinant product (e.g., Refacto, Refacto AF, Xyntha – moroctocog alfa); 0 = the administered product is a full-length recombinant FVIII (e.g., Kogenate FS, Helixate FS, Advate, Recombinate) or a plasma-derived FVIII (e.g., Aafact, Hemofil M). Per-observation (per-row) binary indicator. B-domain-deleted recombinant FVIII proteins retain the FVIII coagulation activity but omit the heavily glycosylated central B-domain; the one-stage clotting assay (OSA) commonly used clinically under-detects B-domain-deleted FVIII activity in plasma by a well-documented ~30-50 % relative to full-length FVIII (mechanism: sub-optimal thrombin-mediated activation kinetics in the OSA reagent milieu), so PK models that pool observations across BDD and full-length products commonly include this indicator as a multiplicative correction on the predicted plasma concentration to reconcile the assay artefact. Note: Xyntha vs Refacto/Refacto AF is a separate potency-calibration contrast within the BDD product family (see FORM_XYNTHA); the two canonicals are independent and can coexist in a single dataset.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (full-length recombinant or plasma-derived FVIII product).
• Source aliases:
  • bdp (paper-prose 0/1 indicator, 1 = B-domain-deleted product) – used in Hazendonk_2016_factor_viii.R (Hazendonk 2016 Methods: C_pred,bdp = C_pred * (1 - theta_bdp) with theta_bdp = 0.34 = 34 % assay-driven under-prediction of the B-domain-deleted product Refacto AF).
• Example models: Hazendonk_2016_factor_viii.R (multiplicative effect on the predicted FVIII plasma concentration: Cc = (central / vc) * (1 - theta_bdp * FORM_FVIII_BDD); reference category 0 = full-length recombinant or plasma-derived FVIII).
• Notes: General scope because the OSA vs CSA assay artefact for B-domain-deleted FVIII is a class-level phenomenon that recurs across BDD-product FVIII popPK models. Distinct from FORM_XYNTHA (which contrasts Xyntha vs Refacto / Refacto AF within the BDD moroctocog product family, capturing the +38 % OSA-vs-CSA potency calibration difference; a subject may have FORM_FVIII_BDD = 1 and FORM_XYNTHA = 0 or 1 depending on which Pfizer moroctocog product they received). Distinct also from ASSAY_OSA (which is the bioanalytical-method indicator: OSA vs CSA); when both FORM_FVIII_BDD and ASSAY_OSA are available, the correct interaction structure is FORM_FVIII_BDD * ASSAY_OSA (only OSA-assayed BDD-product samples show the under-detection), but many perioperative datasets are OSA-only and collapse the correction onto FORM_FVIII_BDD alone (Hazendonk 2016). Per-observation (per-row) when subjects can switch between products across surgical procedures; subject-level otherwise. Mirrors the FORM_<drug>_<variant> family pattern.

FORM_FVIII_FC (canonical for the Fc-fusion (efmoroctocog alfa, Elocta/Eloctate) vs standard half-life factor VIII product indicator)

• Description: 1 = the administered factor VIII (FVIII) concentrate is recombinant FVIII Fc fusion protein (rFVIIIFc, efmoroctocog alfa; Elocta in Europe, Eloctate in North America), an extended half-life (EHL) product; 0 = a standard half-life FVIII concentrate (full-length or B-domain-deleted recombinant, or plasma-derived). Per-dose-record binary indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (standard half-life FVIII product).
• Source aliases: EHL (Allard 2020 ‘EHL (elocta:1 versus others:0)’; Elocta was the only EHL product in that dataset, so FORM_FVIII_FC = EHL).
• Example models: Allard_2020_factorviii.R (log-scale shift on CL, cl = exp(lcl + etalcl + e_form_fviii_fc_cl * FORM_FVIII_FC) * ... with e_form_fviii_fc_cl = -0.394, i.e. 33% lower CL and a ~1.48-fold longer terminal half-life on Elocta).
• Notes: Auto-approved member of the FORM_<drug>_<formulation> family. Named for the Fc-fusion product rather than for “extended half-life” as a class: other EHL FVIII products (PEGylated rurioctocog alfa pegol, turoctocog alfa pegol, damoctocog alfa pegol, efanesoctocog alfa) prolong half-life by different mechanisms and to different extents, so a model fitted with Elocta as its only EHL product does not describe them. Register a sibling FORM_FVIII_<product> indicator for those. Time-varying within subject when patients switch products; set it on each dose record. Independent of FORM_FVIII_BDD (rFVIIIFc is itself B-domain deleted, but the assay-artefact contrast and the half-life contrast are distinct).

DILUTION (canonical for diluted-drug-product indicator)

• Description: 1 = drug diluted (Soehoel 2022 study D2213C00001), 0 = not diluted.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (not diluted).
• Source aliases:
  • dilution (same orientation, no value transformation) – the spelling used in Soehoel_2022_tralokinumab.R’s source paper, recorded in that model’s source_name field.
• Example models: Soehoel_2022_tralokinumab.R (multiplicative effects on bioavailability and absorption rate: fdepot = exp(lfdepot) * (1 + e_dilution_fdepot * DILUTION) and ka = exp(lka) * (1 + e_dilution_ka * DILUTION)).
• Notes: Renamed from the source’s lower-case dilution on 2026-09-09, when the ALL-CAPS convention became machine-enforced; the covariate-effect parameters keep the lower-cased covariate token (e_dilution_<param>) as the whole e_<cov>_<param> family does.

NON_ECZTRA (canonical for non-ECZTRA-trial indicator)

• Description: 1 = not the ECZTRA trial; 0 = ECZTRA.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (enrolled in an ECZTRA phase 3 trial).
• Source aliases:
  • nonECZTRA (same orientation, no value transformation) – the mixed-case spelling used in Soehoel_2022_tralokinumab.R’s source paper, recorded in that model’s source_name field.
• Example models: Soehoel_2022_tralokinumab.R (multiplicative effects on clearance and central volume: cl = ... * (1 + e_non_ecztra_cl * NON_ECZTRA) and vc = ... * (1 + e_non_ecztra_vc * NON_ECZTRA)).
• Notes: Renamed from the source’s mixed-case nonECZTRA on 2026-09-09, when the ALL-CAPS convention became machine-enforced. NON_ECZTRA was taken over STUDY_NON_ECZTRA because the model file’s own rename note named it; the STUDY_* family names the cohort a subject IS in, whereas this indicator is the complement of one.

SEASON2 (canonical for second RSV season at dosing indicator)

• Description: 1 = second RSV season at dosing, 0 = first RSV season.
• Units: (binary)
• Type: binary
• Scope: specific
• Example models: Clegg_2024_nirsevimab.R.
• Notes: Study-specific but semantically general (second-exposure indicator). Promote to general if a second RSV-season model adopts the same semantics.

COHDOSE (canonical for randomized dose cohort (mg/kg))

• Description: Randomized dose cohort expressed in mg/kg. Subject-level (time-fixed) covariate carrying the per-subject cohort dose in a study where each subject remained on a single escalating-cohort dose for the full dosing period.
• Units: mg/kg
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (COHDOSE / ref)^exponent. Reference values observed: 1 mg/kg in Narwal 2013 (human IgG1), 3 mg/kg in Niloy 2026 (preclinical mouse small molecule).
• Source aliases:
  • DOSE – used in Narwal_2013_sifalimumab.R (the paper’s Eq. 3 variable name; renamed to COHDOSE here to avoid colliding with the rxode2/nlmixr2 event-column convention where DOSE or AMT carries the administered dose).
  • Dose_i – used in Niloy_2026_MTMSATrp_mouse.R (Niloy 2026 Methods Eq. 2 notation; the per-mouse mg/kg cohort dose).
• Example models: Narwal_2013_sifalimumab.R (reference 1 mg/kg, exponent 0.0542 on CL), Niloy_2026_MTMSATrp_mouse.R (preclinical mouse; reference 3 mg/kg, exponent -0.30 on CL describing empirical dose-dependent CL over 0.3-10 mg/kg).
• Notes: Scope promoted from specific to general on 2026-06-22 with the Niloy 2026 MTMSA-Trp extraction (second model ratifying the canonical, this time as a preclinical mouse small-molecule dose-cohort covariate with a negative exponent on CL describing decreasing apparent clearance with increasing dose – complementing Narwal 2013’s positive exponent in a Phase Ib human mAb context). The shared functional form (COHDOSE / ref)^exponent on apparent clearance and the shared escalating-cohort design (each subject on one dose for the full study) carry across drug class, species, and study size. For fixed-dose simulations in a weight-based-dosing paper, set COHDOSE = nominal_dose_mg / WT per subject. When the subject receives a per-kg dose label directly (preclinical mg/kg dosing or human mg/kg cohort labels like the MI-CP152 0.3 / 1 / 3 / 10 mg/kg arms), COHDOSE is the mg/kg label.

DOSE (canonical for current administered dose level supplied as a data column)

• Description: Continuous covariate carrying the administered dose level (in mg) as a per-record data column. Two complementary use cases share this canonical:
  1. Per-subject assigned dose – each subject’s fixed assigned dose level (in mg) across the study, used as a power-style or stratified covariate when the population PK model detects a dose-dependent shift in a PK parameter (central volume, clearance, etc.).
  2. Time-varying current administered dose – the current daily dose at the time of the record, used in PD-only models that derive a per-cycle exposure metric (e.g., AUC = DOSE / CLI) from a posthoc-CL covariate without instantiating a PK ODE. The column is set to 0 during off-treatment periods (drug holidays, placebo arm) so the derived exposure becomes 0.
• Units: mg (document per-model in covariateData[[DOSE]]$units if a different dose unit is used).
• Type: continuous
• Reference category: n/a – used with power scaling (DOSE / ref)^exponent for use case (a), or directly inside derived-exposure expressions for use case (b). Reference values observed: 600 mg in Zheng_2016_sifalimumab.R (middle of the 200/600/1200 mg phase IIb dose range).
• Source aliases:
  • Dose – used in Zheng_2016_sifalimumab.R and deCastroSuarez_2020_nimotuzumab.R.
  • DOS – used in the Hansson 2013 sunitinib biomarker / TGI / fatigue PD-model family (DDMODEL00000197 and siblings) as a per-record sunitinib dose column.
• Example models: Zheng_2016_sifalimumab.R (power effect on V1 with exponent 0.06), deCastroSuarez_2020_nimotuzumab.R (binary-indicator usage (DOSE == 50) applying a 53 % decrease in V1 for the 50 mg cohort), Hansson_2013a_sunitinib.R (DDMODEL00000197; time-varying record-level dose feeding AUC = DOSE / CLI), Hansson_2013b_sunitinib.R (DDMODEL00000198; same time-varying AUC = DOSE / CLI form for the tumor growth inhibition model), Schindler_2016_sunitinib.R (DDMODEL00000221; same AUC = DOSE / CLI form, with the daily-dose column toggling between 50 mg/day on-cycle and 0 on dose-holiday records), Schindler_2017_imatinib.R (time-varying daily imatinib dose in mg/day feeding the size and density drug-effect terms as Kdrug,S * (DOSE / 400) * exp(-k * t) and Kdrug,D * (DOSE / 400), with 400 mg/day as the reference normalisation), Girard_2012_pimasertib.R (linear coefficient on the dropout-hazard log-rate: exp(beta * DOSE) Weibull multiplier; per-subject daily dose, observed range 1-255 mg), Wada_2023_sparsentan.R (use case (a): per-record administered dose drives a dose-dependent relative bioavailability Frel = (max(DOSE, 200) / 400)^-0.495 applied as f(depot), giving Frel = 1.41 / 1.00 / 0.71 at 200 / 400 / 800 mg – less-than-dose-proportional exposure, with the 200 mg clamp reproducing the paper’s published lower branch), Comisar_2025_rimegepant.R (use case (a) with BOTH a continuous and a categorical use of the same column in one model: a power term on relative bioavailability f(depot) <- (DOSE/10)^0.191 * ... reproducing greater-than-dose-proportional exposure over 10-150 mg attributed to dose-dependent autoinhibition of CYP3A first-pass metabolism, plus a categorical low-dose indicator on the transit rate constant, ktr * (1 + 0.536 * (DOSE < 50)), which reproduces the control stream’s IF(DOSE.EQ.10) / IF(DOSE.EQ.25) KTRDOSE = (1 + THETA(29)) exactly over the four dose levels studied (10, 25, 75, 150 mg) with 75 mg as the reference), FergusonSells_2022_tadalafil.R (use case (a), time-varying within a patient: pediatric patients stepped from a low to a high once-daily dose, and the per-record dose drives f(depot) <- (DOSE/16.27)^-0.227 * (AGE/52.4)^0.100; the 16.27 mg centring value is printed in the paper’s equation without a stated origin), Irie_2021_favipiravir.R (time-varying LAST ADMINISTERED dose re-evaluated at every dose, power effect on CL/F (DOSE/600)^-0.61 with 600 mg the maintenance dose, reproducing the fall in exposure after the 1600/1800 mg loading day; control-stream data item LAST, carried in micrograms there and converted to mg), Kim_2020_CKD_519.R (per-subject assigned once-daily dose in mg entering the relative bioavailability of an oral CETP inhibitor as Bmax * (1 - DOSE / (BA50 + DOSE)), Kim 2020 Equation 2 with BA50 = 90.1 mg), Jiang_2021_ivosidenib.R (dose per administration, power effect (DOSE / 500)^-0.49 on relative bioavailability of oral ivosidenib), Li_2022_paroxetine.R (use case (a), time-varying within a titrated patient: total daily paroxetine dose in mg/day as a power effect on CL/F, (DOSE / 40)^-1.03, reproducing the greater-than-dose-proportional rise in trough concentration over 20-60 mg/day).
• Notes: Distinct from DOSE_70MG (binary indicator for a specific dose group in a trinary-dose design) and from the rxode2/nlmixr2 event column amt (which carries the administered dose at dose events). For use case (a), the values are typically time-fixed per subject; for use case (b), the values are time-varying with on/off cycling – for sunitinib 4-weeks-on / 2-weeks-off cycling, set DOSE = nominal_daily_mg (e.g., 50) during on-cycles and 0 during off-cycles or for the placebo arm. Per-model covariateData[[DOSE]]$notes should state which use case applies.

NDOSEUNITS (canonical for the number of solid oral dosage units administered in a dose)

• Description: Count of discrete dosage units (capsules, tablets, sachets, vials) making up a single administered dose, supplied as a per-dose-record data column. Deliberately formulation-neutral: the modelled quantity is “how many units were swallowed”, so one canonical serves any paper counting any unit type rather than spawning a per-formulation family. It exists because unit count is a covariate in its own right and is not recoverable from DOSE whenever unit strength varies across the pooled studies – the same 100 mg dose may be one 100 mg unit in one study and four 25 mg units in another. The mechanism is usually a per-unit excipient load: more units means more co-administered surfactant, disintegrant or solubiliser, which raises the bioavailability of a poorly soluble drug independently of the milligrams delivered.
• Units: (count)
• Type: continuous
• Scope: general
• Reference category: n/a. Where a model centres the count, document the centring constant per model. Reference values observed: 5.5 units, the median count among fasted subjects in Krishna_2011_anacetrapib.R.
• Source aliases:
  • N_CAP, NCAP – Krishna 2011 Eqs. 5 and 7 (“the number of capsules”); same orientation as the canonical, no value transformation.
• Example models: Krishna_2011_anacetrapib.R (two simultaneous effects on bioavailability, both gated on FORM_ANACETRAPIB_LFC because the count is only defined for the liquid-filled-capsule formulation: an Emax term 1 - ndoseunits_50 / (ndoseunits_50 + NDOSEUNITS) with ndoseunits_50 = 0.67 units acting in every prandial state, and, in fasted subjects only, the exponential Krishna 2011 Eq. 5 term exp(e_ndoseunits_fdepot * (NDOSEUNITS - 5.5)) with e_ndoseunits_fdepot = 0.07 per unit. Krishna 2011 Population PK Model states the rationale outright: “The number of capsules was considered as a surrogate for the amount of liquid surfactant contained in the capsule unit making up a given dose.” Together the two terms reproduce the paper’s four published unit-count predictions exactly – a two-capsule dose raises bioavailability 25% in fed and 34% in fasted subjects, a six-capsule dose 50% and 113%).
• Notes: Distinct from DOSE (the milligrams administered) and from the FORM_* family (which formulation, not how many units of it). When a model carries both a FORM_* indicator and NDOSEUNITS, state in covariateData[[NDOSEUNITS]]$notes whether the count terms are gated on the formulation: unit count is often meaningless for one arm of a bridging study (a single tablet has no per-capsule excipient load), in which case the FORM_* indicator multiplies the count terms out and the column’s value on those records is immaterial – but supply the physical count anyway so the data stay self-describing. General scope because “number of dosage units per dose” carries the same meaning across drugs, formulations and studies. It enters nonlinear terms (Emax in the count, exponentials in the centred count), so it is treated as a continuous covariate rather than decomposed into per-level binary indicators: the register’s count-decomposition policy applies to counts a paper has categorically binned, not to a count entering a continuous functional form.

CD (canonical for cumulative cladribine dose (time-varying))

• Description: Time-varying cumulative cladribine dose (mg total dose, not body-weight-normalized) administered to each subject up to the current observation time. Stays at zero during the placebo arm and during the pre-dose baseline phase, rises stepwise across the dosing schedule (cladribine is given as short oral pulses), and remains constant between dose events.
• Units: mg
• Type: continuous
• Scope: specific
• Reference category: n/a – used inside EXPS = CD * 104.5 / CRL as an exposure surrogate driving an Emax-style symptomatic effect on disease progression, not as a power-form covariate.
• Source aliases: CD – column name used in the DDMODEL00000223 input dataset (Simulated_Novakovic_2016_multiplesclerosis_cladribine_irt.csv).
• Example models: Novakovic_2017_cladribine.R.
• Notes: Distinct from DOSE (per-subject assigned dose level, time-fixed) and COHDOSE (mg/kg cohort label, time-fixed). CD is the cumulative dose accrued at each timepoint, supplied as a time-varying covariate column rather than via dosing events because the Novakovic 2017 model does not carry an explicit cladribine-PK compartment. Scope: specific because the constant 104.5 inside the exposure-surrogate equation is hard-coded for cladribine in the source.

TRT (canonical for treatment-cohort indicator)

• Description: Per-subject treatment-cohort / treatment-arm integer indicator. Carries the categorical assignment that selects per-arm structural parameters (e.g., per-arm typical tumor-growth rate, per-arm dose-response slope, on-treatment vs placebo gating). Value coding is per-model: each consuming model defines its own integer levels via covariateData[[TRT]]$notes, because the same TRT column header carries different cohort semantics in different studies.
• Units: (categorical / integer-coded)
• Type: categorical
• Scope: general
• Reference category: per-model (each consuming model declares its own reference level in covariateData[[TRT]]$notes).
• Source aliases: TRT, ARM, TRTARM, STUDY_ARM – common column-name variants for the per-subject treatment-arm integer indicator; all map directly to canonical TRT with no value transformation. Per-model integer coding is documented inline.
• Example models: Novakovic_2017_cladribine.R.
  • Novakovic_2017_cladribine.R (DDMODEL00000223): 0 = placebo, 1 = cladribine 3.5 mg/kg cumulative-dose cohort, 2 = cladribine 5.25 mg/kg cumulative-dose cohort. Reference = 0 (placebo). Gates the symptomatic and protective drug-effect terms via TRT >= 1 && t > 0; the categorical level (1 vs 2) is informational because the dose-response is driven by the time-varying CD covariate and the per-cohort dosing schedule, not by TRT itself.
  • Struemper_2025_tumorsize_OS_nsclc.R: 1 = PEMBRO (pembrolizumab, INTR@PID LUNG 037, n = 152), 2 = FELAD (feladilimab, INDUCE-1, n = 52), 3 = CHEMO (docetaxel, Entree Lung Part 2, n = 34), 4 = FELAD+CHEMO (pooled INDUCE-1 + Entree Lung Part 2, n = 78), 5 = IO-COMBO (feladilimab + IO; pooled INDUCE-1 + INDUCE-2, n = 23), 6 = DOSTAR (dostarlimab, GARNET, n = 67), 7 = DOSTAR+CHEMO (PERLA, n = 121), 8 = PEMBRO+CHEMO (PERLA, n = 122), 9 = COBO100MG+DOSTAR cohort B (AMBER, n = 14), 10 = COBO300MG+DOSTAR cohort B (AMBER, n = 41), 11 = COBO900MG+DOSTAR cohort B (AMBER, n = 29), 12 = COBO300MG+DOSTAR cohort D (AMBER, n = 53). Reference = 1 (PEMBRO; the largest single-agent PD-1 cohort and the model’s canonical default). Selects per-arm TVKG and TVKS via a 12-branch if-else cascade and gates the PDL1_TUM effect on ks via a derived has_pd1 indicator equal to 1 when TRT is 1 or in {6, 7, 8, 9, 10, 11, 12}.
  • Stringer_2014_sipoglitazar.R: 0 = placebo (n = 136), 1 = sipoglitazar at any regimen other than 32 mg twice daily (8 mg QD, 16 mg QD, 16 mg BID, 32 mg QD, 64 mg QD), 2 = sipoglitazar 32 mg twice daily, 3 = rosiglitazone 8 mg once daily (n = 72); sipoglitazar arms total n = 572. Reference = 0 (placebo). Carries three distinct pieces of model structure at once: level 2 selects the separately estimated FPG baseline for the 32 mg BID arm (9.02 vs 9.41 mmol/L, Stringer 2014 Table 2 footnote a), level 3 gates the fixed 28% rosiglitazone step effect ROTE on KoutG (no rosiglitazone concentrations were collected, so the comparator effect cannot be exposure-driven), and the 0 / 1-2 / 3 partition selects the HbA1c lifestyle effect LEHB (0.037 placebo, 0.020 sipoglitazar, 0 rosiglitazone). The sipoglitazar DOSE LEVEL is deliberately NOT carried by TRT – it is carried by DOSE_SIPOGLITAZAR_MGD, so a level-1 or level-2 subject must also have a non-zero daily dose set.
  • Zou_2026_pembrolizumab_qol_mbma.R: 0 = pembrolizumab arm (monotherapy or combination), 1 = control arm (the trial’s comparator – chemotherapy, targeted therapy, placebo plus standard of care, or placebo alone; Zou 2026 Section 2.2 pools them all into one level). Reference = 0 (pembrolizumab), which is the REVERSE of the placebo-reference codings above and is why both retained effects are negative. Unlike the three models above this is a STUDY-ARM-level indicator in a model-based meta-analysis, not a per-subject assignment; it shifts logit(Emax) by -0.758 and log(SLP) by -1.4.
• Notes: Four models now use the TRT canonical with different integer codings: Novakovic 2017 (3 levels, placebo reference), Struemper 2025 (12 levels, pembrolizumab reference), Stringer 2014 (4 levels, placebo reference) and Zou 2026 (2 levels, pembrolizumab reference, study-arm rather than per-subject). The column header is shared because both source NONMEM control streams use TRT; the integer-to-cohort mapping is intentionally per-model. Scope upgraded from specific to general on 2026-06-22 when Struemper 2025 added a second consuming model. Future models adding new TRT codings extend the Example-models list rather than registering a new canonical, so long as the column header is TRT; a model that uses a fundamentally different semantics (e.g., a generic on-treatment 0/1 flag distinct from per-arm cohort indexing) should register a new canonical name (e.g., ON_TREATMENT) rather than reusing TRT.

DRUG_ORMU (canonical for Ormutivimab vs HRIG drug-product comparator indicator)

• Description: 1 = subject received Ormutivimab (a recombinant human anti-rabies IgG1 monoclonal antibody, also referred to in the source paper as rHRIG); 0 = subject received plasma-derived human rabies immunoglobulin (HRIG) comparator (or placebo + vaccine without passive antibody). Per-subject (time-fixed) categorical indicator carrying the head-to-head drug-arm assignment in a randomized rabies-vaccine pharmacodynamics study where rabies virus neutralizing antibody (RVNA) activity is modelled with a time-dependent Emax response to the vaccine and the passive antibody product modifies the typical-value Emax / ET50.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (HRIG / no passive antibody).
• Source aliases:
  • antibody type – Zhang 2022 Results section 3.4 covariate-effect prose (“antibody type was determined as the covariate significantly affecting the model”); the source NMTRAN column name is not separately reported.
• Example models: Zhang_2022_ormutivimab.R (additive typical-value shifts on the linear-scale Emax and ET50 of the vaccine-induced RVNA Emax model: Emax_tv = exp(lEmax) + e_drug_ormu_Emax * DRUG_ORMU with e_drug_ormu_Emax = +0.143 IU/mL and ET50_tv = exp(lET50) + e_drug_ormu_ET50 * DRUG_ORMU with e_drug_ormu_ET50 = -3.8 day, yielding a higher and faster vaccine-induced antibody peak in the Ormutivimab arms relative to the HRIG comparator).
• Notes: Distinct from the FORM_* family (within-product formulation-version contrasts of a single drug) because the contrast here is between two biologically distinct products: HRIG is a polyclonal plasma-derived immunoglobulin, while Ormutivimab is a recombinant monoclonal antibody (CHO-cell-produced; the first rhRIG approved in China). Specific scope because the head-to-head HRIG-vs-rHRIG comparator design is tied to the Zhang 2022 phase II rabies-vaccine study. Future head-to-head rhRIG-vs-HRIG popPD models (e.g., SII Rabishield or Twinrab against HRIG) should register a sibling canonical (DRUG_SIIRMAB, DRUG_TWINRAB) rather than overloading DRUG_ORMU; cross-product comparisons that need both indicators in the same dataset can carry them as independent binaries with HRIG as the shared reference.

DRUG_PRED (canonical for prednisone vs fosdagrocorat drug-arm comparator indicator)

• Description: 1 = subject is in the prednisone (oral glucocorticoid comparator) arm; 0 = subject is in the fosdagrocorat (PF-04171327, dissociated agonist of the glucocorticoid receptor) arm or the placebo arm. Per-subject (time-fixed) binary indicator carrying the head-to-head drug-arm assignment in a phase II randomized trial in adults with rheumatoid arthritis where serum bone-formation biomarkers (P1NP, osteocalcin) are modelled with a K-PD framework and the two drugs share the rebound and response-side parameters but carry separate KDE / Imax / EDK50 values.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (fosdagrocorat or placebo).
• Source aliases: derived per subject from the trial-arm assignment; the source paper does not name an NMTRAN column.
• Example models: Shoji_2017_fosdagrocorat_p1np.R, Shoji_2017_fosdagrocorat_oc.R (binary multiplier switching the active KDE and the active sigmoid-Emax inhibition parameters between fosdagrocorat and prednisone: lkde_active = lkde_fos * (1 - DRUG_PRED) + lkde_pred * DRUG_PRED, with analogous switching of logitimax_active / imax_active and ledk50_active; the rebound parameters RBmax, T50 and the response-side parameters Kd, BL, SLP are shared between the drugs and do not multiply by DRUG_PRED).
• Notes: Sibling of DRUG_ORMU (rhRIG vs HRIG) and follows the same “head-to-head drug-arm” pattern: the contrast is between two structurally distinct active comparators in a randomized trial. Specific scope because the fosdagrocorat-vs-prednisone head-to-head is tied to the Shoji 2017 P1NP / OC analyses; future DAGR-vs-prednisone (or DAGR-vs-other glucocorticoid) popPK/PD models can extend this entry’s example_models list, while a contrast between a different test drug and prednisone (e.g., methylprednisolone vs prednisone) should register a sibling canonical rather than reusing DRUG_PRED.

DRUG_BEV (canonical for bevacizumab add-on treatment-arm indicator)

• Description: 1 = subject received bevacizumab in addition to the trial’s chemotherapy backbone; 0 = subject received the chemotherapy backbone alone. Per-subject (time-fixed) binary indicator carrying the add-on randomisation in trials where bevacizumab is given on top of a shared cytotoxic regimen rather than head-to-head against a comparator drug. Used in non-PK / non-exposure survival and disease-progression models where the bevacizumab effect enters the structural equations as a categorical on/off term rather than as an exposure-driven response.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (chemotherapy backbone alone, no bevacizumab).
• Source aliases:
  • Bevacizumab – Claret 2014 (J Clin Pharmacol 54(3):253-257) Table 3 covariate row label; the source arms are “Bev + Chemo” versus “Chemo” in the E4599 study (Claret 2014 Table 1). Coded 1/0.
  • BEV / AVASTIN – common NONMEM column-name variants for a bevacizumab-arm flag.
• Example models: Claret_2014_DCR_PFS_nsclc.R (additive effect on the log-normal AFT location parameter of progression-free survival: mu_pfs += e_bev_mu_pfs * DRUG_BEV with e_bev_mu_pfs = 0.232, i.e. median PFS is exp(0.232) = 1.26 times longer on bevacizumab, over and above the effect already captured by week-8 disease control), Claret_2014_DCR_OS_nsclc.R.
• Notes: Member of the DRUG_<INN> family alongside DRUG_ORMU, DRUG_PRED and DRUG_OBI. Those three carry head-to-head contrasts between two named active products; DRUG_BEV carries an add-on contrast (drug plus backbone versus backbone alone), which is the more common oncology randomisation. The distinction matters for the reference category: the DRUG_BEV reference arm receives active chemotherapy, so this is NOT interchangeable with ON_TREATMENT (active versus placebo) or PLACEBO (placebo-arm membership). Where a source paper pools several bevacizumab-containing arms with differing backbones, document the backbone regimens per arm in covariateData[[DRUG_BEV]]$notes; where a paper differentiates bevacizumab dose levels with distinct parameters, register a per-arm categorical (TRT) instead of overloading this binary.

DOSE_BALCINRENONE_MG (canonical for administered balcinrenone per-administration dose amount)

• Description: Continuous per-dose-record covariate carrying the balcinrenone (AZD9977) dose amount, in mg, associated with the current record. Used when a balcinrenone popPK model makes a PK parameter an explicit function of the administered dose level rather than treating dose only as an event amount – in the founding model, a dose-dependent power function on the first-order absorption rate constant capturing solubility-limited absorption.
• Units: mg
• Type: continuous
• Scope: specific
• Reference category: n/a – used with power scaling (DOSE_BALCINRENONE_MG / ref)^exponent. Reference values observed: 150 mg (Parkinson 2025, the single dose used in the renal-impairment study NCT04469907 and the normalisation anchor at which the reported KA = 0.381 1/h applies).
• Source aliases:
  • DOSE – NONMEM $PK symbol used in Parkinson_2025_balcinrenone.R (Parkinson 2025 Supplementary Materials ‘Final NONMEM Model’: KADOSE = (DOSE/150)**THETA(11); Table 1 footnote Dose KA = (dose/150)**KA~Dose).
• Example models: Parkinson_2025_balcinrenone.R (power effect on the absorption rate constant, ka *= (DOSE_BALCINRENONE_MG / 150)^-0.262, so a 50 mg dose absorbs about 34% faster than the 150 mg reference and a 300 mg dose about 17% more slowly; the paper reports a correspondingly 33% higher dose-normalised Cmax at 50 mg than at 300 mg).
• Notes: A dedicated mg column is required for two independent reasons. First, the general rxode2 constraint documented under DOSE_NMV_MG: a data column literally named DOSE is consumed by rxode2’s event translation, so a model referencing DOSE in model() fails to solve. Second, and specific to this model, Parkinson_2025_balcinrenone.R supplies its amt in nmol (the amount unit the model’s nmol/L concentration states demand) while the published absorption relationship is calibrated in mg; carrying both in one column would silently mix units. Convert with the balcinrenone molar mass 399.4 g/mol (C20H18FN3O5; PubChem CID 118599727, not reported in Parkinson 2025): amt [nmol] = DOSE_BALCINRENONE_MG * 1e6 / 399.4. Same shape as DOSE_MTX_MGM2, which carries the identical molar-amt / mg-covariate split. Sibling of DOSE_NMV_MG, DOSE_EMPA_MGD, DOSE_SEMAGLUTIDE_MG and the rest of the DOSE_<drug>_<units> family; distinct from the DOSE_<N>MG binary dose-level indicators.

DOSE_NMV_MG (canonical for administered nirmatrelvir per-administration dose amount)

• Description: Continuous per-record covariate carrying the nirmatrelvir dose amount, in mg, associated with the current record. Used when a nirmatrelvir popPK model makes a PK parameter an explicit function of the administered dose level rather than treating dose only as an event amount – e.g. a dose-dependent relative-bioavailability power function capturing saturable absorption.
• Units: mg
• Type: continuous
• Scope: specific
• Reference category: n/a – used with power scaling (DOSE_NMV_MG / ref)^exponent. Reference values observed: 300 mg (Chan 2023, the authorized nirmatrelvir dose and the normalisation anchor at which F1 is fixed to 1).
• Source aliases:
  • DOSE – NONMEM $INPUT column name used in Chan_2023_nirmatrelvir.R (Chan 2023 Data S1 header comment: ;AMT: DOSE IN MG; $PK: F1 = TVF1*(DOSE/300)**THETA(11)).
• Example models: Chan_2023_nirmatrelvir.R (power effect on relative bioavailability, F1 *= (DOSE_NMV_MG / 300)^-0.409, so a 150 mg dose carries about 33% higher relative bioavailability than the 300 mg reference; applied on top of the fractional 150-mg-tablet formulation effect FORM_NMV_TAB150).
• Notes: The drug-qualified name is mandatory here rather than the bare DOSE canonical: rxode2’s event translation consumes a data column literally named DOSE, so a model referencing DOSE in model() fails to solve with The following parameter(s) are required for solving: DOSE even when the column is present in the event table. Any new model that needs the administered dose as a model covariate (as opposed to the amt event column) should register a DOSE_<drug>_<units> sibling for the same reason; existing models using the bare DOSE canonical are PD-only or posthoc-CL models that never route the column through an rxode2 event translation. Sibling of DOSE_EMPA_MGD, DOSE_SEMAGLUTIDE_MG, DOSE_CIPARGAMIN_MG, DOSE_TBPPI_MG and the rest of the DOSE_<drug>_<units> family. Distinct from the DOSE_<N>MG binary dose-level indicators.

DOSE_DORAVIRINE_MG (canonical for administered doravirine per-administration dose amount)

• Description: Continuous per-record covariate carrying the doravirine dose amount, in mg, associated with the current record. Used when a doravirine popPK model makes relative bioavailability an explicit function of the administered dose level (less-than-dose-proportional absorption) rather than treating dose only as an event amount.
• Units: mg
• Type: continuous
• Scope: specific
• Reference category: n/a – used to select dose bands. Reference band observed: 30-120 mg (Vaddady 2020, F1 = 1; includes the 100 mg clinical dose).
• Source aliases:
  • dose – Vaddady 2020 Table S1 / Yee 2019 Table 2 row labels F1 <30 mg, F1 30-120 mg (reference), F1 >120 mg; the source dataset column name is not published.
• Example models: Vaddady_2020_doravirine.R (dose-band relative bioavailability: F1 = 1.20 for doses < 30 mg, 1 for 30-120 mg, 0.882 for doses > 120 mg).
• Notes: Member of the DOSE_<drug>_<units> family; the drug-qualified name is required because rxode2’s event translation consumes a data column literally named DOSE (see DOSE_NMV_MG). Set it equal to the amt of the dose records when simulating.

DOSE_1P8MG (canonical for 1.8 mg dose-level indicator)

• Description: 1 = subject or dose record is in the 1.8 mg dose-level cohort, 0 = any other dose level. Route-neutral (SC / IV / oral); typically per-subject in fixed-arm randomizations (e.g., Overgaard 2016’s SCALE Diabetes trial arm assignment 1.8 mg vs 3.0 mg liraglutide). “P” is the decimal-point token used across the DOSE_<N>MG family for fractional doses (analogous to 1P2, 2P4 if those siblings are added in future).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (any non-1.8 mg dose level in the source study; for Overgaard 2016 the reference is the 3.0 mg liraglutide arm).
• Source aliases:
  • Dose 1.8 mg – derived per-subject from the trial-assigned dose level in Overgaard_2016_liraglutide.R (Overgaard 2016 Fig. 2 / Table S1 “Cov. 1.8 mg”; flags subjects in the SCALE Diabetes 1.8 mg arm alongside the 3.0 mg reference).
• Example models: Overgaard_2016_liraglutide.R (multiplicative log-scale effect on CL/F, coefficient +0.02 per Table S1; small and not pharmacokinetically relevant, consistent with dose proportionality in the 1.8-3.0 mg range).
• Notes: Sibling of DOSE_50MG, DOSE_70MG, DOSE_130MG, DOSE_260MG, DOSE_400MG; member of the DOSE_<N>MG family of dose-level indicators. The 1P8 token spells the decimal 1.8 with P = period; if a future paper adds a DOSE_2P4MG or DOSE_0P6MG sibling, follow the same convention. In Overgaard 2016 the indicator marks the SCALE Diabetes 1.8 mg-arm subjects vs the 3.0 mg reference so the paper can quantify (and formally rule out) a dose-linearity break within the SCALE-Diabetes cohort; the estimated log-scale coefficient of +0.02 (95% CI -0.03 to 0.08) confirms dose proportionality.

DOSE_70MG (canonical for 70 mg dose regimen indicator)

• Description: 1 = subject is on the 70 mg SC Q4W dose regimen, 0 = subject is on the 210 or 490 mg SC Q4W regimen.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (210 mg or 490 mg Q4W regimen).
• Source aliases: derived per subject from the trial-assigned dose level.
• Example models: Kotani_2022_astegolimab.R.
• Notes: Zenyatta-study categorical covariate flagging the 70 mg group (lowest dose), modeled as a -15.3% relative change on relative bioavailability. Modeled by Kotani 2022 as 70 mg vs {210 mg, 490 mg} combined reference.

DOSE_50MG (canonical for 50 mg dose-level indicator)

• Description: 1 = subject or dose record is in the 50 mg dose-level cohort, 0 = any other dose level. Route-neutral (covers SC, IV, oral); per-record or per-subject depending on the source design (record-level when a single subject crossed dose levels, subject-level when subjects were randomized to a fixed dose arm).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (any non-50 mg dose level in the source study, e.g. 100-300 mg SC for Othman 2014, 200 mg or 400 mg t.i.d. oral for Jorga 2000).
• Source aliases:
  • derived per dose record from the administered amount (AMT) – used in Othman_2014_daclizumab.R and the Diao 2016 family.
  • I_Dose50mg – subject-level indicator derived from study-arm randomization, used in Jorga_2000_tolcapone_fluctuators.R for the 50 mg t.i.d. tolcapone fluctuator arm.
• Example models: Othman_2014_daclizumab.R (record-level SC), Diao_2016_daclizumab_cd25.R, Diao_2016_daclizumab_cd56bright.R, Diao_2016_daclizumab_treg.R, Jorga_2000_tolcapone_fluctuators.R (subject-level oral t.i.d.).
• Notes: Othman 2014 estimated two separate absolute bioavailabilities because of non-linear dose-normalized exposure at the 50 mg SC dose – F = 0.84 for the therapeutic 100-300 mg SC range and F = 0.57 for the 50 mg SC cohort. Encoded as a record-level indicator so e_dose_50mg_f = 0.57/0.84 - 1 = -0.321 scales bioavailability only on 50 mg SC dose records. For clinical-range simulation (150 mg SC Q4W Phase III regimen) leave DOSE_50MG = 0. The Diao 2016 PK/PD models inherit the Othman 2014 PK backbone verbatim. Jorga 2000 uses the indicator subject-level on the central and peripheral volumes of distribution: (1 + e_dose_50mg_vc_vp * DOSE_50MG) with e_dose_50mg_vc_vp = -0.45 (V is 55% of the 200 mg reference at the 50 mg t.i.d. arm); paired with DOSE_400MG in the same fluctuator model so the 200 mg arm is the joint reference (both indicators = 0).

DOSE_3MG (canonical for 3 mg dose-level indicator)

• Description: 1 = subject or dose record is in the 3 mg dose-level cohort, 0 = any other dose level. Route-neutral; per-record or per-subject depending on the source design (subject-level for Fimbo 2023, a single-dose study in which each participant belongs to exactly one height-assigned dose pool).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (any non-3 mg dose level in the source study; for Fimbo 2023 ivermectin the reference is the pooled 6 / 9 / 12 mg dose pools).
• Source aliases:
  • DOSE3MG – derived in the Fimbo 2023 NONMEM control stream (Data S3) from the dose-pool column: DOSE3MG=0 / IF(DOSEGRP.EQ.3) DOSE3MG=1.
• Example models: Fimbo_2023_ivermectin.R (power-form effect on relative bioavailability: fdepot <- exp(lfdepot + etalfdepot) * (1 + e_dose_3mg_fdepot)^DOSE_3MG with e_dose_3mg_fdepot = 0.4822, so relative F is 48.2% higher in the 3 mg ivermectin pool than in the pooled 6 / 9 / 12 mg reference).
• Notes: Sibling of DOSE_1P8MG, DOSE_10MG, DOSE_20MG, DOSE_50MG, DOSE_70MG, DOSE_130MG, DOSE_260MG and DOSE_400MG; member of the DOSE_<N>MG family of dose-level indicators. Fimbo 2023 treated dose as a CATEGORICAL rather than continuous covariate because the ETA_F1-vs-DOSE plot showed a step function: the medians for the 6, 9 and 12 mg pools lie on a horizontal line and only the 3 mg pool sits above it (Fimbo 2023 Results, “Population pharmacokinetic model”). Because ivermectin MDA dosing is by height pole, the 3 mg pool is also the lightest (mean 19.9 kg) and youngest (mean 6.4 years) stratum, so the dose indicator is strongly confounded with body weight; the model carries allometric weight scaling on CL/F and Vc/F concurrently, and the retained dose effect is the residual bioavailability difference after that scaling. Only 12 of 468 subjects are in the 3 mg pool, hence the 40% RSE and the wide bootstrap CI (0.06-1.28). For routine-MDA simulation set DOSE_3MG = 0 unless the smallest-children pool is specifically being reproduced.

DOSE_10MG (canonical for 10 mg dose-level indicator)

• Description: 1 = subject or dose record is in the 10 mg dose-level cohort, 0 = any other dose level. Route-neutral; per-record or per-subject depending on the source design (subject-level for fixed-arm randomizations such as Zhang 2015’s dolutegravir dose-ranging cohorts).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (any non-10 mg dose level in the source study; for Zhang 2015 dolutegravir the reference is the pooled 25 mg / 50 mg dose levels).
• Source aliases:
  • DOSE (= 1 for the 10 mg dose, 0 for 25 / 50 mg) – used in Zhang_2015_dolutegravir.R (per Zhang 2015 Table 3 footnote, F = 1.21^GEND * 1.24^DOSE).
• Example models: Zhang_2015_dolutegravir.R (exponential effect on bioavailability: f(depot) <- exp(lfdepot + e_dose_10mg_fdepot * DOSE_10MG) with e_dose_10mg_fdepot = log(1.24) = 0.215, so F is 24% higher at the 10 mg dose vs the pooled 25 / 50 mg reference), Gu_2025_rivaroxaban.R (exponential effect on relative bioavailability with e_dose_10mg_fdepot = fixed(log(1.363)), so relative F at the 10 mg rivaroxaban dose is 1.363 times the 15 mg reference; the value was estimated at the base-model stage and then held fixed because only 2 of 105 patients received 10 mg; paired with DOSE_20MG).
• Notes: Sibling of DOSE_50MG, DOSE_70MG, DOSE_130MG, DOSE_260MG, and DOSE_400MG; member of the DOSE_<N>MG family of dose-level indicators. Zhang 2015 reports that continuous dose was tested as a covariate on F but was not significant, and no F difference was found between the 25 and 50 mg dose levels; the higher relative bioavailability at the 10 mg dose is attributed to better dispersion of the lower-strength tablet (Zhang 2015 Discussion).

DOSE_20MG (canonical for 20 mg dose-level indicator)

• Description: 1 = subject or dose record is in the 20 mg dose-level cohort, 0 = any other dose level. Route-neutral; per-record or per-subject depending on the source design (per-record for Gu 2025, whose pooled analysis mixes a 20 mg single-dose healthy-volunteer study with a 10 / 15 / 20 mg once-daily patient cohort).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (any non-20 mg dose level in the source study; for Gu 2025 paired with DOSE_10MG = 0 to select the 15 mg reference dose level).
• Source aliases:
  • F 20mg – Gu 2025 Table 3 row for the relative bioavailability of the 20 mg dose group.
• Example models: Gu_2025_rivaroxaban.R (exponential effect on relative bioavailability: f(depot) <- exp(lfdepot + e_dose_10mg_fdepot * DOSE_10MG + e_dose_20mg_fdepot * DOSE_20MG) with e_dose_20mg_fdepot = log(0.537), so relative F at 20 mg is 0.537 of the 15 mg reference; paired with DOSE_10MG whose fixed effect is log(1.363)).
• Notes: Sibling of DOSE_10MG, DOSE_50MG, DOSE_70MG, DOSE_130MG, DOSE_260MG and DOSE_400MG; member of the DOSE_<N>MG family of dose-level indicators. Gu 2025 attributes the falling relative bioavailability with increasing dose to the limited aqueous solubility of rivaroxaban, a BCS class II compound, and reports 15 mg relative F as 1.86-fold that of 20 mg – consistent with the 1.43-fold value reported for the same contrast by Zhao 2022 in Chinese NVAF patients (Gu 2025 Discussion). Distinct from the continuous per-administration rivaroxaban dose canonicals DOSE_RIV_MGKG (mg/kg, used by the Willmann 2021 pediatric model’s exponential-decay bioavailability function): use the dose-level indicators when the source estimates one relative-F value per fixed dose level, and the continuous column when the source fits a smooth function of dose.

DOSE_400MG (canonical for 400 mg dose-level indicator)

• Description: 1 = subject or dose record is in the 400 mg dose-level cohort, 0 = any other dose level. Route-neutral; per-record or per-subject depending on the source design (subject-level for fixed-arm randomizations such as Jorga 2000’s t.i.d. tolcapone study arms).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (any non-400 mg dose level in the source study; for Jorga 2000 paired with DOSE_50MG = 0 to select the 200 mg t.i.d. reference).
• Source aliases:
  • I_Dose400mg – subject-level indicator derived from study-arm randomization, used in Jorga_2000_tolcapone_fluctuators.R for the 400 mg t.i.d. tolcapone fluctuator arm.
• Example models: Jorga_2000_tolcapone_fluctuators.R (subject-level oral t.i.d.).
• Notes: Sibling of DOSE_50MG and DOSE_70MG; member of the DOSE_<N>MG family of dose-level indicators. Jorga 2000 uses the indicator subject-level on the central and peripheral volumes of distribution: (1 + e_dose_400mg_vc_vp * DOSE_400MG) with e_dose_400mg_vc_vp = +0.40 (V is 140% of the 200 mg reference at the 400 mg t.i.d. arm). The dose-dependent V was an empirical finding in the fluctuator cohort that the authors could not confirm in the nonfluctuator cohort (only 200 and 400 mg arms were enrolled there); the effect plausibly reflects a few high-V outliers in the small-volume cohort rather than a true mechanistic dose-V relationship (Jorga 2000 Discussion).

DOSE_600MG (canonical for 600 mg dose-level indicator)

• Description: 1 = the dose record (or subject, for fixed-arm designs) is at the 600 mg dose level, 0 = any other dose level. Route-neutral; per-record in Zhu_2018_asunaprevir.R, where subjects could move between dose levels mid-study.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (any non-600 mg dose level in the source study; the 100 mg and 200 mg levels in Zhu 2018).
• Source aliases:
  • Dose(600 mg) – the printed covariate-equation term in Zhu_2018_asunaprevir.R.
• Example models: Zhu_2018_asunaprevir.R (per-dose-record indicator on relative bioavailability: F = exp(-0.215 * FORM_TABLET) * exp(0.65 * DOSE_600MG), so bioavailability at the 600 mg level is 1.92-fold the level-independent reference. This encodes the more-than-dose-proportional asunaprevir exposure the authors observed on going from 200 mg BID to 600 mg BID of the Phase II tablet, which they note was especially marked in Japanese subjects. In that pooled dataset 600 mg was administered only as the tablet – 7.1% of subjects – so DOSE_600MG = 1 implies FORM_TABLET = 1 there, and the two effects are not independently identifiable outside the tablet arm).
• Notes: Member of the DOSE_<N>MG family of dose-level indicators; sibling of DOSE_50MG, DOSE_70MG, DOSE_400MG, DOSE_130MG, DOSE_260MG. Scope specific because the reference complement is whatever other dose levels a given study enrolled. A dose-level indicator on bioavailability is the standard empirical encoding of nonlinear dose-exposure behaviour in a model that is otherwise linear; it is NOT a mechanistic saturation term, and it does not interpolate – a simulation at 400 mg gets the reference bioavailability, not something between. Prefer an explicit saturable structure when a paper fits one.

DOSE_130MG (canonical for 130 mg dose-level indicator)

• Description: 1 = subject or dose record is in the 130 mg dose-level cohort, 0 = any other dose level. Route-neutral; per-record or per-subject depending on the source design (per-dose-record for Hajjar 2018’s adult Pompe-disease ATB200+AT2221 study where the AT2221 dose changes across successive dosing occasions within a subject).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (any non-130 mg dose level in the source study; for Hajjar 2018 paired with DOSE_260MG = 0 to select the no-AT2221 reference for ATB200 linear CL).
• Source aliases:
  • derived per-dose-record from the co-administered AT2221 dose amount – used in Hajjar_2018_cipaglucosidase.R as a categorical multiplier on ATB200 linear CL (e_dose130mg_cl ^ DOSE_130MG).
• Example models: Hajjar_2018_cipaglucosidase.R (per-dose-record oral co-medication indicator).
• Notes: Sibling of DOSE_50MG, DOSE_70MG, DOSE_260MG, and DOSE_400MG; member of the DOSE_<N>MG family of dose-level indicators. In Hajjar 2018 the indicator marks the AT2221 (miglustat) co-administration dose paired with the same-occasion 20 mg/kg ATB200 IV infusion; mutual exclusivity with DOSE_260MG is enforced by the study design (a single dose occasion uses at most one AT2221 dose level).

DOSE_260MG (canonical for 260 mg dose-level indicator)

• Description: 1 = subject or dose record is in the 260 mg dose-level cohort, 0 = any other dose level. Route-neutral; per-record or per-subject depending on the source design (per-dose-record for Hajjar 2018’s adult Pompe-disease ATB200+AT2221 study where the AT2221 dose changes across successive dosing occasions within a subject).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (any non-260 mg dose level in the source study; for Hajjar 2018 paired with DOSE_130MG = 0 to select the no-AT2221 reference for ATB200 linear CL).
• Source aliases:
  • derived per-dose-record from the co-administered AT2221 dose amount – used in Hajjar_2018_cipaglucosidase.R as a categorical multiplier on ATB200 linear CL (e_dose260mg_cl ^ DOSE_260MG).
• Example models: Hajjar_2018_cipaglucosidase.R (per-dose-record oral co-medication indicator).
• Notes: Sibling of DOSE_50MG, DOSE_70MG, DOSE_130MG, and DOSE_400MG; member of the DOSE_<N>MG family of dose-level indicators. In Hajjar 2018 the indicator marks the AT2221 (miglustat) co-administration dose paired with the same-occasion 20 mg/kg ATB200 IV infusion; mutual exclusivity with DOSE_130MG is enforced by the study design.

DOSE_18G (canonical for 18 g dose-level indicator)

• Description: 1 = subject or dose record is in the 18 g dose-level cohort, 0 = any other dose level. Gram-scale member of the DOSE_<N><UNIT> dose-level-indicator family, for drugs dosed in grams rather than milligrams.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (any non-18 g dose level; for Darwish 2025 the reference comprises the 6-12 g therapeutic weight-banded doses).
• Source aliases:
  • DoseGrp1 – used in Darwish_2025a_trofinetide.R and Darwish_2025b_trofinetide.R (Darwish 2025 Results, definition of the DoseGrp1_I indicator variable in the typical-value equation for F1).
• Example models: Darwish_2025a_trofinetide.R (proportional shift on oral bioavailability from Darwish 2025 Table 2: fdepot * (1 + e_dose18g_f * DOSE_18G) with e_dose18g_f = -0.132, a 13.2% reduction in F1 at the 18 g supratherapeutic dose. The dose level was studied in the thorough-QTc study ACP-2566-008; the effect captures the less-than-proportional rise in trofinetide exposure above the 6-12 g therapeutic range. Paired with DOSE_24G, which carries the larger 28.4% reduction – founding example), Darwish_2025b_trofinetide.R (proportional shift on oral bioavailability from Darwish 2025 Table 2: fdepot * (1 + e_dose18g_f * DOSE_18G) with e_dose18g_f = -0.132, a 13.2% reduction in F1 at the 18 g supratherapeutic dose. The dose level was studied in the thorough-QTc study ACP-2566-008; the effect captures the less-than-proportional rise in trofinetide exposure above the 6-12 g therapeutic range. Paired with DOSE_24G, which carries the larger 28.4% reduction – founding example).
  • Darwish_2025b_trofinetide.R (the 14-study update of the same sponsor analysis: e_dose18g_f = -0.132, unchanged from the earlier model, Darwish 2025b Table 2).
• Notes: Sibling of the milligram-scale DOSE_1P8MG / DOSE_10MG / DOSE_50MG / DOSE_70MG / DOSE_130MG / DOSE_260MG / DOSE_400MG entries; the unit token is G rather than MG because trofinetide is dosed in grams throughout its literature and spelling this as DOSE_18000MG would be gratuitously unreadable. Follow the same P-for-decimal-point convention as DOSE_1P8MG if a fractional-gram sibling is ever needed. The covariate-effect parameter form is e_dose18g_<param>: unlike the DIS_/FORM_/AE_ families, the DOSE token is retained because it carries the semantic content (the effect is about the dose level) rather than being a bare family prefix. Mutually exclusive with DOSE_24G; both = 0 selects the therapeutic-dose reference. Distinct from the abstract DOSE_HIGH, which names the “in the highest cohort” role without fixing a numerical value – use the numerically-explicit entries when a source estimates a separate effect for each of two or more named supratherapeutic levels, as Darwish 2025 does.

DOSE_24G (canonical for 24 g dose-level indicator)

• Description: 1 = subject or dose record is in the 24 g dose-level cohort, 0 = any other dose level. Gram-scale member of the DOSE_<N><UNIT> dose-level-indicator family.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (any non-24 g dose level; for Darwish 2025 the reference comprises the 6-12 g therapeutic weight-banded doses).
• Source aliases:
  • DoseGrp2 – used in Darwish_2025a_trofinetide.R and Darwish_2025b_trofinetide.R (Darwish 2025 Results, definition of the DoseGrp2_I indicator variable in the typical-value equation for F1).
• Example models: Darwish_2025a_trofinetide.R (proportional shift on oral bioavailability from Darwish 2025 Table 2: fdepot * (1 + e_dose24g_f * DOSE_24G) with e_dose24g_f = -0.284, a 28.4% reduction in F1 at the 24 g supratherapeutic dose – roughly twice the 18 g reduction, consistent with a saturating-absorption interpretation – founding example), Darwish_2025b_trofinetide.R (proportional shift on oral bioavailability from Darwish 2025 Table 2: fdepot * (1 + e_dose24g_f * DOSE_24G) with e_dose24g_f = -0.284, a 28.4% reduction in F1 at the 24 g supratherapeutic dose – roughly twice the 18 g reduction, consistent with a saturating-absorption interpretation – founding example).
  • Darwish_2025b_trofinetide.R (the 14-study update: e_dose24g_f = -0.284, unchanged from the earlier model, Darwish 2025b Table 2).
• Notes: Paired sibling of DOSE_18G; see that entry for the family, unit-token, and covariate-effect-naming rationale. Mutually exclusive with DOSE_18G.

DOSE_HIGH (canonical for highest-dose-cohort binary indicator)

• Description: 1 = subject is in the source paper’s highest-dose cohort (numerical threshold documented per model in covariateData[[DOSE_HIGH]]$notes); 0 = subject is in any lower-dose cohort. Time-fixed per subject in escalating-cohort or parallel-dose designs where each animal / subject remained on a single dose level for the full study and the paper detected a step-function shift in a PK parameter (typically apparent oral clearance or bioavailability) that was significant enough to warrant a separate typical-value estimate for the highest cohort alone.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (any dose cohort below the paper-defined highest-dose threshold).
• Source aliases: derived per subject from the trial- or cohort-assigned dose level.
• Example models: Viberg_2012_AZD6088_rat.R (Viberg 2012 Results, PK model paragraph: ‘The highest dose group deviated in the plots. This might be due to higher bioavailability or saturation of elimination at the highest dose and after estimating different clearance values for this group OFV decreased 32 units’. The 40 umol/kg = ~16.26 mg/kg oral dose group of AZD6088 in male Sprague-Dawley rats is flagged as DOSE_HIGH = 1 and gates a separate typical-value apparent oral clearance lcl_highdose = log(3.92 L/h/kg) in place of the standard-dose lcl = log(10.87 L/h/kg); the initial-efficacy-study cohorts at 1, 2.5, 5, 10, 20 umol/kg are DOSE_HIGH = 0), Maleki_2024_brepocitinib.R (first human ratification: DOSE_HIGH = 1 for an administered daily dose above 100 mg – the 175 and 200 mg phase I arms, 5% of the analysis population – and gates a 30% uplift in relative bioavailability, Frel = 1 * (1 + 0.3 * DOSE_HIGH), per Maleki 2024 Table 3 footnote c. The 100 mg cohort itself sits in the reference group. Same “step-function switch at the top of the dose range” semantics as the Viberg 2012 founding example, but expressed on bioavailability rather than clearance; the paper reached it from single-dose NCA showing dose-dependent PK above 100 mg q.d. plus a correlation between administered dose and the CL/F and Vc/F individual estimates), Kastrissios_2006_apricoxib.R (DOSE_HIGH = 1 for the 400 and 800 mg single-dose cohorts of study 2 and 0 for the 2 to 200 mg cohorts, gating the bracketed two-typical-value apparent-oral-clearance switch of Kastrissios 2006 equation 7 – exp(lcl + (lcl_highdose - lcl) * DOSE_HIGH) selects 34.1 L/h at 0 and 19.5 L/h at 1, the 42.8% reduction the abstract reports as “reduced by 43% at doses greater than 200 mg”. Closest yet to the Viberg 2012 founding semantics of the three, since the switch is on clearance and is encoded as a distinct THETA for the top cohorts; note that this model ALSO carries a smooth saturable relative bioavailability read from podo(depot), and the paper is explicit that “because both apparent oral clearance and Frel were reduced at doses greater than 200 mg, it is difficult to delineate these 2 effects” – so the step indicator and the continuous dose nonlinearity coexist here and must not be conflated), Rohatagi_2008_apricoxib.R (the same Kastrissios 2006 apricoxib PK layer, reproduced inline as the concentration driver of the Rohatagi 2008 dental-pain exposure-response model, with identical semantics and the identical 34.1 / 19.5 L/h switch. DOSE_HIGH is 0 for every patient in that phase 2a cohort, whose apricoxib doses were 10 to 200 mg; it is retained so the embedded PK layer reproduces the published model over its full dose range. Distinct in that file from DOSE_APRICOXIB_MG, which is the continuous per-subject dose regressor of the pharmacodynamic equations 13 to 15), Hughes_2022_brepocitinib.R (DOSE_HIGH = 1 for an administered brepocitinib dose of 175 mg or more – the 175 mg once-daily arm of NCT02310750 and the 200 mg single-dose arms of NCT02310750 and NCT03656952 – gating a 35.1% uplift in relative bioavailability, Frel = 1 * (1 - 0.283 * Fed) * (1 + 0.351 * DOSE_HIGH), per Hughes 2022 Equation 8 and Table 3. Same drug and the same two dose arms as the Maleki_2024_brepocitinib.R entry above, but the two papers state the threshold differently – Hughes 2022 as “>= 175 mg” and Maleki 2024 as “> 100 mg/day” – because neither analysis studied a dose between 100 and 175 mg; the encodings therefore select the same subjects. Hughes 2022 is explicit that the step function is an artefact of that gap: a continuous dose-Frel relationship was fitted and gave minimal improvement, and the Discussion argues the step estimate is likely the maximum of an effect that rises gradually between 100 and 175 mg), DiDeo_2025_tideglusib_healthy.R, DiDeo_2025_tideglusib_dm1.R (DOSE_HIGH = 1 for an administered tideglusib dose above 400 mg, per Di Deo 2025 Table 2, whose rows are split as kA (1/h) vs kA (dose > 400 mg) (1/h) and Bioavailability (doses <= 400 mg) – FIXED to 1 – vs Bioavailability (doses > 400 mg). Notable as the widest use of the indicator so far: one column gates THREE quantities, because the paper splits the proportional residual error on the same threshold as well (Proportional error (doses <= 400 mg) vs (doses > 400 mg)), so the model carries propSd / propSd_highdose alongside lka / lka_highdose and lfdepot / lfdepot_highdose. The threshold applies to the PER-ADMINISTRATION dose rather than the daily dose, which matters in the healthy-elderly study where the 400 mg b.i.d. arm delivers 800 mg/day yet sits in the reference group: DOSE_HIGH = 0 for the 300 mg b.i.d. and 400 mg b.i.d. arms and 1 for the 600, 800, 1000 and 1200 mg q.d. arms. In the DM-1 patient study only the 400 mg q.d. (0) and 1000 mg q.d. (1) arms exist. Mechanistically the paper attributes the step to the limited aqueous solubility of tideglusib – Discussion: “oral bioavailability decreases with increasing doses, even though we have not been able to quantify these changes in a continuous manner. Instead, bioavailability and absorption rate constant were dichotomized into two levels, with exposure data following a 400 mg dose as reference” – the same “could not identify a continuous form, so a step was used” reasoning Hughes 2022 gives above), Koele_2025_btz043.R (DOSE_HIGH = 1 for a daily BTZ-043 dose above 1250 mg – the 1500 and 1750 mg Stage 1 dose-escalation cohorts – gating a multiplicative 0.710 factor on relative bioavailability, frel = ... * 0.710^DOSE_HIGH, per Koele 2025 Table 2 ‘Dose (>1250 mg) on F (%) 71.0 (51.0-98.0)’ and the supplementary control stream IF(DOSE.GT.1250)FDose = THETA(8). Koele 2025 Results: ‘An apparent plateau in bioavailability was observed for doses over 1250 mg daily. No further decrease in the bioavailability was detected between the 1500 and 1750 mg doses (dOFV = -3.3, df = 1)’, attributed in the Discussion to possible saturable absorption. Per dose record rather than per subject in principle, though in this trial each participant stayed on one dose level), Jin_2025_benralizumab.R (DOSE_HIGH = 1 for the 200 mg subcutaneous benralizumab arms – the top of the 2-200 mg SC range used by the two phase II studies in the 12-study pool – gating a 42.5% reduction in absolute subcutaneous bioavailability, Fa1 * exp(-0.554 * DOSE_HIGH), per Jin 2025 Resource 10 row Rfa1Dose. Same step-at-the-top-of-the-dose-range semantics on bioavailability as the two brepocitinib entries above, but the first registered instance in which the step is DOWNWARD – consistent with saturable absorption from the subcutaneous depot at high injected mass. The coefficient additionally carries its own small normally-distributed random effect, omega 0.029 fixed, which the source labels as an IIV on the covariate coefficient itself rather than on a parameter), Abdelgawad_2025_rifampicin.R (DOSE_HIGH = 1 for the LASER-TBM high-dose experimental arms – oral rifampicin 35 mg/kg once daily, or intravenous 20 mg/kg for the first three days – and 0 for the standard-of-care control arm at oral 10 mg/kg. Jointly with DAY28 it selects one of four unconstrained typical values of the maximal intrinsic hepatic clearance (33.1 and 41.4 L/h standard dose at the day-3 and day-28 visits; 46.1 and 70.2 L/h high dose at the same two visits, all as the reported CLint,max * fu product). Viberg 2012 founding semantics – a step-function switch in clearance at the top of the dose range, encoded as distinct THETAs – but note that this model ALSO carries a smooth saturable first-pass extraction, and the paper cannot separate the dose-group step from the arm-level difference in prior rifampicin exposure, so the indicator and the structural nonlinearity coexist here and must not be conflated), Jonsson_2016_tanezumab.R (the ONLY entry so far whose source reference category is the HIGH-dose cohort rather than the low: Jonsson 2016 Equation 3 reads ‘CLDOSE = 1 if dose = 10 mg or 1 + theta12 if dose = 2.5 or 5 mg’, so DOSE_HIGH = 1 marks the 10 mg arm of four parallel-group phase 3 osteoarthritis trials and the published +0.0669 fractional clearance increase is applied against the complement, (1 + e_dose_high_cl * (1 - DOSE_HIGH)), to keep the canonical reference category 0 while preserving the verbatim coefficient. Clearance is 6.69% higher in the pooled 2.5 and 5 mg arms than in the 10 mg arm (Table 3 row ‘Dose on CL’). As in Abdelgawad_2025_rifampicin.R the step coexists with a structural saturable pathway – a parallel Michaelis-Menten elimination arm – and the Discussion is explicit that the dose step is over and above that concentration-dependent nonlinearity, so the two must not be conflated), Francis_2019_rifapentine.R (DOSE_HIGH = 1 for the 125 patients in the RIFAQUIN 1,200-mg-once-weekly continuation-phase arm, the top of the 450 / 600 / 900 / 1,200 mg range pooled across two tuberculosis trials, gating a 13.2 percent REDUCTION in apparent oral clearance, (1 + e_dose_high_cl * DOSE_HIGH) with the coefficient -0.132, per Francis 2019 Table 2 row ‘Effect of group on 1,200-mg dose in RIFAQUIN study on CL (%)’. The clearest case so far that the indicator can encode a dose-AND-FREQUENCY arm rather than a dose level, and it must not be read as concentration-dependent saturation: the 1,200 mg arm is the only once-weekly arm in the pool, and the Discussion attributes the lower clearance to less autoinduction under weekly rather than more frequent dosing – rifapentine induces its own metabolism – so the covariate is confounded with dosing interval by construction. Note the sign is opposite to what a saturable-absorption reading of the other entries would predict, and opposite to Savic et al.’s earlier report of DECREASED rifapentine bioavailability at higher doses, which Francis 2019 explicitly contrasts its own finding against), Chan_2017_ipf_fvc_mbma.R (first MBMA / study-arm-level use, and the first in which the indicator gates an efficacy effect rather than a PK parameter: DOSE_HIGH = 1 selects the pirfenidone 2,403 mg/day maximum effect on placebo-corrected change from baseline %predicted FVC (3.87) in place of the 1,197 mg/day one (2.42), per Chan 2017 Results – ‘the dose-response relationship of pirfenidone was characterized using a step-function, with the high dose (2,403 mg/day) having an estimated greater treatment effect than the low dose (1,197 mg/day)’. Two features distinguish it from the per-subject entries above. The value is a property of a published TRIAL ARM, not of a patient, so it is paired with the PIRFENIDONE arm indicator and is inert (and set to 0) on every one of the other 14 treatments’ arms. And only two dose levels exist in the whole analysis dataset, so the indicator NAMES a level rather than binning a continuous dose at a threshold – there is no mg cut-point to document, unlike the >100 mg, >=175 mg, >400 mg and >1250 mg thresholds of the siblings above), Yamamoto_2017_morphine_rat_pbpk.R (DOSE_HIGH = 1 for the 10 and 40 mg/kg intravenous morphine groups of the Yamamoto 2017 rat CNS PBPK evaluation and 0 for the 4 mg/kg group, per Table 4 footnotes a and b and Table 5 footnotes a and b, which give one set of blood-brain / blood-CSF barrier asymmetry factors and one brain-tissue binding factor at 4 mg/kg – AFout1 11, AFout2 20, AFout3 26, BF 1.8 – and a second, uniformly larger set at 10 and 40 mg/kg – AFout1 20, AFout2 38, AFout3 49, BF 3.9. The supplement carries the two dose groups as two separate NONMEM control streams, S6 and S7, that are otherwise identical. First registered instance in which the indicator gates ACTIVE-TRANSPORT asymmetry factors and a tissue-binding factor rather than a clearance or a bioavailability: the direction – larger efflux asymmetry at the higher doses, so a LOWER unbound brain-to-plasma ratio (Kp,uu 0.38 at 4 mg/kg versus 0.23 at 10-40 mg/kg) – is what saturable active INFLUX at the blood-brain barrier looks like when it is absorbed into a net asymmetry factor, and matches the saturable-influx mechanism Groenendaal 2007 reported for morphine in the same rat dataset. Also the first preclinical single-animal-per-dose-level use since the Viberg 2012 founding example), Marier_2021_teduglutide.R (DOSE_HIGH = 1 for the supra-therapeutic 20 mg teduglutide dose – Marier 2021 Methods ‘dose (therapeutic vs. supra-therapeutic [20 mg])’, the thorough-QT study C09-001 – multiplying the absorption lag time by 1.783, e_dosehigh_tlag^DOSE_HIGH, per Table S8), Marier_2021_teduglutide_ppsv.R (DOSE_HIGH = 1 for the 0.10 mg/kg/day arm, the highest dose of the teduglutide exposure-response data; replaces the Cmax power term on Emax with the multiplier 1 + (-0.225) per Marier 2021 Table S10 ‘Dose effect (0.10 mg/kg) on Emax’; the first use gating a time-course efficacy Emax on a continuous change-from-baseline endpoint), Ahmed_2020_cyclophosphamide.R (DOSE_HIGH = 1 for the 600 mg/m^2 based cyclophosphamide regimens – AC and AC-T, 161 of 267 Ethiopian breast cancer patients – and 0 for the 500 mg/m^2 based FAC regimen. Like Jonsson_2016_tanezumab.R the source reference level is the HIGH dose: Ahmed 2020 Table 4 applies (1 + THETA) to the 500 mg/m^2 regimen, so the model writes (1 + e_dose_high_cl * (1 - DOSE_HIGH)) with -0.323 on clearance and (1 + e_dose_high_vc * (1 - DOSE_HIGH)) with -0.371 on volume. Only two dose levels exist, so the indicator names a level; the regimens also differ in co-administered drugs (5-fluorouracil only in FAC), so the effect is a regimen effect rather than a pure dose nonlinearity), Kim_2020_CKD_519.R (Kim 2020 Table 2: DOSE_HIGH = 1 for the 400 mg once-daily cohort of a 50/100/200/400 mg multiple-ascending-dose study in healthy men, selecting the fixed-zero time-decay constant alpha2 of the relative-bioavailability term FT = exp(-alpha * TIME) in place of the estimated alpha1 = 0.002 1/h shared by the 50-200 mg cohorts; the 400 mg group alone showed no loss of exposure on repeated dosing. First use in which the indicator gates a TIME-dependence of bioavailability rather than a level).
• Notes: Abstract-form sibling of the numerically-explicit DOSE_<N>MG family (DOSE_50MG, DOSE_70MG, DOSE_130MG, DOSE_260MG, DOSE_400MG) and of the drug-suffixed DOSE_HIGH_EFL (Jansson 2008 eflornithine >= 3000 mg/kg indicator): DOSE_HIGH names the abstract “in the highest cohort” role without fixing the mg-threshold in the canonical name, and each model’s covariateData[[DOSE_HIGH]]$notes documents the per-paper dose value that maps to DOSE_HIGH = 1. Distinct from the continuous COHDOSE (mg/kg per-subject dose cohort) and DOSE (per-record administered dose level, mg) canonicals, which are numeric rather than binary; use DOSE_HIGH when the source paper reports a step-function switch in a PK parameter at the top of the dose range (Viberg 2012 pattern: single distinct THETA for the highest cohort, all other cohorts share the standard THETA), use COHDOSE when the source reports a power-form or continuous covariate effect across all cohorts, and use DOSE_HIGH_EFL (or a similar drug-suffixed sibling) when the numerical threshold is enshrined in the canonical name. Scope: specific because the numerical highest-dose value differs per paper; promote to general once a second model registers under the same abstract semantics.

DOSE_REDUCED (canonical for within-course dose-reduction indicator)

• Description: 1 = the subject’s dose of the modelled drug was reduced at some point during the treatment course; 0 = the subject remained on the assigned dose throughout. A per-subject summary of what happened DURING treatment, so it is time-fixed in the sense that one value describes the whole course. Reductions may be protocol-mandated (a data-monitoring-committee decision that moves a whole arm down), toxicity-driven per patient (a haematologic or hepatic stopping rule), or both in the same trial.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no dose reduction).
• Source aliases:
  • dose was reduced – Riggs 2012 Table II row label; no dose reduction is the corresponding reference level in the Figure 2 caption.
• Example models: Riggs_2012_albinterferon_svr_gt1.R, Riggs_2012_albinterferon_svr_gt23.R (logit-scale effects on sustained virologic response of 0.042623 and 0.277399; in the albinterferon phase 3 trials the 1200 ug every-2-weeks arms were moved down to 900 ug after the data-monitoring committee flagged serious pulmonary adverse events, and individual patients were additionally reduced for neutropenia, thrombocytopenia, or anaemia).
• Notes: Semantically DISTINCT from DOSE_HIGH / DOSE_LOW / the DOSE_<N>MG family, all of which record which dose COHORT a subject was assigned to at randomisation. This column records a mid-course modification and is therefore a consequence of what happened on study, not a design variable. That makes it a confounded predictor in any outcome model: patients whose dose was reduced are, by construction, patients who had a toxicity or who were in the higher-dose arm, so a coefficient on this indicator must not be read causally. Riggs 2012 makes exactly that point, reporting a near-null effect on efficacy as the expected consequence of there being no marked exposure-response relationship to begin with. When the exposure metric in the same model is itself computed over the ACTUAL dosing history (as CSS_ALBIFN is), the exposure consequence of the reduction is already absorbed there and this indicator carries only the residual, non-exposure-mediated association. Specific scope because the reduction rules are protocol-specific; document the trial’s reduction criteria in covariateData[[DOSE_REDUCED]]$notes. A model needing the magnitude rather than the fact of the reduction should carry the per-record dose instead.

DOSE_LOW (canonical for lowest-dose-cohort binary indicator)

• Description: 1 = the record’s administered dose is in the source paper’s low-dose group (the specific dose levels documented per model in covariateData[[DOSE_LOW]]$notes); 0 = any higher dose level. Used where a paper detects a step-function shift in a PK parameter at the bottom of the studied dose range and estimates a separate multiplicative effect for the low-dose levels alone, rather than a continuous dose-response across all levels.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (any dose level above the paper-defined low-dose group).
• Source aliases:
  • 10/25 mg dose – used in Comisar_2025_rimegepant.R (Comisar 2025 Table 2 row label ‘10/25mg dose effect on k tr’).
  • dose level – used in Zhang_2016_burosumab.R (Zhang 2016 Table 2 row ‘CL/F (L/day) … for dose of 0.05 mg/kg’; the paper screened ‘dose levels’ as an extrinsic covariate and retained only the low-dose contrast).
• Example models: Comisar_2025_rimegepant.R (founding example; multiplicative fractional effect on the transit absorption rate constant, ktr * (1 + 0.596 * DOSE_LOW) per Comisar 2025 Table 2. DOSE_LOW = 1 for the 10 mg and 25 mg rimegepant dose levels and 0 for 75 mg and 150 mg. The positive sign is an increase in ktr, i.e. faster transit at low dose, per Table 2 footnote a ‘Increase in the transit rate constant for 10-25 mg doses’. The 6 pediatric participants weighing 15 to 30 kg who received 25 mg ODT carry DOSE_LOW = 1), Zhang_2016_burosumab.R (multiplicative fold-effect on apparent clearance, cl * 1.15^DOSE_LOW per Zhang 2016 Table 2, i.e. CL/F 0.321 vs 0.279 L/day at 70 kg. DOSE_LOW = 1 for the 0.05 mg/kg subcutaneous burosumab dose level and 0 for 0.1, 0.3, 0.6 and 1.0 mg/kg. Time-varying within a subject: in the KRN23-INT-001 intradose escalation each participant carries DOSE_LOW = 1 for the first 28-day interval only. Zhang 2016 fit a separate typical CL/F per dose level first (Supplementary Table S2) and pooled 0.1-1.0 mg/kg once that range proved linear).
• Notes: Mirror-image sibling of [[DOSE_HIGH]], and registered on the same rationale: it names the abstract “in the lowest cohort” role without fixing the mg-threshold in the canonical name, with each model’s covariateData[[DOSE_LOW]]$notes documenting the per-paper dose levels that map to 1. Drug-suffixed sibling DOSE_LOW_AMG221 enshrines its threshold in the name instead. Encode as an explicit indicator column rather than deriving it from an inequality on a continuous dose covariate such as [[DOSE_RIMEGEPANT_MG]] – papers define these groups by the studied dose levels, not by a threshold, so an inequality would silently misclassify an intermediate dose that was never studied. Scope: general. The entry was founded at Scope: specific with the condition “promote to general once a second model registers under the same abstract semantics”; Zhang_2016_burosumab.R is that second model (a step-function shift in a PK parameter at the bottom of the studied dose range, against a reference of any higher level), and the abstract semantics and reference category are paper-independent even though the mg values and the affected parameter are not. Note that the indicator is per-dose-record, not per-subject: it is subject-level only when each subject received a single dose level (Comisar 2025), and time-varying in intradose-escalation designs (Zhang 2016).

STUDY1 (canonical for Study-1 cohort indicator)

• Description: 1 = subject enrolled in Study 1 of the Cirincione 2017 pooled analysis, 0 = other. Used to switch the residual-error magnitude per study.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (all other studies; combined with STUDY5 = 0 selects the pooled “other” residual error).
• Source aliases:
  • DVID = "study1" (character-valued study identifier; STUDY1 = as.integer(DVID == "study1")) – legacy form previously used in Cirincione_2017_exenatide.R.
• Example models: Cirincione_2017_exenatide.R.
• Notes: Paired with STUDY5. When both are 0, the subject is in the pooled “other studies” residual-error group.

STUDY5 (canonical for Study-5 cohort indicator)

• Description: 1 = subject enrolled in Study 5 of the Cirincione 2017 pooled analysis, 0 = other. Used to switch the residual-error magnitude per study.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (all other studies).
• Source aliases:
  • DVID = "study5" (character-valued study identifier; STUDY5 = as.integer(DVID == "study5")) – legacy form previously used in Cirincione_2017_exenatide.R.
• Example models: Cirincione_2017_exenatide.R.
• Notes: Paired with STUDY1. When both are 0, the subject is in the pooled “other studies” residual-error group.

STUDY_MD (canonical for Cirincione 2017 AAPS J ER exenatide multi-dose study cohort indicator)

• Description: 1 = subject enrolled in the phase II multi-dose study (weekly SC ER exenatide for 15 weeks) of the Cirincione 2017 AAPS J combined single- and multiple-dose population analysis; 0 = phase II single-dose study (one SC ER exenatide dose). Used to switch the study-specific relative bioavailability (f_rel) value and the log-scale residual-error magnitude per study.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (single-dose phase II study).
• Source aliases: derived per subject from the dose-record structure (a subject with > 1 dose record in the multi-dose phase II study -> 1; subjects with a single dose record in the single-dose phase II study -> 0).
• Example models: Cirincione_2017_exenatide_er.R.
• Notes: Cirincione 2017 AAPS J Table II reports f_rel(single-dose study) = 8.86% and f_rel(MD study) = 15.5%, and log-scale residual SDs 0.684 (single-dose) vs 0.376 (multi-dose). The STUDY_MD indicator selects between them. For the phase III external validation cohort (multi-dose weekly 2 mg ER for 24 weeks), use STUDY_MD = 1 so the multi-dose f_rel and residual magnitudes apply. Specific scope because the indicator is tied to the AAPS J 2017 ER exenatide combined analysis (single-dose phase II + multi-dose phase II) – a future pooled analysis with additional study cohorts would justify its own canonical or a promotion to general.

STUDY_MARIBAVIR_PHASE1 (canonical for phase 1 study cohort indicator in the Sun 2025 maribavir pooled analysis)

• Description: 1 = the concentration record originates from a phase 1 maribavir study in the Sun 2025 pooled population PK analysis (206 individuals, 4231 records: healthy volunteers plus the hepatic-impairment, renal-impairment and stable-renal-transplant cohorts, and the NCT05319353 Japanese-descended / non-Hispanic Caucasian study); 0 = the record originates from a phase 2/3 study (SHP620-202, -203, -302, -303, which includes SOLSTICE and AURORA; 724 individuals, 3200 records). Used ONLY to switch the proportional residual-error magnitude; it has no structural effect on any PK parameter.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (phase 2/3 study).
• Source aliases:
  • STUDY – the NONMEM control-stream study-identifier column. The $ERROR block selects the phase 2/3 proportional term with IF(STUDY.EQ.202.OR.STUDY.EQ.203.OR.STUDY.EQ.303.OR.STUDY.EQ.302), so the binary indicator is derived as STUDY_MARIBAVIR_PHASE1 = 1 for any study identifier outside that set.
• Example models: Sun_2025_maribavir.R (switches the proportional residual SD between 0.265518 for phase 1 and 0.397492 for phase 2/3).
• Notes: Auto-approved member of the STUDY_<id> family, and a direct sibling of [[STUDY_NIPOCALIMAB_PHASE1]] and [[STUDY_BALCINRENONE_PHASE1]] – all three are phase-1-versus-later-phase cohort indicators whose sole role is to switch a study-specific residual-error magnitude in a pooled analysis. Sun 2025 attributes the difference to the different LC-MS assays, with different lower limits of quantification, used in the phase 1 versus the phase 2/3 studies (proportional residual variance 0.0705 vs 0.158, i.e. SD 0.266 vs 0.397). The paper-specific MARIBAVIR token is required because the generic PHASE1 name is ambiguous across pooled analyses that pick opposite reference categories; see the STUDY_NIPOCALIMAB_PHASE1 source-alias note. Distinct from [[DIS_CMV]], which carries the healthy-versus-transplant-recipient contrast structurally on CL/F: the two columns are correlated in the Sun 2025 dataset (every phase 1 participant is DIS_CMV = 0) but encode different things, and a downstream user simulating a phase 1 healthy cohort needs both set to their phase 1 values.

STUDY_PKU015 (canonical for sapropterin PKU-015 pediatric study cohort indicator)

• Description: 1 = subject enrolled in study PKU-015 (pediatric population pharmacokinetic study of sapropterin in infants and young children, 0-6 years old, of the Qi 2014 pooled analysis); 0 = study PKU-004 (adolescent / adult open-label extension study, >= 9 years old). Used to switch the residual-error magnitude per study under the log-transform-both-sides (LTBS) constant-CV residual model.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (PKU-004 adolescent / adult cohort).
• Source aliases: derived per subject from the trial identifier (PKU-015 -> 1, PKU-004 -> 0).
• Example models: Qi_2014_sapropterin.R.
• Notes: Qi 2014 Table 3 reports separate residual-error estimates for the two studies – PKU-004 = 21.1% CV, PKU-015 = 30.2% CV under the LTBS approach. The STUDY_PKU015 indicator selects between them. Specific scope because the indicator is tied to the BioMarin sapropterin clinical-development program (PKU-004 = phase 3b extension, PKU-015 = phase 3b pediatric).

STUDY_EDELWEISS (canonical for the EDELWEISS phase 2b linzagolix study cohort indicator)

• Description: 1 = record from the EDELWEISS phase 2b dose-ranging study (NCT02778399; linzagolix 25-200 mg once daily vs placebo in women with endometriosis-associated pain) of the Pohl 2022 pooled analysis; 0 = record from any other pooled study (KLH1101, 16-OBE2109-011, 17-OBE2109-008, KLH1204). Used to switch the proportional residual-error magnitude per study. Per-record (study-fixed) binary indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (all other pooled studies, which contributed rich PK sampling).
• Source aliases: derived per subject from the trial identifier (EDELWEISS / NCT02778399 -> 1, all other study identifiers -> 0).
• Example models: Pohl_2022_linzagolix.R.
• Notes: Pohl 2022 Table 4 reports two proportional residual-error variances – 0.118 for EDELWEISS and 0.0389 for all other studies – and the STUDY_EDELWEISS indicator selects between them. EDELWEISS is the only phase 2b study in the pool and is the only one that contributed sparse PK (predose plus a single postdose sample at most visits), which is why its residual error is roughly threefold larger in variance than the rich-sampling phase 1 studies. Sibling of STUDY_PKU015 and STUDY_FARLETUZUMAB_PHASE2, which serve the same residual-error-switch role in their own analyses. Specific scope because the indicator is tied to the ObsEva linzagolix clinical-development program.

STUDY_GERECKE (canonical for the Gerecke 2021 PIKRT cohort indicator in the Gotz 2025 pooled fosfomycin analysis)

• Description: 1 = subject enrolled in study B (Gerecke 2021, J Antimicrob Chemother 2021;77:169-173, doi:10.1093/jac/dkab357; eight critically ill patients on prolonged-intermittent kidney replacement therapy with a Genius 90 batch dialysis system) of the Gotz 2025 four-study pooled fosfomycin analysis; 0 = any of studies A (Dimski 2021), C (Huppe 2023) or D (Parker 2015), and 0 for any new patient being simulated. Unlike its STUDY_* siblings, which switch a residual-error magnitude or a relative bioavailability between values, this indicator is a hard structural gate: it multiplies the body-clearance arm by zero, removing that elimination pathway entirely for the cohort.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (studies A, C, D, or a newly simulated patient).
• Source aliases: derived per subject from the pooled-analysis study identifier (study B -> 1; study A / study C / study D -> 0).
• Example models: Gotz_2025_fosfomycin.R (gates the body-clearance arm to zero: cl_body <- diuresis * (1 - STUDY_GERECKE) * exp(lcl_body + etalcl_body) * (CRCL/48.4)^e_crcl_cl_body).
• Notes: Gotz 2025 Table 2 footnote b prints the gate literally as (x 0 if study B), and Results explains it: “CL_body was fixed to 0 for study B since these patients presented no fosfomycin elimination without KRT (Fig. 2), which led to individual CL_body estimates close to zero.” The gate is NOT redundant with the anuria gate that the same model carries on URINE_VOL_24H: study B’s median (IQR) 24-hour urine output was 700 (500-1000) mL, so no study B patient meets that model’s < 100 mL anuria criterion, and both gates have to be carried separately. Study B contributed 8 of the 45 pooled patients (17.8%) and also had no height, BMI, BSA or laboratory markers recorded, which is why Gotz 2025 additionally reports a sensitivity re-fit on a reduced data set (n = 37) that excludes it; Table 2 shows the two parameter sets to be consistent. Set to 0 for simulation – the paper’s own Monte Carlo scenarios (Methods, Simulations; Fig. S6) never invoke it. Specific scope because the indicator is tied to one named cohort of one pooled analysis.

STUDY_BELL (canonical for the Bell 2011 cohort indicator in the de Kock 2018 pooled sulfadoxine-pyrimethamine analysis)

• Description: 1 = the patient is from the Bell et al. trial at Chileka, Malawi (n = 102), one of the four trials pooled in the de Kock 2018 sulfadoxine-pyrimethamine analysis; 0 = the patient is from the Barnes, Tekete or Allen trials. Multiplicative covariate on apparent pyrimethamine clearance.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the Barnes, Tekete and Allen trials, which share the structural pyrimethamine clearance).
• Source aliases:
  • reference 5 – the de Kock 2018 Table 2 row label is “Difference from clearance in reference 5 (%)”, where reference 5 of that paper is Bell et al.; the Results text names the study explicitly (“The study of Bell et al. (5) showed a 54.9% lower pyrimethamine clearance”).
• Example models: deKock_2018_sulfadoxinePyrimethamine.R (cl_pyra <- ... * (1 + e_study_bell_cl_pyra * STUDY_BELL) with e_study_bell_cl_pyra = -0.549, i.e. apparent pyrimethamine clearance 54.9% lower in this cohort; the sulfadoxine arm carries no corresponding effect).
• Notes: Registered as a STUDY_<id> rather than a REGION_<site> indicator because de Kock 2018 attributes the difference to the trial’s bioanalysis, not its geography: Bell et al. “was the only one that assayed whole-blood liquid samples (capillary blood dried-spot samples were assayed in all the other studies), and its samples were assayed in a different lab”, so the paper could not determine whether the shift reflects matrix, assay method or population. Study and site are perfectly confounded here – Chileka is the only Bell et al. site – so there is deliberately no REGION_CHILEKA column; Chileka sits in the reference group for both drugs’ observation scaling, and a Bell et al. subject therefore sets STUDY_BELL = 1 with every REGION_* site indicator at 0. Set to 0 when simulating a new patient: de Kock 2018’s own dose-optimisation simulations use the reference-site parameters. Author-named like STUDY_GERECKE, STUDY_SALEM and STUDY_CHEN, which follow the same convention for pooled analyses whose component trials have no protocol number in the source text.

STUDY_FARLETUZUMAB_PHASE2 (canonical for Phase II study cohort indicator in the Farrell 2012 farletuzumab pooled analysis)

• Description: 1 = subject enrolled in the Phase II study (MORAb-003-002) of the Farrell 2012 pooled farletuzumab analysis; 0 = Phase I study (MORAb-003-001). Used to switch the residual-error magnitude per study.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (Phase I).
• Source aliases:
  • PHASE2 – prior canonical name (pre-2026-06-19 paper-specific-disambiguation standardization). Generic single-token name collided with the parallel PHASE1 canonical from Valenzuela 2025 that picks the opposite reference category.
  • Derived per subject from the trial identifier (MORAb-003-001 -> 0, MORAb-003-002 -> 1).
• Example models: Farrell_2012_farletuzumab.R.
• Notes: Farrell 2012 Table 3 reports separate residual-error estimates for the two studies – Phase I uses a proportional-only model (sigma = 20.5%); Phase II uses a combined additive + proportional model (sigma_prop = 34.9%, sigma_add = 7.94 ug/mL). The STUDY_FARLETUZUMAB_PHASE2 indicator selects between them.

STUDY_PEDIATRIC (canonical for the pediatric-trial cohort indicator in the Chen 2025 hemoporfin pooled pediatric + adult analysis)

• Description: 1 = the subject is from the pediatric port-wine-stain trial of the Chen 2025 hemoporfin analysis (NCT03125057, patients aged 7-14 years, sparse sampling, photodynamic therapy administered); 0 = the subject is from the 2012 adult healthy-volunteer phase I trial (aged 20-45 years, intensive sampling, no photodynamic therapy). Used as the inter-trial (“study effect”) indicator on relative bioavailability, so that the allometric body-size relationship can be estimated across a pooled pediatric + adult data set without the trial difference confounding the exponent.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the adult healthy-volunteer phase I trial). Enters as the additive-on-one form F1 = 1 + theta_study * STUDY_PEDIATRIC (Chen 2025 Equation 2), not as a multiplicative or exponential form – so the fitted theta_study is directly the fractional change in bioavailability, and setting the indicator to 0 recovers F1 = 1 exactly. Value observed: theta_study = -0.175 (Chen 2025 Table 2), i.e. pediatric-trial exposure 17.5% below the allometric prediction.
• Source aliases:
  • STUDY – the $INPUT-derived variable in the Chen 2025 Data S2 final control stream, set from the subject identifier (IF (ID.LT.1000) STUDY=1, IF (ID.GT.1000) STUDY=0).
• Example models: Chen_2025_hemoporfin.R (founding example; f(central) <- 1 + e_study_f * STUDY_PEDIATRIC on a three-compartment intravenous model whose six disposition parameters are otherwise scaled only by fat-free mass).
• Notes: Specific scope until a second model ratifies the name. Member of the STUDY_<id> family; STUDY_PKU015 is the closest structural analogue, being likewise a pediatric-versus-adult trial indicator in a pooled pediatric + adult population PK analysis, though STUDY_PKU015 switches a residual-error magnitude where this one switches bioavailability. Named for the cohort rather than for the NCT number because the paper itself calls the parameter the “study effect” and identifies the cohorts as pediatric versus adult throughout; a second pooled pediatric + adult analysis needing an inter-trial offset of a different kind should register a paper-specific sibling rather than overload this name. Note that Chen 2025 deliberately does not apply this offset in its dose-selection simulations – the authors state that the cause of the study effect (study operations, bioanalytical method, or subject physiology) is unknown, so the recommended pediatric regimens are derived with the indicator effectively set to 0. A downstream user reproducing the paper’s dosing tables must do the same; a user reproducing the observed pediatric concentrations must set it to 1.

STUDY_BALCINRENONE_PHASE1 (canonical for single-dose phase 1 study cohort indicator in the Parkinson 2025 balcinrenone pooled analysis)

• Description: 1 = the participant is from one of the four single-dose phase 1 studies of the Parkinson 2025 pooled balcinrenone analysis (NCT03843060 drug-drug-interaction, NCT03804645 and NCT04798222 bioavailability, NCT04469907 renal impairment – healthy participants and participants with renal impairment but without heart failure); 0 = the participant is from one of the two multiple-dose phase 1b/2b studies in patients with heart failure and chronic kidney disease (NCT03682497, NCT04595370). Used as a multiplicative covariate on apparent clearance.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (phase 1b/2b patient with heart failure and CKD). This matches the paper’s own reference participant, defined in Methods 2.6 as ‘a patient with HF and CKD… with an eGFR of 60 mL/min/1.73 m2 who received balcinrenone in a fasted state’.
• Source aliases:
  • PTSFLAG – NONMEM $PK column name used in Parkinson_2025_balcinrenone.R, coded with the opposite polarity (PTSFLAG = 1 for the patient studies, which are the most common group at 135 of 189 participants). The canonical value is STUDY_BALCINRENONE_PHASE1 = 1 - PTSFLAG. The reference category is unchanged by the recoding: the control stream sets CLPTSFLAG = 1 on the PTSFLAG.EQ.1 branch and (1 + THETA(12)) on the PTSFLAG.EQ.0 branch, so the patient level carries the unit multiplier under either encoding.
• Example models: Parkinson_2025_balcinrenone.R (multiplicative fractional effect cl *= (1 + 1.06 * STUDY_BALCINRENONE_PHASE1), i.e. 2.06-fold higher CL/F in the phase 1 participants, corresponding to 0.48-fold AUCss; Parkinson 2025 Table 1 CL/F~Study = 1.06, 95% CI 0.685-1.44, and Results 3.3).
• Notes: A study-design cohort indicator rather than a mechanistic subject covariate, and the paper says so: the effect survives adjustment for both baseline eGFR and body weight, and the Discussion offers three competing explanations (single-dose phase 1 versus repeated-dose patient designs, i.e. possible time-dependent PK; less accurate dosing- and sampling-time records in the patient studies; and unrecorded food state in the patient studies, where the fed state was assumed). It is partially collinear with DIS_HEALTHY in this dataset but not identical – the phase 1 stratum also contains the renal-impairment cohort NCT04469907, whose participants are not healthy – so DIS_HEALTHY would misdescribe the level. Follows the paper-specific STUDY_<drug>_<phase> convention established by STUDY_NIPOCALIMAB_PHASE1, STUDY_FARLETUZUMAB_PHASE2, STUDY_NMV_PHASE23, STUDY_POSA_PHASE3, STUDY_SULDUR_PHASE2 and STUDY_SULDUR_PHASE3; a generic PHASE1 name is deliberately avoided because different papers pick opposite reference categories (note that STUDY_NIPOCALIMAB_PHASE1 shares the 1-level meaning but inverts nothing, whereas this column inverts its source column).

STUDY_NIPOCALIMAB_PHASE1 (canonical for Phase I study cohort indicator in the Valenzuela 2025 nipocalimab pooled analysis)

• Description: 1 = subject enrolled in a Phase 1 study of the Valenzuela 2025 pooled nipocalimab analysis (MOM-M281-001, MOM-M281-007, MOM-M281-010, EDI1001, EDI1002 – healthy participants); 0 = Phase 2 study (MOM-M281-004 / Vivacity-MG – participants with gMG). Used to switch the proportional PK residual-error magnitude per study phase.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (Phase 2).
• Source aliases:
  • PHASE1 – prior canonical name (pre-2026-06-19 paper-specific-disambiguation standardization). Generic single-token name collided with the parallel PHASE2 canonical from Farrell 2012 that picks the opposite reference category.
  • Derived per subject from the trial identifier (Phase 1 protocols -> 1, NCT03772587 Vivacity-MG -> 0).
• Example models: Valenzuela_2025_nipocalimab.R.
• Notes: Valenzuela 2025 Table 3 reports proportional PK residual 0.0834 (Phase 1) vs 0.367 (Phase 2). Distinct from Farrell 2012 STUDY_FARLETUZUMAB_PHASE2 – the reference category is inverted (Valenzuela 2025 picks Phase 1 as the 1-level).

STUDY_IMEGLIMIN_PHASE1 (canonical for phase I study cohort indicator in the Tomita 2022 imeglimin pooled analysis)

• Description: 1 = subject enrolled in one of the four imeglimin clinical pharmacology studies of the Tomita 2022 pooled analysis (single/multiple ascending dose, Western renal impairment, Japanese renal impairment, and T2DM with renal impairment; 160 subjects); 0 = phase II / III study. Used to switch the residual-error magnitude.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (phase II / III).
• Source aliases: derived per subject from the study (Tomita 2022 Table 1, Table 2 ‘Phase I’ column).
• Example models: Tomita_2022_imeglimin.R (log-scale residual SD 0.190 in phase I vs 0.359 in phase II / III post-dose samples; pre-dose samples use 0.505 via SAMPLE_PREDOSE).
• Notes: Member of the STUDY_<id> family; same role as STUDY_NIPOCALIMAB_PHASE1.

STUDY_WX0593_002 (canonical for the WX-0593-002 food-effect crossover study cohort indicator in the Yang 2025 iruplinalkib pooled analysis)

• Description: 1 = the observation record originates from WX-0593-002 (CTR20190737), the randomized two-period crossover food-effect study in 16 healthy Chinese volunteers dosed 120 mg fasted then fed (or the reverse) with a 7-day washout; 0 = the record originates from one of the three solid-tumor studies in the pooled analysis (WX-0593-001 phase 1 dose escalation/expansion, WX-0593-003 single-arm phase 2, WX-0593-004 randomized phase 3; 376 subjects). Used to switch the proportional residual-error magnitude between the two study groups.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the pooled solid-tumor studies WX-0593-001, -003 and -004).
• Source aliases:
  • Study – Yang 2025 Table 2 splits the residual-error rows by protocol number rather than naming a covariate column; the indicator is derived per record from the protocol identifier (WX-0593-002 -> 1, all others -> 0).
• Example models: Yang_2025_iruplinalkib.R (Yang 2025 Table 2 final model: proportional residual 55.7% for WX-0593-002 against 35.7% for WX-0593-001, -003 and -004; encoded as the indicator-weighted switch propSd <- propSdHv * STUDY_WX0593_002 + propSdPatient * (1 - STUDY_WX0593_002)).
• Notes: Carries no structural or covariate effect – residual error only. Worth recording that the direction is the counter-intuitive one: WX-0593-002 is the densely sampled study (21 timepoints per crossover period, against predose-and-a-couple-of-points trough sampling in the phase 2 and phase 3 studies, Table S1) and the healthiest cohort, yet it carries the larger residual. Yang 2025 does not comment on why. The same indicator-weighted residual-switch construction is used by STUDY_NIPOCALIMAB_PHASE1 in Valenzuela_2025_nipocalimab.R and STUDY_NMV_PHASE23 in Chan_2023_nirmatrelvir.R. Note that the healthy-versus-patient contrast is NOT interchangeable with DIS_HEALTHY here: Yang 2025 screened subject type (healthy subject or patient) as a covariate on the structural parameters and did not retain it, so the only thing that distinguishes the healthy-volunteer study in the final model is this residual magnitude. The two columns would coincide numerically in this particular dataset, since WX-0593-002 enrolled all 16 healthy volunteers and no patients, but they mean different things and the paper’s own Table 2 keys the split on the protocol.

STUDY_MS200095_0028 (canonical for the tepotinib hepatic impairment study MS200095-0028 indicator)

• Description: 1 = record from study MS200095-0028 (NCT03546608), the phase 1 single-dose (500 mg TF2, fed) study in 6 healthy participants and 12 participants with Child-Pugh A or B hepatic impairment; 0 = record from any of the other 11 studies in the pooled tepotinib analysis.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (all other studies).
• Source aliases:
  • Study MS200095-0028 – Xiong 2022 Table 3 row label; study code 0028 in Table 1.
• Example models: Xiong_2022_tepotinib.R (fractional effect on tepotinib apparent clearance, -0.115).
• Notes: Study-level nuisance effect; set to 0 when simulating new patients.

STUDY_AC220019 (canonical for the quizartinib AC220-019 food-effect study cohort indicator in the Kang 2020 pooled analysis)

• Description: 1 = the record originates from AC220-019, the phase 1 food-effect study in 64 healthy volunteers given a single 30 mg quizartinib dose fasted or fed; 0 = the record originates from any of the other seven studies pooled in Kang 2020 (four other healthy-volunteer phase 1 studies and three studies in relapsed/refractory AML).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the other seven studies).
• Source aliases:
  • study == AC220-019 – Kang 2020 names no covariate column; equation 3 defines F1S19_i as the relative bioavailability of ‘the ith participant enrolled in AC220-019’, in place of equation 2 for everyone else.
• Example models: Kang_2020_quizartinib.R (switches the quizartinib relative bioavailability from equation 2 to the study-specific equation 3, F1 = 0.913 * (1 + 0.0509 * FED)).
• Notes: Pair with FED: Kang 2020 estimated the food effect (-51.2% on ka, +5.09% on F1) only inside AC220-019 because no other study recorded meal data comprehensively, so FED should be 0 for every record where STUDY_AC220019 is 0.

STUDY_NMV_PHASE23 (canonical for phase II/III study cohort indicator in the Chan 2023 nirmatrelvir pooled analysis)

• Description: 1 = the observation record originates from the phase II/III EPIC-HR study (NCT04960202; 1087 nonhospitalized symptomatic adults with COVID-19, sparse outpatient PK sampling); 0 = the observation record originates from one of the seven phase I studies in the pooled analysis (150 participants, serial in-clinic sampling). Used to switch the proportional residual-error magnitude per study phase.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (phase I study observation).
• Source aliases:
  • PROT (protocol number; the phase II/III study is protocol 1005) – used in Chan_2023_nirmatrelvir.R (Chan 2023 Data S1 $ERROR: IF (PROT.EQ.1005) W = SQRT(THETA(12)**2 + THETA(7)**2/IPRED**2)).
• Example models: Chan_2023_nirmatrelvir.R (selects propSdPhase23 = 1.39 in place of propSdPhase1 = 0.324 inside the combined additive-plus-proportional residual model; Chan 2023 Table 3 reports proportional error 32.4% for phase I and 139% for phase II/III, with the additive component fixed at the 10 ng/mL LLOQ for both).
• Notes: A record-level study-design property, not a subject-level covariate: it captures the sparse outpatient sampling scheme of EPIC-HR (Chan 2023 Discussion attributes the roughly fourfold higher residual variability to sparse sampling, the outpatient setting, and dosing-time compliance). Collinear with DIS_COVID19 in the Chan 2023 data set – every phase II/III participant has COVID-19 – but the two enter different parts of the model (DIS_COVID19 is a structural covariate on CL; STUDY_NMV_PHASE23 only selects a residual-error magnitude), so both are carried. Follows the paper-specific STUDY_<drug>_<phase> convention established by STUDY_NIPOCALIMAB_PHASE1, STUDY_FARLETUZUMAB_PHASE2, STUDY_POSA_PHASE3, STUDY_SULDUR_PHASE2 and STUDY_SULDUR_PHASE3; a generic PHASE23 name is deliberately avoided because different papers pick opposite reference categories.

STUDY_MOV_PHASE23 (canonical for phase II / phase II-III trial cohort indicator in the Bihorel 2023 molnupiravir pooled analysis)

• Description: 1 = the NHC observation record originates from one of the three trials in participants with COVID-19 – MK-4482-006 (phase IIa, n = 66), MK-4482-001 / MOVe-IN (phase II hospitalized, n = 196) or MK-4482-002 / MOVe-OUT (phase II/III non-hospitalized, n = 845) – all of which used sparse outpatient or ward sampling of 1-5 samples per participant taken around the ninth or tenth dose; 0 = the record originates from the phase I trial MK-4482-004 (100 healthy adults, dense in-clinic sampling of 7-26 samples per participant). Used to switch the proportional residual-error magnitude per study phase.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (phase I trial MK-4482-004 observation).
• Source aliases:
  • STUDY (the trial identifier; the phase I trial is MK-4482-004) – used in Bihorel_2023_molnupiravir.R. Bihorel 2023 Results, “Base structural model development using phase I (stage 1) and phase II data (stage 2)”, lists “distinct residual variability models being estimated for the phase I and II studies” among the stage-2 refinements, and Table 2 carries the two magnitudes on rows labelled “Phase I trials” and “Phase II/III trials”.
• Example models: Bihorel_2023_molnupiravir.R (selects propSdPhase23 = sqrt(0.247) = 0.497 in place of propSdPhase1 = sqrt(0.0652) = 0.255 in the proportional residual model; Bihorel 2023 Table 2 prints the variances 0.247 and 0.0652 with their square roots as 49.7 %CV and 25.5 %CV).
• Notes: A record-level study-design property, not a subject-level covariate: it selects only a residual-error magnitude, so it affects the simulated observation and not the typical-value or individual-prediction profile. Nearly collinear with hospitalization and disease status in the Bihorel 2023 data set – every phase II/III participant has COVID-19 and every phase I participant is healthy – but no disease covariate survived the structural covariate search in that paper, so this indicator is the only place the phase distinction enters the model. The roughly twofold higher residual variability tracks the sampling scheme: 72.8% of the pooled population contributed at most two samples, essentially all of them from the phase II/III trials. Follows the paper-specific STUDY_<drug>_<phase> convention; the closest analogue is STUDY_NMV_PHASE23, the same construct for the Chan 2023 nirmatrelvir COVID-19 antiviral pooled analysis, which likewise splits phase I dense sampling from phase II/III sparse sampling. MOV is the conventional abbreviation for molnupiravir, matching NMV for nirmatrelvir.

STUDY_AALL07P1 (canonical for the Children’s Oncology Group AALL07P1 trial cohort indicator in the Hanley 2017 pediatric bortezomib analysis)

• Description: 1 = subject enrolled in Children’s Oncology Group study AALL07P1 (NCT00873093; phase 2; bortezomib added to a 3-block reinduction regimen in relapsed acute lymphoblastic leukemia; N = 51), 0 = subject enrolled in Children’s Oncology Group study AAML1031 (NCT01371981; phase 3; bortezomib added to a 4-course chemotherapy backbone in de novo acute myelogenous leukemia; N = 53). Used to switch the log-scale residual-error magnitude between the two pooled trials.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (study AAML1031, de novo AML).
• Source aliases: derived per subject from the trial identifier (AALL07P1 -> 1, AAML1031 -> 0).
• Example models: Hanley_2017_bortezomib.R (switches the log-additive residual SD on Cc between expSdAall07p1 (46.8 %CV) and expSdAaml1031 (21.9 %CV); Hanley 2017 Table 2 rows “Residual error for study AALL07P1,%CV” and “Residual error for study AAML1031,%CV”).
• Notes: Subject-level and time-fixed. Carries no structural effect. Disease type (ALL vs AML) was screened as a covariate on CL, V1 and Q3 and was not statistically significant, and VPCs stratified by disease type showed no heterogeneity, so this indicator selects only a residual-error magnitude and does not move the typical-value or individual-prediction profile. The near-twofold difference in residual magnitude tracks the sampling scheme rather than the disease: AALL07P1 sampled predose, 5-15 min, 30-60 min, 4-8 h and ~72 h postdose, whereas AAML1031 additionally sampled 18-30 h, and adding the per-study residual dropped the OFV by more than 52 points (Hanley 2017 Results, “Final model development”). Same construct as STUDY_MOV_PHASE23, STUDY_NMV_PHASE23, STUDY_POSA_PHASE3 and STUDY_ARI_PHASE3 – a pooled-analysis study indicator whose only role is a residual-error switch – but named for the trial acronym rather than a phase, because both contributing trials differ in disease and phase simultaneously and “phase 2 vs phase 3” would not identify the cohort. The screened-but-unretained disease, risk-group and treatment-plan-stratum covariates of the same paper are carried in that model’s covariatesDataExcluded and have no register entries.

STUDY_POSA_PHASE3 (canonical for posaconazole phase 3 patient study cohort indicator in the van Iersel 2018 pooled analysis)

• Description: 1 = subject enrolled in the phase 3 patient study P05615 of the van Iersel 2018 pooled posaconazole solid-tablet PopPK analysis (231 patients at high risk for invasive fungal disease, sparse PK sampling); 0 = subject enrolled in one of the five phase 1 healthy-volunteer studies P04975, P05637, P07764, P07783, P07691 (104 healthy volunteers, rich PK sampling). Used to switch the log-additive PK residual-error magnitude per study phase.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (phase 1 healthy-volunteer studies).
• Source aliases: derived per subject from the trial identifier (P05615 -> 1, any of P04975 / P05637 / P07764 / P07783 / P07691 -> 0).
• Example models: vanIersel_2018_posaconazole.R (switches the log-additive residual SD on Cc between expSd_p1 = 0.42 (phase 1) and expSd_p3 = 0.322 (phase 3), per van Iersel 2018 Table 2 final-model ‘SD (phase 1 studies)’ and ‘SD (phase 3 study)’).
• Notes: Specific scope because the contrast is tied to the van Iersel 2018 posaconazole solid-tablet clinical-development pooled analysis. Drug-specific paper-anchored member of the STUDY_<DRUG>_PHASE<N> family alongside STUDY_NIPOCALIMAB_PHASE1 (Valenzuela 2025) and STUDY_FARLETUZUMAB_PHASE2 (Farrell 2012); distinct from those entries because the reference category (phase 1) and 1-level (phase 3) are paper-specific to van Iersel 2018. Subject-level (time-fixed); set once from the trial identifier on each subject record.

STUDY_ARI_PHASE3 (canonical for the phase 3 study cohort indicator in the Wang 2022 pooled oral-plus-AOM aripiprazole analysis)

• Description: 1 = the observation record comes from the sparse-sampling phase 3 study 31-07-246 of the Wang 2022 pooled aripiprazole population PK analysis (predose samples plus single samples on days 7, 14 and 28 after dosing); 0 = the observation record comes from one of the four rich-sampling phase 1 studies 31-98-206, 31-98-207, CN138020 and 31-05-244. Used to switch the proportional residual-error magnitude per study phase.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (phase 1 study observation record).
• Source aliases: derived per observation record from the study identifier (31-07-246 -> 1; 31-98-206 / 31-98-207 / CN138020 / 31-05-244 -> 0). Wang 2022 Table 1 labels the two magnitudes simply “Phase 1 RV (%CV)” and “Phase 3 RV (%CV)”.
• Example models: Wang_2022_aripiprazole.R (selects propSdPhase3 = 0.2811 in place of propSdPhase1 = 0.2423 inside the proportional residual model; Wang 2022 Table 1 reports 24.23 %CV for phase 1 and 28.11 %CV for phase 3).
• Notes: A record-level study-design property, not a subject-level covariate. Drug-specific paper-anchored member of the STUDY_<DRUG>_PHASE<N> family alongside STUDY_NMV_PHASE23 (Chan 2023), STUDY_POSA_PHASE3 (van Iersel 2018), STUDY_NIPOCALIMAB_PHASE1 (Valenzuela 2025) and STUDY_FARLETUZUMAB_PHASE2 (Farrell 2012); a generic PHASE3 name is deliberately avoided because different papers pick opposite reference categories. Not collinear with route of administration in this analysis: the phase 1 studies contributed both oral and intramuscular aripiprazole-once-monthly records, and the phase 3 study contributed both oral lead-in and AOM records, so a model may carry this indicator alongside route-specific absorption parameters without aliasing. Wang 2022 attributes the larger phase 3 residual to the sparse sampling scheme. A sixth study, phase 3 study 31-07-247, was used only for external validation and contributes no records to the fit, so it has no level in this column.

STUDY_APPRAISE1 (canonical for the APPRAISE-1 phase II acute coronary syndrome study indicator)

• Description: 1 = the observation record comes from APPRAISE-1, the global phase II apixaban study in subjects with acute coronary syndrome (Alexander 2009, Circulation 119:2877-2885; 951 subjects, 2.5 and 10 mg b.i.d. and 10 and 20 mg q.d. for 26 weeks, random samples at weeks 3 and 26); 0 = otherwise. A record-level study-design property, not a subject-level covariate.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0. Paired with STUDY_ARISTOTLE; both indicators 0 selects the pooled stratum of the eight phase I healthy-volunteer studies plus the Japan NVAF phase II and Japan ACS phase II studies.
• Source aliases:
  • STUDYID value 23 – the Cirincione 2018 control-stream branch IF (STUDYID.EQ.23) W=THETA(22).
• Example models: Cirincione_2018_apixaban.R, Cirincione_2018_apixaban_asian_subgroups.R (selects the log-scale residual SD 0.668 in place of the pooled 0.31; APPRAISE-1 carries by far the largest residual variability of the three strata, which the design explains – it is the only contributing study sampled entirely at random times relative to dosing).
• Notes: Well-formed member of the auto-approved STUDY_<id> family, named for the trial acronym rather than a phase number because the apixaban programme contributes three phase II studies with materially different residual magnitudes, so a STUDY_APIXABAN_PHASE2 name would be ambiguous. Distinct from APPRAISE-2, a separate and larger phase III ACS trial that contributed no records to this analysis.

STUDY_ARISTOTLE (canonical for the ARISTOTLE phase III nonvalvular atrial fibrillation study indicator)

• Description: 1 = the observation record comes from ARISTOTLE, the global phase III apixaban-versus-warfarin study in subjects with nonvalvular atrial fibrillation (Granger 2011, N Engl J Med 365:981-992); 0 = otherwise. The pharmacokinetic subset contributing to the Cirincione 2018 analysis is 2,932 subjects dosed 5 mg b.i.d. (2.5 mg b.i.d. in subjects meeting the dose-reduction criteria) for 6 months, each supplying a single random sample at month 2. A record-level study-design property, not a subject-level covariate.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0. Paired with STUDY_APPRAISE1; both indicators 0 selects the pooled stratum of the eight phase I healthy-volunteer studies plus the Japan NVAF phase II and Japan ACS phase II studies.
• Source aliases:
  • STUDYID value 30 – the Cirincione 2018 control-stream branch IF (STUDYID.EQ.30) W=THETA(23).
• Example models: Cirincione_2018_apixaban.R, Cirincione_2018_apixaban_asian_subgroups.R (selects the log-scale residual SD 0.460 in place of the pooled 0.31; ARISTOTLE is the study whose addition defines the paper’s stage 2 analysis, and its one-sample-per-subject sparse design is why the stage 1 models had to be refitted rather than simply extended).
• Notes: Well-formed member of the auto-approved STUDY_<id> family. Distinct from the ARISTOTLE-J study (Ogawa 2011, Circ J 75:1852-1859), the separate Japanese phase II NVAF trial that contributes to the pooled reference stratum of this same model rather than to this indicator.

STUDY_SULDUR_PHASE2 (canonical for Cammarata 2024 sulbactam-durlobactam phase 2 study cohort indicator)

• Description: 1 = subject enrolled in the Phase 2 study CS2514-2017-0003 (complicated urinary tract infection including acute pyelonephritis) of the Cammarata 2024 pooled sulbactam-durlobactam popPK analysis; 0 = otherwise. Paired with STUDY_SULDUR_PHASE3; both indicators 0 selects the six Phase 1 studies (the reference stratum). Used to switch the DURLOBACTAM residual-error magnitude between the three study phases; sulbactam residual variability is not phase-stratified.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 with STUDY_SULDUR_PHASE3 also 0 (Phase 1 studies CS2514-2016-0001, CS2514-2017-0001, CS2514-2017-0002, CS2514-2018-0002, CS2514-2018-0003, ZL-2402-001).
• Source aliases: derived per subject from the trial identifier (CS2514-2017-0003 -> 1).
• Example models: Cammarata_2024_sulbactam_durlobactam.R (selects propSdPhase2 = sqrt(0.0794) = 0.282 for durlobactam plasma observations; no additive residual term applies to Phase 2, per Cammarata 2024 Table 1, which lists a single additive sigma^2 for Phase 1 only), Larson_2026_sulbactam_durlobactam_pediatric_allometry.R, Larson_2026_sulbactam_durlobactam_pediatric_allometry_crcl.R (both pediatric adaptations carry the adult phase-stratified residual-error model over unchanged; poster P-444 changed only the body-size terms and reports no residual-error re-estimation).
• Notes: Specific scope because the contrast is tied to the Cammarata 2024 sulbactam-durlobactam clinical-development pooled analysis. Drug-specific paper-anchored member of the STUDY_<DRUG>_PHASE<N> family alongside STUDY_POSA_PHASE3 (van Iersel 2018), STUDY_ASP8232_PHASE2 (Snelder 2020), STUDY_NIPOCALIMAB_PHASE1 (Valenzuela 2025), and STUDY_FARLETUZUMAB_PHASE2 (Farrell 2012). This is the first member of the family that needs a PAIR of indicators, because the paper stratifies residual error across three phases rather than two; keep both columns rather than an integer PHASE column so the reference stratum stays explicit. Subject-level (time-fixed); set once from the trial identifier on each subject record.

STUDY_B7981032 (canonical for the B7981032 phase 3 long-term ritlecitinib study cohort indicator)

• Description: 1 = the record comes from study B7981032 (NCT04006457), the phase 3 long-term safety and efficacy study of ritlecitinib in adolescents and adults with alopecia areata; 0 = the record comes from study B7931005 (phase 2a, NCT02974868) or B7981015 (phase 2b/3, NCT03732807). Time-fixed per subject within the pooled analysis. Captures the difference in baseline-severity inclusion criteria between the studies: B7931005 and B7981015 required >= 50% scalp hair loss, whereas B7981032 required only >= 25%, so B7981032 participants enter with systematically less severe disease.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (studies B7931005 and B7981015, baseline Severity of Alopecia Tool score >= 50%).
• Source aliases:
  • B7981032 – the protocol number used throughout Huh 2024 (Table 1 “Protocol” rows; Section 3.2 “Study B7981032 effect on BASE”).
• Example models: Huh_2024_ritlecitinib.R (multiplicative fractional shift on the non-AT/AU baseline transformed Severity of Alopecia Tool score: base_nonatau * (1 + e_study_b7981032_base * STUDY_B7981032) with e_study_b7981032_base = -0.645, moving the typical non-AT/AU baseline from 1.92 to 0.68 on the transformed scale; Huh 2024 Table 2. The shift is applied to the non-AT/AU group ONLY, because AT/AU participants have a baseline score of 100 by definition of their disease classification and cannot be moved by an entry criterion).
• Notes: A study-design property rather than a subject physiology covariate: it stands in for the entry-criterion difference, not for anything measured on the participant. Named for the protocol number rather than by phase because the pooled analysis spans three studies of which two share the reference stratum, so a STUDY_<DRUG>_PHASE<N> name in the style of STUDY_POSA_PHASE3 / STUDY_SULDUR_PHASE3 would not identify the contrast (both B7981015 and B7981032 carry a phase 3 component). Ritlecitinib protocol numbers are globally unique identifiers, so no drug prefix is added. If a future ritlecitinib analysis needs to separate B7931005 from B7981015, register a sibling STUDY_B7931005 rather than widening this entry.

STUDY_SULDUR_PHASE3 (canonical for Cammarata 2024 sulbactam-durlobactam phase 3 study cohort indicator)

• Description: 1 = subject enrolled in the Phase 3 study CS2514-2017-0004 (infections caused by Acinetobacter baumannii-calcoaceticus complex) of the Cammarata 2024 pooled sulbactam-durlobactam popPK analysis; 0 = otherwise. Paired with STUDY_SULDUR_PHASE2; both indicators 0 selects the six Phase 1 studies.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 with STUDY_SULDUR_PHASE2 also 0 (Phase 1 studies).
• Source aliases: derived per subject from the trial identifier (CS2514-2017-0004 -> 1).
• Example models: Cammarata_2024_sulbactam_durlobactam.R (selects propSdPhase3 = sqrt(0.203) = 0.451 for durlobactam plasma observations; the Phase 3 additive residual component ‘was determined to not be significant and was consequently removed’, per Cammarata 2024 Results), Larson_2026_sulbactam_durlobactam_pediatric_allometry.R, Larson_2026_sulbactam_durlobactam_pediatric_allometry_crcl.R (both pediatric adaptations carry the adult phase-stratified residual-error model over unchanged).
• Notes: Specific scope; see STUDY_SULDUR_PHASE2 for the full family rationale and the pair-of-indicators convention.

STUDY_CEFIDEROCOL_PHASE2 (canonical for Kawaguchi 2021 cefiderocol phase 2 APEKS-cUTI study cohort indicator)

• Description: 1 = subject enrolled in the phase 2 APEKS-cUTI study (NCT02321800; complicated urinary tract infection or acute uncomplicated pyelonephritis) of the Kawaguchi 2021 pooled cefiderocol popPK analysis; 0 = otherwise.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (for a cUTI patient, the phase 3 CREDIBLE-CR cUTI coefficient applies).
• Source aliases: derived from the subject ID in the source control stream (ID < 2000 = phase 1 or phase 2); acts only jointly with the infection-type code PT = 1.
• Example models: Kawaguchi_2021_cefiderocol.R (with DIS_CUTI = 1, selects the x1.27 cUTI/AUP factor on CL instead of the x0.872 CREDIBLE-CR cUTI factor).
• Notes: Drug-specific paper-anchored member of the STUDY_<DRUG>_PHASE<N> family alongside STUDY_SULDUR_PHASE2. The indicator is only consulted inside the cUTI term, because the phase 1 subjects all have no infection; a user simulating the paper’s cUTI scenarios should leave it at 0, as the paper’s own Monte-Carlo simulations used the CREDIBLE-CR coefficient to represent critically ill patients with carbapenem-resistant infections.

STUDY_PEGCET_PHASE3 (canonical for Crass 2024 pegcetacoplan phase 3 study cohort indicator)

• Description: 1 = subject enrolled in one of the two phase 3 pegcetacoplan studies of the Crass 2024 pooled population PK analysis (PEGASUS, NCT03500549, source STUD 302; or PRINCE, NCT04085601, source STUD 308); 0 = subject enrolled in any of the nine earlier-phase studies in the pool (CP0713-1, CP1014, 101, 102, 401, AIRIS, PHAROAH, PADDOCK, PALOMINO). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the phase 1 / 1b / 2a studies in the pool).
• Source aliases: derived per subject from the trial identifier (STUD 302 or 308 -> 1). The source control stream builds the equivalent flag inline as PNH3=0; IF(STUD.EQ.302) PNH3=1; IF(STUD.EQ.308) PNH3=1.
• Example models: Crass_2024_pegcetacoplan.R (used only in the residual-error switch, never in the structural model: together with DIS_PNH it selects one of three log-scale residual SDs – 0.200 for non-PNH participants, 0.163 for PNH patients in the phase 3 studies, and 0.326 for PNH patients in the phase 1/2 studies PHAROAH / PADDOCK / PALOMINO, per Crass 2024 ESM Table 3 thetas 5-7).
• Notes: The source control stream carries two flags, PNH2 (STUD 202 / 204 / 514, the phase 1/2 PNH studies) and PNH3 (STUD 302 / 308, the phase 3 PNH studies). Because PNH2 and PNH3 partition the PNH cohort exactly, only one canonical indicator is needed: the phase 1/2 stratum is recovered as DIS_PNH * (1 - STUDY_PEGCET_PHASE3). Drug-specific paper-anchored member of the STUDY_<DRUG>_PHASE<N> family alongside STUDY_POSA_PHASE3 (van Iersel 2018), STUDY_SULDUR_PHASE3 (Cammarata 2024), STUDY_ASP8232_PHASE2 (Snelder 2020), STUDY_NIPOCALIMAB_PHASE1 (Valenzuela 2025), and STUDY_FARLETUZUMAB_PHASE2 (Farrell 2012); reference category matches those siblings (the later phase is the 1-level). The residual-error-only role also mirrors STUDY_SULDUR_PHASE3 and STUDY_ASP8232_PHASE2.

STUDY_LEC201 (canonical for lecanemab phase 2 Study 201 cohort indicator)

• Description: 1 = subject enrolled in the lecanemab phase 2 Study 201 (BAN2401-G000-201; dose-finding response-adaptive randomisation across placebo, 2.5 mg/kg biweekly, 5 mg/kg monthly, 5 mg/kg biweekly, 10 mg/kg monthly and 10 mg/kg biweekly, with an 18-month Core phase, a 9-59 month off-treatment gap and an open-label extension); 0 = subject enrolled in the phase 3 Clarity AD trial (Study 301, NCT03887455). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (Clarity AD / Study 301).
• Source aliases:
  • Study – the covariate label used in Bhagunde 2026 Table 1 and Table S5; the source encodes the phase 2 cohort as the 1-level (“Study 201 on baseline GFAP”).
• Example models: Bhagunde_2026_lecanemab_gfap.R (log-linear effect exp(0.388 * STUDY_LEC201) on baseline plasma GFAP).
• Notes: In the founding model the effect is an assay batch effect rather than a biological one – Bhagunde 2026 Section 3.2.3 states “The significance of the study effect was likely due to the assay batch effect since GFAP samples from Study 201 and Clarity AD were analyzed at different times.” Named STUDY_LEC201 rather than a bare STUDY_201 because “Study 201” is a generic sponsor-internal protocol number that would collide across drug programmes; the LEC token pins it to the lecanemab programme, following the drug-token convention of STUDY_DORZA_EARLY, STUDY_SULDUR_PHASE2 and STUDY_PEGCET_PHASE3. The companion Bhagunde 2026 Abeta42/40 and p-tau181 models pool the same two studies but did not retain a study effect, so they do not carry this covariate.

STUDY_ASP8232_PHASE2 (canonical for Snelder 2020 ASP8232 phase 2 study cohort indicator in the pooled TMDD PK-PD analysis)

• Description: 1 = subject enrolled in one of the two phase 2 studies of the Snelder 2020 pooled ASP8232 TMDD PK-PD analysis (VIDI study, NCT02302079, diabetic macular edema; or ALBUM study 8232-CL-0004, NCT02358096, diabetic kidney disease); 0 = subject enrolled in one of the two phase 1 studies (8232-CL-0001 first-in-human healthy volunteers; 8232-CL-0002 renal impairment / T2DM-CKD, NCT02218099). Used to switch the log-additive residual-error magnitude on ASP8232 plasma concentrations and on VAP-1 plasma activity between the phase 1 studies (reference) and the phase 2 studies (paper’s estimated multiplicative factor 1.88 relative to phase 1).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (phase 1 studies 8232-CL-0001 and 8232-CL-0002).
• Source aliases: derived per subject from the trial identifier (8232-CL-3001/VIDI or 8232-CL-0004/ALBUM -> 1; 8232-CL-0001 or 8232-CL-0002 -> 0).
• Example models: Snelder_2020_ASP8232.R (multiplies the log-additive residual SD on ASP8232 total plasma concentration and on VAP-1 plasma activity by factor_phase2 = 1.88 when the indicator is 1, per Snelder 2020 Table 4 ‘Factor res error phase 2 studies’; VAP-1 concentration residual error is not affected because VAP-1 concentration was measured only in the two phase 2 studies).
• Notes: Specific scope because the contrast is tied to the Snelder 2020 pooled ASP8232 clinical-development analysis (four studies in Astellas’ vascular adhesion protein-1 inhibitor programme). Drug-specific paper-anchored member of the STUDY_<DRUG>_PHASE<N> family alongside STUDY_POSA_PHASE3 (van Iersel 2018), STUDY_NIPOCALIMAB_PHASE1 (Valenzuela 2025), and STUDY_FARLETUZUMAB_PHASE2 (Farrell 2012). Reference category matches Farrell 2012 (phase 2 as the 1-level and phase 1 as the reference). Subject-level (time-fixed); set once from the trial identifier on each subject record.

STUDY_FU2022_AZ (canonical for Fu 2022 CVS-CTR atenolol Study 2 (AstraZeneca) cohort indicator)

• Description: 1 = subject enrolled in Study 2 (AstraZeneca Alderley Park, UK; 4 male beagle dogs, 14.2-14.6 kg, 17-22 months old; oral atenolol 0/1/3/10 mg/kg; hemodynamic markers HR, dP/dtmax, and MAP measured but NO cardiac output) of the Fu 2022 CVS-CTR systems-model development pool; 0 = Study 1 (Servier, France; 4 male beagle dogs, 10-15 kg; oral atenolol 0/3/10/30 mg/kg; HR, dP/dtmax, CO, and MAP measured). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (Study 1, Servier site).
• Source aliases:
  • SSID in the NONMEM control stream (SSID = 1 -> Servier -> STUDY_FU2022_AZ = 0; SSID = 2 -> AstraZeneca -> STUDY_FU2022_AZ = 1; SSID = 3 -> GSK external validation, excluded from the estimation dataset via IGNORE=(SSID.EQ.3) and encoded as STUDY_FU2022_AZ = 0 alongside Study 1 for the packaged model).
• Example models: Fu_2022_atenolol_qsp.R (switches six per-study typical values: BSL_HR = 79.4 (S1) vs 77.0 (S2) bpm, V0 = 9.92 (S1) vs 9.15 (S2) mL, BSL_CTRM = 3777 (S1) vs 2422 (S2) mmHg/s, Amp = 0.0931 (S1) vs 0.168 (S2), Hor_HR = 7.86 (S1) vs 19.4 (S2) h, Hor_CTR = 9.82 (S1) vs 21.8 (S2) h; Fu 2022 Table 2 final-model column).
• Notes: Specific scope because the contrast is tied to the multi-site Servier/AstraZeneca beagle-dog telemetry pool used by Fu 2022 to develop the CVS-CTR systems model. Sibling of STUDY_C2201 (Bienczak 2025 ligelizumab), STUDY_ING111521 (Zhang 2015 dolutegravir), STUDY_LBSL (Zhou 2021 belimumab), STUDY_M281_004 (Vivacity-MG nipocalimab), STUDY_MD (Cirincione 2017 ER exenatide multi-dose), STUDY_PKU015 (Qi 2014 sapropterin pediatric), and STUDY_RIV201 (Tammara 2017 rivipansel); member of the STUDY_<name> family of paper-specific study cohort indicators. Subject-level (time-fixed); set once from the trial identifier on each subject record.

STUDY_OPTICLOT_CTR45 (canonical for Preijers 2021 perioperative FVIII Netherlands centre 4-5 residual-error stratum indicator)

• Description: 1 = FVIII sample from one of the two Dutch hemophilia treatment centres numbered 4 and 5 in the OPTI-CLOT perioperative dataset (Hazendonk 2016 / Preijers 2021); 0 = otherwise. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (Netherlands centres 1, 2, 3 when STUDY_GOSH is also 0).
• Source aliases:
  • Centres 4,5 – Preijers 2021 Table 2 residual-variability row label (Hazendonk 2016 Table 4 Center 4, 5). The paper does not say which Dutch centres carry these numbers.
• Example models: Preijers_2021_factor_viii.R (selects additive SD 0.06 IU/mL and proportional 21% instead of the centre 1-3 values 0.12 IU/mL and 19.7%; Preijers 2021 Table 2 final model).
• Notes: Member of the auto-approved STUDY_<id> family, used like STUDY_FU2022_AZ to mark a recruiting site rather than a trial. Acts on residual error only; it does not change the typical-value prediction. Mutually exclusive with STUDY_GOSH. The Hazendonk 2016 model (Hazendonk_2016_factor_viii.R) estimated the same centre split but ships only the centre 1-3 residual error, so it does not use this column.

STUDY_GOSH (canonical for Great Ormond Street Hospital (London) pediatric hemophilia cohort indicator)

• Description: 1 = FVIII sample from the Great Ormond Street Hospital (GOSH), London, retrospective pediatric perioperative cohort (87 children with severe hemophilia A undergoing central-venous-access-device surgery; centre 6 in Preijers 2021); 0 = otherwise. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (Netherlands centres 1, 2, 3 when STUDY_OPTICLOT_CTR45 is also 0).
• Source aliases:
  • Centre 6 – Preijers 2021 Table 2 residual-variability row label.
• Example models: Preijers_2021_factor_viii.R (selects additive SD 0.17 IU/mL and proportional 22%; Preijers 2021 Table 2 final model).
• Notes: Member of the auto-approved STUDY_<id> family, marking a recruiting site that is also a separate data source (the external-validation cohort that Preijers 2021 pooled with the Hazendonk 2016 data). Acts on residual error only. Mutually exclusive with STUDY_OPTICLOT_CTR45.

STUDY_118 (canonical for Nielsen 2015 vigabatrin pediatric Study 118 cohort indicator)

• Description: 1 = subject enrolled in pediatric Study 118 (n = 125; placebo or vigabatrin 20, 60 or 100 mg/kg/day) of the Nielsen 2015 pooled vigabatrin seizure-count analysis; 0 = otherwise. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the pooled adult studies 24 and 25).
• Source aliases:
  • ST118 – used in Nielsen_2015_vigabatrin.R (Nielsen 2015 Equations 10c and 10f preamble, “indicators equal to 1 if ith patient was from Studies 118, 192, or 221, respectively, or zero otherwise”).
• Example models: Nielsen_2015_vigabatrin.R (shifts log overdispersion by -0.671 and the asymptotic time effect by +0.120, and raises the log SD of the baseline-rate random effect by +0.586).
• Notes: Member of the STUDY_<id> auto-approve family of paper-specific study cohort indicators (siblings STUDY_PKU015, STUDY_C2201, STUDY_FU2022_AZ, STUDY1). Because the three pediatric studies enrolled only children and studies 24/25 only adults, STUDY_118 + STUDY_192 + STUDY_221 doubles as the pediatric-cohort flag; the model uses its complement to gate the adults-only Box-Cox transform on the baseline-rate random effect. Study 118 is the cohort whose asymptotic time effect is smaller than adults’ and whose AS shift moved when active-treatment data were added – the paper flags this as unexplained. See also the Study 118 60-kg dosing-weight cap documented under DOSE_VGB_MGD.

STUDY_192 (canonical for Nielsen 2015 vigabatrin pediatric Study 192 cohort indicator)

• Description: 1 = subject enrolled in pediatric Study 192 (n = 55; placebo or weight-banded vigabatrin 0.5-4 g/day) of the Nielsen 2015 pooled vigabatrin seizure-count analysis; 0 = otherwise. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the pooled adult studies 24 and 25).
• Source aliases:
  • ST192 – used in Nielsen_2015_vigabatrin.R (Nielsen 2015 Equation 10c).
• Example models: Nielsen_2015_vigabatrin.R (shifts log overdispersion by -0.940 and the asymptotic time effect by -0.584, and raises the log SD of the baseline-rate random effect by +0.886).
• Notes: Member of the STUDY_<id> auto-approve family; sibling of STUDY_118 and STUDY_221. Study 192’s placebo arm is the single group whose observed seizure rate fell outside the final model’s 90% prediction interval in the paper’s visual predictive check (over-predicted in both the baseline and post-baseline periods); the authors could identify no covariate or trial difference explaining it.

STUDY_221 (canonical for Nielsen 2015 vigabatrin pediatric Study 221 cohort indicator)

• Description: 1 = subject enrolled in pediatric Study 221 (n = 85; placebo or weight-banded vigabatrin 0.5-4 g/day) of the Nielsen 2015 pooled vigabatrin seizure-count analysis; 0 = otherwise. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the pooled adult studies 24 and 25).
• Source aliases:
  • ST221 – used in Nielsen_2015_vigabatrin.R (Nielsen 2015 Equation 10c).
• Example models: Nielsen_2015_vigabatrin.R (shifts log overdispersion by -1.420 and the asymptotic time effect by -0.374, and raises the log SD of the baseline-rate random effect by +0.358).
• Notes: Member of the STUDY_<id> auto-approve family; sibling of STUDY_118 and STUDY_192. Studies 192 and 221 are the two pediatric cohorts with a larger maximum time (placebo) effect than adults, which the paper reconciles with the Rheims meta-analysis finding that placebo responder rates are about 1.9-fold greater in children than adults.

STUDY_NCT00936403 (canonical for the Papathanasiou 2021 somatropin pooled analysis Trial 2 (NCT00936403) cohort indicator)

• Description: 1 = subject from Trial 2 (ClinicalTrials.gov NCT00936403; single-dose NNC126-0083 dose-escalation trial in prepubertal children with GHD, whose Norditropin comparator arm received 0.035 mg/kg/day for 7 days) of the Papathanasiou 2021 pooled Norditropin PK/PD analysis; 0 = otherwise. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (Trial 1 NCT01973244, prepubertal children; or Trial 3 NCT01706783, adults).
• Source aliases:
  • Trial – the paper’s trial identifier (Trial 1 / 2 / 3; Papathanasiou 2021 Table 1 and Table 3 rows ‘Proportional error Trial 1/2/3 (%)’).
• Example models: Papathanasiou_2021_somatropin.R (selects the Trial 2 IGF-I proportional residual error, 8.7%, over Trial 1’s 14.6% among children).
• Notes: Member of the STUDY_<id> auto-approve family, named by the ClinicalTrials.gov identifier because the paper gives the trials no protocol code. The per-trial IGF-I residual errors reflect the different IGF-I assays (Siemens IMMULITE in Trial 2; IDS-iSYS in Trials 1 and 3). Trial 3 is the only adult trial, so the model selects the Trial 3 residual error with CHILD = 0 and needs this single indicator to separate the two pediatric trials.

STUDY_016 (canonical for Loprete 2016 safinamide phase 3 Study 016 cohort indicator)

• Description: 1 = subject enrolled in safinamide phase 3 Study 016 (Borgohain 2014; Parkinson disease with motor fluctuations on stable levodopa, dosed with tablets containing safinamide free base at 50 or 100 mg/day). 0 = safinamide phase 3 Study 015 (Stocchi 2012; early Parkinson disease on a single dopamine agonist, dosed with gelatin capsules containing safinamide methanesulfonate titrated to 100 or 200 mg/day). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (Study 015, whose relative bioavailability is the structural anchor Frel = 1).
• Source aliases:
  • STUD – printed name in Loprete 2016 (Methods, Covariate model). Note the source column is coded 1 for Study 015 and 2 for Study 016, not 0/1, so the value transformation into this canonical is STUDY_016 = STUD - 1.
• Example models: Loprete_2016_safinamide.R (relative bioavailability of the depot, F = 1 * 0.724^STUDY_016, Loprete 2016 Table 3 row F (STUD016) = 0.724; founding example).
• Notes: Specific scope, tied to the Zambon safinamide phase 3 program. Member of the STUDY_<id> auto-approve family of paper-specific study cohort indicators; the bare zero-padded study number follows the STUDY_118 / STUDY_192 / STUDY_221 precedent. Two properties are load-bearing and easy to mis-encode. First, the indicator is confounded, not a population effect: Loprete 2016’s Discussion reports that the ~30% lower exposure in Study 016 matches the molecular-weight difference between safinamide free base and safinamide methanesulfonate, that the same offset appeared between Study 015 and its own follow-up Study 017, and that the two studies’ samples were assayed in different laboratories – so the authors read it as a systematic error and explicitly consider a population or trial-design explanation unlikely. Reading it as a disease-stage or ethnicity effect (Study 016 was 80% Asian and enrolled later-stage patients) is exactly the inference the source rules out. Second, the effect is carried on bioavailability only. Loprete 2016 first fitted the study effect separately on CL/F and Vd/F (model 112, interstudy ratios 1.3 and 1.6) and then replaced it with the single relative-bioavailability factor of model 110/110a, so a model that puts STUDY_016 on clearance instead will not reproduce the published parameters. Because the Table 3 estimates are conditioned on the Study 015 anchor, setting STUDY_016 = 1 is what recovers the Study 016 apparent values quoted in the abstract (CL/F 3.59/0.724 = 4.96 L/h, Vd/F 120/0.724 = 166 L). A future paper reporting a different “Study 016” must register a disambiguated sibling rather than overload this name.

STUDY_C2201 (canonical for Bienczak 2025 ligelizumab study C2201 cohort indicator)

• Description: 1 = subject enrolled in study C2201 (NCT02477332; Novartis Phase 2b ligelizumab dose-finding study in adult CSU patients) of the Bienczak 2025 pooled ligelizumab PopPK analysis; 0 = any other study in the pool (A2103, C2101, C2202, C2302, or C2303). Used to switch the typical CL/F magnitude in study C2201.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (any non-C2201 cohort in the Bienczak 2025 pool: pooled adult / adolescent CSU patients from C2202 / C2302 / C2303 and adult healthy volunteers from A2103 / C2101).
• Source aliases: derived per subject from the trial identifier (C2201 -> 1, else -> 0).
• Example models: Bienczak_2025_ligelizumab.R (Table S6: study C2201 on CL/F = 0.176, log-additive; cl *= exp(0.176) for C2201 subjects).
• Notes: Specific scope because the contrast is tied to the Novartis ligelizumab CSU development program. Subject-level / time-fixed; set once from the trial identifier on each subject record. The C2201 effect was retained in the final model because the residual unexplained CL/F differed between C2201 and the other studies after accounting for body weight, IgE, ADA, and disease-state covariates.

STUDY_J101 (canonical for DS3201-A-J101 valemetostat trial cohort indicator)

• Description: 1 = the record comes from DS3201-A-J101 (NCT02732275), the phase 1 multiple-ascending-dose study of valemetostat in patients with relapsed/refractory non-Hodgkin lymphoma; 0 = any of the other five studies pooled in the Inoue 2025 population PK analysis (J201, VALENTINE-PTCL01, DS3201-A-J107, DS3201-A-J109, DS3201-A-U106). Carried per record rather than per subject because it qualifies an assay, not a patient characteristic.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (any non-J101 study).
• Source aliases:
  • ASSAY – the Inoue 2025 Table 2 parameter label (“DS3201-A-J101 unbound adjustment factor”); the indicator itself is unnamed in the paper and is implied by the study of origin.
• Example models: Inoue_2025_valemetostat.R (does two separate things to UNBOUND valemetostat observations only: rescales the model prediction by the estimated assay factor 0.638, and selects a different proportional residual error, Sigma(3,3) = 0.327 instead of Sigma(2,2) = 0.404. Total valemetostat observations are unaffected by study).
• Notes: Member of the STUDY_<id> family. This is an analytical-method indicator rather than a design or population one: the J101 unbound assay reads systematically low relative to the method used in the later studies, and the popPK model absorbs the bias with a multiplicative factor plus its own residual-error term instead of excluding the data. A consuming simulation that does not distinguish assays should leave this at 0, which returns the model’s unbiased unbound prediction.

STUDY_C208 (canonical for TMC207-C208 bedaquiline trial cohort indicator)

• Description: 1 = subject enrolled in TMC207-C208 (NCT00449644; a randomized, double-blind, placebo-controlled Phase IIb trial of bedaquiline added to a five-drug background regimen in newly diagnosed MDR-TB patients, run in two stages with 8-week and 24-week treatment durations); 0 = subject enrolled in TMC207-C209 (NCT00910871; an open-label, single-arm Phase IIb trial of 24-week bedaquiline plus an individualized background regimen, enrolling both newly diagnosed and treatment-experienced MDR-TB and XDR-TB patients). Used in pooled analyses of the two bedaquiline Phase IIb trials to carry residual between-study differences in trial design.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (TMC207-C209).
• Source aliases: derived per subject from the trial identifier (C208 -> 1, C209 -> 0); the Lin 2024 NONMEM data column is literally named C208.
• Example models: Lin_2024_TB_multistate.R (log-hazard effect shared by all three dropout transitions, lambda14 / lambda24 / lambda34 *= exp(e_study_lambda1424 * STUDY_C208) with e_study_lambda1424 = 0.909188, a hazard ratio of 2.48 for C208 relative to C209).
• Notes: Specific scope because the contrast is tied to the Janssen bedaquiline Phase IIb development program. Subject-level / time-fixed; set once from the trial identifier on each subject record. Lin 2024’s Discussion attributes the higher C208 dropout hazard to trial design and calendar time rather than to any pharmacological difference: patients in a double-blind placebo-controlled trial “might be less willing to return for scheduled follow-ups with few signs of improvement”, and C209 ran after positive C208 results had been reported, so C209 patients “might therefore have higher confidence in drug efficacy”. Cohort split in Lin 2024: C208 n = 195 (49%), C209 n = 207 (51%). Note the studies also differ in what they enrolled – C208 excluded prior anti-TB treatment and enrolled MDR-TB only, while C209 admitted treatment-experienced patients and XDR-TB – so a model carrying this indicator alongside DIS_TB_XDR_STRICT is partially confounding study with resistance stratum by construction.

STUDY_1 (canonical for Duong 2017 Study-1 (newly diagnosed obese T2DM) cohort indicator)

• Description: 1 = subject enrolled in Study 1 (NCT00236600) of the Duong 2017 pooled placebo T2DM analysis (newly diagnosed, treatment-naive, obese T2DM subjects; 6-week placebo run-in followed by 60-week placebo treatment phase with an ancillary weight-loss diet and exercise counselling arm); 0 = subject enrolled in Study 2 (NCT01071850) or Study 3 (NCT01117584) of the same pool (advanced T2DM cohorts on stable diet and exercise, no weight-loss counselling).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (Studies 2 and 3 pooled – the “advanced T2DM” reference: baseline b-cell function logit b0 = 0.677 and no treatment-phase placebo effect on weight (EFPL)).
• Source aliases:
  • STUDY = 1 (character-valued or integer study identifier; STUDY_1 = as.integer(STUDY == 1)).
• Example models: Duong_2016_WHIG_T2DM.R (switches the baseline b-cell function logit b0 = -0.298 for Study 1 vs 0.677 for Studies 2 and 3, and activates the treatment-phase placebo weight effect EFPL only for Study 1 – the “additional placebo effect during the treatment phase for Studies 2 and 3 was not significant” per Duong 2017 Results).
• Notes: Follows the auto-approved STUDY_<id> canonical family. Distinct from the Cirincione 2017 STUDY1 (no underscore) which selects a different study in a different pooled analysis. Subject-level (time-fixed).

STUDY_CLO_BE2 (canonical for Pejcic 2024 clopidogrel bioequivalence Study-2 cohort indicator)

• Description: 1 = subject enrolled in Study 2 of the Pejcic 2024 pooled clopidogrel / clopidogrel-carboxylic-acid analysis (n = 26; sampling to 36 h post-dose, 17 samples per subject per period); 0 = Study 1 (n = 24; sampling to 48 h post-dose, 14 samples per subject per period). Both were 2-treatment, 2-period, 2-sequence crossover bioequivalence studies of a generic 75 mg film-coated clopidogrel tablet against Plavix, dosed as 150 mg (2 tablets) under fasting conditions.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (Study 1).
• Source aliases: derived per subject from the study identifier (Study 2 -> 1, Study 1 -> 0).
• Example models: Pejcic_2024_clopidogrel.R (switches every study-specific quantity in the joint semi-physiological model: mean transit time MTT 0.470 vs 0.410 h, generic relative bioavailability Fgen 1.08 vs 0.960, the first-pass fraction parameter FR1 119 vs 76.8, the IIV magnitudes on F and FR1, the IOV magnitudes on F and MTT, and the proportional residual error 41.95% vs 29.39%).
• Notes: Follows the auto-approved STUDY_<id> canonical family. Subject-level (time-fixed) – each subject participated in exactly one of the two studies. Unusually broad in scope for a study indicator: rather than switching a single parameter, it selects an entire study-specific parameter block, because Pejcic 2024 pooled two datasets collected under different study conditions and found that estimating the absorption / bioavailability / first-pass parameters separately per study lowered OFV and stabilised the model (Discussion paragraph 3). Distinct from the Cirincione 2017 STUDY1 and the Duong 2017 STUDY_1, which index studies in unrelated pooled analyses. Pair with FORM_CLO_GENERIC, which distinguishes the two products within each subject’s crossover, and with OCC, which indexes the two periods.

STUDY_ING111521 (canonical for Zhang 2015 dolutegravir proof-of-concept study cohort indicator)

• Description: 1 = subject enrolled in study ING111521 (phase 2a, proof-of-concept dose-ranging study of dolutegravir monotherapy in HIV-1-infected adults; n = 19 in the Zhang 2015 pooled analysis) of the Zhang 2015 dolutegravir PopPK; 0 = SPRING-1 (phase 2b, n = 141) or SPRING-2 (phase 3, n = 403). Used to switch the typical CL/F magnitude in study ING111521.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (SPRING-1 / SPRING-2 combined adult treatment-naive HIV-1 cohort).
• Source aliases:
  • POC (proof-of-concept) – used in Zhang_2015_dolutegravir.R (= 1 for ING111521, 0 otherwise; per Zhang 2015 Table 3 footnote CL/F = 0.901 * 1.16^SMOK * 1.35^POC * (WT/70)^0.438 * (AGE/40)^0.193 * (BILI/9)^-0.211).
• Example models: Zhang_2015_dolutegravir.R (exponential effect on CL/F: cl *= exp(e_ing111521_cl * STUDY_ING111521) with e_ing111521_cl = log(1.35) = 0.300, so CL/F is 35% higher in study ING111521 than in the SPRING-1 / SPRING-2 reference).
• Notes: Sibling of STUDY_C2201 (Bienczak 2025 ligelizumab), STUDY_LBSL (Zhou 2021 belimumab), STUDY_M281_004 (Vivacity-MG nipocalimab), STUDY_MD (Cirincione 2017 ER exenatide multi-dose), STUDY_PKU015 (Qi 2014 sapropterin pediatric), and STUDY_RIV201 (Tammara 2017 rivipansel); member of the STUDY_<name> family of paper-specific study cohort indicators. The 35% higher CL/F in ING111521 vs SPRING-1 / SPRING-2 is unexplained by available covariates; Zhang 2015 attributes it plausibly to the smaller sample size and less diverse patient population in ING111521 (Zhang 2015 Discussion p. 506). Subject-level (time-fixed); set once from the trial identifier on each subject record.

STUDY_HARROLD_PEG (canonical for Harrold 2020 pegfilgrastim-vs-filgrastim pivotal NHP study indicator)

• Description: 1 = subject enrolled in the pegfilgrastim pivotal NHP study (Harrold 2020 reference 13; pegfilgrastim 300 ug/kg SC on days 1 and 8); 0 = subject enrolled in the filgrastim pivotal NHP study (Harrold 2020 reference 14; filgrastim 10 ug/kg QD SC starting day 1). Used by the Harrold 2020 OS time-to-event sub-model to select between the two study-specific parameter sets in Table III (lambda_ANC, lambda_BC, k_e0); the ANC response sub-model (Table II) is fit on the combined placebo cohorts and does not depend on this indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (filgrastim pivotal NHP study).
• Source aliases:
  • STUDY – the paper-text label for the binary study identifier; derived per subject from the trial-of-origin column in the source dataset.
• Example models: Harrold_2020_radiation_neutropenia.R (selects between the two OS parameter sets in Table III: lambda_ANC -2.15 vs -0.229, lambda_BC -0.347 vs 0.300, k_e0 0.668 vs 0.156 1/day).
• Notes: Specific scope; the indicator is tied to the two Amgen NHP pivotal studies used by Harrold 2020 (filgrastim reference 14 and pegfilgrastim reference 13) and is required because the paper documents an unexplained between-study OS difference that no observed covariate could resolve. Subject-level (time-fixed). Follows the auto-approved STUDY_<id> canonical family.

STUDY_SPZCH (canonical for sporozoite-challenge (SpzCh) study cohort indicator)

• Description: 1 = participant enrolled in a controlled sporozoite challenge (SpzCh) malaria study, in which infection is initiated at the liver stage by inoculating Plasmodium falciparum sporozoites; 0 = participant enrolled in an induced blood stage malaria (IBSM) challenge study, in which infection is initiated directly at the blood stage by inoculating parasitised erythrocytes. Subject-level (time-fixed). In the founding model the indicator does double duty: it scales blood-stage drug potency, EC50,b,SpzCh = EC50,b,IBSM * (1 - theta * STUDY_SPZCH) with theta = 0.83 (Courlet 2023 Table 1 footnote c), and it selects which life-cycle stage is inoculated, so a single model file reproduces both trial designs (liver state seeded with Finc * inoculum when 1, blood state seeded with P0 when 0).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (induced blood stage malaria, IBSM, challenge).
• Source aliases:
  • SpzCh – Courlet 2023 Table 1 footnote c notation (“with SpzCh being equal to 0 or 1 for IBSM and SpzCh populations, respectively”).
• Example models: Courlet_2023_cabamiquine.R (blood-stage EC50 falls from 7.60 ng/mL in the IBSM population to 1.29 ng/mL in the SpzCh population, and the inoculation route switches from blood to liver).
• Notes: Follows the auto-approved STUDY_<id> canonical family, but names a challenge-model design rather than a single named trial, because the IBSM-vs-SpzCh distinction is a standard, reusable split in controlled human malaria infection (CHMI) research and any future CHMI extraction pooling the two designs will need exactly this indicator. The founding paper hypothesises the potency difference is mechanistic rather than a study artefact – cabamiquine damages parasites during the liver stage, so the merozoites subsequently released into blood are less viable and are killed at lower concentrations – which is a further reason to name the covariate after the challenge model rather than after Courlet 2023’s specific protocol numbers.

STUDY_SPR994_104 (canonical for tebipenem pivoxil hydrobromide study SPR994-104 (thorough-QT crossover) cohort indicator)

• Description: 1 = subject enrolled in study SPR994-104 (NCT04238195), the phase 1 randomized, double-blind, placebo- and active-controlled four-way crossover thorough-QT study in which 24 healthy adults received single oral doses of 600 mg and 1200 mg tebipenem pivoxil hydrobromide (TBP-PI-HBr) in crossover fashion; 0 = subject enrolled in any other study of the Ganesan 2023 pooled analysis (SPR994-101, SPR994-102, or the phase 3 SPR994-301 / ADAPT-PO trial). Subject-level (time-fixed).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (studies SPR994-101, SPR994-102, and SPR994-301 / ADAPT-PO).
• Source aliases:
  • S104 – the flag variable named in Ganesan 2023 Eq. 6 (“where S104 is a flag variable to indicate if subject was enrolled in study 104”).
• Example models: Ganesan_2023_tebipenem.R (gates the dose effect on the absorption rate constant: ka * (1 + e_dose_ka * (DOSE / 1200) * STUDY_SPR994_104) with e_dose_ka = -0.478, so Ka is 47.8% lower at 1200 mg and 23.9% lower at 600 mg relative to the same subject’s non-study-104 value, and the term is inert for every other study).
• Notes: Conceptually the same per-study switch family as STUDY1 / STUDY5 / STUDY_LBSL / STUDY_FARLETUZUMAB_PHASE2, but the switch here gates a covariate effect (dose on Ka) rather than a residual-error magnitude or a structural scale factor. Ganesan 2023 restricts the dose effect to this study because it was the only crossover design in the pooled analysis: the same 24 subjects received both dose levels, which limits the between-subject variability that otherwise masks the modest dose-on-absorption-rate signal (the effect was not detectable across the 100-900 mg single doses of study SPR994-101). Must be paired with a DOSE column carrying the administered milligram amount; the product (DOSE / 1200) * STUDY_SPR994_104 is zero for every subject outside study 104.

STUDY_LBSL (canonical for early-phase belimumab LBSL01 / LBSL02 study indicator)

• Description: 1 = subject enrolled in study LBSL01 (NCT00657007) or LBSL02 (NCT00071487) – the two early-phase belimumab studies that used a different ELISA-based bioanalytical assay; 0 = any other belimumab study in the Zhou 2021 pooled analysis. Used to switch CL and V1 magnitudes per study group (effectively an assay / early-development PK adjustment).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (later-phase studies using the electrochemiluminescence assay).
• Source aliases:
  • INDR – used in Zhou_2021_belimumab.R (Zhou 2021 Table 2 footnote: study indicator).
• Example models: Zhou_2021_belimumab.R (multiplicative factors 1.63 on CL and 1.26 on V1 when STUDY_LBSL = 1).
• Notes: Conceptually similar to STUDY1 / STUDY_FARLETUZUMAB_PHASE2 / ELISA / STUDY_NIPOCALIMAB_PHASE1 (per-study switches) but specific to the belimumab program. Subject-level (time-fixed); set from the trial identifier on each subject record.

STUDY_VORI (canonical for Friberg 2012 voriconazole pooled-analysis study indicator)

• Description: Integer-valued (1-5) subject-level identifier of which of the five pooled PK studies of the Friberg 2012 voriconazole integrated population PK analysis a subject belongs to. 1 / 2 / 3 = immunocompromised children (2 to <12 years; Friberg 2012 Table 1 studies 1-3); 4 = immunocompromised adolescents (12 to <17 years; study 4); 5 = healthy adults (22-55 years; study 5). Time-fixed per subject.
• Units: (integer 1-5)
• Type: categorical
• Scope: specific
• Reference category: 5 (healthy adult study; the typical-value reference for ka, Alag, Q, F1 IIV, CL IIV, and residual error).
• Source aliases:
  • STDY_VORI – prior canonical name (pre-2026-06-19 typo correction); a missing-vowel abbreviation of STUDY_VORI.
  • Derived per subject from the Friberg 2012 dataset’s STUDY / STDY identifier column.
• Example models: Friberg_2012_voriconazole.R (drives several effects: -0.382 Study-1 pediatric modifier on Km and Vmax,1; non-adult uplift on Q (+0.637); adolescent ka modifier; non-adult CL IIV scaling (+1.70); F1 IIV magnitude switching between adult and non-adult; per-study residual-error switching across the four levels Study 1, Study 2, Studies 3+4, Study 5).
• Notes: Departs from the binary STUDY1 / STUDY5 / STUDY_PKU015 precedent because the Friberg 2012 analysis uses five distinct studies and four of them carry distinct typical-value or residual-error coefficients (Studies 3 and 4 share one residual-error magnitude). Encoding as a single integer column avoids registering five paired binary indicators; the model file derives (STUDY_VORI == 1) style indicators inline.

STUDY_MODAF (canonical for Darwish 2012 armodafinil / modafinil pooled-analysis study indicator)

• Description: Integer-valued (1-8) subject-level identifier of which study of the Darwish 2012 armodafinil / modafinil pooled analysis a subject belongs to. Levels 1-6 are the pharmacokinetic studies of Darwish 2012 Table I: 1 = armodafinil single dose in healthy volunteers (n = 30); 2 = armodafinil and modafinil single dose in healthy volunteers (n = 89); 3 = modafinil single ascending dose in healthy volunteers (n = 24); 4 = modafinil single 200 mg dose in healthy volunteers (n = 36); 5 = modafinil multiple dose in healthy volunteers (n = 11); 6 = armodafinil and modafinil multiple dose in patients with excessive sleepiness associated with shift work disorder (SWD; n = 94). Levels 7 and 8 are the two pharmacodynamic (MSLT) efficacy trials, which contributed no plasma concentrations: 7 = the armodafinil 150 mg/d SWD trial (Czeisler 2009 Mayo Clin Proc 84:958-972); 8 = the modafinil 200 mg/d SWD trial (Czeisler 2005 N Engl J Med 353:476-486). Time-fixed per subject.
• Units: (integer 1-8)
• Type: categorical
• Scope: specific
• Reference category: 6 for the pharmacokinetic sub-model (the SWD-patient study, whose absorption rate constant is also the one used to predict concentrations for the level-7 and level-8 efficacy-trial patients, who had no PK sampling); 7 for the MSLT hazard sub-model (the armodafinil efficacy trial, against which the study effect delta of the modafinil trial is estimated).
• Source aliases:
  • Study no. – the study-numbering column of Darwish 2012 Table I (levels 1-6). Levels 7 and 8 follow the paper’s own text numbering of the two efficacy trials (“Study 7” and “Study 8” throughout the Methods and Results).
• Example models: Darwish_2012_armodafinil.R (levels 1, 2 and 6-8 – selects the study-specific absorption rate constant and the Study-1 apparent-volume offset of Darwish 2012 Table II), Darwish_2012_modafinil.R (levels 2-8 – selects the study-specific absorption rate constant and switches on the Study-8 baseline-hazard effect delta of Darwish 2012 Table III).
• Notes: Follows the single-integer STUDY_VORI precedent rather than the paired-binary STUDY1 / STUDY5 precedent, because the Darwish 2012 analysis spans eight studies and six of them carry a distinct typical-value coefficient; the model files derive (STUDY_MODAF == 1) style indicators inline. Each drug’s model file carries only the levels in which that drug was actually administered – armodafinil was dosed in studies 1, 2, 6 and 7, modafinil in studies 2-6 and 8 – so a level outside a file’s set simply selects no stratum-specific term. Levels 7 and 8 index the MSLT efficacy trials and are the levels to use when simulating the pharmacodynamic sub-model; setting STUDY_MODAF = 7 for both drugs reproduces the paper’s like-for-like armodafinil-versus-modafinil comparison (Darwish 2012 Figure 8) under a common baseline hazard, because the Study-8 effect delta then evaluates to zero.

STUDY_SALEM (canonical for Derippe 2024 mouse venetoclax Salem 2021 ABBV-167 prodrug study indicator)

• Description: 1 = mouse from Salem 2021 (Mol Cancer Ther 20(6):999-1008), which gave a single 5 mg/kg intravenous dose of the venetoclax prodrug ABBV-167 and measured the venetoclax that appeared from it; 0 = mouse from Eisenmann 2020 (J Chromatogr B 1152:122176), which gave a single 10 mg/kg oral dose of venetoclax itself. Subject-level (time-fixed). The indicator is load-bearing beyond a scale factor because the two structural parameters change physical meaning across its levels: in the Eisenmann cohort ka is a first-order oral absorption rate constant and vc is a volume of distribution, whereas in the Salem cohort ka is the prodrug-to-venetoclax biotransformation rate constant and vc additionally absorbs the unknown fraction biotransformed.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (Eisenmann 2020 oral venetoclax study).
• Source aliases:
  • Study – the column header of the Supplement’s “Mice PK modeling” parameter table, whose levels are printed as “Eisenman et al.” and “Salem”.
• Example models: Derippe_2024_venetoclax_mouse.R (carries both a ka effect, 1.65 vs 0.856 1/h, and a volume effect, 3.56 vs 6.54 L/kg).
• Notes: Follows the auto-approved STUDY_<id> canonical family. Must be paired with SEXF: the Derippe 2024 supplement treats sex as a three-level covariate on volume (female, male, and “unknown / second experiment”), where the third level is exactly the Salem cohort. The model file multiplies the male volume factor by (1 - STUDY_SALEM) so the sex term is inert for Salem rows regardless of what SEXF holds there; set SEXF = 0 for Salem records.

ORAL_VORI (canonical for Friberg 2012 voriconazole observation-during-oral-dose-phase indicator)

• Description: 1 = observation collected when the most recent administered voriconazole dose was oral (powder for oral suspension or tablet); 0 = observation collected when the most recent administered dose was IV. Per-observation (record-level) indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (IV-phase observation).
• Source aliases: derived per observation from the most recent administered dose’s route (POS or tablet -> 1; IV infusion -> 0).
• Example models: Friberg_2012_voriconazole.R (combines with STDY_VORI == 5 to switch the adult residual-error magnitude between IV-only expSdStdy5Iv = 0.0912 and oral sqrt(0.0912^2 + 0.132^2) = 0.160 per Table 3 footnote on the residual-error structure W).
• Notes: Conceptually similar to SAMPLE_INTENSIVE (a generic per-observation switch between estimated residual-error magnitudes); the contrast here is dosing route (oral vs IV) within the same subject’s crossover protocol rather than sampling design. Specific scope because the route-vs-residual-error switch is paper-specific to the Friberg 2012 voriconazole analysis.

SAMPLE_INTENSIVE (canonical for per-observation sampling-intensity indicator)

• Description: 1 = observation belongs to an intensive (rich, post-dose) PK sampling window; 0 = sparse (pre-dose / steady-state trough) sampling. Per-observation (record-level) indicator used to switch the proportional residual-error magnitude when a source paper estimates separate residual errors for intensive vs sparse sampling phases of the same dataset.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (sparse sampling).
• Source aliases: derived per observation from the sampling-design label in the source dataset (intensive 24-h profiles / dense post-dose schedules -> 1; pre-dose troughs and steady-state population samples -> 0).
• Example models: Macpherson_2015_rosuvastatin.R (Table 2 final model: 39.4% CV intensive vs 59.5% CV sparse residual error; switched per observation as propSd <- propSdIntensive * SAMPLE_INTENSIVE + propSdSparse * (1 - SAMPLE_INTENSIVE)).
• Notes: Conceptually similar to STUDY1 / STUDY5 / PHASE1 / ELISA (per-record switches that select between estimated residual-error magnitudes) but the contrast is sampling design rather than study cohort or bioanalytical assay. The indicator is generally applicable: any pediatric / dense-vs-sparse pooled popPK design that estimates two residual errors can carry it. Within-subject variation is permitted (a single subject can have both intensive and sparse observations, as in the Macpherson 2015 CHARON PK-pilot cohort where 12 subjects had a Day-0 intensive profile followed by 2 years of sparse troughs).

SAMPLE_PREDOSE (canonical for per-observation pre-dose-sample indicator)

• Description: 1 = the observation is a plasma sample drawn before the dose at that visit (a pre-dose / trough draw); 0 = a post-dose sample. Per-observation (record-level) indicator used to switch the residual-error magnitude when a source paper estimates a separate residual error for pre-dose samples, typically reflecting uncertainty in the recorded time of the preceding out-patient dose.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (post-dose sample).
• Source aliases: derived per observation from the sample’s nominal time relative to the dose of the same visit (sample before that dose -> 1, after -> 0).
• Example models: Tomita_2022_imeglimin.R (Tomita 2022 Table 3: log-scale residual SD 0.505 for pre-dose samples vs 0.359 for post-dose phase II / III samples and 0.190 for post-dose phase I samples; the pre-dose magnitude takes precedence over the phase I one, an assumption because the paper does not state the precedence).
• Notes: Member of the SAMPLE_* family (SAMPLE_INTENSIVE, SAMPLE_CAPILLARY). Distinct from SAMPLE_INTENSIVE, which partitions intensive vs sparse sampling DESIGN: a sparse phase IIb / III sample can be post-dose, and an intensive phase I profile includes day-N pre-dose draws, so the two partitions differ. Distinct from the meal-timing MEAL_PREDOSE_* indicators and from the CTROUGH concentration value.

SAMPLE_CAPILLARY (canonical for per-observation capillary-vs-venous blood sampling-site indicator)

• Description: 1 = the observation was drawn as a capillary blood sample (finger-prick / heel-prick); 0 = venous blood sample. Per-observation (record-level) indicator used to switch the proportional residual-error magnitude when a source paper estimates separate residual errors for capillary and venous sampling within the same dataset.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (venous sample).
• Source aliases: derived per observation from the sample-matrix / collection-site label in the source dataset (capillary, finger-prick, heel-prick, dried-blood-spot-style microsampling -> 1; venous plasma draws -> 0).
• Example models: Cleary_2023_risdiplam.R (Table 2 final PPK model: 23.4% CV venous vs 34.2% CV capillary proportional residual error, capillary being 3% of the 10,205-record dataset; switched per observation as propSd <- propSdCapillary * SAMPLE_CAPILLARY + propSdVenous * (1 - SAMPLE_CAPILLARY)).
• Notes: Third axis of the per-record residual-switch family, alongside SAMPLE_INTENSIVE (sampling design: rich vs sparse) and ELISA / RIA_ASSAY / IMMUNOASSAY / ASSAY_OSA (bioanalytical assay method). This entry covers the sample matrix / collection site, which none of those capture: in Cleary 2023 both matrices were quantified by the same validated LC-MS/MS method (LLOQ 0.25 ng/mL), so the extra residual variability is attributable to the capillary collection itself rather than to a different assay. Distinct from HCT, RBC, BLOOD_GROUP_O and the other blood-named entries, which are physiological quantities rather than sampling attributes. General scope because capillary microsampling is a standard paediatric-trial technique and any pooled paediatric popPK that estimates a separate capillary residual error can reuse the column. Within-subject variation is permitted (a subject may contribute both capillary and venous observations).

SAMPLE_WHOLEBLOOD, SAMPLE_DBS (canonical for the per-observation plasma / whole-blood / dried-blood-spot sampling-matrix stratum)

• Description: Two binary indicators that together encode a three-level per-observation sampling matrix: SAMPLE_WHOLEBLOOD = 1 for a venous whole-blood measurement, SAMPLE_DBS = 1 for a dried-blood-spot measurement, and both 0 for the venous plasma reference. At most one may be 1 on a given record. Unlike the other members of the per-record switch family, these indicators are not restricted to selecting a residual-error magnitude: they also carry the systematic multiplicative matrix conversion factor that a source paper estimates when one analyte partitions differently between plasma and blood cells, so the same column drives both the predicted concentration and its residual.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: both 0 (venous plasma). Plasma is the natural reference because it is the matrix conventionally reported for popPK parameters, and because a paper that estimates matrix conversion factors states its apparent clearance and volume on one reference scale.
• Source aliases: derived per observation from the sample-matrix label in the source dataset (venous plasma -> both 0; venous whole blood -> SAMPLE_WHOLEBLOOD = 1; dried blood spot / DBS card -> SAMPLE_DBS = 1).
• Example models: Meister_2019_tribendimidine.R (founding example; Meister 2019 Table 2 estimates a matrix conversion factor and a separate residual error for each of the three matrices and each of the two metabolites – dADT -14.5% whole blood and -13.7% DBS, adADT +5.00% whole blood and +7.00% DBS, with dADT log-scale residual SD 0.460 plasma / 0.564 whole blood / 0.623 DBS. Applied as matrix_cc <- 1 + e_wb_cc * SAMPLE_WHOLEBLOOD + e_dbs_cc * SAMPLE_DBS on the concentration and as a parallel three-way switch on the residual SD), Martial_2021_tacrolimus.R (SAMPLE_DBS alone, switching only the proportional residual error between venous whole blood 10.5% and dried blood spot 24.9%, with no conversion factor. Tacrolimus is a whole-blood analyte and the study has no plasma records, so SAMPLE_DBS = 0 means venous whole blood in this model and SAMPLE_WHOLEBLOOD is not read), Francis_2020_lumefantrine.R (SAMPLE_DBS only: capillary dried-blood-spot lumefantrine concentrations are 2.28-fold the venous plasma prediction, Cc *= (1 + (2.28 - 1) * SAMPLE_DBS); no separate residual), Kloosterboer_2020_pipamperone.R (SAMPLE_DBS only – plasma vs DBS, no venous whole blood; the DBS prediction is the linear conversion cal_slope_dbs * Cplasma + cal_int_dbs = 0.33 x + 3.90 ug/L of Kloosterboer 2020 Table 2, with the same residual error as plasma).
• Notes: Registered as a pair rather than a single integer stratum so that each level’s coefficient reads off the source table directly, following the precedent set in the ASSAY_CMIA Notes (“A pooled dataset that mixes CMIA, EMIT and LC-MS/MS needs more than one binary indicator”). Distinct from SAMPLE_CAPILLARY, which is the collection-site axis (capillary finger-prick vs venous draw) and cannot express the contrast these columns exist for: Meister 2019 found whole blood and DBS to be statistically indistinguishable from each other while both differed from plasma, i.e. the discriminating variable is the matrix (plasma vs blood) and not the puncture site, and SAMPLE_CAPILLARY would pool venous plasma with venous whole blood into a single reference level. A study whose only matrix contrast is capillary-vs-venous should keep using SAMPLE_CAPILLARY; a study that separates plasma from whole blood needs these. General scope because plasma / whole-blood / DBS pooling is a standard design in resource-limited-setting trials and in paediatric microsampling, and because nothing in the definition is drug-specific. Within-subject variation is permitted and expected – in the founding example the phase 2a participants contributed all three matrices while the phase 2b participants contributed DBS only. When a data set uses a single matrix throughout, set both columns to that matrix’s level for every row; the other matrices’ conversion factors and residual parameters then become non-identifiable and should be dropped or fixed.

ELISA (canonical for ELISA-vs-electrochemiluminescence bioanalytical assay indicator)

• Description: 1 = the concentration was measured by an enzyme-linked immunosorbent assay (ELISA); 0 = measured by an electrochemiluminescence assay (ECL / ECLIA). Per-observation indicator used to switch the residual-error magnitude (additive, proportional, or log-scale) between the two immunoassay platforms, which typically differ in LLOQ and precision. The two platforms and their LLOQs are paper-specific and MUST be documented in covariateData[[ELISA]]$notes.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (the electrochemiluminescence assay – in both ratifying papers the ECL platform is the newer, lower-LLOQ method used in the later-phase studies, which makes it the natural reference).
• Source aliases: derived per study from the bioanalytical method (Text S2 of Valenzuela 2025; Qi 2024 Sect. 2.3).
• Example models: Valenzuela_2025_nipocalimab.R (Valenzuela 2025 Table 3: additive PK residual 0.445 nmol/L for ELISA, LLOQ 0.150 ug/mL, studies MOM-M281-001 / MOM-M281-007 / MOM-M281-010, vs 0.0342 nmol/L for ECLIA, LLOQ 0.010 ug/mL, studies EDI1001 / EDI1002 / MOM-M281-004), Qi_2024_vosoritide.R (Qi 2024 Table 5: log-scale residual SD 0.665 for the validated ELISA used in studies 111-202 and 111-205, LLOQ 0.391 ug/L, vs 0.610 for the optimized ECL assay used in studies 111-206, 111-301 and 111-302, LLOQ 0.137 ug/L; encoded as expSd <- expSdElisa * ELISA + expSdEcl * (1 - ELISA)).
• Notes: Assay choice is study-fixed in both ratifying papers, but the column is carried per observation so pooled datasets apply the correct residual to each row. Distinct from IMMUNOASSAY (immunoassay of any kind vs an LC-MS/MS reference) and RIA_ASSAY (radioimmunoassay vs LC-MS/MS): those two contrast an immunoassay against chromatography, whereas ELISA contrasts two immunoassay platforms with each other and has no LC-MS/MS arm. Promoted from specific to general scope when the Qi 2024 vosoritide extraction ratified the same ELISA-vs-ECL encoding in an unrelated therapeutic area.

ASSAY_CMIA (canonical for chemiluminescent-microparticle-immunoassay vs EMIT bioanalytical method indicator)

• Description: 1 = the concentration was measured by a chemiluminescent microparticle immunoassay (CMIA); 0 = measured by an enzyme-multiplied immunoassay technique (EMIT). Per-observation indicator used to switch the residual-error magnitudes between the two immunoassay platforms, which differ in precision, linear range and limit of quantification. Both arms are immunoassays: this canonical contrasts two immunoassay platforms with each other and has no chromatographic reference arm. The specific instruments and their LOQs are paper-specific and MUST be documented in covariateData[[ASSAY_CMIA]]$notes.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (EMIT). EMIT is the older enzyme-based platform and the minority method in the founding example; CMIA is the newer chemiluminescent platform and is name-positive so that the column reads true for the assay it is named after. The choice is presentational only – both magnitudes are estimated and the indicator enters as a linear switch, so predictions are unchanged if a downstream user prefers the opposite polarity, provided the coefficients are swapped with it.
• Source aliases:
  • Concentration_Method – the NONMEM $INPUT column in Ding_2026_vancomycin.R, read into the $ERROR block as TYPE and branched with IF (TYPE.EQ.0) THEN; CMIA i1000 ... ELSE; EMIT Viva-E. The source coding is inverted relative to this canonical (0 = CMIA, 1 = EMIT): derive ASSAY_CMIA = 1 - Concentration_Method.
• Example models: Ding_2026_vancomycin.R (founding example; switches both residual magnitudes at once via propSd <- propSdCmia * ASSAY_CMIA + propSdEmit * (1 - ASSAY_CMIA) and the matching addSd line. CMIA carries a combined error, proportional 0.198 plus additive 0.898 mg/L, on an Abbott Architect i1000 with LOQ 3.0 mg/L and a 3.0-100 mg/L linear range; EMIT carries additive error only, 1.83 mg/L, on a Viva-E with LOQ 2.0 mg/L and a 2.0-50 mg/L linear range, its proportional term having been fixed to 0 because the lower bound of its 95% CI was close to zero. CMIA supplied 1062 of 1120 concentrations, 94.8%).
• Notes: Distinct from IMMUNOASSAY and RIA_ASSAY, which contrast an immunoassay of some kind against an LC-MS/MS reference – neither can express a dataset in which both arms are immunoassays and no chromatographic method was used, which is the situation this canonical exists for. Structurally the sibling of ELISA (ELISA vs electrochemiluminescence), the other two-immunoassay-platform contrast in the register; the two are kept separate rather than unified into a generic “assay platform A vs B” column because the platform pair determines which residual magnitudes are comparable across papers. A pooled dataset that mixes CMIA, EMIT and LC-MS/MS needs more than one binary indicator, or an integer stratum column on the STUDY_<id> pattern of STUDY_TACRO_FRANCKE. When a dataset uses a single assay throughout, set the column to that assay’s level for every row; the other platform’s residual parameters then become non-identifiable and should be dropped or fixed.

SIDN (canonical for the opaque study-level nesting column)

• Description: Integer-valued grouping column identifying the clinical trial (or trial group) a subject’s records belong to, used as the nesting level for a second hierarchical level of random effects on top of the per-subject etas. Values 1, 2,…, N. Unlike the STUDY_<id> family, SIDN does not enter any typical-value covariate equation and has no reference category – it only indexes which study-level random-effect draw applies to a record. Use it when the source paper fits a between-study random effect (rather than a fixed per-study shift), whether or not the paper publishes a mapping from trial identifier to level.
• Units: (count)
• Type: categorical
• Scope: general
• Reference category: n/a – SIDN is a nesting level, not a covariate with a reference category.
• Source aliases:
  • SIDN (“secondary study identity number”) – used in Qi_2024_vosoritide.R (Qi 2024 Sect. 3.2).
  • STUD – Rosenborg 2025 supplement Fig. 1, $LEVEL STUD=(9[1]); interstudy variability. Here the mapping is published (levels 1, 2 and 3 are the paper’s studies 1, 2 and 3), so the column is opaque only in Qi 2024’s case; the canonical name is still SIDN because the role – indexing which study-level random-effect draw applies – is identical, and the level index never enters a typical-value equation.
• Example models: Rosenborg_2025_salmeterol.R (Rosenborg 2025 Table 2 row F4_rel_OMEGA_ISV = 0.0293, the between-study SD of relative bioavailability, encoded as etalfdepot_study ~ 0.00085849 | SIDN on top of the subject-level etalfdepot), Qi_2024_vosoritide.R (Qi 2024 Sect. 2.4 and 3.2: the additional hierarchical level eta6 on CL/F and eta7 on V/F, encoded with rxode2/nlmixr2 native nested random effects as etalcl_study ~ 0.066049 | SIDN and etalvc_study ~ 0.000144 | SIDN; Qi 2024 states only three SIDN values were present in the database and does not enumerate which trials map to which value, so the column is deliberately opaque).
• Notes: Distinct from the STUDY_<id> family (binary per-trial indicators that switch a typical value, a residual-error magnitude, or a covariate coefficient) and from OCC (integer occasion / period index for inter-occasion variability within a subject). SIDN groups across subjects: a subject belongs to one study level and all of that subject’s records share the study-level draw. Because it is consumed by the ini() nesting syntax rather than by a model() expression, covariateData[['SIDN']] will not be referenced anywhere in model(). Simulation caveat: rxode2 expands nested random effects into per-level terms, so an event table must contain at least two distinct SIDN values or rxSolve() fails with The following parameter(s) are required for solving: THETA[1], and omega must be passed explicitly (omega = mod$omega) because the nested omega is a list of matrices keyed by level. Ratified canonically alongside the Qi 2024 vosoritide extraction, the first nlmixr2lib model to use nested random effects.

STUDY_M281_004 (canonical for MOM-M281-004 (Vivacity-MG) study indicator)

• Description: 1 = subject enrolled in study MOM-M281-004 (Vivacity-MG; NCT03772587; phase 2 in generalized myasthenia gravis); 0 = any other study in the Valenzuela 2025 pooled analysis. Used to switch the IgG-baseline scaling factor.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-Vivacity-MG studies).
• Source aliases: derived per subject from the trial identifier (NCT03772587 -> 1, else -> 0).
• Example models: Valenzuela_2025_nipocalimab.R.
• Notes: Valenzuela 2025 estimated a slightly lower IgG baseline in MOM-M281-004 participants (baseline scaled by FRIgG0_M281_004 = 0.777 vs. 1 in other studies). Distinct from the disease-state indicator implied by gMG – it is specifically the Vivacity-MG study flag because the IgG baseline factor was only estimated for that study.

STUDY_ABA2_HLA78 (canonical for ABA2 trial HLA 7/8-matched cohort indicator)

• Description: 1 = subject enrolled in the ABA2 hematopoietic-cell-transplant trial (IM101311; NCT01743131) HLA 7/8 (one-allele-mismatched donor) cohort, 0 = any other study in the Takahashi 2023 pooled abatacept population PK analysis (RA/JIA reference and ABA2 8/8 cohort).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-ABA2-7/8 – pooled adult RA / pediatric JIA cohort, plus ABA2 HLA 8/8 cohort which is flagged separately by STUDY_ABA2_HLA88).
• Source aliases: derived per subject from the trial-cohort identifier (Cohort = ABA2 7/8 -> 1, else -> 0). Takahashi 2023 Supplemental Table 4 names the corresponding theta thetaCohort_CL / thetaCohort_Vc.
• Example models: Takahashi_2023_abatacept.R (multiplicative Ratio factors on CL = 0.70 and on Vc = 0.99 vs the RA/JIA reference; values from Takahashi 2023 Supplemental Table 4).
• Notes: Pairs with STUDY_ABA2_HLA88 to reproduce the three-level cohort categorical (RA/JIA, ABA2 7/8, ABA2 8/8) the paper reports as the only retained categorical PK covariate. At most one of STUDY_ABA2_HLA78 and STUDY_ABA2_HLA88 is 1 per subject; both 0 reproduces the RA/JIA reference. Scope: specific because the contrast is tied to the ABA2-vs-RA/JIA pooling design.

STUDY_ABA2_HLA88 (canonical for ABA2 trial HLA 8/8-matched cohort indicator)

• Description: 1 = subject enrolled in the ABA2 hematopoietic-cell-transplant trial (IM101311; NCT01743131) HLA 8/8 (allele-matched donor) cohort, 0 = any other study in the Takahashi 2023 pooled abatacept population PK analysis (RA/JIA reference and ABA2 7/8 cohort).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-ABA2-8/8 – pooled adult RA / pediatric JIA cohort, plus ABA2 HLA 7/8 cohort which is flagged separately by STUDY_ABA2_HLA78).
• Source aliases: derived per subject from the trial-cohort identifier (Cohort = ABA2 8/8 -> 1, else -> 0). Takahashi 2023 Supplemental Table 4 names the corresponding theta thetaCohort_CL / thetaCohort_Vc.
• Example models: Takahashi_2023_abatacept.R (multiplicative Ratio factors on CL = 0.91 and on Vc = 1.32 vs the RA/JIA reference; values from Takahashi 2023 Supplemental Table 4).
• Notes: Pairs with STUDY_ABA2_HLA78. At most one of the two indicators is 1 per subject; both 0 reproduces the RA/JIA reference. Scope: specific.

STUDY_A0081074 (canonical for the pregabalin phase I pediatric PK study A0081074 cohort indicator)

• Description: 1 = record from Pfizer study A0081074, the phase I pharmacokinetic study of pregabalin in pediatric patients aged 1 month to 16 years with focal (partial) onset seizures; 0 = any other study in the Chan 2021 pooled pediatric + adult population PK analysis.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the adult studies and the phase III pediatric study A0081041).
• Source aliases:
  • PROT == 1074 – Chan 2021 final NONMEM control stream (run8.mod, second supplementary file), IF (PROT.EQ.1074) in $ERROR.
• Example models: Chan_2021_pregabalin.R (selects the study-specific residual error only: proportional SD 29.8% with the additive SD held at 0; it overrides the phase I / phase III adult strata and touches no structural or covariate parameter).
• Notes: Member of the STUDY_<id> family, named for the sponsor protocol number the control stream branches on. Pairs with STUDY_A0081041; at most one of the two is 1 per record.

STUDY_A0081041 (canonical for the pregabalin phase III pediatric PERIWINKLE study A0081041 cohort indicator)

• Description: 1 = record from Pfizer study A0081041 (PERIWINKLE, NCT01389596), the phase III randomised placebo-controlled trial of adjunctive pregabalin in pediatric patients aged 4-16 years with focal onset seizures; 0 = any other study in the Chan 2021 pooled pediatric + adult population PK analysis.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the adult studies and the phase I pediatric study A0081074).
• Source aliases:
  • PROT == 1041 – Chan 2021 final NONMEM control stream (run8.mod, second supplementary file), IF (PROT.EQ.1041) in $ERROR.
• Example models: Chan_2021_pregabalin.R (selects the study-specific residual error only: proportional SD 35.0% and additive SD 0.68 ug/mL; it overrides the phase I / phase III adult strata and touches no structural or covariate parameter).
• Notes: Member of the STUDY_<id> family. Pairs with STUDY_A0081074. The PERIWINKLE patients are also the pediatric population of the companion exposure-response model Chan_2021_pregabalin_lsr28.R, which carries that contrast with the general CHILD indicator rather than this study indicator.

STUDY_d2eGFP (canonical for Frohlich 2018 multi-experiment NLME reporter-construct cohort indicator)

• Description: 1 = cell transfected with destabilized eGFP (d2eGFP, mRNA construct pVAXA120-d2EGFP, ~6.6 h protein half-life via C-terminal PEST sequence); 0 = cell transfected with eGFP (mRNA construct pVAXA120-eGFP, ~22.8 h protein half-life). Cohort indicator in the Frohlich 2018 multi-experiment NLME single-cell translation-kinetics analysis used to select between the two reporter-construct cohorts (Experiment 1 eGFP vs Experiment 2 d2eGFP). All structural and ribosomal-binding parameters (mRNA degradation, ribosome-mRNA binding rate x m0, ribosome catalytic rate, R0/m0, fluorescence offset, transfection-onset time, scale x m0) are shared between cohorts; only the protein degradation rate (gamma_eGFP for STUDY_d2eGFP = 0 vs gamma_d2eGFP for STUDY_d2eGFP = 1) and its IIV variance differ.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (eGFP cohort – the stable reporter; this is the cohort with the slower protein degradation rate ~0.030 1/h and the larger between-cell IIV variance on it).
• Source aliases: derived per cell from the experiment identifier in the source dataset (Experiment 1 / eGFP -> 0, Experiment 2 / d2eGFP -> 1; see code.zip project/models/experiments_transfection_ribo.m in the Frohlich 2018 Zenodo deposit doi:10.5281/zenodo.1228899).
• Example models: Frohlich_2018_mRNA_translation.R.
• Notes: Specific scope because the cohort is tied to the Frohlich 2018 in-vitro reporter-mRNA assay (eGFP vs d2eGFP). Cell-level (time-fixed) indicator; set once per cell from the transfection-construct identifier. Conceptually analogous to a single-study two-arm cohort indicator (e.g. STUDY_FARLETUZUMAB_PHASE2, STUDY_NIPOCALIMAB_PHASE1), but the contrast is between two reporter mRNA constructs measured under the same physical / imaging protocol rather than between trial phases.

STUDY_RIV201 (canonical for Tammara 2017 rivipansel phase II SCD study indicator)

• Description: 1 = subject enrolled in the phase II rivipansel study (NCT01119833; “study 201” in the Tammara 2017 integrated population PK analysis; Telen 2015 Blood 125:2656-2664 reports the trial results); 0 = healthy adult volunteers from the three rivipansel phase I studies (studies 101, 102, and 103) pooled into the integrated dataset. Used both as an additive shift on typical clearance (interpreted by the authors as the SCD-hyperfiltration component, ~23%) and to switch the additive / proportional residual-error magnitudes between the two cohorts (Table 1).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (rivipansel phase I healthy-adult studies 101, 102, and study 103 healthy adults with SCD pooled with phase I per Tammara 2017; the paper labels these as “phase I studies” in the residual-error rows of Table 1).
• Source aliases: derived per subject from the trial identifier (NCT01119833 -> 1, else -> 0); the source STUD column in the NONMEM dataset described in Tammara 2017 Table 1 footnote b.
• Example models: Tammara_2017_rivipansel.R (Table 1: additive effect 0.234 on CL via 1 + 0.234 * STUDY_RIV201; selects the cohort-specific additive and proportional residual SDs in model()), Nayak_2021_rivipansel.R (Table 3: fractional increase 0.116 on renal CL via 1 + 0.116 * STUDY_RIV201, the control stream’s IF(STUDY.EQ.1012) SCLF = 1 + THETA(7); B5201012 is the same NCT01119833 trial. The reference level is broader here: 0 covers all seven other fitted studies, including renal- and hepatic-impairment, elderly and stable-SCD cohorts. Also selects the SCD-VOC vs other-study residual-error magnitudes).
• Notes: Specific scope because the contrast is tied to the rivipansel development program. The Tammara 2017 paper interprets the 23% CL increment as a putative hyperfiltration effect of SCD; in simulation use cases targeting the SCD population (the paper’s stated goal) set STUDY_RIV201 = 1 for every subject. Subject-level / time-fixed; set once from the trial identifier on each subject record.

STUDY_LPS30M (canonical for Kutumova 2024 ANP-30-min-after-LPS cohort indicator)

• Description: 1 = albumin nanoparticles administered intravenously 30 min after the intraperitoneal lipopolysaccharide challenge (the paper’s “LPS 30 min” arm, exp 2); 0 otherwise. Cohort indicator in the Kutumova 2024 four-experiment nanoparticle-biodistribution PBPK analysis, used together with STUDY_LPS6H and STUDY_LPS24H to select the arm-specific permeability (PAC) and distribution (P) coefficients of every organ from Table 1. All other fitted parameters (the Hill-function endocytic uptake constants KRESmax / KRES50 / KRESn, the exocytic release constant KRESrelease, the biliary and urinary excretion coefficients, and the radiant-efficiency scale factor k) are shared across all four arms.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (LPS-naive control cohort, exp 1 – mice given nanoparticles only, no LPS).
• Source aliases: derived per animal from the “LPS 30 min” column heading of Kutumova 2024 Table 1 / the ANP-administration time point of 0.5 h after LPS in section 2.4.
• Example models: Kutumova_2024_albuminNanoparticles_mouse_pbpk.R.
• Notes: Specific scope because the cohort is tied to the Kutumova 2024 murine LPS acute-lung-injury protocol. Animal-level (time-fixed) indicator. The three STUDY_LPS* indicators are mutually exclusive; all three zero selects the control arm via f_ctrl <- 1 - STUDY_LPS30M - STUDY_LPS6H - STUDY_LPS24H in model. This encodes an experimental-arm contrast rather than a separate trial, in the same sense as STUDY_d2eGFP (reporter-construct cohorts measured under one imaging protocol).

STUDY_LPS6H (canonical for Kutumova 2024 ANP-6-h-after-LPS cohort indicator)

• Description: 1 = albumin nanoparticles administered intravenously 6 h after the intraperitoneal lipopolysaccharide challenge (the paper’s “LPS 6 h” arm, exp 3); 0 otherwise. Cohort indicator in the Kutumova 2024 four-experiment nanoparticle-biodistribution PBPK analysis, selecting the LPS-6-h column of Table 1 for every organ’s permeability (PAC) and distribution (P) coefficient. This is the arm with peak pulmonary nanoparticle accumulation (lung P = 0.89661, the maximum across all four arms).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (LPS-naive control cohort, exp 1 – mice given nanoparticles only, no LPS).
• Source aliases: derived per animal from the “LPS 6 h” column heading of Kutumova 2024 Table 1 / the ANP-administration time point of 6 h after LPS in section 2.4.
• Example models: Kutumova_2024_albuminNanoparticles_mouse_pbpk.R.
• Notes: Specific scope because the cohort is tied to the Kutumova 2024 murine LPS acute-lung-injury protocol. Animal-level (time-fixed) indicator; mutually exclusive with STUDY_LPS30M and STUDY_LPS24H.

STUDY_LPS24H (canonical for Kutumova 2024 ANP-24-h-after-LPS cohort indicator)

• Description: 1 = albumin nanoparticles administered intravenously 24 h after the intraperitoneal lipopolysaccharide challenge (the paper’s “LPS 24 h” arm, exp 4); 0 otherwise. Cohort indicator in the Kutumova 2024 four-experiment nanoparticle-biodistribution PBPK analysis, selecting the LPS-24-h column of Table 1 for every organ’s permeability (PAC) and distribution (P) coefficient. The paper interprets this arm as partial pulmonary recovery with persisting injury elsewhere (lung P falls from its 6-h peak while splenic PAC keeps rising).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (LPS-naive control cohort, exp 1 – mice given nanoparticles only, no LPS).
• Source aliases: derived per animal from the “LPS 24 h” column heading of Kutumova 2024 Table 1 / the ANP-administration time point of 24 h after LPS in section 2.4.
• Example models: Kutumova_2024_albuminNanoparticles_mouse_pbpk.R.
• Notes: Specific scope because the cohort is tied to the Kutumova 2024 murine LPS acute-lung-injury protocol. Animal-level (time-fixed) indicator; mutually exclusive with STUDY_LPS30M and STUDY_LPS6H. The optimisation constrained PAC to be non-decreasing from control through this arm for the lungs, liver, spleen, and kidneys (section 2.8 equation 9), so a monotone PAC trend across the four indicators is a fitted constraint rather than an emergent result.

STUDY_WAGH_2 (canonical for Wagh 2021 spectinamide 1810 mouse TB study 2 indicator)

• Description: 1 = BALB/c mouse cohort enrolled in study 2 of the Wagh 2021 spectinamide 1810 dose ranging / dose fractionation efficacy program (Wagh 2021 Table 5: study 2 used 4-arm dosing regimens of 50 BID, 100 QD, 166 TIW, 200 BID, and 333 TIW with a 5.37 log CFU baseline at start of treatment); 0 = study 1 (Wagh 2021 Table 4: 24-arm dose-fractionation schedule with a 7.08 log CFU baseline at start of treatment). Used to switch K_kill_max from the study 1 typical value to the 1.15-fold higher study 2 value in the integrated PK/PD model.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (Wagh 2021 study 1: 7.08 log CFU initial bacterial load; the reference K_kill_max = 0.0374 1/h applies).
• Source aliases:
  • STUDY – the Wagh 2021 NONMEM dataset’s study identifier column (1 / 2 in the source; mapped to 0 / 1 in this binary indicator).
• Example models: Wagh_2021_spectinamide_1810_mouse.R (Table 3: study-specific coefficient for K_kill_max = 1.15, RSE 4.0%; applied multiplicatively as killmax = killmax_base * (1 + STUDY_WAGH_2 * (1.15 - 1))).
• Notes: Specific scope because the contrast is tied to the Wagh 2021 study 1 vs study 2 difference in inoculum-induced baseline bacterial load and a downstream effect on observed efficacy. Subject-level / time-fixed; set once per mouse from the source-study identifier on each animal record. Per the paper Discussion, the higher K_kill_max in study 2 (1.15x) is plausibly explained by the lower baseline bacterial load (5.37 vs 7.08 log CFU); see vignette Assumptions and deviations for the per-study residual-error treatment.

STUDY_CT_USA (canonical for the Wang 2024 risperidone LAI US relative-bioavailability trial-group indicator)

• Description: 1 = subject enrolled in one of the two US relative-bioavailability trials of the Wang 2024 RYKINDO (LY03004) / RISPERDAL CONSTA programme – CT-USA-104 (NCT02186769; open-label parallel single-dose 25 and 50 mg Rykindo vs Consta) or CT-USA-102 (NCT02091388; open-label parallel 25 mg every 2 weeks for five injections); 0 = subject enrolled in CT-1S01 (NCT02055287), the open-label single-ascending-dose Rykindo trial (12.5 / 25 / 37.5 / 50 mg). Used to switch the apparent clearance of the risperidone active moiety between the two trial groups.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (CT-1S01, the single-ascending-dose trial, which anchors the reference apparent clearance of 186.7 L/day in males and 153.4 L/day in females).
• Source aliases: derived per subject from the trial identifier (CT-USA-104 or CT-USA-102 -> 1; CT-1S01 -> 0). Wang 2024 Table 2 names the effect in the row label “CL ratio for CT-USA-104 and 102 to CT-1S01”.
• Example models: Wang_2024_risperidone_rykindo.R (log-multiplicative effect on apparent clearance: exp(log(0.675) * STUDY_CT_USA), i.e. a 0.675-fold, 32.5% lower, apparent clearance in the two US trials; Wang 2024 Table 2, %RSE 6.03).
• Notes: Specific scope; tied to the three-trial Luye Pharma risperidone LAI programme. Subject-level (time-fixed); set once from the trial identifier. A group-level indicator spanning two trials, directly analogous to STUDY_DORZA_EARLY and STUDY_LBSL. Applies to the Rykindo model only – the companion Wang_2024_risperidone_consta.R has a single apparent clearance across trials and sexes. Beware a text-versus-table conflict in the source: Wang 2024 Results (“Population Pharmacokinetic Modeling of Rykindo”) describes the reduced clearance as applying to “the multiple-dose study”, i.e. CT-USA-102 alone, whereas the Table 2 row label names both CT-USA-104 and CT-USA-102. The table definition is the one encoded, because it is the final-model parameter definition and because it reproduces the observed CT-USA-104 single-dose exposures (AUC0-t 227 and 399 day*ng/mL at 25 and 50 mg) far better than the text reading does. Distinct from STUDY_CT_USA_102, which is nested inside this group and switches the absorption rate constant rather than clearance; a CT-USA-102 subject has both indicators set to 1. Follows the auto-approved STUDY_<id> canonical family.

STUDY_CT_USA_102 (canonical for the Wang 2024 risperidone LAI multiple-dose trial indicator)

• Description: 1 = subject enrolled in trial CT-USA-102 (NCT02091388), the multiple-dose relative-bioavailability trial of the Wang 2024 RYKINDO (LY03004) / RISPERDAL CONSTA programme, in which 25 mg of either formulation was given intramuscularly every 2 weeks for five injections; 0 = subject enrolled in one of the single-dose trials CT-1S01 (NCT02055287) or CT-USA-104 (NCT02186769). Used to switch the first-order absorption rate constant of the main (third) release phase.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the single-dose trials; for Rykindo this is CT-1S01 and CT-USA-104 pooled, for Consta it is CT-USA-104 alone since Consta was not studied in CT-1S01).
• Source aliases: derived per subject from the trial identifier (CT-USA-102 -> 1, else -> 0). Wang 2024 Table 2 names the two levels in the row labels “KA_CT-1S01 and CT-USA-104” and “KA_CT-USA-102”, with the footnote “KA_CT-USA-104 for Consta”.
• Example models: Wang_2024_risperidone_rykindo.R (log-multiplicative effect on the main-release KA: 0.288 -> 0.380 /day, a 1.319-fold increase), Wang_2024_risperidone_consta.R (0.179 -> 0.271 /day, a 1.514-fold increase).
• Notes: Specific scope; tied to the three-trial Luye Pharma risperidone LAI programme. Subject-level (time-fixed); set once from the trial identifier. Wang 2024 Discussion attributes the faster apparent absorption in the multiple-dose trial to its much sparser PK sampling scheme (trough-focused after the first injection) rather than to any formulation or population difference, so users simulating a clinical dosing scenario should generally leave this at the single-dose reference value of 0. Note that both Wang 2024 models set the elimination rate constant equal to KA (flip-flop kinetics) and derive the apparent central volume as V = CL/KA, so this indicator also shifts the apparent volume. Nested inside STUDY_CT_USA: a CT-USA-102 subject has both indicators set to 1. Follows the auto-approved STUDY_<id> canonical family.

DLVL (canonical for source-protocol dose-level integer indicator)

• Description: Integer dose-level / protocol-arm indicator carried per subject in the DDMODEL00000281 lidocaine bundle’s simulated dataset (Simulated_Lid_B04_ddmore.csv). Values 1-4 (or higher) flag distinct study-protocol dose / regimen tiers; the NA_NA_lidocaine.R model binarises as DLVL > 2 to switch the typical-value baselines for both the GX elimination rate constant K30 and the lidocaine apparent central volume V1 between a “low” (DLVL <= 2) and a “high” (DLVL > 2) regimen.
• Units: (integer-coded categorical)
• Type: categorical
• Scope: specific
• Reference category: the binary form DLVL <= 2 is the reference (THETA(4) for K30 and THETA(14) for V1 in the source .ctl); DLVL > 2 selects the higher-exposure regimen (THETA(5) and THETA(15) respectively).
• Source aliases: DLVL – the column header used in the DDMORE bundle’s .ctl $INPUT and the Simulated_Lid_B04_ddmore.csv data file.
• Example models: NA_NA_lidocaine.R (DDMODEL00000281; binary derivation DLVL_HIGH = as.integer(DLVL > 2) on K30 base and V1 base), Bursi_2017_lidocaine.R (the now-identified linked publication; same derivation, same threshold).
• Notes: The DDMODEL00000281 bundle’s linked publication – recorded here as unidentifiable when this entry was written – has since been identified as Bursi et al. (2017) Eur J Drug Metab Pharmacokinet 42(5):801-814, doi:10.1007/s13318-017-0400-7, extracted as Bursi_2017_lidocaine.R. That paper supplies the physical meaning the integer tiers previously lacked: DLVL is the number of lidocaine 5% medicated plasters applied simultaneously (Fig. 1 legend, “DLVL dose level, i.e., number of plasters applied simultaneously”), taking values 1, 2 or 3 under a protocol permitting up to three non-overlapping plasters for at most 12 h per 24-h period. The > 2 threshold therefore separates the maximum recommended dose (three plasters) from one or two, and the switch on both K30 and V1 encodes a real dose-nonlinearity: Results Sect. 3, “The influence of DLVL can be interpreted as a decrease of systemic bioavailability when the dose, i.e., the number of plasters, increases.” Because the plaster count also sets the zero-order input rate (1800 ug/h per plaster), this covariate enters a simulation twice – once through the dose record and once as the covariate column – and the two must be kept consistent. Scope stays specific because the count is meaningful only for this plaster product. The binary threshold > 2 reproduces both the paper’s Table 3 stratification and the bundle’s .ctl line IF(DLVL.GT.2)P1=0. If a future model needs a different dose-level binarisation or a continuous treatment, register a distinct canonical name rather than overloading DLVL.

S1A2 (canonical for source-protocol CYP1A2 substrate / co-medication categorical indicator)

• Description: Categorical CYP1A2-modifying co-medication / phenotype indicator carried per subject in the DDMODEL00000281 lidocaine bundle’s simulated dataset. Integer code; in the NA_NA_lidocaine.R model the value S1A2 == 3 selects the “CYP1A2 inducer present” sub-cohort (lidocaine N-deethylation to MEGX is CYP1A2-mediated, so the modifier acts on the GX elimination rate constant K30 in the source’s parameterisation). Other integer codes (0, 1, 2) are pooled into the reference.
• Units: (integer-coded categorical)
• Type: categorical
• Scope: specific
• Reference category: S1A2 != 3 (i.e., values 0, 1, 2) – pooled into the reference.
• Source aliases: S1A2 – the column header used in the DDMORE bundle’s .ctl $INPUT and the simulated dataset, with sibling columns D1A2 and H1A2 carried in the data file but dropped via =DROP in the source .ctl.
• Example models: NA_NA_lidocaine.R (DDMODEL00000281; binary derivation S1A2_IND = as.integer(S1A2 == 3) on the GX elimination rate constant K30), Bursi_2017_lidocaine.R (the now-identified linked publication; binary derivation S1A2_IND = (S1A2 == 1) on the same rate constant).
• Notes: The DDMODEL00000281 bundle’s linked publication – recorded here as unidentifiable when this entry was written – has since been identified as Bursi et al. (2017) Eur J Drug Metab Pharmacokinet 42(5):801-814, doi:10.1007/s13318-017-0400-7, extracted as Bursi_2017_lidocaine.R. That paper settles the covariate’s meaning: Sect. 2.7.4 explores “the effect of concomitant medications, such as atenolol (ATEN), a cytochrome P450 1A2 (CYP1A2) substrate, and metoprolol and beta-blockers, as separate group” and states that “Concomitant medication was expressed as binary data (concomitant medication yes or no)”. So the concept is a concomitant CYP1A2-substrate co-medication indicator (competition for the isozyme that N-deethylates lidocaine), not the CYP1A2-induction / smoking indicator this entry previously guessed at; the paper’s Table 3 modifier of +0.852 /h on K30 corresponds to the bundle’s +0.853. Two encodings of the one concept therefore coexist and both are correct for their own source: the paper carries a plain binary yes/no column, while the bundle’s simulated dataset carries an integer code whose level 3 is that column’s “yes” level. Read the Example models derivation to see which encoding a given model expects; supply a 0/1 column for Bursi_2017_lidocaine.R and the integer code for NA_NA_lidocaine.R. Scope stays specific because the integer coding remains dataset-specific. Sibling columns D1A2 and H1A2 are dropped via =DROP in the bundle’s .ctl, so only the level-3 indicator is structurally identifiable from the surviving model code. If a future model needs a richer encoding of CYP1A2 modulation, register a separate canonical (e.g., CYP1A2_IND) rather than overloading S1A2.

DOSE_HIGH_EFL (canonical for high-dose eflornithine indicator)

• Description: 1 = dose record is at the highest oral eflornithine dose level (>= 3000 mg/kg in the Jansson 2008 Sprague-Dawley rat single-dose study), 0 = oral dose at the lower levels (750, 1500, 2000 mg/kg) or any IV dose.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (750-2000 mg/kg oral, or IV at 375 / 1000 mg/kg).
• Source aliases: derived per dose record from the administered amount and route. Jansson 2008 modeled it as a categorical indicator on bioavailability for both enantiomers (Results paragraph; OFV drop -11.4).
• Example models: Jansson_2008_eflornithine_rat.R.
• Notes: Specific scope because the threshold (3000 mg/kg of body weight) and the bioavailability shifts (+14.6% for L-eflornithine, +32.8% for D-eflornithine relative to the 750-2000 mg/kg reference) are intrinsically tied to the Jansson 2008 dose-design and are not transferable to other drugs. Encoded as a binary because the paper’s prose explicitly states linear and power dose-F relationships did not improve the fit. Per-subject indicator in the source data (each rat received exactly one oral dose); for multi-dose simulation, set the indicator per dose record.

DOSE_PEG_300UGKG (canonical for pegfilgrastim 300 ug/kg dose-cohort indicator)

• Description: 1 = subject received pegfilgrastim at 300 ug/kg (the highest pegfilgrastim dose evaluated in Melhem 2018; n = 12 subjects: 8 healthy adult volunteers and 4 adult cancer patients on chemotherapy), 0 = filgrastim or pegfilgrastim at any dose other than 300 ug/kg. Encodes the empirically-observed low-clearance behaviour of the pegfilgrastim 300 ug/kg cohort as a subject-level cohort indicator on CLD (Melhem 2018 Results ‘Model development’ and Table 2 row CLD PEG 300 ug/kg = 0.107 L/h vs CLD PEG = 0.362 L/h). Subject-level (each subject enrolled in one dose-cohort per study); coexists with FORM_GCSF_PEG in the covariate frame and only takes non-zero effect when FORM_GCSF_PEG = 1 (a filgrastim subject with DOSE_PEG_300UGKG = 1 is nonsensical per the paper design).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (filgrastim, or pegfilgrastim at any dose other than 300 ug/kg).
• Source aliases:
  • DOSE_PEG_300 – paper-narrative descriptor for the pegfilgrastim 300 ug/kg cohort in Melhem_2018_g_csf.R.
• Example models: Melhem_2018_g_csf.R.
• Notes: Sibling of DOSE_HIGH_EFL (Jansson 2008 eflornithine highest-dose indicator) and DOSE_400MG / DOSE_50MG / DOSE_70MG / DOSE_130MG / DOSE_260MG (Jorga 2000 / other dose-level indicators). Specific scope because the low-clearance shift (0.107 / 0.362 = fold-decrease 0.295) is intrinsically tied to the Melhem 2018 pooled analysis. Melhem 2018 introduced the effect after graphical analysis of IIV suggested distinct clearance behaviour for the 300 ug/kg pegfilgrastim cohort; the paper notes the small sample size (n = 12) and physiological plausibility of the finding. Encoded as an additive log-scale shift on CLD that is meaningful only when FORM_GCSF_PEG = 1; a study of pegfilgrastim at other dose levels should set DOSE_PEG_300UGKG = 0. Auto-approved sibling of the DOSE_HIGH_<drug> / DOSE_<N>MG cohort-indicator family.

DOSE_HIGH_RIV (canonical for high-dose rivaroxaban indicator)

• Description: 1 = dose record is at the 20 mg-equivalent body-weight-adjusted rivaroxaban dose, 0 = the 10 mg-equivalent body-weight-adjusted rivaroxaban dose. The Willmann 2018 EINSTEIN-Jr phase I paediatric study (NCT01145859) dosed children with body-weight-adjusted amounts targeting adult exposures of either rivaroxaban 10 mg or 20 mg; the popPK model estimates a relative bioavailability F1 = 0.648 for the 20 mg-equivalent dose, with the 10 mg-equivalent dose anchored at F1 = 1 by definition.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (10 mg-equivalent body-weight-adjusted dose; relative bioavailability fixed to 1 in the source paper).
• Source aliases: derived per dose record from the EINSTEIN-Jr phase I dose level assigned to the cohort; the paper’s Methods describe the dose stratification as “two dose levels, equivalent to adult doses of rivaroxaban 10 mg and 20 mg” (Willmann 2018 Methods / Modelling strategy).
• Example models: Willmann_2018b_rivaroxaban.R (multiplicative effect on the depot bioavailability: fdepot <- exp(lfdepot * DOSE_HIGH_RIV) with lfdepot = log(0.648) per Willmann 2018 Table 1, RSE 9.03%; the dose-dependent reduction is consistent with the saturable-solubility behaviour of rivaroxaban reported in the adult patient popPK [reference 21 of Willmann 2018]).
• Notes: Specific scope because the 20 mg-equivalent vs 10 mg-equivalent dose stratification and the 0.648 relative-bioavailability estimate are intrinsically tied to the Willmann 2018 EINSTEIN-Jr phase I paediatric study. Drug-specific member of the DOSE_HIGH_* family alongside DOSE_HIGH_EFL (Jansson 2008 eflornithine high-dose indicator). Per-dose-occasion indicator in principle, although in the Willmann 2018 single-dose study each subject received exactly one rivaroxaban dose.

DOSE_LOW_AMG221 (canonical for low-dose AMG 221 indicator)

• Description: 1 = dose record is the 3 mg oral AMG 221 dose in the Gibbs 2011 phase 1 study; 0 = the 30 or 100 mg oral AMG 221 dose. Sibling of DOSE_HIGH_EFL: same drug-suffixed dose-level indicator family, opposite direction of effect (the reduced-bioavailability tier is the flagged tier).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (30 or 100 mg oral AMG 221, where F1 = 1).
• Source aliases: derived per dose record from the administered amount. Gibbs 2011 Table III + footnote a: F1 = 0.546 at 3 mg vs F1 = 1 fixed at 30 or 100 mg.
• Example models: Gibbs_2011_amg221.R.
• Notes: Specific scope because the threshold (3 mg vs 30/100 mg oral AMG 221 as a suspension in healthy obese adults) and the -45.4% bioavailability shift are intrinsically tied to the Gibbs 2011 dose-design and are not transferable to other drugs. Gibbs 2011 Discussion attributes the reduced 3 mg bioavailability to a possible high-affinity intestinal-metabolism / transport process saturating at higher doses (Caco-2 permeability plus in vitro CYP3A metabolism with Km > 100 uM, so a 30 mg dose is expected to produce intestinal concentrations high enough to saturate intestinal metabolism); the paper flags an alternative Michaelis-Menten dose-F structure that was not fit because only three discrete dose levels were tested. Per-dose-record indicator; observation rows inherit the indicator from the preceding dose.

MEAL_A (canonical for Zvada 2010 meal-A high-fat English breakfast indicator)

• Description: 1 = oral dose administered 30 min after Zvada 2010 meal A (a high-fat English breakfast); 0 = otherwise. Per Zvada 2010 Table 1 meal A consists of 2 rashers of bacon (20 g), 1 fried egg (50 g), 1 slice white toast (30 g) with butter (7 g) and marmalade (10 g), 2 cups decaffeinated coffee (400 ml) with full-cream milk (100 ml) and 2 teaspoons sugar (10 g); 18.9 g protein, 27 g fat, 38 g carbohydrate, 1,966 kJ, 627 g total weight. Distinct from the general FED_HIGHFAT because Zvada 2010 isolates four operationally-distinct meal compositions (A/B/C/D) each with its own bioavailability effect estimate and because the paper’s Discussion explicitly attributes part of meal A’s effect to eggs rather than total fat.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (fasted dosing, equivalent to Zvada 2010 meal E = 200 ml of water; the all-zero state of MEAL_A / MEAL_B / MEAL_C / MEAL_D).
• Source aliases: paper narrative “meal A” (Zvada 2010 Table 1 / Table 3).
• Example models: Zvada_2010_rifapentine.R (additive contribution to TVF = 1 * (1 + sum of per-meal fractional changes); RXF_A = +0.857, RSE 20.1%; Zvada 2010 Table 3).
• Notes: Specific scope because the operational meal definition is paper-specific. Mutually exclusive with MEAL_B, MEAL_C, MEAL_D - at most one of the four indicators can be 1 on a given dosing record; the fasted reference is all four = 0. Future rifapentine / antimycobacterial food-effect studies that reuse the same meal compositions (uncommon - meal compositions are typically protocol-specific) can extend this entry’s example list. Distinct from MEAL_FLAG / SNACK_FLAG (time-varying intra-day flags) because MEAL_A is a per-dose-record indicator that selects between five mutually exclusive meal states.

MEAL_B (canonical for Zvada 2010 meal-B low-fat bulky maize-meal porridge indicator)

• Description: 1 = oral dose administered 30 min after Zvada 2010 meal B (a low-fat bulky maize-meal porridge breakfast); 0 = otherwise. Per Zvada 2010 Table 1 meal B consists of 1.5 cups soft maize meal porridge (375 g cooked) with 3 teaspoons sugar (15 g), 1 cup decaffeinated coffee (200 ml) with full-cream milk (50 ml) and 1 teaspoon sugar (5 g); 6 g protein, 3 g fat, 66 g carbohydrate, 1,285 kJ, 645 g total weight.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (fasted dosing, equivalent to Zvada 2010 meal E).
• Source aliases: paper narrative “meal B” (Zvada 2010 Table 1 / Table 3).
• Example models: Zvada_2010_rifapentine.R (additive contribution to TVF; RXF_B = +0.327, RSE 40.4%; Zvada 2010 Table 3).
• Notes: Specific scope; mutually exclusive with MEAL_A, MEAL_C, MEAL_D. See MEAL_A notes for the per-dose-record indicator pattern.

MEAL_C (canonical for Zvada 2010 meal-C high-fat bulky maize-meal porridge with lard indicator)

• Description: 1 = oral dose administered 30 min after Zvada 2010 meal C (a high-fat bulky maize-meal porridge breakfast with lard); 0 = otherwise. Per Zvada 2010 Table 1 meal C consists of 1.5 cups soft maize meal porridge (375 g cooked) with 3 teaspoons sugar (15 g) and 5 teaspoons of lard (25 g), 1 cup decaffeinated coffee (200 ml) with full-cream milk (50 ml) and 1 teaspoon sugar (5 g); 6 g protein, 28 g fat, 66 g carbohydrate, 2,229 kJ, 670 g total weight.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (fasted dosing, equivalent to Zvada 2010 meal E).
• Source aliases: paper narrative “meal C” (Zvada 2010 Table 1 / Table 3).
• Example models: Zvada_2010_rifapentine.R (additive contribution to TVF; RXF_C = +0.457, RSE 29.3%; Zvada 2010 Table 3).
• Notes: Specific scope; mutually exclusive with MEAL_A, MEAL_B, MEAL_D. Meal C and meal A have similar total fat content but the Zvada 2010 Discussion attributes the substantially smaller effect of meal C (+45.7%) versus meal A (+85.7%) to the absence of eggs in C - supporting the authors’ egg-specific bioavailability hypothesis.

MEAL_D (canonical for Zvada 2010 meal-D low-fat high-fluid chicken noodle soup indicator)

• Description: 1 = oral dose administered 30 min after Zvada 2010 meal D (a low-fat high-fluid chicken noodle soup breakfast); 0 = otherwise. Per Zvada 2010 Table 1 meal D consists of 2 cups reconstituted powdered chicken noodle soup (400 ml), 1 cup decaffeinated coffee (200 ml) with skim milk (50 ml) and 1 teaspoon sugar (5 g); 9 g protein, 4 g fat, 28 g carbohydrate, 774 kJ, 660 g total weight.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (fasted dosing, equivalent to Zvada 2010 meal E).
• Source aliases: paper narrative “meal D” (Zvada 2010 Table 1 / Table 3).
• Example models: Zvada_2010_rifapentine.R (additive contribution to TVF; RXF_D = +0.489, RSE 30.7%; Zvada 2010 Table 3).
• Notes: Specific scope; mutually exclusive with MEAL_A, MEAL_B, MEAL_C. Zvada 2010 Discussion notes that monosodium glutamate (MSG) in the chicken noodle soup may contribute to the effect by accelerating gastric emptying (per Boutry et al., Am J Physiol Endocrinol Metab 2011, on MSG and gastric emptying).

CONMED_ATORVASTATIN_DOSE, CONMED_FLV_DOSE, CONMED_LOV_DOSE, CONMED_PRV_DOSE, CONMED_RSV_DOSE, CONMED_SMV_DOSE, CONMED_EZT_DOSE, CONMED_GEMCABENE_DOSE, CONMED_INH_DOSE, CONMED_TRAMADOL_DOSE, CONMED_VENLAFAXINE_DOSE, CONMED_FLUOXETINE_DOSE (canonical for daily dose of a co-administered named drug)

• Description: Daily dose of the named drug (suffix = INN abbreviation: ATORVASTATIN atorvastatin (spelled out to avoid an ATV collision with atazanavir; see CONMED_ATAZANAVIR), flv fluvastatin, lov lovastatin, prv pravastatin, rsv rosuvastatin, smv simvastatin, ezt ezetimibe, GEMCABENE gemcabene (spelled out; the investigational lipid-altering agent CI-1027 has no established 3-4 letter abbreviation), inh isoniazid, TRAMADOL tramadol (spelled out to keep drug names unambiguous in the MBMA opioid-pain family), VENLAFAXINE venlafaxine and FLUOXETINE fluoxetine (both spelled out; the natural abbreviations VEN / FLU are ambiguous – FLU in particular would collide with fluvastatin’s FLV, fluconazole and influenza-vaccine indicators)). 0 = the named drug is not part of the regimen for this study arm / subject; positive value = total daily dose. Captures the dose-response amplitude in MBMA or co-administered-drug PK/PD models where each drug arm contributes its own dose-effect curve.
• Units: mg/day for the statin / ezetimibe / gemcabene / tramadol / antidepressant series; mg/kg for CONMED_INH_DOSE (paper-specific unit, documented in covariateData[[CONMED_INH_DOSE]]$units).
• Type: continuous
• Scope: specific
• Reference category: 0 (the drug is not given in this arm; equivalent to “this drug-arm contributes 0 to its dose-effect term”).
• Source aliases: DOSE_<drug> legacy form (used in Vargo 2014 statins / ezetimibe MBMA pre-rename, Chen 2017 TB mouse pre-rename); aliases of the canonical CONMED_<drug>_DOSE form rename. CONMED_ATV_DOSE.
• Example models: Vargo_2014_statins_ezetimibe_mbma.R (each CONMED_<statin>_DOSE drives the corresponding statin’s Hill / Emax dose-response curve; per-arm dose level in mg/day, default 0 outside the named statin arm; ezetimibe and statin arms combine via the sub-additive gamma_int term), Chen_2017_TB_MTP_GPDI_mouse.R (CONMED_INH_DOSE drives the isoniazid CL adjustment: cl_inh = cl_inh_lowdose * (1 - slope_inh * (CONMED_INH_DOSE - 12.5))), Mercier_2014_tramadol_tapentadol_mbma.R (CONMED_TRAMADOL_DOSE drives the tramadol Emax-in-dose term in the extent-of-reduction R with ED50 = 184 mg/day), Mandema_2005_gemcabene_mbma.R (the earlier LDL-C MBMA that Vargo 2014 descends from: CONMED_ATORVASTATIN_DOSE / CONMED_RSV_DOSE / CONMED_SMV_DOSE / CONMED_LOV_DOSE / CONMED_PRV_DOSE feed one shared-Emax statin sigmoid through statin-specific ED50s, while CONMED_EZT_DOSE and CONMED_GEMCABENE_DOSE each drive their own Emax / ED50 / Hill dose-response and combine with the statin term through drug-specific interaction coefficients), Maringwa_2025_venlafaxine_fluoxetine_mbma.R (CONMED_VENLAFAXINE_DOSE drives an Emax-in-dose term with ED50 = 29.1 mg/day, while CONMED_FLUOXETINE_DOSE is used only as dose > 0 because no fluoxetine dose-response was identifiable over the studied 20-60 mg/day range – the founding example in this family of a member whose numeric dose is retained for provenance but enters the equations only as an arm indicator).
• Notes: The CONMED_<drug>_DOSE shape replaces the earlier DOSE_<drug> / DOSE_<drug>_<unit> names (which conflated a covariate with a dose-amount column). The atorvastatin slot is spelled out as CONMED_ATORVASTATIN_DOSE rather than CONMED_ATV_DOSE to keep the ATV abbreviation reserved for the binary atazanavir coadministration indicator (CONMED_ATAZANAVIR, source alias ATV) used in HIV-PK extractions like ArabAlameddine 2012 raltegravir. Tramadol is spelled out for the same reason (no established TRM abbreviation collision, but the full drug name keeps the MBMA opioid-pain family unambiguous alongside the sibling TAPENTADOL treatment-arm indicator). New co-medication dose-effect MBMA models should reuse this family and add the appropriate <drug> INN abbreviation; spell out drug names whenever the natural 3-4 letter abbreviation would collide with another CONMED_ entry. The units field is per-paper.

ON_TREATMENT (canonical for generic on-treatment (active-vs-placebo) indicator)

• Description: Generic binary treatment-arm indicator. 1 = subject is in the active-drug arm (or, for time-varying use, the subject is currently on treatment at this record), 0 = subject is in the placebo arm (or, for time-varying use, currently off treatment / pre-treatment / placebo-controlled run-in). Per-subject and time-fixed by default in randomized parallel-group trials; admits a time-varying interpretation for cross-over, run-in, or treatment-suspension designs where the indicator switches at well-defined per-record times. Used in non-PK / non-exposure disease-progression and PD models where the “drug effect” enters the structural equations as a categorical on/off switch rather than as an exposure-driven response, and where the trial design pools or abstracts away dose-level differences within the active arm.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (placebo arm / off treatment).
• Source aliases:
  • Trt (Lee 2011 paper notation: Trt = (0, 1, ...), where the source paper notes the indicator could in principle take more than two levels but only the binary placebo-vs-active contrast is used in the actual analyses) – used in Lee_2011_parkinson_progression.R.
  • Active treatment – Bhatnagar 2024 Table S4 row label for the coefficient; the paper’s Equation 3 writes it beta_trt * TRT. Used in the four Bhatnagar_2024_upadacitinib_asas* models.
• Example models: Lee_2011_parkinson_progression.R (additive shifts on the placebo disease-progression slope and on the placebo short-term symptomatic effect: slope = beta0 + beta1 * ON_TREATMENT + etaslope and symeff = gamma0 + gamma1 * ON_TREATMENT + etasymeff; default ini() parameter values are from the TEMPO study rasagiline arm, so ON_TREATMENT = 1 predicts the rasagiline-active disease trajectory and ON_TREATMENT = 0 predicts the placebo trajectory), Bhatnagar_2024_upadacitinib_asas20_as.R, Bhatnagar_2024_upadacitinib_asas40_as.R, Bhatnagar_2024_upadacitinib_asas20_nraxspa.R, Bhatnagar_2024_upadacitinib_asas40_nraxspa.R (four week-14 ASAS-response logistic regressions in which the treatment indicator is the ONLY predictor: the source evaluated linear and nonlinear exposure models against a treatment-effect model and the treatment-effect model won on AIC, so upadacitinib exposure is absent from the final equations. These are the founding examples of ON_TREATMENT being the whole drug-effect model rather than an adjunct to one, and of the case where the source explicitly SET exposure to zero on the placebo arm, making the indicator and a positive exposure perfectly confounded), Loprete_2016_safinamide_ontime.R (additive 0.728 h shift on the daily-ON-time intercept, intercept = intplac + inttreat * ON_TREATMENT + etaintercept; Loprete 2016 equation 7 and Table 4. A third variant of the exposure-was-tested-and-lost case: the source computed a per-visit average safinamide concentration from the companion population PK model, screened it on both the intercept and the slope of the ON-time trajectory, and retained neither, so the pooled indicator replaces exposure even though a full PK model for the same patients exists in the same paper. It also pools the 50 mg and 100 mg safinamide arms, which the source justifies by observing very little difference between them), Zierhut_2016_hcc_antiangiogenic_os_mbma.R (study-ARM rather than patient level: selects between two parallel intercepts – lmos_aat * ON_TREATMENT + lmos_pbo * (1 - ON_TREATMENT) – in an advanced-hepatocellular-carcinoma median-overall-survival MBMA whose source states its antiangiogenic-therapy and placebo indicators are “mutually exclusive and collectively exhaustive”, so the single canonical column reproduces both of the paper’s indicators and no input row can set them inconsistently).
• Notes: Distinct from TRT (which is the Novakovic 2017 cladribine-cohort categorical indicator with three levels and a paper-specific 1-vs-2 cohort labelling) and from DRUG_ORMU / DRUG_PRED / DRUG_OBI (which are head-to-head drug-product comparator indicators carrying the contrast between two named active drugs in the same trial). Use ON_TREATMENT only when the structural model has no drug-specific PK / exposure and the active arm enters as a pooled on/off switch. When extracting a head-to-head comparator paper (drug A vs drug B) register a sibling DRUG_<X> canonical instead. When the source paper differentiates multiple active dose levels with distinct parameters, register a categorical indicator with paper-specific level labelling (e.g., TRT for the cladribine cohorts) rather than overloading ON_TREATMENT. Per-model covariateData[[ON_TREATMENT]]$notes should document which drug the active arm corresponds to in that file’s default ini() values, the trial design (parallel-group vs cross-over), and any time-varying interpretation.

PLACEBO (canonical for placebo-arm membership indicator)

• Description: Binary indicator of randomized placebo-arm membership: 1 = subject is in the placebo arm, 0 = active-treatment arm. Time-fixed per subject. Used to select placebo-arm-specific structural parameters (e.g., a separately estimated baseline) in models that ALSO carry an exposure-driven active-drug effect, so the placebo indicator is a covariate on a structural parameter rather than the drug-effect switch itself.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (active-treatment arm)
• Source aliases: none.
• Example models: Hansson_2013_sunitinib_dbp.R (switches the typical baseline diastolic blood pressure dbp0 from 71.8 mmHg in the active arm to 77.6 mmHg in the placebo run-in arm via exp(ldbp0 + e_placebo_dbp0 * PLACEBO); the sunitinib AUC drug effect on kin is modelled separately and exposure-driven), Lalovic_2020_lemborexant_somnolence.R, Hammer_2020_acetaminophen.R (0.524-fold multiplier on acetaminophen clearance in the saline-placebo arm, whose subjects had residual acetaminophen concentrations, via e_placebo_cl^PLACEBO).
• Notes: Distinct from ON_TREATMENT (the generic pooled active-vs-placebo on/off switch used when the active arm REPLACES exposure in the structural model, reference 0 = placebo): PLACEBO has the opposite polarity (1 = placebo) and is used alongside, not instead of, an exposure-driven drug effect, marking the placebo subpopulation for an arm-specific structural parameter. If a model needs the pooled on/off drug switch, use ON_TREATMENT (= 1 - PLACEBO) instead.

SEMESTER (canonical for paired-season indicator (winter or spring vs summer or fall))

• Description: Calendar paired-season indicator: 1 = winter or spring (December 21 to June 19), 0 = summer or fall (June 20 to December 20). Time-fixed per subject (or per simulation) based on the calendar date at which the observation occurred.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (summer or fall)
• Source aliases:
  • SEMESTER1 – used in Gonzalez-Sales 2015 testosterone circadian model with the same encoding (winter or spring = 1, summer or fall = 0).
• Example models: GonzalezSales_2015_testosterone.R (+8.09% multiplicative effect on Base; Table III).
• Notes: Gonzalez-Sales 2015 split season into a paired SEMESTER1 / SEMESTER2 covariate (SEMESTER1 = 1 if winter or spring; SEMESTER2 = 1 if fall or winter), keeping only SEMESTER1 in the final model. The canonical column omits the 1 suffix because the second indicator was not retained; the source-paper naming is recorded as a source alias. Calendar boundaries are reported verbatim in the paper Methods (winter: Dec 21 - Mar 19; spring: Mar 20 - Jun 19; summer: Jun 20 - Sep 21; fall: Sep 22 - Dec 20). Scope: specific until a second endogenous-rhythm model adopts the same encoding; at that point promote to general. Distinct from SEASON2 (RSV-second-exposure indicator) which has different semantics.

CEN (canonical for Wahlby 2004 pefloxacin study-centre indicator)

• Description: Binary study-centre indicator carried over from the Karlsson MO, Sheiner LB 1993 (J Pharmacokin Biopharm 21(6):735-750) pefloxacin analysis that Wahlby 2004 re-fit. The source paper does not specify which of the two recruiting centres corresponds to CEN = 1 vs CEN = 0; the indicator captures an unexplained between-centre shift in clearance.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (one of the two recruiting centres; the source paper does not state which).
• Source aliases:
  • CEN – Wahlby 2004 source-column convention; used in Wahlby_2004_pefloxacin.R.
• Example models: Wahlby_2004_pefloxacin.R (multiplicative effect on CL via exp[0.19 * CEN]; CEN = 1 increases CL by ~21 percent relative to CEN = 0).
• Notes: Specific scope; the underlying centre identity is paper-defined and not generalisable. Users assembling a virtual cohort should treat CEN as a sensitivity covariate or fix it at 0 (the reference centre) when between-centre exploration is not relevant.

FORM_AXI_XLI (canonical for axitinib Form XLI vs Form IV crystal polymorph indicator)

• Description: 1 = subject received axitinib as crystal polymorph Form XLI (the marketed commercial polymorph used in the approved tablet); 0 = subject received axitinib as crystal polymorph Form IV (the earlier Phase I polymorph and the typical-value bioavailability reference). Per-dose-occasion categorical indicator: a single subject may carry both indicator values across the within-subject crossover formulation arms in studies 7 and 9.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (Form IV; the typical-value F reference in Garrett 2014 Table 3, F = 0.465 in the fed state).
• Source aliases:
  • FORM – used in Garrett_2014_axitinib.R (Garrett 2014 Methods ‘Statistical analysis’ and Results ‘Full model and final model’ describe the formulation as a categorical {Form IV, Form XLI} with Form XLI = 1 as the test polymorph and Form IV = 0 as the typical-value reference).
  • Form – used in Rini_2013_axitinib.R (Rini 2013 Table 1 records the polymorph per treatment arm and Table 2 carries the row “Form XLI on F”; same orientation as the canonical).
• Example models: Garrett_2014_axitinib.R (linear-proportional effect on bioavailability per Garrett 2014 Table 3 final estimates: f = exp(lfdepot) * (1 + e_fast_f * (1 - FED)) * (1 + e_xli_f * FORM_AXI_XLI) with e_xli_f = -0.150 (15% reduction in F for Form XLI relative to Form IV in the fed state); Form XLI has no retained effect on ka, CL, Vc, Q, or Vp), Rini_2013_axitinib.R (the pooled healthy-volunteer-plus-patient analysis that precedes Garrett 2014: e_xli_f = -0.121 per Rini 2013 Table 2, and the fasting effect on F is GATED to Form IV records – f = exp(lfdepot) * (1 + e_fast_f * (1 - FED) * (1 - FORM_AXI_XLI)) * (1 + e_xli_f * FORM_AXI_XLI) – because Rini 2013 Table 2 footnote g states that no food effect was observed with Form XLI, giving F = 0.402 for Form XLI in either prandial state).
• Notes: Specific scope because the Form XLI vs Form IV crystal polymorph contrast is tied to the axitinib drug-product-development comparison in Garrett 2014. Crystal polymorph difference (same dosage form, same drug substance, different crystal lattice with different solubility / dissolution behaviour) rather than a dosage-form difference (tablet vs capsule vs solution) – distinct from FORM_TABLET, FORM_CAPSULE, FORM_SOLUTION, and the rest of the dosage-form FORM_* family. Mirrors the existing FORM_SAR_DP2 (sarilumab drug-product 2), FORM_ISA_P2F2 (isatuximab P2F2 drug material), FORM_LINAG_TAB1 (linagliptin tablet 1), FORM_VISMO_PHASEI (vismodegib Phase I capsule), FORM_ITR_SUBA (SUBA-itraconazole vs Sporanox capsule), FORM_TAC_IR (tacrolimus IR vs prolonged release), FORM_DOX_DORYX_MPC (Doryx MPC tablet), and FORM_THEO_APNECUT (theophylline Apnecut) entries under the FORM_* family of drug-specific drug-product-version / polymorph indicators. Set to 0 to simulate the Form IV early-phase reference; set to 1 to simulate the Form XLI marketed product (the routine clinical-use default).

STUDY_B1831090 (canonical for Abrantes 2017 moroctocog study B1831090 cohort indicator)

• Description: 1 = subject enrolled in study B1831090 (phase I bioequivalence study of Refacto; n = 18, severe and moderately severe hemophilia A, rich sampling, CSA assay) of the Abrantes 2017 pooled population PK analysis; 0 = any other study in the 13-study pool (B1831003, B1831004, B1831015, B1831053, B1831054, B1831061, B1831066, B1831067, B1831068, B1831070, B1831071, B1831077). Used to switch the typical CL magnitude downward for the B1831090 cohort while keeping the rest of the pooled-population PK parameters applicable to the remaining 12 studies.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (any non-B1831090 study in the Abrantes 2017 pool).
• Source aliases:
  • STUD – used in Abrantes_2017_moroctocog.R (Abrantes 2017 Table 2 footnote b: study indicator on CL).
• Example models: Abrantes_2017_moroctocog.R (multiplicative effect on CL: (1 - 0.347 * STUDY_B1831090) so B1831090 subjects have ~34.7% lower CL than the other 12 studies; Abrantes 2017 Table 2 also reports a 95% CI of -40.7% to -24.2%).
• Notes: Specific scope because the contrast is tied to the Abrantes 2017 pooled-analysis design. Subject-level / time-fixed; set once from the trial identifier on each subject record. Inclusion in the model allows the typical PK parameters to describe the remaining 12 studies, with B1831090 captured by the indicator (Abrantes 2017 Discussion).

TINF (canonical for the duration of an intravenous infusion)

• Description: Duration of the intravenous infusion of the dose record, in hours. Continuous. Used as a covariate when a source analysis estimates an effect of the administration protocol on PK parameters; this is separate from, and in addition to, the physical infusion duration carried on the dose record itself.
• Units: h
• Type: continuous
• Scope: general
• Reference category: none (continuous). Source analyses typically enter it as a linear deviation from a paper-specific reference duration, so record the reference in the model’s covariateData[["TINF"]]$notes rather than assuming a library-wide default.
• Source aliases:
  • Tinf – used in Schreib_2024_busulfan.R (Table 3 covariates theta_V3 and theta_k6, both entering as (Tinf - 3 h)).
  • TSEOI – time since the end of the infusion in the Zhou 2021 Figure S3 control stream, where a sample is flagged as during the infusion when TSEOI.EQ.0; Zhou_2021_remimazolam.R reconstructs that flag as tad(central) < TINF * 60 because rxode2 does not expose the infusion duration.
• Example models: Zhou_2021_remimazolam.R (not a covariate on a PK parameter: it selects the during-infusion branch of a venous:arterial concentration ratio on the venous output only, with a bolus entered as TINF = 0), Schreib_2024_busulfan.R (exponential effects on both the central volume, exp(0.226 * (TINF - 3)), and the elimination rate constant, exp(-0.161 * (TINF - 3)); the center changed its standard busulfan infusion from 3 h to 4 h in October 2014, so only two values occur in that cohort and the covariate term is 0 or 1).
• Notes: Matches the near-universal pharmacometric symbol T_inf. Supply this as an explicit data column even though the event table already encodes the infusion duration – rxode2 takes the physical infusion duration from the dur / rate fields of the dose record and does not expose it to model(), so a covariate effect on infusion duration needs its own column. Opposite-signed effects on volume and on the elimination rate constant (as in Schreib 2024) are a recognised signature of a one-compartment model absorbing a distribution phase that a shorter infusion exposes – clearance is then nearly independent of infusion duration; note that interpretation when it applies rather than reading the two effects as independent physiology.

T_PUMP (canonical for the duration of a breast-milk pumping (expression) session)

• Description: Time the lactating subject spent expressing milk with a breast pump to produce the milk sample, i.e. the duration of the collection procedure itself. Continuous. In lactation PK the covariate is not a nuisance sampling detail: milk fat content rises through a single expression as hind-milk is released, so a longer pumping session yields a fattier, more lipid-rich sample and a correspondingly different concentration of the analyte, water-soluble drugs being diluted and lipophilic drugs concentrated. Recorded per sample, so it is time-varying within subject in principle; the founding model treats it as a per-subject constant because each woman’s sessions were of similar length and only one duration is tabulated per individual.
• Units: h
• Type: continuous
• Scope: general
• Reference category: none (continuous). Source analyses centre it at a cohort-typical duration and enter the deviation linearly, so record the centring value in the model’s covariateData[["T_PUMP"]]$notes rather than assuming a library-wide default. Melander 2025 centres at 0.22 h (13.2 min).
• Source aliases:
  • TPUMP – used in Melander_2025_cetirizine.R (the NONMEM covariate symbol carried in the published covariate equation, VTpump = 1 + VTpump1 * (TPUMP - 0.22); same quantity in the same units, no value transformation).
• Example models: Melander_2025_cetirizine.R (linear centred effect on the apparent milk volume of distribution, vc = 19.9 * (1 + 3.47 * (T_PUMP - 0.22)); longer pumping increases the apparent milk volume, which the authors attribute to the higher fat content of milk expressed later in a session and cetirizine’s water solubility. Sole covariate retained by a stepwise search that also screened maternal age, maternal weight, BMI, breastfeeding exclusivity and infant age).
• Notes: Units are hours, matching the general library convention and the equation text of the founding paper, even though lactation studies conventionally quote pumping durations in minutes (Melander 2025 reports a cohort mean of 14.9 min, range 2-50 min, i.e. 0.033-0.833 h); convert minutes to hours when building a covariate column rather than registering a minute-scaled sibling. Per the T_<event> canonical family the T_ prefix denotes a procedure-related time covariate and, as with T_CPB (cardiopulmonary-bypass run duration), the “event” here is the pumping session itself and the column carries its duration – not a time since or until it. Distinct from TINF (duration of an intravenous infusion, a dose-administration rather than a sample-collection procedure) and from TPP (time postpartum). A study that instead records which milk fraction was sampled (fore-milk vs hind-milk vs whole-breast) is measuring a related construct through a categorical lens and should register its own indicator rather than being coerced into this continuous column.

STUDY_LEFAMULIN_PHASE1 (canonical for Bian 2024 lefamulin phase 1 study cohort indicator)

• Description: 1 = subject enrolled in one of the Phase 1 studies (n = 98 healthy adults) of the pooled lefamulin popPK model-building dataset; 0 = otherwise. Paired with STUDY_LEFAMULIN_PHASE2; both indicators 0 selects the Phase 3 CABP studies (n = 622), which are the reference stratum. Shifts clearance, distributional clearance to the first peripheral compartment, and the first peripheral volume.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 with STUDY_LEFAMULIN_PHASE2 also 0 (Phase 3 community-acquired bacterial pneumonia studies). Bian 2024 Equations 1, 3 and 4 all state it explicitly: “When PHASE = 3, CLPHASE = 1”.
• Source aliases: derived per subject from the integer PHASE column of the pooled dataset (PHASE == 1 -> 1).
• Example models: Bian_2024_lefamulin_original_ppb.R, Bian_2024_lefamulin_higher_ppb.R (multiplies CL by 1.766 / 1.710, CLd1 by 2.12 / 1.788, and Vp1 by 2.75 / 1.889 in the original- and higher-plasma-protein-binding fits respectively).
• Notes: Specific scope because the contrast is tied to the lefamulin clinical-development pooled analysis. Drug-specific paper-anchored member of the STUDY_<DRUG>_PHASE<N> family alongside STUDY_SULDUR_PHASE2 / STUDY_SULDUR_PHASE3 (Cammarata 2024), STUDY_POSA_PHASE3 (van Iersel 2018), STUDY_ASP8232_PHASE2 (Snelder 2020), STUDY_NIPOCALIMAB_PHASE1 (Valenzuela 2025) and STUDY_FARLETUZUMAB_PHASE2 (Farrell 2012). Like the sulbactam-durlobactam pair this needs TWO indicators because the source stratifies across three phases; keeping both columns rather than an integer PHASE column keeps the reference stratum explicit. Subject-level (time-fixed). The underlying source is the pooled foreign dataset also used by Zhang 2019 (doi:10.1093/jac/dkz088) and the FDA XENLETA review (NDA 211672/211673), so the same two columns apply to any further extraction from that lineage.

STUDY_LEFAMULIN_PHASE2 (canonical for Bian 2024 lefamulin phase 2 study cohort indicator)

• Description: 1 = subject enrolled in the Phase 2 acute bacterial skin and skin structure infection study (n = 129) of the pooled lefamulin popPK model-building dataset; 0 = otherwise. Paired with STUDY_LEFAMULIN_PHASE1; both indicators 0 selects the Phase 3 CABP studies. Shifts clearance, distributional clearance to the first peripheral compartment, and the first peripheral volume.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 with STUDY_LEFAMULIN_PHASE1 also 0 (Phase 3 community-acquired bacterial pneumonia studies).
• Source aliases: derived per subject from the integer PHASE column of the pooled dataset (PHASE == 2 -> 1).
• Example models: Bian_2024_lefamulin_original_ppb.R, Bian_2024_lefamulin_higher_ppb.R (multiplies CL by 1.827 / 1.707, CLd1 by 1.44 / 1.192, and Vp1 by 1.985 / 1.28 in the original- and higher-plasma-protein-binding fits respectively).
• Notes: Specific scope; see STUDY_LEFAMULIN_PHASE1 for the full family rationale and the pair-of-indicators convention.

INJSITE_BUTTOCK (canonical for SC injection-site = buttock indicator)

• Description: 1 = subject’s SC dose injected into the buttock, 0 = abdomen (the universal SC reference site across the popPK literature) or a different non-buttock injection site. Per-dose-record OR per-subject covariate flagging the SC injection site when a population analysis estimates site-specific absorption parameters.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (abdomen).
• Source aliases: paper narrative “buttock” / “abdomen” injection-site labels driving site-specific absorption parameters.
• Example models: Glatard_2025_octreotide.R (fractional change -0.527 on the fast-release mean absorption time MAT_fast of the CAM2029 octreotide depot relative to abdominal injection, per Table 3 / Eq 11; supplied per dose record alongside INJSITE_THIGH, both 0 meaning abdomen).
• Notes: Specific scope because the buttock-vs-abdomen contrast is paper-specific. Third sibling of the INJSITE_<site> family alongside INJSITE_ARM (arm-vs-abdomen) and INJSITE_THIGH (thigh-vs-abdomen); the three indicators are mutually exclusive and all-zero denotes the abdomen reference. Distinct from ROUTE_IV (IV vs SC route, not within-SC site) and from DEVICE_AI (autoinjector vs prefilled syringe, device rather than anatomical site). In Glatard 2025 the buttock effect rests on only 1.5% of observations (from 5.1% of participants) and the authors explicitly advise caution in interpreting it – carry that caveat forward when reusing the estimate.

FORM_OCTREOTIDE_IR (canonical for immediate-release octreotide formulation indicator)

• Description: 1 = record belongs to immediate-release (IR) subcutaneous octreotide, 0 = the CAM2029 sustained-release subcutaneous octreotide depot (FluidCrystal injection depot).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (CAM2029 sustained-release depot).
• Source aliases: “octreotide IR” / “CAM2029” treatment-group labels in Glatard 2025.
• Example models: Glatard_2025_octreotide.R (selects between the two formulation-specific log-scale residual error terms of Table 3 – 0.393 for CAM2029 vs 0.204 for octreotide IR, each carrying its own exponential IIV).
• Notes: Member of the FORM_<drug>_<formulation> family. In the founding model this indicator does not route the dose – the dosing compartment does that (depot / depot2 for CAM2029’s two parallel release processes, depot3 for octreotide IR) – it selects the residual-error stratum. Time-varying within a participant: in trial HS-19-664 healthy volunteers received four octreotide IR doses, then after a washout of at least 6 days received CAM2029, so the same subject contributes records under both values.

PRIOR_OCTREOTIDE (canonical for prior/ongoing octreotide (or lanreotide) therapy at a pre-first-dose baseline record)

• Description: 1 = the record is a pre-first-study-dose baseline sample from a participant already receiving octreotide LAR or lanreotide autogel therapy, 0 = octreotide-naive participant, or any record at or after the first study dose.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (octreotide-naive, or any on-treatment record).
• Source aliases: “Pre-treatment” treatment-group label in Glatard 2025 Online Resource 2; “pre-treated participants with octreotide” in Sect. 2.3.1.
• Example models: Glatard_2025_octreotide.R (gates the estimated additive baseline rbase = 0.433 ng/mL, the authors’ “fudge factor” for the octreotide concentration present before the first CAM2029 dose in participants entering the phase 3 trials on stable somatostatin-receptor-ligand therapy).
• Notes: Member of the registered PRIOR_<drug or class> family (cf. PRIOR_STATIN, PRIOR_EZE, PRIOR_TNF, PRIOR_IPI, PRIOR_TAXANE), but unlike those per-subject indicators this one is supplied per record, because the parameter it gates describes a specific baseline observation rather than a subject-level exposure history. Online Resource 2 of the founding paper reports exactly 1.0 observation per participant in the “Pre-treatment” group (22 of 46 participants in HS-18-633; 49 of 95 in HS-19-647), i.e. the single pre-dose baseline sample. At such a record the model prediction arising from study dosing is identically zero, so adding the baseline to the prediction and substituting it for the prediction are equivalent; set the indicator to 0 for ordinary forward simulation. Distinct from CONMED_<INN> (a genuinely concomitant medication continued during the study) because the prior therapy is stopped when study treatment begins.

STUDY_TLV_PHASE2 (canonical for Van Wart 2025 telavancin phase 2 study cohort indicator)

• Description: 1 = subject enrolled in a Phase 2 study of the Van Wart 2025 pooled telavancin popPK analysis (the Phase 2 cSSSI studies and the Phase 2 study in patients with uncomplicated Staphylococcus aureus bacteremia); 0 = otherwise. Paired with STUDY_TLV_PHASE3; both indicators 0 selects the pooled Phase 1 and Phase 4 stratum (the reference). Used to switch the proportional (constant-coefficient-of-variation) residual-error magnitude between study phases; the additive residual component is shared across all phases.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 with STUDY_TLV_PHASE3 also 0 (the twelve Phase 1 studies in healthy subjects, the Phase 1 study in obese subjects, and the Phase 4 study TLV-2014-020 in otherwise healthy subjects with renal impairment).
• Source aliases: derived per subject from the study identifier and its phase (Van Wart 2025 Table S1 lists the phase of each of the 21 studies).
• Example models: VanWart_2025_telavancin.R (selects propSdPhase2 = sqrt(0.0169) = 0.130 for plasma observations).
• Notes: Specific scope because the contrast is tied to the Van Wart 2025 telavancin clinical-development pooled analysis. Drug-specific paper-anchored member of the STUDY_<DRUG>_PHASE<N> family alongside STUDY_SULDUR_PHASE2 / STUDY_SULDUR_PHASE3 (Cammarata 2024), STUDY_POSA_PHASE3 (van Iersel 2018), STUDY_ASP8232_PHASE2 (Snelder 2020), STUDY_NIPOCALIMAB_PHASE1 (Valenzuela 2025) and STUDY_FARLETUZUMAB_PHASE2 (Farrell 2012). Second member of the family needing a PAIR of indicators for a three-way phase stratification, after STUDY_SULDUR_PHASE2/3; keep both columns rather than an integer PHASE column so the reference stratum stays explicit. Note the wrinkle that distinguishes this pair from the Cammarata pair: here the reference stratum pools two phases (1 and 4), because Van Wart 2025 states “Phases 1 and 4 were combined into a single value” – so a Phase 4 subject sets both indicators to 0 and there is deliberately no STUDY_TLV_PHASE4 column. Subject-level (time-fixed); set once from the study identifier on each subject record.

STUDY_TLV_PHASE3 (canonical for Van Wart 2025 telavancin phase 3 study cohort indicator)

• Description: 1 = subject enrolled in a Phase 3 study of the Van Wart 2025 pooled telavancin popPK analysis (the Phase 3 cSSSI studies and the two Phase 3 studies in patients with hospital-acquired or ventilator-associated bacterial pneumonia); 0 = otherwise. Paired with STUDY_TLV_PHASE2; both indicators 0 selects the pooled Phase 1 and Phase 4 stratum.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 with STUDY_TLV_PHASE2 also 0 (pooled Phase 1 and Phase 4 studies).
• Source aliases: derived per subject from the study identifier and its phase (Van Wart 2025 Table S1).
• Example models: VanWart_2025_telavancin.R (selects propSdPhase3 = sqrt(0.0441) = 0.210 for plasma observations – the largest of the three phase magnitudes, consistent with the sparser and more variable sampling in the Phase 3 patient studies).
• Notes: Specific scope; see STUDY_TLV_PHASE2 for the full family rationale, the pair-of-indicators convention, and why the reference stratum pools Phases 1 and 4.

STUDY_AZTAVI_PHASE2 (canonical for Xie 2025 aztreonam-avibactam phase 2 residual-error stratum indicator)

• Description: 1 = the observation belongs to the Phase 2 stratum of the Xie 2025 pooled aztreonam-avibactam popPK analysis; 0 = otherwise. One of a trio (STUDY_AZTAVI_PHASE2, STUDY_AZTAVI_PHASE3, STUDY_AZTAVI_PHASE23); all three 0 selects the Phase 1 stratum, which is the reference. Used to switch the AZTREONAM proportional residual-error magnitude between study strata; avibactam residual variability is not stratified.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 with STUDY_AZTAVI_PHASE3 and STUDY_AZTAVI_PHASE23 also 0 (Phase 1 stratum).
• Source aliases: derived per subject / record from the study phase.
• Example models: Xie_2025_aztreonam_avibactam.R (selects propSdPhase2 = 0.224 for aztreonam observations; Xie 2025 Table S3 row Prop RSV_ATM phase 2 = 22.4%).
• Notes: Specific scope because the contrast is tied to the Xie 2025 aztreonam-avibactam pooled analysis. Member of the STUDY_<DRUG>_PHASE<N> family alongside STUDY_SULDUR_PHASE2 / STUDY_SULDUR_PHASE3 (Cammarata 2024), STUDY_POSA_PHASE3 (van Iersel 2018), STUDY_ASP8232_PHASE2 (Snelder 2020), STUDY_NIPOCALIMAB_PHASE1 (Valenzuela 2025) and STUDY_FARLETUZUMAB_PHASE2 (Farrell 2012). This is the first member needing a TRIO of indicators, because Xie 2025 Table S3 reports four aztreonam proportional magnitudes; keep the separate columns rather than an integer PHASE column so the reference stratum stays explicit.

STUDY_AZTAVI_PHASE3 (canonical for Xie 2025 aztreonam-avibactam phase 3 residual-error stratum indicator)

• Description: 1 = the observation belongs to the Phase 3 stratum of the Xie 2025 pooled aztreonam-avibactam popPK analysis (studies REVISIT / NCT03329092 and ASSEMBLE / NCT03580044); 0 = otherwise. Paired with STUDY_AZTAVI_PHASE2 and STUDY_AZTAVI_PHASE23.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 with the other two phase indicators also 0 (Phase 1 stratum).
• Source aliases: derived per subject / record from the study phase.
• Example models: Xie_2025_aztreonam_avibactam.R (selects propSdPhase3 = 0.403 for aztreonam observations; Xie 2025 Table S3 row Prop RSV_ATM phase 3 = 40.3%).
• Notes: Specific scope; see STUDY_AZTAVI_PHASE2 for the full family rationale.

STUDY_AZTAVI_PHASE23 (canonical for Xie 2025 aztreonam-avibactam pooled phase-2/3 residual-error stratum indicator)

• Description: 1 = the observation belongs to the pooled phase-2/3 stratum of the Xie 2025 aztreonam residual-error model; 0 = otherwise. Paired with STUDY_AZTAVI_PHASE2 and STUDY_AZTAVI_PHASE3.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 with the other two phase indicators also 0 (Phase 1 stratum).
• Source aliases: Prop RSV_ATM phase 2/3 – the Xie 2025 Table S3 row label, which is the only place the stratum is named.
• Example models: Xie_2025_aztreonam_avibactam.R (selects propSdPhase23 = 0.533; Xie 2025 Table S3 row Prop RSV_ATM phase 2/3 = 53.3%).
• Notes: Specific scope; see STUDY_AZTAVI_PHASE2 for the family rationale. Stratum membership is not defined by any available source. Xie 2025 Table S3 reports this fourth aztreonam proportional magnitude alongside separate phase 2 (theta21) and phase 3 (theta22) magnitudes, and neither the paper, its supplement, nor the predecessor Das 2024 says which records make up the pooled phase-2/3 stratum as distinct from the other two. The column is registered so the published parameter set can be carried complete rather than silently dropping theta23; assigning records to it is left to the user, and the founding model’s vignette records the gap in its Errata. Because this selects only a residual-error magnitude, it affects the simulated observation and not the typical-value or individual-prediction profile.

STUDY_CAZAVI_PHASE2 (canonical for Li 2019 ceftazidime-avibactam phase 2 residual-error stratum indicator)

• Description: 1 = the observation belongs to the Phase 2 stratum of the Li 2019 pooled ceftazidime-avibactam popPK analysis (the Phase 2 cIAI and cUTI studies); 0 = otherwise. Paired with STUDY_CAZAVI_PHASE3; both 0 selects the Phase 1 stratum, which is the reference and the only stratum carrying an additive residual component for either analyte.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 with STUDY_CAZAVI_PHASE3 also 0 (Phase 1 stratum).
• Source aliases:
  • PHASE = 2 – study-phase level in the Li 2019 NONMEM control streams (supplementary Data S1); the avibactam stream sets its PH2 flag from the equivalent test STDY.EQ.2001.OR.STDY.EQ.2002.
• Example models: Li_2019_avibactam.R (selects propSdPhase2 = 0.492; Li 2019 Table 2 row theta19: Proportional variability, phase II), Li_2019_ceftazidime.R (pooled with Phase 3 onto a single propSdPhase23 = 0.338 / addSdPhase23; Li 2019 Table 1 rows Proportional error, phase II and phase III). Franzese_2022_ceftazidime.R / Franzese_2022_avibactam.R (pediatric + adult refit; this indicator stays restricted to the two ADULT phase 2 studies, STDY = 2001/2002. The pediatric phase 2 studies are carried by STUDY_CAZAVI_PED_PHASE2, because the two Franzese analytes stratify them differently – ceftazidime pools them into its phase II/III residual stratum, avibactam leaves them in its phase I stratum).
• Notes: Specific scope because the contrast is tied to the Li 2019 ceftazidime-avibactam pooled analysis. Member of the STUDY_<DRUG>_PHASE<N> family alongside STUDY_AZTAVI_PHASE2 / STUDY_AZTAVI_PHASE3 / STUDY_AZTAVI_PHASE23 (Xie 2025, the successor analysis in the same development programme), STUDY_SULDUR_PHASE2 / STUDY_SULDUR_PHASE3 (Cammarata 2024) and STUDY_POSA_PHASE3 (van Iersel 2018). The two Li 2019 analytes stratify differently – ceftazidime pools Phase 2 with Phase 3 into one residual stratum while avibactam keeps all three apart – so the two indicators are registered as a pair and the ceftazidime model consumes only their sum. Keep the columns separate rather than collapsing them to a single phase-2/3 flag, so one simulated covariate set can drive both companion models.

STUDY_CAZAVI_PHASE3 (canonical for Li 2019 ceftazidime-avibactam phase 3 residual-error stratum indicator)

• Description: 1 = the observation belongs to the Phase 3 stratum of the Li 2019 pooled ceftazidime-avibactam popPK analysis (RECLAIM 1/2, RECLAIM 3, RECAPTURE 1/2, REPRISE and REPROVE); 0 = otherwise. Paired with STUDY_CAZAVI_PHASE2.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 with STUDY_CAZAVI_PHASE2 also 0 (Phase 1 stratum).
• Source aliases:
  • PHASE = 3 – study-phase level in the Li 2019 NONMEM control streams (supplementary Data S1).
• Example models: Li_2019_avibactam.R (selects propSdPhase3 = 0.363; Li 2019 Table 2 row theta20: Proportional variability, phase III), Li_2019_ceftazidime.R (pooled with Phase 2; see STUDY_CAZAVI_PHASE2). Franzese_2022_ceftazidime.R / Franzese_2022_avibactam.R (pediatric + adult refit; same semantics).
• Notes: Specific scope; see STUDY_CAZAVI_PHASE2 for the full family rationale and for why the pair is not collapsed into one flag.

STUDY_CAZAVI_PED_PHASE2 (canonical for the Franzese 2022 ceftazidime-avibactam pediatric phase 2 study indicator)

• Description: 1 = the record belongs to one of the two pediatric phase 2 studies of the Franzese 2022 pooled ceftazidime-avibactam analysis (cIAI, NCT02475733, and cUTI, NCT02497781; multiple-dose, children aged 3 months to < 18 years); 0 = any other study, including the single-dose pediatric phase 1 study NCT01893346 and every adult study.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (any other study).
• Source aliases:
  • PHASE = 2 with STDY = 15 or 16 – in the Franzese 2022 final control streams (Supplementary Methods) the pediatric phase 2 studies carry PHASE = 2, which the ceftazidime $ERROR block tests directly (IF(PHASE.EQ.2.OR.PHASE.EQ.3)), while the avibactam $ERROR block tests the adult study numbers (IF (STDY.EQ.2001.OR.STDY.EQ.2002) PH2=1). The study numbers are inferred from the control streams’ explicit STDY.EQ.15 pediatric-cIAI branch and their PEDFLG set STDY = 14, 15, 16.
• Example models: Franzese_2022_ceftazidime.R (selects the pooled phase II/III residual stratum, propSdPhase23 = 0.374 / addSdPhase23 = 2.56 mg/L, Table S2 theta10 / theta11). Not referenced by Franzese_2022_avibactam.R, whose phase I residual stratum (combined error, Table S3 theta17 / theta18) covers these records.
• Notes: Specific scope; member of the STUDY_<DRUG>_PHASE<N> family alongside STUDY_CAZAVI_PHASE2 / STUDY_CAZAVI_PHASE3. Kept separate from STUDY_CAZAVI_PHASE2 so a single covariate data set can drive both Franzese 2022 analytes, which place these records in different residual strata. Purely a residual-error switch: it carries no structural effect (the pediatric study-level thetas CLPED, V1PED, CLPP-STUDY15 and V1POP-STUDY15 are all held at 0 in the final avibactam control stream).

STUDY_SIPO_PHASE2 (canonical for Stringer 2013 sipoglitazar phase II patient-study residual-error stratum indicator)

• Description: 1 = the observation belongs to the pooled phase II type-2-diabetes patient studies (EC201 and EC202) of the Stringer 2013 sipoglitazar analysis; 0 = the observation belongs to the phase I healthy-volunteer study (006), which is the reference stratum. Used to switch the proportional residual-error magnitude between the two populations and to switch on the inter-individual variability that Stringer 2013 places on the phase II residual error.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (phase I healthy-volunteer study 006, dense single-dose sampling).
• Source aliases: derived per record from the study identifier (006 -> 0; EC201 / EC202 -> 1).
• Example models: Stringer_2013_sipoglitazar.R, Stringer_2013_sipoglitazar_mixture.R (select propSdPhase2 = sqrt(0.167) = 0.409 with IIV 76.9% CV instead of propSdPhase1 = sqrt(0.0552) = 0.235; Stringer 2013 Table 2 and supplementary Table A1).
• Notes: Specific scope because the contrast is tied to the Stringer 2013 pooled analysis. Member of the STUDY_<DRUG>_PHASE<N> family alongside STUDY_AZTAVI_PHASE2 / STUDY_AZTAVI_PHASE3 (Xie 2025), STUDY_SULDUR_PHASE2 / STUDY_SULDUR_PHASE3 (Cammarata 2024), STUDY_LEFAMULIN_PHASE1 / STUDY_LEFAMULIN_PHASE2 (Bian 2024) and STUDY_NIPOCALIMAB_PHASE1 (Valenzuela 2025). Only one indicator is needed here because the analysis has exactly two residual-error strata. Stringer 2013 explains the split in its supplementary methods: the phase II samples were all intended as troughs but the actual time after dose was never recorded, so the phase II residual carries both assay error and an unrecorded-time component, and an IIV term on that residual dropped the objective function by 508 points and the residual error from 131% to 24%.

STUDY_UPA_PHASE1 (canonical for Klunder 2019 upadacitinib phase I study-stratum indicator)

• Description: 1 = the record belongs to one of the four phase I studies of the Klunder 2019 pooled upadacitinib analysis (healthy volunteers plus 14 subjects with mild to moderate rheumatoid arthritis, dense serial sampling); 0 = the record belongs to one of the eight phase II, phase IIb/III or phase III rheumatoid-arthritis studies, which are the reference stratum. Switches BOTH the interindividual-variability magnitudes on CL/F and Vc/F AND the proportional residual-error magnitude between the two strata. It does not change any typical value.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the pooled phase II / IIb-III / III patient studies, which supply 3982 of the 4170 subjects and are therefore the bulk stratum).
• Source aliases:
  • Ph23flag – the NONMEM control-stream variable, built as a chain of IF(STDY.EQ.<n>) Ph23flag=1 tests over the nine phase II/III study numbers (13537, 13550, 13542, 13549, 14663, 15555, 14465, 13545, 16098). The canonical is the COMPLEMENT of this flag: STUDY_UPA_PHASE1 = 1 - Ph23flag. The control stream is published as Appendix S1 of the successor analysis Bhatnagar 2024 (Clin Transl Sci 17:e13733), which inherits the Klunder 2019 model.
  • STDY – the raw study-number column the flag is derived from.
• Example models: Klunder_2019_upadacitinib.R (selects etalcl_ph1 (ISV 20.5%) over etalcl_ph23 (36.5%), etalvc_ph1 (24.4%) over etalvc_ph23 (53.0%), and propSdPhase1 = 0.344 over propSdPhase23 = 0.543; Klunder 2019 Table 3).
• Notes: Specific scope because the stratum boundary is tied to the Klunder 2019 pooled upadacitinib analysis. Member of the STUDY_<DRUG>_PHASE<N> family alongside STUDY_SIPO_PHASE2 (Stringer 2013), STUDY_AZTAVI_PHASE2 / STUDY_AZTAVI_PHASE3 (Xie 2025), STUDY_SULDUR_PHASE2 / STUDY_SULDUR_PHASE3 (Cammarata 2024), STUDY_LEFAMULIN_PHASE1 / STUDY_LEFAMULIN_PHASE2 (Bian 2024) and STUDY_NIPOCALIMAB_PHASE1 (Valenzuela 2025). Like STUDY_SIPO_PHASE2 it needs only one indicator because the analysis has exactly two strata, and like it the indicator drives both a variance and a residual magnitude; unlike it, the phase I level rather than the patient level is the flagged one, so the polarity is reversed relative to that entry. Klunder 2019 Results quantifies both splits: separating the ISV cut the objective function by 152 points and separating the proportional residual error cut it by a further 1108 points. The paper attributes the larger phase II/III variability to the wider inclusion criteria and the sparse, less-controlled sampling of the later trials. Distinct from a disease-state covariate: phase I is overwhelmingly but not exclusively healthy (14 of the 188 phase I subjects had RA), and the RA-vs-healthy contrast is carried separately by DIS_RA on CL/F, so the two columns are not interchangeable and a model needs both.

STUDY_CHRONOS3 (canonical for the CHRONOS-3 (Bayer study 17067) copanlisib trial cohort indicator)

• Description: 1 = subject enrolled in CHRONOS-3, the randomized double-blind phase III trial of copanlisib plus rituximab in relapsed indolent non-Hodgkin lymphoma (Bayer study 17067, NCT02367040); 0 = subject enrolled in any of the eight phase I / phase II copanlisib studies pooled alongside it (12871, 15205, 16270, 16349 parts A and B / CHRONOS-1, 16790, 16866, 17792, 18041). Carries two distinct roles in the founding model: a structural clearance shift, and selection of the phase III residual-error magnitude.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (any of the eight pooled phase I / phase II copanlisib studies).
• Source aliases:
  • Study 17067 – the sponsor study number, used as the row label in Morcos 2023 Table 1 and as the theta subscript in Table 2 (theta_17067CL).
  • CHRONOS-3 – the trial acronym, used throughout the Morcos 2023 narrative for the same cohort.
• Example models: Morcos_2023_copanlisib.R (fractional shift of -0.184 on clearance, i.e. 18.4% lower CL, per Morcos 2023 Table 2 theta_17067CL; and selection of the phase III log-scale residual SD expSd_phase3 = sqrt(0.632) = 0.7950 (CV 93.8%) over the pooled phase I/II expSd_phase12 = sqrt(0.176) = 0.4195 (CV 43.9%) for samples drawn more than 20 min after the start of an infusion).
• Notes: Specific scope because the contrast is one named trial within one development program. Sibling of the STUDY_<DRUG>_PHASE<N> residual-error-stratum family (STUDY_AZTAVI_PHASE2 / STUDY_AZTAVI_PHASE3, STUDY_SULDUR_PHASE2 / STUDY_SULDUR_PHASE3, STUDY_SIPO_PHASE2), but named for the trial rather than the phase because the same indicator also carries a structural clearance effect and because CHRONOS-3 is the only phase III study in the pool. Confounding caveat that must travel with any reuse: rituximab was co-administered only in study 17067, so Morcos 2023 Table S2 records that a rituximab-comedication covariate is entirely confounded with this study indicator, and the Discussion declines to attribute the 18.4% clearance reduction to rituximab (“this could not be robustly assigned to co-administration with rituximab due to differences in study design and patient population compared with the other studies”). Do not re-label this column as a rituximab-comedication effect.

STUDY_CIAI_PH2 (canonical for the phase 2 cIAI study cohort indicator in the ceftazidime-avibactam / aztreonam-avibactam model lineage)

• Description: 1 = subject enrolled in the Phase 2 complicated-intra-abdominal-infection study of the ceftazidime-avibactam development program (labelled “Study2002” in Xie 2025 Table S3); 0 = otherwise. A study-cohort effect that is retained across successive models in this lineage and stacks ON TOP OF the infection-type covariate DIS_CIAI for those subjects rather than replacing it.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (any other study).
• Source aliases:
  • Study2002 – printed row label in Xie_2025_aztreonam_avibactam.R (Table S3 rows Study2002 on CL_AVI, Study2002 on Vc_AVI).
  • Population effect on CL / Vc (cIAI, phase II) – the corresponding row labels in Das 2024 supplementary Table 5, which is where the proportional X*(1 + theta) form comes from.
• Example models: Xie_2025_aztreonam_avibactam.R (proportional shifts of +0.89 on avibactam CL and +1.64 on avibactam Vc; aztreonam carries no term, and the indicator is 0 for every aztreonam-avibactam subject and for the whole simulated population of the paper), Li_2019_avibactam.R (the FOUNDING paper of this lineage and the source of the row labels quoted above: proportional shifts of +1.92 on Vc, Table 2 theta9, and +0.406 on CL, Table 2 theta10. Its semantics differ from Xie 2025’s: in Li 2019 the Phase 2 cIAI volume shift REPLACES the Phase 3 cIAI shift rather than stacking on it, so a model that adds both double-counts the cohort. The clearance shift has no Phase 3 counterpart and so is a plain addition in both papers). Franzese_2022_avibactam.R (pediatric + adult refit of Li_2019_avibactam.R, same REPLACE semantics on Vc: +2.17 on Vc, Table S3 theta9, and +0.431 on CL, theta10. The indicator is 0 for the PEDIATRIC phase 2 cIAI study, which takes the phase 3 cIAI Vc shift theta12 and no CL shift – control stream IF(STDY.EQ.15) V1POP=(1 + THETA(12))).
• Notes: Specific scope because the cohort is a single named trial in one development program. A subject in this cohort therefore receives both shifts.

STUDY_TACRO_FRANCKE (canonical for Francke 2025 tacrolimus transplant-centre x bioanalytical-method residual-error stratum)

• Description: Integer-valued (1-7) per-observation identifier of the combination of transplant centre and tacrolimus measurement method to which a whole-blood concentration record belongs, in the Francke 2025 international multicentre tacrolimus population PK analysis. The seven strata pair the four contributing centres (Erasmus MC Rotterdam, LUMC Leiden, Bellvitge University Hospital Barcelona, Cliniques Universitaires St Luc Brussels) with the two bioanalytical methods (immunoassay, LC-MS/MS). Record-level, not subject-level: a single recipient can contribute records under more than one code if their centre changed analytical method during follow-up.
• Units: (integer 1-7)
• Type: categorical
• Scope: specific
• Reference category: None. Every observation belongs to exactly one stratum, and each stratum selects one of five estimated residual-error magnitudes; no stratum is a typical-value reference.
• Source aliases:
  • DATA – the column name used in the Francke 2025 Supplementary Data S3 $ERROR block (IF (DATA.EQ.1.OR.DATA.EQ.2) Y=IPRED + ERR1*EPS(1), etc.).
  • CENTER – the column name listed in the same control stream’s $INPUT record. The published stream does not reconcile the two names; they refer to the same stratification.
• Example models: Francke_2025_tacrolimus.R, Francke_2025_tacrolimus_startingdose.R (both select the log-scale residual SD per observation across five levels: codes 1-2 -> Rotterdam immunoassays, codes 3-4 -> Rotterdam and Leiden LC-MS/MS, code 5 -> Barcelona immunoassay, code 6 -> Barcelona LC-MS/MS, code 7 -> Brussels immunoassay).
• Notes: Follows the STUDY_VORI precedent (a single integer column rather than a family of paired binary indicators) because the stratification has seven levels and the model file can derive (STUDY_TACRO_FRANCKE == n) indicators inline. Francke 2025 Section 2.2.1 states that seven centre-by-method combinations were formed and that “residual errors were combined where possible based on analytical method and fit of the model”, which is why the seven codes collapse onto five estimated magnitudes. The source paper does not state what distinguishes code 1 from code 2, or code 3 from code 4; the most plausible reading is two immunoassay platforms within Rotterdam and the Rotterdam-versus-Leiden split of a shared LC-MS/MS magnitude. Because each member of those pairs selects the same estimated SD, the ambiguity has no effect on model predictions and a user may code either member. For a new single-centre LC-MS/MS dataset, code 3 (or 4) is the closest analogue. Distinct from the general IMMUNOASSAY canonical, which captures only the bioanalytical method and cannot express the centre dimension that Francke 2025 stratifies jointly on. Auto-approved member of the STUDY_<id> canonical family.

DOSE_RIMEGEPANT_MG (canonical for administered rimegepant per-administration dose amount)

• Description: Continuous per-dose-record covariate carrying the administered rimegepant dose amount in mg. Drives the nonlinear relative bioavailability of oral rimegepant, which increases greater-than-dose-proportionally over the studied 10 to 150 mg range – attributed to dose-dependent autoinhibition of CYP3A-mediated first-pass metabolism, rimegepant being a weak in-vitro CYP3A4 inhibitor.
• Units: mg
• Type: continuous
• Reference category: n/a – used with power scaling (DOSE_RIMEGEPANT_MG / 10)^exponent; the 10 mg reference is the lowest studied dose level.
• Source aliases:
  • Dose – used in Comisar_2025_rimegepant.R (Comisar 2025 Table 2 row label ‘Dose effect on F 1’).
• Example models: Comisar_2025_rimegepant.R (founding example; fdepot <- 1 * (DOSE_RIMEGEPANT_MG / 10)^0.192 * (1 + (-0.331) * FED). The power form and the 10 mg reference are printed verbatim on the structural schematic of the upstream adult-only model, Comisar 2025 CPT:PSP Figure 2, as F1 = (Dose/10)^0.191; relative bioavailability at the marketed 75 mg dose is therefore 1.47-fold that at 10 mg. Paired with the separate binary DOSE_LOW indicator, which carries the step change in the transit rate constant at the 10 and 25 mg levels).
• Notes: Auto-approved member of the DOSE_<drug>_<units> family. Use this rather than the bare [[DOSE]] canonical: rxode2’s event-table translator etTrans consumes a covariate column literally named DOSE (in any casing) and never exposes it to model, so a model reading the bare name fails at solve time with The following parameter(s) are required for solving: DOSE. The drug-suffixed name sidesteps that entirely. Distinct from [[DOSE_LOW]] and the DOSE_<N>MG indicator family, which are binary dose-level flags rather than the numeric dose amount.

USG_CORRECTED (canonical for urine-specific-gravity-corrected concentration indicator)

• Description: 1 = the urine concentration on this record was corrected for urine specific gravity (USG) to normalise for the subject’s hydration status at the time of collection, 0 = no USG value was recorded for the sample so no correction was applied. A per-record sample-handling flag, not a subject-level attribute: the same subject can contribute both corrected and uncorrected records.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (concentration not corrected for specific gravity).
• Source aliases:
  • USG – Thoueille 2026 Data S1 $INPUT column, branched in $PK as IF (USG.EQ.0)... IF (USG.EQ.1)....
• Example models: Thoueille_2026_salmeterol.R (gates the inter-individual variability on the estimated urine-production rate: urprod <- exp(lurprod + etalurprod * (1 - USG_CORRECTED)), reproducing the source stream’s two-branch $PK in which ETA(9) (variance 0.429, 73% CV) carries the uncorrected branch and ETA(10) is fixed to 0 for the corrected branch. Correcting for USG removes the hydration-status component of urine-production variability, so no residual IIV was supported once the correction had been applied. The paper’s own Figure 3 / Table 3 / Table S1 simulations are USG-corrected, i.e. USG_CORRECTED = 1).
• Notes: The alternative was to carry only the uncorrected branch’s eta and relegate the corrected branch to prose in the model file and vignette Errata; that was rejected because it would absorb a known, recorded experimental distinction into unexplained variability and would silently mis-simulate the paper’s own headline figures. The correction itself is defined by the WADA technical document as corrected = (1.020 - 1) / (USG + 0.002 - 1) x observed, where 1.020 is the specific-gravity threshold and 0.002 accounts for measurement uncertainty; the formula is a data transform applied before modelling and is not itself parameterised, so only the applied / not-applied indicator is carried as a covariate. General scope because USG correction is a standard anti-doping and occupational-toxicology urine-normalisation procedure with a stable meaning independent of any one paper. A future model that needs the raw specific gravity as a continuous predictor rather than the applied / not-applied flag should register a separate continuous USG canonical; do not overload this binary.

STUDY_SALM (canonical for source-study identifier in the Thoueille 2026 pooled salmeterol analysis)

• Description: Source-study identifier in the six-study pooled inhaled-salmeterol doping-control analysis of Thoueille 2026. Values in the final model: 1 = Jacobson & Hostrup 2022, 2 = Jessen 2021, 4 = Jacobson 2017 (enantioselective), 5 = Hostrup 2012, 6 = Deventer 2011, 7 = Petrou et al. (unpublished). Value 3 (Bozzolino 2019) was excluded from the analysis for unreliable dose and collection timing.
• Units: (categorical)
• Type: categorical
• Scope: specific
• Reference category: n/a – the indicator selects a residual-error magnitude tier only; it does not shift any structural parameter.
• Source aliases:
  • SID – Thoueille 2026 Data S1 $INPUT column, branched throughout $ERROR.
• Example models: Thoueille_2026_salmeterol.R (selects the urine residual-error magnitude tier that Table 2 regroups as low / mid / high: parent urine uses the low tier for SID 1 / 5 / 6, the mid tier for 2 / 7 and the high tier for 4; metabolite urine uses the mid tier for 1 / 5 / 7 and the high tier for 2. The two plasma residual errors are common to all contributing studies).
• Notes: Well-formed member of the auto-approved STUDY_<id> covariate family, so no separate naming ruling was required; registered here for the same reason as its siblings STUDY_VORI, STUDY_TACRO_FRANCKE and STUDY_SALEM – the semantics are tied to one paper’s pooled-study set, hence scope specific. Note that the numbering is the source dataset’s own SID coding and is not contiguous: 3 is absent from the final model because that study was excluded. Set STUDY_SALM to any of 1 / 5 / 6 to simulate with the lowest reported urine residual magnitudes, or to 2 / 7 for the mid tier. As with STUDY_VORI, the indicator is consumed only inside the residual-error expression, so it must still be supplied to rxSolve even for a typical-value simulation.

CONMED_ENOXAPARIN_DOSE, CONMED_ARDEPARIN_DOSE, CONMED_BEMIPARIN_DOSE, CONMED_DALTEPARIN_DOSE, CONMED_NADROPARIN_DOSE, CONMED_REVIPARIN_DOSE, CONMED_TINZAPARIN_DOSE, CONMED_APIXABAN_DOSE, CONMED_BETRIXABAN_DOSE, CONMED_EDOXABAN_DOSE, CONMED_RAZAXABAN_DOSE, CONMED_RIVAROXABAN_DOSE, CONMED_LY517717_DOSE, CONMED_PD0348292_DOSE, CONMED_YM150_DOSE, CONMED_AVE5026_DOSE, CONMED_FONDAPARINUX_DOSE, CONMED_DABIGATRAN_DOSE, CONMED_XIMELAGATRAN_DOSE, CONMED_DESIRUDIN_DOSE, CONMED_SR123781A_DOSE, CONMED_HEPARIN_DOSE, CONMED_WARFARIN_DOSE (canonical for per-arm total daily dose of a named anticoagulant in a VTE-prophylaxis dose-response meta-analysis)

• Description: Per-study-arm TOTAL DAILY dose of the named anticoagulant. 0 = that drug is not the agent given in this arm; an arm in which every column of the family is 0 is a placebo arm. Suffix is the INN (or the development code where the agent never received an INN: LY517717, PD0348292, YM150, AVE5026, SR123781A). Members of the CONMED_<drug>_DOSE family; each column drives its own drug’s dose-response term in a model-based meta-analysis, exactly as the statin members of that family do.
• Units: Per drug, and NOT interchangeable. mg/day for enoxaparin, apixaban, betrixaban, edoxaban, razaxaban, rivaroxaban, LY517717, PD0348292, YM150, AVE5026, fondaparinux, dabigatran, ximelagatran, desirudin and SR123781A; IU/kg/day (weight-normalised anti-Xa activity) for ardeparin, nadroparin and tinzaparin; kIU/day (thousands of anti-Xa IU) for bemiparin, dalteparin, reviparin and unfractionated heparin; and achieved INR for warfarin. The unit of each column must match the unit of the matching ED50 parameter in the model. Record the per-column unit in covariateData[[<name>]]$units.
• Type: continuous
• Scope: specific
• Reference category: 0 (this drug was not the agent given in this arm, so it contributes nothing to the dose-response). All columns 0 = placebo arm.
• Source aliases: “Dose” / “Daily dose range” in per-drug meta-analysis summary tables.
• Example models: Mandema_2011_anticoagulants_mbma.R (each column drives its drug’s sigmoid-Emax efficacy term and linear bleeding term through the drug’s own ED50; per-drug daily-dose ranges are in Mandema 2011 Table 1).
• Notes: CONMED_WARFARIN_DOSE carries an achieved INR rather than a milligram dose because warfarin is titrated to an INR target and the source models it that way (Mandema 2011 Table 1 gives the warfarin “daily dose range” as 2.2-2.5 INR and the Supplementary Data gives “ED50 Warfarin (INR)”); it is distinct from INR_BASE, which is a subject-level PRE-medication baseline INR used as a PK/PD covariate. CONMED_HEPARIN_DOSE is a per-arm total DAILY prophylactic dose in kIU/day and is distinct from DOSE_UFH_UH, a concomitant continuous-infusion heparin dose RATE in U/h. CONMED_RIVAROXABAN_DOSE is a per-arm total daily mass dose and is distinct from DOSE_RIV_MGKG, a per-administration weight-normalised paediatric dose. Certoparin appears in the Mandema 2011 Methods search list but no certoparin trial met the inclusion criteria, so no CONMED_CERTOPARIN_DOSE is registered. Regimen variants that the source treated as a separate potency are carried by a sibling FORM_<drug>_<regimen> indicator rather than by a second dose column (see FORM_XIMELAGATRAN_PRESURG).

FORM_XIMELAGATRAN_PRESURG (canonical for ximelagatran pre-surgery melagatran regimen indicator)

• Description: 1 = the ximelagatran arm gave ximelagatran in combination with subcutaneous melagatran STARTED BEFORE surgery; 0 = ximelagatran administered alone (the model reference).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (ximelagatran administered alone).
• Source aliases: “Ximelagatran started before surgery” as a separate drug row in a meta-analysis parameter table.
• Example models: Mandema_2011_anticoagulants_mbma.R (multiplies the ximelagatran ED50 by 0.365 when set to 1, i.e. the pre-surgery melagatran combination is about 2.7-fold more potent; Mandema 2011 Supplementary Data row “ED50 Ximelagatran started before surgery / ED50 Ximelagatran”).
• Notes: Study-arm-level regimen indicator, not a subject-level covariate, and inert when CONMED_XIMELAGATRAN_DOSE is 0. Sibling of FORM_FLV_BID_XR and FORM_LOV_BID_XR under the FORM_<drug>_<regimen> family pattern: like those, it encodes a regimen that the source modelled as an ED50 RATIO on an existing drug rather than as a separate dose column. Mandema 2011 justifies the split as “a statistically significant difference in potency between these two treatments (P < 0.001)”.

STUDY_DE (canonical for the Riggs 2013 empagliflozin Study-D / Study-E cohort indicator)

• Description: 1 = subject’s observation comes from Riggs 2013 Study D (NCT00789035, 12-week Phase II) or Study E (NCT00749190, 12-week metformin add-on Phase II); 0 = Study A (EudraCT 2007-000654-32, 8-day), Study B (NCT00558571, 4-week) or Study C (NCT00885118, 4-week, Japanese patients). Selects which of two estimated residual-error pairs applies to the record.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (Studies A, B and C – the shorter early-phase studies, which had the lower residual variability).
• Source aliases:
  • STUDY – Riggs 2013 Methods (“Separate residual variance terms were included during model development to account for the greater variability in the empagliflozin concentrations observed from Studies D and E compared with Studies A, B, and C”) and the Table 2 residual-variance block, which is split into “Studies A, B, and C” and “Studies D and E” sub-rows. Used in Riggs_2013_empagliflozin.R.
• Example models: Riggs_2013_empagliflozin.R (switches the combined residual error only: proportional 19.6% CV + additive 0.179 nmol/L for Studies A/B/C versus proportional 35.7% CV + additive 0.010 nmol/L for Studies D/E; no structural or disposition parameter depends on it).
• Notes: Specific scope; tied to the five-trial Boehringer Ingelheim empagliflozin Phase I/II programme pooled by Riggs 2013. Supply per observation record rather than per subject in principle, though in this analysis every subject contributes to exactly one study so it is effectively subject-level. Unusually for the STUDY_<id> family this indicator carries no fixed-effect meaning – it exists purely to stratify the residual error, which is why the model computes the proportional and additive SDs as propSd_studyabc * (1 - STUDY_DE) + propSd_studyde * STUDY_DE and passes those model-computed variables into prop() / add(). Directly analogous to STUDY_DORZA_EARLY, which likewise uses one binary indicator to name a group of studies rather than a single trial. Follows the auto-approved STUDY_<id> canonical family.

TRT_T2DM_NAIVE, TRT_T2DM_NONNAIVE, TRT_T2DM_ADDON, TRT_T2DM_MIXED (canonical for the four-level type-2-diabetes treatment-history stratum of a trial arm)

• Description: Four mutually exclusive binary indicators recording the antihyperglycemic treatment history of a type-2-diabetes study arm. Exactly one is 1 on any record. The levels are those defined in Yao 2023 Model development: TRT_T2DM_NAIVE = patients naive to oral hypoglycemic agents; TRT_T2DM_NONNAIVE = patients who had taken hypoglycemic agents but had undergone a washout period of more than 2 weeks; TRT_T2DM_ADDON = patients receiving the study drug on top of ongoing antihyperglycemic combination treatment; TRT_T2DM_MIXED = arms that pooled naive and non-naive patients or did not report treatment history. The stratum selects the magnitude of the placebo response on FPG and on HbA1c, which differ materially between them: FPG rises under placebo in every stratum except add-on, where concomitant agents pull it down.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: none – this is a set of four indicators with no reference level, because the source estimates a separate placebo maximum in each stratum rather than a reference plus three offsets. A model consuming them multiplies each stratum-specific parameter by its indicator and sums.
• Source aliases:
  • treatment type (categorical column with levels naive, non-naive, add-on, mixed) – Yao 2023 Model development and Table 2 rows Pfmax1-4 / Phmax1-4. Decompose into the four indicators: TRT_T2DM_NAIVE = as.integer(treatment_type == "naive"), and so on.
• Example models: Yao_2023_sglt2_endpoints_mbma.R (selects Pfmax1-4 on the FPG placebo response and Phmax1-4 on the HbA1c placebo response; Yao 2023 reports “Treatment type is not significant on other parameters here”, so no other parameter is stratified), Sato_2024_sglt2_hba1c_mbma.R (uses only TRT_T2DM_NAIVE and TRT_T2DM_ADDON, which is sufficient because Sato 2024 resolves the treatment history of every one of its 295 arms and so never needs the TRT_T2DM_MIXED level: naive is (NAIVE 1, ADDON 0), non-naive is (0, 0) and add-on is (0, 1), splitting 51 / 53 / 191 with no co-occurrence. The two levels act on different parameters – TRT_T2DM_NAIVE multiplies Emax by 1.230 while TRT_T2DM_ADDON shifts the placebo term by -0.156 – rather than jointly selecting one stratum-specific parameter as in Yao 2023).
• Notes: General scope because the four-level naive / non-naive / add-on / mixed stratification is the standard framing of prior antihyperglycemic therapy in T2DM efficacy trials, not a Yao-2023-specific construct. The set is deliberately NOT decomposed into PRIOR_* and CONMED_* indicators: naive, non-naive and add-on map onto (no prior therapy), (prior therapy, washed out) and (concomitant therapy), but TRT_T2DM_MIXED is a meta-analysis artefact with no patient-level referent – it records that the published arm does not resolve treatment history – and so cannot be reconstructed from prior- and concomitant-therapy flags. Because that level exists only at the arm level, the whole set is arm-level in an MBMA even though the first three levels have individual-level meaning. Distinct from LINE_1L, whose entry warns against overloading it for finer resolution than 1L vs 2L+; distinct from TRT_<INN> (TRT_EPHEDRINE, TRT_CAFEDRINE_THEODRENALINE), which name the drug a subject received rather than the therapeutic context it was added to; and distinct from CONMED_METFORMIN, which names one specific background agent rather than the presence of any. The T2DM token is part of the name because “naive” and “add-on” mean materially different things in other therapeutic areas; a future oncology or immunology paper needing an analogous stratum should register its own disease-qualified family rather than reuse this one. Family, four-level structure and the T2DM qualifier ratified by the maintainers (PMC10088079, 2026-09-05), whose ruling recorded TRT_ as an existing three-entry family and this set as a family-conforming extension: “treatment-history strata are disease-specific (naive/add-on mean different things in oncology or HIV), and dropping it would create a collision the moment a second indication needs the same axis.”

FED_MISSING (canonical for prandial-state-not-recorded indicator)

• Description: 1 = the prandial state in force at this dose record was not captured by the study, 0 = the prandial state was recorded (fasted, fed, or any of the meal-composition levels). Completes the FED_* family for pooled analyses in which a substantial share of dose records carry no food-status information and the analyst estimates the unrecorded stratum as its own level rather than dropping the records or imputing them into an observed level. The distinguishing feature is that the associated absorption and bioavailability estimates are a nuisance average over an unknown mixture of true prandial states, so they describe the missingness stratum and not any physiological meal condition.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (prandial state recorded). Mutually exclusive with the meal-composition indicators in the sense that a record with FED_MISSING = 1 has no known meal composition, so FED / FED_HIGHFAT / FED_LOWFAT must all be 0 on that record and must not be read; the fasted reference is likewise derived as the complement only among records with FED_MISSING = 0.
• Source aliases:
  • FED2 = -99 – Du 2025 repotrectinib $INPUT sentinel for an unrecorded prandial state; the source column pools four levels (0 fasted, 1 fed, 2 modified fasted, -99 unknown) and the control stream branches on the sentinel exactly as it branches on the observed levels (IF (FED2.EQ.-99) TVKA = THETA(8)).
• Example models: Du_2025_repotrectinib.R (selects a fourth absorption rate constant, 0.193 1/h – the fastest of the four – and a fourth bioavailability, 0.533, essentially the fasted value; Du 2025 Table 2 KAUnknown and F1Unknown. The stratum is large: 434 of 715 prandial records, 60.6%, per Table 1. Du 2025 Discussion states that “the unknown food effect was also estimated as a nuisance parameter”, which is what allowed those records to be retained rather than excluded), Kuchimanchi_2022_diroximelFumarate.R (source column FAT = 4, Table 2 ‘UNK’, a meal of unknown fat content, recorded only for the phase III patients with multiple sclerosis, who were told to take the drug with or without food. It speeds absorption – fractional changes +0.843 and +0.399 on the two metabolite absorption rate constants – and selects its own residual SDs. Because only patients carry it, it is confounded with the patient-status covariate, as the paper’s supplement notes).
• Notes: General scope: “food status was not collected for part of the pooled dataset” is a design-independent property of retrospective pooled popPK analyses, not a property of one protocol, and the register already carries a well-populated <COVARIATE>_MISSING family on exactly this pattern (AAG_MISSING, ADA_MISSING, CYP2C9_MISSING, PDL1_TUM_MISSING, TUM_EGFR_MUT_MISSING, UGT2B15_MISSING). This indicator is not interchangeable with a fasted or fed flag and must never be collapsed into one. Its estimates are properties of a missingness mechanism – in Du 2025 the unknown-state absorption rate constant is the fastest of the four levels and its bioavailability is closest to fasted, a combination that no single real prandial state produces – so folding it into an observed level would misreproduce every prediction that touches those records. A model that carries this indicator should set it to 0 for all prospective simulations of a defined prandial state; it exists to reproduce the published fit, not to describe a meal. Distinct from FED_PATIENTSELECTED, where the meal is unstandardised but its occurrence is recorded; here the occurrence itself is unknown. Contrast Bihorel_2023_molnupiravir.R, which faced the same problem and chose the opposite resolution – mixture modelling with the high-fat fraction fixed at 25%, hard-coded into the analysis dataset – so a paper that imputes rather than estimates the unrecorded stratum should carry FED_HIGHFAT with a provenance note instead of this indicator.

TRT_BENRALIZUMAB (canonical for benralizumab treatment-arm indicator)

• Description: Binary treatment-arm indicator: 1 = the subject was randomised to benralizumab, 0 = the subject was randomised to placebo. A patient-level property of the randomised assignment, not of a trial arm in a meta-analysis.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (placebo).
• Source aliases:
  • TRTAN – the actual-treatment column of Jin 2025 Resource 3, coded ‘0: placebo, 1: benralizumab 30 mg every 4 weeks [Q4W], 2: benralizumab 30 mg every 8 weeks [Q8W]’; the binary indicator is derived as TRTAN > 0, which collapses the two active regimens because the model that uses it carries no dose or regimen term.
• Example models: Jin_2025_benralizumab_fev1.R (gates the time-driven Emax * t / (T50 + t) treatment effect on the pre-bronchodilator FEV1 endpoint on and off, so that the placebo arm is described by the baseline plus the exponential-onset placebo term alone).
• Notes: Specific scope. Follows the TRT_<INN> convention established by TRT_EPHEDRINE: use TRT_<INN> when the drug is the study treatment under investigation, CONMED_<INN> when it is background co-medication, and a bare INN only for MBMA trial-arm indicators. An explicit arm indicator is needed only when the drug effect is not concentration-driven. In an exposure-response model the Emax term vanishes of its own accord in the placebo arm because the predicted concentration is zero, so no indicator is required (see Zhang_2025_dupilumab_fev1.R, and the sibling Jin_2025_benralizumab_aaer.R, which both rely on that). Jin_2025_benralizumab_fev1.R needs the indicator precisely because its FEV1 exposure-response relationship was flat over the studied dose range and the authors therefore re-expressed the treatment effect as a function of TIME rather than of concentration – a time-driven effect does not switch itself off for placebo subjects.

TRT_ANTIPD1 (canonical for anti-PD-1 antibody in the regimen indicator)

• Description: Binary indicator: 1 = the regimen includes an anti-PD-1 monoclonal antibody (nivolumab, pembrolizumab), as monotherapy or combined with another agent; 0 = it does not.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no anti-PD-1 antibody in the regimen).
• Source aliases:
  • FactorPD1 / factor-PD-1 – Shulgin 2020 Methods (‘set to 1 if a PD-1 inhibitor drug was given as monotherapy or in combination, and to 0 otherwise’) and Supplemental Tables 1-3.
  • Factor_PD-1 / FactorPD1 – Zhang 2022 Equations 3 and 5 and its supplementary dataset (same definition).
• Example models: Zhang_2022_ici_irae_grade3_mbma.R (study-arm level in a non-small cell lung cancer irAE MBMA; enters only as the product with potency-normalized anti-CTLA-4 exposure, adding 0.0176 per unit to the grade >= 3 irAE logit slope), Shulgin_2020_ici_trae34_mbma.R, Shulgin_2020_ici_gi_imae34_mbma.R, Shulgin_2020_ici_hepatic_imae34_mbma.R (study-arm level; enters only as the product with potency-normalized anti-CTLA-4 exposure, steepening the exposure slope of grade 3/4 adverse-event rates four- to eight-fold), Shulgin_2020_ici_trae34_covariate_mbma.R (additionally gates the chemotherapy effect: chemotherapy raises the logit only when an anti-PD-1 antibody is present).
• Notes: Class-level member of the TRT_ family (compare the modality indicator TRT_ANTIPCSK9_SIRNA), used because the source pools two anti-PD-1 antibodies under one indicator. Unlike CONMED_IPI_ANY / COMBO_NIVO, which flag a SECOND agent added to a backbone drug, this flag is 1 for anti-PD-1 MONOTHERAPY as well, so the anti-PD-1 + anti-CTLA-4 combination is identified by TRT_ANTIPD1 = 1 together with a positive anti-CTLA-4 exposure, not by this column alone. Register a drug-specific TRT_<INN> instead when a source distinguishes nivolumab from pembrolizumab.

TRT_ANTIPDL1 (canonical for anti-PD-L1 antibody in the regimen indicator)

• Description: Binary indicator: 1 = the regimen includes an anti-PD-L1 monoclonal antibody (atezolizumab, durvalumab, avelumab), as monotherapy or combined with another agent; 0 = it does not.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no anti-PD-L1 antibody in the regimen).
• Source aliases:
  • Factor_PD-L1 / FactorPDL1 – Zhang 2022 Methods (‘set to 1 if a PD-L1 inhibitor was given as monotherapy or in combination with a CTLA-4 inhibitor, and to 0 when not given’), Equations 2-5 and the supplementary dataset.
• Example models: Zhang_2022_ici_irae_mbma.R (study-arm level; -0.3484 on the any-grade irAE logit relative to the anti-PD-1 reference), Zhang_2022_ici_irae_grade3_mbma.R (-0.6271 on the grade >= 3 irAE logit, plus 1.3525 through its product with anti-CTLA-4 presence for the anti-CTLA-4 + anti-PD-L1 combination).
• Notes: Anti-PD-L1 sibling of TRT_ANTIPD1, with the same class-level pooling and the same reading: it is 1 for anti-PD-L1 MONOTHERAPY as well, so it is not an add-on flag like COMBO_DURVA. The anti-CTLA-4 + anti-PD-L1 combination is identified by TRT_ANTIPDL1 = 1 together with a positive anti-CTLA-4 exposure (CAV > 0). Register a drug-specific TRT_<INN> instead when a source distinguishes the individual anti-PD-L1 antibodies; Zhang 2022 screened the drug within a target class and found no difference.

TRT_CORTICOSTEROID, TRT_IMMUNOSUPPRESSANT, TRT_RAS_BLOCKER, TRT_ANTIPLATELET, TRT_OMEGA3_FA, TRT_OTHER_IGAN (canonical for IgA-nephropathy drug-class treatment-arm indicators)

• Description: Binary study-arm drug-class indicators for IgA-nephropathy treatment: 1 = the arm received a drug of that class, 0 otherwise. TRT_CORTICOSTEROID = systemic corticosteroid; TRT_IMMUNOSUPPRESSANT = non-steroid immunosuppressant (e.g. mycophenolate mofetil, cyclophosphamide, azathioprine, calcineurin inhibitors); TRT_RAS_BLOCKER = renin-angiotensin system blocker (ACE inhibitor or angiotensin receptor blocker); TRT_ANTIPLATELET = antiplatelet agent; TRT_OMEGA3_FA = N-3 (omega-3) polyunsaturated fatty acids (fish oil); TRT_OTHER_IGAN = any agent outside those five classes, or a combination of drugs from different classes. Exactly one indicator is 1 on a drug arm and all six are 0 on a placebo arm.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (placebo arm, or a different drug class).
• Source aliases:
  • Corticosteroids, Immunosuppressant, RAS blockers, Antiplatelet agents, N-3 fatty acids, Other drugs – Yu 2022 Tables 1-3 class labels, which follow the Japanese Society of Nephrology 2014 guideline classification.
• Example models: Yu_2022_igaNephropathy_mbma.R (study-arm level; each indicator selects its class Emax on the change from baseline in urinary protein excretion, and any indicator being 1 switches the arm from the placebo Emax / ET50 to the drug Emax / shared drug ET50).
• Notes: Specific scope. Class-level members of the TRT_ family, following TRT_ANTIPD1 and TRT_ANTIPCSK9_SIRNA: the source pools many agents under each class, so no TRT_<INN> or bare-INN arm indicator applies. TRT_OTHER_IGAN carries a disease suffix because “other” is defined by exclusion from the five named IgA-nephropathy classes and has no meaning outside that classification; a different disease’s residual class should register its own name. Distinct from CONMED_STEROID, which flags background corticosteroid use in a model of some other drug; here the corticosteroid is the studied treatment of the arm.

TRT_SB16 (canonical for SB16 denosumab-biosimilar treatment-arm indicator)

• Description: Binary treatment-arm indicator: 1 = the subject received SB16 (a proposed denosumab biosimilar, approved as Obodence / Xbryk in the EU and Ospomyv / Xbryk (denosumab-dssb) in the US), 0 = the subject received reference denosumab. A patient-level randomised-treatment indicator. In Choi 2025 the EU-sourced and US-sourced reference products were pooled into a single reference group, so the 0 level spans both.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (reference denosumab, EU- and US-sourced pooled). Choi 2025 justify pooling the two reference sources because the same pivotal clinical data supported approval of the originator in both regions.
• Source aliases:
  • treatment group (SB16 vs DEN) – Choi 2025 Methods 2.7 and Results 3.3.
• Example models: Choi_2025_denosumab.R (multiplicative effect on CL/F of 0.9982, i.e. a 0.18% lower apparent clearance on SB16 than on reference denosumab).
• Notes: Specific scope (a per-product arm indicator). Member of the TRT_<product> family; named for the development code rather than the INN because both arms receive denosumab and the INN alone cannot distinguish them – the same reasoning that gives the anti-HBV siRNA arms their development-code names. This indicator carries a deliberately non-significant effect. Choi 2025 found treatment group was not a statistically significant covariate on any PK or PD parameter and excluded it from the covariate-selected model, then re-introduced it on CL/F specifically to drive the comparative biosimilarity simulation (Methods 2.3, Results 3.3). Consumers reproducing the covariate-selected final model should leave TRT_SB16 at its reference level 0; consumers reproducing the paper’s Table 5 biosimilarity comparison should vary it. Future biosimilar-vs-reference extractions should ratify a sibling TRT_<code> canonical rather than overloading this one, since the effect size is product- and analysis-specific.

TRT_PF06439535 (canonical for PF-06439535 bevacizumab-biosimilar treatment-arm indicator)

• Description: Binary treatment-arm indicator: 1 = the subject received PF-06439535 (Pfizer’s bevacizumab biosimilar, approved as Zirabev / bevacizumab-bvzr), 0 = the subject received reference bevacizumab sourced from the European Union (Avastin). A patient-level randomised-treatment indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (bevacizumab-EU, the reference product).
• Source aliases:
  • DP (PF-06439535=1, Bevacizumab-EU=2) – Li 2020 Figure 2 axis label; the final-model equations apply the effect ‘(1.02 in PF-06439535)’ on CL and ‘(1.07 in PF-06439535)’ on V1.
• Example models: Li_2020_bevacizumab.R (multiplicative theta^COV effect, Li 2020 Eq. 5: 1.02 on CL and 1.07 on V1).
• Notes: Specific scope; sibling of TRT_SB16 in the TRT_<product> family, named for the development code because both arms receive bevacizumab. This indicator carries a deliberately non-significant effect: Li 2020 retained drug product in the final model to quantify similarity even though the stepwise covariate search did not select it (Table 2 footnote b); both 95% bootstrap CIs include unity. Consumers reproducing the typical bevacizumab-EU patient should leave it at 0.

IGA_MOD_PCT (canonical for moderate-IGA-severity cohort prevalence percentage)

• Description: Study-arm-level percentage (0-100) of the enrolled cohort whose baseline Investigator’s Global Assessment (IGA) was 3 (moderate) rather than 4 (severe). Continuous covariate scaled in percent (not fraction). An arm enriched for severe disease (a LOW value here) leaves more room for improvement, so this column typically carries a negative coefficient on a drug-effect maximum.
• Units: %
• Type: continuous
• Scope: specific
• Reference category: 0% (all-severe arm). Note that models centre this covariate on a study-set median rather than on zero – Takechi 2025 centres on 59.0%.
• Source aliases: PPM / “proportion of patients classified as moderate by IGA” (Takechi 2025 Table 4 covariate label). Frequently published as its COMPLEMENT: Takechi 2025 Table 3 prints IGA4 at baseline (%), the severe share, so IGA_MOD_PCT = 100 - IGA4%.
• Example models: Takechi_2025_nemolizumab_mbma_ppnrs.R (fractional-deviation coefficient -1.84 on the nemolizumab maximum logit-scale effect, centred on the 59.0% median of the 13 meta-analysed arms, consumed as IGA_MOD_PCT / 100).
• Notes: MBMA study-arm-level covariate – an aggregate property of a trial arm, NOT a per-subject characteristic. Distinct in shape from the register’s nearest severity neighbours HEPIMP_MOD and RENALIMP_MOD, which are per-subject binary indicators; this is a continuous proportion of participants, which is why it carries the _PCT suffix. Family precedent: DIS_CHD_PERCENT (Vargo 2014); direct siblings TUMTP_SQUAM_PCT, PS_ECOG_0_PCT, RACE_ASIAN_PCT (Franzese 2026). The _PCT suffix and percent scaling were ratified by the maintainers over a proposed IGA_MOD_FRAC so that every arm-level proportion in the register shares one convention, with the percent-to-fraction conversion done inline in model() and therefore visible at the point of use. Not an alias of K_CTRL, which is a trial-design input to a sample-size calculator rather than a baseline-severity descriptor. The IGA instrument is a 4-point severity scale; a companion per-subject IGA score column would be a separate canonical.

DIS_AF (canonical for nonvalvular atrial fibrillation cohort indicator)

• Description: 1 = record belongs to a cohort or study arm of patients with nonvalvular atrial fibrillation (AF), 0 = non-AF cohort (the model reference). Used when a pooled multi-indication analysis carries the AF cohort as its own binary partition of the data.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (non-AF cohort; which indication that reference actually is must be stated per model).
• Source aliases:
  • STU = 3001 – study-number encoding of the AF cohort in Willmann_2018a_rivaroxaban.R (the ROCKET AF phase III PK substudy). As for DIS_ACS, an integrated multi-indication popPK partitions by study number because indication and study coincide.
• Example models: Willmann_2018a_rivaroxaban.R (multiplicative study / indication effect on rivaroxaban CL/F: 0.849^DIS_AF, i.e. 15.1% lower apparent clearance than the VTE-treatment reference; RSE 3.48%, bootstrap 95% CI 0.793-0.900; n = 161 patients / 800 observations).
• Notes: Sibling of DIS_ACS and DIS_VTE_P within the DIS_<indication> family; mutually exclusive with them per record in the founding model. Distinct from an anticoagulation-treatment indicator: this flags the enrolled indication, not the drug. Willmann 2018a is explicit that its indication effects are not mechanistically resolved – the Discussion calls them “likely multifactorial”, mixing real differences in patient health status against study-specific peak / trough sampling windows that emphasise different parts of the concentration-time profile – so a DIS_AF coefficient estimated in one pooled analysis should not be transported to another study design without re-estimation.

DIS_VTE_P (canonical for venous-thromboembolism-prevention (post-orthopaedic-surgery thromboprophylaxis) cohort indicator)

• Description: 1 = record belongs to a cohort receiving anticoagulation for prevention of venous thromboembolism after elective major orthopaedic surgery (hip or knee replacement), 0 = other cohort (the model reference). Deliberately distinct from a VTE-treatment cohort, in which the patient already has an acute thrombus: the two differ in the acute-surgical physiology that drives clearance, which is exactly what the founding model estimates.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (non-VTE-prevention cohort; in the founding model the reference is the VTE-treatment cohort).
• Source aliases:
  • STU in (10944, 10945, 11527) – study-number encoding of the VTE-prevention cohort in Willmann_2018a_rivaroxaban.R (ODIXa-Hip2, ODIXa-Knee and ODIXa-OD-Hip respectively).
• Example models: Willmann_2018a_rivaroxaban.R (multiplicative study / indication effect on rivaroxaban CL/F that is time-varying within the cohort: 1.04 at or before 72 h after the first dose and 1.29 thereafter, both relative to the VTE-treatment reference; the paper attributes the rise to clearance increasing over the first three days after major orthopaedic surgery).
• Notes: Sibling of DIS_ACS and DIS_AF within the DIS_<indication> family; mutually exclusive with them per record in the founding model. The founding model pairs the indicator with a time split rather than a second column – the <= 72 h / > 72 h distinction is evaluated from the solver time inside model(), so an event table using this covariate must place the first dose at time 0. Note also that Willmann 2018a excluded the immediate postsurgical phase from its analysis dataset (patients there separate into slow and fast absorbers of rivaroxaban), so the coefficient describes the post-exclusion window only and must not be extrapolated back to the first hours after surgery.

STUDY_NAI114346 (canonical for the GSK NAI114346 intravenous zanamivir cardiac-conduction study indicator)

• Description: 1 = subject enrolled in study NAI114346 (NCT01353729; randomized placebo-controlled four-way crossover thorough-QT study of single 600 mg and 1200 mg intravenous zanamivir doses in 39 healthy adults), 0 = other.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the other seven pooled phase I-III intravenous zanamivir studies).
• Source aliases: Study (NAI114346) / STDY – Zuo 2020 Table 2 row “V1/V2 ~ Study (NAI114346)”; Figure 1 labels it STDY.
• Example models: Zuo_2020_zanamivir.R (founding example; multiplicative effect of 0.729 on both V1 and V2).
• Notes: A healthy-volunteer study, so the effect does not enter any patient simulation (Zuo 2020 Discussion); set to 0 for prospective simulations.

STUDY_SEACAT (canonical for the SEACAT 2.4.1 / 2.4.2 artemether-lumefantrine study cohort indicator)

• Description: 1 = participant of the South African SEACAT (South East African Combination Anti-malarial Therapy) pharmacokinetic studies 2.4.1 or 2.4.2 of artemether-lumefantrine in HIV-infected and uninfected adults; 0 = participant of any other study in the pooled analysis. Per-subject (study-fixed) binary indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (any other study of the Francis 2020 WWARN pooled set; the bioavailability reference is the InterACT and SEACAT evening doses).
• Source aliases: study name SEACAT 2.4.1 / SEACAT 2.4.2 in Francis 2020 Tables 1-2 (both studies share the same term; no value transformation).
• Example models: Francis_2020_lumefantrine.R (dose-occasion effects on relative bioavailability selected together with OCC: first (morning) dose F x (1 - 0.486) when OCC = 1, consecutive morning doses F x (1 - 0.772) when OCC is odd and > 1, evening doses at the reference).
• Notes: Francis 2020 attributes the morning/evening contrast to diurnal variation and to the different fat content of the meals given with each dose. The indicator marks the study; the dose-occasion selection lives in OCC. Sibling of the other STUDY_<id> members.

STUDY_UGANDA (canonical for the pooled Uganda artemether-lumefantrine studies 1-4 cohort indicator)

• Description: 1 = participant of any of the four Ugandan artemether-lumefantrine pharmacokinetic studies (Byakika-Kibwika et al.; Lamorde et al.; Walimbwa et al.) pooled in the Francis 2020 WWARN meta-analysis, where doses were taken with a standard Ugandan breakfast; 0 = participant of any other study in that pooled analysis. Per-subject (study-fixed) binary indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (any other study of the Francis 2020 pooled set).
• Source aliases: study names Uganda study 1 to Uganda study 4 in Francis 2020 Tables 1-2 (all four share one term).
• Example models: Francis_2020_lumefantrine.R (fractional effect on relative bioavailability of every dose, F x (1 - 0.269)).
• Notes: Four studies pooled into one level because their bioavailability estimates did not differ. Sibling of the other STUDY_<id> members.

STUDY_NIGERIA1 (canonical for the Nigeria study 1 artemether-lumefantrine cohort indicator)

• Description: 1 = participant of Nigeria study 1 (Parikh et al.) in the Francis 2020 WWARN meta-analysis, a multiple-dose study sampling after the observed 6th (last) dose with the previous doses self-administered; 0 = participant of any other study in that pooled analysis. Per-subject (study-fixed) binary indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (any other study of the Francis 2020 pooled set).
• Source aliases: study name Nigeria study 1 in Francis 2020 Tables 1-2.
• Example models: Francis_2020_lumefantrine.R (two dose-occasion effects selected together with OCC: relative bioavailability of the 6th (morning) dose F x (1 - 0.608) when OCC = 6, and an estimated 4.30 h delay of the unobserved 5th dose when OCC = 5, shared with STUDY_USHV).
• Notes: Distinct from Nigeria study 2 of the same analysis, which is identified by its dried-blood-spot matrix (SAMPLE_DBS) rather than a study term. Sibling of the other STUDY_<id> members.

STUDY_USHV (canonical for the U.S. healthy-volunteer artemether-lumefantrine study cohort indicator)

• Description: 1 = participant of the U.S. healthy-volunteer artemether-lumefantrine drug-interaction study (German et al. 2009) pooled in the Francis 2020 WWARN meta-analysis; 0 = participant of any other study in that pooled analysis. Per-subject (study-fixed) binary indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (any other study of the Francis 2020 pooled set).
• Source aliases: U.S. healthy volunteer study in Francis 2020 Tables 1-2.
• Example models: Francis_2020_lumefantrine.R (estimated 4.30 h delay of the unobserved 5th dose when OCC = 5, shared with STUDY_NIGERIA1; bioavailability not different from the reference).
• Notes: Sibling of the other STUDY_<id> members.

STUDY_0059 (canonical for Lo 2021 revefenacin phase II Study 0059 cohort indicator)

• Description: 1 = record from Theravance revefenacin Study 0059 (Lo 2021 “Study 1”, NCT03064113; n = 32 patients with moderate to severe COPD, single nebulized doses of 350 and 700 ug in a four-period crossover), 0 = any other study in the pooled analysis. Per-record (study-fixed) binary indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the other two phase II studies 0091 and 0117 and the two phase III studies 0126 and 0127).
• Source aliases:
  • Study 1 – Lo 2021 Methods 2.1.1 and Table 2 row ‘Study 1 effect on F1’.
  • Study 0059 – FDA Clinical Pharmacology Review of NDA 210598, Table 4.1.2.4.1 row ‘Effect of Study 0059 on Bioavailability’.
• Example models: Lo_2021_revefenacin.R (multiplies the relative bioavailability of nebulized revefenacin by 0.553 in the form F1 = TV x 0.553^STUDY_0059, reflecting the lower exposures observed in that study; the FDA reviewer suggests a different nebulizer as a possible cause).
• Notes: Member of the STUDY_<id> auto-approve family; the id keeps the sponsor’s four-digit protocol number with its leading zero so it cannot be read as a generic cohort count. Set to 0 when simulating the marketed product; the paper does not attribute the effect to any patient characteristic.

STUDY_242_07_208 (canonical for delamanid trial 242-07-208 indicator)

• Description: 1 = the record comes from Otsuka trial 242-07-208 (NCT02573350), the open-label phase II extension of trial 242-07-204 in which MDR-TB patients received delamanid plus an optimized background regimen for up to 26 more weeks; 0 = any other trial in the pooled analysis.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (trials 242-07-204, 242-08-210 and 242-09-213).
• Source aliases:
  • trial 208 – Wang 2020 Table 1 and Table 4 row sigma pr,7208.
• Example models: Wang_2020_delamanid.R (selects the trial-208 proportional residual SD, 41.8% vs 26.7% in the other trials).
• Notes: Well-formed member of the auto-approved STUDY_<id> family, named with the full Otsuka protocol number because the short form 208 is not unique across sponsors.

STUDY_242_09_213 (canonical for delamanid phase III trial 242-09-213 indicator)

• Description: 1 = the record comes from Otsuka phase III trial 242-09-213 (NCT01424670; delamanid 100 mg BID for 8 weeks then 200 mg QD for 18 weeks plus an optimized background regimen in MDR-TB); 0 = any other trial in the pooled analysis.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (trials 242-07-204, 242-07-208 and 242-08-210).
• Source aliases:
  • trial 213 – Wang 2020 Table 1 and Table 4 row sigma add,9213.
• Example models: Wang_2020_delamanid.R (selects the trial-213 additive residual SD, 44.2 ng/mL vs 1.55 ng/mL in the other trials).
• Notes: Well-formed member of the auto-approved STUDY_<id> family, named with the full Otsuka protocol number because the short form 213 is not unique across sponsors.

FORM_IMEGLIMIN_CONVENTIONAL_TABLET (canonical for imeglimin conventional (non-optimised) tablet formulation indicator)

• Description: 1 = imeglimin conventional tablet; 0 = capsule or optimised tablet. Used with FORM_CAPSULE; the reference formulation (both indicators 0) is the optimised tablet of the Japanese phase IIb / III studies.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (capsule or optimised tablet).
• Source aliases: derived from the study formulation in Tomita 2022 Table 1 (conventional tablet: single/multiple ascending dose study and Western phase IIb).
• Example models: Tomita_2022_imeglimin.R (fractional effects on lag time, +0.719, and, shared with the capsule, on ka, +0.298).
• Notes: Member of the FORM_<drug>_<formulation> family. Mutually exclusive with FORM_CAPSULE in the Tomita 2022 dataset.

Occasion / period (IOV)

OCC (canonical for the integer-valued occasion / period column)

• Description: Integer-valued occasion / period indicator for inter-occasion-variability (IOV) modelling. Values 1, 2,…, N identify the occasion to which each observation belongs (typically a dosing visit, study period, or sampling occasion). Time-varying within subject; constant within an occasion.
• Units: (count)
• Type: categorical
• Scope: general
• Reference category: n/a – OCC is decomposed inside model() into mutually-exclusive binary indicators, e.g., oc1 <- (OCC == 1), oc2 <- (OCC == 2),…, that are then multiplied against per-occasion eta* slots.
• Source aliases:
  • OCC – used in Jonsson_2011_ethambutol.R (DDMODEL00000220 NMTRAN $INPUT column; values 1..4).
• Example models: Jonsson_2011_ethambutol.R (4-occasion IOV on log-CL; cl <- exp(lcl + etalcl + oc1 * etalcl_oc1 + oc2 * etalcl_oc2 + oc3 * etalcl_oc3 + oc4 * etalcl_oc4) * (WT/50)^0.75, where each etalcl_oc<k> is a separate ~ fix(0.127) after the first to encode NONMEM $OMEGA BLOCK(1) SAME), Aregbe_2012_alvespimycin.R (5-occasion BOV on Q2 and V1), Oosten_2016_fentanyl.R (10-occasion IOV on transdermal Ka; only OCC >= 1 records carry IOV, sc / non-transdermal records pass OCC = 0 so all indicators zero out), Jonsson_2011_ethambutol_ddmore.R (4-occasion IOV on log-CL; cl <- exp(lcl + etalcl + oc1 * etalcl_oc1 + oc2 * etalcl_oc2 + oc3 * etalcl_oc3 + oc4 * etalcl_oc4) * (WT/50)^0.75, where each etalcl_oc<k> is a separate ~ fix(0.127) after the first to encode NONMEM $OMEGA BLOCK(1) SAME), Chen_2023_nemonoxacin.R (2-occasion IOV on log-CL distinguishing the single-dose occasion (OCC = 1) from steady state 72 h after the first of multiple doses (OCC = 2); iov_cl <- oc1 * etaiov_cl_1 + oc2 * etaiov_cl_2 with etaiov_cl_2 ~ fix(0.017) encoding the shared pi^2 of NONMEM $OMEGA BLOCK(1) SAME), Waterhouse_2024_vedolizumab.R (7-occasion IOV on log-CL across the phase 3 VEDO-3035 vedolizumab dosing schedule; occasions 2-7 ~ fix(0.0315) per NONMEM $OMEGA BLOCK(1) SAME, and records preceding the first dose take OCC = 1), Stoschus_2025_phenobarbital.R (8-occasion IOV on log-CL; the source is a MONOLIX fit that defines an occasion as the interval between consecutive therapeutic-drug-monitoring samples, cohort median 48 h, and reports one shared IOV magnitude without stating an occasion count, so eight 48-h occasions are encoded to span 0-384 h and cover the paper’s 336-h steady-state assessment time, with occasions 2-8 ~ fixed(0.127331)), Mascarenhas_2015_pentadecanoic_triheptadecanoic.R (4-occasion between-occasion variability on absorption mean transit time and on relative bioavailability, carried independently for each of the two co-administered analytes – etaiov_mtt_<k> / etaiov_fdepot_<k> for pentadecanoic acid and etaiov_mtt_hepta_<k> / etaiov_fdepot_hepta_<k> for heptadecanoic acid, occasions 2-4 ~ fixed(...) per NONMEM $OMEGA BLOCK(1) SAME; the malabsorption blood test was repeated on up to 4 separate occasions at least 5 days apart), Leeds_2013_tecovirimat_cyno.R (2-occasion IOV on log-Ka and log-CL in cynomolgus monkeys, where the occasion is the animal’s infection status – OCC = 1 the pre-infection dosing occasion, OCC = 2 the post-inoculation dosing occasion – with etaiov_ka_2 / etaiov_cl_2 ~ fixed(...) encoding the shared pi^2; Leeds 2013 does not state an occasion count, and the two-occasion reading comes from the paper attributing the IOV to infection status and printing it on exactly the two rows that carry Infected / Uninfected sub-rows), Ding_2026_vancomycin.R (4-occasion IOV on both log-CL and log-V – iov_cl and iov_vc built from the same occ1..occ4 indicators – with occasions 2-4 ~ fixed(...) per NONMEM $OMEGA BLOCK(1) SAME. Like Stoschus 2025, the source reports the IOV magnitudes (Table 3 IOV_CL 0.089 and IOV_V 0.136 on the SD scale, and a 2-df OFV drop for Model 8 confirming exactly two IOV variances) without ever stating an occasion count or how an occasion was delimited; four are encoded, covering the cohort’s median of three samples per patient. The occasion column is an extraction-side construct here: the deposited dosing control stream is a MAX=0 prediction stream that drops IOV entirely and has no occasion column in its $INPUT), Bihorel_2023_molnupiravir.R (2-occasion IOV on the log absorption mean transit time – iov_mtt <- oc1 * etaiov_mtt_1 + oc2 * etaiov_mtt_2 with etaiov_mtt_2 ~ fixed(0.147049) encoding the shared pi^2 of NONMEM $OMEGA BLOCK(1) SAME. Unusually among the entries above, the occasion count is stated outright by the source rather than inferred: the Bihorel 2023 Table 2 note reads “The different occasions for IOV in MTT were labeled as occasion 1 and occasion 2. Participants were assumed to have only one occasion for IOV in MTT, except those enrolled in the food effect and MAD part of MK-4482-004”, and the note prints separate shrinkage for each occasion (27.6% and 7.8%). Every participant with COVID-19, and the single-ascending-dose part of the phase I trial, therefore take OCC = 1. IOV rather than IIV carries this parameter because the paper found IIV in MTT “reduced to zero when associated with IOV”), Tsirizani_2025_ritonavir.R (2-occasion BOV on relative bioavailability, absorption lag, zero-order absorption duration and ka, with the occasion-2 eta of each pair ~ fixed(...) per NONMEM $OMEGA BLOCK(1) SAME. Two features are unusual. First, the occasion is defined by WITNESSING rather than by visit or period: OCC = 1 is the last dose taken before the intensive PK window, unobserved by study staff, and OCC = 2 the dose given under direct observation with a standardised breakfast. Second, the occasion indicator is load-bearing beyond multiplexing etas – it scales one of the BOV standard deviations and gates a covariate effect. The bioavailability BOV SD is inflated 2.14-fold on occasion 1 (oc1 * sd_ratio_fdepot_occ1 * etaiov_fdepot_1 + oc2 * etaiov_fdepot_2, reproducing IF(OCC.EQ.1)BOVBIO = ETA(9)*EBOV) because neither the amount nor the timing of an unwitnessed dose is known; and the 3.61-fold evening-dose lag effect is gated on the product CONMED_LOPINAVIR * oc1, since the twice-daily lopinavir arm is the only one whose occasion-1 dose was taken at night. A useful precedent that an unwitnessed-dose occasion may warrant an inflated variance rather than exclusion of the records), Hong_2013_glucose_insulin_HGC.R (2-occasion IOV on log-CLG and log-VG in a hyperglycaemic-clamp glucose-insulin model, one occasion per treatment period of the crossover; iov_clg <- oc1 * etaiov_clg_1 + oc2 * etaiov_clg_2 with the occasion-2 eta of each pair ~ fixed(...) per NONMEM $OMEGA BLOCK(1) SAME), Hong_2013_glucose_insulin_MTT.R (4-occasion IOV on the same two parameters in the meal-tolerance-test companion model, the four occasions being two crossover periods x palosuran / placebo; occasions 2-4 ~ fixed(...). The MTT IOV magnitudes exceed the HGC ones because they span four occasions and absorb intra-individual variability across the two treatment arms), Assmus_2022_emodepside.R (2-occasion IOV on the log absorption mean transit time, applied only to the multiple-ascending-dose study – OCC = 1 days 0-6 of dosing, OCC = 2 day 7 onwards, etaiov_mtt_2 ~ fixed(0.0692) per NONMEM $OMEGA BLOCK(1) SAME; single-dose studies carry OCC = 0 so both indicators zero out, reproducing the S1 Code IF(STUDY_ID.EQ.2) gate), Sasaki_2022_delamanid.R (4-occasion IOV on relative bioavailability and on the absorption mean transit time; the source defines an occasion only as ‘the i-th dosing occasion’ without a count, so four occasions are an extraction-side construct, with occasions 2-4 ~ fixed(...)), Francis_2020_lumefantrine.R (OCC = dose number within the artemether-lumefantrine course, 12 slots: per-dose BOV on F, MTT and ka, and selector for the dose-occasion bioavailability effects of STUDY_SEACAT / STUDY_NIGERIA1 and the unobserved-dose delay of STUDY_NIGERIA1 / STUDY_USHV), Zou_2022_pretomanid.R (4-occasion IOV on log relative bioavailability, log KA and log MTT, one occasion per period of a single-dose four-period crossover with 7-day washouts; occasions 2-4 ~ fixed(...) per NONMEM $OMEGA BLOCK(3) SAME, the BLOCK(3) having zero off-diagonals), vanEijk_2022_paclitaxel.R (2-occasion BOV on relative gut bioavailability of oral paclitaxel – each dose administration is an occasion, and no subject in the analysis data had more than two PK occasions – plus the 2-occasion within-subject variability on the mean absorption time and absorption-time dispersion of the embedded Yu 2020 ritonavir inverse Gaussian input; occasion-2 etas ~ fixed(...) at the occasion-1 variance).
• Notes: OCC is the recommended canonical for new IOV-using models – the binary OOC1..OOCN indicators below remain canonical for legacy / pre-existing models that ship the data already-decomposed.

VIRTUAL_DOSE (canonical for the virtual pre-study-exposure dose-record indicator)

• Description: 1 = this dose record is a virtual (hypothetical) bolus that stands in for residual drug from exposure before the study, rather than a real administration; 0 = a real dose. Per-record indicator, needed only when the virtual bolus enters the same compartment as a real dose and so cannot be told apart by cmt.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (a real dose).
• Source aliases:
  • PREDOSE – Olsson Gisleskog 2021 ESM IV control stream (F1 = THETA(7)*EXP(ETA(5)) when PREDOSE.EQ.1, else 1).
• Example models: OlssonGisleskog_2021_nicotine_iv.R (a 1-mg bolus into central at the start of the 36-h washout, with bioavailability equal to the estimated pre-washout nicotine dose, 4.90 mg, describes residual nicotine from pre-study smoking: f(central) <- (1 - VIRTUAL_DOSE) + VIRTUAL_DOSE * fcentral).
• Notes: A modelling device, not a patient attribute. The oral, buccal and transdermal siblings of the founding model dose central only with the virtual bolus and so encode it as f(central) without this flag. Omit the virtual record (and keep the flag 0) to simulate a drug-naive subject.

OOC1, OOC2, OOC3, OOC4 (canonical for mutually-exclusive crossover occasion indicators)

• Description: Mutually exclusive occasion indicators for a crossover / multi-period design. Exactly one is 1 per observation.
• Units: (binary)
• Type: binary
• Scope: specific
• Example models: Xie_2019_agomelatine.R, Salinger_2019_pretomanid.R (three-occasion interoccasion variability on a CORRELATED relative-bioavailability / clearance PAIR rather than on a single parameter: iovlfdepot and iovlcl are each built from the same OOC1 / OOC2 / OOC3 indicators, and the two etas of a given occasion share a 2x2 covariance block whose occasions 2 and 3 are fixed() to the occasion-1 values per NONMEM $OMEGA BLOCK(2) SAME. Occasions are defined per contributing study from trial length and sampling schedule – Salinger 2019 supplemental Table S1 lists them study by study, e.g. weeks 2 / 8 / 16 for Nix-TB and days 1-7 / 8-13 / 14-29 for the longer phase 1 studies – so the indicator’s meaning is study-dependent and a pooled data set must be assembled per Table S1).
• Notes: Renamed from the source file’s lower-case ooc1..ooc4 on 2026-09-09, when the ALL-CAPS convention became machine-enforced; the source spellings are recorded in Xie_2019_agomelatine.R’s source_name fields. This remains the already-decomposed legacy form – new models should prefer the integer-valued OCC canonical above and decompose into binary indicators inside model.

MONTH1 (canonical for first-month-of-treatment landmark indicator)

• Description: Binary within-subject landmark indicator: 1 = the observation falls within the first 30 days after treatment initiation, 0 = subsequent months. Time-varying within subject – gates a transient step change in typical-value CL that the source paper attributes to higher Erwinia asparaginase clearance in the first month of pediatric ALL therapy.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (month 2 onwards).
• Source aliases: none known; Sassen 2017 reports the first-month-vs-after contrast as a single multiplicative coefficient on TVCL.
• Example models:
  • Sassen_2017_crisantaspase.R (multiplicative shift on CL: cl <- exp(lcl + etalcl) * (WT/70)^0.75 * (1 + e_month1_cl * MONTH1) with e_month1_cl = 0.14, encoding the 14% higher CL in the first month of treatment relative to subsequent months).
• Notes: Specific scope because the 1-month (30-day) cutoff is tied to Sassen 2017’s pediatric Erwinia asparaginase pharmacology (transient higher CL early in treatment; mechanism not established in the source paper). Future studies that test a similar within-subject step change with a different cutoff (e.g., 2 weeks or 6 weeks) should register a new canonical name. Distinct from OCC and ooc<n> (which decompose multi-occasion sampling for IOV), from DAY14 (which uses a 14-day cutoff for malnutrition-recovery contrasts), and from CYCLE (which is a dose-number counter, not a single binary landmark). Data assemblers can derive MONTH1 = as.integer(time_post_treatment_start_days < 30) for a regularly-sampled multi-month study.

DAY2 (canonical for day-after-first-dose landmark indicator)

• Description: Binary within-subject landmark indicator: 1 = the record falls on study day 2 or later, 0 = the record falls on study day 1, the first dosing day. Time-varying within subject – gates a step change in typical-value PK parameters between the very first dose and every subsequent day of a repeated-dose regimen. The smallest-cutoff member of the DAY<n> family: where DAY4, DAY8, DAY14, DAY21 and DAY28 contrast two discrete sampling visits, DAY2 contrasts the first dose against everything after it.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (study day 1, the first dosing day).
• Source aliases:
  • DAY – the $INPUT column in the Mulford 2026 Supporting Information Model Code carries the integer study day; the three covariate branches are written IF(DAY>1) TVKA = TVKA * (1 + THETA(17)/100), IF(DAY>1) TVCL = ... and IF(DAY>1) TVVC = ..., so the canonical 0/1 indicator is DAY2 = as.integer(study_day >= 2).
  • Day > 1-effect – Mulford 2026 Table 1 row label for all three effects.
  • Q1 – Guidi 2019 Table 2 footnote defines the artesunate bioavailability effect as (1 + theta_day_F1 * Q1) “with Q1 = 0 for the first treatment day; 1 for subsequent therapy days”, which is the canonical indicator verbatim. The Guidi 2019 Methods text indexes the treatment day from 0 (day 0 vs days 1 and 2) while its Table 3 indexes from 1 (DAY = 1 vs DAY > 1); both spellings denote the same first-day-vs-rest contrast and both map to DAY2.
• Example models:
  • Mulford_2026_vonoprazan.R, Mulford_2026_vonoprazan_maturation.R (three simultaneous multiplicative shifts from one indicator: ka *= (1 + e_day2_ka * DAY2) with e_day2_ka = -0.208, cl *= (1 + e_day2_cl * DAY2) with e_day2_cl = -0.115, and vc *= (1 + e_day2_vc * DAY2) with e_day2_vc = -0.116. Founding example.)
  • Guidi_2019_artesunate.R (multiplicative shift on relative bioavailability: fdepot *= (1 + e_day2_fdepot * DAY2) with e_day2_fdepot = -0.29), Guidi_2019_mefloquine.R (the indicator instead selects between two separately estimated absorption rate constants rather than scaling one, lka_day <- (1 - DAY2) * lka_day1 + DAY2 * lka_day2 with Ka = 0.17 /h on the first treatment day and 0.40 /h afterwards). Both come from the same 3-day fixed-dose artesunate-mefloquine course in African children, where the 1 level averages over treatment days 2 and 3 only – a much tighter and differently-constituted window than the founding paper’s day-7 / day-14 sampling, and the reason this entry stays Scope: specific rather than being promoted to general. The mefloquine model is also the precedent that DAY2 may gate a stratum-suffixed pair of parameters instead of a single multiplicative coefficient, which is the right encoding when the source reports two parallel point estimates rather than a reference plus an offset.
• Notes: Registered under the naming pattern the DAY14 entry prescribes by name (“future studies that test a similar within-subject step change with a different cutoff (day 7, day 28, etc.) should register a new canonical”), and a well-formed member of the same DAY<n> landmark-indicator family as DAY4, DAY8, DAY14, DAY21 and DAY28. Specific scope because what the 1 level averages over is design-dependent: in the founding paper the post-day-1 samples were drawn on day 7 (and day 14 in the adolescent study VPED-102), and Mulford 2026 Results 3.1 states outright that a continuous relationship could not be identified from that design (“Because plasma concentrations were mainly observed on Day 1 and again on Day 7, we could not identify a continuous relationship but instead used a binary change for day > 1”), so the coefficient is the day-1-to-steady-state contrast and carries no information about when within that window the change occurs. This is exactly the situation the DAY28 entry describes as the right use of the family – a source that abandoned a continuous time-on-treatment model for a per-visit step. Distinct from MULTI_DOSE_PT, which flags a study PHASE (a multiple-dose period contrasted against a single-dose run-in or a healthy-volunteer arm) and is often subject-level; DAY2 is a per-record study-day threshold within one continuous regimen. Distinct from OCC and OOC<n>, which decompose multi-occasion sampling for inter-occasion variability and carry per-occasion eta slots rather than a typical-value step. Distinct from MONTH1, whose polarity is inverted (1 = the EARLY window). When simulating, set DAY2 = 0 throughout a single-dose study, and 0 over the first 24 h then 1 thereafter for a once-daily regimen; a steady-state projection uses DAY2 = 1.

DAY4 (canonical for day-4-of-therapy landmark indicator)

• Description: Binary within-subject landmark indicator: 1 = the observation falls on day 4 (or later) of treatment, 0 = the observation falls on day 1 of treatment. Time-varying within subject – gates a step change in a typical-value PK parameter between the first day of therapy and the expected steady-state sampling day.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (day 1 of therapy).
• Source aliases:
  • 4th day – Zavrelova 2025 writes the term as beta_CL_4th day [if 4th day] in the Table 2 parameter list and as [if 4th day] in the printed covariate equation.
• Example models:
  • Zavrelova_2025_linezolid.R (multiplicative log-scale shift on CL: cl <- exp(lcl + e_age_cl * AGE + (e_day4_cl + etae_day4_cl) * DAY4 + etalcl) with e_day4_cl = -0.40, encoding the exp(-0.40) = 0.67 fold change – a 33% CL reduction – between day 1 and day 4 of intravenous linezolid therapy in hematooncological patients with sepsis).
• Notes: Specific scope because the day-4 cutoff is the source study’s therapeutic-drug-monitoring sampling landmark (concentrations drawn on day 1 and again on day 4, the expected steady-state day), not a mechanistic threshold: Zavrelova 2025 states in the Discussion that the timing of the CL change between days 1 and 4 could not be identified from the design and “probably occurs” by around day 2. Distinct from OCC and ooc<n> (which decompose multi-occasion sampling for IOV and carry per-occasion eta slots rather than a typical-value step), from DAY14 and MONTH1 (same landmark-indicator family, different cutoffs and different source pharmacology), and from CYCLE (a dose-number counter, not a binary landmark). Data assemblers can derive DAY4 = as.integer(time_post_treatment_start_days >= 4) – or, for a study that samples only the two landmark days, directly from the sampling-day label.

DAY8 (canonical for day-8-of-treatment landmark indicator)

• Description: Binary within-subject landmark indicator: 1 = the observation falls on day 8 of treatment, 0 = the observation falls on day 1 of treatment. Time-varying within subject – gates a step change in a typical-value PD parameter between the first dosing day and the second intensive-sampling day of a first-in-human dose-escalation study.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (day 1 of treatment).
• Source aliases:
  • Dosing day (Day 8 versus Day 1) – Mukker 2026 Table S1 row label for the additive fixed effect on the C-DeltaHR intercept.
• Example models:
  • Mukker_2026_tuvusertib_HR.R (additive shift on the linear-mixed-effects intercept of the concentration-DeltaHR model: e_day8_e0 = 2.84 bpm, encoding a heart rate 2.84 bpm higher on day 8 than on day 1 at matched tuvusertib concentration and nominal post-dose time).
• Notes: Specific scope because the day-8 landmark is the DDRiver Solid Tumors 301 Part A1 second intensive PK-ECG sampling day, not a mechanistic threshold. Distinct from DAY4 and DAY14 (same landmark-indicator family, different cutoffs and different source pharmacology), from OCC / ooc<n> (which carry per-occasion eta slots rather than a typical-value step), and from CYCLE (a dose-number counter, not a binary landmark). Note that in the founding model this indicator sits alongside T_LASTDOSE, which carries the within-day nominal post-dose hour: DAY8 distinguishes which sampling day an ECG belongs to, T_LASTDOSE where within that day it falls. Data assemblers can derive DAY8 = as.integer(treatment_day >= 8) for a study that samples only the two landmark days.

DAY14 (canonical for day-14-post-treatment-initiation landmark indicator)

• Description: Binary within-subject landmark indicator: 1 = the observation falls on or after day 14 of treatment, 0 = the observation falls before day 14. Time-varying within subject – gates a step change in a typical-value PK or PD parameter that the source paper estimates separately for the first two weeks of treatment and for day 14 onwards. The mechanism is paper-defined and recorded per model: nutritional recovery plus auto-induction of hepatic metabolism in the Archary 2019 / MATCH trial of severely malnourished HIV-infected children; a higher maximal biomarker inhibition after 14 days of repeated dosing in Taneja 2019.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (before day 14; e.g., day 1 of treatment).
• Source aliases: none known; Archary 2019 reports the day-1-vs-day-14 contrast directly in the parameter table (separate typical-value rows for “day 1” and “day 14”); Taneja 2019 Table 3 labels it ‘Imax at day 14 and later of the FIH study, MAD part’.
• Example models:
  • Archary_2019_abacavir.R (multiplicative additive shift on CL/F: cl <- exp(lcl + etalcl) * (WT/7)^0.75 * (1 + e_day14_cl * DAY14) with e_day14_cl = 0.760, encoding the day-1 typical CL/F = 3.33 -> day-14 CL/F = 5.86 L/h per 7 kg step).
  • Archary_2019_lamivudine.R (multiplicative additive shift on ka: ka <- exp(lka) * (1 + e_day14_ka * DAY14) * exp(etalka) with e_day14_ka = 0.133, encoding the day-1 typical ka = 0.30 -> day-14 ka = 0.34 /h step).
  • Taneja_2019_glpg1690.R (additive shift of 0.2043 on the probit of Imax, gated to one study part as e_day14_imax * STUDY_FIH_MAD * DAY14, raising Imax from 90.8% to 93.7%; the gate is required because the 12-week proof-of-concept study in the same pooled analysis also samples past day 14 without carrying the effect).
• Notes: Promoted to general scope because the cutoff and the within-subject-landmark shape recur independently of the Archary 2019 re-feeding window; the mechanism behind the step is paper-specific and belongs in covariateData[[DAY14]]$notes. When the effect applies to only one study or study part of a pooled analysis, gate it with the matching STUDY_<id> indicator rather than minting a compound canonical. Studies that test a similar within-subject step change with a different cutoff (day 7, day 28, etc.) should register a new canonical (e.g., DAY28; see the sibling DAY21). Distinct from OCC and ooc<n> (which decompose multi-occasion sampling for IOV), from TRT_PHASE (which gates active-vs-baseline study-phase contributions), and from EARLY_ART (which is a between-subject randomization-arm indicator, not a within-subject landmark). Data assemblers can derive DAY14 = as.integer(time_post_treatment_start_days >= 14) for a regularly-sampled multi-day study.

DAY21 (canonical for day-21-of-cycle landmark indicator)

• Description: Binary within-subject landmark indicator: 1 = the observation was collected at the day-21 visit of the treatment cycle, 0 = the observation was collected at the day-1 (first-dose) visit of the same cycle. Time-varying within subject – gates a step change in a typical-value model parameter between the first dose of a cycle and the end-of-dosing (steady-state) visit of that cycle.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (day 1 of the cycle).
• Source aliases:
  • visit – Zhou 2025 writes the term as theta_4 * Visit_k in the printed C-QTc equation, where “theta_4 is the fixed effect associated with cycle 1 day 21 (taking cycle 1 day 1 as the reference visit)” and k indexes the visit (cycle 1 day 1 or cycle 1 day 21); the parameter tables label the row Cycle 1 day 21 visit.
• Example models:
  • Zhou_2025_fruquintinib_QTcP_M11.R (additive shift on the linear-mixed-effects DeltaQTcP intercept, e_day21_e0 = -1.39 ms; the cycle 1 day 21 ECG visit follows 21 consecutive daily 5 mg doses of fruquintinib in a 28-day cycle, so it is the steady-state visit).
  • Zhou_2025_fruquintinib_QTcF_M11.R (same term on the Fridericia-corrected endpoint, e_day21_e0 = -1.50 ms).
• Notes: Specific scope because the day-21 cutoff is the source study’s dosing-schedule landmark – fruquintinib is given on days 1-21 of a 28-day cycle, so day 21 is the last dosing day of the cycle rather than a mechanistic threshold. Distinct from CYCLE (a dose-number counter across cycles, not a within-cycle binary landmark): a multi-cycle study would carry both, CYCLE identifying which cycle and DAY21 identifying the within-cycle visit. Distinct from OCC and ooc<n>, which decompose multi-occasion sampling for inter-occasion variability and carry per-occasion eta slots rather than a typical-value step. Also distinct from NTIME, which resolves the within-day nominal hour after dose: in the founding models the two are orthogonal and both appear, DAY21 selecting the visit and NTIME the hour within it. Data assemblers can derive DAY21 = as.integer(day_within_cycle >= 21) – or, for a study that samples only the two landmark visits, directly from the visit label. Not every model that tests the term retains it: Zhou_2025_fruquintinib_QTcP_parent.R documents it in covariatesDataExcluded because the source authors dropped it from that fit (95% CI included 0; likelihood-ratio-test p = 0.1318).

DAY28 (canonical for day-28-of-treatment landmark indicator)

• Description: Binary within-subject landmark indicator: 1 = the record belongs to the day-28 pharmacokinetic visit of the study, 0 = the record belongs to the earlier baseline / early-treatment visit that the same study samples (day 3 in the founding model). Time-varying within subject – gates a step change in a typical-value model parameter between the early and the late sampling visit, typically standing in for enzyme autoinduction, nutritional recovery, or another slow process that has run to completion by the second visit.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the study’s earlier PK visit; day 3 in Abdelgawad_2025_rifampicin.R).
• Source aliases:
  • PK_VISIT_2 – the NONMEM control-stream column in the Abdelgawad 2025 supplementary material, which carries the literal visit day (values 3 and 28) rather than a 0/1 indicator; the covariate branches are written IF (RIFHIGH.EQ.0.AND.PK_VISIT_2.EQ.3) and IF (RIFHIGH.EQ.1.AND.PK_VISIT_2.EQ.28). The canonical column is the 0/1 indicator, so a data assembler derives DAY28 = as.integer(pk_visit_day >= 28).
  • visit 2 / Visit 2 – Abdelgawad 2025 Table 2 row labels (“Standard dose: visit 2”, “High dose: visit 2”).
• Example models:
  • Abdelgawad_2025_rifampicin.R (selects, jointly with DOSE_HIGH, one of four stratum-suffixed typical values of the maximal intrinsic hepatic clearance: 33.1 L/h standard dose day 3, 41.4 L/h standard dose day 28, 46.1 L/h high dose day 3, 70.2 L/h high dose day 28, all as the reported CLint,max * fu product).
  • Calderin_2025_pyrazinamide.R (multiplicative shift on clearance: cl <- exp(lcl + etalcl) * (FFM/45)^0.75 * (1 + e_day28_cl * DAY28) with e_day28_cl = 0.302, i.e. pyrazinamide clearance is 30.2% higher at day 28 than at day 3, reducing exposure in both plasma and CSF over the first month of antituberculosis therapy).
• Notes: Registered under the naming pattern the DAY14 entry prescribes by name (“future studies that test a similar within-subject step change with a different cutoff (day 7, day 28, etc.) should register a new canonical (e.g., DAY28)”), and a well-formed member of the same DAY<n> landmark-indicator family as DAY4, DAY8, DAY14 and DAY21. Specific scope because the day-28 cutoff is a sampling-schedule landmark of the source trial rather than a mechanistic threshold, and because what the 0 level means depends on which earlier visit the study sampled. Distinct from OCC and ooc<n>, which decompose multi-occasion sampling for inter-occasion variability and carry per-occasion eta slots rather than a typical-value step – the founding model carries both, OCC (1-5) driving the between-occasion etas on absorption and DAY28 selecting the visit-specific typical clearance. Distinct from DAY21, whose 0 level is day 1 of a treatment cycle rather than an early on-treatment visit, and from DAY14, whose semantics are “on or after day 14 of treatment” across a continuously sampled study rather than “at the second of two discrete PK visits”. A study whose autoinduction is resolvable in continuous time should prefer a time-since-treatment-start covariate over this indicator; DAY28 is the right column when the source paper explicitly abandoned a continuous autoinduction model for separate per-visit typical values, as Abdelgawad 2025 did after neither an exponential time-on-treatment model nor an enzyme-turnover model converged.

NTIME (canonical for the protocol-scheduled (nominal) hour after the most recent dose)

• Description: Protocol-scheduled (“nominal”) time, in whole hours after the most recent dose, of the sampling slot a record belongs to. An integer categorical index over the study’s fixed set of post-dose sampling hours, with hour 0 (pre-dose) as the reference level. Time-varying within subject. Carries a set of free, non-periodic per-timepoint FIXED effects – the standard ICH E14 / Garnett concentration-QTc design element, where the nominal-time coefficients absorb the drug-independent diurnal drift of the ECG endpoint over the sampling day. Decomposed inside model into mutually-exclusive binary indicators multiplied by per-timepoint covariate-effect parameters.
• Units: h
• Type: categorical
• Scope: general
• Reference category: 0 (the pre-dose nominal time point). NTIME = 0 sets every derived indicator to zero and recovers the reference intercept. The typical model() decomposition is nt1 <- (NTIME == 1);...; ntime_effect <- e_ntime1_e0 * nt1 +..., following the pattern the OCC entry prescribes (“decompose into binary indicators inside model()”).
• Source aliases:
  • NTIME – the term name in the Zhou 2025 supplementary model-development logs, glossed “NTIME, nominal time point after the most recent dose” (Supplementary Tables 1-3 footnotes).
  • NTime – the parameter-table row prefix in Zhou 2025 Table 2 and Supplementary Tables 4-5 (NTime = 1.. NTime = 4), glossed “NTime, nominal time in hours”.
  • theta_3(NTime_j) – the printed-equation notation, described as “the fixed effect associated with nominal time (one value estimated nonreference per nominal time point)”.
• Example models: Zhou_2025_fruquintinib_QTcP_M11.R, Zhou_2025_fruquintinib_QTcF_M11.R, Zhou_2025_fruquintinib_QTcP_parent.R (all three carry the 0, 1, 2, 3, 4 h nominal-time grid of the FRESCO-2 ECG substudy, with four estimated coefficients e_ntime1_e0 .. e_ntime4_e0 against the 0 h reference).
• Notes: General scope: nominal-time fixed effects are a standard element of the ICH E14 / Garnett C-QTc design, so any future concentration-QTc extraction will need this column regardless of drug or indication, and one integer column generalises to any number of sampling hours (a study with 1, 2, 4, 6, 8 and 24 h slots needs no new canonical). NOMINAL, not actual. The value is the protocol slot, not the elapsed time since dosing: a sample drawn at 0.4 h actual time after dose belongs to the 1 h nominal slot if that is the protocol slot it was collected in. Do NOT derive this column from rxode2’s tad() – that conflates the two and silently substitutes a hidden assumption for an auditable covariate. Because the model() decomposition tests exact equality, supply whole-number hours. Deliberately distinct from three neighbouring canonicals. (1) TCLOCK is a CONTINUOUS wall-clock hour of day feeding a 24-hour PERIODIC function; NTIME is a discrete set of free, non-periodic dummy coefficients at protocol hours after dose. (2) OCC is an occasion index tied to per-occasion eta slots for inter-occasion variability; NTIME carries FIXED effects and no random effects. (3) The T_<event> family members (T_FIRSTDOSE, T_ENTRY) are continuous elapsed-time-since-event covariates rather than a categorical slot index. Also orthogonal to the DAY<n> landmark family: in the founding models DAY21 selects the visit and NTIME the hour within it, and both appear together. Name and general scope ratified by the maintainers (PMC12555101, 2026-08-28) alongside the Zhou 2025 fruquintinib C-QTc extraction; the integer-plus-decomposition form was chosen over four separate binary canonicals precisely so that a C-QTc study with a different sampling grid needs no new register entry.

CYCLE (canonical for dose-number / treatment-cycle counter)

• Description: Dose-number / treatment-cycle counter (1 = first dose or cycle, 2 = second,…). Integer count, time-varying across a multi-dose / multi-cycle treatment course, incremented at each new administration.
• Units: (count)
• Type: count
• Scope: general
• Reference category: n/a – used either with a power-covariate form CYCLE^Fm (Fm typically negative) to capture cycle-over-cycle decline in a derived quantity such as ADC-to-payload conversion fraction (Li 2017 brentuximab vedotin), or with a piecewise indicator CYCLE == 1 vs CYCLE >= 2 to capture a step change in PK between the first and subsequent administrations (Hong 2025 datopotamab deruxtecan; Huynh 2026 VRC07-523LS).
• Source aliases:
  • CYCLE – used in Li_2017_brentuximab.R, Hong_2025_datopotamab.R, Lu_2022_patritumab.R, and Huynh_2026_VRC07523LS.R with the same canonical name.
  • MTXNUM – Zhang 2015’s counter of methotrexate chemotherapy courses, glossed by the paper as “the number of methotrexate chemotherapy cycles before MTX infusion” but tabulated with a MINIMUM of 1 rather than 0 in a dataset that must contain first courses, which identifies it as a 1-based counter of the current course rather than a count of strictly prior ones. Used in Zhang_2015_methotrexate.R. Note that the paper separately screened NUM, the count of chemotherapy cycles of any drug (median 8, range 1-28); the two were strongly correlated and only the drug-specific counter was retained, so NUM is documented under covariatesDataExcluded in that model rather than mapped to this canonical.
• Example models:
  • Li_2017_brentuximab.R (exponent on the fraction of ADC that converts to MMAE by proteolytic degradation, Fm = -0.261, to reflect tumor-burden reduction across successive treatment cycles).
  • Zhang_2015_methotrexate.R (founding LINEAR-in-count example, and the founding example of an UNCENTERED covariate term: enters central clearance as the bare count (1 - 0.0183 * CYCLE) with no reference subtraction, so the paper’s tabulated typical CL of 6.20 L/h is the value extrapolated to CYCLE = 0, which never occurs in the data. At the cohort median of two courses the typical clearance is 5.97 L/h, which is also much closer to the covariate-free base model’s 5.81 L/h – that agreement is what confirms the uncentered form was transcribed correctly. The mechanism the paper offers is cumulative renal impairment from repeated high-dose methotrexate).
  • Hong_2025_datopotamab.R (cycle-1 vs cycle-2+ piecewise scaling Factor1 = 0.696 on the DAR equation that drives DXd formation rate from total Dato-DXd elimination).
  • Lu_2022_patritumab.R (cycle-1 vs cycle-2+ piecewise scaling Factor1 = theta = 0.648 on the payload-to-intact-drug ratio PIR that scales DXd release rate from intact ADC).
• Notes: Must be >= 1 throughout (CYCLE^Fm is undefined at 0; the piecewise form requires CYCLE to be a positive integer at every observation row). Distinct from ooc<n> binary-occasion indicators: CYCLE is an integer count, not a mutually-exclusive set of indicator columns. Data-assembly helper: set CYCLE = floor((TIME - TIME_FIRST_DOSE) / cycle_length_days) + 1 for a fixed-interval dosing regimen.

T_FIRSTDOSE (canonical for time elapsed since the first dose of the treatment course)

• Description: Time elapsed since the subject’s very first administration of the modelled drug, i.e. since the start of the treatment course rather than since the most recent dose. Time-varying within subject and monotonically increasing across the whole record; it does NOT reset at each dosing event (that quantity is time-after-dose, which rxode2 supplies natively via tad and which must not be confused with this column). Used to drive slow, treatment-duration-dependent drift in a PK parameter – autoinduction, adherence decay, disease-state resolution, or a gradual change in absorption or bioavailability – over weeks to months, on a timescale far longer than a dosing interval.
• Units: h
• Type: continuous
• Scope: general
• Reference category: n/a – enters as a continuous time-decay regressor, typically inside an exponential of the form 1 + a * exp(-k * T_FIRSTDOSE / 24) where the division by 24 converts the canonical hours to days so that the published rate constant k retains its per-day units.
• Source aliases:
  • TAF (Eechoute 2012 imatinib; defined by Yang 2025 Table 1 abbreviation list as ‘TAF time from first dose administration (h)’).
• Example models: Eechoute_2012_imatinib.R (drives a joint exponential decay of the absorption rate constant, ka = 0.699 * (1 + 1.18 * exp(-0.0256 * T_FIRSTDOSE / 24)), and of relative bioavailability, F = 1 + 0.482 * exp(-0.0256 * T_FIRSTDOSE / 24), with a shared 0.0256/day rate constant giving a ~27-day half-life for the drift; both quantities fall towards their asymptotes over the first months of continuous imatinib therapy in GIST), Willmann_2024_elinzanetant.R, Bienczak_2016_nevirapine.R.
• Notes: Carried in hours to match the units$time = "h" convention of the PK models that use it, so a model whose published decay constant is expressed per day must divide by 24 at the call site (visible in the example above) rather than storing days in the column. Distinct from T_ENTRY (a per-subject offset locating study entry on a shared integration axis, not a treatment-duration clock), from CYCLE (an integer dose/cycle counter rather than a continuous time), and from MONTH1 / DAY14 (binary landmark indicators derived from a treatment-duration clock rather than the clock itself). For a single-course simulation whose first dose is at time = 0, T_FIRSTDOSE equals time; the column exists so that records beginning mid-course, or subjects with a pre-study treatment history, can be represented honestly.

T_CONMED (canonical for time of the first perpetrator (concomitant) drug administration relative to model time zero)

• Description: Time of the first administration of a co-administered perpetrator drug, measured on the model integration axis (model time zero is the first dose of the modelled drug). Time-fixed per subject. Used to derive the time since the perpetrator dose inside model() as t - T_CONMED, which selects time-interval-specific drug-drug-interaction effects without carrying a time-varying covariate column. Member of the auto-approved T_<event> family.
• Units: h
• Type: continuous
• Scope: general
• Reference category: n/a – enters as the offset of the perpetrator-time axis. Reference values observed: 2 h for every subject in Nassar 2022 (perpetrator given 2 h after the start of the midazolam bolus plus infusion).
• Source aliases:
  • none as a column – Nassar 2022’s NONMEM control stream hard-codes the perpetrator windows on its midazolam-time axis (TIME.GT.2.AND.TIME.LE.3 for the first voriconazole hour), so T_CONMED = 2 and TIME - 2 is the paper’s “time relative to perpetrator administration”.
• Example models: Nassar_2022_midazolam.R (locates the time-interval fractional changes in midazolam clearance for oral and intravenous voriconazole, efavirenz and rifampicin, e.g. tconmed <- t - T_CONMED and e_conmed_efv_cl_t2to3 * (tconmed > 2) * (tconmed <= 3); the windows are left-open, right-closed as in the control stream).
• Notes: General scope. Pair it with the indicator of the perpetrator it times (CONMED_<INN> or a qualified member); it is irrelevant when every perpetrator indicator is 0. For a repeatedly dosed perpetrator it marks the FIRST dose, so windows after later doses are expressed relative to the first (Nassar 2022 rifampicin windows (22, 24] … (30, 34] h, second dose at 24 h). Distinct from T_FIRSTDOSE (time since the first dose of the modelled drug, a time-varying clock) and from T_INCISION (a non-drug event on the same time-fixed-offset pattern, which it mirrors).

T_LASTDOSE (canonical for nominal time elapsed since the most recent dose)

• Description: Nominal (protocol-scheduled) time elapsed since the subject’s most recent administration of the modelled drug. Resets to 0 at each dosing event, in contrast to T_FIRSTDOSE, which measures time since the start of the treatment course and never resets. Exists for PD-only models that carry no dosing records – an algebraic exposure-response model driven by an externally supplied concentration column has no dose events for rxode2’s native tad to key off, so the time-after-dose clock has to be supplied as a covariate. Models that DO carry dosing records must use tad and must not duplicate it into this column.
• Units: h
• Type: continuous
• Scope: general
• Reference category: n/a – may enter continuously, or (as in the founding model) be decomposed inside model into a set of mutually exclusive nominal-timepoint class effects when the source paper estimates a separate intercept shift at each scheduled sampling time.
• Source aliases:
  • nominal time after dose / Time_j (Mukker 2026 Equation 1 and Table 2; the C-QTc analysis dataset carried the nominal post-dose hour of each triplicate-ECG timepoint, with levels 0, 1, 2 and 3 h).
• Example models: Mukker_2026_tuvusertib_QTcF.R, Mukker_2026_tuvusertib_HR.R (the four Garnett-model nominal-timepoint intercept shifts are selected by binning T_LASTDOSE to its nearest scheduled level inside model(): i0 <- (T_LASTDOSE < 0.5), i1 <- (T_LASTDOSE >= 0.5) * (T_LASTDOSE < 1.5), and so on).
• Notes: General scope because time since the most recent dose is a paper-independent quantity. Carried in hours to match the units$time = "h" library convention, like its T_<event> siblings T_FIRSTDOSE, T_ECMO and T_POST_HEMODIAL. Carries the nominal scheduled time, not the actual elapsed time, whenever the source paper estimates class effects at nominal timepoints – the two differ by protocol-window slippage and the class-effect model is defined on the nominal grid. Distinct from T_FIRSTDOSE (treatment-course clock, monotonically increasing, does not reset), from TCLOCK (wall-clock hour of day feeding a 24-hour periodic function – note that a nominal-timepoint class effect may be interpreted by its authors as capturing diurnal variation, as in Mukker 2026, without being a periodic function of clock time), and from OCC / ooc<n> (which decompose multi-occasion sampling for IOV and carry per-occasion eta slots rather than typical-value intercept shifts).

T_ENTRY (canonical for per-subject study-entry time on the model integration axis)

• Description: Per-subject study-entry time, expressed on the same integration-time axis the model is solved on (years for AD disease-progression models). The dataset’s TIME column carries the integration-time variable (with TIME = 0 the integration origin, typically well before any subject’s disease onset); T_ENTRY records, per subject, the integration-time value at which the subject’s first observation falls. Used in models whose pre-study and post-study time semantics differ – e.g., a disease-progression activation function defined on global disease time plus a placebo-effect term defined on time-since-study-entry need access to both reference clocks within the same model block.
• Units: year (or whatever time unit the model’s units$time field declares)
• Type: continuous
• Scope: specific
• Reference category: n/a – T_ENTRY is a per-subject time-offset covariate that anchors the time-since-study-entry clock used by post-entry-only model terms (placebo / learning effects, study-design transient drops).
• Source aliases: none standardized; this canonical originates with the Delor 2013 AD disease-progression extraction. The source paper’s NONMEM dataset handles the global / study-time split internally (the dataset is staged with TIME = study time and the model uses a fixed offset to align with DOT); when porting to rxode2 / nlmixr2 the per-subject offset is exposed as a covariate so simulation event tables can carry it explicitly. Berkhout 2015 EPIC placebo-arm osteoporosis QSP uses the same pattern with the source NONMEM column STDA (start day of placebo treatment relative to menopause onset) mapped to T_ENTRY.
• Example models: Delor_2013_alzheimer.R (placebo-term clock: t_pl_raw <- time - T_ENTRY; post_entry <- t_pl_raw > 0; placebo_term <- pl_indiv * (1 - exp(-kpl_indiv * t_pl_raw * post_entry)) * post_entry), Berkhout_2015_osteoporosis_placebo_qsp.R (PCa placebo function gated by t >= T_ENTRY with delayed-onset / slow-offset dip in the RANK-RANKL-OPG occupancy factor).
• Notes: Scope: specific because the variable is paper-domain-bound – it only makes sense for models that operate on a global disease-time axis distinct from per-subject study-entry timing. For a typical-value reproduction the user supplies T_ENTRY per subject in the simulation event table; a reasonable construction is T_ENTRY = DOT_individual + a few years of established disease so the patient is observed during disease progression (see the Delor 2013 vignette for the construction used to reproduce Figures 2-4).

T_SCAN_TO_DOSE (canonical for the delay between the baseline assessment and the first dose)

• Description: Per-subject elapsed time between the baseline disease assessment that defines the model’s time origin (a screening tumour scan, a baseline biomarker draw) and the first administration of the modelled treatment. Time-fixed per subject and non-negative. Used by disease-progression / tumour-growth models whose integration axis is anchored at the baseline assessment rather than at the first dose, so that the drug-driven arm of the model must be held inert over the untreated screening-to-dosing interval – typically via a max(0, t - T_SCAN_TO_DOSE) term or an equivalent t >= T_SCAN_TO_DOSE gate on the treated sub-state. The untreated arm (natural disease progression) runs from t = 0 regardless.
• Units: day (or whatever time unit the model’s units$time field declares; carry it in the model’s own time unit so the published equation’s constants keep their units).
• Type: continuous
• Scope: general
• Reference category: n/a – a per-subject time offset, not a covariate effect coefficient. T_SCAN_TO_DOSE = 0 recovers the no-delay case (dosing on the day of the baseline assessment) and is the sensible default for a typical-value simulation when the source paper does not report the distribution.
• Source aliases:
  • t1 – Zou 2020 Methods Eq 3 (time of the first leuprorelin dose on an axis anchored at the baseline PSA draw).
  • delay – Chatterjee 2016 main-article Methods, tumor size NLME model structure (“‘delay’ the delay between baseline and the first dose”).
• Example models: Chatterjee_2016_pembrolizumab.R (gates the first-order decay of the treatment-sensitive tumour sub-state: d/dt(shrink) <- -kdeath * shrink * (t >= T_SCAN_TO_DOSE), reproducing the published exp(-kdeath * max(0, time - delay)) term, while the resistant sub-state grows from t = 0), Zou_2020_leuprorelin.R (switches the drug-sensitive PSA sub-state from pre-treatment growth kge_sens to leuprorelin kill kse at time >= T_SCAN_TO_DOSE, reproducing Zou 2020’s t_s = min(t, t1), t_k = max(0, t - t1); source symbol t1, time of the first leuprorelin dose, measured from the baseline PSA draw).
• Notes: Well-formed member of the auto-approved T_<event> canonical family. Distinct from T_FIRSTDOSE, which is a time-VARYING treatment-duration clock measured forward from the first dose; T_SCAN_TO_DOSE is a time-FIXED scalar offset locating the first dose on an axis whose origin is the baseline assessment. Also distinct from T_ENTRY, which locates a subject’s study entry on a shared global disease-time axis. A source paper may report this quantity as an estimated population parameter (a pharmacologic onset delay) rather than as data; that is a different construct and belongs in ini(), not here – Chatterjee 2016 tested exactly such an estimated onset delay, dropped it for high shrinkage, and retained only the data-derived scan-to-dose offset that this column carries.

TCLOCK (canonical for wall-clock time-of-day covariate feeding a 24-hour periodic (circadian) function)

• Description: Wall-clock time of day at each observation, expressed on a 0-24 hour scale. Distinct from the integration-time axis time (which carries time after first dose or another simulation-anchored reference) – TCLOCK records the local hour of day at which a sample was drawn / an observation was recorded so a 24-hour periodic function of the form cos(2*pi * (TCLOCK + 24 - Tmax) / 24) can be evaluated at each record. Time-varying per record if samples are drawn at multiple times of day within a subject; per-subject time-fixed for study designs whose samples are always drawn at a consistent local time (e.g., morning pre-dose).
• Units: hour of day (0-24)
• Type: continuous
• Scope: general
• Reference category: n/a – enters as an argument to a periodic function whose period is 24 hours. The natural per-model reference is the time-of-maximum parameter Tmax in that periodic function; setting TCLOCK = Tmax collapses the circadian factor to its peak, and TCLOCK = Tmax + 12 (mod 24) collapses it to its trough.
• Source aliases:
  • tclock / clocktime – Hoefman 2021 (Hoefman_2021_asp8232.R) equation 3 notation for the wall-clock time argument of the eGFR CysC circadian amplitude term.
• Example models: Hoefman_2021_asp8232.R (drives the eGFR CysC circadian oscillation (1 + Ampli * cos(2*pi * (TCLOCK + 24 - tmax) / 24)) with Ampli = 0.0783 (theta_32) and tmax = 10.5 h (theta_33)).
• Notes: General scope because any circadian / diurnal PD model whose within-day rhythm is parameterised by a periodic function of the observation’s clock time will need this canonical – the concept generalises across drugs, biomarkers, and disease areas. Distinct from time (the integration-time axis of the model, on which ODE progression is solved) and from T_ENTRY (per-subject study-entry offset used to align a per-subject placebo clock to a global disease-time clock). Data-assembly helper: if the source dataset carries a time-of-day column encoded as HH:MM:SS or minutes-since-midnight, convert to decimal hours before populating TCLOCK. If a subject’s records are drawn at a consistent time of day and TCLOCK is unknown per record, a sensible default is TCLOCK = Tmax (peak of the circadian wave) or the paper’s stated typical sampling time.

DOSETIME_EVENING (canonical for evening dose-administration-time indicator in a clinical chronopharmacokinetic covariate model)

• Description: 1 = the dose on this record was administered during the evening clock-time window; 0 = the dose was administered outside it (morning or afternoon). A per-DOSE-RECORD property of when the drug was taken, not a per-subject attribute and not a property of when a sample was drawn. Use it when a source paper bins wall-clock dosing time into windows and estimates a shift in an absorption parameter for one of those bins. The exact window boundaries are paper-defined and MUST be recorded in covariateData[[DOSETIME_EVENING]]$notes, because “evening” is not a standardised interval – Cirincione 2018 defines it as 5:01 pm to 12:00 am, against a reference that pools morning (12:01 am to 11:00 am) and afternoon (11:01 am to 5:00 pm).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (dose administered outside the paper’s evening window). State which windows the reference pools: a paper with a three-level dosing-time factor that estimates only the evening level folds the other two into the reference, and a reader cannot recover that from the column name.
• Source aliases:
  • AMPM – the Cirincione 2018 $INPUT column, a three-level factor annotated in the control stream as 1:12am-11am, 2:5pm-12am, 3:11am-5pm. Decompose to DOSETIME_EVENING = as.integer(AMPM == 2), following the register’s standing rule that a multi-level categorical is carried as binary indicators. The control stream also derives an afternoon indicator (QAMPM3), but its effect is not evaluable in the final models and is fixed to zero, so no DOSETIME_AFTERNOON column is minted here; add one only when a paper actually estimates it.
• Example models: Cirincione_2018_apixaban.R, Cirincione_2018_apixaban_asian_subgroups.R (fractional-change effect on the absorption rate constant, ka = theta1 * (1 + e_dosetime_evening_ka * DOSETIME_EVENING) with e_dosetime_evening_ka = -0.433: apixaban taken in the evening is absorbed 43 percent more slowly than the same dose taken in the morning or afternoon. The authors judged this clinically unimportant given twice-daily dosing, but it is a well-estimated effect – -0.433 +/- 0.0236 – and it is retained in all three reported final models), Kuchimanchi_2022_diroximelFumarate.R (source column PM, 1 = evening dose of a twice-daily regimen; fractional changes on both metabolite absorption rate constants, -0.592 on monomethyl fumarate and -0.267 on 2-hydroxyethyl succinimide, an additive 1.96 h absorption lag on 2-hydroxyethyl succinimide, and its own residual SDs, all fixed at phase I estimates. The paper defines no clock-time window: the flag marks the evening dose of the BID schedule. The control stream zeroes it outside study A102, whose serial post-evening-dose profiles were the only ones considered informative), Lalovic_2020_lemborexant.R (lemborexant; the source calls it ‘bedtime’ / ‘nighttime’ dosing and gives no clock-time window – it is the evening dose of an insomnia drug taken at bedtime in the phase 2/3 trials, versus daytime dosing in the extensively sampled phase 1 studies. Multiplies the zero-order absorption duration D1 by 2.33, fixed from the base model).
• Notes: Promoted to general scope with the Kuchimanchi 2022 diroximel fumarate model, the second independent paper to use the concept (Cirincione 2018 bins wall-clock dosing time; Kuchimanchi 2022 flags the evening dose of a twice-daily schedule). Record the paper’s own definition of ‘evening’ per model. Three boundaries are worth stating. First, this is not TCLOCK: TCLOCK is a continuous wall-clock hour feeding a 24-hour periodic function evaluated at an OBSERVATION, whereas DOSETIME_EVENING is a binary bin attached to a DOSE. A paper that resolves a continuous circadian rhythm should use TCLOCK, not a coarser indicator. Second, it is not PERIOD_ACTIVE, which records the light-dark phase of a nocturnal laboratory rodent’s dosing time; species chronobiology inverts the mapping from clock time to behavioural phase, so the two must not be pooled. Third, it is not a formulation, adherence or observed-dosing indicator: the covariate says WHEN the dose was taken, not WHETHER it was taken or WHAT was taken, so it is distinct from the FORM_<drug>_<formulation> family and from the unobserved-dose indicators used in the adherence-adjustment models.

OUTPATIENT (canonical for outpatient (vs inpatient) treatment-setting indicator)

• Description: 1 = the dose was taken while the subject was treated as an outpatient; 0 = the subject was hospitalized (inpatient) at the time of the dose. Time-varying at the dose record: in long-treatment trials a subject can move from inpatient to outpatient care.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (inpatient / hospitalized).
• Source aliases:
  • OUT – Wang 2020 Figure 1B label OUT dose on F1 and Table 4 row F1,out.
• Example models: Wang_2020_delamanid.R (multiplicative 1.09 on the relative bioavailability of delamanid in MDR-TB patients; patients were hospitalized all, part or none of the time depending on the trial).
• Notes: A treatment-setting property, not a disease-severity marker: distinct from DIS_CRITILL (ICU / critical illness) and from HOSPRA (hospitalization for a pulmonary exacerbation at a spirometry visit). The founding paper attributes the effect to differences in administration conditions (food intake around dosing) rather than to illness severity.

FORM_ELZ_HARDGEL (canonical for the elinzanetant hard-gelatin capsule formulation indicator)

• Description: 1 = the oral elinzanetant dose was given as the hard-gelatin capsule; 0 = the dose was given as the soft-gel capsule, as the aqueous oral suspension, or intravenously. Per-dose-record indicator distinguishing the hard gel capsule used in the early phase I food-effect study and in the RELENT-1 proof-of-mechanism study from the soft-gel capsule that was carried into the phase III programme.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (soft-gel capsule, aqueous suspension and intravenous solution pooled; source FORM levels 1, 2 and 4).
• Source aliases:
  • FORM3 – used in Willmann_2024_elinzanetant.R. Willmann 2024 Data S1 defines FORM3 = 1 when FORM = 3, else 0, where the source data file’s FORM column codes the formulation and level 3 is the hard gel capsule.
• Example models: Willmann_2024_elinzanetant.R (multiplicative effects on absorption only: ka *= (1 + e_form_elz_hardgel_ka * FORM_ELZ_HARDGEL) with e_form_elz_hardgel_ka = -0.264, a 26% lower absorption rate constant, and tlag *= (1 + e_form_elz_hardgel_tlag * FORM_ELZ_HARDGEL) with e_form_elz_hardgel_tlag = 0.548, a 55% longer lag time; the extent of absorption is unaffected).
• Notes: Scope: specific because the reference level pools three elinzanetant-protocol formulations rather than expressing a general pharmaceutical contrast. Distinct from FORM_CAPSULE (capsule vs non-capsule oral liquid), which cannot express this comparison because both the test and the reference formulation here are capsules – the contrast is hard-gelatin shell versus soft-gel shell. If a later extraction needs the same hard-gel-versus-soft-gel contrast for a different drug, promote a general FORM_HARDGEL canonical and record this entry as an alias rather than adding a second drug-specific name.

T_TRT (canonical for total duration of study treatment received)

• Description: The total length of the treatment course a subject actually received, from the first dose to the last, in the source paper’s reporting unit. A per-subject summary of the whole course - NOT an elapsed-time-since-event regressor like T_FIRSTDOSE, and not a per-record covariate. Member of the T_<event> family in the same “total duration of a named event” sense as T_CPB (total cardiopulmonary bypass duration), rather than the “time since” sense of T_ECMO or T_DIAG_DIAB. Its usual role is in an outcome or exposure-response model, where it separates patients who completed the planned course from those who stopped early for futility, toxicity, or withdrawal.
• Units: weeks. Document per-model via covariateData[[T_TRT]]$units when a paper reports days, months, or cycles.
• Type: continuous
• Scope: general
• Reference category: n/a – supplied as a duration. Papers commonly consume it DICHOTOMIZED against the protocol’s full planned course; binarize inside model() so the threshold stays auditable, as the ALP entry prescribes for above-ULN indicators. Full-course values observed: 48 weeks (Riggs 2012 chronic HCV genotype 1) and 24 weeks (Riggs 2012 chronic HCV genotype 2/3). Dichotomization is not the only consuming form: Kemal 2026 instead keeps the duration continuous and passes it through a saturable Emax-type term T_TRT / (ET50 + T_TRT), motivated by the expectation that the effect of time on response rate persists only initially. Both forms are applied inside model(), which is what keeps the column itself form-neutral.
• Source aliases:
  • treatment duration – Riggs 2012 Methods and Table II row label.
• Example models: Riggs_2012_albinterferon_svr_gt1.R (dichotomized as trtReduced <- (T_TRT < 48), giving a logit-scale effect of -2.235680 on sustained virologic response - the largest single coefficient in that model), Riggs_2012_albinterferon_svr_gt23.R (same covariate against the shorter genotype 2/3 course, trtReduced <- (T_TRT < 24), effect -2.612296), Kemal_2026_nemtabrutinib_bor.R (carried CONTINUOUSLY in days through the saturable term T_TRT / (et50 + T_TRT) with et50 fixed at 200 days, logit-scale maximum effect 8.7637 on best overall response).
• Notes: The dichotomization threshold is protocol-specific and can differ between two models extracted from the SAME paper, as the Riggs 2012 pair shows - which is a good reason to carry the duration continuously in the column and apply the cut-point in model rather than shipping a pre-computed binary. Beware the confounding this covariate usually carries: a short course is very often a consequence of the outcome being modelled (early discontinuation for lack of efficacy or for an adverse event), so a large negative coefficient on treatment duration in an efficacy model is not evidence that longer treatment causes response. Record in covariateData[[T_TRT]]$notes what the source says about that confounding - Riggs 2012, for instance, argues explicitly that exposure and duration are not confounded in its SVR model because Cavg had approached steady state before the futility assessments. Distinct from T_FIRSTDOSE (time elapsed since the first dose, a within-course time axis) and from TRT / TRT_PHASE (which treatment or which phase, not how long).

T_PREVOBS (canonical for time elapsed since the subject’s previous observation occasion)

• Description: Time between the current observation record and that subject’s immediately preceding observation record. A per-record, subject-specific gap length – NOT time since a fixed study origin. It appears in discrete-transition / Markov disease-progression models, where the probability and magnitude of a score change between two consecutive assessments naturally scale with how long the interval was, and in any analysis whose visit schedule is irregular enough that the transition interval is not implied by the visit number.
• Units: day. Document per-model via covariateData[[T_PREVOBS]]$units when a paper reports hours, weeks, or cycles.
• Type: continuous
• Scope: general
• Reference category: n/a – entered uncentred in the founding models. At a subject’s first modelled occasion the gap is measured back to the baseline/admission assessment.
• Source aliases:
  • TLS – the NONMEM $INPUT name in BOTH Karlsson 2010 control streams. Read the aliased column, not the paper’s legend: the Supplement B dataset legend defines TLS as “Time since start of study (i.e. time since stroke) in days”, and the tabulated data falsify that. In the NIHSS excerpt subject 2 has DAY 8 -> TLS 7, DAY 33 -> TLS 25 (= 33 - 8), DAY 62 -> TLS 29 (= 62 - 33); in the BI excerpt subject 1 has DAY 29 -> TLS 22, DAY 60 -> TLS 31, DAY 89 -> TLS 29, i.e. non-monotonic in DAY, which a time-since-stroke clock cannot be.
• Example models: Karlsson_2010_acute_stroke_nihss.R (on the probability of not reaching a fully recovered score, e_t_prevobs_p_notmax = 0.0277 per day), Karlsson_2010_acute_stroke_bi.R (on the logit-scale relative-decline magnitude, e_t_prevobs_decl = -0.0332 per day; founding examples).
• Notes: Member of the T_<event> family in the “time since” sense of T_ECMO / T_LASTDOSE, not the “total duration” sense of T_CPB / T_TRT. The distinction that matters most in practice is against the study-origin clock: a paper can carry BOTH, and Karlsson 2010 does – its Barthel Index model puts T_PREVOBS on the relative decline and time-since-stroke on the not-maximum probability, and Table II lists “Influence of time since previous observation” and “Influence of time since baseline” as separate rows. Where the study-origin clock coincides with the model’s integration axis (as it does there, the source stream setting TSS = DAY), encode it as rxode2’s time rather than duplicating it as a covariate column, and reserve T_PREVOBS for the gap. Distinct from T_FIRSTDOSE and T_LASTDOSE, which are keyed to dosing events rather than to observation records, and from TAD; a drug-free disease-progression model has no dose events at all yet still needs this covariate. Ratified canonically alongside the Karlsson 2010 acute-stroke disease-progression extraction.

MIX_PDI (canonical for binary mixture-model class indicator: PDI inhibition-of-synthesis vs PDS stimulation-of-elimination thrombocytopenia mechanism)

• Description: Per-subject latent mixture-model class indicator from the Tsuji 2017 linezolid platelet PKPD model. 1 = subject classified to the PDI mechanism (linezolid inhibits platelet synthesis via a linear effect on the proliferation rate of the formation compartment); 0 = subject classified to the PDS mechanism (linezolid stimulates platelet elimination via an Emax effect on the circulating-compartment first-order loss rate). Not a measured clinical covariate – the mixture assignment is a posterior class assignment from the NONMEM $MIXTURE block (Tsuji 2017 Equation 8 and Methods, “Mixture model” subsection).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (PDS = stimulation of elimination; the 3% minority class in the source cohort).
• Source aliases: MIXTURE – NONMEM $MIXTURE block class index in the Tsuji 2017 estimation run (component 1 = PDI, component 2 = PDS); the binary indicator is MIX_PDI = as.integer(MIXTURE == 1).
• Example models: Tsuji_2017_linezolid.R (gates the drug-effect term that enters the platelet ODE chain: MIX_PDI = 1 activates SLOPE * Cc on the formation rate, MIX_PDI = 0 activates SMAX * Cc / (SC50 + Cc) on the circulating-compartment elimination rate).
• Notes: The population probability of MIX_PDI = 1 is the estimated population mixture fraction FPOP_inhibit = 0.969 (Tsuji 2017 Table 2; 95% CI 0.867-1.00, 78/80 patients in the source dataset). For typical-value simulation set MIX_PDI = 1 (dominant clinical phenotype, slower 2-week nadir; Figure 5 left panel) or MIX_PDI = 0 (rare immune-mediated-like phenotype, faster 2-day nadir; Figure 5 right panel). For population simulation, draw MIX_PDI ~ Bernoulli(0.969) per subject. Scope: specific because the binary semantics are tied to Tsuji 2017’s two-mechanism platelet model; mixture indicators from future papers that share the same biological dichotomy may extend this entry, but mixture indicators from unrelated dichotomies should register a new canonical name (e.g., MIX_FAST_ELIM for a fast/slow eliminator mixture).

MIX_FAST_ELIM (canonical for binary mixture-model class indicator: fast (high-CL) vs slow (low-CL) eliminator subpopulation)

• Description: Per-subject latent mixture-model class indicator from a $MIXTURE NONMEM block describing a bimodal apparent-clearance distribution. 1 = subject classified to the high-CL / fast-eliminator subpopulation; 0 = subject classified to the low-CL / slow-eliminator subpopulation. Not a measured clinical covariate – the mixture assignment is a posterior latent-class index from the NONMEM $MIXTURE block. One assignment per subject, time-fixed.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (low-CL / slow-eliminator subpopulation).
• Source aliases:
  • MIXTURE – NONMEM $MIXTURE block class index in the Morris 2011 estimation run (component 1 = high-CL with P = 0.251, component 2 = low-CL with P = 0.749); the binary indicator is MIX_FAST_ELIM = as.integer(MIXTURE == 1).
• Example models: Morris_2011_telapristone.R (selects the telapristone CL/F typical value between the high-CL anchor exp(lcl_pop1) = 11.6 L/h and the low-CL anchor exp(lcl_pop2) = 3.34 L/h; the shared IIV etalcl ~ 0.200 is applied multiplicatively to whichever typical the mixture-class selects), Wilkins_2011_isoniazid.R (gates the typical apparent CL/F via the power-form covariate effect e_mix_fast_elim_cl on CL/F: cl = exp(lcl + e_mix_fast_elim_cl * MIX_FAST_ELIM + etalcl) * (WT/70)^0.75 * (1 + e_hiv_pos_cl * HIV_POS) with lcl = log(9.70) (slow-eliminator reference) and e_mix_fast_elim_cl = log(21.6/9.70) approx 0.8005), Cazaubon_2019_mitotane.R (Monolix latent covariate lcat, modality 2 = ‘ultrafast metabolizer’ with plcat_2 = 0.115; MIX_FAST_ELIM = as.integer(lcat == 2); log-scale CL shift e_mix_fast_elim_cl = 1.12, i.e. 3.06-fold higher CL).
• Notes: Pre-named by the MIX_PDI register entry’s Notes block (“mixture indicators from unrelated dichotomies should register a new canonical name (e.g., MIX_FAST_ELIM for a fast/slow eliminator mixture)”). Population probability of MIX_FAST_ELIM = 1 is recorded per-model in covariateData[[MIX_FAST_ELIM]]$notes; in Morris 2011 it is 0.251 (Table II, “Probability” row, RSE 61.0%). The Morris 2011 Discussion attributes the bimodal CL distribution to polymorphic CYP3A5 (functional CYP3A5 present in 10-40% of Caucasians, ~50% of African Americans, ~33% of Asians) but does not test the genotype hypothesis directly; the mixture indicator is therefore a latent-class label, not a CYP3A5-genotype indicator. For typical-value simulation set MIX_FAST_ELIM = 1 (fast eliminator; lower steady-state exposure, ~12 h elimination half-life per Morris 2011 Results) or MIX_FAST_ELIM = 0 (slow eliminator; higher steady-state exposure, ~35 h half-life). For population simulation, draw MIX_FAST_ELIM ~ Bernoulli(0.251) per subject. Scope: general because the fast/slow-eliminator dichotomy is a recurring popPK concept across drugs metabolized by polymorphic enzymes (CYP2C92/3, CYP2C192, CYP3A53, NAT2 slow/fast acetylators); future popPK papers fitting a $MIXTURE on CL should re-use this canonical rather than introducing per-drug-specific MIX_<drug>_FAST names. The reference category (= 0 = slow eliminator) is chosen so the binary numerically matches the paper’s mixture-indicator orientation.

MIX_LARGE_BASE (canonical for binary mixture-model class indicator: larger-baseline vs smaller-baseline tumor-lesion subpopulation)

• Description: Per-subject latent mixture-model class indicator from the Schindler 2017 imatinib-GIST liver-metastases joint tumor-dynamics model. 1 = subject classified to the larger-baseline subpopulation (typical lesion 1 baseline 76.6 mm MTD / 161 mL Vactual / 187 mL Vellipsoid; estimated population probability Ppop1 = 0.348 in Schindler 2017 Table 2 row ‘Ppop1’); 0 = subject classified to the smaller-baseline subpopulation (typical lesion 1 baseline 20.9 mm MTD / 3.45 mL Vactual / 3.93 mL Vellipsoid). Not a measured clinical covariate – the mixture assignment is the per-subject latent-class index from a NONMEM $MIXTURE block describing the observed bimodal distribution of baseline tumor sizes (Schindler 2017 Methods, “Maximum transaxial diameter, actual volumes, and ellipsoidal volume models” subsection: “Semiparametric distributions and mixture models were investigated to describe the observed bimodal distribution in baseline MTD, Vactual, and Vellipsoid”). One assignment per subject is shared across the three size models (MTD, Vactual, Vellipsoid) and across the subject’s up-to-two lesions; density (D0) is not mixture-indexed.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (smaller-baseline subpopulation; majority class at 65.2 % of the source cohort).
• Source aliases: MIXTURE – NONMEM $MIXTURE block class index in the Schindler 2017 estimation run (component 1 = larger-baseline subpop with P = 0.348, component 2 = smaller-baseline subpop); the binary indicator is MIX_LARGE_BASE = as.integer(MIXTURE == 1).
• Example models: Schindler_2017_imatinib.R (gates the per-lesion typical baselines S0_<metric>_l<k>_typ between subpopulation-1 and subpopulation-2 anchors, and selects between subpopulation-specific IIV etas etalS0_<metric>_pop1 and etalS0_<metric>_pop2 whose log-scale variances differ across mixture classes).
• Notes: Schindler 2017 Methods reports 22 % of patients with one target lesion vs 39 % of patients with two target lesions were assigned to subpopulation 1, suggesting the larger-baseline class is enriched among multi-lesion patients. For typical-value simulation set MIX_LARGE_BASE = 1 to reproduce the larger-baseline phenotype (dominates Figure 2 right panel typical-individual trajectory) or MIX_LARGE_BASE = 0 to reproduce the smaller-baseline phenotype. For population simulation, draw MIX_LARGE_BASE ~ Bernoulli(0.348) per subject. Scope: specific because the binary semantics are tied to Schindler 2017’s two-class baseline-tumor-size mixture; future tumor-burden mixture models that share the same large-vs-small baseline dichotomy may extend this entry, while mixture indicators from unrelated dichotomies (e.g., fast / slow tumor-growth phenotypes) should register a new canonical name (MIX_FAST_GROW, etc.).

MIX_LARGE_RUV (canonical for binary mixture-model class indicator: larger-additive-RUV vs smaller-additive-RUV subpopulation)

• Description: Per-subject latent mixture-model class indicator from the Kappelhoff 2005 ritonavir popPK model. 1 = subject classified to the minority larger-additive-RUV subpopulation P2 (additive residual SD = 0.199 mg/L; 35.2 % of the source cohort); 0 = subject classified to the majority smaller-additive-RUV subpopulation P1 (additive residual SD = 0.0600 mg/L; 64.8 % of the source cohort). Both subpopulations share the same 15.4 % proportional residual error component and the same structural PK; only the additive RUV magnitude differs. Not a measured clinical covariate – the mixture assignment is the per-subject latent-class index from Kappelhoff 2005’s NONMEM $MIX block (paper Methods, “Basic pharmacokinetic model” subsection: “Subpopulations were estimated using the $MIX function in the control stream”).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (smaller-additive-RUV subpopulation P1; majority class at 64.8 % of the source cohort).
• Source aliases: $MIX class assignment – NONMEM $MIX block class index in the Kappelhoff 2005 estimation run (component 1 = P1 small-RUV with P = 0.648, component 2 = P2 large-RUV with P = 0.352); the binary indicator is MIX_LARGE_RUV = as.integer($MIX == 2).
• Example models: Kappelhoff_2005_ritonavir.R (gates the active additive residual SD inside model(): add_sd_eff <- addSd_p1 * (1 - MIX_LARGE_RUV) + addSd_p2 * MIX_LARGE_RUV; the proportional component is shared).
• Notes: Kappelhoff 2005 Discussion notes the mechanism behind the two-population residual structure could not be identified; the mixture was retained for goodness-of-fit (Delta-OFV = -48, P < 0.001 vs single-population RUV). For typical-value simulation set MIX_LARGE_RUV = 0 (dominant subpopulation). For population simulation, draw MIX_LARGE_RUV ~ Bernoulli(0.352) per subject. Scope: specific because the binary semantics are tied to Kappelhoff 2005’s two-class additive-RUV mixture; future popPK models that share the same large-vs-small additive-RUV dichotomy may extend this entry, while mixture indicators from unrelated dichotomies should register a new canonical name (e.g., MIX_LARGE_PROPRUV for a mixture on the proportional component instead).

MIX_LARGE_PROPRUV (canonical for binary mixture-model class indicator: larger-proportional-RUV vs smaller-proportional-RUV subpopulation)

• Description: Per-subject latent mixture-model class indicator for a two-class mixture on the PROPORTIONAL (constant-coefficient-of-variation) residual error component. 1 = subject assigned to the minority large-residual subpopulation; 0 = subject assigned to the majority small-residual subpopulation. Structural parameters, covariate effects and inter-individual variability are shared across the two classes; only the residual magnitude switches. Not a measured clinical covariate – the assignment is the per-subject latent-class index of a NONMEM $MIX / $MIXTURE block. The per-model covariateData[[MIX_LARGE_PROPRUV]]$notes must state the two residual magnitudes and the estimated class probability, because both are paper-specific.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (smaller-proportional-RUV subpopulation; the majority class in both founding designs).
• Source aliases: $MIX class assignment / $MIXTURE class assignment – the NONMEM mixture-block class index; the binary indicator is MIX_LARGE_PROPRUV = as.integer($MIX == <large-residual component>).
• Example models: Jonsson_2016_tanezumab.R (Jonsson 2016 Equation 2, Y = Yhat + eps1 if subpopulation 1 or eps2 if subpopulation 2; the base model used log transformation of both sides so both components are log-scale SDs and enter as lnorm(). Small-residual class 13% CV at probability 0.763 (Table 2 row ‘Mixture probability with low RSV’), large-residual class 54% CV at probability 0.237. Gated inside model() as expSdMix <- expSd_p1 * (1 - MIX_LARGE_PROPRUV) + expSd_p2 * MIX_LARGE_PROPRUV. The class-specific eta shrinkages Jonsson reports – CL 11%/10%, V1 15%/23%, VM 66%/71%, V2 57%/79% for the low/high classes – confirm the assignment is at the subject level rather than per observation).
• Notes: Registered under the name the MIX_LARGE_RUV entry reserves for exactly this dichotomy (“mixture indicators from unrelated dichotomies should register a new canonical name (e.g., MIX_LARGE_PROPRUV for a mixture on the proportional component instead)”). Distinct from MIX_LARGE_RUV, whose mixture is on the ADDITIVE residual component with a shared proportional component – the two are not interchangeable, because a model can in principle mix on either component or on both, and a reader must be able to tell which from the column name alone. A log-transform-both-sides residual counts as proportional for the purposes of this canonical: the source’s own wording (“approximately constant coefficient of variation model”) is the test, not whether the file encodes it as prop() or lnorm(). For typical-value simulation set MIX_LARGE_PROPRUV = 0; for population simulation draw MIX_LARGE_PROPRUV ~ Bernoulli(1 - p_small) per subject.

MIX_LAGGED_ABS (canonical for binary mixture-model class indicator: lagged-absorption vs no-lag absorption subpopulation)

• Description: Per-subject latent mixture-model class indicator from the Bonate 2004 apomine population PK model. 1 = subject classified to the lagged-absorption Group 1 subpopulation (estimable lag time and faster first-order absorption rate constant ka); 0 = subject classified to the no-lag Group 2 subpopulation (lag time fixed at 0 and slower ka). Not a measured clinical covariate – the mixture assignment is the per-subject latent-class index from a NONMEM $MIXTURE block (Bonate 2004 Methods “Base model development”, equation (1.4), and Results paragraph 1).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (Group 2 = no-lag minority class; 3 % of the source cohort).
• Source aliases: MIXTURE – NONMEM $MIXTURE block class index in the Bonate 2004 estimation run (component 1 = lagged-absorption Group 1 with P = 0.970, component 2 = no-lag Group 2 with P = 0.030); the binary indicator is MIX_LAGGED_ABS = as.integer(MIXTURE == 1).
• Example models: Bonate_2004_apomine.R (gates ka between two subpopulation-specific typical values and gates the lag time on the depot compartment between Group 1 = exp(ltlag) and Group 2 = 0; the shared etalka log-normal IIV scales whichever group’s ka is active for the subject), Frymoyer_2013_mycophenolic_acid.R (gates only the lag time on the depot compartment between Group 2 = exp(ltlag) = 1.96 h and Group 1 = 0; ka, F and disposition parameters are shared between the two classes per Frymoyer 2013 “Allowing other absorption parameters (ka and F) to vary for the two subpopulations by using an extended mixture model did not improve the overall model or predictions”), Kamp_2023_tacrolimus.R (gates only the lag time on the depot compartment of a meltdose-tacrolimus transit-absorption model between the lagged class = exp(ltlag) = 2.29 h and the no-lag class = 0, as tlag <- exp(ltlag) * MIX_LAGGED_ABS; ktr and the disposition parameters are shared between the two classes. The lagged class is the majority here: Kamp 2023’s Table 2 / Table S2 footnotes define the estimated $MIX proportion of 26% as the fraction without a lag, so MIX_LAGGED_ABS ~ Bernoulli(0.74). The $PK IF-block as printed in that paper’s Supplementary Data S1 maps the 26% onto the 2.29 h lag instead, contradicting both the footnotes and its own $THETA comments; the footnote reading is adopted because the paper’s own Figure 4C – the median of 1000 simulated profiles from the final model – sits on the baseline through 3 h and peaks near 6.1 h, which only the 74%-lagged reading reproduces), Wright_2025_colchicine.R (gates the lag time on the central compartment of a zero-order-absorption colchicine model between the lagged class = exp(ltlag) = 1.3 h and the no-lag class = 0, as tlag <- exp(ltlag) * MIX_LAGGED_ABS; D1, F1 and the disposition parameters are shared between the two classes. Not a $MIXTURE class index – Wright 2025 reports that a $MIX model for the bimodal lag “was not supported”, so the class was assigned deterministically in the analysis dataset as a LAGFLAG column set to 1 for the 17 of 78 NZ Gout Study participants whose steady-state Cmax was at or below their Cmin. The 1.3 h lag is itself fixed – it is the longest post-dose sampling time recorded in that study, not an estimate), Michalickova_2020_riociguat.R (gates the lag time on the depot compartment of a one-compartment riociguat + M1 parent-metabolite model between the lagged class = exp(ltlag) = 2.95 h and the no-lag class = 0, as tlag <- exp(ltlag) * MIX_LAGGED_ABS; ka and all disposition parameters are shared. Like Wright 2025 the class is not a $MIXTURE index: the six of 49 patients with unexpectedly high riociguat concentrations late after the last dose (ID 6, 10, 11, 19, 25, 28) were identified by visual inspection of the raw data and the lag was estimated for them only, so the observed prevalence is 6/49 = 0.122).
• Notes: The population probability of MIX_LAGGED_ABS = 1 is the estimated mixture fraction P(Group 1) = 1 / (1 + exp(P1)) = 0.970 (Bonate 2004 Table 3 with P1 = -3.47). For typical-value simulation set MIX_LAGGED_ABS = 1 to reproduce the dominant 97 % phenotype (lagged-absorption with ka = 1.77 /h and lag = 0.821 h); set MIX_LAGGED_ABS = 0 to reproduce the rare 3 % no-lag phenotype (ka = 0.361 /h, no lag). For population simulation, draw MIX_LAGGED_ABS ~ Bernoulli(0.970) per subject. Scope: specific because the binary semantics are tied to Bonate 2004’s two-class absorption mixture; future popPK papers that share the same lagged-vs-no-lag absorption-mixture dichotomy may extend this entry, while mixture indicators from unrelated dichotomies (e.g., fast vs slow elimination subpopulations) should register a new canonical name (MIX_FAST_ELIM, etc.). The class assignment need not come from a $MIXTURE block: Wright_2025_colchicine.R uses the same column for a lagged-vs-no-lag dichotomy that its authors assigned deterministically in the analysis dataset (from the observed Cmax <= Cmin pattern) after a $MIX mixture failed to support the bimodality. The covariate semantics – 1 = lagged-absorption class, 0 = no lag – are identical either way; what changes is only how a downstream user should draw it for simulation (a fitted mixture fraction versus an observed class prevalence, documented in each model’s covariateData[[MIX_LAGGED_ABS]]$notes).

MIX_HIGH_VP (canonical for binary mixture-model class indicator: high-peripheral-volume vs typical-peripheral-volume subpopulation)

• Description: Per-subject latent mixture-model class indicator from the Bonate 2004 apomine population PK model. 1 = subject classified to the high-peripheral-volume subpopulation (Bonate 2004 Study 2 healthy-male multiple-dose subjects with Vp many-fold higher than the rest of the cohort, modelled as a multiplicative shift Vp_i = TVV3 * theta_Study2 with theta_Study2 = 23.5); 0 = subject classified to the typical-Vp subpopulation. Not a measured clinical covariate – the mixture assignment was identified by Bonate 2004 from a bimodal empirical-Bayesian Vp histogram and back-fitted as a dichotomous study-membership multiplier (Bonate 2004 Results paragraph 1 and Discussion).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (typical-Vp majority class; 34 / 38 = 89.5 % of the source model-development cohort).
• Source aliases: STUDY2 – the paper modelled the high-Vp subgroup as a dichotomous indicator on Study 2 membership (the 4 healthy-male multiple-dose subjects in Study 2); the binary indicator is MIX_HIGH_VP = as.integer(STUDY == 2) for the source cohort.
• Example models: Bonate_2004_apomine.R (multiplicative log-additive effect on Vp: vp = exp(lvp + e_high_vp_vp * MIX_HIGH_VP + etalvp) with e_high_vp_vp = log(23.5) = 3.157).
• Notes: The population probability of MIX_HIGH_VP = 1 is 4 / 38 = 0.105 in the Bonate 2004 model-development set. For typical-value forward simulations set MIX_HIGH_VP = 0 (the recommended default); the paper notes that simulations with and without the multiplier showed minimal differences in concentration-time profiles. Set MIX_HIGH_VP = 1 only to reproduce the Bonate 2004 Study 2 healthy-male multiple-dose subgroup specifically. Scope: specific because the binary semantics are tied to Bonate 2004’s anomalous Study 2 high-Vp subgroup; future popPK papers that retain a similar dichotomous high-Vp subgroup multiplier may extend this entry, while peripheral-volume mixture indicators from unrelated dichotomies should register a new canonical name.

MIX_VAC_RELAPSE (canonical for binary mixture-model class indicator: tumor-relapse vs cure subpopulation after a single-dose immunotherapy / vaccine)

• Description: Per-subject latent mixture-model class indicator from the Parra-Guillen 2013 cancer-vaccine tumor-dynamics model. 1 = subject classified to the relapser subpopulation (transient vaccine-elicited inhibitory signal, SVAC degradation rate k2 = k1 estimated; tumor regrowth observed after initial response; estimated population probability 1 - P(1) = 0.156 in Parra-Guillen 2013 Table I row ‘P(1) = 0.844 FIX’); 0 = subject classified to the responder / cure subpopulation (permanent vaccine-elicited inhibitory signal, SVAC degradation rate k2 = 0 FIX; complete tumor regression maintained for the duration of the experiment). Not a measured clinical covariate – the mixture assignment is the per-subject latent-class index from a NONMEM $MIXTURE block (Parra-Guillen 2013 Methods ‘Data Evaluation, Biological Assumptions and Mathematical Model’ subsection bullet (e); Discussion ‘A challenge when using a mixture model’ paragraph). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (responder / cure subpopulation; majority class at 84.4 % in the Parra-Guillen 2013 source cohort and assumed equal in the Medina-Echeverz 2014 IL-12 applicability cohort).
• Source aliases: MIXTURE – NONMEM $MIXTURE block class index in the Parra-Guillen 2013 estimation run (component 1 = responder / cure subpop with P(1) = 0.844 FIX, component 2 = relapser subpop); the binary indicator is MIX_VAC_RELAPSE = as.integer(MIXTURE == 2) (orientation flipped from the NONMEM convention so the named state corresponds to the deviation from the dominant cure phenotype, following the Schindler 2017 / Tsuji 2017 MIX_* naming pattern).
• Example models: ParraGuillen_2013_cyaaE7.R, ParraGuillen_2013_il12.R (gates the SVAC degradation rate inside Eq. 3: k2 <- MIX_VAC_RELAPSE * k1 so MIX_VAC_RELAPSE = 0 freezes the inhibitory signal indefinitely (cure) and MIX_VAC_RELAPSE = 1 makes it decay at rate k1 (relapse). The same indicator is reused across both drug files because the biological dichotomy is identical (responders vs relapsers to immunotherapy) and the paper fixed P(1) = 0.844 from CyaA-E7 when re-fitting IL-12).
• Notes: P(1) = 0.844 was obtained from a sub-analysis of only the 4 / 7 / 11-day post-inoculation vaccine-administration groups (where tumor-size-induced resistance had not yet emerged); it was held fixed during the full-cohort and IL-12 re-estimations because the resistance term in later vaccine-administration arms masked the true mixture proportion (Discussion p. 803). For typical-value simulation set MIX_VAC_RELAPSE = 0 (dominant cure phenotype, complete tumor regression; Figure 3 ‘light grey’ individual predictions) or MIX_VAC_RELAPSE = 1 (relapser phenotype, tumor regrowth; Figure 3 ‘dark grey’ individual predictions). For population simulation, draw MIX_VAC_RELAPSE ~ Bernoulli(1 - 0.844) = Bernoulli(0.156) per subject. Discussion p. 803 notes the mixture is empirical: ‘no statistical differences on tumour size at the moment of vaccine administration were found between both populations (p > 0.05)’. The cure semantics (‘a sub-population of mice able to trigger only a temporal tumour response was postulated to describe the relapse observed in a few mice… the vaccine is able to trigger a permanent immune response (probably mediated by memory T-cells) in the cured mouse population, but no in all the mice’) are tied to immune-memory immunotherapy specifically and may not generalise to cytotoxic-chemotherapy relapse mixtures, so scope is specific. Future immunotherapy / vaccine / oncolytic-virus papers that share the same cure-vs-relapse dichotomy at the level of a vaccine-elicited inhibitory signal degradation rate may extend this entry; mixture indicators from unrelated dichotomies (e.g., fast-acquired-resistance vs slow-acquired-resistance to cytotoxic chemotherapy) should register a new canonical name.

MIX_SLOW_ELIM_NVP (canonical for binary mixture-model class indicator: slow-vs-fast nevirapine apparent-oral-clearance subpopulation)

• Description: Per-subject latent mixture-model class indicator from the Svensson 2012 nevirapine multi-source mega-model. 1 = subject classified to the minority slow-eliminator subpopulation (typical CL/F = 1.45 L/h at FFM 42 kg reference; estimated population probability 17.3 % in Svensson 2012 Table 2 ‘Probability (%) of belonging to pop. 2’); 0 = subject classified to the majority fast-eliminator subpopulation (typical CL/F = 3.12 L/h, 82.7 %). Not a measured clinical covariate – the mixture assignment is the per-subject latent-class index from the Svensson 2012 NONMEM mixture block. The paper Discussion paragraph 4 attributes the slow class biologically to CYP2B6 516TT homozygotes (‘The estimate of the probability of belonging to the low-clearance population (17.3 %) agrees well with the reported proportion of CYP2B6-516-TT homozygotes in the South African population, implying biological plausibility of two populations with different clearance rates’); the model itself does not require genotype data as input.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (fast-eliminator majority class; 82.7 % of the source cohort).
• Source aliases: MIXTURE – NONMEM $MIXTURE block class index in the Svensson 2012 estimation run (component 1 = fast with P = 0.827, component 2 = slow with P = 0.173); the binary indicator is MIX_SLOW_ELIM_NVP = as.integer(MIXTURE == 2).
• Example models: Svensson_2012_nevirapine.R (gates the typical apparent oral clearance between subpopulation-specific anchors: cl_typ <- exp(lcl_fast) * (1 - MIX_SLOW_ELIM_NVP) + exp(lcl_slow) * MIX_SLOW_ELIM_NVP, with lcl_fast = log(3.12) and lcl_slow = log(1.45); the single 24.9 % BSV on CL/F is shared across the two mixture sub-populations).
• Notes: For typical-value simulation set MIX_SLOW_ELIM_NVP = 0 (the recommended default) to reproduce the majority fast-eliminator phenotype; set MIX_SLOW_ELIM_NVP = 1 to reproduce the rarer slow phenotype that the paper associates with CYP2B6 516TT homozygotes. For population simulation, draw MIX_SLOW_ELIM_NVP ~ Bernoulli(0.173) per subject. Scope: specific because the drug-specific _NVP suffix scopes the binary semantics to nevirapine CL/F mixtures; future fast/slow CYP2B6-driven elimination mixtures for other antiretrovirals (e.g., efavirenz, which shares CYP2B6 as the major elimination enzyme) should register sibling canonicals (MIX_SLOW_ELIM_EFV, etc.) rather than reuse this entry. The Bienczak 2016 and Schipani 2011 nevirapine extractions in the registry use CYP2B6 genotype indicators directly (CYP2B6_IM / CYP2B6_SM / CYP2B6_USM / SNP_CYP2B6_RS3745274_T_COUNT) and are NOT mixture models – they assume the user supplies known genotype; MIX_SLOW_ELIM_NVP is the orthogonal pathway for the Svensson 2012 paper which estimates the phenotype distribution without genotype data.

MIX_SLOW_ELIM_EFV (canonical for binary mixture-model class indicator: slow-vs-typical efavirenz apparent-oral-clearance subpopulation)

• Description: Per-subject latent mixture-model class indicator from the Chala 2023 NONMEM $MIX block. 1 = subject classified to the minority slow-eliminator subpopulation, whose typical CL/F is 0.2829 times that of the main population (about 3.5-fold lower); 0 = subject classified to the majority main population. Not a measured clinical covariate – the assignment is the per-subject latent-class index from the $MIX block, and no covariate in the paper’s screen identified the class. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (majority main population; 92.5% of the source cohort).
• Source aliases:
  • MIXNUM – NONMEM $MIX block class index in the Chala 2023 estimation run ($MIX NSPOP=2; P(1) = THETA(11); P(2) = 1 - P(1), with MIXPOP1 = 1 when MIXNUM.EQ.1 and CLMIX = THETA(12)**MIXPOP1); component 1 is the slow class with P = 0.07511, so the binary indicator is MIX_SLOW_ELIM_EFV = as.integer(MIXNUM == 1).
• Example models: Chala_2023_efavirenz.R (multiplicative fold-factor on apparent oral clearance raised to the 0/1 indicator power, ... * e_mix_slow_elim_efv_cl^MIX_SLOW_ELIM_EFV with e_mix_slow_elim_efv_cl = 0.2829; the single 35.4% CV IIV on CL/F is shared across both mixture classes).
• Notes: Registered as the efavirenz sibling that the MIX_SLOW_ELIM_NVP Notes block pre-named (“future fast/slow CYP2B6-driven elimination mixtures for other antiretrovirals (e.g., efavirenz …) should register sibling canonicals (MIX_SLOW_ELIM_EFV, etc.) rather than reuse this entry”). Estimated population probability of MIX_SLOW_ELIM_EFV = 1 is 0.07511 (RSE 71%; bootstrap median 0.09, 95% CI 0.014-0.272; Chala 2023 Table 2 ‘Proportion of an unknown subpopulation among the studied population’). For typical-value simulation set MIX_SLOW_ELIM_EFV = 0 (the recommended default); for population simulation draw MIX_SLOW_ELIM_EFV ~ Bernoulli(0.075) per subject, which is what Chala 2023 did for its own Figure 3 Monte-Carlo simulations (Methods: “randomly assigned to an unknown subpopulation at the probability of 8%”). Unlike MIX_SLOW_ELIM_NVP, this mixture is orthogonal to – not a substitute for – genotype covariates: Chala_2023_efavirenz.R carries CYP2B66 and ABCB1 rs3842 genotype effects on CL/F and* this latent class simultaneously, and the Chala 2023 Discussion speculates the class “might represent unstudied covariates like other CYP2B6 and CYP2A6 polymorphisms”, neither of which was genotyped. Note the reference-category convention differs from the general MIX_FAST_ELIM column, where 1 marks the fast class; the _SLOW_ELIM_ names put the minority slow class at 1 so that 0 is the ordinary simulation default.

MIX_ELEV_IL6 (canonical for binary cycle-level mixture-model class indicator: elevated-IL-6-production-surge active vs no-surge subpopulation)

• Description: Per (subject, chemotherapy cycle) latent mixture-model class indicator: 1 = this cycle has an elevated IL-6 production surge (the IL-6 surge function g_IL6(t) is active and adds the relative-surge term (1 + g_IL6(t)) to the IL-6 production rate); 0 = no elevated IL-6 production in this cycle (the surge function is zeroed and IL-6 follows its baseline turnover only). Not a measured clinical covariate – the gate implements the NONMEM $MIXTURE block frequency that the Netterberg 2018 model estimated from the data. Time-fixed within a single chemotherapy cycle; redrawn independently between cycles 1 and 4 in the paper’s cohort.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no elevated IL-6 production this cycle; the complement of the surge-active subpopulation).
• Source aliases: MIXTURE – NONMEM $MIXTURE block class index in the Netterberg 2018 estimation run; the binary indicator is MIX_ELEV_IL6 = as.integer(elevation-active component). The paper reports 16 subpopulations across the 2 cycles x 2 biomarkers indicator combinations, but the marginal IL-6-elevation probability per cycle is a single value (Table 2 Pelevation,IL-6 = 63.4 %, RSE 10 %).
• Example models: Netterberg_2018_breast_cancer_FN_biomarkers.R, Netterberg_2018_breast_cancer_FN_tte_preFN.R, Netterberg_2018_breast_cancer_FN_tte_atFN.R (gates the IL-6 surge function: g_il6 <- MIX_ELEV_IL6 * sa_il6_i / (((t - pt_il6_i) / sw_il6_i)^4 + 1), so MIX_ELEV_IL6 = 0 zeroes the surge contribution while MIX_ELEV_IL6 = 1 activates the Netterberg 2018 Equation 1 Lorentzian surge centred at PT_IL-6).
• Notes: General scope because acute-phase biomarker surge mixtures recur in cytotoxic-chemotherapy IL-6 / CRP turnover models (any future paper modelling a Friberg-style IL-6 surge with a NONMEM mixture-component frequency should reuse this canonical). For typical-value simulation set MIX_ELEV_IL6 = 1 (the dominant 63.4 % per-cycle phenotype); for population simulation, draw MIX_ELEV_IL6 ~ Bernoulli(0.634) per (subject, cycle). The companion MIX_ELEV_CRP indicator gates the CRP surge function in the same model.

MIX_ELEV_CRP (canonical for binary cycle-level mixture-model class indicator: elevated-CRP-production-surge active vs no-surge subpopulation)

• Description: Per (subject, chemotherapy cycle) latent mixture-model class indicator: 1 = this cycle has an elevated CRP production surge (the CRP surge function g_CRP(t) is active); 0 = no elevated CRP production surge (the surge function is zeroed, but CRP production may still rise via the IL-6 -> CRP linear-regulation Slope coefficient when MIX_ELEV_IL6 = 1). Not a measured clinical covariate – the gate implements the NONMEM $MIXTURE block frequency that the Netterberg 2018 model estimated from the data. Time-fixed within a single chemotherapy cycle; redrawn independently between cycles in the paper’s cohort.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no elevated CRP production surge this cycle; CRP elevation may still occur via the IL-6 RCFB regulation through Slope).
• Source aliases: MIXTURE – NONMEM $MIXTURE block class index in the Netterberg 2018 estimation run; the binary indicator is MIX_ELEV_CRP = as.integer(elevation-active component). Per Netterberg 2018 Results, the per-cycle marginal probability is 44.3 % (Table 2 Pelevation,CRP, RSE 20 %), lower than the IL-6 marginal because the IL-6 -> CRP linear regulation contributes additional CRP elevations even when the dedicated CRP surge is off.
• Example models: Netterberg_2018_breast_cancer_FN_biomarkers.R, Netterberg_2018_breast_cancer_FN_tte_preFN.R, Netterberg_2018_breast_cancer_FN_tte_atFN.R (gates the CRP surge function: g_crp <- MIX_ELEV_CRP * sa_crp_i / (((t - pt_crp_i) / sw_crp_i)^4 + 1), so MIX_ELEV_CRP = 0 zeroes the dedicated CRP-surge contribution while MIX_ELEV_CRP = 1 activates the Netterberg 2018 Equation 1 Lorentzian surge centred at PT_CRP = PT_IL-6 + PT_CRP+).
• Notes: General scope because acute-phase biomarker surge mixtures recur in cytotoxic-chemotherapy IL-6 / CRP turnover models. For typical-value simulation set MIX_ELEV_CRP = 1 to reproduce a CRP-elevation cycle; for population simulation, draw MIX_ELEV_CRP ~ Bernoulli(0.443) per (subject, cycle). The paper observes that the actual fraction of CRP elevations is higher than Pelevation,CRP because of the IL-6 RCFB coupling through Slope; the canonical refers strictly to the dedicated surge function and not the coupled-elevation path.

MIX_MONO_SLOW (canonical for binary mixture-model class indicator: monophasic-slow tumour-shrinkage subpopulation)

• Description: Per-subject latent mixture-model class indicator from a four-class longitudinal tumour-size (TGI) mixture model. 1 = subject classified to the “monophasic slow” class, whose target-lesion burden declines steadily at the base first-order kill rate with no static residual; 0 = otherwise. One assignment per subject, time-fixed. Not a measured clinical covariate – the assignment is a posterior latent-class index from a NONMEM $MIX block. Used together with MIX_BIPHASIC and MIX_MONO_FAST: exactly one of the three may be 1, and all three 0 denotes the reference “escape” (progressor) class.
• Units: (binary)
• Type: binary
• Scope: general (oncology)
• Reference category: 0; the model-wide reference class is “escape” (all three of MIX_MONO_SLOW, MIX_BIPHASIC, MIX_MONO_FAST equal to 0).
• Source aliases: MIXNUM = 2 / EST = 2 – Chatterjee 2017 NONMEM $MIX class index and the EST column of its output table; the indicator is MIX_MONO_SLOW = as.integer(MIXNUM == 2).
• Example models: Chatterjee_2017_pembrolizumab_mixture.R (selects the base kill rate KD with no static “deep” tumour portion: supplementary Table 1 row “Monophasic slow”; estimated class prevalence 0.389, main-article Table 2 P2).
• Notes: A four-level latent class is carried as three binary indicators rather than one integer class index, per the library’s decomposed-indicator convention for categorical covariates. The three indicators are mutually exclusive; a data assembler that sets two of them to 1 will silently produce a nonsense kill-rate multiplier, so validate the sum (0 or 1) when building a dataset. For simulation, draw the class from the model’s own multinomial probability outputs (prob_escape, prob_monophasic_slow, prob_biphasic, prob_monophasic_fast), which Chatterjee 2017 fits as a second stage on the class logits. General scope because four-pattern response mixtures (progressor / slow responder / fast responder / biphasic responder) recur across immuno-oncology TGI analyses.

MIX_BIPHASIC (canonical for binary mixture-model class indicator: biphasic tumour-shrinkage-then-plateau subpopulation)

• Description: Per-subject latent mixture-model class indicator from a four-class longitudinal tumour-size (TGI) mixture model. 1 = subject classified to the “biphasic” class, whose target-lesion burden falls quickly and then plateaus on a static residual tumour mass; 0 = otherwise. One assignment per subject, time-fixed. Companion to MIX_MONO_SLOW and MIX_MONO_FAST; all three 0 denotes the reference “escape” (progressor) class.
• Units: (binary)
• Type: binary
• Scope: general (oncology)
• Reference category: 0; the model-wide reference class is “escape” (all three mixture indicators 0).
• Source aliases: MIXNUM = 3 / EST = 3 – Chatterjee 2017 NONMEM $MIX class index; the indicator is MIX_BIPHASIC = as.integer(MIXNUM == 3).
• Example models: Chatterjee_2017_pembrolizumab_mixture.R (the only class carrying a non-zero static “deep” tumour compartment BASEL2, together with the accelerated kill rate KD x 4.17: supplementary Table 1 row “Biphasic”; estimated class prevalence 0.255, main-article Table 2 P3).
• Notes: The distinguishing structural feature is the static residual, not the kill rate – this class shares its kill-rate multiplier with MIX_MONO_FAST and differs only by the presence of the non-shrinking tumour portion. Chatterjee 2017 introduced it because “in many patients, tumor size was stable for long durations after an initial drop in tumor size, a pattern that is uncharacteristic for conventional chemotherapies with which relapse would be more common”, which is also why the resistance term of the Claret 2009 model it was derived from was fixed to zero. A durable-plateau class is the signature of checkpoint-inhibitor response, so expect to reuse this indicator in other anti-PD-1 / anti-PD-L1 TGI mixtures.

MIX_MONO_FAST (canonical for binary mixture-model class indicator: monophasic-fast tumour-shrinkage subpopulation)

• Description: Per-subject latent mixture-model class indicator from a four-class longitudinal tumour-size (TGI) mixture model. 1 = subject classified to the “monophasic fast” class, whose target-lesion burden declines at an accelerated first-order kill rate with no static residual; 0 = otherwise. One assignment per subject, time-fixed. Companion to MIX_MONO_SLOW and MIX_BIPHASIC; all three 0 denotes the reference “escape” (progressor) class.
• Units: (binary)
• Type: binary
• Scope: general (oncology)
• Reference category: 0; the model-wide reference class is “escape” (all three mixture indicators 0).
• Source aliases: MIXNUM = 4 / EST = 4 – Chatterjee 2017 NONMEM $MIX class index; the indicator is MIX_MONO_FAST = as.integer(MIXNUM == 4).
• Example models: Chatterjee_2017_pembrolizumab_mixture.R (accelerated kill rate KD x 4.17 with the static “deep” tumour portion fixed to 0: supplementary Table 1 row “Monophasic fast”; estimated class prevalence 0.0628, main-article Table 2 P4).
• Notes: The rarest of the four classes in the founding cohort (about 6%), which is why the source paper’s Discussion notes that distinguishing mixture categories from sparse tumour-size data (some subjects contributed only two or three measurements) is difficult and was one of the reasons the authors later moved to a continuous-parameter “consolidated” structure (Chatterjee_2017_pembrolizumab_consolidated.R). Treat a per-subject class assignment from such a model as an uncertain posterior, not a phenotype.

MIX_LONG_MTT (canonical for binary mixture-model class indicator: long-mean-transit-time vs reference-mean-transit-time absorption subpopulation)

• Description: Per-subject latent mixture-model class indicator for a two-class mixture on the mean transit time (MTT) of a Savic-style transit-absorption chain. 1 = subject classified to the long-MTT subpopulation (slower gamma-distributed absorption input; MTT multiplied by a paper-specific factor greater than 1); 0 = subject classified to the reference-MTT subpopulation. Not a measured clinical covariate – the assignment is the per-subject latent-class index from a NONMEM $MIX / $MIXTURE block. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (the reference-MTT majority class; 83.9 % of the Fimbo 2023 source cohort).
• Source aliases:
  • MIXNUM / MIXEST – NONMEM $MIX block class index in the Fimbo 2023 estimation run (NSPOP=2, P(1) = THETA(11) = 0.1606 is the long-MTT class); the binary indicator is MIX_LONG_MTT = as.integer(MIXNUM == 1).
• Example models: Fimbo_2023_ivermectin.R (multiplies the typical mean transit time of the Savic transit chain: mtt <- exp(lmtt + etalmtt) * (1 + e_mix_long_mtt_mtt)^MIX_LONG_MTT with e_mix_long_mtt_mtt = 0.9695, so the long-MTT subpopulation absorbs with a 97.0 % longer mean transit time; the shared etalmtt log-normal IIV applies to whichever class the subject is assigned to).
• Notes: Distinct from MIX_LAGGED_ABS, which encodes a lag-time-present vs lag-time-absent absorption dichotomy (Bonate 2004, Frymoyer 2013) in models that carry an explicit alag(); MIX_LONG_MTT is for mixtures on the mean transit time of a transit-compartment chain, where there is no lag-time parameter at all and the delay is gamma-distributed. Registering a separate canonical follows the standing instruction in the MIX_PDI / MIX_LAGGED_ABS Notes that “mixture indicators from unrelated dichotomies should register a new canonical name”. The population probability of MIX_LONG_MTT = 1 is recorded per-model in covariateData[[MIX_LONG_MTT]]$notes; in Fimbo 2023 it is 0.1606 (Table 2 row “Proportion of a subpopulation with a different typical MTT”, RSE 30 %, bootstrap median 0.15, 95 % CI 0.07-0.30). For typical-value simulation set MIX_LONG_MTT = 0 (the majority phenotype); for population simulation draw MIX_LONG_MTT ~ Bernoulli(p) per subject. Because MTT shifts the shape and timing of the absorption input but not its total area, this mixture moves Tmax and Cmax while leaving AUC unchanged – a useful sanity check when validating a model that carries it. Scope: general because a two-class mixture on transit-chain MTT is a recurring popPK device for cohorts with heterogeneous gastric emptying or formulation dispersion, and future papers fitting a $MIX on MTT should reuse this canonical rather than introducing per-drug names. Ratified canonically alongside the Fimbo 2023 ivermectin extraction.

MIX_FAST_ABS (canonical for binary mixture-model class indicator: fast vs slow first-order absorption-rate subpopulation)

• Description: Per-subject latent mixture-model class indicator for a two-class mixture on the first-order absorption rate constant. 1 = subject classified to the fast-absorption subpopulation (in the founding example ka is fixed at a large value so the dose reaches central almost at once); 0 = subject classified to the slow-absorption subpopulation, which keeps the estimated lka and its IIV. Not a measured clinical covariate – the assignment is the per-subject latent-class index from a NONMEM $MIXTURE block. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (the slow-absorption majority class; P = 0.75 in Kruizinga 2022).
• Source aliases:
  • Prob. slow group – Kruizinga 2022 Table 2 (theta Prob. slow group = 0.75 is the probability of the slow class, so P(MIX_FAST_ABS = 1) = 1 - 0.75 = 0.25).
• Example models: Kruizinga_2022_clonazepam.R (founding example; selects the absorption rate constant: ka <- exp(lka + etalka) * (1 - MIX_FAST_ABS) + exp(lka_fastabs) * MIX_FAST_ABS, with the slow-class lka = log(1.106) 1/h carrying etalka and the fast-class lka_fastabs = fixed(log(100)) 1/h carrying no IIV).
• Notes: Distinct from MIX_LAGGED_ABS (lag time present vs absent), MIX_LONG_MTT (a mixture on the mean transit time of a transit chain) and MIX_FAST_ELIM (a mixture on clearance); here only the first-order absorption rate constant differs between the classes. Distinct also from the parallel two-route release pair lka_fast / lka_slow, in which EVERY subject absorbs through both routes at once – under MIX_FAST_ABS each subject absorbs through one route only, chosen by class. Oriented like MIX_FAST_ELIM so that the majority class is the reference 0, even though the founding paper tabulates the probability of the slow class. The population probability of MIX_FAST_ABS = 1 is recorded per-model in covariateData[[MIX_FAST_ABS]]$notes. For typical-value simulation set MIX_FAST_ABS = 0; for population simulation draw MIX_FAST_ABS ~ Bernoulli(p) per subject. Because only the input rate changes, the mixture moves Tmax and Cmax but leaves AUC unchanged. Scope: general because a fast/slow absorption-rate mixture is the usual remedy for a bimodal etalka distribution (Kruizinga 2022 Results 3.1: an etalka variance of 0.66 with a bimodal distribution fell to 0.16 once the mixture was added). Ratified by the maintainers alongside the Kruizinga 2022 clonazepam extraction.

MIX_FAST_GROW (canonical for latent mixture-model class indicator: fast-tumor-growth subpopulation)

• Description: Per-subject latent mixture-model class indicator for the fast-tumor-growth phenotype in a longitudinal tumor-burden mixture model. 1 = subject classified to the fast tumor-growth (rapid early progression) subpopulation; 0 = subject classified to some other tumor-dynamic phenotype. Not a measured clinical covariate – the assignment is the per-subject latent-class index from a NONMEM $MIXTURE block. Time-fixed per subject. In mixtures with more than two classes this indicator is paired with one or more sibling MIX_* indicators (e.g. MIX_NO_GROW) so that an all-zero pattern denotes the reference class; a three-class mixture is therefore encoded as two binaries rather than a single integer class index, matching the register’s standing preference for decomposed binary indicators over integer-coded categories.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (not the fast-growth class). In the founding Feng 2019 three-class mixture the reference pattern MIX_FAST_GROW = 0 and MIX_NO_GROW = 0 denotes the intermediate tumor-growth-and-shrinkage subpopulation.
• Source aliases: MIXTURE – NONMEM $MIXTURE block class index; in the Feng 2019 estimation run component 1 is the fast-TG subpopulation, so MIX_FAST_GROW = as.integer(MIXTURE == 1).
• Example models: Feng_2019_ipilimumab.R (selects the fast-TG anchors for baseline tumor burden, tumor-growth rate, and tumor-shrinkage rate constant, and selects the fast-TG correlated IIV block etaltg_fast + etalts_fast).
• Notes: Pre-named by the MIX_LARGE_BASE register entry’s Notes block (“mixture indicators from unrelated dichotomies (e.g., fast / slow tumor-growth phenotypes) should register a new canonical name (MIX_FAST_GROW, etc.)”). Population probability of MIX_FAST_GROW = 1 is recorded per-model in covariateData[[MIX_FAST_GROW]]$notes; in Feng 2019 it is TP1 / (1 + TP1 + TP2) = 1.20 / 3.078 = 0.390 (Table 2 footnote e). Distinct from MIX_LARGE_BASE, which indexes a bimodal baseline size mixture rather than a growth-rate phenotype; a model may in principle carry both. General scope because a fast-vs-other tumor-growth phenotype mixture recurs across oncology tumor-dynamics models (immune-checkpoint inhibitors in particular, where a subset of patients progress rapidly while others reach a durable steady state).

MIX_NO_GROW (canonical for latent mixture-model class indicator: no-growth (durable steady-state tumor burden) subpopulation)

• Description: Per-subject latent mixture-model class indicator for the no-growth phenotype in a longitudinal tumor-burden mixture model. 1 = subject classified to the no-growth subpopulation, in which the linear tumor-growth rate is structurally zero and tumor burden decays from baseline to a non-zero steady-state plateau; 0 = subject classified to some other tumor-dynamic phenotype. Not a measured clinical covariate – the assignment is the per-subject latent-class index from a NONMEM $MIXTURE block. Time-fixed per subject. Paired with MIX_FAST_GROW in three-class mixtures, where the all-zero pattern denotes the reference (intermediate) class.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (not the no-growth class). In the founding Feng 2019 three-class mixture the reference pattern MIX_FAST_GROW = 0 and MIX_NO_GROW = 0 denotes the intermediate tumor-growth-and-shrinkage subpopulation.
• Source aliases: MIXTURE – NONMEM $MIXTURE block class index; in the Feng 2019 estimation run component 2 is the no-growth subpopulation, so MIX_NO_GROW = as.integer(MIXTURE == 2).
• Example models: Feng_2019_ipilimumab.R (selects the no-growth anchors for baseline tumor burden, tumor-shrinkage rate constant, and the steady-state tumor-burden plateau TBss, forces the linear growth term to zero, and selects the no-growth diagonal IIVs etalts_nogrow / etaltbss_nogrow).
• Notes: The structural signature of this class is that the linear growth term is fixed to zero and a steady-state plateau parameter (ltbss, the asymptote the decaying tumor burden approaches) is estimated in its place – an immunotherapy-specific response pattern that the pre-immunotherapy Wang-type tumor-dynamics models cannot express (Feng 2019 Methods, “TGD model”: “This model was modified to improve the description of TB time profiles that asymptotically approach a steady-state value”). Population probability of MIX_NO_GROW = 1 is recorded per-model in covariateData[[MIX_NO_GROW]]$notes; in Feng 2019 it is 1 / (1 + TP1 + TP2) = 1 / 3.078 = 0.325 (Table 2 footnote e). General scope for the same reason as MIX_FAST_GROW: a durable-response / no-progression latent class is a recurring feature of immune-checkpoint-inhibitor tumor-dynamics models. Distinct from MIX_VAC_RELAPSE, which indexes a cure-vs-relapse dichotomy in a preclinical single-dose vaccine model and gates a signal-degradation rate rather than a growth term.

MIX_RAMP_REL (canonical for binary mixture-model class indicator: linearly time-ramping vs constant first-order slow depot release)

• Description: Per-subject latent mixture-model class indicator for which slow-release kinetics a long-acting depot follows in a subject. 1 = the slow-release depot empties with a first-order rate constant that grows linearly with time after the first dose, k(t) = ka_slope * t; 0 = the slow-release depot empties with a constant first-order rate ka_slow. Not a measured clinical covariate – the assignment is the per-subject latent-class index from a NONMEM $MIXTURE block. Time-fixed per subject. The dichotomy concerns the SLOW release route only; a fast first-order route running in parallel is shared by both classes.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (constant first-order slow release). In the founding Snelder 2019 EC403 cohort this is the 40.8% class.
• Source aliases:
  • POP / IPOP – NONMEM $MIXTURE class index in the Snelder 2019 Supplement 4 control stream, where SLP takes the ramp slope in one class and the constant rate in the other.
• Example models: Snelder_2019_leuprorelin.R (founding example; selects between ka_slope * max(0, tafd()) and ka_slow as the rate constant emptying depot2, the slow-release depot of a leuprorelin-SR 6-month formulation).
• Notes: Population probability of MIX_RAMP_REL = 1 is recorded per-model in covariateData[[MIX_RAMP_REL]]$notes; in Snelder 2019 it is 1 - 0.408 = 0.592 (Table 2 prop Pop1 = 0.408, RSE 18.9%, read together with Results 3.1: “41% and 59% of the subjects were assigned to population 1 (1st-order release) and population 2 (time dependent release), respectively”). For population simulation draw MIX_RAMP_REL ~ Bernoulli(0.592); set it to 0 or 1 to reproduce either release phenotype as a typical value. Registered as a new member of the MIX_* family on the instruction in MIX_LAGGED_ABS’s notes that “mixture indicators from unrelated dichotomies … should register a new canonical name”. Distinct from MIX_LAGGED_ABS, which switches an absorption LAG on and off while both classes keep a constant first-order rate; here neither class has a lag and the distinction is the functional FORM of the release rate. Specific scope because the semantics are tied to a parallel fast/slow depot-release structure; future long-acting-injectable models sharing the same ramp-vs-constant dichotomy may extend this entry. Ratified 2026-09-02 alongside the Snelder 2019 leuprorelin extraction (PMC6533438; ratified by the maintainers).

MIX_EARLY_REL (canonical for binary subpopulation indicator: early extended-release vs typical multi-phase release profile of a long-acting depot)

• Description: Per-subject subpopulation indicator for the shape of the release profile after a long-acting injectable depot. 1 = subject belongs to the early extended-release subpopulation, whose exposure is dominated by an early, prolonged release phase (a large first-delayed-release fraction) followed by a small late release; 0 = subject shows the typical multi-phase profile (small initial burst followed by up to three sequential delayed release peaks). Not a measured clinical covariate: in the founding model the assignment comes from an unsupervised shape-respecting k-means clustering (R package kmlShape, generalised Frechet distance) of each subject’s concentration profile, normalised by its own mean concentration, run BEFORE the population fit and then entered in the NONMEM dataset as a fixed categorical covariate. It is therefore a subpopulation class like the $MIX-based members of this family, but it is determined by the observed data rather than estimated as a latent mixture probability. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (typical multi-phase release; paper “cluster 1”, 103 of 118 subjects in Kapralos 2021).
• Source aliases:
  • cluster – Kapralos 2021 dataset column coded 1 or 2 (Equation 7: Parameter_pop = theta1 + (cluster - 1) * theta2); MIX_EARLY_REL = cluster - 1.
• Example models: Kapralos_2021_octreotide.R (founding example; additive shifts on the typical apparent clearance, -8.61 L/h, and on the log-ratio coordinates of the second and third release fractions, +3.06 and -0.523, of a four-process octreotide LAR release model).
• Notes: Population frequency of MIX_EARLY_REL = 1 is recorded per-model in covariateData[[MIX_EARLY_REL]]$notes; in Kapralos 2021 it is 15/118 = 12.7% (Table 1), and an external evaluation on the test-product arm of the same study gave 16/119 (Section 3.1). For population simulation draw MIX_EARLY_REL ~ Bernoulli(15/118); set it to 0 for a typical subject. Because the class is read off each subject’s observed profile, it cannot be known before dosing, and a model carrying it predicts a new subject only through that population frequency. Registered as a new member of the MIX_* family on the instruction in MIX_LAGGED_ABS’s notes that “mixture indicators from unrelated dichotomies … should register a new canonical name”. Distinct from MIX_RAMP_REL (functional form of a slow first-order depot rate) and MIX_LONG_MTT (a single transit chain’s mean transit time): here the classes differ in how the dose is SPLIT across parallel release processes, and in apparent clearance. Specific scope because the dichotomy is tied to a multi-process PLGA-microsphere release model; future long-acting-injectable models that cluster subjects on the same early-vs-typical release shape may extend this entry.

MIX_LOW_EXPOSURE (canonical for binary subpopulation indicator: low-exposure vs normal-exposure subjects identified before the fit)

• Description: Per-subject subpopulation indicator. 1 = subject identified by the analysts as having systematically lower drug exposure than the rest of the population; 0 = normal-exposure subject. Not a measured clinical covariate and not a NONMEM $MIXTURE posterior: in the founding model the class was assigned by visual inspection of the concentration data before the population fit and entered in the dataset as a fixed flag, on which study-specific bioavailability shifts were then estimated. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (normal-exposure subject).
• Source aliases:
  • LOW subscript in Bihorel 2021 Table 1 (F1 SAD,LOW, F1 MAD,LOW, F2 SAD,LOW, F2 MAD,LOW) and F1study,low / F2study,low in Figure 1 – the fold changes applied when MIX_LOW_EXPOSURE = 1.
• Example models: Bihorel_2021_BMS986166.R (founding example; multiplies the BMS-986166 relative bioavailability F1 by 0.655 in Study IM018001 or 0.471 in Study IM018003, and the fraction of the molar dose absorbed as BMS-986166-P, F2, by 1.25 or 1.82).
• Notes: Population frequency of MIX_LOW_EXPOSURE = 1 is 6/48 = 12.5% in Bihorel 2021 (Results; Discussion). No subject characteristic (age, weight, renal or hepatic function markers, ethnicity) separated the two classes, and the Discussion reads the opposite-direction shifts (F1 below 1, F2 above 1) as a possible sign of greater pre-systemic conversion of the prodrug in low-exposure subjects. For population simulation draw MIX_LOW_EXPOSURE ~ Bernoulli(0.125); set it to 0 for a typical subject. Because the class is read off each subject’s observed profile, a model carrying it predicts a new subject only through that population frequency. Registered as a member of the MIX_* family, the same way as MIX_EARLY_REL, which is also a class fixed before the fit rather than estimated as a latent mixture probability.

MIX_LOW_EXPANSION (canonical for binary mixture-model class indicator: low-expansion vs reference-expansion CAR-T cell subpopulation)

• Description: Per-subject latent mixture-model class indicator for a two-class mixture on the baseline maximum expansion rate of CAR-T cells. 1 = subject assigned to the low-expansion subpopulation, whose typical baseline maximum expansion rate Vmax1,base is markedly reduced (by ~92% in the founding example) and which carries no retained interindividual variability and no CD4/CD8-ratio covariate effect; 0 = subject assigned to the reference-expansion population. Not a measured clinical covariate – the assignment is the per-subject latent-class index from a NONMEM $MIXTURE block. Time-fixed per subject. The two classes share the prior-ASCT covariate effect but differ in the baseline expansion rate and its covariate/IIV structure.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (reference-expansion population; the majority class).
• Source aliases: MIXTURE – NONMEM $MIXTURE block class index in the Mueller-Schoell 2021 estimation run (the reference-expansion population is the majority component with MIXP = 0.803); the binary indicator is MIX_LOW_EXPANSION = as.integer(MIXTURE == low-expansion component). The per-patient assignment is tabulated in Supplementary Table S1 (“Low expansion-subpopulation [yes/no]”).
• Example models: MuellerSchoell_2021_axicabtageneCiloleucel_qsp.R (selects the baseline maximum expansion rate Vmax1,base between the reference anchor exp(lvmax1_base_ref) = 0.00846 and the low-expansion anchor exp(lvmax1_base_low) = 0.000700; the reference branch additionally carries the day-7 CD4/CD8 power effect and the 150% CV IIV, both of which the low-expansion branch drops; Equations 10-11 and Table 1).
• Notes: Registered as a new member of the MIX_* family on the standing instruction that mixture indicators from unrelated dichotomies register a new canonical name. Population probability of MIX_LOW_EXPANSION = 1 is recorded per-model in covariateData[[MIX_LOW_EXPANSION]]$notes; in Mueller-Schoell 2021 it is 1 - MIXP = 1 - 0.803 = 0.197 (Table 1 MIXP = 0.803, RSE 11%; 4 of 19 patients). For typical-value simulation set MIX_LOW_EXPANSION = 0 (dominant reference phenotype); for population simulation draw MIX_LOW_EXPANSION ~ Bernoulli(0.197) per subject. Distinct from the tumour-dynamics growth-phenotype mixtures (MIX_FAST_GROW, MIX_NO_GROW), which index tumour-growth latent classes; here the mixture is on the CAR-T CELL expansion capacity, a distinct biological dichotomy. General scope because a low-vs-reference cellular-expansion latent class is expected to recur across adoptive cell-therapy (CAR-T, TIL, TCR-T) cellular-kinetic models, where a subset of patients reliably fails to expand.

SEIZURE_ACUTE (canonical for acute focal-seizure-activity indicator in preclinical brain-distribution studies)

• Description: 1 = the animal is undergoing acute focal seizure activity (typically chemoconvulsant-induced, e.g., intrahippocampal pilocarpine perfusion in a microdialysis design) during the modelled observation window; 0 = no acute seizure activity. Captures the experimental flag used by preclinical BBB / biophase-distribution PK studies to test whether seizure-driven changes in blood-brain barrier permeability or brain redistribution alter brain PK parameters relative to non-seizing control animals. Time-fixed per animal in single-condition allocations (each rat in one arm); a time-varying form is permitted when a study induces transient seizures during a sampling window and the model’s author chooses to gate the covariate only over the seizure interval (document in covariateData[[SEIZURE_ACUTE]]$notes).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no acute seizure activity; non-seizing control animal).
• Source aliases:
  • A (values inverted: source A = 1 means non-seizing control and A = 0 means seizing, per Clinckers 2008 Methods ‘Both A and B were set to 1 in control animals; to 0 and 1, respectively, in seizing animals’; the canonical encoding flips the polarity via SEIZURE_ACUTE = 1 - A) – used in Clinckers_2008_MHD_rat.R.
• Example models: Clinckers_2008_MHD_rat.R (mutually-exclusive selection of the seizure-specific biophase volume V3b in place of the control V3a inside the V3 expression of the one-compartment-plus-biophase model for MHD in male Wistar rats; v3 = exp(lv3a + etalv3a) * (1 - SEIZURE_ACUTE) * (1 - EFFLUX_INHIB) + exp(lv3b) * SEIZURE_ACUTE * (1 - EFFLUX_INHIB) + exp(lv3c + etalv3c) * (1 - SEIZURE_ACUTE) * EFFLUX_INHIB).
• Notes: Specific scope because the indicator’s semantics are tied to acute, induced focal seizure activity in preclinical BBB / biophase PK designs; clinical-trial canonicals for epilepsy (CONMED_AED, CONMED_EIAED, PDV previous-period seizure count) describe distinct concepts and should not be conflated. Future preclinical studies that test a similar seizure-vs-non-seizure contrast should extend the example list; promote to general scope once a second model legitimately ratifies the name. Mutually exclusive with EFFLUX_INHIB in Clinckers 2008 (each rat is allocated to exactly one of {control, seizure, efflux-inhibition}); the model() encoding uses (1 - SEIZURE_ACUTE) * (1 - EFFLUX_INHIB) as the control multiplier, so any data record that asserts both flags as 1 would zero out the V3a term – data assemblers should preserve mutual exclusivity unless a model is explicitly designed for the cross-condition case.

SEIZURE_LOCUS_PARTIAL (canonical for partial (focal) versus generalized seizure-locus indicator)

• Description: 1 = the patient’s epilepsy has a partial (focal) seizure locus, 0 = generalized seizure locus. The axis is the anatomical ORIGIN of the seizure under the International League Against Epilepsy (ILAE) classification – focal / partial onset versus generalized onset – and it is a subject-level clinical diagnosis, time-fixed per subject in practice. Use it for clinical epilepsy popPK / PD models that retain the partial-versus-generalized contrast as a covariate.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (generalized seizure locus).
• Source aliases:
  • partial seizure – the Nakashima 2015 Eq 6 term label (“partial seizure = 1 if a partial seizure was present, and 0 if partial seizures were absent”) and Table 3 slope row label Partial seizure.
  • Seizure locus – the Nakashima 2015 Table 1 / Table 2 covariate-screen label, whose levels are Generalize and Partial.
• Example models: Nakashima_2015_valproic_acid.R (modifier of the exposure slope of a logistic exposure-response model for an over-50% reduction in seizure frequency: e_partial_slope = +2.41 per 100 ug/mL of predicted trough valproate, i.e. partial-onset patients have a less negative concentration-response slope than generalized-onset patients; Nakashima 2015 Table 3).
• Notes: Named for the CLASSIFICATION AXIS (_LOCUS_) rather than as a bare SEIZURE_PARTIAL, because epilepsy papers routinely tabulate two orthogonal categorical axes and the founding source screens both: seizure locus (partial versus generalized, this canonical) and seizure type or aetiology (symptomatic versus cryptogenic, which was screened and rejected – Nakashima 2015 Table 2, difference of objective function 3.05 and 0.67, both not significant). Collapsing them into one name would make a future extraction ambiguous, so a symptomatic-versus-cryptogenic indicator must be registered under its own name if a model ever needs one. Name ratified by the maintainers (PMC9833507, 2026-09-02) in preference to the shorter SEIZURE_PARTIAL for exactly that reason. Distinct from [[SEIZURE_ACUTE]], which is a preclinical experimental flag for whether an animal is undergoing chemoconvulsant-induced acute seizure activity during a brain-distribution study – an experimental condition, not a clinical diagnosis – and distinct from baseline seizure-count covariates such as the NSP3M_* family, which are continuous or count measures of seizure burden rather than of seizure origin. Registered with general scope on first sighting because the ILAE focal-versus-generalized dichotomy is a standard, universally applicable clinical classification rather than a paper-defined construct; the per-model covariateData[[SEIZURE_LOCUS_PARTIAL]]$notes should still record the source’s own wording and the cohort split, since papers differ in how they assign patients with both focal and generalized seizure types. Cohort caution from the founding source: the cohort is heavily partial-dominant (67 of 77, 87.0%), leaving only 10 generalized patients in the reference group, and the +2.41 estimate has a bootstrap confidence interval of -0.53 to 7.82 that spans zero (Nakashima 2015 Table 1 and Table 3).

EFFLUX_INHIB (canonical for local efflux-transporter-inhibitor co-perfusion indicator in preclinical brain-distribution studies)

• Description: 1 = the animal’s brain microdialysis probe is co-perfused with an efflux-transporter inhibitor (e.g., verapamil 5 mM for P-glycoprotein / MDT blockade in Clinckers 2008) at the site of brain sampling, 0 = no local efflux-transporter inhibitor co-perfusion. Captures the experimental flag used by preclinical BBB / biophase-distribution PK studies to test whether local pharmacological blockade of brain efflux transporters alters brain PK parameters relative to non-blocked control animals. Time-fixed per animal in single-condition allocations.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: n/a – T_ENTRY is a per-subject time-offset covariate that anchors the time-since-study-entry clock used by post-entry-only model terms (placebo / learning effects, study-design transient drops).
• Source aliases: none standardized; this canonical originates with the Delor 2013 AD disease-progression extraction. The source paper’s NONMEM dataset handles the global / study-time split internally (the dataset is staged with TIME = study time and the model uses a fixed offset to align with DOT); when porting to rxode2 / nlmixr2 the per-subject offset is exposed as a covariate so simulation event tables can carry it explicitly.
• Example models: Delor_2013_alzheimer.R, Clinckers_2008_MHD_rat.R (mutually-exclusive selection of the verapamil-specific biophase volume V3c in place of the control V3a inside the V3 expression: v3 = v3a*(1-SEIZURE_ACUTE)*(1-EFFLUX_INHIB) + v3b*SEIZURE_ACUTE*(1-EFFLUX_INHIB) + v3c*(1-SEIZURE_ACUTE)*EFFLUX_INHIB).
• Notes: Scope: specific because the variable is paper-domain-bound – it only makes sense for models that operate on a global disease-time axis distinct from per-subject study-entry timing. For a typical-value reproduction the user supplies T_ENTRY per subject in the simulation event table; a reasonable construction is T_ENTRY = DOT_individual + a few years of established disease so the patient is observed during disease progression (see the Delor 2013 vignette for the construction used to reproduce Figures 2-4).

M3M3FBS_PRESENT (canonical for m-3M3FBS phospholipase-C-activator pretreatment indicator in the Grzesk 2016 vascular-reactivity ex-vivo design)

• Description: 1 = the isolated, perfused tail artery preparation was pretreated with the phospholipase-C activator 2,4,6-trimethyl-N-[3-(trifluoromethyl)phenyl]benzenesulfonamide (m-3M3FBS) at 1e-5 mol/L prior to the agonist concentration-response titration; 0 = no m-3M3FBS pretreatment (control arm). Time-fixed per artery / concentration-response curve (each artery is allocated to exactly one of the control or +m-3M3FBS arms). Captures the binary experimental flag distinguishing the two CRC sets of Grzesk 2016 Table I (control vs +m-3M3FBS) and selects the m-3M3FBS-shifted (EC50, Emax) parameter pair inside model.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (control artery without m-3M3FBS pretreatment).
• Source aliases: none – Grzesk 2016 Tables I and II label the two arms “controls” vs “+m-3M3FBS” in prose; the model column is the canonical M3M3FBS_PRESENT.
• Example models: Grzesk_2016_m3M3FBS.R (binary selector that swaps the per-agonist control (EC50, Emax) pair for the +m-3M3FBS pair in the sigmoidal Emax CRC; the agonist identity is selected by the companion AGONIST_CODE).
• Notes: Specific scope because the indicator’s semantics are tied to m-3M3FBS pretreatment in the Grzesk 2016 vascular-reactivity ex-vivo design. The Grzesk-group sibling papers (Biomed Rep 2 / 2014 pertussis toxin; Mol Med Report 5 / 2012 calcium blockers; etc.) test different pretreatments and would register their own pretreatment indicators rather than reuse this name. Distinct from CONMED_* (clinical-trial-level concomitant medication indicators) because m-3M3FBS is an experimental ex-vivo pretreatment, not a clinically administered comedication.

PERIOD_ACTIVE (canonical for binary diurnal-period indicator in preclinical chronopharmacology / diurnal-variation PK-PD studies)

• Description: Binary indicator of the diurnal (light-dark) period during which the experimental time-of-administration falls. 1 = active period (lights-off / dark phase for nocturnal laboratory rodents; the animal’s behaviourally active wake window); 0 = resting period (lights-on / light phase; the animal’s behaviourally quiescent sleep window). In the standard 12:12 light-dark cycle used by Kervezee 2014 (Wistar rats housed under LD 12:12 with lights-off defined as Zeitgeber Time ZT12), the resting-period cohort is dosed at ZT0-ZT12 and the active-period cohort at ZT12-ZT24. Time-fixed per experimental subject (each animal is dosed at a single ZT, so the derived period membership is a per-subject scalar in the fitting cohort). The paper aggregates six ZT levels (ZT0/4/8/12/16/20) into the binary resting-vs-active split for the covariate model, so PERIOD_ACTIVE is the collapsed two-level version of the six-level ZT design; the raw ZT is not carried as a separate canonical covariate.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (resting period / lights-on). In Kervezee 2014 the effect enters as a period-specific selection of the P-gp component of five brain-distribution clearances (CL_DBR-PL,Pgp, CL_PL-ECF,Pgp, CL_ECF-PL,Pgp, CL_PL-LV,Pgp) and of the CSF bulk-flow rate Q_CSF, with separate resting-period and active-period point estimates rather than a single reference-plus-multiplier form (Kervezee 2014 Table II reports both periods side-by-side).
• Source aliases:
  • time of administration (P-gp) – Kervezee 2014 Table I prose label for the covariate hypothesis tested against the five P-gp-mediated clearance parameters; free-form phrase, not a data column name.
  • time of administration (Q_CSF) – Kervezee 2014 Table I prose label for the covariate hypothesis tested against CSF bulk flow.
• Example models: Kervezee_2014_quinidine_rat.R (nine-compartment brain-distribution PBPK for i.v. quinidine in male Wistar rats; PERIOD_ACTIVE = 1 selects the active-period estimates for CL_DBR-PL,Pgp = 659 uL/min, CL_PL-ECF,Pgp = 18.3 uL/min, CL_ECF-PL,Pgp = 3.98 uL/min, CL_PL-LV,Pgp = 2.44 uL/min, and Q_CSF = 0.227 uL/min; PERIOD_ACTIVE = 0 selects the resting-period estimates 228, 17, 3.14, 1.55, and 0.522 uL/min respectively).
• Notes: Specific scope because the binary active-vs-resting collapse is meaningful only for chronopharmacology / diurnal-variation preclinical PK-PD designs that pool experimental time-of-administration into two phase cohorts; a future paper that resolves circadian rhythmicity as a continuous cosinor / harmonic function of clock time (e.g., a cosine effect on P-gp expression peaking near ZT18) would use a separate canonical (a continuous clock-time covariate; register with a name like TIME_OF_DAY when the founding paper lands) rather than reusing PERIOD_ACTIVE. Distinct from SEIZURE_ACUTE / EFFLUX_INHIB / M3M3FBS_PRESENT (paper-defined binary experimental-condition indicators in preclinical models that also live in this section) because those flag pharmacological / disease-state interventions rather than a natural chronobiological phase. Downstream promotion to general scope only if a second preclinical or clinical chronopharmacology model adopts the same binary active-vs-resting collapse with equivalent semantics; nocturnal-vs-diurnal-species handling (rats are nocturnally active, humans are diurnally active) must be documented in covariateData[[PERIOD_ACTIVE]]$notes per model because the mapping ZT / clock-time -> {active, resting} depends on species. Kervezee 2014 does not observe circadian rhythmicity in constant darkness so the paper is careful to call the variation “diurnal” rather than “circadian” – the canonical name PERIOD_ACTIVE avoids that debate by labelling the level (active-vs-resting) rather than the driver (diurnal-vs-circadian).

SPECIES_RABBIT (canonical for New Zealand White rabbit species indicator in cross-species / animal-to-human translation models)

• Description: 1 = the subject is a New Zealand White (NZW) rabbit; 0 = the subject belongs to another species in the same model. Time-fixed per subject. Used by joint cross-species models that carry one parameter set per species in a single model file and select among them with mutually-exclusive species indicators, rather than splitting the paper’s analysis into one file per species. Pairs with SPECIES_MACAQUE; the convention is that all species indicators being 0 selects the model’s reference species (human, in the founding example).
• Units: (binary)
• Type: binary
• Scope: general – animal-to-human translation (FDA Animal Rule, first-in-human allometric projection, preclinical-to-clinical bridging) is a recurring model class and future extractions are expected to reuse this indicator family.
• Reference category: 0 (not a NZW rabbit; in a model whose reference species is human, both SPECIES_RABBIT and SPECIES_MACAQUE equal to 0 selects the human parameter set).
• Source aliases:
  • SPECIES (a categorical column with levels such as “rabbit” / “monkey” / “human”) – decomposed into one binary indicator per non-reference species, following the same one-hot pattern used by the RACE_<GROUP> family. Used in Nagy_2017_obiltoxaximab.R.
• Example models: Nagy_2017_obiltoxaximab.R (selects the NZW rabbit column of Nagy 2017 Supplementary Table S1 – CL 0.0263 L/day, Vc 0.114 L, Vp 0.0744 L, Q 0.119 L/day, Ka 0.961 /day, F1 0.899 – together with the rabbit-specific 3.165 kg allometric reference weight and the rabbit Michaelis-Menten arm Vmax 0.912 mg/day, Km 10.4 ug/mL), Cheng_2026_levamisole_2cm.R, Cheng_2026_levamisole_mpbpk.R (select the rabbit weight-normalised clearances 1.09 and 1.46 L/h/kg from Cheng 2026 Table 1 at the paper’s 3 kg reference weight; the mPBPK file also selects the rabbit blood volume 152.7 mL from Supplemental Table S1A).
• Notes: A species indicator is a subject attribute, not an experimental intervention, but it lives in this section because its mechanical role – a mutually-exclusive binary that swaps one parameter set for another inside model() – matches the existing preclinical indicators SEIZURE_ACUTE, EFFLUX_INHIB, M3M3FBS_PRESENT and PERIOD_ACTIVE. Each species carries its own allometric reference weight, so a model using this family must select wtRef by the same indicators it uses to select the structural parameters; do not normalise every species to a single reference weight. Data assemblers must preserve mutual exclusivity (no record may set more than one SPECIES_* indicator to 1), because the selection arithmetic isHuman <- 1 - SPECIES_RABBIT - SPECIES_MACAQUE would otherwise go negative and silently produce nonsense rather than erroring. Register a sibling canonical (SPECIES_RAT, SPECIES_DOG, SPECIES_MOUSE,…) rather than overloading this name when a new species appears. Distinct from the model-file metadata field population$species, which is free text describing the studied population and is not a data column. When a paper’s species-specific models are genuinely independent fits with no shared structure, the library’s default remains one model file per species (e.g. An_2012_mitoxantrone_mouse_pbpk.R / An_2012_mitoxantrone_human_pbpk.R); this family is for the case where a single file is the more faithful or more useful representation.

SPECIES_MACAQUE (canonical for cynomolgus macaque species indicator in cross-species / animal-to-human translation models)

• Description: 1 = the subject is a cynomolgus macaque (Macaca fascicularis); 0 = the subject belongs to another species in the same model. Time-fixed per subject. The non-human-primate member of the SPECIES_<NAME> one-hot family; see SPECIES_RABBIT for the family conventions.
• Units: (binary)
• Type: binary
• Scope: general – the cynomolgus macaque is the standard non-human primate in monoclonal-antibody and biologics preclinical programmes, so cross-species biologics extractions are expected to reuse this indicator.
• Reference category: 0 (not a cynomolgus macaque; in a model whose reference species is human, both SPECIES_RABBIT and SPECIES_MACAQUE equal to 0 selects the human parameter set).
• Source aliases:
  • SPECIES (categorical column, level “monkey” / “macaque” / “cyno”) – decomposed to a binary indicator. Used in Nagy_2017_obiltoxaximab.R.
• Example models: Nagy_2017_obiltoxaximab.R (selects the cynomolgus macaque column of Nagy 2017 Supplementary Table S1 – CL 0.0191 L/day, Vc 0.134 L, Vp 0.123 L, Q 0.0890 L/day, Ka 3.89 /day, F1 0.895 – with the macaque-specific 2.88 kg allometric reference weight and the macaque Michaelis-Menten arm Vmax 0.275 mg/day, Km 3.21 ug/mL).
• Notes: In the founding example the macaque nonlinear-elimination component is also the one carried over to the human infected projection (Nagy 2017 Methods: “the nonlinear clearance model component from the cynomolgus macaque model was added to the human population PK model, allometrically scaled to human body size”). That carry-over is a property of the model, not of this covariate: the indicator still reads 0 for the human subject, and the model handles the transfer by scaling the macaque Vmax from the macaque reference weight rather than the human one. Any future model that borrows a component across species should document the borrowing in covariateData notes so the reference-weight choice is auditable.

SPECIES_DUCK (canonical for domestic duck species indicator in cross-species / animal-to-human translation models)

• Description: 1 = the subject is a domestic duck (Anas platyrhynchos domesticus); 0 = the subject belongs to another species in the same model. Time-fixed per subject. An avian member of the SPECIES_<NAME> one-hot family; see SPECIES_RABBIT for the family conventions (mutual exclusivity, all-indicators-zero selects the reference species, per-species reference weights).
• Units: (binary)
• Type: binary
• Scope: general – ducks are a recurring veterinary-pharmacology species and interspecies meta-analyses of antiparasitic and antimicrobial drugs routinely pool avian with mammalian cohorts.
• Reference category: 0 (not a duck; in a model whose reference species is human, all SPECIES_* indicators equal to 0 selects the human parameter set).
• Source aliases:
  • SPECIES (categorical column, level “duck”) – decomposed into a binary indicator following the one-hot pattern of the RACE_<GROUP> family. Used in Cheng_2026_levamisole_2cm.R / Cheng_2026_levamisole_mpbpk.R.
• Example models: Cheng_2026_levamisole_2cm.R (selects the duck weight-normalised clearance 0.20 L/h/kg from Cheng 2026 Table 1, joint 2CM column, at the paper’s 2.5 kg reference weight), Cheng_2026_levamisole_mpbpk.R (selects the duck clearance 0.216 L/h/kg from Cheng 2026 Table 1, joint mPBPK column, together with the duck blood-volume fraction 86.3 mL/kg from Supplemental Table S1A).
• Notes: Birds and mammals differ materially in basal metabolic rate and renal morphology, and Cheng 2026 found ducks to be a clearance outlier that had to be dropped from the simple allometric regression – a reason to keep each avian species as its own indicator rather than collapsing to a single SPECIES_BIRD. Pairs with SPECIES_CHICKEN.

SPECIES_CHICKEN (canonical for domestic chicken species indicator in cross-species / animal-to-human translation models)

• Description: 1 = the subject is a domestic chicken (Gallus gallus domesticus); 0 = the subject belongs to another species in the same model. Time-fixed per subject. An avian member of the SPECIES_<NAME> one-hot family; see SPECIES_RABBIT for the family conventions.
• Units: (binary)
• Type: binary
• Scope: general – the chicken is the most-studied avian species in veterinary pharmacokinetics and in food-animal residue-depletion work, so cross-species extractions are expected to reuse this indicator.
• Reference category: 0 (not a chicken; in a model whose reference species is human, all SPECIES_* indicators equal to 0 selects the human parameter set).
• Source aliases:
  • SPECIES (categorical column, level “chicken” / “broiler breeder” / “prelay hen”) – decomposed into a binary indicator. Used in Cheng_2026_levamisole_2cm.R / Cheng_2026_levamisole_mpbpk.R.
• Example models: Cheng_2026_levamisole_2cm.R (selects the chicken weight-normalised clearance 0.93 L/h/kg from Cheng 2026 Table 1, joint 2CM column, at the paper’s 4.5 kg reference weight), Cheng_2026_levamisole_mpbpk.R (selects the chicken clearance 2.59 L/h/kg from Cheng 2026 Table 1, joint mPBPK column, the chicken blood-volume fraction of 10 percent of body weight from Supplemental Table S1A, and the chicken-specific tissue-to-plasma partition coefficient Kp,chicken = 3.38 from Table 3).
• Notes: In the founding example this indicator does double duty: it selects both a species-specific clearance and a species-specific distribution parameter (Kp), because chickens showed markedly higher tissue partitioning than the pooled cross-species value. A model that needs to distinguish laying stage (prelay versus peak production, whose reported bioavailabilities differ 61 percent versus 88 percent in Cheng 2026 Supplemental Table S9) should register a separate physiological-state covariate rather than splitting this species indicator. Pairs with SPECIES_DUCK.

SPECIES_GOAT (canonical for domestic goat species indicator in cross-species / animal-to-human translation models)

• Description: 1 = the subject is a domestic goat (Capra hircus); 0 = the subject belongs to another species in the same model. Time-fixed per subject. A small-ruminant member of the SPECIES_<NAME> one-hot family; see SPECIES_RABBIT for the family conventions.
• Units: (binary)
• Type: binary
• Scope: general – goats are a standard small-ruminant species in veterinary antiparasitic pharmacokinetics and appear alongside sheep in most food-animal interspecies analyses.
• Reference category: 0 (not a goat; in a model whose reference species is human, all SPECIES_* indicators equal to 0 selects the human parameter set).
• Source aliases:
  • SPECIES (categorical column, level “goat”) – decomposed into a binary indicator. Used in Cheng_2026_levamisole_2cm.R / Cheng_2026_levamisole_mpbpk.R.
• Example models: Cheng_2026_levamisole_2cm.R (selects the goat weight-normalised clearance 0.38 L/h/kg from Cheng 2026 Table 1, joint 2CM column, at the paper’s 18 kg reference weight), Cheng_2026_levamisole_mpbpk.R (selects the goat clearance 0.359 L/h/kg from Cheng 2026 Table 1, joint mPBPK column, together with the goat blood-volume fraction 70 mL/kg from Supplemental Table S1A).
• Notes: Distinct from SPECIES_SHEEP even though goats and sheep are both small ruminants frequently studied in the same publication: Cheng 2026 estimated clearances that differ by roughly a factor of three between them (0.38 versus 0.99 L/h/kg in the joint 2CM), and goats were the one species whose oral bioavailability departed from the otherwise consistent 50-80 percent cross-species range. Do not collapse the two into a SPECIES_RUMINANT indicator.

SPECIES_DOG (canonical for domestic dog species indicator in cross-species / animal-to-human translation models)

• Description: 1 = the subject is a domestic dog (Canis lupus familiaris); 0 = the subject belongs to another species in the same model. Time-fixed per subject. A member of the SPECIES_<NAME> one-hot family; see SPECIES_RABBIT for the family conventions.
• Units: (binary)
• Type: binary
• Scope: general – the dog is one of the two standard non-rodent toxicology species and is ubiquitous in preclinical-to-clinical bridging, veterinary pharmacokinetics, and allometric projection. This is the name named in the SPECIES_RABBIT Notes as an expected sibling.
• Reference category: 0 (not a dog; in a model whose reference species is human, all SPECIES_* indicators equal to 0 selects the human parameter set).
• Source aliases:
  • SPECIES (categorical column, level “dog” / “beagle”) – decomposed into a binary indicator. Used in Cheng_2026_levamisole_2cm.R / Cheng_2026_levamisole_mpbpk.R.
• Example models: Cheng_2026_levamisole_2cm.R (selects the dog weight-normalised clearance 0.49 L/h/kg from Cheng 2026 Table 1, joint 2CM column, at the paper’s 20.7 kg reference weight), Cheng_2026_levamisole_mpbpk.R (selects the dog clearance 0.464 L/h/kg from Cheng 2026 Table 1, joint mPBPK column, together with the dog blood-volume fraction 84 mL/kg from Supplemental Table S1A).
• Notes: The indicator carries no breed information – a model whose dog cohort is breed-specific in a load-bearing way (for example an MDR1 / ABCB1 nt230(del4) collie cohort) should add a separate genotype or breed covariate rather than splitting this indicator. In the founding example the dog profile declined mono-exponentially, so the two-tissue mPBPK distribution parameters were poorly identified for this species; that is a property of the source data, not of the covariate, and is documented in the model files’ covariateData notes.

SPECIES_SHEEP (canonical for domestic sheep species indicator in cross-species / animal-to-human translation models)

• Description: 1 = the subject is a domestic sheep (Ovis aries), including lambs; 0 = the subject belongs to another species in the same model. Time-fixed per subject. A small-ruminant member of the SPECIES_<NAME> one-hot family; see SPECIES_RABBIT for the family conventions.
• Units: (binary)
• Type: binary
• Scope: general – sheep are a standard food-animal and large-animal-surgery species and recur in antiparasitic, antimicrobial, and residue-depletion pharmacokinetics.
• Reference category: 0 (not a sheep; in a model whose reference species is human, all SPECIES_* indicators equal to 0 selects the human parameter set).
• Source aliases:
  • SPECIES (categorical column, level “sheep” / “ewe” / “lamb”) – decomposed into a binary indicator. Used in Cheng_2026_levamisole_2cm.R / Cheng_2026_levamisole_mpbpk.R.
• Example models: Cheng_2026_levamisole_2cm.R (selects the sheep weight-normalised clearance 0.99 L/h/kg from Cheng 2026 Table 1, joint 2CM column, at the paper’s 26 kg reference weight), Cheng_2026_levamisole_mpbpk.R (selects the sheep clearance 0.994 L/h/kg from Cheng 2026 Table 1, joint mPBPK column, together with the sheep blood-volume fraction 59 mL/kg from Supplemental Table S1A).
• Notes: Lambs are recorded under this indicator rather than a separate canonical, because body weight – already carried as WT – is the covariate that distinguishes them; a model that needs a maturation effect distinct from body size should add an age covariate rather than a second species indicator. See SPECIES_GOAT Notes for why goats and sheep are kept as separate indicators.

SPECIES_PIG (canonical for domestic pig species indicator in cross-species / animal-to-human translation models)

• Description: 1 = the subject is a domestic pig (Sus scrofa domesticus), including minipigs; 0 = the subject belongs to another species in the same model. Time-fixed per subject. A member of the SPECIES_<NAME> one-hot family; see SPECIES_RABBIT for the family conventions.
• Units: (binary)
• Type: binary
• Scope: general – the pig and minipig are recurring non-rodent translational species for oral absorption and dermal work, and the pig is also a major food animal in residue-depletion pharmacokinetics.
• Reference category: 0 (not a pig; in a model whose reference species is human, all SPECIES_* indicators equal to 0 selects the human parameter set).
• Source aliases:
  • SPECIES (categorical column, level “pig” / “swine” / “minipig”) – decomposed into a binary indicator. Used in Cheng_2026_levamisole_2cm.R / Cheng_2026_levamisole_mpbpk.R.
• Example models: Cheng_2026_levamisole_2cm.R (selects the pig weight-normalised clearance 0.27 L/h/kg from Cheng 2026 Table 1, joint 2CM column, at the paper’s 39.2 kg reference weight), Cheng_2026_levamisole_mpbpk.R (selects the pig clearance 0.371 L/h/kg from Cheng 2026 Table 1, joint mPBPK column, the pig blood-volume fraction 60 mL/kg from Supplemental Table S1A, and the pig-specific tissue-to-plasma partition coefficient Kp,pig = 5.62 from Table 3).
• Notes: Like SPECIES_CHICKEN, this indicator selects both a species-specific clearance and a species-specific distribution parameter in the founding example, because pigs showed a tissue-to-plasma partition coefficient roughly four times the pooled cross-species value despite unremarkable plasma protein binding. Minipigs are recorded under this indicator rather than a separate canonical; Cheng 2026 notes that minipig distribution behaviour is not well characterised, so a future model that needs to separate minipig from domestic pig should register SPECIES_MINIPIG and document why the split is load-bearing.

CNSREG_PFC (canonical for medial-prefrontal-cortex sampling-region indicator in regional-CNS PK-PD models)

• Description: 1 = the observation was sampled from the medial prefrontal cortex; 0 = it was sampled from the reference CNS region. Member of the CNSREG_<region> family, which identifies the anatomical central-nervous-system region (brain nucleus, cortical area, or spinal-cord segment) a per-observation measurement was taken from, in models that fit one joint structural model across several CNS sampling sites and let a region indicator select region-specific typical values. Per-observation rather than per-subject: a single animal commonly contributes one sample from each region.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the model’s reference CNS region; the amygdala in the founding example).
• Source aliases:
  • BRAIN AREA (categorical column, level PFC) – decomposed to a binary indicator. Used in AlfoseaCuadrado_2024_reserpine_rat.R.
• Example models: AlfoseaCuadrado_2024_reserpine_rat.R (selects the prefrontal-cortex typical value of the monoamine precursor production rate, kpin = (exp(lkpin) * (1 - CNSREG_PFC - CNSREG_SC) + exp(lkpin_pfc) * CNSREG_PFC + exp(lkpin_sc) * CNSREG_SC) * exp(etalkpin), with kin = 6.97 (amygdala), 2.10 (prefrontal cortex) and 1.78 (spinal cord) mg/L/h sharing one 97 % inter-animal variability term).
• Notes: The CNSREG_ prefix was chosen over a BRAIN_ prefix so that spinal-cord levels are described correctly – the founding example’s three levels are two brain regions plus the lumbar spinal cord, and the spinal cord is not brain. Distinct from the brain_<region> compartment namespace in compartment-names.md (brain_cortex, brain_hippocampus,…), which names ODE states holding a drug concentration in a region; CNSREG_<region> is a data column marking which region an observation came from and carries no mass balance. Also distinct from REGION_<geography> (geographic enrollment region), TUMTP_<type> (tumour histology) and SAMPLE_<design> (blood-collection design flags on residual error). Data assemblers must preserve mutual exclusivity – no record may set more than one CNSREG_* indicator to 1 – because the reference-level arithmetic 1 - CNSREG_PFC - CNSREG_SC would otherwise go negative and silently produce nonsense rather than erroring. Register a sibling canonical (CNSREG_HIPPOCAMPUS, CNSREG_STRIATUM, CNSREG_NAC,…) rather than overloading this name when a new region appears; promote the family to general scope once a second model ratifies it.

CNSREG_SC (canonical for spinal-cord sampling-region indicator in regional-CNS PK-PD models)

• Description: 1 = the observation was sampled from the spinal cord (the lumbar portion in the founding example); 0 = it was sampled from the reference CNS region. Sibling of CNSREG_PFC in the CNSREG_<region> family. Per-observation rather than per-subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the model’s reference CNS region; the amygdala in the founding example).
• Source aliases:
  • BRAIN AREA (categorical column, level SC) – decomposed to a binary indicator. Used in AlfoseaCuadrado_2024_reserpine_rat.R. Note the source column is named for brain area even though this level is spinal cord.
• Example models: AlfoseaCuadrado_2024_reserpine_rat.R (selects the spinal-cord typical value of the monoamine precursor production rate kin = 1.78 mg/L/h).
• Notes: See CNSREG_PFC for the family’s naming rationale, the distinction from the brain_<region> compartment namespace, and the mutual-exclusivity requirement.

INH_MCT_KM_RATIO (canonical for the inhibitor-to-Ki concentration ratio acting on the Michaelis constant of a saturable transport process)

• Description: Dimensionless steady-state inhibitor concentration ratio R = [I] / Ki applied as a (1 + R) multiplier on the Michaelis constant Km of a saturable (transporter-mediated) process. Together with the companion INH_MCT_CONC_RATIO it selects the inhibition mechanism in the standard enzyme / transporter inhibition algebra: with INH_MCT_KM_RATIO = R and INH_MCT_CONC_RATIO = 0 the rate law is COMPETITIVE inhibition; with both set to R it is NONCOMPETITIVE; with INH_MCT_KM_RATIO = 0 and INH_MCT_CONC_RATIO = R it is UNCOMPETITIVE. Both at 0 recovers the uninhibited rate law exactly, so a user who never sets either covariate gets the paper’s fitted model. R is a ratio, not a concentration – the inhibitor is assumed to be at steady state, so no inhibitor PK compartment is required.
• Units: (unitless ratio)
• Type: continuous
• Scope: specific
• Reference category: 0 (no inhibitor present; the rate law reduces to the uninhibited Michaelis-Menten form).
• Source aliases:
  • R (defined as [I]/Ki) – Felmlee 2010 Methods (“where R equals I/Ki and the inhibitor is administered at steady-state”) and eqs. (11)-(13). Used in Felmlee_2010_ghb_rat.R.
• Example models: Felmlee_2010_ghb_rat.R (multiplies km_reab in the saturable renal-reabsorption clearance reabsorption <- vmax_reab / (km_reab * (1 + INH_MCT_KM_RATIO) + Culf1 * (1 + INH_MCT_CONC_RATIO)); the paper simulates R = 1, 10 and 100 in Table II and Fig. 3 and corroborates the competitive case experimentally with L-lactate co-administration in Table III).
• Notes: Specific scope because the pair is tied to a monocarboxylate-transporter (MCT) inhibition design; a paper inhibiting a different transporter or an enzyme should register its own INH_<TARGET>_* pair rather than overload the MCT name, since the Ki that defines R is target-specific. The two-covariate encoding was ratified in preference to a single ratio plus a mechanism-code switch (maintainers’ ruling, PMC2895455, 2026-09-02) because selecting the branch inside model() would otherwise need if/else or magic-number arithmetic; the two-multiplier form encodes all three published mechanisms in one branch-free expression. Always set both members of the pair explicitly in a simulation event table – leaving one unset is what silently converts a competitive simulation into an uncompetitive one.

INH_MCT_CONC_RATIO (canonical for the inhibitor-to-Ki concentration ratio acting on the substrate-concentration term of a saturable transport process)

• Description: Dimensionless steady-state inhibitor concentration ratio R = [I] / Ki applied as a (1 + R) multiplier on the SUBSTRATE CONCENTRATION term in the denominator of a saturable (transporter-mediated) rate law. Companion to INH_MCT_KM_RATIO; see that entry for the mechanism-selection table (competitive / noncompetitive / uncompetitive) and the reduction to the uninhibited model when both are 0. Meaningless alone – the two are always declared and set together.
• Units: (unitless ratio)
• Type: continuous
• Scope: specific
• Reference category: 0 (no inhibitor present; the substrate term is unmodified).
• Source aliases:
  • R (defined as [I]/Ki) – Felmlee 2010 Methods and eqs. (12)-(13). Used in Felmlee_2010_ghb_rat.R.
• Example models: Felmlee_2010_ghb_rat.R (multiplies the proximal-tubule ultrafiltrate concentration Culf1 in the saturable renal-reabsorption clearance; set to 0 for the competitive simulations the paper actually tabulates, and to R for the noncompetitive and uncompetitive mechanisms the paper reports as “data not shown”).
• Notes: Specific scope, same rationale as INH_MCT_KM_RATIO. Note that the two covariates carry the SAME source symbol R – the source paper writes one ratio and three separate equations, whereas the canonical encoding writes one equation and routes R to one or both multipliers. That is a deliberate re-factoring of the paper’s algebra, not two distinct measured quantities, and the model file’s covariateData notes must say so. Ratified canonically alongside the Felmlee 2010 GHB extraction (maintainers’ ruling, PMC2895455, 2026-09-02).

HCV_GT1B (canonical for hepatitis C virus genotype-1 subtype indicator: GT1B vs GT1A)

• Description: Binary indicator of the HCV genotype-1 subtype assigned to the subject. 1 = patient infected with HCV genotype 1B (GT1B); 0 = patient infected with HCV genotype 1A (GT1A; the reference subtype in the source paper’s IC50 estimates). Time-fixed per subject because the HCV subtype is determined at the time of infection and does not change over the modelled treatment window. Distinct from the broader genotype number (HCV genotype 1, 2, 3,…) because the GT1A vs GT1B contrast within genotype 1 carries meaningful PK/PD differences (replicon susceptibility, resistance-associated substitutions, drug-resistance trajectory) for direct-acting antivirals; a future model that needs a genotype-2-vs-1 contrast or a 4-level genotype indicator should register a sibling canonical rather than overloading HCV_GT1B.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no human serum; the active fraction factive is forced to 1).
• Source aliases:
  • HS – Garonzik 2016 (paper Methods + Table 2, “% Human Serum”; the model column was renamed from the short HS to the spelled-out canonical HUMAN_SERUM_PCT on 2026-05-27 to align with the register’s naming standards and avoid an ambiguous two-letter abbreviation).
• Example models: Garonzik_2016_daptomycin.R, Wang_2018_daclatasvir_asunaprevir.R (multiplicative effect on IC50 of both drugs via the fixed scaling factors SCL_IC50_DCV = 0.18 and SCL_IC50_ASV = 0.30; the encoding is ic50_dcv_t0 = exp(lic50_dcv_gt1a + etalic50_dcv) * scl_ic50_dcv^HCV_GT1B and ic50_asv_t0 = exp(lic50_asv_gt1a + etalic50_asv) * scl_ic50_asv^HCV_GT1B. Also switches the DCV resistance coefficient Kr_DCV between 0.43 /day for GT1A and 0.13 /day for GT1B; Kr_ASV is the same for both subtypes), Canini_2018_setrobuvir.R (switches setrobuvir sigmoid-Emax EC50, Hill coefficient, and viral clearance rate c between the two per-genotype typical values and their independent IIVs).
• Notes: Specific scope because the discrete serum-percentage levels and the associated factive estimates are tied to the Garonzik 2016 daptomycin in-vitro design. An in-vitro experimental condition rather than a human pop-PK covariate; HUMAN_SERUM_PCT values outside the studied discrete set make factive = 0 inside the model (a deliberately conspicuous failure rather than silent interpolation). The spelled-out name follows the register’s anti-abbreviation principle (cf. the DIS_BUNIONECTOMY entry’s avoidance of DIS_BUN). Future in-vitro protein-binding-versus-serum experiments should extend the example list.

HCV_GT2 (canonical for hepatitis C virus genotype-2 indicator)

• Description: Binary indicator of the HCV genotype assigned to the subject. 1 = patient infected with HCV genotype 2; 0 = any other genotype. Time-fixed per subject, because the infecting genotype is determined at the time of infection and does not change over a treatment window. The two can coexist in one dataset.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0. In the multi-indicator encoding described in Notes the implied reference is HCV genotype 1, which is the dominant genotype in Western chronic-HCV trial populations and the reference in every registered example.
• Source aliases: paper narrative “HCV genotype 2” / “genotype 2” subgroup labels, and NONMEM-style dataset columns such as GT2 or a multi-level GENO column that must be one-hot expanded.
• Example models: Riggs_2012_albinterferon.R (multiplicative effects on all three disposition parameters of subcutaneous albinterferon alfa-2b: 0.971 on CL/F, 0.952 on V/F, and 1.14 on ka, per Table I. Supplied alongside HCV_GT3 and HCV_GT4 with HCV genotype 1 as the common reference; 66% of the 1984-patient cohort carried genotype 1).
• Notes: Use one indicator per non-reference genotype (HCV_GT2, HCV_GT3, HCV_GT4,…), all mutually exclusive, with genotype 1 as the implied reference when all are 0 – the same one-hot pattern as the RACE_<GROUP> family. Do NOT encode genotype as a single integer-valued column: a popPK covariate model estimates a separate coefficient per genotype, and an integer column would impose a spurious ordering. When a paper’s reference genotype is not genotype 1, register the genotype-1 indicator HCV_GT1 explicitly rather than re-purposing one of these, and state the reference in covariateData[[...]]$reference_category. Sibling canonicals for other hepatitis viruses (HBV genotypes) should be registered separately when first encountered.

HCV_GT3 (canonical for hepatitis C virus genotype-3 indicator)

• Description: Binary indicator: 1 = patient infected with HCV genotype 3; 0 = any other genotype. Time-fixed per subject. Member of the one-hot HCV_GT<n> family described under HCV_GT2.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (HCV genotype 1 in the registered examples; state the paper’s reference per model).
• Source aliases: paper narrative “HCV genotype 3” / “genotype 3” subgroup labels; a one-hot expansion of a multi-level GENO column.
• Example models: Riggs_2012_albinterferon.R (multiplicative effects 0.994 on CL/F, 0.891 on V/F, and 0.854 on ka, per Table I).
• Notes: See HCV_GT2 for the family encoding rules. Genotypes 2 and 3 are frequently pooled in chronic-HCV trial design (a shared 24-week treatment duration versus 48 weeks for genotype 1) even when the popPK model estimates them separately, as in Riggs 2012 – so a paper’s trial stratum “genotype 2/3” is not the same thing as either of these indicators, and a pooled genotype-2-or-3 covariate should be registered separately as HCV_GT23 if a model ever needs one.

HCV_GT4 (canonical for hepatitis C virus genotype-4 indicator)

• Description: Binary indicator: 1 = patient infected with HCV genotype 4; 0 = any other genotype. Time-fixed per subject. Member of the one-hot HCV_GT<n> family described under HCV_GT2.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (HCV genotype 1 in the registered examples; state the paper’s reference per model).
• Source aliases: paper narrative “HCV genotype 4” / “genotype 4” subgroup labels; a one-hot expansion of a multi-level GENO column.
• Example models: Riggs_2012_albinterferon.R (multiplicative effects 0.886 on CL/F, 0.595 on V/F, and 0.202 on ka, per Table I – but only 5 of 1984 patients carried genotype 4, and the authors retained the indicator specifically so that the resulting imprecision would quantify the lack of information rather than be reported as an insignificant effect, which is what a stepwise covariate search would have concluded (Discussion). The ka coefficient’s bootstrap 95% CI spans 0.102 to 1.4).
• Notes: See HCV_GT2 for the family encoding rules. Genotype 4 is rare in the North American and European trial populations that dominate the chronic-HCV popPK literature, so a coefficient on this indicator is very often estimated from a handful of subjects. Record the genotype-4 subject count in covariateData[[HCV_GT4]]$notes and treat simulations at HCV_GT4 = 1 as extrapolation.

Envenomation / venom source

SNAKEFAMILY_ELAPID (canonical for snake-family categorical indicator of venom source)

• Description: Binary indicator of the snake-family origin of the venom bolus delivered by a single bite event. 1 = bite from a snake of the family Elapidae (front-fanged elapids – cobras, kraits, mambas, sea snakes, Australian terrestrial elapids such as taipans / brown snakes / death adders); 0 = bite from a snake of the family Viperidae (true vipers and pit vipers, including Bothrops, Crotalus, Daboia / Vipera russelli, Vipera aspis / berus / ammodytes, Bitis, Hypnale, Cerastes). The covariate is a per-bite-event property of the dose source (the snake), not a property of the patient; in a dataset of envenomed patients each subject’s row(s) carry one value across the entire follow-up.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (healthy opioid-naive).
• Source aliases: none.
• Example models: Mann_2022_respiratory_physiology.R, Sanhajariya_2018_snake_venom.R (Sanhajariya 2018 Table A1 covariate model: f(central) <- exp(lfdepot + e_snakefamily_elapid_fdepot * SNAKEFAMILY_ELAPID + etalfdepot) with lfdepot = log(1) fixed and e_snakefamily_elapid_fdepot = log(0.569)).
• Notes: Scope: specific because the two parameter sets are tied to the Mann 2022 chronic-vs-naive opioid pharmacology calibration. A future model that captures a graded tolerance (e.g., a continuous “tolerance index”) rather than a two-class binary should register a separate continuous canonical.

CONMED_ELACRIDAR (canonical for concomitant elacridar (GF120918) P-glycoprotein inhibitor co-administration indicator)

• Description: 1 = subject co-administered elacridar (GF120918, a third-generation P-glycoprotein / breast-cancer-resistance-protein inhibitor used experimentally to block efflux at the blood-brain barrier), 0 = vehicle-only / no concomitant elacridar. Time-fixed per subject in the Groenendaal 2007 source study, where the inhibitor infusion starts 120 min before the morphine dose and runs past the last brain-ECF observation; time-varying use in future per-occasion studies is permitted.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (vehicle co-infusion, no elacridar).
• Source aliases:
  • GF120918 – the GlaxoSmithKline development code, which is how the compound is named throughout the Groenendaal 2007 paper (and in most of the 1998-2010 BBB-transport literature). Per the standing generic-INN-over-development-code rule the canonical column uses the INN ELACRIDAR; a model whose source paper says “GF120918” sets covariateData[[CONMED_ELACRIDAR]]$source_name to record the paper’s own term.
  • Paper-specific treatment-arm indicator – Groenendaal 2007 Table 1 distinguishes a “GF120918” arm from a “Vehicle” arm; paper equation 8 defines the factor as 1 when GF120918 is co-infused and 0 when vehicle is co-infused.
• Example models: Groenendaal_2007_morphine_brain_rat.R (multiplicative exponential effect on the BBB active-efflux rate constant only: kefflux_bbb = exp(lkefflux_bbb + e_conmed_elacridar_kefflux_bbb * CONMED_ELACRIDAR + etalkefflux_bbb) with e_conmed_elacridar_kefflux_bbb = log(0.0113 / 0.0195) = -0.5455, the 42% reduction in keff reported in the paper abstract. The paper tested GF120918 on the blood three-compartment disposition and on the passive-diffusion and active-influx terms as well, and retained it on keff alone.).
• Notes: Scope: specific because the only source is the Groenendaal 2007 rat morphine BBB-transport paper and the column’s meaning is tied to an experimental P-gp blockade design rather than a general clinical-coadministration indicator. Per-model covariateData[[CONMED_ELACRIDAR]]$notes must document the dose / regimen used (Groenendaal 2007: a 1 min 6 mg/kg IV bolus in DMSO followed by a continuous 25 ng/min infusion in DMSO / glucose / cyclodextrine 5/5/10% in saline, giving a measured mean steady-state blood concentration of 214 +/- 16 ng/mL) and any per-subject vs per-record time-varying convention. Directly analogous to [[CONMED_PROBENECID]] (Xie 2000) and [[CONMED_TARIQUIDAR]] (Syvanen 2011 / 2012, Westerhout 2013) – all three are experimentally-administered transporter inhibitors used to dissect a substrate’s brain-vs-plasma distribution, encoded as a binary co-administration indicator regardless of the inhibitor’s clinical use case.

CONMED_TARIQUIDAR (canonical for concomitant tariquidar (P-glycoprotein inhibitor) co-administration indicator)

• Description: 1 = subject co-administered tariquidar (a third-generation P-glycoprotein / breast-cancer-resistance-protein inhibitor used as an experimental probe to block efflux at the blood-brain barrier), 0 = vehicle-only / no concomitant tariquidar. Time-fixed per subject in the Syvanen 2011 source study (single 15 mg/kg IV bolus given 20-30 min before the PET tracer injection means the indicator is operationally constant across the 60-minute PET acquisition window); time-varying use in future per-occasion studies is permitted.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no tariquidar co-administration; vehicle is 5% glucose in saline).
• Source aliases:
  • Paper-specific treatment-arm indicator – used in Syvanen_2011_verapamil_rat.R (the paper distinguishes a tariquidar-treatment arm from a vehicle arm; the canonical column is 1 for the 21 tariquidar-treated rats and 0 for the 21 vehicle-treated rats).
• Example models: Syvanen_2011_verapamil_rat.R (multiplicative theta^COVARIATE form on three structural parameters: vp = exp(lvp + e_conmed_tariquidar_vp * CONMED_TARIQUIDAR) with e_conmed_tariquidar_vp = log(1.20) = 0.1823 (20% increase in plasma peripheral 1 volume); vbr1 = ... * exp(e_conmed_tariquidar_vbr1 * CONMED_TARIQUIDAR + ...) with e_conmed_tariquidar_vbr1 = log(2.41) = 0.8796 (2.41-fold increase in fast-exchange brain compartment volume); qin = exp(lqin + e_conmed_tariquidar_qin * CONMED_TARIQUIDAR) with e_conmed_tariquidar_qin = log(12.0) = 2.4849 (12.0-fold increase in BBB influx clearance). The paper screened tariquidar as a covariate on Qout as well but did not retain it.), Syvanen_2012_quinidine_rat.R (multiplicative exponential effects on CL, Q_out, Q_in, and f2: e_conmed_tariquidar_cl = log(0.835) = -0.180; e_conmed_tariquidar_qout = log(0.366) = -1.005; e_conmed_tariquidar_qin = log(2.65) = 0.975; e_conmed_tariquidar_f2 = log(5.66) = 1.733. Combined Q_in / Q_out shift raises the pseudo-equilibrium brain ECF : plasma ratio 7.2-fold and the combined f2 shift raises the total-brain : plasma ratio about 40-fold per paper Results p91), Westerhout_2013_quinidine.R (P-gp split into separate passive CL_X,p and P-gp-mediated CL_X,P-gp clearances per the paper Appendix mass-balance equations; the P-gp components are switched off when CONMED_TARIQUIDAR = 1 via the multiplier pgp_active = 1 - CONMED_TARIQUIDAR and contribute additively (to efflux clearances CL_X-PL) or subtractively (to influx clearances CL_PL-X) when CONMED_TARIQUIDAR = 0; 1.9-fold P-gp effect on systemic CL_E from Table 4 SBPK combined column, with full bidirectional P-gp components on plasma <-> brain_deep / brain_ecf / brain_csf_lv exchanges and P-gp absent at the CM in this final model).
• Notes: Scope: specific because the only source is the Syvanen 2011 rat PET BBB-transport paper and the canonical column meaning is intrinsically tied to the experimental P-gp blockade design rather than a general clinical-coadministration indicator. Per-model covariateData[[CONMED_TARIQUIDAR]]$notes must document the dose / regimen of tariquidar used (Syvanen 2011: 15 mg/kg IV bolus in 3 mL/kg of 5% glucose in saline, administered 20-30 min before the radiotracer) and any per-subject vs per-record time-varying convention. Analogous to [[CONMED_PROBENECID]] in the Xie 2000 rat BBB-transport paper – both are experimentally-administered transporter inhibitors used to dissect a tracer’s brain-vs-plasma distribution, encoded as a binary co-administration indicator regardless of the inhibitor’s clinical use case. Future preclinical-PET extractions that test tariquidar (or a related Pgp inhibitor like elacridar / zosuquidar with comparable mechanism) on a different probe-substrate should reuse this canonical when the inhibitor is tariquidar; structurally distinct P-gp inhibitors should register a separate canonical with the inhibitor’s INN in the name (e.g., CONMED_ELACRIDAR).

DIS_POSTSE_KAINATE (canonical for the post-status-epilepticus state induced by kainic-acid pre-treatment in rats)

• Type: binary
• Description: 1 = animal underwent kainic-acid-induced status epilepticus (SE) some number of days prior to the modelled observation window, 0 = animal received saline (or another non-epileptogenic control). Time-fixed per animal in the Syvanen 2011 source study (the SE-induction-to-PET interval is a per-arm design constant – 7 days post-induction at scanning); the canonical column models the chronic post-SE state at a defined post-induction interval, NOT the acute SE episode itself.
• Example models: Syvanen_2012_quinidine_rat.R, Syvanen_2011_verapamil_rat.R.

L_ANTAGONIST_pM (canonical for time-varying opioid-antagonist effect-site concentration input to the Mann 2022 binding layer)

• Description: Time-varying opioid-antagonist effect-site concentration in picomolar (pM), supplied as a data covariate to the multi-ligand competitive mu-receptor binding model. Antagonist analogue of L_OPIOID_pM. In a composed Mann 2022 + Laffont 2024 / 2025 chain, the upstream antagonist PK layer (Laffont_2024_naloxone or Laffont_2024_nalmefene) is post-processed in the vignette by (a) converting time to minutes, (b) convolving plasma concentration with the Mann 2022 ke0 = 0.001774 1/s effect-site equilibration (carried into Laffont 2024 Supp Table S3 unchanged for both nalmefene and naloxone), and (c) converting ng/mL to pM via the antagonist’s free-base molecular weight (naloxone 327.37 g/mol, nalmefene 339.43 g/mol); the resulting per-subject time series is supplied as this covariate.
• Units: pM (picomolar)
• Type: continuous
• Scope: specific
• Reference category: n/a.
• Source aliases: none.
• Example models: Mann_2022_mu_receptor_binding.R.
• Notes: Scope: specific. Same pM-unit / Table-S2-Kon-unit alignment requirement as L_OPIOID_pM.

CAR_OPIOID (canonical for time-varying fraction of mu-opioid receptors bound by an agonist input to the Mann 2022 physiology layer)

• Description: Time-varying fraction (0..1) of mu-opioid receptors bound by an opioid agonist. The Mann 2022 respiratory-physiology layer consumes this as a data covariate to drive opioid-induced reductions in wakefulness drive (W - Wmax * CAR^P3) and in chemoreflex drives (factor 1 - CAR^P1). In the composed Mann 2022 chain, this is the RL_op output of Mann_2022_mu_receptor_binding.R; in standalone physiology-only use, the user supplies CAR_OPIOID as a time-varying data column.
• Units: fraction (0..1)
• Type: continuous
• Scope: specific
• Reference category: n/a; 0 = no receptor occupancy = baseline ventilation.
• Source aliases: RL_op / CAR (binding-model output name).
• Example models: Mann_2022_respiratory_physiology.R.
• Notes: Scope: specific because the semantics are anchored to mu-opioid receptor occupancy in the Mann 2022 translational chain. Future opioid-pharmacology models that consume a different receptor-occupancy concept (e.g., kappa-opioid or delta-opioid) should register a separately named canonical with the receptor subtype in the name (e.g., CAR_KAPPA).

Q_TOTAL_LPM (canonical for total cardiac output input to the Mann 2022 opioid-PK shock-state Q_Scale feedback)

• Description: Time-varying total cardiac output Qb + Qt (cerebral + peripheral-tissue blood flow), in L/min, supplied to the Mann 2022 IV-opioid PK models so they can evaluate the FDA delaymymod.c lines 358-368 shock-state Q_Scale feedback: Q_Scale = 1 + 1 / (1 + exp((1.6 - Q_TOTAL_LPM / 4.87) / 0.05)), clamped to [1, 2]. Q_Scale scales the effective central volume of distribution down (vc_eff = vc / Q_Scale) so that hyperperfusion-driven concentration of opioid in the central / biophase compartment is captured during overdose-induced chemoreflex hyperperfusion. Without this feedback the standalone PK model under-estimates effect-site concentration in shock conditions and produces PaO2 troughs too shallow to reach the cardiac-arrest threshold.
• Units: L/min
• Type: continuous
• Scope: specific
• Reference category: 4.87 L/min (FDA delaymymod.c baseline Q_0; gives Q_Scale = 1, no amplification). Standalone PK use should leave Q_TOTAL_LPM at this baseline.
• Source aliases: q_total / Q_total / Q (physiology layer state name).
• Example models: Mann_2022_fentanyl_iv.R, Mann_2022_carfentanil_iv.R.
• Notes: Scope: specific because the 4.87 / 1.6 / 0.05 numerics are anchored to the FDA delaymymod.c Q_0 baseline and the empirical sigmoid centred at Q/Q_0 = 1.6. In the composed Mann 2022 chain Q_TOTAL_LPM is the Q_total = qb + qt output of Mann_2022_respiratory_physiology.R; in standalone PK-only use the user supplies Q_TOTAL_LPM = 4.87 (or whatever fixed baseline appropriate).

OPIOID_PATIENT_TYPE (canonical for opioid-naive vs chronic-opioid-user indicator in the Mann 2022 respiratory-depression PD layer)

• Description: Binary indicator selecting the pharmacodynamic-sensitivity parameter set in the Mann 2022 respiratory-physiology layer. 0 = healthy opioid-naive volunteer (P1 = 2.875, P3 = 0.9); 1 = chronic opioid user with established tolerance (P1 = 4.226, P3 = 1.323). P2 (metabolism exponent) is shared across both patient types at 0.06319. The naive vs chronic split is empirically calibrated against Algera 2021 (Clin Pharmacol Ther 2021;109(3):637-645) and Stoeckel 1982 (Br J Anaesth 1982;54(10):1087-1095); the numeric P1, P3 values are taken from FDA simulateToGetOD_IM.R lines 185-192 (Mann 2022 reference implementation).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (healthy opioid-naive).
• Source aliases: none.
• Example models: Mann_2022_respiratory_physiology.R.
• Notes: Scope: specific because the two parameter sets are tied to the Mann 2022 chronic-vs-naive opioid pharmacology calibration. A future model that captures a graded tolerance (e.g., a continuous “tolerance index”) rather than a two-class binary should register a separate continuous canonical.

DIS_HF_OR_LF_SEV (canonical for severe heart failure OR severe liver failure pooled indicator)

• Description: 1 = subject has severe heart failure (low cardiac output or pulmonary oedema) OR severe liver failure (advanced cirrhosis), 0 = neither. Pooled indicator used by source papers that observe similar PK effects from end-stage heart and end-stage liver disease (mechanistically: both reduce hepatic perfusion and cytochrome-mediated drug metabolism) and combine them into a single covariate when the per-group counts are too small for separate estimation. Mild and moderate heart or liver dysfunction are excluded by definition; the indicator captures only the severe ends of each axis.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no severe heart failure and no severe liver failure; includes normal function and mild/moderate dysfunction in either axis).
• Source aliases:
  • severe HF or LF – narrative composite used in Fattinger_1991_quinidine.R (Table 1 row ‘CL_nonrenal for patients with severe HF or LF’; severe HF defined as low output or pulmonary oedema, n = 2; severe LF defined as serum bilirubin > 30 umol/L AND prothrombin time < 60% of normal, n = 3; pooled because the two effects on non-renal CL were of similar magnitude and per-group counts were small).
• Example models: Fattinger_1991_quinidine.R (multiplicative reduction of the non-renal CL arm of total apparent CL from 12.6 L/h to 6.8 L/h when set to 1; log-multiplicative effect e_dis_hf_or_lf_sev_cl_nonren = log(6.8/12.6) = log(0.5397); Table 1 reduction in objective function 10.8, P < 0.005).
• Notes: Scope kept specific because the pooled HF-or-LF semantic is paper-specific (Fattinger 1991 pools the two severe end-organ-failure axes because the cohort had n = 2 + 3 subjects in those groups and similar effect sizes). A future model that retains severe HF and severe LF as separate covariates (with enough subjects in each to estimate them independently) should use the existing HEPIMP_SEV canonical for the severe-hepatic axis and register a parallel CARDIMP_SEV canonical for the severe-cardiac axis rather than reuse this pooled indicator. The DIS_HF_OR_LF_SEV pooling preserves the load-bearing convention of the source paper without forcing later users to artificially separate the two axes when the published evidence base lumps them.

FORM_QUIN_SR (canonical for slow-release quinidine bisulphate vs immediate-release quinidine sulphate formulation indicator)

• Description: 1 = subject received slow-release quinidine bisulphate (e.g., Kinidin duriles, Astra; oral slow-release tablets); 0 = subject received immediate-release quinidine sulphate (e.g., Chinidin sulfuricum, Siegfried; oral immediate-release). Per-dose-occasion indicator: in mixed-formulation cohorts a single subject may receive both formulations across dose records, with FORM_QUIN_SR set on each dose record to identify the formulation. The indicator drives the structural switch between formulation-specific zero-order absorption durations and relative bioavailability (slow-release QBS shows ~6 h duration of absorption vs ~1.4 h for immediate-release QS, and ~1.36-fold higher relative bioavailability vs the QS reference).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (immediate-release quinidine sulphate; the typical-value absorption-duration and bioavailability reference, with F = 1 fixed by convention per Fattinger 1991 abstract item 3).
• Source aliases: none – the source paper narratives the formulation as “quinidine sulphate (Chinidin sulfuricum)” vs “slow release quinidine bisulphate (Kinidin duriles)” without a single NMTRAN column name; the model column carries the canonical orientation directly.
• Example models: Fattinger_1991_quinidine.R (structural switch driving dur(central) between exp(ldur_qs + etaldur_qs) = 1.37 h * exp(eta) and exp(ldur_qbs) = 6.0 h, and f(central) between 1 (QS reference) and exp(lfdepot) = 1.36 (QBS); the IIV on QS absorption duration applies only when FORM_QUIN_SR = 0 per Methods page 282).
• Notes: Specific scope because the quinidine sulphate-vs-bisulphate contrast is tied to the specific drug. Drug-specific member of the FORM_* family alongside FORM_TAC_IR (tacrolimus immediate-release vs prolonged-release), FORM_LINAG_TAB1 (linagliptin tablet 1), FORM_VISMO_PHASEI (vismodegib Phase I dry-blend capsule), FORM_THEO_APNECUT (Apnecut vs theophylline-alcohol), FORM_ASV_LIQUID (asunaprevir liquid), and FORM_ABA_PHASE2 (abatacept SC Phase-2). Doses must be entered in mg of quinidine BASE (apply the Windholz 1983 stoichiometric factors of 0.829 mg base per mg quinidine sulphate and 0.663 mg base per mg quinidine bisulphate before passing to the model); the f(central) term then captures only the formulation-driven absorption difference, not the stoichiometric salt-vs-base conversion.

FORM_VPA_SR (canonical for sustained-release valproic acid tablet formulation indicator)

• Description: 1 = the valproic acid dose was given as a sustained-release tablet; 0 = the per-paper reference oral formulation (oral syrup in Zhang_2023_*). Per-dose-record indicator. Selects the sustained-release absorption rate constant in models that FIX a formulation-specific Ka set rather than estimating absorption.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (oral syrup). Pairs with FORM_TABLET to encode the three-level syrup / conventional-tablet / sustained-release-tablet stratification the valproate literature uses, with syrup as the reference level (both indicators 0) – the same derived-reference pattern as FORM_TABLET + FORM_CAPSULE in Kleideiter_2017_cebranopadol.R.
• Source aliases:
  • SR tablet – used in Zhang_2023_valproic_acid_base.R and Zhang_2023_valproic_acid_exponent.R (Supplementary Table S3 footnote: “Ka was fixed to 2.64, 1.57, 0.46 for syrup, conventional tablet and SR tablet, respectively”).
  • HS – Jiang 2007 uses an inverted indicator (HS = 1 for SR tablet) inside an additive Ka expression, Ka = 0.251 + 2.24 * (1 - HS); tabulated in Zhang 2023 Table 1. A model extracting Jiang 2007 directly should populate FORM_VPA_SR and negate in the model body rather than registering an inverted sibling.
• Example models: Zhang_2023_valproic_acid_base.R, Zhang_2023_valproic_acid_exponent.R (both encode ka <- exp(lka + e_form_tablet_ka * FORM_TABLET + e_form_vpa_sr_ka * FORM_VPA_SR) with lka FIXED at log(2.64) for the syrup reference and both shifts FIXED at the log-ratio of the published formulation-specific Ka values), Zhang_2023_valproic_acid_onebindingsite.R, Zhang_2023_valproic_acid_langmuir.R, Zhang_2023_valproic_acid_ddemax.R, Zhang_2023_valproic_acid_nonsaturable.R (all six models of the Zhang 2023 comparison encode ka <- exp(lka + e_form_tablet_ka * FORM_TABLET + e_form_vpa_sr_ka * FORM_VPA_SR) with lka FIXED at log(2.64) for the syrup reference and both shifts FIXED at the log-ratio of the published formulation-specific Ka values), Zhang_2024_valproic_acid.R (a two-level cohort – oral solution or sustained-release tablet only – so FORM_TABLET is absent and the model encodes ka <- exp(lka + e_form_vpa_sr_ka * FORM_VPA_SR) with lka FIXED at log(2.64); Zhang 2024 cites the same Ding 2015 source for the fixed 2.64 / 0.46 1/h pair), Wang_2024_valproic_acid.R (also a two-level cohort – oral syrup or sustained-release tablet – encoding ka <- exp(lka + e_form_vpa_sr_ka * FORM_VPA_SR) with lka FIXED at log(2.64); Wang 2024 cites Mei 2018 rather than Ding 2015 for the same fixed 2.64 / 0.46 1/h pair, confirming that the two constants circulate as a single literature convention rather than as one paper’s estimate), TeixeiradaSilva_2022_valproic_acid.R (three-level Spanish cohort – oral solution, gastro-resistant tablet, modified-release coated tablet – encoding ka <- exp(lka + e_form_tablet_ka * FORM_TABLET + e_form_vpa_sr_ka * FORM_VPA_SR) with lka FIXED at log(2.64) for the oral-solution reference; unlike the Ding 2015 triple, this paper fixes the tablet values at 0.78 1/h (gastro-resistant, FORM_TABLET = 1) and 0.38 1/h (modified-release coated, FORM_VPA_SR = 1), citing Methaneethorn 2017, per Teixeira-da-Silva 2022 Methods 2.3 and Supplementary Table S2).
• Notes: Specific scope because the fixed Ka triple (2.64 / 1.57 / 0.46 1/h) is a valproate-specific literature convention propagated across several paediatric valproate popPK papers (Ding 2015, Gu 2021 and thence Zhang 2023; Teixeira-da-Silva 2022 shares the 2.64 1/h solution value but fixes the tablet pair at 0.78 / 0.38 1/h instead). Distinct from FORM_QUIN_SR (quinidine bisulphate slow-release, where the contrast additionally carries a salt-vs-base stoichiometric conversion) and from the modified-release FORM_FLV_BID_XR / FORM_LOV_BID_XR pair (fluvastatin / lovastatin extended-release, where the indicator also encodes a dosing-frequency change). Trough-only therapeutic-drug-monitoring datasets identify absorption poorly, which is exactly why the source papers fix rather than estimate these constants – do not re-estimate lka from a trough-only cohort.

FORM_PAROXETINE_IR (canonical for paroxetine immediate-release tablet vs sustained-release tablet indicator)

• Description: 1 = the paroxetine dose was given as an immediate-release tablet; 0 = the dose was given as a sustained-release (enteric-coated controlled-release, 25 mg) tablet. Per-record indicator: in TDM cohorts the formulation is recorded per concentration and can change within a patient.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (sustained-release tablet; the typical-value reference for V/F in Li 2022).
• Source aliases:
  • formulation – used in Li_2022_paroxetine.R (Li 2022 Methods 2.3.2: “COV = 1 for immediate-release tablet; COV = 0 for sustained release tablet”). The canonical keeps the source orientation.
• Example models: Li_2022_paroxetine.R (linear fractional effect on V/F, V/F = 8850 * (1 - 0.666 * FORM_PAROXETINE_IR), so the immediate-release tablet has V/F = 2956 L and the sustained-release tablet 8850 L).
• Notes: Specific scope because the column is tied to the two marketed oral paroxetine tablet products. Li 2022’s Discussion describes the effect with the opposite orientation (“Compared with the IR-T, the SR-T resulted in a decrease in V/F of about 66.6%”), but the Methods coding is the one the model’s own typical-value simulations (Figure 3A, Figure 5) reproduce, so the Methods orientation is the authoritative one. Kim 2015 (Kim_2015_paroxetine.R) pooled immediate- and controlled-release paroxetine without a formulation covariate. A future paroxetine model that adds a third product (e.g., an oral suspension) should register a sibling indicator rather than overload this one.

FORM_LNP_SM102 (canonical for SM-102 ionizable-lipid RNA-lipid-nanoparticle formulation indicator)

• Description: 1 = the RNA-lipid-nanoparticle (LNP) dose was formulated with SM-102 as its ionizable lipid; 0 = the per-paper reference ionizable lipid (DLin-MC3-DMA / MC3 in Wang_2024_*). Subject- / arm-level indicator selecting an entire ionizable-lipid-specific parameter column. In an LNP the ionizable lipid is roughly half the particle by mole and governs plasma-protein adsorption, receptor-mediated uptake, particle disassembly and biodegradability, so swapping it changes the disposition parameters rather than only a bioavailability factor – which is why the indicator selects several parameters at once.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (DLin-MC3-DMA / MC3, the ionizable lipid of the first approved siRNA-LNP drug patisiran and the standard against which newer lipids are benchmarked). Pairs with FORM_LNP_LIPID5 to encode the three-level MC3 / SM-102 / Lipid 5 stratification, with MC3 as the reference level when both indicators are 0 – the same derived-reference pattern as FORM_TABLET + FORM_CAPSULE in Kleideiter_2017_cebranopadol.R.
• Source aliases:
  • SM-102 – the column heading of the Wang 2024 Figure 4A fitted-parameter table.
• Example models: Wang_2024_ionizableLipid_rat_pbpk.R (selects the spleen and “other organs” permeabilities, the “other organs” elimination rate, and the Eq. (8) scaling factors on the uptake, disassembly and metabolism rates; the liver permeability is deliberately not switched because Wang 2024 section 3.1.2 reports that “the permeability rate of MC3 applies to SM-102”).
• Notes: Follows the auto-approved FORM_<drug>_<formulation> canonical family, with LNP standing for the delivery system rather than for a single INN because the same ionizable lipid is used across many different RNA payloads. Distinct from the small-molecule FORM_* dosage-form indicators (FORM_TABLET, FORM_CAPSULE), which contrast presentations of one drug substance; here the indicator identifies a different chemical entity within the carrier. Register a sibling FORM_LNP_<lipid> canonical for each further ionizable lipid rather than overloading this entry.

FORM_LNP_LIPID5 (canonical for Lipid 5 ionizable-lipid RNA-lipid-nanoparticle formulation indicator)

• Description: 1 = the RNA-LNP dose was formulated with Lipid 5 as its ionizable lipid; 0 = the per-paper reference ionizable lipid (MC3 in Wang_2024_*).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (DLin-MC3-DMA / MC3). Paired with FORM_LNP_SM102; both 0 selects MC3, and the two are mutually exclusive.
• Source aliases:
  • Lipid 5 – the column heading of the Wang 2024 Figure 4A fitted-parameter table.
• Example models: Wang_2024_ionizableLipid_rat_pbpk.R (selects all three permeabilities including the liver one, which Wang 2024 section 3.1.2 reports “had to be down-regulated to get a satisfactory fitting” unlike SM-102, plus the “other organs” elimination rate and the Eq. (8) uptake / disassembly / metabolism scaling factors).
• Notes: Sibling of FORM_LNP_SM102 under the FORM_<drug>_<formulation> family. Lipid 5 and SM-102 differ only in the position of one ester linker, which is why the two share several parameters but not the liver permeability.

FORM_LNP_DMAPBLP78 (canonical for 78 nm DMAP-BLP RNA-lipid-nanoparticle formulation indicator)

• Description: 1 = the RNA-LNP dose was formulated with the ionizable lipid DMAP-BLP at a particle diameter of about 78 nm; 0 = the per-paper reference formulation (MC3 at about 80 nm in Wang_2024_*). This indicator carries a joint lipid-and-size contrast rather than size alone, because the source study changed both together.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (MC3 LNP at about 80 nm). Pairs with FORM_LNP_DMAPBLP45 to encode the three-level 80 nm MC3 / 78 nm DMAP-BLP / 45 nm DMAP-BLP stratification, with the MC3 arm as the reference level when both indicators are 0.
• Source aliases:
  • DMAP-BLP – the column heading of the Wang 2024 Figure 5B fitted-parameter table.
  • DMAP-DLP – the spelling used in Wang 2024 Results section 3.1.3 and Figure S6 for the same lipid; the Methods and figure-table spelling DMAP-BLP is canonical here.
• Example models: Wang_2024_ionizableLipid_mouse_pbpk.R (selects the spleen and “other organs” permeabilities and the plasma dissociation rate kf = 0.0414 1/h; the liver permeability, uptake rate and “other organs” elimination rate are inherited from the 80 nm arm under the Figure 5B “#” footnote).
• Notes: Follows the auto-approved FORM_<drug>_<formulation> family. Particle diameter is embedded in the indicator name rather than carried as a separate continuous covariate because the source paper fitted a discrete parameter set per particle-size arm rather than a size-parameterised relationship; a future model that fits a continuous size-permeability relationship should register a continuous SIZE_LNP_NM covariate instead of extending this family.

FORM_LNP_DMAPBLP45 (canonical for 45 nm DMAP-BLP RNA-lipid-nanoparticle formulation indicator)

• Description: 1 = the RNA-LNP dose was formulated with the ionizable lipid DMAP-BLP at a particle diameter of about 45 nm; 0 = the per-paper reference formulation (MC3 at about 80 nm in Wang_2024_*).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (MC3 LNP at about 80 nm). Paired with FORM_LNP_DMAPBLP78; both 0 selects the MC3 arm, and the two are mutually exclusive.
• Source aliases:
  • DMAP-BLP – the column heading of the Wang 2024 Figure 5C fitted-parameter table.
• Example models: Wang_2024_ionizableLipid_mouse_pbpk.R (selects an independently re-fitted parameter set: a much higher uptake rate of 125.06 vs 5.24 mL/mmol/h, a lower liver permeability, and the plasma dissociation rate kf = 0.1089 1/h from Figure 6A).
• Notes: Sibling of FORM_LNP_DMAPBLP78. The contrast between the two is the paper’s central size comparison: the smaller particle is taken up into the liver much faster yet knocks down its target gene less effectively.

FORM_LNP_MC3 (canonical for DLin-MC3-DMA ionizable-lipid RNA-lipid-nanoparticle formulation indicator)

• Description: 1 = the RNA-LNP was formulated with DLin-MC3-DMA (MC3) as its ionizable lipid; 0 = the per-paper reference ionizable lipid (C12-200 in Wang_2024_ionizableLipid_hela_qsp.R). Use this indicator when MC3 is a comparator rather than the reference; when MC3 is the reference, leave the sibling FORM_LNP_* indicators at 0 instead of adding this column.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the per-paper reference ionizable lipid; C12-200 in Wang 2024’s cellular model, which is the formulation the shared trafficking rates were fitted to). Pairs with FORM_LNP_L319 to encode the three-level C12-200 / MC3 / L319 stratification.
• Source aliases:
  • MC3 – the column heading of the Wang 2024 Figure 9D parameter-comparison table.
• Example models: Wang_2024_ionizableLipid_hela_qsp.R (selects the MC3 endosomal-escape rate krel = 0.0058 1/h; MC3’s own cytoplasmic-delivery fraction frel was not measured and is assumed equal to the L319 value of 0.5).
• Notes: Sibling of FORM_LNP_SM102 / FORM_LNP_LIPID5 / FORM_LNP_L319 under the FORM_<drug>_<formulation> family. MC3 is the reference level in Wang 2024’s in-vivo models but a comparator in its cellular model, which is why it needs its own indicator; the reference level of an FORM_LNP_* set is always documented per model rather than assumed.

FORM_LNP_L319 (canonical for L319 ionizable-lipid RNA-lipid-nanoparticle formulation indicator)

• Description: 1 = the RNA-LNP was formulated with L319 as its ionizable lipid; 0 = the per-paper reference ionizable lipid (C12-200 in Wang_2024_ionizableLipid_hela_qsp.R).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (C12-200). Paired with FORM_LNP_MC3; both 0 selects C12-200, and the two are mutually exclusive.
• Source aliases:
  • L319 – the column heading of the Wang 2024 Figure 9D parameter-comparison table.
• Example models: Wang_2024_ionizableLipid_hela_qsp.R (selects the L319 endosomal-escape rate krel = 0.0157 1/h; L319 is the only formulation for which the cytoplasmic-delivery fraction frel was measured directly, at 0.5).
• Notes: Sibling of FORM_LNP_MC3 under the FORM_<drug>_<formulation> family.

FORM_WIXELA_INHUB (canonical for the Wixela Inhub fluticasone propionate / salmeterol dry-powder-inhaler test-product indicator)

• Description: 1 = the inhalation was taken from the Wixela Inhub dry powder inhaler (Viatris / Mylan), the generic test product in the Rosenborg 2025 bioequivalence analysis; 0 = Advair Diskus (GSK), the brand reference product. Per-dose-record indicator – in a two-way crossover the same subject carries 1 on one period’s dose row and 0 on the other’s.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (Advair Diskus reference inhaler), which is the relative-bioavailability anchor F = 1.
• Source aliases:
  • TREA – Rosenborg 2025 supplement Sect. 1: “Treatment alternative, test (TREA=1) and reference (TREA=2) formulation”. The supplement’s NONMEM code derives TREA1 / TREA2 indicators from it; FORM_WIXELA_INHUB is TREA1.
• Example models: Rosenborg_2025_fluticasone_300ug.R, Rosenborg_2025_fluticasone_750ug.R, Rosenborg_2025_fluticasone_1500ug.R, Rosenborg_2025_salmeterol.R (gates both the monophasic absorption rate constant, ka <- exp(lka_test * FORM_WIXELA_INHUB + lka_ref * (1 - FORM_WIXELA_INHUB) + etalka), and relative bioavailability, f(depot) <- exp((lfdepot + etalfdepot) * FORM_WIXELA_INHUB)).
• Notes: Both arms are dry powder inhalers delivering the same fixed-dose combination of fluticasone propionate and salmeterol, so the contrast is a product / device difference rather than a dosage-form difference; reusing FORM_POWDER would be uninformative when both arms are powders. The same indicator serves the fluticasone propionate and the salmeterol models because a single inhalation delivers both actives from one device. Unlike FORM_NOSCAPINE_TEST, which gates bioavailability alone, this indicator also switches the absorption rate constant: Rosenborg 2025 Table 2 reports a separate k41(test) and k41(ref) per model, sharing one inter-individual random effect. Specific scope because the contrast identifies two named commercial inhaler products.

FORM_NOSCAPINE_TEST (canonical for the reformulated noscapine oral-suspension test-product indicator)

• Description: 1 = the noscapine dose was given as the reformulated oral suspension (InfectoPharm, Heppenheim) that served as the test product in the Chen 2024 bioequivalence study; 0 = the Nipaxon 5 mg/mL oral suspension (McNeil, Solna) reference product. Per-dose-record indicator – in a crossover design the same subject carries 1 on one period’s dose row and 0 on the other’s.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (Nipaxon reference oral suspension), which is the relative-bioavailability anchor F = 1.
• Source aliases:
  • formulation (test vs reference) – Chen 2024 (Sect. 2.1 describes the two products; Table 4 reports the single formulation parameter F1 (%) = 82.8).
• Example models: Chen_2024_noscapine.R (gates relative bioavailability only: f(transit1) <- exp((lf1 + etalf1) * FORM_NOSCAPINE_TEST) with lf1 = log(0.828), so reference doses get F = 1 exactly while test doses get 82.8% carrying 34.1% CV inter-individual variability).
• Notes: Both products are oral suspensions of the same strength, so the contrast is a product difference, not a dosage-form difference – hence the drug-plus-product naming rather than reuse of FORM_SUSPENSION (which distinguishes a suspension from another dosage form and would be uninformative when both arms are suspensions). Bioavailability is the only parameter that differs by formulation: Chen 2024 Sect. 3.3.2 reports that “only apparent bioavailability differed between test and reference preparations”, and the Discussion attributes it to “a lower amount of drug released from the suspension” rather than to an absorption-rate difference. Contrast with FORM_QUIN_SR / FORM_VPA_SR, where the formulation indicator switches an absorption-rate or duration parameter as well. Specific scope because the contrast identifies two named commercial noscapine products.

FORM_SEP4199_IR, FORM_SEP4199_SOLUTION, FORM_SEP4199_CR10, FORM_SEP4199_CR15, FORM_SEP4199_CR20, FORM_SEP4199_CR25, FORM_SEP4199_CR25_FED, FORM_SEP4199_CR40, FORM_SEP4199_MUPS30, FORM_SEP4199_MUPS225 (canonical for the ten SEP-4199 nonracemic-amisulpride formulation prototypes)

• Description: Ten mutually exclusive and exhaustive binary indicators identifying which of the ten SEP-4199 (nonracemic amisulpride, a fixed 85:15 aramisulpride:esamisulpride ratio) oral formulation prototypes a dose was given as. Exactly one is 1 on any dose record. FORM_SEP4199_IR = the immediate-release tablet; FORM_SEP4199_SOLUTION = the oral solution used by the PET studies; FORM_SEP4199_CR10 / _CR15 / _CR20 / _CR25 / _CR40 = controlled-release tablets carrying 10, 15, 20, 25 and 40 percent of the rate-controlling polymer; FORM_SEP4199_CR25_FED = the 25 percent controlled-release tablet administered with food; FORM_SEP4199_MUPS30 / _MUPS225 = multiple-unit pellet system capsules at 30 and 22.5 percent polymer. Each indicator selects a complete parameter set rather than a shift off a reference: a bioavailability triple (the fractions absorbed through each of three parallel absorption processes) and a Weibull in vitro dissolution pair (rate scaling and shape). A coarser dosage-form class – solution, tablet or capsule – is derived inside model from these ten and selects the three absorption rate constants.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: n/a. The set is exhaustive and each member carries its own independently estimated parameters, so there is no reference level whose effect is fixed at zero; the indicators multiplex a ten-way categorical rather than contrasting against a baseline. FORM_SEP4199_IR is nonetheless the clinical reference, being the formulation carried through the Phase II bipolar depression study (NCT03543410) against which the controlled-release prototypes were compared.
• Source aliases:
  • RGN – the NONMEM $INPUT regimen column of both Hopkins 2024 control streams, an integer 1..10 dispatched through IF (RGN.EQ.k) blocks. The mapping is RGN1 = _CR10, RGN2 = _CR25, RGN3 = _IR, RGN4 = _SOLUTION, RGN5 = _CR15, RGN6 = _CR25_FED, RGN7 = _MUPS30, RGN8 = _MUPS225, RGN9 = _CR20, RGN10 = _CR40. Re-express as FORM_SEP4199_<name> = as.integer(RGN == k).
• Example models: Hopkins_2024_amisulpride.R (founding example; each indicator selects f(depot1) / f(depot2) / f(depot3) and the ra / gam1 Weibull dissolution pair, and FORM_SEP4199_SOLUTION additionally switches the Weibull dissolution term off entirely because a solution has no solid-state dissolution step).
• Notes: Members of the auto-approved FORM_<drug>_<formulation> family. Named on the development code SEP-4199 rather than on the INN amisulpride because the entity is the fixed nonracemic 85:15 ratio rather than racemic amisulpride, and because these are specific prototype formulations of that entity; a future racemic-amisulpride formulation model should register its own siblings rather than reuse these. Time-varying within subject: Study 5 used a partial-crossover design in which one subject received several prototypes across dosing periods, so the indicator set must be written per dose record and is naturally paired with OCC. The fed arm is carried here as its own formulation rather than through a separate food covariate, because the source estimates a distinct bioavailability triple and Weibull pair for it rather than a food coefficient; FED is consequently present in the NONMEM $INPUT but never referenced in $PK, and the model records it in covariatesDataExcluded. Distinct from FORM_TABLET / FORM_CAPSULE / FORM_SOLUTION, which contrast dosage forms generically: seven of these ten are tablets and two are capsules, so a dosage-form indicator alone cannot identify the parameter set.

Receptor-binding ligand selection and pharmacological-chain inputs

OPIOID_ID (canonical for opioid-agonist ligand selector in the Mann 2022 mu-receptor binding panel)

• Description: Integer 1..13 selecting which ligand from the Mann 2022 Supplement 1 Table S2 binding-kinetic panel (1..12) or the Laffont 2024 Supplementary Table S3 scaling-derived extension (13 = nalmefene) occupies the opioid-agonist slot of a multi-ligand competitive mu-opioid-receptor binding model at simulation time. The same compiled binding model can simulate any agonist in the panel without re-instantiation by varying OPIOID_ID per subject (or per simulation arm). Mapping (preserved verbatim in Mann_2022_mu_receptor_binding.R::ini): 1 = alfentanil; 2 = buprenorphine; 3 = butyryl fentanyl; 4 = carfentanil; 5 = fluorobutyryl fentanyl; 6 = fentanyl; 7 = fluoroisobutyryl fentanyl; 8 = furanyl fentanyl; 9 = isobutyryl fentanyl; 10 = naloxone; 11 = remifentanil; 12 = sufentanil; 13 = nalmefene (added from the Laffont 2024 Supp Table S3 scaling approach over Cassel 2005). Out-of-range values cause every dispatch indicator to evaluate to 0, which zeros the agonist slot rates rather than throwing an error.
• Units: (categorical)
• Type: categorical
• Scope: specific
• Reference category: n/a (every value selects a distinct ligand parameter set).
• Source aliases: none (new canonical introduced with Mann 2022 extraction).
• Example models: Mann_2022_mu_receptor_binding.R.
• Notes: Scope: specific because the integer-to-ligand mapping is anchored to Mann 2022 Table S2 row order, with the Laffont 2024 / Laffont 2025 13th-ligand extension preserving that anchoring (nalmefene appended at index 13, not inserted alphabetically). The maintainers authorised this 13th-ligand expansion on 2026-05-29; it was carried out after confirming the nalmefene Kon/Koff/n values are explicitly published as final values in Laffont 2024 Supplementary Table S3 (a primary source).

ANTAGONIST_ID (canonical for opioid-antagonist ligand selector in the Mann 2022 mu-receptor binding panel)

• Description: Integer 1..13 selecting which ligand from the Mann 2022 Supplement 1 Table S2 (1..12) or the Laffont 2024 Supplementary Table S3 scaling-derived extension (13 = nalmefene) occupies the antagonist slot of the multi-ligand competitive mu-opioid-receptor binding model at simulation time. Uses the same integer-to-ligand mapping as OPIOID_ID. Mann 2022 itself uses only naloxone (10) as the antagonist; Laffont 2024 / Laffont 2025 add nalmefene (13) as the second commercially-relevant intranasal antagonist option. The wider 1..13 range is retained so downstream tasks can flip slots without code changes (e.g., partial-agonist buprenorphine 2 in the antagonist slot to model receptor occupancy without full agonist effect).
• Units: (categorical)
• Type: categorical
• Scope: specific
• Reference category: n/a (every value selects a distinct ligand parameter set).
• Source aliases: none.
• Example models: Mann_2022_mu_receptor_binding.R.
• Notes: Scope: specific. Same row-order semantics as OPIOID_ID (Mann 2022 Table S2 rows 1..12 verbatim; Laffont 2024 Supp Table S3 nalmefene at row 13). Out-of-range values zero the antagonist slot rates safely.

L_OPIOID_pM (canonical for time-varying opioid-agonist effect-site concentration input to the Mann 2022 binding layer)

• Description: Time-varying opioid-agonist effect-site (biophase) concentration in picomolar (pM), supplied as a data covariate to the multi-ligand competitive mu-receptor binding model. The pM unit matches the per-second Kon units (pM^-n s^-1) tabulated in Mann 2022 Supplement 1 Table S2; the binding ODE inside Mann_2022_mu_receptor_binding.R consumes this column directly. In a composed Mann 2022 chain, the upstream IV-opioid PK layer (Mann_2022_fentanyl_iv.R or Mann_2022_carfentanil_iv.R) exposes its effect-site Ce_pM as this covariate; in standalone use, the user supplies it as a time-varying data column on the subject’s records.
• Units: pM (picomolar)
• Type: continuous
• Scope: specific
• Reference category: n/a.
• Source aliases: Ce_pM (upstream PK output name when composed in-chain).
• Example models: Mann_2022_mu_receptor_binding.R.
• Notes: Scope: specific because the unit choice (pM) is tied to the Mann 2022 binding-rate parameterisation. A future binding model that uses nM or mg/L should register a separately named canonical (e.g., L_OPIOID_nM) rather than reusing this name with a different unit, because the binding model’s downstream math depends on the unit match.

AGONIST_CODE (canonical for vasoactive-agonist ligand selector in the Grzesk 2016 vascular-reactivity sigmoidal Emax CRC model)

• Description: Integer 1..4 selecting which vasoactive agonist from the Grzesk 2016 Table I CRC panel occupies the agonist slot of the sigmoidal Emax concentration-response model at simulation time. The same compiled model can simulate any of the four agonists by varying AGONIST_CODE per record (or per simulation arm). Mapping (preserved verbatim in Grzesk_2016_m3M3FBS.R::ini parameter names): 1 = phenylephrine (PHE; alpha1-adrenergic receptor agonist); 2 = arg-vasopressin (AVP; V1 vasopressin receptor agonist); 3 = mastoparan-7 (heterotrimeric G-protein direct activator); 4 = Bay K8644 (L-type voltage-gated calcium channel agonist). Out-of-range values cause every dispatch indicator to evaluate to 0, which zeros the sigmoidal Emax expression rather than throwing an error.
• Units: (categorical)
• Type: categorical
• Scope: specific
• Reference category: n/a (every value selects a distinct (EC50, Emax) parameter pair; companion M3M3FBS_PRESENT selects within each agonist between the control and +m-3M3FBS sub-pairs).
• Source aliases: none – Grzesk 2016 Tables I and II label the four agonists in prose (“PHE”, “AVP”, “mastoparan-7”, “Bay K8644”); the model column is the canonical AGONIST_CODE.
• Example models: Grzesk_2016_m3M3FBS.R (Grzesk 2016 Table I + Table II Phase 2: per-agonist EC50 (M/L) and maximal-perfusion-pressure Emax (mmHg) pairs; companion M3M3FBS_PRESENT flips between control and m-3M3FBS-pretreatment sub-pairs).
• Notes: Specific scope because the integer-to-agonist mapping is anchored to Grzesk 2016 Table I row order. The Grzesk-group sibling papers (Biomed Rep 2 / 2014 pertussis toxin; Mol Med Report 5 / 2012 calcium blockers; Exp Ther Med 4 / 2012; etc.) reuse the same isolated-perfused-tail-artery scaffold with overlapping agonist panels (PHE / AVP / mastoparan-7 are common across the series; the third “channel” probe varies); a future extraction of any of those papers can extend the integer mapping at row 5+ rather than re-introduce a parallel selector canonical. Distinct from OPIOID_ID / ANTAGONIST_ID (Mann 2022 opioid-binding panel) because the agonists here are not within a single receptor-binding family – PHE, AVP, mastoparan-7, and Bay K8644 hit four mechanistically distinct contractile pathways (alpha1, V1, G-protein direct, L-type Ca channel), and the model uses agonist-specific (EC50, Emax) pairs rather than a shared binding-kinetic parameter set.

STUDY_TOMORROW (canonical for TOMORROW (1199.30) IPF phase II study indicator)

• Description: 1 = subject enrolled in the TOMORROW phase II trial (NCT00514683; Richeldi 2011 NEJM), the BI 1199.30 protocol that evaluated nintedanib 50 mg QD, 50 mg BID, 100 mg BID, or 150 mg BID vs placebo over 52 weeks in idiopathic pulmonary fibrosis (IPF); 0 = subject from any other study in the Schmid 2017 nintedanib pooled-PK analysis (NSCLC phase II, LUME-Lung 1, or LUME-Lung 2). Used as a per-subject categorical indicator for the trial-effect covariate on nintedanib first-order absorption rate ka.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-TOMORROW; in Schmid 2017 the trial-effect reference for ka is the LUME-Lung 1 + LUME-Lung 2 phase III NSCLC pair).
• Source aliases: derived per subject from the trial identifier.
• Example models: Schmid_2017_nintedanib.R (combined with STUDY_NSCLC_NIN_PH2 to drive the ka phase-II trial-group effect = 2.20x; also captures the ka2 BIBF 1202 metabolite phase-II trial-group effect = 0.756x).
• Notes: Specific scope because the trial-effect grouping is paper-specific.

STUDY_NSCLC_NIN_PH2 (canonical for nintedanib NSCLC phase II (Reck 2011) study indicator)

• Description: 1 = subject enrolled in the nintedanib phase II NSCLC trial reported by Reck et al. 2011 (Annals of Oncology 22(6):1374-1381; BI 1199.4 protocol), which compared nintedanib 150 or 250 mg BID monotherapy in relapsed advanced NSCLC; 0 = subject from any other study in the Schmid 2017 pooled-PK analysis (TOMORROW IPF phase II, LUME-Lung 1, or LUME-Lung 2). Used as a per-subject categorical indicator for two distinct trial-effect covariates on F1 (relative bioavailability) and ka (absorption rate).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-NSCLC-phase-II; in Schmid 2017 the trial-effect reference for F1 is the LUME-Lung 1 + TOMORROW pair, and the reference for ka / ka2 is the LUME-Lung 1 + LUME-Lung 2 pair).
• Source aliases: derived per subject from the trial identifier.
• Example models: Schmid_2017_nintedanib.R (combined with STUDY_LUMELUNG2 to drive the F1 trial-group effect = 1.30x; combined with STUDY_TOMORROW to drive the ka phase-II trial-group effect = 2.20x and the ka2 BIBF 1202 metabolite phase-II trial-group effect = 0.756x).
• Notes: Specific scope. The Schmid 2017 paper’s prose refers to this trial as “NSCLC phase II [23]”; the BI protocol code (1199.4) is not explicit in the paper but is the standard BIBF 1120 NSCLC phase II Reck 2011 study.

STUDY_LUMELUNG2 (canonical for LUME-Lung 2 (nintedanib + pemetrexed NSCLC phase III) study indicator)

• Description: 1 = subject enrolled in LUME-Lung 2 (NCT00806819; Hanna 2016 Lung Cancer 102:65-73; BI 1199.14 protocol), the phase III trial of nintedanib 200 mg BID + pemetrexed 500 mg/m^2 vs placebo + pemetrexed in second-line non-squamous NSCLC; 0 = subject from any other study in the Schmid 2017 pooled-PK analysis (TOMORROW IPF phase II, NSCLC phase II, or LUME-Lung 1). Used as a per-subject categorical indicator paired with STUDY_NSCLC_NIN_PH2 for the F1 trial-group effect.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (non-LUME-Lung-2).
• Source aliases: derived per subject from the trial identifier.
• Example models: Schmid_2017_nintedanib.R (combined with STUDY_NSCLC_NIN_PH2 to drive the F1 trial-group effect = 1.30x; acts as the reference cohort for the ka / ka2 trial-effect).
• Notes: Specific scope.

DOSE_AGT_UG (canonical for per-observation angiotensin challenge dose in ug of angiotensin II equivalents)

• Description: Per-observation intravenous-bolus dose of exogenous angiotensin administered as a pharmacologic probe during an angiotensin-challenge phase I protocol, expressed as ug of angiotensin II equivalents. For an angiotensin II bolus this is the literal injected mass; for an angiotensin I bolus, multiply by Q = 0.78 (molar-weight ratio between angiotensin I and angiotensin II) before populating this column. Each row in the event table carries the dose given immediately before the observed BP peak was sampled.
• Units: ug (Ang II equivalents)
• Type: continuous
• Scope: specific
• Reference category: none (continuous).
• Source aliases:
  • D x Q – used in BuchwalderCsajka_1999_angiotensin.R (Buchwalder-Csajka 1999 Table 1 row 5 column header, where D is the raw angiotensin dose in ug and Q is the molar-weight conversion factor).
• Example models: BuchwalderCsajka_1999_angiotensin.R (drives the algebraic Emax dose-response E = Emax * DOSE_AGT_UG / (DOSE_AGT_UG + ED50) separately for SBP and DBP).
• Notes: Per-observation challenge-dose column for the algebraic Emax angiotensin-challenge PD model. Differs from the standard rxode2 AMT / EVID = 1 dosing column because the model is purely algebraic (no PK, no ODEs); the dose enters the model() block as a covariate symbol rather than via a dosing event. Specific scope because the column’s semantics (Q-corrected angiotensin dose) are tied to angiotensin-challenge protocols; future angiotensin-challenge extractions may reuse this canonical.

DOSE_ISOPROTERENOL_UG (canonical for per-observation isoproterenol challenge dose in ug)

• Description: Per-observation intravenous-bolus dose of isoproterenol (isoprenaline) administered as a beta-adrenergic agonist probe during an isoproterenol sensitivity test (IST), expressed in ug. Each observation row in the event table carries the isoproterenol dose given immediately before the heart-rate response was sampled; rows outside an IST challenge carry 0, which recovers the drug-free baseline heart rate.
• Units: ug
• Type: continuous
• Scope: specific
• Reference category: none (continuous). Set to 0 for observation rows with no isoproterenol challenge; the Emax term then vanishes and the readout reduces to the baseline E0.
• Source aliases:
  • D – used in Hwang_2023_carvedilol.R (Hwang 2023 equations (3) and (4) and Fig. 1, where D is the isoproterenol dose driving the direct-effect Emax heart-rate response).
• Example models: Hwang_2023_carvedilol.R (drives the competitive-antagonism direct-effect Emax heart-rate model HR = E0 + Emax * DOSE_ISOPROTERENOL_UG / (ED50 * (1 + Cc / IC50) + DOSE_ISOPROTERENOL_UG), in which the carvedilol plasma concentration inflates the apparent isoproterenol ED50).
• Notes: Per-observation challenge-dose column for algebraic Emax agonist-challenge PD models, following the DOSE_AGT_UG precedent (angiotensin challenge; Buchwalder-Csajka 1999). Differs from the standard rxode2 AMT / EVID = 1 dosing column because the challenge agonist has no PK compartment in the model: the dose enters model() as a covariate symbol rather than via a dosing event. Hwang 2023 Fig. 1 draws an isoproterenol compartment with an elimination rate constant k, but no value for k is reported anywhere in the paper and the final-model equation printed in the same figure is algebraic in the dose D, so the dose-as-covariate encoding is the one the published parameters support. Specific scope because the semantics are tied to isoproterenol sensitivity-test protocols; future beta-blocker IST extractions should reuse this canonical.

AUC_EMPA (canonical for per-subject steady-state AUC of empagliflozin)

• Description: Per-subject (time-fixed) steady-state AUC of empagliflozin over the q24h dosing interval, supplied as a static drug-exposure covariate to PD models that consume the upstream popPK without instantiating a PK ODE. The source authors generate AUCss values from individual empirical Bayes estimates of an upstream empagliflozin popPK analysis (Mondick 2018 plus additional EASE-2 / EASE-3 data on file in the M-EASE-2 application) and pass them as a static covariate column to the M-EASE-2 PD model.
• Units: nmol*h/L (document per-model via covariateData[[AUC_EMPA]]$units if a different exposure unit is reported). Yao_2023_sglt2_endpoints_mbma.R reports the same quantity in ng*h/mL; empagliflozin’s molecular weight is 450.91 g/mol, so 1 ngh/mL = 2.218 nmolh/L. The unit is load-bearing because it must match the EC50 / AUC50 the consuming model carries.
• Type: continuous
• Scope: specific
• Reference category: n/a – enters via direct Emax form emax_i * AUC_EMPA / (auc50 + AUC_EMPA) on the HbA1c drug-effect term. Set AUC_EMPA = 0 to recover the placebo arm (drug-effect term vanishes; HbA1c follows baseline + placebo drift only). Reference values observed: AUC50 = 498 nmol*h/L (Johnston 2019 Table 2 typical-value AUC at half-maximal effect; corresponds approximately to the simulated median AUCss for the 2.5 mg QD arm).
• Source aliases:
  • AUCss (or AUCSS,i) – the printed variable name in Johnston 2019 Equation 1 (the individual steady-state AUC entering the Emax expression).
• Example models: Johnston_2019_empagliflozin.R (M-EASE-2 exposure-response PD model for HbA1c in adults with type 1 diabetes; AUC_EMPA drives the direct Emax HbA1c-reduction term), Johnston_2021_empagliflozin_MEASE1.R (semi-mechanistic exposure-response PD model on TDID / MDG / HbA1c in adults with T1D; AUC_EMPA drives the direct Emax reductions on both TDID and MDG, which in turn drive HbA1c), Yao_2023_sglt2_endpoints_mbma.R (study-arm-level steady-state 24 h AUC in ng*h/mL rather than nmol*h/L, driving the shared-Emax / drug-specific-EC50 model of the dUGEc translational biomarker in a three-drug SGLT2-inhibitor MBMA; EC50 = 841 ng*h/mL).
• Notes: Specific scope because the column meaning is tied to empagliflozin and to a q24h steady-state AUC convention. Sibling drug-specific AUC canonicals (AUC_CARBO, AUC_GEM, AUC_BAST_FW, AUC_PAZO, AUC_GCV, AUC_RTV, AUC_DAPA, AUC_CANA) follow the same AUC_<DRUG> naming pattern. A future PK/PD model that uses a different empagliflozin exposure metric (trough concentration, q12h-interval AUC) should register a parallel canonical rather than overload AUC_EMPA. Distinct from DOSE_EMPA_MGD, which carries the daily dose rather than the exposure it produces; Baron_2016_empagliflozin.R computes an AUCss internally from DOSE_EMPA_MGD and so does not use this column.

AUC_CANA (canonical for steady-state 24 h AUC of canagliflozin)

• Description: Canagliflozin area under the plasma concentration-time curve over the once-daily 24 h dosing interval at steady state, supplied as a static drug-exposure column to PD / disease-endpoint models that consume an upstream population PK analysis without instantiating a PK ODE. Study-arm-level in model-based meta-analyses (one value per published trial arm); per-subject in individual-level exposure-response analyses.
• Units: ng*h/mL. Document per-model via covariateData[[AUC_CANA]]$units. Must be on the same scale as the EC50 the consuming model carries.
• Type: continuous
• Scope: specific
• Reference category: n/a – enters via an Emax form Emax * AUC_CANA / (EC50 + AUC_CANA). Set to 0 for placebo arms and for arms treated with a different SGLT2 inhibitor; the drug term then vanishes exactly. Reference value observed: EC50 = 2310 ngh/mL for the change in urinary glucose excretion normalised by baseline FPG (Yao 2023 Table 2); the 300 mg label dose gives 25000 ngh/mL from the companion popPK (1000 * 300 mg / 12.0 L/h).
• Source aliases:
  • AUC0-24h – the printed name in Yao 2023 Equation 2, Figure 2 and the PK/PD model section. Same units and orientation as the canonical, so no value transformation.
• Example models: Yao_2023_sglt2_endpoints_mbma.R (study-arm-level steady-state 24 h AUC driving the shared-Emax / drug-specific-EC50 model of the dUGEc translational biomarker).
• Notes: Specific scope because the column meaning is tied to canagliflozin and to the once-daily steady-state 24 h AUC convention. Member of the AUC_<DRUG> family; sibling of AUC_EMPA and AUC_DAPA. Compute it from modellib('Yao_2023_canagliflozin_mbma') as total daily dose divided by CL/F. Note that the Yao 2023 canagliflozin apparent clearance (12.0 L/h) is roughly a third of the value implied by the labelled single-dose AUC, so AUC_CANA values derived from that model are correspondingly larger than a label-derived AUC; the paper’s canagliflozin EC50 is scaled to the same internal exposure, so the pair is self-consistent and must not be mixed with an externally derived AUC. A future canagliflozin model using a different exposure metric should register a parallel canonical rather than overload this one.

AUC_DAPA (canonical for per-subject steady-state 24 h AUC of dapagliflozin)

• Description: Per-subject (time-fixed) steady-state AUC of dapagliflozin over the q24h dosing interval, supplied as a static drug-exposure covariate to exposure-response models that consume an upstream population PK analysis without instantiating a PK sub-model. Sokolov 2023 Table 1 records the values as “taken from a previously performed population PK analysis (Melin et al., 2022)” and passes them to the PD model as a Monolix regressor.
• Units: ng*h/mL (document per-model via covariateData[[AUC_DAPA]]$units if a different exposure unit is reported).
• Type: continuous
• Scope: specific
• Reference category: n/a – enters via inhibitory Imax forms 1 - imax * AUC_DAPA / (auc50 + AUC_DAPA) on the basal insulin dose, average daily glucose and HbA1c ratios. Set AUC_DAPA = 0 to recover the placebo arm exactly (every Imax term vanishes). Reference values observed: IAUC50 = 38.8 ngh/mL (basal insulin dose) and 67.4 ngh/mL (glucose, and the direct HbA1c effect fixed to the same value) per Sokolov 2023 Table 3. Mean AUC by dose (Sokolov 2023 Figure 10 caption): 51.4, 130.6, 294.5 and 594.3 ng*h/mL at 1, 2.5, 5 and 10 mg once daily.
• Source aliases:
  • AUC – the printed variable name in Sokolov 2023 Supplementary Equations 1-3 and the Table 1 measurement row “Dapagliflozin AUC”.
• Example models: Sokolov_2023_dapagliflozin.R (semi-mechanistic exposure-response model on the basal insulin dose, CGM glucose and HbA1c ratios in adults with type 1 diabetes; AUC_DAPA drives all three Imax terms), Yao_2023_sglt2_endpoints_mbma.R (merged from a duplicate register entry during merge dedup).
• Notes: Specific scope because the column meaning is tied to dapagliflozin and to a q24h steady-state AUC convention. Member of the AUC_<DRUG> family (AUC_CARBO, AUC_GEM, AUC_GCV, AUC_PAZO, AUC_RTV, AUC_EMPA, AUC_HTBZ, AUC_LEN); AUC_EMPA is the closest structural analogue – likewise a per-subject steady-state SGLT2-inhibitor AUC carried out of an upstream popPK into a downstream type 1 diabetes exposure-response model. Distinct from CINH, the time-varying plasma SGLT-inhibitor concentration regressor used in renal-glucose-reabsorption QSP models, and from CONMED_SGLT2I, the binary concomitant-SGLT2-inhibitor indicator. A future dapagliflozin model driven by a different exposure metric (trough concentration, Cmax, dose-normalised AUC) should register a parallel canonical rather than overload AUC_DAPA.

AUC_HTBZ (canonical for daily AUC of the valbenazine metabolite [+]-alpha-HTBZ at steady state)

• Description: Area under the plasma concentration-time curve of [+]-alpha-dihydrotetrabenazine ([+]-alpha-HTBZ, NBI-98782) over the once-daily 24 h dosing interval at steady state. [+]-alpha-HTBZ is the sole active metabolite of the prodrug valbenazine and carries essentially all of the VMAT2-inhibitory activity, so it – not the parent – is the exposure driver of valbenazine exposure-response models. Supplied per observation record as a data column; 0 for placebo-arm records, which makes any Emax term in it vanish exactly.
• Units: ng*h/mL. Must be in the same units as the model’s EC50 so the Emax term is dimensionless. Document per-model via covariateData[[AUC_HTBZ]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – enters via the direct Emax form emax * AUC_HTBZ / (ec50 + AUC_HTBZ). Reference values observed: EC50 = exp(7.50) = 1808 ngh/mL (Nguyen 2025 Table 3 theta 6, printed back-transformed as 1820). Nguyen 2025 Table 2 gives the Huntington’s-disease cohort median (5th; 95th percentile) values as 176 (107; 328), 402 (244; 749), 657 (399; 1225) and 931 (566; 1736) ngh/mL at valbenazine 20, 40, 60 and 80 mg once daily; the E-R simulations explored 0-3000 ng*h/mL.
• Source aliases:
  • AUC[+]-alpha-HTBZ – the printed variable name in the Nguyen 2025 Emax equation (Methods, “Exposure-Efficacy Modeling”).
  • AUC 24,ss – the Nguyen 2025 Table 2 column heading for the same quantity.
• Example models: Nguyen_2025_valbenazine_tmc.R (drives the Emax drug effect on the change from baseline in the UHDRS Total Maximal Chorea score in Huntington’s-disease chorea; emax = -14.2 TMC points with lec50 = 7.50).
• Notes: Specific scope because the column meaning is tied to this metabolite and to the q24h steady-state AUC convention. Member of the AUC_<DRUG> family (AUC_CARBO, AUC_GEM, AUC_GCV, AUC_PAZO, AUC_RTV, AUC_VERUB, AUC_ADU, AUC_DON, AUC_GAN, AUC_LEC, AUC_IBRU, AUC_LCM, AUC_CBZ, AUC_AMPH, AUC_LEN, AUC_EMPA); AUC_LEN is the closest structural analogue, likewise a per-interval steady-state AUC carried out of an upstream population PK run into a downstream PD run as a data column. Named for the METABOLITE rather than for the administered drug (contrast AUC_LEN, AUC_EMPA) because the parent valbenazine exposure is explicitly NOT the driver – Nguyen 2025 selected the metabolite AUC, and a future valbenazine model driven by parent exposure would need a separate AUC_VBZ sibling. Companion to the htbz metabolite suffix registered in compartment-names.md; the upstream model that generates this column is modellib('Nguyen_2025_valbenazine'). A future model using a different [+]-alpha-HTBZ exposure metric (Cmax, Cmin, Cavg – all three are tabulated in Nguyen 2025 Table 2 and gave consistent E-R profiles) should register a parallel canonical rather than overload AUC_HTBZ.

CONMED_CYP3A4_INH_STRONG (canonical for concomitant strong CYP3A4 inhibitor coadministration indicator)

• Description: 1 = subject / dose-record with concomitant coadministration of a strong CYP3A4 inhibitor (FDA/EMA classification of “strong”: AUC increase >= 5-fold for a sensitive CYP3A4 substrate; representative agents include itraconazole, ketoconazole, posaconazole, voriconazole, clarithromycin, and ritonavir-boosted regimens), 0 = no strong CYP3A4 inhibitor coadministration during the observation window (whether no CYP3A4 inhibitor at all, or a weak/moderate inhibitor only). Time-varying per record. Companion canonical to the broader CONMED_CYP3A4_INH that stratifies specifically by FDA/EMA inhibitor-strength tier – distinct from CONMED_CYP3A4_INH_HI (which stratifies by cumulative exposure duration rather than by inhibitor-strength category).
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no strong CYP3A4 inhibitor coadministration; subject may still be on a weak or moderate CYP3A4 inhibitor).
• Source aliases:
  • CYP3A4 inhibitors – used in Courlet_2021_amlodipine.R (Courlet 2021 Table 2 ‘theta CYP3A4 inhibitors’; antiretroviral strong CYP3A4 inhibitors per the FDA list: ritonavir- or cobicistat-boosted darunavir, ritonavir-boosted atazanavir, cobicistat-boosted elvitegravir).
  • DDICYPSH – used in Mitra_2026_ziftomenib.R (Kura Oncology KOMET-001 + KO-MEN-003 NONMEM control-stream column; strong CYP3A4 inhibitor coadministration during a ziftomenib dose record; dominant driver in the R/R AML cohort is prophylactic antifungal azole use).
• Example models: Mitra_2026_ziftomenib.R (multiplicative effects: 0.459x on parent ziftomenib CL/F, 0.195x on KO-739 CL, 0.449x on KO-516 CL when CONMED_CYP3A4_INH_STRONG = 1; encoded as additive shifts on the log(CL) scale per the paper’s NONMEM PK block), Kemal_2026_nemtabrutinib.R (multiplicative effect on CL/F: (1 + -0.0119 * CONMED_CYP3A4_INH_STRONG), i.e. 1.2% lower CL/F under strong CYP3A4 inhibition; RSE 774%, 95% CI includes zero – retained in the full covariate model), Jiang_2021_ivosidenib.R (pooled with CONMED_CYP3A4_INH_MOD into ‘other strong or moderate CYP3A4 inhibitors’, 0.92-fold CL/F, excluding the three separately estimated azoles; see the CONMED_CYP3A4_INH_MOD example), Courlet_2021_amlodipine.R (linear multiplicative effect on amlodipine CL/F: (1 - 0.49 * CONMED_CYP3A4_INH_STRONG), a 96% higher AUC0-24 with boosted ARVs in people living with HIV; Courlet 2021 Table 2).
• Notes: Companion canonical to CONMED_CYP3A4_INH (which pools any inhibitor strength into a single 0/1 indicator) and to CONMED_CYP3A4_INH_HI / CONMED_CYP3A4_INH_LO (which stratify by cumulative exposure duration rather than by inhibitor-strength category). Future extractions that need the sibling moderate- and weak-strength indicators should register CONMED_CYP3A4_INH_MOD and CONMED_CYP3A4_INH_WEAK following the same pattern.

CONMED_PANCYP_INH (canonical for concomitant hypothetical pan-CYP inhibitor (CYP1A2 + CYP2C8 + CYP2C9 + CYP3A4) indicator)

• Description: 1 = the victim drug is coadministered with a perpetrator that potently inhibits CYP1A2, CYP2C8, CYP2C9 and CYP3A4 simultaneously; 0 = no CYP modulator coadministered. Registered for the worst-case-inhibition scenario a physiologically based drug-interaction analysis constructs when it needs an upper bound on victim exposure and no single marketed drug inhibits every contributing CYP. The perpetrator is therefore usually hypothetical rather than a named INN, which is why the indicator is named for the inhibited enzyme set rather than for a drug.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no CYP modulator coadministered). Mutually exclusive with the per-perpetrator CONMED_<INN> flags and with the strength-tiered CONMED_CYP3A4_INH* family in any single record.
• Source aliases:
  • a hypothetical compound as perpetrator with potent inhibition against CYP1A2, CYP2C8, CYP2C9, CYP3A4 (Ki = 0.015 uM) – Ezuruike 2018 Table 4 [Table 3 in print] footnote b, the “Complete CYP inhibition” column.
• Example models: Ezuruike_2018_ethinylestradiol.R (the column drives BOTH the relative CYP3A4 activity term and a separate relative activity for CYP1A2 + CYP2C8 + CYP2C9; because the source gives the hypothetical inhibitor the same Ki of 0.015 umol/L it gives ketoconazole against CYP3A4, the model reuses ketoconazole’s back-solved activity of 0.222 for all four enzymes, making this arm an out-of-sample prediction that lands within 1.9% of the three published AUC cells).
• Notes: Deliberately distinct from three neighbours. (1) CONMED_CYP3A4_INH and its strength-tiered siblings (_STRONG, _MOD, _HI, _LO) record inhibition of CYP3A4 only, so a victim with a substantial non-CYP3A4 CYP fraction cannot express its worst case through them. (2) CONMED_CYP3A4_PGP_INH pools two mechanisms (one enzyme, one transporter) in a union indicator; this column pools four enzymes of the same mechanism class. (3) A named CONMED_<INN> flag should be preferred whenever the perpetrator is a real drug – reach for this column only when the source’s perpetrator is explicitly hypothetical or is described purely by its inhibited enzyme set. The per-model covariateData entry must record which enzymes are inhibited and with what potency, because “pan-CYP” is not self-defining: a future paper inhibiting a different CYP set (adding CYP2D6, say) should state the difference in its own entry rather than silently reuse this one with a different meaning.

DOSE_PTM_MG (canonical for administered pretomanid per-administration dose amount)

• Description: Administered oral dose of the nitroimidazooxazine antituberculosis agent pretomanid (formerly PA-824), in mg, for the current dose record. Per-dose-record covariate; must equal the amt of the corresponding dosing record. Not a PK covariate in the usual sense – the amount already appears on the dose record via amt – but it is required as an explicit regressor because pretomanid bioavailability is saturable in dose, so the dose amount has to be readable inside model to compute f(depot).
• Units: mg
• Type: continuous
• Scope: specific
• Reference category: n/a – enters via the saturable bioavailability f(depot) = Fmax / (1 + DOSE_PTM_MG / ED50) with Fmax assumed 1 and ED50 = 554 mg in humans (Mehta 2023 Table 1). A 200 mg dose therefore gives F = 1 / (1 + 200/554) = 0.735.
• Source aliases:
  • dose – used in Mehta_2023_pretomanid_mpbpk.R (Mehta 2023 ESM S2, which writes doseIn = Fmax*dose/(1 + dose/ED50) followed by f(depot) = doseIn/dose; the packaged model uses the algebraically identical single-expression form, which avoids dividing by the dose amount).
• Example models: Mehta_2023_pretomanid_mpbpk.R (drives the dose-dependent bioavailability of the translational lung-lesion mPBPK model; simulated regimen 200 mg once daily, clinical validation data spanning 50-1200 mg).
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family (siblings: DOSE_EMPA_MGD, DOSE_CIPARGAMIN_MG, DOSE_PHT_MGKGD, DOSE_RIVAROXABAN_MG). Units suffix is MG because pretomanid is dosed as a per-administration amount rather than a daily total.

DOSE_RIVAROXABAN_MG (canonical for administered rivaroxaban per-administration dose amount)

• Description: Administered oral dose of the direct factor Xa inhibitor rivaroxaban, in mg, for the current dose record. Per-dose-record covariate; must equal the amt of the corresponding dosing record. As with the other members of this family the amount is already carried by amt, and the column exists only so that a dose-dependent bioavailability can be evaluated inside model().
• Units: mg
• Type: continuous
• Scope: specific
• Reference category: n/a – enters via the dose-dependent relative bioavailability f(depot) = Fmin + (Fmax - Fmin) * exp(-ln(2)/D50 * DOSE_RIVAROXABAN_MG) with Fmin = 0.590, Fmax fixed at 1.25 and D50 = 14.4 mg (Willmann 2018a Table 3 and Supplementary Eq. 1). The function is anchored so that the 10 mg reference dose returns F = 0.998, i.e. 1 – the paper fixes relative bioavailability at 1.0 for 10 mg (Table 3 footnote c). Representative values: F = 1.175 at 2.5 mg, 0.998 at 10 mg, 0.842 at 20 mg, 0.746 at 30 mg.
• Source aliases:
  • DOSE – the NONMEM control-stream column name in Willmann_2018a_rivaroxaban.R (Willmann 2018a Supplementary Data S9, run12 $INPUT), which is a separate column from AMT in that dataset and is used only in the F1 expression.
• Example models: Willmann_2018a_rivaroxaban.R (drives the dose-dependent relative bioavailability of the integrated multi-indication adult popPK model; pooled doses span 2.5 mg once daily to 30 mg twice daily).
• Notes: Follows the DOSE_<DRUG>_<UNITS> auto-approve family (siblings: DOSE_PTM_MG, DOSE_EMPA_MGD, DOSE_CIPARGAMIN_MG). Units suffix is MG, not MGD, because the value is one administration and not the daily total – this distinction is load-bearing here, since much of the pooled dataset is twice-daily and the mechanism being modelled (solubility-limited absorption of a BCS class 2 compound) saturates per administration. A 15 mg b.i.d. arm therefore takes DOSE_RIVAROXABAN_MG = 15, not 30. Contrast the sibling DOSE_PTM_MG, whose saturable-F form is hyperbolic (Fmax / (1 + DOSE/ED50)) rather than the exponential decay to a non-zero Fmin asymptote used here; the two are not interchangeable functional forms even though both consume a DOSE_<DRUG>_MG column.

INSDOSE_BL (canonical for baseline total daily insulin dose per body weight)

• Description: Per-subject (time-fixed) total daily exogenous insulin dose at study baseline, normalised to body weight. Captures the insulin requirement of patients on background insulin therapy (a typical input covariate in T1DM / T2DM popPK/PD models where the empagliflozin / SGLT-2 inhibitor / GLP-1 analogue effect is layered on top of pre-existing insulin treatment). Distinct from INS_BL (baseline fasting plasma insulin concentration, a measured biomarker in pmol/L) – INSDOSE_BL is the administered dose level, not the resulting plasma level.
• Units: U/kg/day (international units of insulin per kilogram body weight per day). Document per-model via covariateData[[INSDOSE_BL]]$units if a paper reports total daily units without body-weight normalisation.
• Type: continuous
• Scope: specific
• Reference category: n/a – enters via power form (INSDOSE_BL / ref)^e_insdose_bl_<param>. Reference values observed: 0.660 U/kg/day (Johnston 2019 reference-patient description).
• Source aliases:
  • IDB – “Insulin daily dose at Baseline” abbreviation used in Johnston 2019 Table 2.
• Example models: Johnston_2019_empagliflozin.R (M-EASE-2 covariate effect on the typical baseline HbA1c and on Emax in T1DM patients with background insulin therapy; effect exponents 0.0141 on baseline HbA1c and 0.0552 on Emax, both with CIs that cross zero – the covariate was retained as part of the full-random-effects covariate model rather than dropped), Johnston_2021_empagliflozin_popPK.R (Johnston 2021 empagliflozin popPK covariate on CL/F in adults with T1D; effect exponent 0.0469 with CI that spans zero – retained as part of the full-covariate model).
• Notes: Specific scope because the column meaning is tied to baseline total-daily-dose accounting in insulin-treated diabetes populations; future T1DM / T2DM popPK/PD extractions can reuse this canonical and document the per-model units / dose-counting convention in covariateData[[INSDOSE_BL]]$notes. A model that uses time-varying insulin dose (rather than baseline only) should register a parallel canonical (e.g., INSDOSE). Companion concept to INS_BL (plasma insulin concentration) and to HBA1C (glycemic control).

INSDT_CSII (canonical for binary CSII vs MDI insulin delivery type indicator)

• Description: Binary indicator for the subject’s insulin delivery type at study entry. 1 = continuous subcutaneous insulin infusion (CSII; insulin pump therapy); 0 = multiple daily injections (MDI; subcutaneous injection regimen) reference. Per-subject and time-fixed within the analysis window (regimen switches during the study are not represented).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 = MDI (multiple daily injections).
• Source aliases:
  • INSDT – “Insulin Dose Type” abbreviation used in Johnston 2019 Table 2 (the two-level categorical column decomposed into the binary CSII indicator with MDI as the reference category).
• Example models: Johnston_2019_empagliflozin.R (M-EASE-2 covariate multiplier on baseline HbA1c, Emax, and placebo rate; CSII shifts placebo drift upward by a factor of 1.47 vs MDI – a moderate effect on the placebo-adjusted HbA1c change), Johnston_2021_empagliflozin_MEASE1.R (M-EASE-1 covariate multiplier on Emax_MDG in T1D; the reported CSII / MDI effect is very close to 1 (0.995) and is documented as retained-but-negligible in the full-covariate model).
• Notes: Specific scope because the CSII-vs-MDI contrast and reference category are study-specific. Follows the <COLUMN>_<LEVEL> decomposition pattern used for REGI_BID / TUMTP_GLIO. A future paper that retains the same INSDT covariate but with more than two levels (e.g., CSII vs MDI vs continuous glucose monitoring + open-loop algorithm) would register sibling canonicals (e.g., INSDT_CGM) rather than overload this entry.

LSCI (canonical for amplitude of an inverse-Bateman lifestyle-change / placebo effect on energy intake in body-composition modelling)

• Description: Per-subject (or per-study-arm) amplitude of the inverse-Bateman lifestyle-change (LSC) effect on energy intake (EI) in body-composition models. Drives the magnitude of the placebo / dietary-restriction-induced fractional reduction in EI; combined with onset rate Kdiet and reduction rate Kred to form the LSC effect LSCeff(t) = 1 - LSCI * Kdiet * (exp(-Kdiet * t) - exp(-Kred * t)) / (Kred - Kdiet) applied multiplicatively to EI. Subjects in the same study arm typically share the same LSCI value; the column is time-fixed at study entry (a single LSCI per subject per arm).
• Units: fraction (0..1; can be negative for arms that report a paradoxical increase in EI)
• Type: continuous
• Scope: specific
• Reference category: none – LSCI = 0 corresponds to no lifestyle effect (the model collapses to pure body composition + drug effects).
• Source aliases:
  • LSCI – used directly in Bosch_2024_glp1ra_bodyweight.R (Bosch 2024 supplement S10 THETA7..THETA16 are per-study fitted LSCI values).
• Example models: Bosch_2024_glp1ra_bodyweight.R (introduces the canonical; per-study fitted values 0..0.548 across diet and GLP-1RA studies, with -0.0288 reported for the Blundell 2017 arm).
• Notes: Specific scope because the column’s semantics are tied to body-composition / energy-balance modelling. Future extractions that combine the Hall body composition framework with other drug classes or other lifestyle-intervention protocols may reuse this canonical; document the per-paper values in covariateData[[LSCI]]$notes so users can match the source-paper LSCI to the simulated arm.

WM_IBT (canonical for weight-management + intensive-behavioural-treatment intervention indicator)

• Description: Binary indicator that the subject is enrolled in a study arm with weight management and intensive behavioural treatment (e.g., the STEP-trial-family lifestyle-intervention protocols described in Wadden 2021 / Garvey 2022 / Rubino 2022). Used in QSP body-composition models to gate a body-weight-dependent activity effect on the exercise component of physical activity energy expenditure. Subjects in arms without an explicit weight-management + IBT protocol carry WM_IBT = 0.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no weight management + intensive behavioural treatment).
• Source aliases:
  • IFLAG – used in Bosch_2024_glp1ra_bodyweight.R (Bosch 2024 supplement S10 IFLAG > 0 gates the activity effect).
• Example models: Bosch_2024_glp1ra_bodyweight.R (introduces the canonical; gates the body-weight-dependent activity effect on PAE for STEP-family arms in the Bosch 2024 dataset).
• Notes: Specific scope because the canonical name is tied to body-composition QSP modelling where the activity effect is empirically estimated for IBT-instrumented arms. Future extractions of similar lifestyle-intervention-aware body-composition or weight-loss-trajectory models may reuse this canonical; mark the per-paper IBT-intensity definition in covariateData[[WM_IBT]]$notes so users can distinguish “weekly clinic counselling + diary review” intensity (STEP 3) from “minimal-counselling -500 kcal/d” intensity (STEP 5 / STEP 8).

STRAIN_C57BI6 (canonical for C57BI/6 mouse-strain indicator)

• Description: Binary within-species mouse-strain indicator: 1 = subject is a C57BI/6 mouse, 0 = subject is a mouse of the reference strain (in the founding example, NMRI). Used to gate strain-specific fractional multipliers on structural PK parameters within a mouse popPK model when the source paper pools two strains and reports a fractional strain effect.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (NMRI mouse in Larsen 2018; other reference strains permitted for future extractions provided the per-paper reference is documented in covariateData[[STRAIN_C57BI6]]$notes).
• Source aliases: none.
• Example models: Larsen_2018_factorviia_mouse.R (Larsen 2018 pooled C57BI/6 + NMRI mice with fractional C57BI/6 effect on V and CL for rFVIIa: V_C57BI6 = V * 0.588, CL_C57BI6 = CL * 0.87 per Table 2).
• Notes: Analogous to the RACE_<GROUP> naming family for human race indicators: STRAIN_<GROUP> gives each within-species strain-of-interest its own binary indicator, with the reference strain implicitly encoded as 0. A future mouse popPK model that pools additional strains (e.g., BALB/c, C3H) should register a parallel canonical (STRAIN_BALBC, STRAIN_C3H,…) rather than overload this one.

STRAIN_NUDE (canonical for athymic-nude mouse-strain indicator)

• Description: Binary within-species mouse-strain indicator: 1 = subject is an Hsd:Athymic Nude-Foxn1nu mouse, 0 = subject is a severe-combined-immunodeficient (SCID, C.B-17/IcrHan(R)Hsd-Prkdcscid) mouse. Used when a preclinical popPK analysis pools the two immunocompromised strains routinely used for xenograft work and estimates strain-specific absorption, bioavailability or study-scaling parameters.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (SCID mouse in DeJongh 2025; other reference strains permitted for future extractions provided the per-paper reference is documented in covariateData[[STRAIN_NUDE]]$notes).
• Source aliases:
  • STR – used in DeJongh_2025_azd7648_mouse.R (AZD7648 NONMEM dataset; STR = 0 SCID, STR = 1 nude).
  • STRN – used in DeJongh_2025_olaparib_mouse.R (olaparib NONMEM dataset; STRN = 1 SCID, STRN = 2 nude, so the indicator is STRN - 1).
• Example models: DeJongh_2025_azd7648_mouse.R (gates both the absorption rate constant – estimated at 2.77 1/h in SCID, fixed at 9.9 1/h in nude mice for want of absorption-phase samples – and a -52% relative bioavailability in nude mice), DeJongh_2025_olaparib_mouse.R (splits the study S1734 relative-bioavailability factor by strain).
• Notes: Member of the STRAIN_<GROUP> family established by STRAIN_C57BI6, which explicitly invites parallel canonicals for additional strains. Note the two DeJongh 2025 source datasets encode the same biological covariate on different numeric scales (0/1 vs 1/2); the canonical column is always the 0/1 indicator, and the per-model source_name records which raw column it came from.

STRAIN_SHR (canonical for spontaneously-hypertensive-rat strain indicator)

• Description: Binary within-species rat-strain indicator: 1 = subject is a spontaneously hypertensive rat (SHR), 0 = subject is a normotensive Wistar-Kyoto (WKY) rat. The WKY strain is the generally accepted normotensive control strain for the SHR, so the two form a natural hypertensive / normotensive contrast within one species. Used when a preclinical analysis pools both strains and estimates strain-specific baseline values for the quantities the model turns over.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (normotensive WKY rat in Snelder 2014; other reference strains permitted for future extractions provided the per-paper reference is documented in covariateData[[STRAIN_SHR]]$notes).
• Source aliases:
  • Strain – the Table 1 study-overview column in Snelder 2014, whose levels are the literal strings SHR and WKY.
  • The _SHR / _WKY suffixes on the Snelder 2014 Table 5 baseline parameter rows (BSL_HR_SHR / BSL_HR_WKY, etc.), which is where the covariate is actually expressed in that paper.
• Example models: Snelder_2014_cardiovascular_rat.R (selects between the two Table 5 baseline triples BSL_HR / BSL_MAP / BSL_CO – 310 / 155 / 69.0 in SHR versus 323 / 102 / 129 in WKY rats – with every other system parameter shared across the strains).
• Notes: Member of the STRAIN_<GROUP> family established by STRAIN_C57BI6, which explicitly invites parallel canonicals for additional strains; this is the family’s first rat member, the two existing ones (STRAIN_C57BI6, STRAIN_NUDE) being mouse strains. The family is species-agnostic by construction – it is the within-species strain that is named, not the species – so no species qualifier is added to the name; a future model needing to distinguish, say, a mouse SHR-analogue would register its own member. The indicator gates baselines, not mechanism. In the founding example the strain difference in feedback strength that the paper reports (feedback about twofold higher in WKY than in SHR) is NOT a strain term: feedback is a power function of the individual baseline MAP, so the strain effect on feedback emerges from the lower WKY baseline rather than from this column. A model that used STRAIN_SHR to switch a mechanism directly would be doing something the founding example deliberately avoided, and should say so in its covariate notes. Distinct from DIS_HYPERT, which is a human medical-history comorbidity flag: STRAIN_SHR identifies a genetic line whose hypertension is one of several correlated phenotypes, and it is time-fixed by genotype rather than by clinical history.

STRAIN_SD (canonical for Sprague-Dawley rat strain indicator)

• Description: Binary within-species rat-strain indicator: 1 = subject is a Sprague-Dawley (SD) rat, 0 = subject belongs to the paper’s comparator strain. The comparator is a per-paper fact recorded in each model’s covariateData[["STRAIN_SD"]]$notes: the Wistar-Kyoto lineage (normotensive WKY, or spontaneously hypertensive SHR, which was bred from WKY) in Sang 2021; the inbred Lewis rat in White 2022.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (Wistar-Kyoto / SHR in Sang 2021; Lewis in White 2022).
• Source aliases:
  • Species – the regressor in the Sang 2021 deposited Mlxtran code, coded 1 for the Sprague-Dawley literature studies; the code comments the switched term as ‘Drug Effect on MC (Wistar)’.
  • Strain / ‘strain covariate’ – the Phoenix NLME covariate in White 2022 (Methods 2.8.1: ‘strain (Lewis = 0, SD = 1)’); same orientation, no value transformation.
• Example models: Sang_2021_doxorubicin_rat_qsp.R (multiplies the doxorubicin myocardial-compliance effect by 1 - STRAIN_SD, so SD rats show the systolic-dysfunction phenotype – falling LVEF, rising LVEDV – and WKY / SHR rats the diastolic-dysfunction phenotype with falling LVEDV), White_2022_alfaxalone_rat_pop1.R and White_2022_alfaxalone_rat_pop2.R (exponential strain effects on alfaxalone clearance (population 1 only) and volume of distribution, Lewis reference).
• Notes: Member of the STRAIN_<GROUP> family (STRAIN_C57BI6, STRAIN_NUDE, STRAIN_SHR). Like STRAIN_SHR, the 1-level always means the named strain while the 0-level comparator is set by the paper, so check the consuming model’s notes before pooling datasets across models. In Sang 2021 the strain and the dysfunction phenotype coincide study by study (studies 11-13 SD, study 14A-C WKY / SHR), and the paper describes the switch by phenotype; the column records the strain, which is the observable data item.

DOSE_AZD7648_MGKGD (canonical for concomitant AZD7648 total daily dose per kg body weight)

• Description: Total daily dose of co-administered AZD7648 (a DNA-dependent protein kinase inhibitor) in mg per kg body weight per day. Carried on every record so a drug-drug-interaction equation can read the co-medication dose level without back-computing it from the AZD7648 event records. It is the total across the dosing day, so a 75 mg/kg twice-daily regimen carries the value 150. Zero for olaparib monotherapy and for vehicle controls.
• Units: mg/kg/day
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as the driver of the Emax-shaped clearance-interaction term CmS = 1 - DOSE_AZD7648_MGKGD / (DOSE_AZD7648_MGKGD + CmD50) of DeJongh 2025 Equations 7-8, with CmD50 = 82.5 mg/kg/day. DOSE_AZD7648_MGKGD = 0 gives CmS = 1 (no interaction, olaparib clearance unchanged); DOSE_AZD7648_MGKGD = 82.5 halves olaparib clearance.
• Source aliases:
  • CMDDOS – used in DeJongh_2025_olaparib_mouse.R (olaparib population-PK NONMEM dataset, co-medication dose column).
  • AZD7648 / DOSE_AZD (mg/kg) – used in DeJongh_2025_azd7648_olaparib_xenograft_mouse.R (PK-PD dataset, supplementary file 7 of DeJongh 2025).
• Example models: DeJongh_2025_olaparib_mouse.R, DeJongh_2025_azd7648_olaparib_xenograft_mouse.R.
• Notes: Sibling canonical in the DOSE_<DRUG>_MGKGD family for daily co-medication doses, as anticipated by the DOSE_RTV_MGKG notes (“Drug-self-dose covariates for other drugs should register sibling canonicals (e.g., DOSE_<DRUG>_MGKG for per-administration or DOSE_<DRUG>_MGKGD for daily)”). Specific scope because the CmD50 value is tied to the DeJongh 2025 mouse analysis, and because only one source study (S1734) contained olaparib arms with and without AZD7648 co-treatment.

STUDY_S1143 (canonical for DeJongh 2025 olaparib mouse study S1143 cohort indicator)

• Description: 1 = animal enrolled in mouse study S1143 (2, 5, 10, 50 or 100 mg/kg olaparib as a single oral dose in nude mice); 0 = the reference study S11448. Selects the study-specific relative oral bioavailability of olaparib.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (study S11448, the 20 mg/kg intravenous / 50 mg/kg oral study in nude mice that anchors absolute bioavailability for the whole pooled analysis).
• Source aliases: derived from the STUDY column of the olaparib NONMEM dataset (STUDY == 1143 -> 1).
• Example models: DeJongh_2025_olaparib_mouse.R (selects lfdepot_s1143, F1 = 0.174).
• Notes: DeJongh 2025 Results: “Relative oral bioavailability was found to vary substantially between some studies, even within the same mice strain, and was accounted for in the final PK model by fitting a descriptive inter-study scaling factor on this parameter.” One binary per non-reference study, following the decomposed-binary-indicator convention rather than a single integer study code.

STUDY_S1721 (canonical for DeJongh 2025 olaparib mouse study S1721 cohort indicator)

• Description: 1 = animal enrolled in mouse study S1721 (50, 75 or 100 mg/kg olaparib as single and multiple oral doses in nude mice); 0 = the reference study S11448. Selects the study-specific relative oral bioavailability of olaparib.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (study S11448).
• Source aliases: derived from the STUDY column of the olaparib NONMEM dataset (STUDY == 1721 -> 1).
• Example models: DeJongh_2025_olaparib_mouse.R (selects lfdepot_s1721, F1 = 0.489).
• Notes: See STUDY_S1143.

STUDY_S1734 (canonical for DeJongh 2025 olaparib mouse study S1734 cohort indicator)

• Description: 1 = animal enrolled in mouse study S1734 (100 mg/kg olaparib as single and multiple oral doses in SCID and nude mice, with and without AZD7648 co-treatment); 0 = the reference study S11448. Selects the study-specific relative oral bioavailability of olaparib, which S1734 estimates separately per strain – so this indicator is combined with STRAIN_NUDE.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (study S11448).
• Source aliases: derived from the STUDY column of the olaparib NONMEM dataset (STUDY == 1734 -> 1).
• Example models: DeJongh_2025_olaparib_mouse.R (selects lfdepot_s1734scid, F1 = 0.598, or lfdepot_s1734nude, F1 = 0.238).
• Notes: S1734 is the only source study containing olaparib arms both with and without AZD7648 co-treatment, so it is the sole study informing the CmD50 drug-drug-interaction parameter; the source control stream applies the interaction term conditionally on this study. The library model applies the interaction unconditionally, which is equivalent because DOSE_AZD7648_MGKGD is 0 outside the co-treated arms.

STUDY_S1770 (canonical for DeJongh 2025 olaparib mouse study S1770 cohort indicator)

• Description: 1 = animal enrolled in mouse study S1770 (multiple oral olaparib doses with and without AZD7648 co-treatment in SCID mice); 0 = the reference study S11448. Selects the study-specific relative oral bioavailability of olaparib.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (study S11448).
• Source aliases: derived from the STUDY column of the olaparib NONMEM dataset (STUDY == 1770 -> 1).
• Example models: DeJongh_2025_olaparib_mouse.R (selects lfdepot_s1770, F1 = 1.48).
• Notes: See STUDY_S1143. Values above 1 are possible because the source parameterises olaparib clearance in absolute L/h while the volumes are weight-normalised L/kg; the per-study bioavailability factor absorbs that scale mismatch as well as genuine formulation / exposure differences.

STUDY_S1816 (canonical for DeJongh 2025 olaparib mouse study S1816 cohort indicator)

• Description: 1 = animal enrolled in mouse study S1816 (multiple oral olaparib doses with AZD7648 co-treatment in SCID mice); 0 = the reference study S11448. Selects the study-specific relative oral bioavailability of olaparib.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (study S11448).
• Source aliases: derived from the STUDY column of the olaparib NONMEM dataset (STUDY == 1816 -> 1).
• Example models: DeJongh_2025_olaparib_mouse.R (selects lfdepot_s1816, F1 = 4.18).
• Notes: This is the factor the authors carried into the DeJongh 2025 tumour PK-PD model (DeJongh_2025_azd7648_olaparib_xenograft_mouse.R fixes lfdepot_olaparib at 4.1817), because the xenograft studies had no olaparib PK of their own. See STUDY_S1770 on why the value exceeds 1.

IP_FA (canonical for tablet-transit inflection-point time from fundus to antrum)

• Description: Individual inflection-point time (h) at which the sigmoid step function governing tablet movement from the fundus to the antrum equals 0.5 (paper Equation 1 form: STEP(t) = 1 / (1 + exp(-SIG * (t - IP)))). Used in the Gastro-Intestinal Transit Time (GITT) absorption model of Henin 2012 to drive per-subject tablet residence time in the fundus. The paper samples IP per subject from a fixed log-normal distribution IP = MRT * exp(eta) with eta ~ N(0, VRT) and MRT / VRT taken from the upstream Bergstrand 2009 Markov-chain fit (Table II of Henin 2012): MRT_fundus = 0.4 h (fasted) / 1.04 h (fed), VRT_fundus = 0.46 / 1.09 h^2 (CV 100%).
• Units: h
• Type: continuous
• Scope: specific
• Reference category: NULL – IP_FA is a continuous per-subject residence time; the sigmoid step is centred at IP_FA regardless of whether the subject is fed or fasted, but the sampling distribution is fed / fasted stratified.
• Source aliases:
  • IP_F_A – used directly in the Henin 2012 model file annotations for the fundus-to-antrum inflection point.
• Example models: Henin_2012_felodipine.R (founding example; used in the felodipine GITT extraction for the extended-release tablet transit).
• Notes: Specific scope because the canonical name is tied to the GITT / MMM-derived semi-mechanistic absorption modelling framework where the tablet position is modelled by sigmoid STEP functions with per-subject inflection points. Only the “no return to fundus” subpopulation is encoded in the Henin 2012 extraction (7/12 felodipine subjects per the paper); the paper’s 3-component mixture with 0 / 1 / 2 antrum-to-fundus returns is documented as a deviation in the vignette. Future GITT-family extractions (Bergstrand 2009 upstream, or 2020+ mechanistic-absorption papers extending the same STEP-function idiom) should reuse this canonical.

IP_APSI (canonical for tablet-transit inflection-point time from antrum to proximal small intestine)

• Description: Individual inflection-point time (h) at which the sigmoid step function governing tablet movement from the antrum (or the enteric-coated-tablet whole-stomach exit) to the proximal small intestine equals 0.5. Central variable in the Gastro-Intestinal Transit Time (GITT) model (Henin 2012 Equation 1). For the extended-release felodipine model the population distribution is fed / fasted stratified: MRT_antrum = 0.32 h (fasted) / 1.58 h (fed), VRT_antrum = 0.15 / 2.50 h^2 (CV 100%). For enteric-coated diclofenac the same covariate represents the combined stomach transit (mean ~ 2 h ranging 1.5-3 h per paper Results).
• Units: h
• Type: continuous
• Scope: specific
• Reference category: NULL – continuous per-subject residence time.
• Source aliases:
  • IP_A_PSI – used in the Henin 2012 model file annotations for the antrum-to-proximal-small-intestine inflection point.
• Example models: Henin_2012_felodipine.R (extended-release with drug release in fundus + antrum + PSI + DSI + colon), Henin_2012_diclofenac.R (enteric-coated with stomach transit only, no drug release in stomach; IP_APSI represents the gastric emptying event).
• Notes: Specific scope. IP_APSI’s exact operational meaning differs slightly between the two Henin 2012 models: felodipine’s IP_APSI is the antrum -> PSI transition (fundus and antrum are separate compartments), while diclofenac’s IP_APSI is a lumped stomach -> PSI transition (fundus and antrum are not distinguished because the enteric coating prevents any drug release in the stomach). Both use the same STEP-function machinery and the same “no return to fundus” subpopulation simplification.

IP_PSI_DSI (canonical for tablet-transit inflection-point time from proximal to distal small intestine)

• Description: Individual inflection-point time (h) at which the sigmoid step function governing tablet movement from the proximal small intestine to the distal small intestine equals 0.5. Not fed / fasted stratified in the Henin 2012 GITT framework. Sampled per subject as IP_PSI_DSI = MRT_psi * exp(eta) with MRT_psi = 1.17 h, VRT_psi = 1.37 h^2 (CV 50%) per Table II.
• Units: h
• Type: continuous
• Scope: specific
• Reference category: NULL – continuous per-subject residence time.
• Source aliases: none.
• Example models: Henin_2012_felodipine.R, Henin_2012_diclofenac.R.
• Notes: Specific scope. Time-zero-referenced: IP_PSI_DSI is the absolute clock time (relative to dose administration) at which the sigmoid switch crosses 0.5. The effective PSI residence time is IP_PSI_DSI - IP_APSI.

IP_DSI_C (canonical for tablet-transit inflection-point time from distal small intestine to colon)

• Description: Individual inflection-point time (h) at which the sigmoid step function governing tablet movement from the distal small intestine to the colon equals 0.5. Not fed / fasted stratified in the Henin 2012 GITT framework. Sampled per subject as IP_DSI_C = MRT_dsi * exp(eta) with MRT_dsi = 1.22 h, VRT_dsi = 1.48 h^2 (CV 58%) per Table II.
• Units: h
• Type: continuous
• Scope: specific
• Reference category: NULL – continuous per-subject residence time.
• Source aliases: none.
• Example models: Henin_2012_felodipine.R, Henin_2012_diclofenac.R.
• Notes: Specific scope. Time-zero-referenced; effective DSI residence time is IP_DSI_C - IP_PSI_DSI.

CONMED_LAMOTRIGINE (canonical for concomitant lamotrigine coadministration indicator)

• Description: 1 = subject is coadministered lamotrigine (phenyltriazine antiepileptic and mood stabiliser, cleared predominantly by UGT1A4 glucuronidation) at the pharmacokinetic observation, 0 = no concomitant lamotrigine. Time-varying when comedication start / stop events are captured; time-fixed at the analysis baseline in sources that record comedication only as a per-record flag.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant lamotrigine).
• Source aliases: none.
• Example models: Zhang_2024_sertraline.R (screened on sertraline CL/F and V/F via the paper’s categorical-covariate form and not retained; declared in covariatesDataExcluded to preserve the screen – founding example).
• Notes: Named-drug member of the CONMED_<INN> family. Distinct from the class-level CONMED_AED (any antiepileptic) and CONMED_EIAED (enzyme-INDUCING antiepileptic): lamotrigine is not a clinically meaningful CYP inducer, so a model that pools it into CONMED_EIAED would be mis-specified. Also distinct from CONMED_UGT_INH, which is the pooled UGT-inhibitor indicator – lamotrigine is a UGT substrate rather than a UGT inhibitor. Use the named indicator when the source paper screens lamotrigine specifically, and the class indicator when the source pools antiepileptics.

CONMED_QUETIAPINE (canonical for concomitant quetiapine coadministration indicator)

• Description: 1 = subject is coadministered quetiapine (dibenzothiazepine second-generation antipsychotic, cleared predominantly by CYP3A4) at the pharmacokinetic observation, 0 = no concomitant quetiapine. Time-varying when comedication start / stop events are captured; time-fixed at the analysis baseline in sources that record comedication only as a per-record flag.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant quetiapine).
• Source aliases: none.
• Example models: Zhang_2024_sertraline.R (screened on sertraline CL/F and V/F via the paper’s categorical-covariate form and not retained; declared in covariatesDataExcluded to preserve the screen – founding example).
• Notes: Named-drug member of the CONMED_<INN> family. Quetiapine is a CYP3A4 substrate and a weak CYP2D6 inhibitor; it is neither a strong CYP3A4 inhibitor nor an inducer, so it should not be folded into CONMED_CYP3A4_INH / CONMED_CYP3A4_IND. Register sibling named canonicals (CONMED_OLANZAPINE, CONMED_RISPERIDONE,…) rather than reusing this one when a source screens a different antipsychotic, and reserve a future class-level CONMED_ANTIPSYCHOTIC for sources that pool the class.

CONMED_VENLAFAXINE (canonical for concomitant venlafaxine coadministration indicator)

• Description: 1 = subject is coadministered venlafaxine (serotonin-norepinephrine reuptake inhibitor, cleared predominantly by CYP2D6 to O-desmethylvenlafaxine with a CYP3A4 minor route) at the pharmacokinetic observation, 0 = no concomitant venlafaxine. Time-varying when comedication start / stop events are captured; time-fixed at the analysis baseline in sources that record comedication only as a per-record flag.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant venlafaxine).
• Source aliases: none.
• Example models: Zhang_2024_sertraline.R (screened on sertraline CL/F and V/F via the paper’s categorical-covariate form and not retained; declared in covariatesDataExcluded to preserve the screen – founding example).
• Notes: Named-drug member of the CONMED_<INN> family. Relevant to psychotropic popPK models both as a coadministered antidepressant and as a CYP2D6 substrate that can compete with other CYP2D6-cleared comedications. Register sibling named canonicals for other antidepressants screened by name rather than reusing this one; reserve a future class-level CONMED_ANTIDEPRESSANT for sources that pool the class.

STUDY_VANCO_CENTER2 (canonical for Shen 2024 vancomycin second-study-center indicator)

• Description: 1 = subject enrolled at the second of the two study centers pooled in the Shen 2024 Southern Chinese pediatric vancomycin analysis (Table 1 column N2, n = 176 patients / 249 concentrations); 0 = subject enrolled at the first center (Table 1 column N1, n = 210 patients / 272 concentrations). Selects between the two additive residual-error magnitudes the paper estimated per center.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (center N1).
• Source aliases: derived per subject from the recruiting-hospital identifier. Shen 2024 does not state which of Baoan Women’s and Children’s Hospital / Shenzhen Children’s Hospital is N1 and which is N2, so the indicator is anchored to the Table 1 column label rather than to a hospital name.
• Example models: Shen_2024_vancomycin.R (switches the additive residual SD between addSdCenter1 = 4.64 mg/L and addSdCenter2 = 4.53 mg/L, per Shen 2024 Table 3 rows RV1 / RV2 and Results “RV was best characterized by an additive error model for each of the two study centers”).
• Notes: Specific scope; tied to the two-hospital Shen 2024 vancomycin cohort. Member of the auto-approved STUDY_<id> family, and the first member keyed on a recruiting center rather than on a trial phase or protocol number – the STUDY_<DRUG>_PHASE<N> naming of STUDY_POSA_PHASE3 / STUDY_NIPOCALIMAB_PHASE1 / STUDY_SULDUR_PHASE2 does not apply because both Shen 2024 centers contributed the same kind of routine-TDM observations under one protocol. Subject-level (time-fixed). Distinct from the model’s age strata: center N1 has median age 0.95 y and N2 median 4.32 y, but both centers contribute patients on both sides of the 2-year clearance cutoff, so this indicator must not be conflated with the AGE <= 2 stratum switch. A future multi-center analysis of a different drug should register its own STUDY_<drug>_CENTER<N> sibling rather than reuse this entry, because the center numbering is source-table-specific.

STUDY_HUASHAN (canonical for Chen 2021 tacrolimus Huashan Hospital cohort indicator)

• Description: 1 = subject from the Huashan Hospital (Shanghai) cohort pooled in the Chen 2021 adult kidney transplant tacrolimus analysis (52 patients, 620 trough concentrations, May 2009-December 2013, EMIT assay converted to CMIA equivalents); 0 = subject from the Changhai Hospital (Shanghai) cohort (90 patients, 758 troughs, 2016, CMIA assay). Selects between the two centre-specific exponential residual errors.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (Changhai Hospital).
• Source aliases: CH / HS – Chen 2021 Table 2 rows ‘CH exponential error’ and ‘HS exponential error’.
• Example models: Chen_2021b_tacrolimus.R (switches the log-scale residual SD between expSd_changhai = sqrt(0.0606) and expSd_huashan = sqrt(0.0887)).
• Notes: Member of the auto-approved STUDY_<id> family, keyed on a recruiting centre like [[STUDY_VANCO_CENTER2]]; the paper names the centres, so the indicator is anchored to the hospital name. Subject-level (time-fixed). It carries no structural effect, but in the source data it is confounded with Wuzhi capsule use (no Huashan patient received it) and with the assay (EMIT converted to CMIA), so it must not be read as a proxy for either.

STUDY_SPD555303 (canonical for van Schaick 2016 prucalopride SPD555-303 phase 3 study cohort indicator)

• Description: 1 = subject enrolled in SPD555-303, the phase 3 multiple-dose trial of prucalopride in children with functional constipation (NCT01330381; 137 patients in the PK dataset, sparse sampling of one post-first-dose and two steady-state concentrations); 0 = subject enrolled in PRU-USA-12, the phase 1 richly sampled single-dose study (38 patients, mean 12.6 samples each). Selects among the study-specific typical clearance, the study-specific clearance IIV, and the study-specific log-scale residual-error magnitude that van Schaick 2016 estimated separately for the two pooled studies.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (PRU-USA-12 phase 1 single-dose study).
• Source aliases: derived per subject from the trial identifier (SPD555-303 -> 1, PRU-USA-12 -> 0).
• Example models: vanSchaick_2016_prucalopride_pediatric.R (switches CL between lcl_spd = log(20.1) and lcl_pru = log(22.9) L/h/70 kg, the CL IIV between omega^2 0.1191 and 0.0151, and the log-scale residual SD between expSdSPD = 0.35 and expSdPRU = 0.142, per van Schaick 2016 Table 3).
• Notes: Specific scope; tied to the two-study van Schaick 2016 pediatric prucalopride pooled analysis. Member of the auto-approved STUDY_<id> family, keyed on the protocol number rather than on a trial phase, because the paper labels its Table 3 parameter rows with the protocol identifiers (CL_PRU-USA-12, CL_SPD555-303) and not with phase. Unusual among the STUDY_* indicators in that it switches a structural parameter (typical CL) and not only variance terms: most siblings (STUDY_MD, STUDY_POSA_PHASE3, STUDY_VANCO_CENTER2, STUDY_NMV_PHASE23) select a residual-error magnitude alone. The paper’s justification is that “a slightly lower typical value for CL in SPD555-303 appeared to provide a better description of the trial data… only 12% lower than in PRU-USA-12, but in combination with a considerably larger interindividual variability in CL, it provided a better description of the observed prucalopride concentrations” (Discussion), and it attributes the larger SPD555-303 residual error to that study’s sparse sampling. Subject-level (time-fixed). Set STUDY_SPD555303 = 1 to reproduce the paper’s own dose-selection simulations, which the Results state were “performed with clearance based on SPD555-303”.

CONMED_MEROPENEM (canonical for concomitant meropenem coadministration indicator)

• Description: 1 = subject is coadministered meropenem (carbapenem beta-lactam, eliminated predominantly by renal filtration and tubular secretion with partial hydrolysis by dehydropeptidase-I) at the pharmacokinetic observation, 0 = no concomitant meropenem. Time-varying when comedication start / stop events are captured; time-fixed at the analysis baseline in sources that record comedication only as a per-record flag.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant meropenem).
• Source aliases: none.
• Example models: Shen_2024_vancomycin.R (screened on vancomycin CL because it was coadministered in more than 10% of patients – 43.76% of records – and not retained; declared in covariatesDataExcluded to preserve the screen – founding example).
• Notes: Named-drug member of the CONMED_<INN> family. Relevant to renally cleared antibacterial popPK models as a potential competitor for tubular secretion. Register sibling named canonicals for other antibacterials screened by name rather than reusing this one; use the class-level [[CONMED_CARBAPENEM]] for sources that pool the carbapenems, or CONMED_ABX for sources that pool all other antibacterials.

CONMED_CARBAPENEM (canonical for concomitant carbapenem-class coadministration indicator)

• Description: 1 = subject is coadministered any carbapenem beta-lactam (meropenem, ertapenem, imipenem/cilastatin, doripenem, biapenem) at the pharmacokinetic observation, 0 = no concomitant carbapenem. A class-level pooled indicator for sources that record carbapenem exposure as a single flag and estimate one shared effect for the class rather than a per-agent effect. Time-varying when comedication start / stop events are captured; time-fixed at the analysis baseline in sources that record comedication only as a per-record flag.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant carbapenem).
• Source aliases:
  • CBP – used in Zhang_2024_valproic_acid.R (Zhang 2024 Table 1 and the Abstract clearance equation; the paper defines CBP as meropenem or ertapenem pooled).
• Example models: Zhang_2024_valproic_acid.R (exponential effect on valproate CL/F: cl *= exp(1.50 * CONMED_CARBAPENEM), i.e. exp(1.50) = 4.48-fold higher apparent clearance – a 448.2% increase – in the 22 of 443 patients (4.9%) coadministered a carbapenem, of whom 18 received meropenem and 3 ertapenem; founding example).
• Notes: Class-level counterpart to the named-agent canonicals CONMED_MEROPENEM and CONMED_MER; the name was reserved in CONMED_MEROPENEM’s notes for exactly this “source pools the class” case. Use CONMED_CARBAPENEM only when the source paper itself estimates a single pooled class effect and does not resolve the individual agents; use the per-INN indicators when the paper names and estimates effects for specific agents. Distinct from CONMED_ABX (which pools all non-modelled antibacterials, not one mechanistically coherent class). The carbapenem-valproate interaction that motivates this canonical is mechanistically specific rather than a generic antibacterial confounder: carbapenems inhibit acylpeptide hydrolase, blocking the deglucuronidation of valproate glucuronide back to valproate, and the resulting fall in valproate exposure is large enough that pooling meropenem with ertapenem is pharmacologically defensible (Zhang 2024 Discussion; Suzuki 2016; Li 2021). A future source that reports agent-resolved carbapenem effects should populate the per-INN canonicals instead of, or alongside, this one.

CONMED_CEFOPERAZONE (canonical for concomitant cefoperazone coadministration indicator)

• Description: 1 = subject is coadministered cefoperazone (third-generation cephalosporin, eliminated predominantly by biliary excretion) at the pharmacokinetic observation, 0 = no concomitant cefoperazone. Time-varying when comedication start / stop events are captured; time-fixed at the analysis baseline in sources that record comedication only as a per-record flag.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant cefoperazone).
• Source aliases: none.
• Example models: Shen_2024_vancomycin.R (screened on vancomycin CL because it was coadministered in more than 10% of patients – 25.91% of records – and not retained; declared in covariatesDataExcluded to preserve the screen – founding example).
• Notes: Named-drug member of the CONMED_<INN> family. Cefoperazone is unusual among cephalosporins in being cleared mainly by the biliary rather than the renal route, so it should not be folded into a renal-competition class indicator. Register sibling named canonicals for other cephalosporins screened by name rather than reusing this one.

CONMED_CEFTRIAXONE (canonical for concomitant ceftriaxone coadministration indicator)

• Description: 1 = subject is coadministered ceftriaxone (third-generation cephalosporin, highly protein-bound, eliminated by both renal and biliary routes) at the pharmacokinetic observation, 0 = no concomitant ceftriaxone. Time-varying when comedication start / stop events are captured; time-fixed at the analysis baseline in sources that record comedication only as a per-record flag.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant ceftriaxone).
• Source aliases: none.
• Example models: Shen_2024_vancomycin.R (screened on vancomycin CL because it was coadministered in more than 10% of patients – 25.72% of records – and not retained; declared in covariatesDataExcluded to preserve the screen – founding example), Li_2021_vancomycin.R (retained: multiplies vancomycin CL by 1.46 in infants with septicemia, applied as 1.46^CONMED_CEFTRIAXONE; source column DC).
• Notes: Named-drug member of the CONMED_<INN> family. Ceftriaxone’s high albumin binding makes it a plausible displacement interactant as well as a coeliminated antibacterial, which is why sources screen it by name rather than pooling it into a class indicator.

CONMED_MANNITOL (canonical for concomitant mannitol coadministration indicator)

• Description: 1 = subject is coadministered mannitol (osmotic diuretic, freely filtered at the glomerulus and not reabsorbed) at the pharmacokinetic observation, 0 = no concomitant mannitol. Time-varying when comedication start / stop events are captured; time-fixed at the analysis baseline in sources that record comedication only as a per-record flag.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant mannitol).
• Source aliases: none.
• Example models: Shen_2024_vancomycin.R (screened on vancomycin CL because it was coadministered in more than 10% of patients – 12.67% of records – and not retained; declared in covariatesDataExcluded to preserve the screen – founding example), Wei_2022_vancomycin.R (retained: multiplies vancomycin CL by exp(0.13 * CONMED_MANNITOL), a 14% increase, in Chinese adults after neurosurgery; source column Mannitol).
• Notes: Named-drug member of the CONMED_<INN> family. Mechanistically distinct from the antibacterial CONMED_<INN> entries: mannitol raises urine flow and can transiently increase the renal clearance of filtered drugs, so it is screened in renally cleared popPK models as a renal-haemodynamic perturbation rather than as a metabolic interactant. Register CONMED_FUROSEMIDE and other named diuretics separately rather than reusing this entry; reserve a future class-level CONMED_DIURETIC for sources that pool the class.

CP_ABEMACICLIB_UM (canonical for instantaneous abemaciclib plasma concentration as a time-varying CNS-PBPK input function)

• Description: Instantaneous total abemaciclib concentration in systemic plasma, supplied directly as a time-varying covariate column rather than computed from a coupled PK model. Serves as the input function (“forcing function”) of a permeability-limited CNS PBPK model: it drives delivery into the brain-blood compartment through the cerebral-blood-flow term Qbrain * (CP_ABEMACICLIB_UM - Cbb) and is never depleted by CNS uptake, so systemic and CNS pharmacokinetics are one-way coupled. Values must be interpolated linearly between the supplied times.
• Units: umol/L (= uM; abemaciclib MW = 506.6 g/mol so 1 umol/L = 0.507 mg/L = 507 ng/mL).
• Type: continuous
• Scope: specific
• Reference category: n/a – enters linearly as the donor concentration of the cerebral blood flow term. Reference peak: the population-mean profile of Wickramasinghe 2025 Table S2, determined in glioblastoma patients on twice-daily oral abemaciclib, reaches 0.4191 umol/L (about 212 ng/mL) at 124.61 h with a median inter-peak interval of 11.4 h.
• Source aliases:
  • Plasma1 (Wickramasinghe 2025 Table S2 input-file template; the paired sampling times are the Time1 column).
  • Cart (Data S1 of the same paper, the arterial concentration symbol in the brain-blood differential equation).
• Example models: Wickramasinghe_2025_abemaciclib_cns_pbpk.R (founding example; drives all nine states of the 9-CNS spatial brain / brain-tumour PBPK model).
• Notes: Specific scope; abemaciclib-specific. Follows the established CP_<drug>_<units> precedent (CP_RIF_UM, CP_GDC_UM, CP_GLASDEGIB_NGML, CP_RITUXIMAB_UGML), using the full INN rather than an abbreviation. Distinct in role from the perpetrator-drug members of the family (CP_RIF_UM, CP_GDC_UM), which enter a competitive-inhibition denominator: this one is a physiological input function, so a downstream user who substitutes their own profile changes the model’s driving exposure rather than the strength of an interaction. The 9-CNS ODE system is linear and homogeneous in concentration, so any self-consistent concentration unit may be used provided the supplied plasma profile and the interpreted compartment outputs share it; umol/L is the unit implied by the abemaciclib values in the Wickramasinghe 2025 template, whose own figure axes are labelled only “Concentrations”.

CP_BMS911543_UM (canonical for the BMS-911543 concentration in an in-vitro enzyme incubation, as the driver of a Michaelis-Menten velocity or of a mechanism-based inactivation rate)

• Description: Concentration of BMS-911543, a selective small-molecule JAK2 inhibitor developed as Bristol-Myers Squibb development code BMS-911543 and never assigned an INN, supplied directly as a covariate column rather than computed from a coupled PK model. Founding uses are all in-vitro: as the substrate concentration S driving the Michaelis-Menten velocity of metabolite-M1 formation in three recombinant CYP incubations, and as the initial inhibitor concentration I driving the hyperbolic inactivation-rate term of a CYP1A2 time-dependent-inhibition assay. The column is defined generally, so a plasma or unbound site-of-interaction concentration can use the same name; the per-model covariateData entry states which sense is required.
• Units: umol/L (= uM).
• Type: continuous
• Scope: specific
• Reference category: n/a – enters either a Michaelis-Menten velocity vmax * CP_BMS911543_UM / (km + CP_BMS911543_UM) or the inactivation-rate term kinact * CP_BMS911543_UM / (ki + CP_BMS911543_UM). Set to 0 for the no-substrate or vehicle-control condition, at which both forms collapse to 0 and, in the inactivation model, the relative-activity state stays at 1 for all time. Reference values: 0.1-10 uM across the metabolism assay, and 0, 0.39, 0.78, 1.56, 3.125, 6.25, 12.5, 25.0 uM across the inactivation assay (Zhou 2015 Methods).
• Source aliases:
  • Substrate Conc. (Zhou 2015 Figure 1 x-axis label; values in uM) – the metabolism arm.
  • I (Zhou 2015 Eq. 2, “a specific initial inhibitor concentration”) and BMS-911543 Concentration (Figure 2b x-axis label) – the inactivation arm.
• Example models: Zhou_2015_bms911543_rcyp1a2.R, Zhou_2015_bms911543_rcyp3a4.R, Zhou_2015_bms911543_rcyp2j2.R (substrate concentration driving M1-formation velocity in recombinant CYP1A2, CYP3A4 and CYP2J2; km 2.1, 1.4 and 1.3 uM respectively), Zhou_2015_bms911543_cyp1a2_tdi.R (preincubation inhibitor concentration driving mechanism-based inactivation of CYP1A2; ki 2.9 uM, kinact 1.4 per h).
• Notes: Specific scope; BMS-911543-specific. Follows the established CP_<drug>_<units> family (CP_RIF_UM, CP_GDC_UM, CP_CARBAMAZEPINE_UM, CP_ABEMACICLIB_UM), and the same reuse-for-a-non-plasma-matrix pattern that CP_RIF_UM and CP_CARBAMAZEPINE_UM established for hepatocyte culture medium and CP_TUVUSERTIB_NGML for a patch-clamp bath. The name keeps the development code because no INN exists: the compound was discontinued after the phase 1/2a myelofibrosis study. The two senses in the founding paper are NOT interchangeable even though both are buffer concentrations in uM: the metabolism arm was run at 0.25 mg/mL microsomal protein and the inactivation arm at 1.0 mg/mL, so the fractions unbound in incubation differ and only the metabolism arm has a reported one (fumic 0.78, itself a Simcyp prediction rather than a measurement). The sibling model Zhou_2015_bms911543_hlm.R does NOT use this column: its incubation protein concentration is reported, so the substrate is carried as an ODE state and dosed rather than supplied as a covariate. A user driving the inactivation model from an in-vivo profile must supply the unbound concentration at the site of enzyme interaction and should note that Zhou 2015 itself found the in-vitro ki of 2.9 uM had to be raised to about 11 uM to reproduce the clinical accumulation.

CP_STA_UM (canonical for the schisantherin A concentration in an in-vitro CYP inhibition incubation)

• Description: Concentration of schisantherin A (STA), the most abundant lignan of the Wuzhi capsule (a standardised ethanolic extract of Schisandra sphenanthera fruit, co-prescribed with tacrolimus in Chinese transplant care), supplied directly as a covariate column rather than computed from a coupled PK model. Founding use is in-vitro: the inhibitor concentration I driving both the reversible-inhibition term 1/(1 + CP_STA_UM/ki) and the mechanism-based inactivation-rate term kinact * CP_STA_UM / (KI + CP_STA_UM) of a CYP3A4/CYP3A5 inhibition assay. The column is defined generally, so a plasma or unbound site-of-interaction concentration can use the same name; the per-model covariateData entry states which sense is required.
• Units: umol/L (= uM).
• Type: continuous
• Scope: specific
• Reference category: n/a – enters either a reversible-inhibition factor 1/(1 + CP_STA_UM/ki) or an inactivation-rate term kinact * CP_STA_UM / (KI + CP_STA_UM). Set to 0 for the vehicle-control condition, at which the reversible factor is 1 and, in the inactivation model, the relative-activity state stays at 1 for all time. Reference values: 0, 0.125, 0.25, 0.5 uM (reversible assay) and 0, 0.25, 0.5, 1, 2 uM (inactivation assay) (He 2021 Methods 4.3, 4.4).
• Source aliases:
  • I (He 2021, the inhibitor concentration in the Dixon-plot and double-reciprocal-plot analyses).
• Example models: He_2021_schisantherinA_cyp3a.R (reversible inhibition of CYP3A4 ki 0.15 uM and CYP3A5 ki 0.11 uM, plus time-dependent inactivation of CYP3A4 kinact 0.11 /min, KI 2.45 uM).
• Notes: Specific scope; schisantherin-A-specific. Follows the established CP_<drug>_<units> family (CP_RIF_UM, CP_BMS911543_UM, CP_CARBAMAZEPINE_UM) and the same reuse-for-a-non-plasma-matrix pattern that CP_RIF_UM established for a culture medium and CP_BMS911543_UM for a microsomal incubation buffer. The token is the INN-less lignan common name (STA); no INN exists for a single plant-extract constituent. Distinct from the concomitant-medication indicator [[CONMED_WUZHI]], which is a 0/1 flag for whole-capsule coadministration in a clinical popPK model rather than a numeric single-constituent incubation concentration.

CP_SIA_UM (canonical for the schisandrin A concentration in an in-vitro CYP inhibition incubation)

• Description: Concentration of schisandrin A (SIA), a lignan of the Wuzhi capsule (a standardised ethanolic extract of Schisandra sphenanthera fruit, co-prescribed with tacrolimus in Chinese transplant care), supplied directly as a covariate column rather than computed from a coupled PK model. Founding use is in-vitro: the inhibitor concentration I driving both the reversible-inhibition term 1/(1 + CP_SIA_UM/ki) and the mechanism-based inactivation-rate terms kinact * CP_SIA_UM / (KI + CP_SIA_UM) of a CYP3A4/CYP3A5 inhibition assay. The column is defined generally, so a plasma or unbound site-of-interaction concentration can use the same name; the per-model covariateData entry states which sense is required.
• Units: umol/L (= uM).
• Type: continuous
• Scope: specific
• Reference category: n/a – enters either a reversible-inhibition factor 1/(1 + CP_SIA_UM/ki) or an inactivation-rate term kinact * CP_SIA_UM / (KI + CP_SIA_UM). Set to 0 for the vehicle-control condition, at which the reversible factor is 1 and, in the inactivation model, the relative-activity states stay at 1 for all time. Reference values: 0, 2.4, 7.2, 12 uM (reversible assay) and 0, 2, 4, 8, 16 uM (inactivation assay) (He 2021 Methods 4.3, 4.4).
• Source aliases:
  • I (He 2021, the inhibitor concentration in the Dixon-plot and double-reciprocal-plot analyses).
• Example models: He_2021_schisandrinA_cyp3a.R (reversible inhibition of CYP3A5 ki 8.74 uM, plus time-dependent inactivation of CYP3A4 kinact 0.019 /min, KI 2.54 uM and CYP3A5 kinact 0.014 /min, KI 2.07 uM).
• Notes: Specific scope; schisandrin-A-specific. Same CP_<drug>_<units> family and non-plasma-matrix reuse pattern as [[CP_STA_UM]], with which it is the sibling column from the same paper (schisandrin A is the weaker of the two Wuzhi lignans). The token is the INN-less lignan common name (SIA). Distinct from the concomitant-medication indicator [[CONMED_WUZHI]], which is a 0/1 whole-capsule flag rather than a numeric single-constituent incubation concentration.

CP_SZA_UM (canonical for the schizandrol A concentration in an in-vitro CYP inhibition incubation)

• Description: Concentration of schizandrol A (SZA), a lignan of the Wuzhi capsule (a standardised ethanolic extract of Schisandra sphenanthera fruit, co-prescribed with tacrolimus in Chinese transplant care), supplied directly as a covariate column rather than computed from a coupled PK model. Founding use is in-vitro: the inhibitor concentration I driving the mechanism-based inactivation-rate term kinact * CP_SZA_UM / (KI + CP_SZA_UM) of a CYP3A4 inhibition assay. SZA shows little reversible inhibition, so no reversible term consumes it. The column is defined generally, so a plasma or unbound site-of-interaction concentration can use the same name; the per-model covariateData entry states which sense is required.
• Units: umol/L (= uM).
• Type: continuous
• Scope: specific
• Reference category: n/a – enters an inactivation-rate term kinact * CP_SZA_UM / (KI + CP_SZA_UM). Set to 0 for the vehicle-control condition, at which the relative-activity state stays at 1 for all time. Reference values: 0, 10, 16, 20, 25, 32 uM (pooled-CYP3A and CYP3A4 inactivation assays) and 0, 10, 20, 30, 40, 50 uM (CYP3A5 inactivation assay) (He 2022 Methods 4.4).
• Source aliases:
  • I (He 2022, the inhibitor concentration in the double-reciprocal-plot analyses).
• Example models: He_2022_schizandrolA_cyp3a.R (time-dependent inactivation of CYP3A4 kinact 0.024 /min, KI 15.38 uM; no CYP3A5 TDI and no reversible inhibition).
• Notes: Specific scope; schizandrol-A-specific. Same CP_<drug>_<units> family and non-plasma-matrix reuse pattern as [[CP_STA_UM]] / [[CP_SIA_UM]], the sibling lignan columns from the earlier He 2021 paper on the OTHER two Wuzhi lignans. The token is the INN-less lignan common name (SZA = schizandrol A, distinct from the schisandrin A of [[CP_SIA_UM]] and the schisantherin A of [[CP_STA_UM]]). Distinct from the concomitant-medication indicator [[CONMED_WUZHI]], which is a 0/1 whole-capsule flag rather than a numeric single-constituent incubation concentration.

CP_SZB_UM (canonical for the schizandrol B concentration in an in-vitro CYP inhibition incubation)

• Description: Concentration of schizandrol B (SZB), the more potent of the two schizandrol lignans of the Wuzhi capsule (a standardised ethanolic extract of Schisandra sphenanthera fruit, co-prescribed with tacrolimus in Chinese transplant care), supplied directly as a covariate column rather than computed from a coupled PK model. Founding use is in-vitro: the inhibitor concentration I driving both the reversible-inhibition term 1/(1 + CP_SZB_UM/ki) and the mechanism-based inactivation-rate terms kinact * CP_SZB_UM / (KI + CP_SZB_UM) of a CYP3A4/CYP3A5 inhibition assay. The column is defined generally, so a plasma or unbound site-of-interaction concentration can use the same name; the per-model covariateData entry states which sense is required.
• Units: umol/L (= uM).
• Type: continuous
• Scope: specific
• Reference category: n/a – enters either a reversible-inhibition factor 1/(1 + CP_SZB_UM/ki) or an inactivation-rate term kinact * CP_SZB_UM / (KI + CP_SZB_UM). Set to 0 for the vehicle-control condition, at which the reversible factors are 1 and the relative-activity states stay at 1 for all time. Reference values: 0, 1, 2, 4 uM (reversible assays) and 0, 0.1, 0.2, 0.5, 1, 2 uM (pooled-CYP3A inactivation), 0, 0.5, 1, 2, 4, 8 uM (CYP3A4), 0, 2, 4, 8, 16 uM (CYP3A5) (He 2022 Methods 4.3, 4.4).
• Source aliases:
  • I (He 2022, the inhibitor concentration in the Dixon-plot and double-reciprocal-plot analyses).
• Example models: He_2022_schizandrolB_cyp3a.R (reversible inhibition of CYP3A4 ki 2.18 uM and CYP3A5 ki 2.03 uM, plus time-dependent inactivation of CYP3A4 kinact 0.37 /min, KI 0.69 uM and CYP3A5 kinact 0.009 /min, KI 0.5 uM).
• Notes: Specific scope; schizandrol-B-specific. Same CP_<drug>_<units> family and non-plasma-matrix reuse pattern as [[CP_SZA_UM]], with which it is the sibling column from the same paper (schizandrol B is the more potent of the two schizandrol lignans). The token is the INN-less lignan common name (SZB). Distinct from the concomitant-medication indicator [[CONMED_WUZHI]], which is a 0/1 whole-capsule flag rather than a numeric single-constituent incubation concentration.

HEPARIN_RT (canonical for heparin-chromatography retention time of the administered antibody)

• Description: Retention time (minutes) at which the administered monoclonal antibody elutes from a heparin affinity column under a defined salt gradient, measured at the centre of the elution peak. A high-throughput surrogate for charge-mediated nonspecific binding, and the strongest in-vitro predictor of fast mAb clearance in the Liu 2023 developability panel. Enters the model as the driver of a sigmoidal relationship for the antibody-specific pinocytosis-uptake coefficient F1 of the Shah & Betts platform PBPK model.
• Units: min
• Type: continuous
• Scope: specific
• Reference category: NULL – continuous.
• Source aliases:
  • Heparin_RT – Liu 2023 Supplementary Table S1 column name.
• Example models: Liu_2023_mAb_mouse_pbpk.R.
• Notes: Assay-condition dependent, so retention times are only comparable within a single column chemistry and gradient. Liu 2023 used a HiTrap Heparin High Performance 1 mL column (Cytiva 17040601), 0.4 mg antibody loaded in 50 mM Tris pH 7.6 / 5 mM NaCl, washed 5 column volumes, then eluted on a linear 5-400 mM NaCl gradient over 20 column volumes; the reported value is the elution-peak centre. Observed range in that panel: 2.92-31.5 min (56-antibody training set) and 15.1-32.6 min (14-antibody validation set); the paper’s flag threshold for a fast-clearing antibody is 16.5 min. A related but distinct readout, Heparin_pB_buffer (percent B buffer at elution), was measured in the same runs and is highly correlated with HEPARIN_RT; it is not registered here because no extracted model uses it.

SNP_SLC22A1_RS2282143 (canonical for SLC22A1 (OCT1) 1022C>T (rs2282143) variant carrier indicator)

• Description: Binary genotype indicator for the SLC22A1 (OCT1) c.1022C>T single-nucleotide polymorphism, rs2282143 (protein change P341L / p.Pro341Leu), a decreased-function allele of the hepatic basolateral organic cation transporter 1. 1 = at least one variant (T) allele present; 0 = homozygous wild-type 1022CC. Time-fixed per subject (germline genotype).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (homozygous wild-type 1022CC).
• Source aliases:
  • GENO – Khwarg 2024 (paper Results ‘Effect of SLC22A1 genetic polymorhism on OCT1 activity’: “The SLC22A1 genotype (CC genotype, GENO = 0; CT genotype, GENO = 1) was applied as an exponent of FUP”, and Figure 2 legend).
  • SLC22A1 1022C>T / 1022C>T – Khwarg 2024 (paper Table 1, which keys each assayed SNP by rsID and gives 9 CC + 4 CT + 0 TT among the 13 genotyped subjects).
• Example models: Khwarg_2024_proguanil.R (exponent on the OCT1-uptake multiplier inside the well-stirred hepatic extraction ratio: eh = fub * clint * e_snp_oct1_clint^SNP_SLC22A1_RS2282143 / (qbh + fub * clint * e_snp_oct1_clint^SNP_SLC22A1_RS2282143) with e_snp_oct1_clint = 0.416; CT heterozygotes have 0.42-fold the CC hepatocyte uptake, raising proguanil and lowering cycloguanil exposure; Khwarg 2024 Table 3 FUP with RSE 54.1%).
• Notes: rsID-based canonical name because the source paper reports every assayed SNP by rsID (Table 1) alongside the nucleotide-position form. The sibling SLC22-family entries SNP_SLC22A2_808GT (OCT2) and SNP_SLC22A4_917CT (OCTN1) use the position-based form only because their source (Yoon 2013) reported no rsID; see the SNP_SLC22A2_808GT Notes, which anticipate exactly this rsID-form registration. Because the Khwarg 2024 cohort contained no 1022TT homozygotes (Table 1), the indicator resolves heterozygote-vs-wild-type; a future source observing TT homozygotes should either record the variant-allele count under a paired SNP_SLC22A1_RS2282143_T_COUNT canonical (following SNP_CYP2B6_RS3745274_T_COUNT) or document the pooling in covariateData[[SNP_SLC22A1_RS2282143]]$notes. Khwarg 2024 also assayed rs202220802 (1258-1260delATG), rs12208357 (181C>T), rs34059508 (1393G>C), rs55918055 (262T>C) and rs4646277 (848C>T); all 13 subjects were homozygous wild type for these, so no canonicals are registered for them. Allele frequency of the 1022C>T variant was 16.8% by Hardy-Weinberg in this cohort, against a reported 16.7% in Koreans, 11.0% in Chinese, 5.5% in Vietnamese and 2.0% in German populations (Khwarg 2024 Discussion), so this variant is materially more relevant in East Asian cohorts.

CONC_PMB_MGL (canonical for static in-vitro polymyxin B concentration driving an antibacterial PD model)

• Description: Static (time-invariant) TOTAL polymyxin B concentration applied to the medium of an in-vitro bacterial experiment, supplied as an exogenous covariate that drives bacterial kill. Distinct from a state-derived plasma concentration (Cc) and from the CP_<DRUG> plasma-PD-driver family: this is an applied experimental concentration in the in-vitro matrix. Where the medium binds the drug, this covariate carries the TOTAL concentration and the model derives the unbound concentration internally.
• Units: mg/L
• Type: continuous
• Scope: specific
• Reference category: n/a – 0 mg/L is the drug-free growth control. Lacroix 2025 studied total PMB 0.5 to 128 mg/L for A. baumannii AB121-D0 and 8 to 512 mg/L for AB122-D12.
• Source aliases: none standardized (Lacroix 2025 writes C in Eq 1 and C_T in Fig. S2).
• Example models: Lacroix_2025_polymyxinB_AB121D0.R, Lacroix_2025_polymyxinB_AB122D12.R (founding examples; static total PMB per time-kill tube, converted to the unbound concentration Cu = C_T x fu that drives the Emax / sigmoidal-Emax kill rate).
• Notes: Specific scope because the value is bound to polymyxin B and to the in-vitro time-kill assay design. Member of the in-vitro applied-drug-concentration CONC_<DRUG>_MGL family (siblings CONC_DOR_MGL, CONC_MEM_MGL, CONC_RIF_MGL, CONC_INH_MGL, CONC_IPM_MGL, CONC_TOB_MGL). PMB was shown stable over the 30-h time-kill experiment, so the concentration is genuinely time-invariant and needs no PK ODE state. Pairs with MUCIN_PRESENT, which selects whether the medium binds PMB at all.

MUCIN_PRESENT (canonical for mucin supplementation of the in-vitro growth medium)

• Description: 1 = the in-vitro growth medium was supplemented with 1% (w/v) porcine-stomach mucin, the gel-forming glycoprotein that is the key solid component of airway mucus; 0 = unsupplemented medium (control arm). Time-fixed per experiment (each time-kill tube belongs to exactly one arm). Mucin is a binding matrix as well as a growth-condition modifier, so models typically use this indicator twice: to switch on a medium-binding unbound-fraction submodel, and to select mucin-specific values of the pharmacodynamic parameters it perturbs.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (medium without mucin).
• Source aliases: none – Lacroix 2025 writes Mucin in main-text Eq 4 and supplemental Fig. S2.
• Example models: Lacroix_2025_polymyxinB_AB121D0.R (founding example; gates the Eq 1 unbound-fraction model and selects emax_r_camhb vs emax_r_mucin), Lacroix_2025_polymyxinB_AB122D12.R (gates the same binding model and selects knet_camhb / knet_mucin and ec50_camhb / ec50_mucin).
• Notes: General scope because mucin supplementation is a standard, drug-agnostic in-vitro manipulation used to mimic the airway or gastrointestinal mucus barrier; polymyxins, aminoglycosides and fluoroquinolones are all reported to bind it. Register a distinct covariate rather than reusing this one if a source varies the mucin PERCENTAGE as a continuous exposure (Lacroix 2025 tested a single 1% level and modelled it as a binary categorical, matching the physiological composition of mucus in mild cystic fibrosis). Structurally analogous to M3M3FBS_PRESENT, the binary ex-vivo pretreatment indicator, whose notes anticipate sibling experimental-condition indicators being registered under their own names rather than reusing that entry. Distinct from CONMED_<INN> (clinically administered comedications) because mucin is a medium constituent, not a drug given to a subject.

CONMED_ZOLBETUXIMAB (canonical for concomitant zolbetuximab coadministration indicator)

• Description: 1 = the index chemotherapy dose was administered in the presence of zolbetuximab (a recombinant chimeric IgG1 monoclonal antibody targeting claudin 18.2, CLDN18.2), 0 = the same chemotherapy administered alone. Time-varying by design in the studies that use it: the ILUSTRO Cohort 2 schedule deliberately delayed the Cycle 1 zolbetuximab infusion to Day 3 so that Cycle 1 chemotherapy PK could be sampled without zolbetuximab, with Cycle 2 chemotherapy then given in its presence. The column therefore belongs on the dosing / observation records rather than being a subject-level baseline characteristic.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (chemotherapy alone, without zolbetuximab).
• Source aliases:
  • X_zolbetuximab – the symbol used in the Yamada 2025 covariate equation P = theta_p * (1 + theta_zolbetuximab * X_zolbetuximab) (Methods, “Evaluation of zolbetuximab effect as a covariate”).
• Example models: Yamada_2025_oxaliplatin.R (fractional effect -0.123 shared across V1, V2 and V3, i.e. a 12.3% reduction in each distribution volume of free platinum; retained by forward selection at alpha = 0.01 and backward elimination at alpha = 0.001 with dOFV -13.834, P = 0.0002 – founding example), Yamada_2025_fluorouracil.R (screened on 5-FU CL and V, neither significant, declared in covariatesDataExcluded to preserve the screen).
• Notes: Named-drug member of the CONMED_<INN> family, used as the perpetrator indicator in monoclonal-antibody / cytotoxic-chemotherapy drug-drug-interaction popPK analyses. Note the direction of the interaction question: the covariate flags the presence of the mAb on the PK parameters of the small-molecule victim drug (5-FU, oxaliplatin), not the reverse. When a source reports the effect on the mAb itself, the appropriate covariate is a chemotherapy-backbone indicator (e.g. CONMED_CHEMO) on the mAb model instead.

CONMED_ZOPICLONE (canonical for concomitant zopiclone coadministration indicator)

• Description: 1 = the patient was co-prescribed zopiclone (a cyclopyrrolone non-benzodiazepine hypnotic), 0 = not. Time-fixed per subject in the founding example, which draws the comedication list once from a retrospective inpatient medical log; a study that records zopiclone starting or stopping during the sampling window should carry the column on the observation records instead.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no concomitant zopiclone), which is the larger cohort in the founding example (73 of 81 patients) and carries the structural typical value.
• Source aliases:
  • ZOP – the symbol used in the Han 2025 covariate equation CL/F = 29.6 * (weight/70)^0.75 * (1 - 0.254 * ZOP) (Results, Equations 6-7: “ZOP denoted zopiclone and when schizophrenia patients took ZOP, ZOP denoted 1, otherwise ZOP denoted 0”).
  • Zopiclone tablets – the row label in the Han 2025 comedication table (Table 2), coded Category 0/1 on the same orientation as this canonical.
• Example models: Han_2025_clozapine.R (linear-shift effect on apparent oral clearance: cl *= (1 + e_zop_cl * CONMED_ZOPICLONE) with e_zop_cl = -0.254, i.e. concomitant zopiclone lowers clozapine CL/F by 25.4%; founding example).
• Notes: Named-drug member of the CONMED_<INN> family, used as the perpetrator indicator in a comedication-screening popPK analysis. The founding paper attributes the interaction to competition for CYP3A4, which metabolises both clozapine (alongside CYP1A2) and zopiclone. Use CONMED_ZOPICLONE only for the perpetrator role; a popPK model of zopiclone itself does not need a covariate for the drug being modelled. Cohort prevalence in the founding example was 8 of 81 patients (9.9%, Table 2), so the -0.254 estimate rests on a small exposed subgroup – its bootstrap 95% interval [-0.408, -0.009] very nearly touches zero and should be quoted alongside the point estimate. Zopiclone was also among the 30 comedications screened by Zhang_2024_olanzapine.R (same research group, same centre, overlapping inpatient population), where it was not retained on olanzapine clearance; olanzapine is cleared mainly by CYP1A2 and UGT1A4 rather than CYP3A4, so the two results are consistent rather than contradictory. Promote to general scope once a second model registers zopiclone as a covariate.

CONMED_TISACEL (canonical for scheduled tisagenlecleucel CAR T-cell product indicator)

• Description: 1 = the subject’s scheduled CD19-directed CAR T-cell product is tisagenlecleucel (tisa-cel), 0 = axicabtagene ciloleucel (axi-cel). Time-fixed per subject. Used in models of lymphodepleting chemotherapy given before CAR T-cell infusion, where the product a patient is scheduled to receive both selects the lymphodepletion dosing regimen and marks the treatment cohort.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (axicabtagene ciloleucel), which is the larger cohort in the founding example and carries the structural typical value.
• Source aliases:
  • CAR_T / CART – two-level categorical coded 1 = axi-cel, 2 = tisa-cel in VarelaGonzalezAller_2025_fludarabine.R (Table 2 footnote). Re-express onto the canonical binary orientation as CONMED_TISACEL = CART - 1; a source that codes the levels in the opposite order needs the complementary transformation, so verify the coding statement before mapping.
• Example models: VarelaGonzalezAller_2025_fludarabine.R (founding example; axi-cel-referenced multiplicative factor 3.9/4.4 = 0.8864 on the non-renal clearance arm of fludarabine, re-expressed from the paper’s two-level categorical clearance intercept; 38 axi-cel / 18 tisa-cel).
• Notes: Named-product member of the CONMED_<INN> family; both axicabtagene ciloleucel and tisagenlecleucel are INNs. Timing caveat, important for interpretation: in lymphodepletion popPK models the cell product is infused on day 0 while the conditioning chemotherapy is given on the preceding days, so the product has not been administered during any modelled observation. An effect on a conditioning drug’s PK is therefore a planned-treatment / cohort marker – a proxy for unmeasured differences between the populations selected for each product (disease biology, prior therapies, tumour burden) – and not a pharmacokinetic drug-drug interaction; the founding paper says so explicitly in its Discussion. Record that reading in the per-model covariateData[[CONMED_TISACEL]]$notes rather than implying a mechanistic interaction. Keep the per-product dose difference (30 vs 25 mg/m^2 fludarabine phosphate in the founding example) in the event table, not in this covariate. Register sibling named canonicals (CONMED_LISOCEL, CONMED_BREXUCEL, CONMED_IDECEL, CONMED_CILTACEL,…) for other cell products rather than overloading this entry, and reserve a future CONMED_CART_ANY for sources that pool CAR T-cell therapy as a class.

FORM_MLX_NANOCAPSULE (canonical for meloxicam-loaded polymeric-nanocapsule formulation indicator)

• Description: 1 = the subject received meloxicam formulated as poly(epsilon-caprolactone) polymeric nanocapsules (NC-MLX; interfacial-polymer-deposition nanocapsules, mean diameter 326 nm, zeta potential -26.2 mV, polysorbate-80 coated, capric/caprylic-triglyceride core), 0 = the subject received the free-meloxicam comparator (MLX dissolved in 60% v/v PEG-400 then diluted with water). Per-subject (treatment-arm) categorical indicator in a parallel-arm oral single-dose design; both arms receive the same 0.2 mg/kg meloxicam dose, so the indicator marks the delivery system rather than the amount administered.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (free meloxicam solution; the typical-value reference for both Tlag2 and V2 in the Teixeira 2026 final model).
• Source aliases:
  • NC-MLX – the arm label used throughout Teixeira 2026 (Table 2 “Covariate Effect in the Parameter” column, and the beta_Tlag2_NC-MLX / beta_V2_NC-MLX coefficient names).
  • Formulation – the covariate name used in the Teixeira 2026 supplementary Table S2 model-selection table (footnote c).
• Example models: Teixeira_2026_meloxicam_dog.R (exponential effects on the second-absorption lag time and on the apparent peripheral volume per Teixeira 2026 Table 2: tlag2 = exp(ltlag2 + e_form_nanocapsule_tlag2 * FORM_MLX_NANOCAPSULE + etaltlag2) with e_form_nanocapsule_tlag2 = 0.74, giving 1.22 h for free MLX and 1.22 * exp(0.74) = 2.56 h for NC-MLX; and vp = exp(lvp + e_form_nanocapsule_vp * FORM_MLX_NANOCAPSULE + etalvp) * WT with e_form_nanocapsule_vp = 1.11, giving 0.134 L/kg for free MLX and 0.134 * exp(1.11) = 0.407 L/kg for NC-MLX).
• Notes: Specific scope because the contrast is between one particular nanoencapsulated drug product and its own free-drug solution comparator, not a generic dosage-form category. Distinct from the FORM_LNP_* family (FORM_LNP_SM102, FORM_LNP_MC3,…), which indexes the ionizable lipid of an RNA lipid nanoparticle where the drug is a nucleic-acid cargo; here the nanocarrier encapsulates a small molecule and the reference level is the same small molecule in solution. Distinct from FORM_SOLUTION / FORM_SUSPENSION, where both levels are conventional dosage forms. The mechanism the founding paper attributes the effect to is sustained release from the polymeric shell plus reduced opsonization of polysorbate-coated nanocapsules, which delays the second absorption wave and enlarges the apparent peripheral distribution volume; clearance is unaffected. Future polymeric-nanocapsule / nanosphere formulations of other small molecules should register a sibling FORM_<INN>_NANOCAPSULE canonical rather than reusing this drug-specific entry, unless a cross-drug polymeric-nanocarrier canonical is ratified first.

FORM_DRL17822_ASD (canonical for the DRL-17822 amorphous solid dispersion formulation indicator)

• Description: 1 = the dose record is the amorphous solid dispersion (ASD) oral formulation of the CETP inhibitor DRL-17822, 0 = the nanocrystal (NC) oral formulation. Time-varying per dose record, because the food-by-formulation crossover (Goulooze 2020 study 4, parts A and B) gave both formulations to the same subjects.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (nanocrystal formulation).
• Source aliases:
  • FORM – the $INPUT column of the Goulooze 2020 supplement control stream (0 = nanocrystal, 1 = ASD).
• Example models: Goulooze_2020_DRL17822.R (the formulation does not act through its own coefficient; it selects, together with FED, FED_LOWFAT and FED_HIGHFAT, one of four relative-bioavailability bins: ASD fasted = 0.151 and NC fasted = 0.056 of the reference (NC after a high-fat or continental breakfast), and ASD after a low- or high-fat breakfast = 0.532, the same bin as NC after a low-fat breakfast. The product FED_HIGHFAT * FORM_DRL17822_ASD also scales ka by 0.599).
• Notes: Follows the FORM_<drug>_<formulation> family. Specific scope because the contrast is between two particular products of one development compound. A future ASD formulation of another drug should register its own FORM_<INN>_ASD entry.

TANNER_1 (canonical for Tanner Stage 1 (prepubertal) indicator)

• Description: 1 = the subject is at Tanner Stage 1 of sexual maturation (prepubertal), 0 = the subject is at Tanner Stage 2 or above. Time-fixed at study entry per subject. Used when a paediatric source paper reports a pubertal-maturation effect as the two-level Tanner Stage split its trial protocol used to assign dose and formulation, rather than as a continuous function of AGE, PNA or a Tanner ordinal 1-5.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (Tanner Stage 2 or above), which is the larger cohort in the founding example and carries the structural typical value.
• Source aliases:
  • Tanner Stage – two-level categorical reported as “Tanner Stage 1” vs “Tanner Stage >= 2” in Knebel_2013_atorvastatin.R (Table 2 rows “F1-Tanner Stage 1” and “F1-Tanner Stage 2”). Map onto the canonical binary orientation as TANNER_1 = as.integer(TannerStage == 1). A source that reports the effect on the mature stratum instead needs the complementary indicator; register a sibling TANNER_GE2 rather than silently flipping the reference.
• Example models: Knebel_2013_atorvastatin.R (founding example; multiplicative factor 0.752 on relative bioavailability F1, Tanner-Stage-2-or-above-referenced; 15 Tanner Stage 1 / 24 Tanner Stage 2-or-above patients).
• Notes: Sibling in spirit to the AGE_GE70 / ECOG_GE1 binary-threshold family: a paper-defined developmental band decomposed to a single indicator rather than carried as a continuous or ordinal covariate. Confounding caveat, important for interpretation: paediatric trials routinely key the dose level and the formulation to the Tanner Stage stratum, so the indicator is usually a composite proxy rather than a clean maturation effect. The founding paper says so explicitly – “the Tanner Stage classification in this analysis represents a composite of potential factors such as age, different atorvastatin doses, formulation, and/or treatment compliance” (Knebel 2013 Discussion) – with Tanner Stage 1 patients receiving 5 mg of a chewable tablet and Tanner Stage 2-or-above patients 10 mg of the marketed tablet. Record that reading in the per-model covariateData[[TANNER_1]]$notes; keep the per-stratum dose difference in the event table, not in this covariate, and do not restate a confounded stage effect as a formulation (FORM_<drug>_<formulation>) or dose (DOSE_<drug>_<units>) effect unless the paper actually separates them. When a source models the full Tanner ordinal (stages 1-5) or a genital/breast/pubic-hair sub-scale, register a distinct canonical rather than collapsing it onto this binary. Scope is specific until a second paediatric model ratifies the name; at that point promote to general.

FORM_DUTOGLIPTIN_ECCAPSULE (canonical for the enteric-coated dutogliptin capsule formulation indicator)

• Description: 1 = the dutogliptin dose was given as the enteric-coated hard-shell capsule (drug-coated microcrystalline-cellulose beads in an enteric-coated capsule shell, PROT103 period 4), 0 = any other dutogliptin oral formulation in the pooled analysis (uncoated capsule, aqueous solution, or tartrate-salt tablet). Per-dose covariate, not a subject-level baseline characteristic: PROT103 was a four-period crossover, so the same subject contributes both coated and uncoated records.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0. Paired with FORM_TABLET on the same model so that both indicators = 0 selects the pooled capsule / aqueous-solution reference, which carries the structural lka_cap typical value.
• Source aliases:
  • “enteric-coated capsule” – the formulation level named in Marier 2014 Results (“the enteric-coated capsules displayed much slower absorption (0.0204 hour^-1)”) and in Supplemental Table S1, PROT103 (“Capsules were enteric-coated for the fourth period only”). Marier 2014 does not print a source column name.
• Example models: Marier_2014_dutogliptin.R (founding example; selects the third of three formulation-specific first-order absorption rate constants, ka = 0.0204 1/h for the enteric-coated capsule versus 1.55 1/h for the pooled uncoated-capsule / aqueous-solution reference and 0.558 1/h for the tartrate-salt tablet – a 76-fold slower absorption than the reference. Formulation acts on ka only: Marier 2014 tested formulation and fed/fasting status on the lag time and on relative bioavailability as well and retained neither, so both arms share F and the lag time. Only 6 subjects received the enteric-coated capsule, and the paper’s bootstrap median for this ka was 16.3% above the point estimate for exactly that reason).
• Notes: Drug-scoped member of the FORM_<drug>_<formulation> family rather than a generic FORM_EC_CAPSULE, because the register does not yet carry a general enteric-coated-capsule canonical and the Marier 2014 grouping is paper-specific in two ways: the uncoated capsule is pooled with the aqueous solution into the reference (Marier 2014 reports the solution ka only as “comparable to that of the capsules”, with no separate printed value), and the tablet is a tartrate salt while the capsule is free base. Distinct from FORM_CAPSULE, which contrasts a capsule against a per-paper non-capsule comparator; here the contrast is between two capsule presentations of the same drug that differ only in the enteric coat. Distinct from FORM_DOX_DORYX_MPC (a modified-acid-resistance delayed-release tablet versus a conventional delayed-release tablet) and from the FORM_TABLET / FORM_SOLUTION / FORM_CAPSULE generic trio. A future paper contrasting enteric-coated against uncoated capsules of a different drug should either register its own drug-scoped sibling or propose promoting a generic FORM_EC_CAPSULE canonical; do not overload this entry.

MW (canonical for molecular weight of the modelled drug supplied as a data column)

• Description: Molecular weight of the drug (or drug moiety) the record refers to, supplied to the model as a data column rather than hard-coded as a parameter. Use this when one model file spans several different molecules – a cross-drug meta-analysis, an in-vitro potency panel, or any model whose equations convert between a mass-per-volume concentration and a molar concentration and therefore need a per-record molecular weight. A single-drug PBPK or popPK model should keep its molecular weight as an ini parameter or a literal constant instead; MW exists for the case where the molecule varies from record to record.
• Units: g/mol (equivalently Da). Document per model via covariateData[[MW]]$units.
• Type: continuous
• Scope: general
• Reference category: n/a – enters as a unit-conversion factor. The most common form is the ng/mL to nM conversion C_nM = C_ngmL * 1000 / MW; the mirror conversion is C_ngmL = C_nM * MW / 1000.
• Source aliases:
  • MW – Panday 2025 Table S2 parameter rows and Supplemental Methods Eq. S1; same orientation and units as the canonical.
  • MolWt, MOLWT, M_r – common alternative spellings in physicochemical data tables.
• Example models: Panday_2025_sert_tremor_cavg_mbma.R, Panday_2025_sert_tremor_cmax_mbma.R (g/mol; per-drug molecular weights spanning 20 SERT inhibitors, 221.34 for tapentadol to 441.5 for vilazodone, used in Eq. S1 to put the reported plasma concentration on the same molar basis as the nanomolar in vitro SERT potency; founding example).
• Notes: Do NOT use MW as a stand-in for a salt-factor or free-base correction – those are a separate concern and belong in the dosing records or in an explicit ini conversion parameter. Where a paper reports the molecular weight of a salt form but doses on a free-base basis, record which basis the value is on in covariateData[[MW]]$notes.

KPUU_BRAIN (canonical for the unbound-brain to unbound-plasma partition coefficient supplied as a data column)

• Description: Steady-state ratio of the unbound drug concentration in brain to the unbound drug concentration in plasma (K_p,uu), supplied to the model as a data column. This is the transporter-aware brain-penetration descriptor: a value near 1 indicates passive equilibration, values below 1 indicate net efflux (P-gp / BCRP), and values above 1 indicate net uptake. Distinct from the total brain-to-plasma ratio K_p, which is confounded by nonspecific binding in both matrices; the two relate as K_p,uu = K_p * f_u,brain / f_u,plasma. Use this covariate when the molecule varies from record to record (a cross-drug CNS meta-analysis or potency panel); a single-drug CNS-PBPK model should keep its partition coefficient as an ini parameter in the lkp_<tissue> family instead.
• Units: (unitless ratio)
• Type: continuous
• Scope: general
• Reference category: n/a – enters multiplicatively on the unbound plasma concentration to give the unbound brain concentration: C_br,ub = C_pl,ub * KPUU_BRAIN.
• Source aliases:
  • Kpuu, K_p,uu – Panday 2025 Table S2 (Kpuu and unbound brain to plasma ratio parameter rows) and Supplemental Methods Eq. S2b; same orientation as the canonical.
  • Kp,uu,brain, Kpuu_brain – expanded spellings common in CNS drug-discovery literature.
• Example models: Panday_2025_sert_tremor_cavg_mbma.R, Panday_2025_sert_tremor_cmax_mbma.R (unitless; per-drug values across 20 SERT inhibitors, 0.04 for milnacipran to 2.3 for tramadol and tapentadol, mostly rodent-derived, driving Eq. S2b; founding example).
• Notes: Values in the CNS literature are predominantly rodent-derived even when the clinical endpoint being modelled is human – record the source species in covariateData[[KPUU_BRAIN]]$notes, because the cross-species assumption is usually the dominant uncertainty in the predicted brain exposure. Where a source reports only K_p and a brain fraction unbound, convert with K_p * f_u,brain / f_u,plasma and say so in the notes; where the brain fraction unbound was measured in brain homogenate, check whether the source applied a lysosomal pH-partition correction (Friden et al.) before the conversion, since an uncorrected homogenate value overpredicts f_u,brain for basic drugs. If a future extraction needs the analogous descriptor for another barrier-protected tissue, prefer a sibling canonical in the same shape (KPUU_CSF, KPUU_TUMOR) rather than overloading this entry.

IC50_SERT (canonical for in vitro serotonin-reuptake-transporter inhibitory potency supplied as a data column)

• Description: Half-maximal inhibitory concentration of the drug against the serotonin reuptake transporter (SERT, SLC6A4), measured in vitro and expressed relative to the FREE (unbound) drug concentration in the assay. Supplied as a per-record data column so that one model can span many SERT inhibitors. Used as the denominator of a target-coverage metric, coverage = C_unbound_at_site / IC50_SERT.
• Units: nM (document per model via covariateData[[IC50_SERT]]$units; uM is also common in the source literature and must be converted).
• Type: continuous
• Scope: general
• Reference category: n/a – enters as the denominator of a target-coverage ratio.
• Source aliases:
  • SERT IC50, IC50_SERT, SERT_IC50 – Panday 2025 Table S2 parameter rows and Supplemental Methods Eqs. S7 and S8; same orientation as the canonical.
  • SERT K_D / SERT kD – a DIFFERENT quantity (equilibrium binding affinity rather than functional uptake-inhibition potency). Panday 2025 ran a parallel analysis substituting K_D for IC50 in the same equation; because that substitution changes only the supplied value and the accompanying fitted parameter set, it is recorded here as an alias with a mandatory note rather than as a separate canonical. State explicitly in covariateData[[IC50_SERT]]$notes which of the two a given model’s values are, because the fitted EC50 differs substantially between them.
• Example models: Panday_2025_sert_tremor_cavg_mbma.R, Panday_2025_sert_tremor_cmax_mbma.R (nM; free-corrected geometric-mean potencies across 20 SERT inhibitors, 0.34 nM for paroxetine to about 3220 nM for tapentadol and tramadol; founding example).
• Notes: Values must be on a free-drug basis before use. Where the source assay contained serum or albumin, the measured potency is a total-drug potency and has to be corrected to the unbound assay concentration first; where the assay was protein-free, the measured value is already the free potency. Record which applies per model. Where several literature values exist for one drug, the established convention (Panday 2025) is to use their geometric mean. Proposed as the founding member of an IC50_<TARGET> family so that future off-target-safety or target-coverage extractions can register siblings (IC50_HERG, IC50_NET, IC50_DAT) without a fresh naming decision; <TARGET> is the upper-case common abbreviation of the molecular target.

IC50_CTLA4 (canonical for in vitro CTLA-4 ligand-binding inhibitory potency of an anti-CTLA-4 antibody supplied as a data column)

• Description: Half-maximal inhibitory concentration of an anti-CTLA-4 antibody (ipilimumab, tremelimumab) for inhibition of ligand (CD80/CD86) binding to human CTLA-4, measured in vitro. Supplied as a per-record data column so that one model can pool several anti-CTLA-4 antibodies on a common potency-normalized exposure scale, C_norm = CAV / IC50_CTLA4.
• Units: nM (document per model via covariateData[[IC50_CTLA4]]$units; only the ratio CAV / IC50_CTLA4 enters a model, so both columns must share one unit – the Zhang 2022 models use ng/mL as their source prints).
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as the denominator of a potency-normalized exposure ratio.
• Source aliases:
  • IC50 – Shulgin 2020 Supplemental Methods table (ipilimumab 13.3 nM, tremelimumab 0.65 nM; human CTLA-4 expressed on CHO or HEK cells). The exposure axis the Shulgin 2020 coefficients were fitted on corresponds to 1.33 nM for ipilimumab – every ipilimumab regimen plots 10-fold higher in the paper’s Figure 2 than the printed 13.3 nM gives – so those models document 1.33 nM as the value to supply.
• Example models: Shulgin_2020_ici_trae34_mbma.R, Shulgin_2020_ici_trae34_covariate_mbma.R, Shulgin_2020_ici_gi_imae34_mbma.R, Shulgin_2020_ici_hepatic_imae34_mbma.R (study-level logit meta-regressions of immune checkpoint inhibitor adverse-event rates; founding example), Zhang_2022_ici_irae_mbma.R, Zhang_2022_ici_irae_grade3_mbma.R (supplied in ng/mL as printed in Zhang 2022 Supplementary Table S3: ipilimumab 200, tremelimumab 95.15; 200 ng/mL is about 1.35 nM for a 148 kDa IgG, the same ipilimumab value the Shulgin 2020 fits correspond to).
• Notes: Member of the IC50_<TARGET> family founded by IC50_SERT. Must be strictly positive on every record, including records whose regimen contains no anti-CTLA-4 antibody (where the paired CAV is 0); a zero denominator produces NaN rather than a zero exposure. The ratio assumes the potencies of the pooled antibodies were measured in comparable assays – the Shulgin 2020 values differ twenty-fold, so supplying one drug’s potency with the other’s concentration moves the normalized exposure by more than an order of magnitude.

IC50_NEURAMINIDASE (canonical for the neuraminidase-inhibition IC50 of an infecting influenza virus supplied as a data column)

• Description: Half-maximal inhibitory concentration of a neuraminidase-inhibitor antiviral against the neuraminidase (NA) enzyme of the influenza virus a subject is infected with, measured in vitro by fluorometric NA-inhibition assay. Supplied as a per-subject (time-fixed) data column so that one exposure-response model can span several infecting strains: unlike IC50_SERT, where the target is fixed and the potency varies across drugs, here the drug is fixed and the potency varies across viruses. A property of the virus, not of the patient.
• Units: nM (document per model via covariateData[[IC50_NEURAMINIDASE]]$units; uM is also used in the source literature and must be converted).
• Type: continuous
• Scope: general
• Reference category: paper-specific; state the reference level in covariateData[[IC50_NEURAMINIDASE]]$reference_category. In the founding models the reference is 0.18 nM (influenza A/Texas/36/91 (H1N1)) and the single comparison level is 16.76 nM (influenza B/Yamagata/16/88).
• Source aliases:
  • IC50 – the printed variable name in Rayner 2013 Tables 2, 5 and 6 (written IC50 (nM) as a table row label); same orientation, no value transformation.
• Example models: Rayner_2013_oseltamivir_symptomscore.R, Rayner_2013_oseltamivir_symptomalleviation.R, Rayner_2013_oseltamivir_shedding.R (oseltamivir carboxylate against influenza A/Texas/36/91 (H1N1), 0.18 +/- 0.11 nM, and influenza B/Yamagata/16/88, 16.76 +/- 4.10 nM; founding examples).
• Notes: Member of the IC50_<TARGET> family founded by IC50_SERT, whose Notes declare the family extensible without a fresh naming decision, with <TARGET> the upper-case common abbreviation of the molecular target. The literal abbreviation for influenza neuraminidase is NA, which would give IC50_NA – rejected because a data column of that name reads as “IC50 not available” to any R user, so the target is spelled out here. Follow that precedent for any future target whose abbreviation collides with an R reserved word or missing-value literal. A model with only two observed IC50 levels has fitted a two-level categorical contrast, not a slope: a consuming model must recover the indicator from the printed reference value (highIc50 <- (IC50_NEURAMINIDASE > ic50_ref)) rather than interpolating, and must say in covariateData[[...]]$notes that intermediate values are unsupported. When the IC50 column is perfectly collinear with study identity – which happens whenever each study inoculated one strain – record that in the notes; Rayner 2013 describes its own IC50 covariate as “a surrogate for study name or virus type”, and the whole point of retaining it is to show that strain identity does not drive the exposure-response relationship. Do not use this column for an antiviral whose target is not neuraminidase (register the appropriate IC50_<TARGET> sibling), and do not overload it with an EC50 from a plaque-reduction or hollow-fibre infection model, which measures inhibition of viral replication rather than of the enzyme and takes systematically different values.

CMAX (canonical for maximum drug plasma concentration over a dosing interval supplied as a data column)

• Description: Maximum plasma concentration of the modelled drug over a dosing interval, supplied as a data column. The peak-exposure counterpart of CAV: use CMAX when the model’s exposure-response driver is the peak rather than the interval average. As with CAV, the value is an input the user supplies – typically an individual empirical-Bayes prediction from a previously published population PK model, or a study-arm-level summary in a model-based meta-analysis – not a quantity the model derives.
• Units: ng/mL (document per model via covariateData[[CMAX]]$units; mg/L, ug/mL and nM are all common in source papers).
• Type: continuous
• Scope: specific
• Reference category: n/a – enters via a sigmoidal Emax form. AUC_ANDRO = 0 makes the sigmoid vanish exactly and leaves the baseline effect. Reference values observed (Songvut 2026 Table 2, 30 mg andrographolide per dose): single-dose day-1 AUC(0-4 h) 28.54 +/- 8.72 ugh/L, AUC(0-inf) 32.81 +/- 8.91 ugh/L; day-5 AUC(0-4 h) under q8h dosing 30.12 +/- 15.83 ugh/L with an observed range of roughly 8-57 ugh/L (Figure 5).
• Source aliases:
  • Cmax, CMAX, Cmax,ss – standard spellings.
  • plasma steady state total Cmax, steady state plasma total Cmax, plasma steady state cmax – Panday 2025 Table S2 parameter-row labels; same orientation as the canonical, in ng/mL, on a TOTAL (bound + unbound) basis.
  • CMAXSS / Css,max – Majid 2024 supplement Text S2 $INPUT / $PRED and Table 2; the steady-state maximum serum concentration of a monoclonal antibody, in ug/mL.
• Example models: Majid_2024_lecanemab_ariae.R (ug/mL, individual TOTAL steady-state maximum serum lecanemab concentration as an empirical-Bayes prediction from the companion population PK model modellib('Majid_2024_lecanemab'); enters a binomial logistic exposure-safety regression LINEARLY and UNTRANSFORMED as 0.00666 * CMAX, odds ratio 1.95 per 100 ug/mL, for the incidence of amyloid-related imaging abnormalities with edema/effusion (ARIA-E) – note this is the linear-scale counterpart of the log-scale Chen 2021 entries below, and the units are load-bearing because Majid 2024 Table 2 reports the same effect a second time per ug/deciliter), Panday_2025_sert_tremor_cmax_mbma.R (ng/mL, study-arm-level steady-state total Cmax across 20 SERT inhibitors; converted to a molar unbound brain concentration via Eqs. S1 and S2b and then to a SERT target coverage driving an Emax model of tremor incidence), Chen_2021_lorlatinib_hypercholesterolemia.R (ng/mL, individual TOTAL maximum observed lorlatinib plasma concentration prior to the adverse event, Cmax,event, from the upstream modellib('Chen_2021_lorlatinib') population PK model with time-varying clearance; enters a binomial logistic exposure-safety regression on the NATURAL LOG scale as 1.659 * log(CMAX), odds ratio 5.256 per e-fold, for grade >= 3 hypercholesterolemia – note the averaging window ends at the adverse event and so is subject-specific rather than a fixed dosing interval), Chan_2025_atezolizumab_sae.R (ug/mL, individual TOTAL model-predicted Cycle-1 maximum serum atezolizumab concentration as an empirical-Bayes prediction from the companion population PK model modellib('Chan_2025_atezolizumab'); enters a binomial logistic exposure-safety regression LINEARLY, UNCENTRED and RESCALED as 0.210 * (CMAX / 100), odds ratio 1.23 per 100 ug/mL, for serious adverse events – the relationship is NOT significant (p = 0.441), and the Cycle-1 rather than steady-state landmark was chosen deliberately by the source “to minimize the potential effect of response-dependent time-varying clearance”), Gidal_2018_eslicarbazepine_somnolence.R (adjunctive eslicarbazepine acetate in focal-onset seizures; the only one of that paper’s three adverse-event models for which the PEAK rather than the interval AUC was the statistically significant exposure metric, log-odds -0.000178 per ng/mL), Assmus_2022_emodepside_teae_interest.R, Assmus_2022_emodepside_eye_disorder.R, Assmus_2022_emodepside_nervous_system_disorder.R, Assmus_2022_emodepside_teae_drug_related.R (ng/mL, individual TOTAL venous-plasma emodepside peak over the regimen received, predicted by the companion population PK model modellib('Assmus_2022_emodepside'); enters four separate binary logistic exposure-safety regressions LINEARLY and UNCENTRED, log-odds 0.0077, 0.0085, 0.0063 and 0.0064 per ng/mL for drug-related eye-or-nervous-system TEAEs, eye disorders, nervous system disorders and any drug-related TEAE – Assmus 2022 S6 and S7 Tables), Marier_2021_teduglutide_ppsv.R (ng/mL, individual steady-state teduglutide Cmax at the randomized dose, derived from the companion population PK model modellib('Marier_2021_teduglutide'); enters a time-course Emax model of the change from baseline in prescribed parenteral support volume as the power function (CMAX / 30)^0.684 on Emax, Marier 2021 Table S10; the 30 ng/mL normaliser is taken from the authors’ Table S14 sensitivity model because Table S10 does not print it), Shemesh_2020_atezolizumab_orr_cmax.R, Shemesh_2020_atezolizumab_aeg34_cmax.R, Shemesh_2020_atezolizumab_aesi_cmax.R (ug/mL, individual TOTAL Cycle-1 end-of-infusion serum atezolizumab concentration after the first 1200 mg IV dose, an empirical-Bayes prediction from the Stroh 2017 intravenous popPK model; enters three binomial logistic regressions in high-tumour-mutational-burden patients LINEARLY, UNCENTRED and RESCALED per 100 ug/mL, none significant; coefficients recovered from the plotted fitted curves of Shemesh 2020 Figure S6).
• Notes: Specific scope for the same reason as CAV – the value is intrinsically tied to the modelled drug, with no shared meaning across drugs or studies. Each model’s covariateData[[CMAX]]$notes must state (a) whether the value is a TOTAL or an UNBOUND concentration, (b) whether it is a single-dose or steady-state Cmax, and (c) how it was derived (empirical-Bayes from a named upstream popPK model, a study-arm summary, or a label-reported value). Distinct from CMAX_M1, which is specifically the month-1 maximum concentration in its founding model. Where a model needs both the peak and the interval average, use CMAX and CAV together rather than inventing a combined column; where it needs the trough, use CTROUGH (next entry).

CTROUGH (canonical for trough drug plasma concentration over a dosing interval supplied as a data column)

• Description: Trough (pre-dose) plasma concentration of the modelled drug, supplied as a data column. The third member of the peak / average / trough exposure-metric triple alongside CMAX and CAV: use CTROUGH when the model’s exposure-response driver is the trough rather than the peak or the interval average. As with CMAX and CAV, the value is an input the user supplies – typically an individual empirical-Bayes prediction from a previously published population PK model, or a study-arm-level summary in a model-based meta-analysis – not a quantity the model derives.
• Units: ng/mL (document per model via covariateData[[CTROUGH]]$units; mg/L, ug/mL and nM are all common in source papers).
• Reference category: n/a – a continuous exposure regressor. It may enter linearly, on a log scale, or through a saturable Emax / Imax form; the source paper’s transform must be reproduced exactly, because the model intercept absorbs both the transform and the concentration unit. Reference values observed: 114.97 ng/mL mean (SD 40.28), the cycle-1 lorlatinib trough at 100 mg once daily in Chen 2021.
• Type: continuous
• Scope: specific
• Source aliases:
  • Ctrough, CTROUGH, Ctrough,ss, Ctrough ss – standard spellings.
  • Cmin, Cmin,ss – the classical-NCA spelling of the same quantity for a drug at steady state. Note that for an extravascular drug the minimum concentration over an interval is the END-of-interval value, so Cmin and Ctrough coincide; for an intravenous infusion they need not, and a paper that distinguishes them should have its convention recorded in covariateData[[CTROUGH]]$notes.
  • Ctrough,P1, Ctrough cycle 1 – trough at the end of the first treatment cycle rather than at steady state (Chen 2021). Same orientation as the canonical; the averaging landmark is what differs and belongs in the per-model notes.
• Example models: Liu_2024_saf189s_orr.R, Liu_2024_saf189s_pfs.R, Liu_2024_saf189s_dor.R, Chen_2021_lorlatinib_teae_grade3.R (ng/mL, individual TOTAL lorlatinib trough at steady state from the upstream modellib('Chen_2021_lorlatinib') population PK model with time-varying clearance; enters a binomial logistic exposure-safety regression on the NATURAL LOG scale as 1.167 * log(CTROUGH), odds ratio 3.214 per e-fold, for any grade >= 3 treatment-emergent adverse event – founding example), Han_2024_ainuovirine_adr_ctrough.R (ng/mL, individual TOTAL steady-state ainuovirine trough from the companion modellib('Han_2024_ainuovirine') population PK model; enters a binomial logistic exposure-safety regression linearly and uncentred as 0.0040 * CTROUGH – Han 2024 states the linear form outright, so the coefficient is per ng/mL rather than per e-fold, in contrast to the log-scale Chen 2021 founding example), Han_2024_ainuovirine_virologic_ctrough.R (same column and same linear uncentred form driving the parallel, non-significant exposure-efficacy regression for HIV-RNA < 50 copies/mL at week 48, 1.65 + 0.0038 * CTROUGH, Han 2024 Table 4), Parasrampuria_2025_temsavir_hivrna_day8.R, Parasrampuria_2025_temsavir_resp05log_day8.R, Parasrampuria_2025_temsavir_resp1log_day8.R (ng/mL, individual TOTAL steady-state plasma temsavir concentration at the END OF THE 12 h DOSING INTERVAL – the source writes it Ctau – derived post hoc from the companion modellib('Parasrampuria_2025_temsavir') population PK model and computed from the simulated profiles with PKNCA; enters all three exposure-response models through a SATURABLE Emax form Ctau/(EC50 + Ctau) rather than linearly or on a log scale, with EC50 between 64 and 88 ng/mL, so the three fits sit on the plateau of the curve at the 433 ng/mL cohort-median exposure), Chan_2025_atezolizumab_pfs.R (ug/mL, individual TOTAL model-predicted Cycle-1 trough serum atezolizumab concentration as an empirical-Bayes prediction from the companion population PK model modellib('Chan_2025_atezolizumab'); enters a Cox proportional-hazards relative-hazard model for progression-free survival LINEARLY, UNCENTRED and RESCALED as -0.002 * (CTROUGH / 10) – the first Cox rather than logistic consumer of this column, and effectively a null effect, with a 95 percent CI of -0.052 to 0.047 straddling zero symmetrically), Hartmann_2026_nintedanib_fvcpp.R, Hartmann_2026_nintedanib_fvcz.R (nM, individual TOTAL model-predicted steady-state nintedanib trough as an empirical-Bayes prediction from the companion pediatric population PK model modellib('Hartmann_2026_nintedanib'); enters both FVC disease-progression models through a saturable Emax form emax * CTROUGH / (ec50 + CTROUGH) acting on the rate of change of the endpoint rather than on its level, with EC50 8.05 nM for FVC percent predicted and 8.12 nM for the FVC Z-score. Two firsts for this column: the unit is nM rather than a mass concentration – the source reports every nintedanib exposure in nM and the EC50 unit is printed as nM, so supplying ng/mL would inflate the driver by the 1.853 ng/mL-per-nM factor – and the column is genuinely time-varying within a subject, stepping with each protocol dose adjustment and dropping to 0 during a treatment interruption and for every placebo subject, which is what makes the zero-exposure arm fall out of the same equation), Moein_2025_etrolizumab_induction_clinrem.R, Moein_2025_etrolizumab_induction_endoimp.R, Moein_2025_etrolizumab_induction_endorem.R, Moein_2025_etrolizumab_maintenance_clinrem.R, Moein_2025_etrolizumab_maintenance_endoimp.R, Moein_2025_etrolizumab_maintenance_endorem.R (ug/mL, individual TOTAL etrolizumab trough predicted at week 4 after a SINGLE dose from the companion modellib('Moein_2025_etrolizumab') population PK model; enters all six landmark binomial logistic regressions LINEARLY and UNCENTRED, so the intercept is the placebo probability and the slope is per ug/mL. The SINGLE-DOSE landmark is deliberate and is the sharpest illustration of why this entry’s Notes require the landmark to be recorded: etrolizumab clearance FALLS with time in a way that correlates with clinical improvement, so an on-treatment trough would be confounded by the outcome it is being regressed on, and the authors back-predicted a one-dose trough to break that circularity. The induction models use the patient’s assigned 105 or 210 mg dose while the maintenance models always use 105 mg, so the two columns are NOT interchangeable), Komatsu_2016_penicillinG_clinical_success.R (ug/mL, MEASURED TOTAL serum penicillin G trough in adults with viridans-streptococcal infective endocarditis; the only entry so far whose column is an observed assay value rather than an empirical-Bayes prediction from an upstream population PK model, which is why its landmark is simply “the trough drawn during therapy” – Komatsu 2016 sampled immediately before a dose and 2 or 3 h after it. It is also the first entry in which the column does NOT enter the linear predictor directly: it is first divided by a model mic parameter to form the dimensionless PK/PD index Cmin/MIC, per the AUCMIC_TYLO split-the-ratio rule, and the resulting index enters LINEARLY and UNCENTRED as 0.0524 * (CTROUGH / mic), so the slope is per unit of ratio and not per ug/mL. Supplying this model a trough without also setting mic to the isolate’s susceptibility silently rescales the driver), Hibma_2026_elranatamab_crs.R (ng/mL, individual FREE – not total – elranatamab concentration on Cycle 1 Day 4, i.e. the trough of the 12 mg first step-up priming dose interval taken just before the 32 mg second step-up dose, which for this slowly absorbed SC antibody is also the interval Cmax; an empirical-Bayes prediction from the companion target-binding population PK model modellib('Hibma_2026_elranatamab'), where it is the free-drug output Cc at time 3 days. Enters a landmark binomial logistic regression for any-grade cytokine release syndrome after the first step-up dose on the NATURAL LOG scale, 1.35 * log(CTROUGH), odds ratio 3.86 per e-fold – the first consumer whose landmark is a single priming dose rather than steady state or end of cycle 1), Shemesh_2020_atezolizumab_orr_ctrough.R, Shemesh_2020_atezolizumab_aeg34_ctrough.R, Shemesh_2020_atezolizumab_aesi_ctrough.R (ug/mL, individual TOTAL serum atezolizumab trough at the END OF CYCLE 1 – the source’s ‘cycle 1 Cmin’, taken at the day-1 cycle-2 pre-dose time – as an empirical-Bayes prediction from the Stroh 2017 intravenous popPK model; enters three binomial logistic regressions in high-tumour-mutational-burden patients LINEARLY, UNCENTRED and RESCALED per 10 ug/mL, none significant, the objective-response slope being drawn exactly flat; coefficients recovered from the plotted fitted curves of Shemesh 2020 Figure 3B and Figure S6).
• Notes: Specific scope for the same reason as CMAX and CAV – the value is intrinsically tied to the modelled drug, with no shared meaning across drugs or studies. Each model’s covariateData[[CTROUGH]]$notes must state (a) whether the value is a TOTAL or an UNBOUND concentration, (b) which landmark the trough is taken at (single dose, end of cycle 1, steady state), and (c) how it was derived (empirical-Bayes from a named upstream popPK model, a study-arm summary, or a label-reported value). The landmark matters wherever clearance is time-varying: lorlatinib auto-induces its own metabolism over roughly the first week, so its steady-state trough is materially lower than its day-1 trough, and substituting one for the other biases any model fitted on the log scale. Registered on the ratified reading that a trough is not an alias of an interval average – several AUC_<DRUG> entries (AUC_LEN, AUC_GCV, AUC_RTV, AUC_EMPA, AUC_ISTRA, AUC_OLM) and CMAX_APTM already direct that a model using a different exposure metric “should register a parallel canonical rather than overload” theirs, and this is that parallel canonical for the trough. Named CTROUGH rather than CMIN to match the token used by the founding paper and by regulatory exposure-response practice generally, with Cmin carried as a source alias. Ratified canonically by the maintainers on 2026-09-02 alongside the Chen 2021 lorlatinib exposure-response extraction. The FOUNDING consumer is Chen_2021_lorlatinib_teae_grade3 (ng/mL, individual TOTAL lorlatinib trough at steady state from the upstream Chen_2021_lorlatinib population PK model with time-varying clearance; a binomial logistic exposure-safety regression on the natural-log scale, 1.167 * log(CTROUGH), odds ratio 3.214 per e-fold, for any grade >= 3 treatment-emergent adverse event). It is named here in prose rather than in the Example models list because the model file was not yet in the library when this entry was written, and checkNamingRegisters() requires every cited example file to exist; add it to the Example models list once the file is in the library.

DBP (canonical for diastolic blood pressure)

• Description: Subject diastolic blood pressure, in mmHg. Captured at baseline or serially during a study. The diastolic sibling of SBP, completing the covariate pair that mirrors the dbp / sbp PD-state pair in compartment-names.md. Used where hemodynamic state is a covariate on a non-PK endpoint (blood-pressure exposure-response, disease progression, survival) or on clearance. Document baseline-vs-time-varying status per model in covariateData[[DBP]]$notes.
• Units: mmHg
• Type: continuous
• Scope: general
• Reference category: n/a – used with power scaling (DBP / ref)^exponent, linear-deviation forms (1 + e * (DBP - ref)), or additive-log forms beta * log(DBP). Reference values observed: 102 mmHg (Song 2013 Table 2 pooled-Phase-III mean baseline seated trough diastolic pressure in treated hypertensive patients); 80 mmHg is the population-adult clinical normal midpoint.
• Source aliases:
  • Baseline BP – Song 2013 Table 3 / Table S6 row label for the baseline seated trough diastolic pressure entering both the placebo and the olmesartan / amlodipine drug-effect terms of the diastolic exposure-response model; same orientation, no value transformation.
  • SeDBP – seated trough diastolic blood pressure, the Song 2013 measurement convention.
  • DBP – source-paper column name; same orientation, no value transformation.
• Example models: Song_2013_olmesartan_amlodipine_hydrochlorothiazide_dbp.R (baseline seated trough DBP; enters the placebo term as (DBP / 102)^3.19 and the olmesartan and amlodipine drug-effect terms as (DBP / 102)^2.46 and (DBP / 102)^4.12 – founding example).
• Notes: Keep the two as separate columns rather than a single BP column with a type flag: papers routinely fit systolic and diastolic responses as independent models with different covariate sets (Song 2013 retains an age effect on the olmesartan effect for diastolic only, and a sex effect on the amlodipine effect for systolic only), so a downstream dataset must be able to carry both simultaneously. Distinct from DBP_REL, which carries a drug-induced relative change from baseline rather than an absolute pressure.

AUC_OLM (canonical for per-subject steady-state AUC of olmesartan)

• Description: Per-subject steady-state area under the olmesartan plasma concentration-time curve over the 24 h dosing interval, AUCss, computed as Dose / (CL/F) from the individual post hoc Bayesian clearance estimate of an olmesartan population PK model.
• Units: ng*h/mL (document per-model via covariateData[[AUC_OLM]]$units if a different exposure unit is reported).
• Type: continuous
• Scope: specific
• Reference category: n/a – enters via a saturable Emax form emax_om * AUC_OLM / (AUC_OLM + eauc50_om) on the blood-pressure drug-effect term. Set AUC_OLM = 0 to drop olmesartan from the combination (the olmesartan mono-effect and every interaction term containing it vanish). Reference values observed: EAUC50 = 1850 ngh/mL (diastolic) and 1590 ngh/mL (systolic) in Song 2013 Table 3.
• Source aliases:
  • AUCOM / AUC_OM – Song 2013 Results notation for the olmesartan steady-state exposure entering the exposure-response models.
• Example models: Song_2013_olmesartan_amlodipine_hydrochlorothiazide_dbp.R, Song_2013_olmesartan_amlodipine_hydrochlorothiazide_sbp.R (saturable Emax olmesartan arm of the CS-8635 triple-combination blood-pressure exposure-response models – founding examples).
• Notes: Specific scope because the column meaning is tied to olmesartan and to the q24h steady-state AUC convention. Member of the AUC_<DRUG> family (AUC_CARBO, AUC_GEM, AUC_GCV, AUC_LCM, AUC_CBZ, AUC_PAZO, AUC_RTV, AUC_EMPA, AUC_IBRU, AUC_VERUB,…). The upstream model that generates this column is modellib('Song_2013_olmesartan'); because Song 2013 defines AUCss as Dose / (CL/F) exactly, the column can be computed in closed form from that model’s clearance without integrating a profile. A future model using a different olmesartan exposure metric (Cmax, Ctrough, a q12h-interval AUC) should register a parallel canonical rather than overload AUC_OLM.

AUC_AML (canonical for per-subject steady-state AUC of amlodipine)

• Description: Per-subject steady-state area under the amlodipine plasma concentration-time curve over the 24 h dosing interval, AUCss, computed as Dose / (CL/F) from the individual post hoc Bayesian clearance estimate of an amlodipine population PK model.
• Units: ng*h/mL (document per-model via covariateData[[AUC_AML]]$units if a different exposure unit is reported).
• Type: continuous
• Scope: specific
• Reference category: n/a – enters via a saturable Emax form emax_aml * AUC_AML / (AUC_AML + eauc50_aml) on the blood-pressure drug-effect term. Set AUC_AML = 0 to drop amlodipine from the combination. Reference values observed: EAUC50 = 453 ngh/mL (diastolic) and 309 ngh/mL (systolic) in Song 2013 Table 3.
• Source aliases:
  • AUCAML – Song 2013 Results notation for the amlodipine steady-state exposure entering the exposure-response models.
• Example models: Song_2013_olmesartan_amlodipine_hydrochlorothiazide_dbp.R, Song_2013_olmesartan_amlodipine_hydrochlorothiazide_sbp.R (saturable Emax amlodipine arm of the CS-8635 triple-combination blood-pressure exposure-response models – founding examples).
• Notes: Specific scope because the column meaning is tied to amlodipine and to the q24h steady-state AUC convention. Member of the AUC_<DRUG> family. The upstream model that generates this column is modellib('Song_2013_amlodipine'). Song 2013 found a saturable amlodipine exposure-response relationship where the same group’s earlier work had used a linear one, attributing the change to the enrichment of the dataset with the CS8635-A-U301 (TRINITY) study – so a model that reuses this column with a linear slope is describing a different, earlier analysis and should say so in its notes.

AUC_HCTZ (canonical for per-subject steady-state AUC of hydrochlorothiazide)

• Description: Per-subject steady-state area under the hydrochlorothiazide plasma concentration-time curve over the 24 h dosing interval, AUCss, computed as Dose / (CL/F) from the individual post hoc Bayesian clearance estimate of a hydrochlorothiazide population PK model.
• Units: ng*h/mL (document per-model via covariateData[[AUC_HCTZ]]$units if a different exposure unit is reported).
• Type: continuous
• Scope: specific
• Reference category: n/a – enters via a LINEAR slope form slope_hctz * AUC_HCTZ / 1000 on the blood-pressure drug-effect term, the slope being reported per 1000 ngh/mL. Set AUC_HCTZ = 0 to drop hydrochlorothiazide from the combination. Reference values observed: slopes -3.3 mmHg (diastolic) and -9.38 mmHg (systolic) per 1000 ngh/mL in Song 2013 Table 3.
• Source aliases:
  • AUCHCTZ – Song 2013 Results notation for the hydrochlorothiazide steady-state exposure entering the exposure-response models.
• Example models: Song_2013_olmesartan_amlodipine_hydrochlorothiazide_dbp.R, Song_2013_olmesartan_amlodipine_hydrochlorothiazide_sbp.R (linear hydrochlorothiazide arm of the CS-8635 triple-combination blood-pressure exposure-response models – founding examples).
• Notes: Specific scope because the column meaning is tied to hydrochlorothiazide and to the q24h steady-state AUC convention. Member of the AUC_<DRUG> family. The upstream model that generates this column is modellib('Song_2013_hydrochlorothiazide'). Unlike its olmesartan and amlodipine siblings the hydrochlorothiazide arm is unsaturated over the studied 12.5-25 mg dose range, so the per-1000-unit slope convention matters: supplying the raw AUC to a model expecting the divided form overstates the diuretic effect by three orders of magnitude.

STUDY_CS8635_A_U301 (canonical for CS8635-A-U301 (TRINITY) study indicator)

• Description: 1 = subject enrolled in CS8635-A-U301 (TRINITY), the pivotal phase III factorial trial of the olmesartan / amlodipine / hydrochlorothiazide triple fixed-dose combination CS-8635, 0 = another study in the pooled exposure-response analysis.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the other two pooled phase III studies, CS8663-A-U301 and 866-318). The three study indicators partition the pooled dataset and sum to 1 for every subject.
• Source aliases:
  • Study CS8635-A-U301 – Song 2013 Table 3 / Tables S6 and S7 placebo-effect row label.
• Example models: Song_2013_olmesartan_amlodipine_hydrochlorothiazide_dbp.R, Song_2013_olmesartan_amlodipine_hydrochlorothiazide_sbp.R (selects the study-specific placebo effect, -3.80 mmHg diastolic / -4.20 mmHg systolic; in the systolic model it also gates the Hispanic-ethnicity effect on the placebo term).
• Notes: Member of the STUDY_<id> family. Song 2013 estimated the placebo effect as a separate numerical value per contributing study rather than as a single pooled intercept, so all three indicators must be supplied together; a dataset that leaves all three at 0 produces a zero placebo effect, which is not a state the model was fitted to.

STUDY_CS8663_A_U301 (canonical for CS8663-A-U301 (COACH) study indicator)

• Description: 1 = subject enrolled in CS8663-A-U301 (COACH), the pivotal phase III factorial trial of the olmesartan / amlodipine dual fixed-dose combination, 0 = another study in the pooled exposure-response analysis.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (CS8635-A-U301 and 866-318).
• Source aliases:
  • Study CS8663-A-U301 – Song 2013 Table 3 / Tables S6 and S7 placebo-effect row label.
• Example models: Song_2013_olmesartan_amlodipine_hydrochlorothiazide_dbp.R (selects the -3.57 mmHg placebo effect and gates the Black-race effect on the placebo term, which Song 2013 retained only within this study), Song_2013_olmesartan_amlodipine_hydrochlorothiazide_sbp.R (selects the -3.45 mmHg placebo effect).
• Notes: Member of the STUDY_<id> family. Note the study-gated covariate: the Black-race effect on the diastolic placebo response (-0.607) applies only to subjects in this study, so the indicator does double duty as a placebo selector and as an interaction gate.

STUDY_866_318 (canonical for 866-318 study indicator)

• Description: 1 = subject enrolled in 866-318 (written SE866-318 in the parameter tables), the phase III placebo-controlled factorial trial of the olmesartan / hydrochlorothiazide dual fixed-dose combination, 0 = another study in the pooled exposure-response analysis.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (CS8635-A-U301 and CS8663-A-U301).
• Source aliases:
  • Study SE866-318 – Song 2013 Table 3 / Tables S6 and S7 placebo-effect row label.
  • 866-318 – the form used in the Song 2013 main text and in Table 2.
• Example models: Song_2013_olmesartan_amlodipine_hydrochlorothiazide_dbp.R, Song_2013_olmesartan_amlodipine_hydrochlorothiazide_sbp.R (selects the study-specific placebo effect, -6.08 mmHg diastolic / -5.26 mmHg systolic – markedly larger than in the other two studies).
• Notes: Member of the STUDY_<id> family. The leading digit in the canonical name is preceded by STUDY_ so the identifier remains a valid R name; the source study code has no alphabetic prefix.

DOSE_PHAGE_PFU (canonical for administered bacteriophage dose in plaque-forming units)

• Description: The bacteriophage dose administered on the current dose record, expressed in plaque-forming units (PFU). Per-dose-record covariate. Used by phage PBPK models whose state system is carried in percent-injected-dose (%ID) rather than PFU: the dose event itself is then always amt = 100, and this column supplies the PFU scale that the mononuclear-phagocyte-system carrying capacity is calibrated in.
• Units: PFU
• Type: continuous
• Scope: specific
• Reference category: n/a – enters the MPS carrying-capacity conversion A_RES [%ID per 1e5 phagocytes] = 10^A_MPS * 100 / DOSE_PHAGE_PFU, where A_MPS is the literature capacity in log10(PFU / 1e5 cells). Because the capacity is a fixed absolute number of phage particles, raising the dose lowers the fraction of the dose the MPS can hold, which is the model’s only source of dose non-linearity.
• Source aliases:
  • Mean Phage Dose (Log10 PFU) – Echterhof 2026 Table S1 reports the dose on the log10 scale (10.7 Luz24, 10.1 OMKO1, 11.9 PAML-31-1); the covariate column carries the untransformed PFU value, so supply 10^x.
• Example models: Echterhof_2026_phage_mouse_pbpk.R (founding example), Echterhof_2026_phage_rat_pbpk.R, Echterhof_2026_phage_human_pbpk.R.
• Notes: Well-formed member of the auto-approved DOSE_<DRUG>_<UNITS> family (cf. DOSE_MTX_MGM2, DOSE_EMPA_MGD). A dedicated column is required rather than the generic DOSE canonical for exactly the reason given in the DOSE_MTX_MGM2 entry: amt is in %ID while the capacity relation is calibrated in PFU, so carrying both in one column would silently mix units. <DRUG> is PHAGE rather than a strain name because the capacity term is a property of the phagocyte, not of the phage strain – Echterhof 2026 applies the same A_MPS to Luz24, OMKO1 and PAML-31-1. A model that carries its states natively in PFU does not need this column at all.

STUDY_PAML31 (canonical for the PAML-31-1 bacteriophage study-arm indicator)

• Description: 1 = the PAML-31-1 phage study arm; 0 = the Luz24 or OMKO1 arms, which share one typical value. Time-fixed per subject. Selects the covariate-adjusted active surface clearance in the Echterhof 2026 phage PBPK.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (Luz24 and OMKO1 pooled), which carries the structural typical value CL_Active.
• Source aliases:
  • PAML-31 – the categorical covariate level named in Echterhof 2026 Results (“a categorical covariate to describe a PAML-31-specific reduction in CL_Active”).
• Example models: Echterhof_2026_phage_mouse_pbpk.R (founding example), Echterhof_2026_phage_rat_pbpk.R, Echterhof_2026_phage_human_pbpk.R.
• Notes: Well-formed member of the auto-approved STUDY_<id> family. Encoding caveat: Echterhof 2026 Table 1 reports the covariate coefficient theta_PAML31 = -1.94 alongside both an unadjusted CL_Active (1.29e-2 L/h/kg) and a covariate-adjusted CL_Active,PAML (1.34e-3 L/h/kg), and no standard transform connects the three (exp(-1.94) = 1.85e-3, and the two tabulated clearances are 9.63-fold apart while the prose claims 10.7-fold). The founding models therefore switch DIRECTLY between the two tabulated clearances and never evaluate theta; a future extraction that recovers a coherent coefficient should record the resolution here.

STUDY_PAPAYA (canonical for Lau 2026 paracetamol PAPAYA oral study cohort indicator)

• Description: 1 = subject enrolled in Study 1 of the Lau 2026 pooled paracetamol analysis (the PAPAYA study: single 1000 mg oral paracetamol suspension in adults with and without obesity, Albert Schweitzer Hospital / Erasmus MC, the Netherlands, no fasting restrictions); 0 = otherwise.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0. Paired with STUDY_CHEN; when both are 0 the subject is in Study 3 (van Rongen intravenous), which carries the reference structural clearances and the reference residual-error magnitudes.
• Source aliases:
  • FLG3 – the NONMEM flag in the deposited control stream (Supplementary Materials 3), set by IF (ID.GT.100.AND.ID.LT.300) FLG3=1.
• Example models: Lau_2026_paracetamol.R (founding example).
• Notes: Selects two things and only two things: the Study 1 gallbladder-emptying schedule (first emptying 60 min post-dose, second 240 min, each lasting 6 min – the earliest of the three studies because Study 1 participants had no fasting restrictions and food intake accelerates gallbladder motility) and the Study 1 fold increase in the combined paracetamol residual error (2.25), together with the shared oral fold on the PCM-GLU residual error (0.872) that Studies 1 and 2 have in common. Study 1 carries no structural clearance multiplier – it shares the reference intrinsic and elimination clearances with Study 3. This asymmetry with STUDY_CHEN is deliberate and load-bearing: the Lau 2026 backward elimination retained study effects only where Study 2 (higher BMI at comparable total body weight) differed.

STUDY_CHEN (canonical for Lau 2026 paracetamol Chen oral-liquid study cohort indicator)

• Description: 1 = subject enrolled in Study 2 of the Lau 2026 pooled paracetamol analysis (the previously published Chen et al. study: single 1500 mg oral paracetamol liquid in patients with obesity scheduled for laparoscopic Roux-en-Y gastric bypass, United States, fasted); 0 = otherwise.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0. Paired with STUDY_PAPAYA; when both are 0 the subject is in Study 3 (van Rongen intravenous), the reference cohort.
• Source aliases:
  • FLG1 – the NONMEM flag in the deposited control stream (Supplementary Materials 3), set by IF (ID.GT.300) FLG1=1.
• Example models: Lau_2026_paracetamol.R (founding example).
• Notes: The most load-bearing of the two indicators. It selects the Study 2 gallbladder-emptying schedule (480 and 660 min post-dose, the latest of the three studies, consistent with the fasted state), four structural multipliers – 1.94 on the oxidative intrinsic hepatic clearance, 0.650 on the PCM-GLU elimination clearance, 0.826 on the PCM-SUL elimination clearance and 1.59 on the PCM-CYS & PCM-MER elimination clearance – the switch of the PCM-CYS & PCM-MER residual error from proportional to purely additive (0.288 umol/L), and the Study 2 fold increase in the combined paracetamol residual error (1.87). Lau 2026 Discussion attributes the structural multipliers to the higher BMI of the Study 2 cohort at total body weights comparable to the other studies, hypothesising metabolic-associated fatty liver disease (MAFLD) as the mechanism; the covariate is therefore a cohort marker standing in for an unmeasured hepatic-status covariate rather than a study-conduct artefact, and that reading should be preserved in the per-model covariateData notes.

SELFADMIN (canonical for self-administered (not directly observed) dosing-occasion indicator)

• Description: 1 = the dosing occasion was self-administered, i.e. taken by the patient or a caregiver without direct supervision by study or clinic staff; 0 = the dosing occasion was directly observed therapy (DOT). Per-DOSING-OCCASION and time-varying within a subject – a subject typically contributes both observed and unobserved occasions. Acts as an adherence proxy on relative oral bioavailability: an apparent reduction in F on unobserved occasions is the pharmacokinetic signature of missed or partial doses and should be interpreted as such rather than as a true absorption effect unless the source argues otherwise.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (directly observed therapy). Encoded multiplicatively as e_selfadmin_fdepot^SELFADMIN so the reference occasion is unmodified.
• Source aliases:
  • theta Self-administered DP – Wallender 2021 Table 2 / Eq. 2 parameter label. Papers that code the complement (1 = directly observed) must invert to this canonical and record the inversion in covariateData[[SELFADMIN]]$notes.
• Example models: Wallender_2021_piperaquine.R (multiplicative 0.397 on relative oral bioavailability when the second and third daily dihydroartemisinin-piperaquine doses were taken at home rather than in clinic, i.e. 60% lower apparent bioavailability; the authors attribute this to missed doses rather than to absorption), Kengo_2025_atazanavir.R, Kengo_2025_atazanavir_a5231.R (a VARIANCE-model role rather than a typical-value one: the indicator does not shift bioavailability at all, it multiplies the between-occasion etas of ka, mean transit time AND bioavailability by 1.63 (DERIVE) or 1.7 (A5231) for doses that were not directly observed, encoding the source control stream’s IF (OBS.EQ.0) THEN BOVKA = SCALE_BOV*BOVKA ... branch. The sign convention is unchanged (1 = unobserved), and the effect is still adherence / dose-timing uncertainty; what differs is that the uncertainty is expressed as inflated occasion-to-occasion spread instead of a mean shift, which is the appropriate encoding when a trial’s pre-dose samples follow an unwitnessed home dose), Burns_2015_tenofovir.R (a third role, and the most explicit: the indicator selects between a relative bioavailability FIXED at 0.5 with a FIXED variance of 10 on the unobserved dose, and 1 on the observed one. This is Gibiansky’s adherence-adjustment method, in which the wide fixed prior is deliberately near-uninformative so each subject’s empirical Bayes estimate of F absorbs missed doses (low) or extra doses (high) from their own pre-dose concentration; the reported estimates spanned 0.015 to 4.17. Note the consequence for downstream simulation: set the column to 0 on every dose to model full compliance, because sampling that fixed variance produces the “exaggerated prediction intervals” the source paper itself reports and works around), Palmer_2025_moxifloxacin.R (the effect is on the IOV MAGNITUDE rather than on a typical value: iov_scale <- e_selfadmin_iovscale^SELFADMIN with e_selfadmin_iovscale = 1.70 multiplies the interoccasion-variability standard deviation on BOTH bioavailability and mean absorption time, per the control stream’s UNOBS_EFF = 1 + THETA(10)*UNOBS, F1 = 1*EXP(IOVF*UNOBS_EFF) and MAT = TVMAT*EXP(IOVMAT*UNOBS_EFF). In the CATALYST trial the unobserved occasions are the predose samples arising from the previous day’s at-home dose, and the paper attributes the inflated variability to caregivers reporting imprecise dosing times).
• Notes: Distinct from a continuous adherence fraction (percentage of prescribed doses taken), which would warrant a separate continuous canonical if a future paper reports one. Distinct from OCC, the integer occasion index used to multiplex inter-occasion variability – the two are commonly carried together, as in Wallender 2021, where each DP course is both an IOV occasion and either observed or unobserved. The polarity is deliberately set to self-administered = 1 so that the coefficient is a bioavailability PENALTY and maps 1:1 onto how sources report it (a fold change below 1 relative to observed dosing), avoiding a reference-category inversion at transcription time. Ratified 2026-09-02 alongside the Wallender 2021 piperaquine extraction (maintainers’ decision, PMC8602248).

TRANSM_HIGH_2015 (canonical for the 2015 high-malaria-transmission calendar-period indicator)

• Description: 1 = the record falls within the 2015 annual high-malaria-transmission period at the study site; 0 = otherwise. Time-varying within subject. In Wallender 2021 the high-transmission period is defined as 1 March to 31 August annually in Tororo, Uganda; the complement (all low-transmission months across all study years, pooled) is the reference.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 – pooled low-transmission periods. Enters as a multiplicative adjustment on the baseline malaria hazard.
• Source aliases:
  • theta Transmission period 2015 – Wallender 2021 Table 2 parameter label.
• Example models: Wallender_2021_piperaquine_malaria.R (multiplicative 1.29 on the baseline incident-malaria hazard).
• Notes: Specific scope because the calendar boundaries and the year-specific magnitudes are tied to one site’s transmission history and its indoor-residual-spraying schedule (bendiocarb through 2015, pirimiphos-methyl from 2016), which is why the three annual multipliers span 1.29 to 7.83 rather than sharing a single seasonal value. A model at a different site, or one estimating a single season-independent high-transmission effect, must not reuse these columns; register a site- or study-specific member of the same family instead. Members of the family: TRANSM_HIGH_2015, TRANSM_HIGH_2016, TRANSM_HIGH_2017. The three indicators are mutually exclusive one-hot columns; setting all three to 0 selects the low-transmission reference. Distinct from SEASON2 (RSV second-season indicator) and SEMESTER (paired winter/spring vs summer/fall indicator). Ratified 2026-09-02 alongside the Wallender 2021 piperaquine extraction (maintainers’ decision, PMC8602248).

TRANSM_HIGH_2016 (canonical for the 2016 high-malaria-transmission calendar-period indicator)

• Description: 1 = the record falls within the 2016 annual high-malaria-transmission period at the study site; 0 = otherwise. Time-varying within subject. Same calendar definition and reference as TRANSM_HIGH_2015.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 – pooled low-transmission periods.
• Source aliases:
  • theta Transmission period 2016 – Wallender 2021 Table 2 parameter label.
• Example models: Wallender_2021_piperaquine_malaria.R (multiplicative 5.20 on the baseline incident-malaria hazard).
• Notes: See TRANSM_HIGH_2015.

TRANSM_HIGH_2017 (canonical for the 2017 high-malaria-transmission calendar-period indicator)

• Description: 1 = the record falls within the 2017 annual high-malaria-transmission period at the study site; 0 = otherwise. Time-varying within subject. Same calendar definition and reference as TRANSM_HIGH_2015.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 – pooled low-transmission periods.
• Source aliases:
  • theta Transmission period 2017 – Wallender 2021 Table 2 parameter label.
• Example models: Wallender_2021_piperaquine_malaria.R (multiplicative 7.83 on the baseline incident-malaria hazard).
• Notes: See TRANSM_HIGH_2015.

Simulation / methodology covariates (SIMCOV_<role> family)

Methodology papers demonstrate an estimator on synthetic data, and their covariates are frequently meaning-free by construction – a standard-normal draw with no clinical referent, present only so the method has something to estimate a coefficient for. Such a column has no clinical concept to name it after, so it cannot be mapped onto any of the demographic, laboratory or disease-state canonicals above.

The SIMCOV_<role> family covers exactly this case. The <role> suffix records how the column enters the model, not what it measures (there is nothing to measure) and not whether the data column itself varies with time:

• SIMCOV_TI – enters the model directly, as a time-invariant effect.
• SIMCOV_TV – enters the model through a time-varying transform.

Reach for this family only when the source covariate is genuinely semantically empty. A synthetic covariate that still denotes a real physical quantity – a simulated AUC, a simulated body weight – takes the ordinary canonical for that quantity, or a specific-scope sibling such as AUC_BAST_FW.

SIMCOV_TI (canonical for a meaning-free simulation covariate with a time-invariant effect)

• Description: A semantically-empty per-subject covariate used in a simulation or methodology study to demonstrate estimation of a time-invariant covariate effect. Carries no clinical interpretation; in the founding example it is a standard-normal draw (mean 0, SD 1) entering the log hazard directly.
• Units: (z-score; standard normal, mean 0, SD 1 in the founding example – record the actual generating distribution per model via covariateData[["SIMCOV_TI"]]$units)
• Type: continuous
• Scope: general
• Reference category: n/a – continuous, and centred at 0 by construction when the generating distribution is standard normal, so the covariate contributes nothing at its own mean.
• Source aliases:
  • COV1 – Lin 2026 Methods 2.1 and the $INPUT column of Data S1/S2; per-subject constant column in the deposited Data S3 example dataset.
• Example models: Lin_2026_sc1cmt_coxTte.R (founding example; enters the Cox hazard ratio as beta1 * SIMCOV_TI with beta1 = 0.3).
• Notes: General scope, ratified 2026-09-02 (PMC13106229) so that future methodology extractions reuse the family rather than minting a per-paper name for a column that is interchangeable by construction. The maintainers considered and rejected source-tied specific-scope names on the AUC_BAST_FW pattern: AUC_BAST_FW is tied to a paper because it still denotes a real physical quantity whose value only makes sense for that hypothetical drug, whereas a meaning-free standard-normal draw carries no paper-specific semantics at all and so cannot collide. Paired with SIMCOV_TV; a study demonstrating more than one covariate of the same role should number them (SIMCOV_TI2, …) rather than overload either name.

SIMCOV_TV (canonical for a meaning-free simulation covariate with a time-varying effect)

• Description: A semantically-empty per-subject covariate used in a simulation or methodology study to demonstrate estimation of a time-varying covariate effect. Carries no clinical interpretation; in the founding example it is a standard-normal draw (mean 0, SD 1) that enters the log hazard multiplied by an explicit function of time.
• Units: (z-score; standard normal, mean 0, SD 1 in the founding example – record the actual generating distribution per model via covariateData[["SIMCOV_TV"]]$units)
• Type: continuous
• Scope: general
• Reference category: n/a – continuous, and centred at 0 by construction when the generating distribution is standard normal.
• Source aliases:
  • DSSTB1 – the $INPUT column name in the Kim 2026 deposited NONMEM control streams, glossed there as “DSST baseline 1h”.
• Example models: Kim_2026_zolpidem_dsst.R, Kim_2026_zolpidem_crt.R, Kim_2026_zolpidem_vas.R, Kim_2026_zolpidem.R, Lin_2026_sc1cmt_coxTte.R (founding example; enters the Cox hazard ratio as beta2 * SIMCOV_TV * log(t + 20) with beta2 = 0.05).
• Notes: Member of the _BL per-subject-baseline suffix family, registered alongside CRT_BL and VAS_SEDATION_BL from the same founding paper. Deliberately NOT named SCORE_DSST_BL: the SCORE_ prefix belongs to the composite clinical rating instruments (SCORE_ADAS_COG, SCORE_MMSE) that are reported as an instrument total without a pre-dose/post-dose baseline distinction, whereas the _BL family is specifically the pre-dose readout entering a PD parameter. Register a separate SCORE_DSST canonical if a future model needs a time-varying (non-baseline) DSST column. Ratified canonically on 2026-09-01 alongside the Kim 2026 zolpidem extraction.

VAS_SEDATION_BL (canonical for per-subject baseline self-rated sedation on a visual analog scale)

• Description: Subject’s pre-dose baseline self-rated sedation / drowsiness marked on a visual analog scale, carried into a model as a time-fixed per-subject covariate. A subjective patient-reported outcome, in contrast to the objective instruments DSST_BL (cognitive) and CRT_BL (psychomotor). Higher values indicate greater perceived sedation.
• Units: mm on a 100 mm scale (document the scale length and anchor wording per-model in covariateData[[VAS_SEDATION_BL]]$notes; a 0-10 cm or 0-100 point scale maps onto the same canonical with units documented per-model).
• Type: continuous
• Scope: general
• Reference category: n/a – enters either as a ratio power term (VAS_SEDATION_BL / <ref>)^e_vas_sedation_bl_<param> or as a centered linear term. Reference value observed: 30.5 mm (Kim 2026 cohort median in 30 healthy Korean volunteers).
• Source aliases:
  • VASB1 – the $INPUT column name in the Kim 2026 deposited NONMEM control streams, glossed there as “VAS baseline 1h”.
• Example models: Kim_2026_zolpidem_vas.R, Kim_2026_zolpidem_dsst.R, Kim_2026_zolpidem_crt.R, Kim_2026_zolpidem.R (founding extraction; screened as a candidate covariate and not retained anywhere, so declared in covariatesDataExcluded in each – in the PK-VAS model the Day -1 baseline trajectory is instead estimated structurally as the four spline knots logitrbase_t1..t4, of which the 1 h knot is the model’s counterpart to this observed column).
• Notes: Member of the _BL per-subject-baseline suffix family. The canonical name identifies the INSTRUMENT (sedation VAS), not merely the scale type, precisely so it cannot collide with the other VAS-scaled entries already in this register that measure different constructs – BLPHYVAS (physician’s global assessment of disease activity) and PGA_PT (patient’s global assessment of arthritis). Any future VAS-based baseline must likewise name its construct (VAS_PAIN_BL, VAS_NAUSEA_BL, …) rather than registering a bare VAS_BL. Ratified canonically on 2026-09-01 alongside the Kim 2026 zolpidem extraction.

AUC_ANDRO (canonical for andrographolide AUC over the 0-4 h post-dose window)

• Description: Per-subject plasma andrographolide (AP1) area under the concentration-time curve over the 0-4 h post-dose window, used as the drug-exposure driver of pharmacodynamic models of standardized Andrographis paniculata preparations. Andrographolide is the labelled reference diterpenoid for A. paniculata products under Thai FDA quality-control requirements and the compound these analyses standardise on, even though the co-occurring diterpenoid 14-deoxy-11,12-didehydroandrographolide (AP3) reaches substantially higher plasma exposure. Time-fixed per subject in the founding example, which pairs one day-5 AUC value with one day-1-to-day-5 viral-load reduction; a longitudinal PK/PD model would instead update it per dosing interval.
• Units: ug*h/L (equivalently ng*h/mL). Must be in the same units as the model’s auc50 so the sigmoid AUC^hill / (auc50^hill + AUC^hill) is dimensionless. Document per-model via covariateData[[AUC_ANDRO]]$units.
• Type: continuous
• Scope: specific
• Reference category: n/a – enters via a sigmoidal Emax form. AUC_ANDRO = 0 makes the sigmoid vanish exactly and leaves the baseline effect. Reference values observed (Songvut 2026 Table 2, 30 mg andrographolide per dose): single-dose day-1 AUC(0-4 h) 28.54 +/- 8.72 ugh/L, AUC(0-inf) 32.81 +/- 8.91 ugh/L; day-5 AUC(0-4 h) under q8h dosing 30.12 +/- 15.83 ugh/L with an observed range of roughly 8-57 ugh/L (Figure 5).
• Source aliases:
  • AUC0-4h / AUC(0-4h, day 5) – printed forms in Songvut 2026 Table 2, Section 2.4 and the Figure 5 x-axis label. The 0-4 h truncation is not arbitrary: plasma andrographolide falls below the 4.69 ng/mL lower limit of quantification by about 4 h post-dose, and the study restricted on-site time for infectious patients, so the window covers essentially the whole quantifiable profile.
• Example models: Songvut_2026_andrographolide.R (drives the four-parameter sigmoidal Emax exposure-response curve for log10 SARS-CoV-2 RdRp viral-load reduction in mild COVID-19; auc50 = 29.80 ug*h/L, hill = 8).
• Notes: Specific scope because the column meaning is tied to andrographolide as the analyte and to the 0-4 h truncated-AUC convention. Member of the AUC_<DRUG> family (AUC_CARBO, AUC_GEM, AUC_GCV, AUC_PAZO, AUC_RTV, AUC_VERUB, AUC_ADU, AUC_DON, AUC_GAN, AUC_LEC, AUC_IBRU, AUC_LCM, AUC_CBZ, AUC_AMPH, AUC_LEN, AUC_EMPA, …), whose entries direct that a new drug or a new exposure convention take a sibling canonical rather than overload an existing one. Accordingly, a future A. paniculata model driven by a different metric (Cmax, AUC(0-inf), a steady-state 0-8 h interval AUC) or by a different diterpenoid should register a parallel canonical – AUC_ANDRO_INF, AUC_AP3 and so on – rather than reuse this name; note in particular that AUC(0-4 h) and AUC(0-inf) are NOT interchangeable here even though they are numerically close, because the 4 h truncation is the convention the founding model’s auc50 was estimated against. There is no published population PK model for andrographolide from this formulation, so downstream users must supply the column from observed concentrations by non-compartmental analysis. Ratified canonically alongside the Songvut 2026 andrographolide exposure-response extraction.

DOSE_APIXABAN_MG (canonical for the administered apixaban dose amount)

• Description: Apixaban dose amount administered on this dose record, in milligrams of apixaban free base. Carried as a model covariate (not merely as the event-table amt) because the source analysis makes relative bioavailability an explicit function of the dose level, so the dose amount has to be readable inside model().
• Units: mg
• Type: continuous
• Scope: specific
• Reference category: n/a – enters a dose-dependent relative-bioavailability function anchored at the lowest studied dose. Reference value observed: 2.5 mg (Cirincione 2018, where Frel is fixed to 1 at 2.5 mg and the 47.5 mg denominator of ((DOSE - 2.5) / 47.5)^gamma makes the estimated I50 the fractional reduction at a 50 mg dose).
• Source aliases:
  • DOSE – the Cirincione 2018 NONMEM control-stream variable used in the $PK branch IF(DOSE.EQ.2.5) THEN F1=1 ELSE F1 = 1 - IMF*((DOSE-2.5)/47.5)**GAM.
• Example models: Cirincione_2018_apixaban.R, Cirincione_2018_apixaban_asian_subgroups.R (drives the less-than-proportional exposure increase the authors attribute to decreased dissolution at higher doses; dose range 2.5 to 50 mg across eight phase I, three phase II and one phase III study. At the 2.5 mg and 5 mg b.i.d. doses approved for NVAF the reduction is small – Frel is 1 and about 0.94 respectively – so this term matters mainly for reproducing the phase I dose-ranging arms).
• Notes: Well-formed member of the auto-approved DOSE_<drug>_<units> family. Distinct from CONMED_APIXABAN_DOSE, which records apixaban as a COMEDICATION arm dose in the Vargo-style anticoagulant dose-response meta-analyses; here apixaban is the analysed drug. Implementation note for simulation: because the covariate is read inside model() it must be supplied as a data column on the dose records, and it must agree with the event table’s amt – a mismatch silently applies the wrong bioavailability to the right dose.

DOSE_APRICOXIB_MG (canonical for administered apricoxib single-dose amount)

• Description: Administered single oral dose of the selective cyclooxygenase-2 inhibitor apricoxib (development code CS-706), in mg, with 0 for a placebo subject. Time-fixed per subject in the founding study (each patient with acute postoperative dental pain received one dose of placebo or 10, 50, 100 or 200 mg). Not a pharmacokinetic covariate – the amount already appears on the depot dose record via amt, and the PK layer’s own dose non-linearities read it from there via podo(depot) and the separate DOSE_HIGH indicator. Required as its own column because Rohatagi 2008 uses the dose as a subject-level regressor in three algebraic pharmacodynamic equations: P(MPR) = P0 + (Pmax - P0) * Dose / (Dose + D50PMPR) (equation 13), log(TMPR / TRescue) = LTR0 + LTRmax * Dose / (Dose + D50LTR) (equation 14) and log(TMPR) = LT0 + LTSlope * Dose (equation 15).
• Units: mg
• Type: continuous
• Scope: specific
• Reference category: n/a – enters uncentred, so Dose = 0 is the placebo reference and returns the published placebo values directly: P(MPR) = P0 = 16 percent and TMPR = exp(LT0) = 1.6 h.
• Source aliases:
  • Dose – used in Rohatagi_2008_apricoxib.R (Rohatagi 2008 equations 13 to 15, “Dose_i is the patient’s dose”; studied levels 0 [placebo], 10, 50, 100 and 200 mg).
• Example models: Rohatagi_2008_apricoxib.R (drives the saturable meaningful-pain-relief probability of equation 13, the saturable onset-time ratio of equation 14, and the log-linear onset time of equation 15).
• Notes: Well-formed member of the auto-approved DOSE_<DRUG>_<UNITS> family (siblings: DOSE_CIPARGAMIN_MG, DOSE_CABAMIQUINE_MG, DOSE_TBPPI_MG, DOSE_RIMEGEPANT_MG). A dedicated column rather than the generic DOSE canonical for two reasons: DOSE is consumed by etTrans() and is not visible inside model(), and a placebo subject has no dose record at all, so podo(depot) cannot distinguish “placebo” from “before the first dose” while an explicit 0 in this column can. APRICOXIB is the INN of the molecule the source papers name only by the Sankyo development code CS-706, per the library’s generic-name-over-development-code convention. Distinct from DOSE_HIGH, which is the Kastrissios 2006 supratherapeutic-cohort indicator gating the two-typical-value apparent oral clearance and is always 0 in the Rohatagi 2008 dental-pain cohort. Populate this column with the same amount the depot dose record carries.

STUDY_NO16853 (canonical for the Bruno 2012 metastatic-breast-cancer study NO16853 cohort indicator)

• Description: Binary indicator for the randomized phase II noninferiority study NO16853 in the Bruno 2012 pooled metastatic-breast-cancer analysis: 1 = NO16853, 0 = the pivotal phase III study SO14999. Time-fixed per subject. Study NO16853 compared capecitabine 825 mg/m^2 twice daily with the registered 1,250 mg/m^2 dose, both plus docetaxel 75 mg/m^2; SO14999 established the registered dose against single-agent docetaxel roughly a decade earlier.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (study SO14999).
• Source aliases:
  • STUD – Bruno 2012 NONMEM $INPUT column; the control stream tests IF(STUD.EQ.2), i.e. STUD = 2 is NO16853 and STUD = 1 is SO14999.
  • Study effect – Bruno 2012 Table 1 and Table 2 covariate row label.
• Example models: Bruno_2012_capecitabine_docetaxel_tgi.R (selects the study-specific additive residual error SD, sqrt(332) = 18.22 mm for SO14999 versus sqrt(112) = 10.58 mm for NO16853), Bruno_2012_capecitabine_docetaxel_os.R, Bruno_2012_capecitabine_docetaxel_pfs.R (log-normal accelerated-failure-time effects on log median OS and PFS, entered on Bruno 2012’s 1 / 2 code as e_studyno16853_tmed * (1 + STUDY_NO16853)).
• Notes: The indicator carries three distinct effects across the paper’s three models, which is why it is a single column rather than three. Bruno 2012 attributes the longer survival in NO16853 to a change in the standard of care between the two studies (the prognostic factors in the model could not explain it) and the longer PFS partly to the different progression-assessment criteria – WHO in SO14999, RECIST in NO16853 – which is also the stated reason for the larger residual error in the older study. The clinical trial simulations in the paper were conditioned on NO16853, so STUDY_NO16853 = 1 is the natural setting when re-running the published framework.

CONC_SOTRO_168H (canonical for the individual sotrovimab serum concentration at the 168-hour landmark from an upstream population PK fit)

• Description: Individual serum concentration of the anti-SARS-CoV-2 monoclonal antibody sotrovimab at the fixed 168 h (7 day) landmark after a single intravenous or intramuscular dose, carried as a per-subject exposure metric that drives a downstream exposure-response model. Not an observation: Sager 2023 integrated each patient’s predicted concentration-time profile from the final population PK model using that patient’s own empirical Bayes parameter estimates, then read the profile at the nominal 168 h time point. Member of the CONC_ family, in the drug-and-landmark-specific form CONC_<ANALYTE>_<LANDMARK> – the analyte and the landmark hour are both carried in the name rather than out-of-band in per-model notes, which is the ambiguity the family exists to avoid. Distinct from the in-vitro applied-concentration siblings CONC_<DRUG>_MGL (a static experimental concentration in an assay medium) in being a model-predicted in-vivo quantity at a nominal clinical time. Also deliberately not a member of the CSS_<DRUG> family (CSS_RBV, CSS_ALBIFN, CSS_NEMTA, CSS_DFO, CSS_CVC), whose members are all averaged steady-state or on-treatment exposures under repeated dosing: this is a single-dose concentration at one instant, so no averaging and no steady state is involved. And distinct from the CP_<DRUG>_<UNITS> family, whose members are instantaneous and time-varying, fed per event record; CONC_SOTRO_168H is time-fixed, one value per subject.
• Units: ug/mL
• Type: continuous
• Scope: specific
• Reference category: n/a – enters the exposure-response model as an UNCENTRED linear term on the logit scale, e_conc_sotro_168h_logitprog * CONC_SOTRO_168H, so the intercept is the log-odds of progression at zero sotrovimab exposure. That is why Sager 2023 labels the intercept the “model-estimated placebo response” even though the source study COMET-TAIL had no placebo arm. Reference values observed (Sager 2023 Table S10, reproduced in the vignette): roughly 60 ug/mL after 500 mg i.v., 21 ug/mL after 500 mg i.m. and 11 ug/mL after 250 mg i.m.
• Source aliases:
  • CP168 – Sager 2023 $INPUT column name in MODEL CODE S2 of the Supporting Information; the paper’s prose and Table 3 write it as “C 168h” / “concentration at 168 h”.
• Example models: Sager_2023_sotrovimab_progression.R (the sole exposure term of the COVID-19 progression logistic regression; logit slope -0.02037 per ug/mL, Table 3), Sager_2023_sotrovimab.R.
• Notes: Specific scope because the concentration scale and the 168 h landmark are both tied to this analysis. Sager 2023 screened a family of alternative exposure metrics for the same endpoint – predicted concentrations at 24, 48, 72, 96 and 168 h, the corresponding partial AUCs, and AUC through day 28 – and found only AUC(0-day28), C(96h) and C(168h) to be statistically significant. C(168h) was carried into the final model because the median time from randomization to a progression event was 5.5 days, making a one-week exposure the clinically meaningful one; the C(96h) base model was rejected because it under-predicted the observed progression rate in the 250 mg intramuscular arm. Do not substitute a 96 h concentration, an AUC or a Cmax into this column: the slope is specific to the 168 h landmark. A future analysis reporting an exposure-response on a different landmark for the same drug registers a sibling CONC_SOTRO_<LANDMARK> rather than overloading this entry.

NRISK_GT1 (canonical for the more-than-one-protocol-defined-risk-factor stratum indicator)

• Description: 1 = the subject carries more than one protocol-defined risk factor for the trial’s progression / deterioration endpoint; 0 = one or fewer. Subject-level baseline indicator, formed by dichotomising an integer count of risk conditions at a cut point of 1. The name states the quantity – a count of risk factors, dichotomised – rather than any one source column’s spelling, and is deliberately NOT scoped to a disease: a “more than one risk factor” stratum recurs across indications, and burning the founding study’s disease into the name would wrongly imply the concept is specific to it. What IS study-specific is the content of the risk-factor list, which each model must document in its own covariateData[[NRISK_GT1]]$notes.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (one or fewer risk factors). In the founding study (Sager 2023, COMET-TAIL) the split is 630 of 902 (69.8%) at the reference and 272 of 902 (30.2%) carrying the indicator.
• Source aliases:
  • RISKCATN – Sager 2023 $INPUT column in MODEL CODE S2, tagged under ;--catcovs-; the dichotomised form of the underlying risk-condition count RFNUMCND. The paper’s prose and Table 3 write it as “number of risk factors (<=1 vs. >1)” and “RISKCATN = 1”.
• Example models: Sager_2023_sotrovimab_progression.R (additive shift of +1.887 on the progression logit intercept; raises the model-estimated zero-exposure progression probability from 1.52% to 9.25%, with no effect on the exposure slope).
• Notes: Specific scope for now because only one paper uses it; promotion to general scope is appropriate after a second corroborating paper, and the disease-neutral name is chosen precisely so that promotion needs no rename. Each model’s covariateData[[NRISK_GT1]]$notes MUST state the exact risk-factor list and cut point its protocol used, because those differ between trials – in Sager 2023 the count is INCLUSIVE of age and body mass index (source flags RFAGE55N for age >= 55 years and RFBMI30N for BMI > 30 kg/m^2, alongside diabetes, chronic kidney disease and congestive heart failure), which is what separates it from the separately screened “number of other risk factors” column NOTHRISK that excludes them. The direction of coding is settled by three independent lines in the source and should not be reversed: the shift is positive and therefore must attach to the higher-risk group; mixing the two final intercepts over the published 69.8% / 30.2% split at zero exposure reproduces the single-intercept base model of Table S9 (0.0386 against 0.0461) whereas the reversed coding does not (0.0692); and Sager 2023 Results describe the covariate as acting on the “model-estimated placebo response”. Distinct from MORTRISK_HIGH, which is a paediatric severe-malnutrition mortality-risk composite scored by the Berkley 2003 criteria, and from ORG_FAIL_COUNT, which is retained as an integer count rather than dichotomised.

Molecule-level physicochemical / developability attributes

STUDY_EMPA_A (canonical for Riggs empagliflozin study A cohort indicator)

• Description: 1 = patient enrolled in study A of the pooled empagliflozin analyses of Riggs 2013 / Riggs 2014 (EudraCT 2007-000654-32; phase I, randomised, double-blind, placebo-controlled, 8 days, 2.5 / 10 / 25 / 100 mg once daily or placebo, n = 48, Germany); 0 = otherwise. Subject-level (time-fixed).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0. One of the mutually exclusive five-member set STUDY_EMPA_A.. STUDY_EMPA_E; exactly one is 1 per subject.
• Source aliases:
  • Study A – the paper-text label used throughout Riggs 2014 Methods, Table 1 and Table 2.
• Example models: Riggs_2014_empagliflozin.R (selects the typical baseline FPG 7.85 mmol/L, Table 2 theta_1, and is the reference level for the baseline-UGE, gamma_base, Umax, Ustim50 and C*50 parameters).
• Notes: Follows the auto-approved STUDY_<id> canonical family. The five members are named for the drug rather than for the letter alone because the letters A-E are generic and would collide with any other paper that labels its cohorts the same way. Study A contributed no HbA1c to the Riggs 2014 exposure-response analysis (Table 1 footnote), so no baseline-HbA1c theta exists for it. The identical five-study set underlies the companion population PK analysis (Riggs 2013 J Clin Pharmacol 53:1028-1038), so a future extraction of that paper should reuse these five canonicals rather than register siblings.

STUDY_EMPA_B (canonical for Riggs empagliflozin study B cohort indicator)

• Description: 1 = patient enrolled in study B of the pooled empagliflozin analyses of Riggs 2013 / Riggs 2014 (NCT00558571; phase I, randomised, double-blind, placebo-controlled, parallel-group, 4 weeks, 10 / 25 / 100 mg once daily or placebo, n = 78, Germany); 0 = otherwise. Subject-level (time-fixed).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0. One of the mutually exclusive five-member set STUDY_EMPA_A.. STUDY_EMPA_E.
• Source aliases:
  • Study B – the paper-text label used throughout Riggs 2014 Methods, Table 1 and Table 2.
• Example models: Riggs_2014_empagliflozin.R (selects baseline FPG 8.50 mmol/L, Table 2 theta_2, and baseline HbA1c 7.18 pct, theta_23, and scales the baseline UGE by theta_20 = 0.320; shares the reference gamma_base, Umax, Ustim50 and C*50 with study A).
• Notes: Follows the auto-approved STUDY_<id> canonical family. See [[STUDY_EMPA_A]] for the naming rationale and the shared five-study set.

STUDY_EMPA_C (canonical for Riggs empagliflozin study C cohort indicator)

• Description: 1 = patient enrolled in study C of the pooled empagliflozin analyses of Riggs 2013 / Riggs 2014 (NCT00885118; phase II, randomised, double-blind, placebo-controlled, parallel-group, 4 weeks, 1 / 5 / 10 / 25 mg once daily or placebo, n = 100, Japan, Japanese patients only); 0 = otherwise. Subject-level (time-fixed).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0. One of the mutually exclusive five-member set STUDY_EMPA_A.. STUDY_EMPA_E.
• Source aliases:
  • Study C – the paper-text label used throughout Riggs 2014 Methods, Table 1 and Table 2.
• Example models: Riggs_2014_empagliflozin.R (the only study with its own value for every urinary-glucose-excretion and potency parameter: baseline UGE x theta_16 = 0.632, gamma_base x theta_17 = 1.16, Umax x theta_14 = 1.11, Ustim50 x theta_15 = 1.58 and C50 x theta_19 = 0.169, giving a study C AUC50 of 106 nmolh/L against the 626 nmol*h/L reference; also selects baseline FPG 8.76 mmol/L and baseline HbA1c 7.85 pct).
• Notes: Follows the auto-approved STUDY_<id> canonical family. This is the load-bearing member of the set: it is the only Japanese cohort and the only one whose potency parameters all differ from the pooled reference. Riggs 2014 Discussion attributes the greater low-dose response in study C partly to the lower body weight of the Japanese cohort (mean 67.9 kg against 81.4 - 94.6 kg elsewhere) and therefore its higher dose-normalised exposure, but notes that confounding precludes an exact explanation – so the indicator is a cohort marker standing in for a partly unmeasured population difference rather than a study-conduct artefact. See [[STUDY_EMPA_A]] for the naming rationale.

STUDY_EMPA_D (canonical for Riggs empagliflozin study D cohort indicator)

• Description: 1 = patient enrolled in study D of the pooled empagliflozin analyses of Riggs 2013 / Riggs 2014 (NCT00789035; phase IIb, randomised, double-blind, 12 weeks, 5 / 10 / 25 mg once daily, placebo or open-label metformin, multinational; n = 324 contributed to the exposure-response analyses after excluding the open-label metformin arm and two patients without PK samples); 0 = otherwise. Subject-level (time-fixed).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0. One of the mutually exclusive five-member set STUDY_EMPA_A.. STUDY_EMPA_E.
• Source aliases:
  • Study D – the paper-text label used throughout Riggs 2014 Methods, Table 1 and Table 2.
• Example models: Riggs_2014_empagliflozin.R (selects baseline FPG 9.30 mmol/L, Table 2 theta_3, and baseline HbA1c 7.85 pct, theta_24; shares the reference C50 with studies A and B, whose pooled estimate is the primary AUC50 = 626 nmolh/L).
• Notes: Follows the auto-approved STUDY_<id> canonical family. Study D contributed no urinary-glucose-excretion observations, so the UGE parameters are not identified for it and fall back to the study A reference. Study D was empagliflozin monotherapy whereas study E was on background metformin; Riggs 2014 Discussion cites that difference as one candidate explanation for the ~2-fold difference between their AUC50 estimates. See [[STUDY_EMPA_A]] for the naming rationale.

STUDY_EMPA_E (canonical for Riggs empagliflozin study E cohort indicator)

• Description: 1 = patient enrolled in study E of the pooled empagliflozin analyses of Riggs 2013 / Riggs 2014 (NCT00749190; phase IIb, randomised, double-blind, parallel-group, 12 weeks, 1 / 5 / 10 / 25 / 50 mg once daily, placebo or open-label sitagliptin, on background metformin, multinational, n = 424 after excluding the open-label sitagliptin arm); 0 = otherwise. Subject-level (time-fixed).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0. One of the mutually exclusive five-member set STUDY_EMPA_A.. STUDY_EMPA_E.
• Source aliases:
  • Study E – the paper-text label used throughout Riggs 2014 Methods, Table 1 and Table 2.
• Example models: Riggs_2014_empagliflozin.R (selects baseline FPG 9.49 mmol/L, Table 2 theta_4, and baseline HbA1c 7.89 pct, theta_25, and scales C50 by theta_21 = 1.93, giving the study E AUC50 of 1210 nmolh/L against the 626 nmol*h/L reference).
• Notes: Follows the auto-approved STUDY_<id> canonical family. Study E contributed no urinary-glucose-excretion observations, so the UGE parameters fall back to the study A reference. Its roughly two-fold higher AUC50 is the single largest inter-study difference in the Riggs 2014 model and drives the paper’s sensitivity analysis: the same 10 and 25 mg doses give 80 / 90 pct of maximal response on the pooled A + B + D estimate but only 65 / 82 pct on the study E estimate. See [[STUDY_EMPA_A]] for the naming rationale.

STUDY_PROACTIVE (canonical for the PROACTIVE olaparib-boosting trial cohort indicator)

• Description: 1 = subject enrolled in PROACTIVE (NCT05078671), the randomised cross-over trial of reduced-dose olaparib 100 mg twice daily boosted with cobicistat 150 mg twice daily against olaparib 300 mg twice daily monotherapy in patients with solid tumours; 0 = subject enrolled in one of the other three cobicistat studies pooled into the same analysis (DATE-4, PANNA or PRACTICAL, all cobicistat 150 mg once daily). Subject-level (time-fixed).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0, which pools DATE-4 (healthy volunteers boosting atazanavir), PANNA (postpartum women with HIV boosting elvitegravir) and PRACTICAL (rheumatoid arthritis, boosting tofacitinib) into a single reference stratum.
• Source aliases:
  • OLAP – the NONMEM data item in Overbeek 2025 Online Resource Material 1 ($INPUT ID TIME EVID DV AMT WEIGHT ELVI ATAZ TOFA OLAP MDV CMT DROP II SS), applied as CLINT=(THETA(3)*VL)*(THETA(4)**OLAP)*EXP(ETA(4)).
  • PROACTIVE – the study acronym used in Overbeek 2025 Table 1, Fig. 2 and Fig. 3.
• Example models: Overbeek_2025_cobicistat.R (multiplicative power-form effect raising cobicistat unbound intrinsic clearance 1.21-fold, the only one of four candidate study indicators retained in the final model).
• Notes: Well-formed member of the auto-approved STUDY_<id> canonical family, named for the trial acronym in keeping with STUDY_NOVA / STUDY_QUADRA / STUDY_ODYSSEY. Overbeek 2025 tested a binary indicator for each of its four studies on both prehepatic bioavailability and intrinsic clearance (Eq 7); all four improved the fit univariately and this one was retained. Read it as a cohort effect, not as a drug-drug interaction. Although Table 2 labels the coefficient CLint-olaparib and the data item is named OLAP, the paper’s Discussion attributes the higher intrinsic clearance to selection of a population with relatively high CYP3A activity – the same patients had unusually low olaparib exposure at the standard dose – and the indicator is simultaneously confounded with the twice-daily cobicistat regimen and the solid-tumour population. A model needing a genuine “olaparib is coadministered” perpetrator flag should use a CONMED_<INN> column instead; the sibling model Overbeek_2025_olaparib.R uses [[CONMED_COBICISTAT]] for the reciprocal effect of cobicistat on olaparib, which is a drug-drug interaction. The two columns are therefore not mirror images of one another and must not be collapsed.

CRT_BL (canonical for per-subject baseline choice reaction time)

• Description: Subject’s pre-dose baseline choice reaction time (CRT), a psychomotor latency measured before any study drug is given and carried into the model as a time-fixed per-subject covariate. Choice reaction time is the interval between a stimulus and the correct discriminative motor response; slower (larger) values indicate poorer baseline psychomotor function. Used in CNS-depressant PD models where a subject’s intrinsic psychomotor speed modulates drug sensitivity, entering the potency and/or steepness parameters as a ratio-normalised power term or a centered linear term. Distinct from the model’s own estimated lrbase / BASE parameter: CRT_BL is an observed data column, whereas BASE is a structural parameter of the same endpoint.
• Units: msec (document any per-model deviation in covariateData[[CRT_BL]]$notes).
• Type: continuous
• Scope: general
• Reference category: n/a – enters either as a ratio power term (CRT_BL / <ref>)^e_crt_bl_<param> or as a centered linear term 1 + e_crt_bl_<param> * (CRT_BL - <ref>). Reference value observed: 438.5 msec (Kim 2026 cohort median in 30 healthy Korean volunteers).
• Source aliases:
  • CRTB1 – the $INPUT column name in the Kim 2026 deposited NONMEM control streams, glossed there as “CRT baseline 1h” because the study’s baseline battery was run at 1 h on Day -1.
• Example models: Kim_2026_zolpidem_crt.R (founding example; retained on BOTH EC50, as the power term (CRT_BL/438.5)^-1.28, and HILL, as the linear term 1 + 0.00637 * (CRT_BL - 438.5)), Kim_2026_zolpidem.R, Kim_2026_zolpidem_dsst.R, Kim_2026_zolpidem_vas.R (screened but not retained; declared in covariatesDataExcluded to preserve the screen).
• Notes: Member of the _BL per-subject-baseline suffix family, on the QTC_BL precedent – a derived per-subject instrument readout taken before dosing that enters PD parameters as a centred or ratio-normalised term. The clock time at which the baseline battery was administered is study-specific (1 h on Day -1 in the founding example) and belongs in the per-model covariateData[[CRT_BL]]$notes, not in the canonical name; do NOT register parallel canonicals such as CRT_BL_1H. The name records the instrument, not the units, so a study reporting CRT in seconds uses the same canonical with units documented per-model.

IPSS_BL (canonical for per-subject baseline International Prostate Symptom Score)

• Description: Subject’s observed International Prostate Symptom Score (IPSS; the sum of the seven symptom questions, 0-35, excluding the separate quality-of-life item) at the baseline / randomisation visit, carried into a model as a time-fixed per-subject covariate. Higher scores indicate more severe lower urinary tract symptoms (0-7 mild, 8-19 moderate, 20-35 severe). Used in BPH / LUTS drug-disease models both as the initial condition of the modelled ipss state and as a covariate on disease-progression parameters. Distinct from the screening IPSS, which precedes a placebo run-in and is a different quantity (regression to the mean between screening and baseline).
• Units: IPSS points (0-35)
• Type: continuous
• Scope: general
• Reference category: n/a – enters as a centred linear term 1 + e_ipss_bl_<param> * (IPSS_BL - <ref>). Reference value observed: 16 points (D’Agate 2020 and D’Agate 2021 pooled-population median).
• Source aliases:
  • IPSS0 – the $INPUT column in the D’Agate 2020 Supporting Information control stream (A_0(1) = obsIPSS, obsIPSS = IPSS0).
  • IPSSb, B_VARIABLE – D’Agate 2020 Table 2 and Figure 2 labels.
• Example models: DAgate_2020_bph_ipss_mbma.R (founding example; initial condition ipss(0) <- IPSS_BL, and linear effect on the disease progression rate DISP * (1 + 0.027 * (IPSS_BL - 16))), DAgate_2021_bph_aurs_mbma.R (log-linear effect on the AUR/S hazard, exp(log(1.04) * (IPSS_BL - 16))).
• Notes: Member of the _BL per-subject-baseline instrument-readout family, on the QTC_BL / CRT_BL / DSST_BL precedent: an observed pre-treatment readout of the same instrument the model predicts, used as a data column rather than an estimated baseline parameter. A model that instead ESTIMATES the baseline IPSS should use an lrbase-style parameter, not this column. A screening-visit IPSS, if ever needed as a covariate, must be a separate canonical (e.g. IPSS_SCR), never overloaded onto this one.

DSST_BL (canonical for per-subject baseline digit symbol substitution test score)

• Description: Subject’s pre-dose baseline score on the digit symbol substitution test (DSST), a pencil-and-paper or computerised cognitive test of motor speed, attention, working memory and visuoperceptual ability, scored as the number of correct substitutions completed in a fixed interval. Higher scores indicate better cognitive performance – the opposite polarity to CRT_BL, where higher is worse. Carried into a model as a time-fixed per-subject covariate on PD parameters. Distinct from the model’s own estimated lrbase / BASE parameter for the same endpoint.
• Units: score (number of correct substitutions; document the test duration and any per-model scaling in covariateData[[DSST_BL]]$notes).
• Type: continuous
• Scope: general
• Reference category: n/a – enters either as a ratio power term (DSST_BL / <ref>)^e_dsst_bl_<param> or as a centered linear term 1 + e_dsst_bl_<param> * (DSST_BL - <ref>). Reference value observed: 72.5 score (Kim 2026 cohort median in 30 healthy Korean volunteers).
• Source aliases:
  • DSSTB1 – the $INPUT column name in the Kim 2026 deposited NONMEM control streams, glossed there as “DSST baseline 1h”.
• Example models: Kim_2026_zolpidem_dsst.R, Kim_2026_zolpidem_crt.R, Kim_2026_zolpidem_vas.R, Kim_2026_zolpidem.R (founding extraction; screened as a candidate covariate on the PK and all three PD endpoints and not retained on any, so declared in covariatesDataExcluded in each – the DSST baseline is instead estimated structurally as the BASE parameter with age as its covariate).
• Notes: Member of the _BL per-subject-baseline suffix family, registered alongside CRT_BL and VAS_SEDATION_BL from the same founding paper. Deliberately NOT named SCORE_DSST_BL: the SCORE_ prefix belongs to the composite clinical rating instruments (SCORE_ADAS_COG, SCORE_MMSE) that are reported as an instrument total without a pre-dose/post-dose baseline distinction, whereas the _BL family is specifically the pre-dose readout entering a PD parameter. Register a separate SCORE_DSST canonical if a future model needs a time-varying (non-baseline) DSST column.

CONMED_QUINIDINE (canonical for concomitant quinidine (P-glycoprotein inhibitor) coadministration indicator)

• Description: 1 = subject coadministered quinidine during the observation interval, 0 = no concomitant quinidine. Quinidine is the classical probe P-glycoprotein inhibitor used in dedicated DDI studies to isolate a P-gp contribution from a CYP3A4 contribution; it is also a strong CYP2D6 inhibitor, so a paper using quinidine as a CYP2D6 probe should document which mechanism its effect is attributed to in the per-model notes. Distinct from the class-level [[CONMED_PGP_INH]] indicator: use this canonical when a paper singles quinidine out by name. Per-subject time-varying when a DDI study spans on / off quinidine periods; time-fixed on the on-treatment day of a fixed-sequence DDI arm.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (no concomitant quinidine).
• Source aliases:
  • Quinidine – used in Schlachter_2026_atogepant.R (Schlachter 2026 Table 1 cotreatment row, 25 of 1356 subjects; the dedicated phase 1 study 3101-103-002, Table S1).
• Example models: Schlachter_2026_atogepant.R (single multiplicative fractional effect of -0.285 on apparent clearance per Schlachter 2026 Table 2 and the Section 3.1 CL/F display equation, a 28.5% reduction that the Discussion attributes to P-gp inhibition. Unlike the itraconazole and rifampicin arms of the same model, quinidine carries no effect on relative bioavailability, which is the signature that distinguishes a pure transporter-inhibition DDI from one with a gut-CYP3A4 first-pass component).
• Notes: Auto-approved member of the CONMED_<INN> family (INN = quinidine). Sibling of [[CONMED_ITRACONAZOLE]] and [[CONMED_RIFAMPICIN]]. Do NOT confuse with [[FORM_QUIN_SR]], which is a quinidine formulation indicator in a model where quinidine is the victim drug rather than the perpetrator. Per-model notes should record the quinidine regimen, since P-gp inhibition is concentration-dependent.

CONMED_RIFAMPICIN_SD, CONMED_RIFAMPICIN_MD (canonical for the single-dose and multiple-dose concomitant rifampicin states, carried as separate indicators)

• Description: A mutually exclusive pair splitting concomitant rifampicin coadministration into its two mechanistically opposite phases. CONMED_RIFAMPICIN_SD = 1 on records where a single rifampicin dose has been given, the phase dominated by acute OATP1B1 (and OATP1B3) inhibition, which raises victim exposure. CONMED_RIFAMPICIN_MD = 1 on records after multiple rifampicin doses, the phase dominated by CYP3A4 (and P-gp) induction, which lowers victim exposure. Both are 0 when no rifampicin is given. Per-dose-record rather than per-subject: a subject in a fixed-sequence DDI study carries _SD on the single-dose occasion and _MD on the multiple-dose occasion. A record with both set to 1 is not a valid state.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 for both (no concomitant rifampicin).
• Source aliases:
  • Rifampin after first dose / Rifampin following multiple doses – the Schlachter 2026 Table 2 row labels (“rifampin” is the US adopted name for the INN rifampicin).
• Example models: Schlachter_2026_atogepant.R (founding example; four effects in total, two per phase, per Schlachter 2026 Table 2 and the Section 3.1 display equations. Single dose: CL/F x (1 - 0.128) and Frel x (1 + 1.42), i.e. 12.8% lower clearance and a 2.4-fold higher relative bioavailability. Multiple doses: CL/F x (1 + 0.818) and Frel x (1 - 0.248), i.e. a 1.82-fold higher clearance and a 24.8% lower relative bioavailability. The sign reversal on both parameters between the two phases is exactly what makes a single pooled indicator unable to express the model), Scotcher_2021_gadoxetate_rat_pbpk.R (preclinical rat PBPK; CONMED_RIFAMPICIN_SD = 1 for gadoxetate given 1 h after a single 10 mg/kg intravenous rifampicin dose, selecting the separately estimated rifampicin-phase active hepatic uptake clearance 0.095 L/h (control 2.38 L/h, a 96% OATP inhibition) and biliary clearance 0.08 L/h (control 0.07 L/h) of Scotcher 2021 Table 4; _MD is not used).
• Notes: Registered as a pair rather than as qualifiers on a single indicator because the two phases act in opposite directions: pooling them into one [[CONMED_RIFAMPICIN]] column would average a 1.82-fold induction against a 2.4-fold inhibition and destroy both effects. Use the unqualified [[CONMED_RIFAMPICIN]] when a paper reports only the steady-state induction effect (the common case) and reserve this pair for papers that resolve the acute inhibition phase separately, which requires a dedicated single-dose-plus-multiple-dose DDI design. Shape precedent: the [[CONMED_PANCRELIPASE_PRE30]] / _POST30 / _POST60 dose-timing qualifier family. The INN spelling RIFAMPICIN is used in preference to the US rifampin for consistency with the existing [[CONMED_RIFAMPICIN]] entry.

DOSE_ATOGEPANT_MG (canonical for the administered atogepant dose level)

• Description: Per-dose-record administered atogepant dose amount in mg, carried as a covariate column in its own right so that it can drive dose-dependent absorption and bioavailability terms. Continuous; studied over 10 to 300 mg.
• Units: mg
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as the ratio power terms (DOSE_ATOGEPANT_MG / 60)^e_dose_atogepant_mg_<param>. Reference dose 60 mg, the marketed once-daily dose.
• Source aliases:
  • Dose – the Schlachter 2026 Table 2 covariate rows “Exponential dose effect on Frel” and “Exponential dose effect on ka”.
• Example models: Schlachter_2026_atogepant.R (two power-model effects per the Schlachter 2026 Section 3.1 display equations: Frel = (dose/60 mg)^0.119, reproducing the paper’s statement that Frel is approximately 1.24-fold higher at 60 mg than at 10 mg since (60/10)^0.119 = 1.24, and Tk0 = 0.908 * (dose/60 mg)^0.199. The second estimate appears in Table 2 under the label “Exponential dose effect on ka” because the model derives ka = Fk0 / (Tk0 * (1 - Fk0)), so a covariate applied to the zero-order duration propagates into the first-order rate constant; the display equation shows it applied to Tk0).
• Notes: Auto-approved member of the DOSE_<drug>_<units> family. As with every member of that family, the column must not be named DOSE: a covariate column literally named DOSE (any casing) is consumed by rxode2’s etTrans() and never reaches model(), so the dose amount must be supplied both as the event-table amt and as this separately-named covariate column.

DOSE_NICOTINE_MG (canonical for the nominal nicotine dose of the product used on an occasion)

• Description: Per-dose-record nominal nicotine content of the product used, in mg. For inhaled products this is the nominal nicotine ISO yield (0.5 mg for the Tobacco Heating System, 0.1-1.5 mg for conventional cigarettes depending on the individual brand); for nicotine replacement products it is the labelled nicotine content (1 mg for the regular nasal spray, 2 mg for the mentholated gum). Continuous, and time-varying in crossover designs where a subject uses more than one product. It is emphatically NOT the systemically available amount: how much of the nominal content is actually released and absorbed is exactly what the relative-bioavailability term it drives is there to estimate.
• Units: mg
• Type: continuous
• Scope: specific
• Reference category: n/a – enters as the ratio power term (DOSE_NICOTINE_MG / 0.5)^e_dose_nicotine_mg_frel. The reference 0.5 mg is the nominal nicotine ISO yield of the Tobacco Heating System, the reference product.
• Source aliases:
  • nicotine ISO yield / nicotine dose – Marchand 2017 (paper Sect. 2.2 dataset composition, Suppl. Table 1 primary-covariate list, and the Table 4 Frel equation, which labels the term “nicotine ISO yield” even though the same column carries the labelled nicotine content of the non-inhaled products).
• Example models: Marchand_2017_nicotine.R (power effect on relative bioavailability, Frel = (DOSE_NICOTINE_MG / 0.5)^-0.573 * ..., so a doubling of the nominal nicotine dose reduces Frel by 33 percent; Marchand 2017 Table 4 and the Sect. 4 self-titration rationale), OlssonGisleskog_2021_nicotine_mouthspray.R (power effect on the fraction of the spray dose swallowed, (DOSE_NICOTINE_MG / 2)^0.0928; the source column is the actual dose ADOSE, equal to the nominal dose for a spray), OlssonGisleskog_2021_nicotine_gum.R (power effect (DOSE_NICOTINE_MG / 2)^0.507 on the oromucosal ka, and the denominator of the release-rate calculation with DOSE_NICOTINE_RELEASED_MG), OlssonGisleskog_2021_nicotine_lozenge.R (a 2-mg lozenge shifts the logit of the fraction swallowed by -0.217 relative to 4 mg).
• Notes: Auto-approved member of the DOSE_<drug>_<units> family. As with every member, the column must not be named DOSE: a covariate column literally named DOSE (any casing) is consumed by rxode2’s etTrans() and never reaches model(), so the nominal dose must be supplied both as the event-table amt and as this separately-named covariate column. The value on a record must equal that record’s amt when the model is simulated in mg, because the same number is doing two jobs – setting the administered amount and indexing the dose-dependent bioavailability. Distinct from the product-type indicators FORM_NICOTINE_CC / FORM_NICOTINE_NNS / FORM_NICOTINE_GUM, which carry the product’s identity rather than its nicotine content; Marchand 2017 fits both simultaneously, so nasal spray is not merely “a 1 mg product” but also carries its own bioavailability offset.

DOSE_NICOTINE_RELEASED_MG (canonical for the individual amount of nicotine released from a chewing gum)

• Description: Per-dose-record amount of nicotine, in mg, actually released from a nicotine chewing gum during use, measured as the nominal content minus the residual nicotine assayed in the chewed gum. Distinct from the nominal content DOSE_NICOTINE_MG: typically 64-79% of it.
• Units: mg
• Type: continuous
• Scope: specific
• Reference category: n/a – continuous; enters through the ratio to DOSE_NICOTINE_MG.
• Source aliases:
  • ADOSE – Olsson Gisleskog 2021 ESM gum control stream and dataset (Sect. 2.2: ‘the actual dose … calculated as the difference between the average nicotine content in the formulation batch and the residual nicotine after use’).
• Example models: OlssonGisleskog_2021_nicotine_gum.R (sets the first-order release rate from the gum, Krel = -log(1 - DOSE_NICOTINE_RELEASED_MG / DOSE_NICOTINE_MG) / 0.5, so exactly this amount leaves the gum in the 0.5-h chewing period; Eq. 1).
• Notes: Auto-approved member of the DOSE_<drug>_<units> family with a qualifier, like DOSE_TAF_MG (amount delivered in vivo by an implant). The dose records still carry the nominal content as amt; this column only drives the release rate.

FORM_ATOGEPANT_EARLYTAB (canonical for the early phase 1 atogepant tablet formulation indicator)

• Description: 1 = dose administered as one of the early phase 1 atogepant tablet presentations, 0 = the Formulation 5 tablet used in the phase 2b/3 and phase 3 studies and carried into the marketed product. Per-dose-record indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (the Formulation 5 tablet), which supplies 1274 of 1356 subjects (94.0%) and 100% of the phase 2b/3, phase 3 and both external-validation cohorts (Schlachter 2026 Table S2).
• Source aliases:
  • Formulation – the Schlachter 2026 Table S3 covariate row and the Table 2 row “Formulation 2 tablet effect on ka/formulation 4 tablet effect on ka”.
• Example models: Schlachter_2026_atogepant.R (founding example; single multiplicative fractional effect of -0.353 on the zero-order absorption duration per the Schlachter 2026 Section 3.1 display equation, a 35% shorter Tk0, which propagates into the derived ka and is therefore labelled an effect “on ka” in Table 2).
• Notes: Auto-approved member of the FORM_<drug>_<formulation> family. Registered as a single aggregate early-tablet indicator rather than as per-prototype indicators (the [[FORM_SEP4199_IR]] shape) because the source is internally inconsistent about which early formulations carry the effect: the Section 3.1 display equation says “phase 1 / formulations 3 and 4”, the Section 3.1 prose says “the formulation 1 tablet used in early phase 1 studies”, and the Table 2 row alias says “Formulation 2 tablet / formulation 4 tablet” – a label carried over from the Phase 1 Model column, where two separate estimates of -0.44 and -0.42 sat directly on ka. All three statements agree the effect belongs to the early phase 1 tablets and not to Formulation 5, and Formulations 1 to 4 together are only 82 of 1356 subjects, so the aggregate indicator is the only reading the source supports unambiguously. Register per-prototype siblings if a future source resolves the enumeration.

FORM_ABROCITINIB_SUSP (canonical for the abrocitinib oral suspension formulation indicator)

• Description: 1 = dose given as the abrocitinib oral suspension (drug substance in 0.5% w/v methylcellulose), 0 = a tablet formulation. Per-dose-record indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0. With all three FORM_ABROCITINIB_* indicators at 0 the dose is the phase II 100 mg tablet (D1400040).
• Source aliases:
  • FORMS = 1 – Wojciechowski 2022 Online Resource 2 formulation column (1 = suspension, 2 = phase II 100 mg tablet, 3 = phase II 10 and 50 mg tablets, 4 = phase III 100 mg tablet); the companion column FORM (2 = tablet, 3 = suspension) selects the tablet-only absorption lag.
• Example models: Wojciechowski_2022_abrocitinib.R (founding example; multiplies the first-order absorbed-amount cap amt_fo by 1 + 1.17 and removes the 0.183 h tablet lag).
• Notes: Auto-approved member of the FORM_<drug>_<formulation> family. Kept drug-specific rather than reusing [[FORM_SUSPENSION]] (an extemporaneously prepared suspension of a solid form) because here it is a manufactured development formulation contrasted against three tablet generations.

FORM_ABROCITINIB_TAB_PH2B (canonical for the abrocitinib phase IIb 10 mg and 50 mg tablet formulation indicator)

• Description: 1 = dose given as the abrocitinib 10 mg (D1500091) or 50 mg (D1500093) round film-coated tablets used in the phase IIb atopic-dermatitis study B7451006, 0 = any other formulation. Per-dose-record indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (see [[FORM_ABROCITINIB_SUSP]] for the all-zero reference).
• Source aliases:
  • FORMS = 3 – Wojciechowski 2022 Online Resource 2; Table 2 rows ‘Effect of phase IIb tablets on F’ and ‘… on Ak1’.
• Example models: Wojciechowski_2022_abrocitinib.R (founding example; additive shift of -1.02 on logit absolute bioavailability and the first-order amount cap multiplied by 1 - 0.68).
• Notes: Auto-approved member of the FORM_<drug>_<formulation> family.

FORM_ABROCITINIB_TAB_PH3 (canonical for the abrocitinib phase III 100 mg film-coated tablet formulation indicator)

• Description: 1 = dose given as the abrocitinib 100 mg round film-coated tablet (D1700147) used in the phase III programme and later phase I studies, 0 = any other formulation. Per-dose-record indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (see [[FORM_ABROCITINIB_SUSP]] for the all-zero reference). The source paper’s simulation reference scenario uses this tablet, so set it to 1 to reproduce that scenario.
• Source aliases:
  • FORMS = 4 – Wojciechowski 2022 Online Resource 2; Table 2 row ‘Effect of phase III tablet on F’.
• Example models: Wojciechowski_2022_abrocitinib.R (founding example; additive shift of -0.766 on logit absolute bioavailability).
• Notes: Auto-approved member of the FORM_<drug>_<formulation> family.

STUDY_CGP_PK_02, STUDY_CGP_MD_01, STUDY_ATOGEPANT_PHASE3 (canonical for the three non-reference residual-error study strata of the atogepant population PK analysis)

• Description: Mutually exclusive per-record study indicators selecting among the four study-specific proportional residual error magnitudes of the Schlachter 2026 atogepant Phase 3 Model. STUDY_CGP_PK_02 = the 40-subject phase 1 single-dose itraconazole DDI study, the only contributing study providing both plasma and dried-blood-sample concentrations. STUDY_CGP_MD_01 = the 463-subject phase 2b/3 dose-ranging efficacy study in patients with episodic migraine, which contributed 1431 sparsely sampled dried-blood-sample observations. STUDY_ATOGEPANT_PHASE3 = either pivotal phase 3 study, ADVANCE (3101-301-002, episodic migraine, in the development dataset) or PROGRESS (3101-303-002, chronic migraine, in the external-validation dataset); one indicator covers both because the source reports a single shared estimate for them. When all three are 0 the record falls in the reference stratum, “all phase 1 studies except CGP-PK-02”.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 for all three (any phase 1 study other than CGP-PK-02).
• Source aliases:
  • Study – the Schlachter 2026 Table 2 residual-error rows “sigma prop”, “sigma prop (study CGP-PK-02)”, “sigma prop (study CGP-MD-01)” and “sigma prop (ADVANCE and PROGRESS studies)”.
• Example models: Schlachter_2026_atogepant.R (founding example; selects among proportional residual SDs of 0.307 for the reference stratum, 0.228 for CGP-PK-02, 0.585 for CGP-MD-01 and 0.491 for the phase 3 studies, assembled in model() as a weighted sum of the three indicators and their complement. The ordering tracks sampling richness: the sparsely sampled phase 2b/3 dried-blood-sample study carries the largest residual error and the richly sampled phase 1 DDI study the smallest).
• Notes: Auto-approved members of the STUDY_<id> family, on the [[STUDY_EDELWEISS]] and [[STUDY_POSA_PHASE3]] precedents. STUDY_ATOGEPANT_PHASE3 is named for the drug plus phase rather than for either trial acronym precisely because it spans two trials; had the source reported separate estimates, STUDY_ADVANCE and STUDY_PROGRESS would have been the right pair.

COPD FEV1 meta-analysis columns

Registered with the Yang_2026_copd_fev1_adipd_mbma / Yang_2026_copd_fev1_ipd extraction. The first group are combined aggregated-data + individual-patient-data (ADIPD) meta-analysis infrastructure columns that are not specific to COPD and are expected to recur in any combined-evidence model; the rest are COPD-specific.

DTYPE_AGGREGATED (canonical for the aggregated-vs-individual record-type indicator in a combined AD + IPD meta-analysis)

• Description: 1 = the record is an AGGREGATED-data observation (one arm-mean endpoint value read from a published trial report), 0 = the record is an INDIVIDUAL-patient observation (one subject’s value). Marks which of the two evidence types a row carries when both are fitted in one model.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (individual-patient record).
• Source aliases:
  • DTYPE – the $INPUT column in the Yang 2026 combined control stream, coded 1 = aggregated and 2 = individual; re-coded to a 0/1 binary for the canonical column so it reads as an ordinary indicator.
• Example models: Yang_2026_copd_fev1_adipd_mbma.R (switches the baseline between a log-normal and its normal-approximated arm mean, switches the residual error between a sample-size-weighted additive model and a power model, and gates the subject-level random effects).
• Notes: Distinguishes the two evidence types WITHIN one data set; it is not a study-design descriptor and says nothing about randomization or blinding. A combined-evidence model typically uses it in three places at once – the parameter-mean transformation (an arm mean of a log-normally distributed parameter is not log-normal, so an aggregated record needs the normal approximation of the central limit theorem to avoid aggregation bias), the residual-error model (an arm mean of N patients is far less noisy than one patient), and the random-effect structure (subject-level etas are meaningless on an arm mean). Pairs with N_ARM, which supplies the N those transformations need. Register a separate canonical rather than overloading this one if a future model needs more than two evidence types (for example reconstructed individual data as a third level).

MEAS_POSTBD (canonical for the post-bronchodilator spirometry measurement indicator)

• Description: 1 = this spirometry record was measured AFTER administration of a short-acting bronchodilator, 0 = the record is a pre-bronchodilator (trough) measurement. A record-level, not subject-level, attribute.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (pre-bronchodilator trough measurement).
• Source aliases:
  • POSTBD – the $INPUT column in the Yang 2026 control streams.
• Example models: Yang_2026_copd_fev1_adipd_mbma.R (a post-bronchodilator record is predicted from a baseline raised by the median absolute reversibility of 0.18 L and from only the estimated fraction rel_postbd = 0.531 of the long-acting bronchodilator effect, because a short-acting bronchodilator has already recruited part of the available reversibility).
• Notes: Load-bearing in any pooled airways-disease analysis because trials differ in whether they report trough or post-bronchodilator FEV1, and the two are not interconvertible without a reversibility assumption. Do NOT use it to mark a study that reports only post-bronchodilator values – it is per record, and a single trial commonly reports both. Distinct from FEV1_PBD_ANCHOR, which addresses a different reconciliation problem (studies whose absolute FEV1 was reconstructed from a post-bronchodilator BASELINE plus a change from baseline). Scope: specific until a second airways model ratifies the semantics.

FEV1_PBD_ANCHOR (canonical for the reported FEV1 value consumed by a post-bronchodilator baseline reconciliation term)

• Description: The record’s own reported FEV1 value, supplied as an input column, for records belonging to studies whose absolute FEV1 was reconstructed as (post-bronchodilator baseline + change from baseline) while the model’s baseline is on the pre-bronchodilator scale. 0 for every record to which the reconciliation does not apply, and 0 for all simulation.
• Units: L
• Type: continuous
• Scope: specific
• Reference category: 0 (no reconciliation; the correction term vanishes).
• Source aliases:
  • FEV1 used as a right-hand-side data item – the Yang 2026 combined control stream writes POSTBDCORR = FEV1 * (1-THETA(53)) for six named studies with POSTBD = 0, where FEV1 is the DV itself.
• Example models: Yang_2026_copd_fev1_adipd_mbma.R (multiplied by 1 - postbd_recon = 0.116 and added to the pre-bronchodilator prediction for six of the 298 aggregated studies).
• Notes: Exists to reproduce an ESTIMATION-TIME data-reconciliation idiom and must be 0 for forward simulation. The source multiplies the record’s own DEPENDENT VARIABLE into the prediction, which rxode2 cannot express (the observation is unknown while solving) and which is not meaningful for simulation; exposing the value as an explicit input column is the faithful translation, and setting it to 0 reproduces the published model exactly for every record outside those six studies. Registered rather than hidden inside the model so that the dependence is auditable. A future extraction meeting the same idiom for a different endpoint should register a parallel <ENDPOINT>_PBD_ANCHOR-style column rather than overloading this one, since the anchor is endpoint-specific by construction.

OCS_NONRESPONDER (canonical for the oral-corticosteroid-non-responder-only study-population indicator)

• Description: 1 = the trial enrolled ONLY patients documented not to respond to oral corticosteroids, 0 = an unselected study population. A study-level design covariate.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (unselected study population).
• Source aliases:
  • OCNR – the derived flag in the Yang 2026 combined control stream, set by IF (REF.EQ.636) OCNR = 1 with the comment “only non-responders to OCS are included, virtually zero effect”.
• Example models: Yang_2026_copd_fev1_adipd_mbma.R (forms rel_cs = 1 - OCS_NONRESPONDER, which zeroes the INHALED-corticosteroid component of the anti-inflammatory effect while leaving the non-steroid anti-inflammatory agents – roflumilast, cilomilast, AZD9668, PH797804 – untouched).
• Notes: Carries no estimated parameter of its own; it is a pure structural switch, which is why it is a covariate rather than a fitted effect. Selecting on oral-corticosteroid non-response is an enrichment strategy that predicts near-zero inhaled-corticosteroid benefit, so a meta-analysis pooling such a trial with unselected trials must exclude it from the steroid effect or the steroid Emax is biased downward. Distinct from INCL_EXAC_REQUIRED, the other study-level entry-criterion covariate registered here, which DOES carry an estimated effect.

INCL_EXAC_REQUIRED (canonical for the exacerbation-history trial entry-criterion indicator)

• Description: 1 = the trial’s inclusion criteria required a documented history of disease exacerbations for enrolment, 0 = no such requirement. A study-level design covariate marking trials enriched for exacerbating patients.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no exacerbation-history entry requirement).
• Source aliases:
  • INCL – the $INPUT column in the Yang 2026 control streams.
• Example models: Yang_2026_copd_fev1_adipd_mbma.R (estimated effect on baseline FEV1, e_incl_exac_base = -0.0213, and an additional role inside the fixed-coefficient regressions that impute missing background-therapy fractions).
• Notes: A STUDY DESIGN attribute, not a patient characteristic: it describes who the trial was willing to enrol, and is therefore constant within a study. Distinct from NEXAC12M, which is a per-patient count of exacerbations in the previous 12 months; a trial can require an exacerbation history without any patient-level count being reported, which is the common case in published aggregated data. Patients meeting an exacerbation-history criterion have lower lung function, so a meta-analysis pooling enriched and unenriched trials needs this term or the enriched trials’ baselines look anomalously low. In Yang_2026_copd_fev1_adipd_mbma it is 0 on every individual-patient record, because neither individual-patient trial applied the criterion.

BGTHER_ICS_RUNIN_PCT, BGTHER_LABA_RUNIN_PCT, BGTHER_LAAC_RUNIN_PCT, BGTHER_ICS_MAINT_PCT, BGTHER_LABA_MAINT_PCT, BGTHER_LAAC_MAINT_PCT (canonical for the percentage of a study arm on a named class of background respiratory therapy, by trial period)

• Description: Percentage of the patients in a study arm receiving background (non-randomized) therapy of the named drug class, during the named trial period. ICS = inhaled corticosteroid, LABA = long-acting beta-2 agonist, LAAC = long-acting anticholinergic. RUNIN = the run-in period before randomized treatment begins; MAINT = the randomized maintenance period. Members of the BGTHER_<class>_<period>_PCT family.
• Units: % of patients in the arm (0-100, NOT a 0-1 fraction)
• Type: continuous
• Scope: specific
• Reference category: 0 (no background therapy of that class).
• Source aliases:
  • ICSpRUNIN, LABApRUNIN, LAACpRUNIN, ICSpMAINT, LABApMAINT, LAACpMAINT – the $INPUT columns in the Yang 2026 control streams. Those columns carry two in-band sentinels that the canonical columns preserve: 9999 = the class was not used at all (treated as 0) and 7777 = the class was used but the fraction was not reported, which activates the source’s fixed-coefficient logistic imputation.
• Example models: Yang_2026_copd_fev1_adipd_mbma.R (each class’s background contribution is valued at a reference compound for that class – fluticasone propionate b.i.d. for ICS, salmeterol for LABA, tiotropium at its 18 ug/day reference dose for LAAC – scaled by the arm fraction; the run-in fractions additionally set the starting point of each class’s effect-onset time course, so an arm already on a class starts part-way up its onset curve).
• Notes: An ARM-LEVEL PREVALENCE, not a per-patient indicator: in aggregated meta-analysis data the published trial reports what proportion of an arm was on background therapy, never who. That is why the family is _PCT and continuous rather than a set of binary CONMED_* indicators, and why the value enters the model linearly as a fraction of a reference compound’s effect. Distinct from the CONMED_<INN>_DOSE family, which carries RANDOMIZED treatment doses; background therapy is what patients were already taking and is present in placebo arms too, so a model must not gate it on active treatment. The run-in / maintenance split is load-bearing because trials commonly withdraw or standardise background therapy at randomization, so the two periods can carry very different fractions. Percent rather than fraction is deliberate and matches the source columns; a model consuming these divides by 100.

DIS_COPD_GOLD (canonical for the GOLD spirometric severity stage of a COPD patient)

• Description: Global Initiative for Chronic Obstructive Lung Disease (GOLD) spirometric severity stage as an ordinal 1-4 category: 1 = mild, 2 = moderate, 3 = severe, 4 = very severe, assigned from FEV1 percent predicted. Per-patient and time-fixed at screening.
• Units: (ordinal stage 1-4)
• Type: categorical
• Scope: specific
• Reference category: n/a – enters as a centred linear deviation 1 + e_gold_<param> * (DIS_COPD_GOLD - ref) or a piecewise-linear (hockey-stick) form with a knee at the reference. Reference value observed: stage 3 (Yang 2026, the cohort median, used both as the centring constant and as the hockey-stick knee).
• Source aliases:
  • FL_COPD – the $INPUT column in the Yang 2026 control streams, glossed in Supporting Information Table S1 footnote 1 as the “% predicted GOLD Stage Category at the screening phase”. Sentinel -99 marks a missing stage, which the source replaces with the median stage 3.
• Example models: Yang_2026_copd_fev1_ipd.R (hockey-stick with separate slopes below and above stage 3 on baseline FEV1 and on vilanterol efficacy, plus a single slope on the disease-progression slope), Yang_2026_copd_fev1_adipd_mbma.R (a SINGLE linear slope on baseline, the hockey stick having been deliberately linearized away to remove a source of aggregation bias).
• Notes: Carried as a SINGLE ORDINAL COLUMN, deliberately not as pre-binned stage indicators, following the rationale recorded for SMOKE_TTFC_SCORE: the founding models fit piecewise-LINEAR effects in the raw stage number, which indicator columns cannot reproduce, and one column can serve either a linear-in-stage or a level-by-level parameterisation. Distinct from DIS_COPD, a binary has-COPD-or-not indicator, and from FEV1_PCTPRED, the continuous measurement the stage is binned from – a model with access to the continuous value should prefer it, since the staging discards information. GOLD also publishes an A-D / A-E symptom-and-exacerbation grouping that is NOT this variable; register a separate canonical if a model needs it. The two founding models show the same covariate fitted with different functional forms in the same paper, so always read the per-model covariateData[[DIS_COPD_GOLD]]$notes for the form and centring rather than assuming. Typed categorical rather than ordinal: this file’s ## Entry schema vocabulary is continuous | binary | categorical | count and has no ordinal member, and SMOKE_TTFC_SCORE – the precedent this entry follows – records its own ordinal-ness in the Units: slot ((ordinal score 0-3)) while typing itself categorical. The Units: line above does the same here.

DIS_COPD_GOLD_LOW, DIS_COPD_GOLD_HIGH (canonical for the lowest and highest GOLD spirometric stage admitted by a trial’s inclusion criteria)

• Description: The bounds of the GOLD spirometric severity range a trial was willing to enrol, each as an ordinal 1-4 category. Study-level design covariates that stand in for a per-patient severity distribution when only aggregated data are available.
• Units: (ordinal stage 1-4)
• Type: categorical
• Scope: specific
• Reference category: n/a – typically combined into the arm’s mean stage as (DIS_COPD_GOLD_LOW + DIS_COPD_GOLD_HIGH) / 2 and then centred. In the background-therapy imputation regressions of the founding model they are centred separately, at 2 and 4 respectively.
• Source aliases:
  • LOWDS and HIGHDS – the $INPUT columns in the Yang 2026 control streams.
• Example models: Yang_2026_copd_fev1_adipd_mbma.R (the aggregated arm’s mean severity is the midpoint of the two, which is the paper’s deliberate linearization of the published model’s separate lowest-and-highest severity effects; they also drive the fixed-coefficient logistic regressions that impute missing background-therapy fractions).
• Notes: Used on AGGREGATED-data records in place of DIS_COPD_GOLD, because a published trial reports its inclusion range rather than each patient’s stage. A trial admitting a single stage sets both columns equal. The midpoint is an approximation to the arm’s mean severity and will be wrong for a trial whose enrolment was skewed within its admitted range; the founding paper adopts it explicitly to keep the aggregated and individual covariate models on the same scale and thereby avoid aggregation bias (Section 2.3.1). Registered as a pair rather than as a single range-width column because the two bounds enter the imputation regressions with different centrings. Typed categorical for the same reason as DIS_COPD_GOLD: the ordinal-ness is recorded in the Units: slot, this file’s schema vocabulary having no ordinal member.

CONMED_ACLIDINIUM_DOSE, CONMED_ARFORMOTEROL_DOSE, CONMED_AZD9668_DOSE, CONMED_BATEFENTEROL_DOSE, CONMED_BEA2180_DOSE, CONMED_BECLOMETHASONE_DOSE, CONMED_BUDESONIDE_DOSE, CONMED_CILOMILAST_DOSE, CONMED_FLUTICASONEFUROATE_DOSE, CONMED_FLUTICASONEPROPIONATE_DOSE, CONMED_FORMOTEROL_DOSE, CONMED_GLYCOPYRRONIUM_DOSE, CONMED_GSK233705_DOSE, CONMED_INDACATEROL_DOSE, CONMED_MOMETASONE_DOSE, CONMED_OLODATEROL_DOSE, CONMED_PH797804_DOSE, CONMED_REVEFENACIN_DOSE, CONMED_ROFLUMILAST_DOSE, CONMED_SALMETEROL_DOSE, CONMED_TIOTROPIUM_DOSE, CONMED_UMECLIDINIUM_DOSE, CONMED_VILANTEROL_DOSE (canonical for the per-arm total daily dose of a named respiratory agent in a COPD dose-response meta-analysis)

• Description: Total daily dose of the named agent assigned to the study arm (per-arm meta-analysis) or to the individual subject (subject-level trial simulation). 0 = the agent was not given as a randomized treatment in this arm, so it contributes nothing; every column of the family 0 = a placebo (or background-therapy-only) arm. Members of the CONMED_<INN>_DOSE family, alongside its rheumatoid-arthritis, psoriasis, statin and anticoagulant members.
• Units: ug/day of TOTAL DAILY dose throughout this family, including for the orally dosed agents (roflumilast, cilomilast), so that one unit serves every column and matches the unit of each drug’s ED50. Note that a b.i.d. regimen therefore appears as twice its per-administration dose – budesonide 160 ug b.i.d. is 320, formoterol 9 ug b.i.d. is 18, aclidinium 400 ug b.i.d. is 800.
• Type: continuous
• Scope: specific
• Reference category: 0 (this agent was not given as a randomized treatment in this arm).
• Source aliases:
  • avdostot1 / avdostot2 / avdostot3 – the Yang 2026 control streams carry up to three treatment slots per arm, each a (drgNoN, avdostotN) pair of a drug code and its average total daily dose. The wide per-drug columns of this family are the same information keyed by drug name instead of by slot: the source’s IF (drgNo1.EQ.25) VILDOSE = avdostot1 idiom becomes a CONMED_VILANTEROL_DOSE column directly.
• Example models: Yang_2026_copd_fev1_adipd_mbma.R (all 23 columns; each drives its own drug’s Emax or constant-effect term, which are then summed within drug class and combined across the three bronchodilator classes through a power interaction), Yang_2026_copd_fev1_ipd.R (CONMED_VILANTEROL_DOSE and CONMED_FLUTICASONEFUROATE_DOSE only).
• Notes: Reference doses against which the founding model quotes each drug’s reference efficacy: aclidinium 200 (q.d.) / 800 (b.i.d.), arformoterol 50, budesonide 320, formoterol 18, indacaterol 75, glycopyrronium 100, roflumilast 500, tiotropium 18 (HandiHaler) / 5 (Respimat), vilanterol 25, revefenacin 175, olodaterol 5, batefenterol 400, fluticasone furoate 100 ug/day. Ten of the 23 carry NO dose-response at all – beclomethasone, cilomilast, fluticasone propionate, mometasone, salmeterol, umeclidinium, AZD9668, GSK233705, BEA2180 and PH797804 have a single constant effect, and the model keys those on the column being NON-ZERO, following the idiom already recorded for CONMED_MTX_DOSE in Mandema_2011_biologicDMARDs_mbma.R. The column still carries the dose so the arm is self-describing; do not read dose-proportionality into those ten. CONMED_FLUTICASONEPROPIONATE_DOSE and CONMED_FLUTICASONEFUROATE_DOSE are DIFFERENT MOLECULES with different effects and different dose-responses and must never be merged into a single “fluticasone” column; the founding source’s own Table S3 mislabels one as a ratio of the other, which is exactly the confusion the two distinct names prevent. CONMED_ARFORMOTEROL_DOSE carries no parameters of its own in the founding model, which assumes arformoterol’s Emax equals formoterol’s and its ED50 is half of formoterol’s because it is the single active enantiomer of racemic formoterol. Several columns name development codes rather than INNs (AZD9668, BEA2180, GSK233705, PH797804) because the source identifies those investigational compounds only by code; if one is later assigned an INN, register the INN name as an alias rather than renaming the column.

FORM_ACLIDINIUM_BID, FORM_MOMETASONE_BID, FORM_OLODATEROL_BID, FORM_BATEFENTEROL_BID (canonical for the twice-daily-regimen indicator of a named inhaled respiratory agent)

• Description: 1 = the arm received the named agent on a twice-daily schedule, 0 = once daily (the model reference). Selects between two separately parameterised versions of one drug’s effect. Members of the FORM_<drug>_<variant> family, alongside FORM_FLV_BID_XR and FORM_LOV_BID_XR.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (once-daily regimen).
• Source aliases:
  • dosfreqNo1 / dosfreqNo2 / dosfreqNo3 – the Yang 2026 control streams code 1 = q.d. and 2 = b.i.d. per treatment slot; the binary indicator is dosfreqNoN == 2 for the slot carrying that drug.
• Example models: Yang_2026_copd_fev1_adipd_mbma.R (aclidinium selects a separate b.i.d. reference efficacy AND a separate b.i.d. ED50; mometasone multiplies the q.d. effect by rel_mome_bid = 0.784; olodaterol and batefenterol each switch from a q.d. Emax dose-response to a distinct CONSTANT b.i.d. effect).
• Notes: The regimen is not reducible to the total daily dose: the founding model gives the b.i.d. arms their own potency parameters, not merely twice the dose, because a split dose has a different trough-concentration profile and trough FEV1 is the endpoint. Note the four members are used differently – a separate Emax/ED50 pair, a multiplicative ratio, and a switch from a dose-response to a flat effect – so always read the per-model covariateData entry rather than assuming a common form. The founding control stream records that q.d. and b.i.d. were “never given in ambiguous combination in dataset” for each of these drugs, which is what makes a single binary sufficient. Umeclidinium deliberately has NO member of this family because the founding model assigns its q.d. and b.i.d. regimens the same effect.

FORM_TIOTROPIUM_SMI (canonical for the tiotropium soft-mist-inhaler device indicator)

• Description: 1 = tiotropium was delivered by the Respimat soft-mist inhaler, 0 = by the Spiriva HandiHaler dry-powder inhaler (the model reference).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (HandiHaler dry-powder inhaler).
• Source aliases:
  • dosfrm1SM / dosfrm2SM / dosfrm3SM – the Yang 2026 control streams set a single SMI flag if any treatment slot is a soft-mist formulation.
• Example models: Yang_2026_copd_fev1_adipd_mbma.R (selects the 5 ug/day reference dose and the Respimat reference efficacy of 0.120 L in place of the 18 ug/day HandiHaler pair at 0.122 L; the Respimat ED50 is then DERIVED from the HandiHaler Emax and the Respimat reference efficacy rather than estimated, via ed50_smi = 5 * (emax_dpi / effref_smi - 1)).
• Notes: A genuine formulation / device covariate, unlike its sibling FORM_TIOTROPIUM_OPENLABEL. The two devices deliver very different nominal doses for a similar effect (5 versus 18 ug/day), so a meta-analysis pooling them on nominal dose without this indicator would badly misestimate tiotropium potency. The founding control stream records that there are “no occurances of tio being given as HandiHaler with another drug as SMI”, which is what allows a single arm-level flag rather than a per-slot one.

FORM_TIOTROPIUM_OPENLABEL (canonical for the open-label tiotropium administration indicator)

• Description: 1 = tiotropium was given open-label, 0 = blinded (the model reference).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (blinded administration).
• Source aliases:
  • DRGOL1 / DRGOL2 / DRGOL3 – the Yang 2026 control streams set an open-label flag per treatment slot.
• Example models: Yang_2026_copd_fev1_adipd_mbma.R (scales the tiotropium reference efficacy by rel_tio_ol = 0.918 while keeping the blinded ED50, i.e. open-label tiotropium is estimated to perform about 8% worse than the same drug given blinded).
• Notes: Strictly a TRIAL-CONDUCT covariate rather than a formulation, and is registered in the FORM_<drug>_<variant> family only because it selects between two variants of one drug’s effect in exactly the way the other members do; the alternative of a general OPENLABEL covariate was rejected because the estimated effect is a tiotropium-specific potency ratio, not a generic unblinding bias, and pooling it across drugs would be unsupported. The direction is counterintuitive – open-label administration reducing rather than inflating the measured effect – and reflects that the open-label tiotropium arms served as active comparators in trials of other drugs rather than as the trial’s own test arm. The founding control stream records that “blinded tio never given with another drug being OL in dataset”.

FORM_NICOTINE_CC (canonical for the conventional-cigarette nicotine-product indicator)

• Description: 1 = the nicotine dose on this record was taken from a conventional (combusted) cigarette; 0 = any other nicotine product. Per-dose-occasion indicator. One of three mutually exclusive product-type indicators (FORM_NICOTINE_CC, FORM_NICOTINE_NNS, FORM_NICOTINE_GUM) whose common reference level – all three zero – is the Tobacco Heating System (THS), the reference product of the Marchand 2017 analysis.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (Tobacco Heating System). THS is the bioavailability anchor: Marchand 2017 Table 3 fixes Frel-THS to 1 and reports every other product as a log-scale difference from it.
• Source aliases:
  • nature of product – Marchand 2017 Suppl. Table 1 (a categorical primary covariate) and Table 4, where the levels appear as the separate estimates dFreld-CC, dFreld-NNS and dFreld-GUM.
• Example models: Marchand_2017_nicotine.R (Frel_CC = Frel_THS * exp(0.0189), i.e. cigarette bioavailability is 102 percent of THS at the same nicotine ISO yield, body weight and subject characteristics; Marchand 2017 Table 4).
• Notes: Auto-approved member of the FORM_<drug>_<formulation> family. Registered as a product-type rather than a dosage-form contrast because the family’s defining question – does this record’s delivery form change the drug’s input kinetics or extent of absorption – is exactly what separates a combusted cigarette from a heated-tobacco stick, a nasal spray and a chewing gum. Orthogonal to FORM_NICOTINE_MENTHOL, which flags the mentholated variant of whichever product is in use: a mentholated cigarette record carries both indicators. Do not fold the product identity into DOSE_NICOTINE_MG; Marchand 2017 fits the nominal-dose power term and these indicators simultaneously, so the two carry genuinely different information.

FORM_NICOTINE_NNS (canonical for the nicotine-nasal-spray product indicator)

• Description: 1 = the nicotine dose on this record was given as the regular nicotine nasal spray; 0 = any other nicotine product. Per-dose-occasion indicator. Sibling of FORM_NICOTINE_CC and FORM_NICOTINE_GUM with the same all-zero Tobacco Heating System reference.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (Tobacco Heating System).
• Source aliases:
  • nature of product – Marchand 2017 Suppl. Table 1 and Table 4 (level NNS, estimate dFreld-NNS).
• Example models: Marchand_2017_nicotine.R (Frel_NNS = Frel_THS * exp(-1.42), i.e. nasal-spray bioavailability is 24 percent of THS; Marchand 2017 Table 4).
• Notes: Auto-approved member of the FORM_<drug>_<formulation> family. The nasal spray is not an inhaled product, yet Marchand 2017 found zero-order absorption described it adequately alongside the inhaled products (Sect. 3.2), so this indicator modifies only the extent of absorption and not the absorption model: unlike FORM_NICOTINE_GUM it carries no effect on the zero-order duration Tdur.

FORM_NICOTINE_GUM (canonical for the nicotine-gum product indicator)

• Description: 1 = the nicotine dose on this record was given as nicotine chewing gum; 0 = any other nicotine product. Per-dose-occasion indicator. Sibling of FORM_NICOTINE_CC and FORM_NICOTINE_NNS with the same all-zero Tobacco Heating System reference.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (Tobacco Heating System).
• Source aliases:
  • nature of product – Marchand 2017 Suppl. Table 1 and Table 4 (level GUM, estimates dFreld-GUM and dTdurd-GUM).
• Example models: Marchand_2017_nicotine.R (two effects: Frel_GUM = Frel_THS * exp(-0.489), 61 percent of THS bioavailability; and Tdur = 5.30 * exp(2.14) = 45 min against 5.3 min for every other product, the authors’ explicit allowance for chewing taking longer than inhaling – Marchand 2017 Table 4 and Sect. 3.2).
• Notes: Auto-approved member of the FORM_<drug>_<formulation> family. The only product-type indicator in the set that modifies the absorption rate structure as well as its extent, which is why the three indicators are registered separately rather than as a single categorical column. In the founding dataset every gum occasion was the mentholated variant, so FORM_NICOTINE_MENTHOL is also 1 on those records and the two menthol effects apply on top of the gum effects; a future dataset with regular-flavour gum would separate them.

FORM_NICOTINE_MENTHOL (canonical for the mentholated nicotine-product variant indicator)

• Description: 1 = the nicotine product used on this record was its mentholated variant; 0 = the regular, non-mentholated variant. Per-dose-occasion indicator, and deliberately orthogonal to the product-type indicators: mentholated heated-tobacco sticks, mentholated cigarettes and mentholated gum all carry a 1 alongside whichever product indicator applies.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (regular, non-mentholated variant).
• Source aliases:
  • STUDY – the integer study column of the Choi 2016 NONMEM $INPUT (deposited control stream PSP4-5-532-s008), where STUDY = 1 is the SAD trial and STUDY = 2 the S&MAD trial; STUDY_SMAD = as.integer(STUDY == 2).
• Example models: Choi_2016_gcc4401c.R, Marchand_2017_nicotine.R (V1/F = 70.0 * exp(0.0912), a 9.5 percent larger apparent central volume, and Tdur = 5.30 * exp(0.0530), a 5 percent longer zero-order absorption duration; Marchand 2017 Table 4).
• Notes: Carries four distinct effects in the founding model, which is why it is a single indicator rather than a residual-error stratum. (i) Apparent central volume / assay accuracy. The control stream sets S2 = V2/1000 and, for STUDY.EQ.2, S2 = (V2/1000)/THETA(10) with THETA(10) = 0.82 – so predicted plasma concentrations in the S&MAD study are 18% lower than in the SAD study for the same amount. The paper attributes this to the two trials using different bioanalytical laboratories (Choi 2016 Eqs. 7-8 and Table 2 footnote d, and DeltaMOFV -21.037 for its inclusion). Because the scaling divides the volume, it is encoded as e_study_smad_vc rather than as a separate assay parameter. (ii) The sigmoidicity gamma of the coagulation-factor-X PD model (1.07 SAD versus 0.60 S&MAD, Table 3a). (iii) The EC50 of both prothrombin-time PD models (426 versus 1350 ng/mL for INR, Table 3d; 563 versus 1450 ng/mL for seconds, Table 3e). (iv) The Emax of the aPTT PD model (16.9 versus 20.4 s, Table 3f). The paper notes the PD differences “might be partially ascribed to study difference” without identifying a mechanism. The companion rivaroxaban model Choi_2016_rivaroxaban.R does not carry this covariate because that arm ran only in the S&MAD study. Do not reuse this name for an unrelated trial’s single-and-multiple-ascending-dose cohort – SMAD here is the Choi 2016 trial’s own label, and a new paper needs its own STUDY_<id> member.

FORM_NICOTINE_FRESHMINT (canonical for the coated Freshmint/Freshfruit nicotine chewing-gum indicator)

• Description: 1 = the nicotine chewing gum used on this record was the coated Nicorette Freshmint or Freshfruit gum; 0 = Nicorette classic gum. Per-dose-record indicator within a gum-only model.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (Nicorette classic chewing gum).
• Source aliases:
  • FORM (FORM.NE.2, control-stream flag FORMFL) – Olsson Gisleskog 2021 ESM gum control stream.
• Example models: OlssonGisleskog_2021_nicotine_gum.R (oromucosal ka at 2 mg 9.29 1/h for the coated gum vs 26.5 1/h for classic gum; Table 5 footnote d).
• Notes: Auto-approved member of the FORM_<drug>_<formulation> family. Not a sibling of FORM_NICOTINE_GUM (Marchand 2017), which contrasts gum with the Tobacco Heating System across product types; this one contrasts two gum variants within a gum-only model.

FORM_NICOTINE_NICORETTE_LOZENGE (canonical for the Nicorette-vs-NiQuitin nicotine lozenge indicator)

• Description: 1 = the nicotine lozenge used on this record was a Nicorette lozenge (McNeil AB); 0 = a NiQuitin lozenge (GlaxoSmithKline). Per-dose-record indicator within a lozenge-only model.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (NiQuitin lozenge).
• Source aliases:
  • TMT (codes 20001, 20003, 20005 and 20006, control-stream term EFF_FRSW1) – Olsson Gisleskog 2021 ESM lozenge control stream; Table 5 row ‘Nicorette on Frsw’.
• Example models: OlssonGisleskog_2021_nicotine_lozenge.R (adds 0.412 to the logit of the fraction of the dose swallowed).
• Notes: Auto-approved member of the FORM_<drug>_<formulation> family. The product name is spelled out with the dosage form because ‘Nicorette’ is the brand of several nicotine products (gum, patch, spray, inhaler); the indicator is only meaningful among lozenges.

FORM_NICOTINE_INVISIPATCH (canonical for the Nicorette Invisipatch-vs-Nicorette patch indicator)

• Description: 1 = the transdermal patch used on this record was the Nicorette Invisipatch (10-25 mg/16 h); 0 = the Nicorette patch (5-15 mg/16 h). Per-dose-record indicator within a patch-only model.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (Nicorette patch).
• Source aliases:
  • FORM (FORM.EQ.8 is the Nicorette patch, control-stream flag FORMFL) – Olsson Gisleskog 2021 ESM transdermal control stream; ‘NNTP’ in the paper’s figures.
• Example models: OlssonGisleskog_2021_nicotine_transdermal.R (switches the first-order release fraction Fr1 71.9% vs 40.0%, the duration fraction Frdur1 96.2% vs 44.5% of 16 h, and a 0.53-h lag on the first-order pathway).
• Notes: Auto-approved member of the FORM_<drug>_<formulation> family. The two patches differ in design: the Invisipatch skin-contact layer carries nicotine from the start, while the Nicorette patch adhesive layer contains little at release (Olsson Gisleskog 2021 Discussion).

SNP_CES1_RS71647871 (canonical for CES1 rs71647871 (p.Gly143Glu) variant-allele carrier indicator)

• Description: Binary germline indicator for the CES1 nonsynonymous variant rs71647871 (c.428G>A, p.Gly143Glu), which abolishes much of the catalytic activity of carboxylesterase 1. 1 = carrier of at least one A (variant) allele; 0 = GG wild-type homozygote. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (rs71647871 GG wild-type homozygote).
• Source aliases:
  • X4A – used in Lyauk_2016_methylphenidate.R. Value-inverted: the source NONMEM column codes 1 = wild-type and 0 = the GA variant (IF(X4A.EQ.1) CLX4A = 1 / IF(X4A.EQ.0) CLX4A = ( 1 + THETA(13))), plus the sentinel 999 for missing. Convert on ingestion with SNP_CES1_RS71647871 = as.integer(X4A == 0). The paper’s own printed typical-value equation already uses the carrier-equals-1 orientation that this canonical adopts.
• Example models: Lyauk_2016_methylphenidate.R (linear proportional effect on apparent oral clearance, cl <- ... * (1 + e_snp_ces1_rs71647871_cl * SNP_CES1_RS71647871) * ... with e_snp_ces1_rs71647871_cl = -0.587; GA carriers have 58.7% lower CL/F, giving a 2.43-fold higher d-methylphenidate AUC0-inf, Lyauk 2016 Table 1 theta 9 and Figure 4).
• Notes: Member of the auto-approving SNP_<GENE>_<RSID> germline-pharmacogenomics family, following the family default that 1 = mutant allele present. General scope: rs71647871 is the one CES1 variant with replicated cross-drug effects, reported for oseltamivir, clopidogrel, quinapril and enalapril as well as methylphenidate (Lyauk 2016 Discussion and references 25-27), so the same column will recur in any CES1-substrate analysis. Only heterozygotes were observed in the founding cohort (6 of 116 genotyped subjects); a future paper resolving AA homozygotes separately should register a _HOM sibling on the SNP_ABCG2_RS2231142_HET / _HOM pattern rather than overloading this carrier indicator. Pair with SNP_CES1_RS71647871_MISSING when the source handles ungenotyped subjects with an EXTRA / EST missing-data parameter. Distinct from SNP_CES1_RS115629050, an independent CES1A1 variant that the founding paper confirmed is in linkage equilibrium with this one (Supplementary Figure S1), so both may be carried simultaneously.

SNP_CES1_RS71647871_MISSING (canonical for CES1 rs71647871 genotype-missing indicator)

• Description: Binary indicator that a subject’s CES1 rs71647871 genotype could not be determined. 1 = genotype missing; 0 = genotype measured (whether wild-type or variant). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: general
• Reference category: 0 (rs71647871 genotype successfully measured). Mutually exclusive with SNP_CES1_RS71647871 = 1.
• Source aliases:
  • X4A sentinel value 999 – used in Lyauk_2016_methylphenidate.R (IF(X4A.EQ.999) CLX4A = ( 1 + THETA(12))). Convert on ingestion with SNP_CES1_RS71647871_MISSING = as.integer(X4A == 999).
• Example models: Lyauk_2016_methylphenidate.R (linear proportional effect on apparent oral clearance, e_snp_ces1_rs71647871_missing_cl = -0.157; the 5% of subjects without a genotype call behave as though modestly enriched for the variant, Lyauk 2016 Table 1 theta 10).
• Notes: Member of the established <COV>_MISSING missing-indicator family (ADA_MISSING, CYP2C9_MISSING, SNP_CYP2C19_RS3814637_MISSING, SNP_CYP3A4_RS2242480_MISSING). The modelling role here is the EXTRA (also called EST) method of Keizer 2012 and Johansson & Karlsson 2013 for data missing not at random: rather than imputing a genotype, the estimation adds one typical-value multiplier per covariate that contains missing data, so ungenotyped subjects form their own effect stratum. Record which role the flag plays – a shift of the typical value (as here) versus a shift of the between-subject variance – in covariateData[[SNP_CES1_RS71647871_MISSING]]$notes. A source that instead drops ungenotyped subjects, or imputes them to wild-type, needs no such column.

SNP_CES1_RS115629050 (canonical for CES1 rs115629050 (p.Ala270Ser) variant-allele carrier indicator)

• Description: Binary germline indicator for the CES1A1 nonsynonymous variant rs115629050 (p.Ala270Ser), located in exon 7. 1 = carrier of at least one variant (G) allele at the T>G position, i.e. a TG heterozygote or GG homozygote; 0 = wild-type homozygote. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (rs115629050 wild-type homozygote).
• Source aliases:
  • X7A – used in Lyauk_2016_methylphenidate.R. Source coding matches this canonical directly (IF(X7A.EQ.0) CLX7A = 1 wild-type, IF(X7A.EQ.1) CLX7A = ( 1 + THETA(15)) variant), with the sentinel 999 for missing; no value inversion, in contrast to the sibling SNP_CES1_RS71647871 in the same model.
• Example models: Lyauk_2016_methylphenidate.R (linear proportional effect on apparent oral clearance, e_snp_ces1_rs115629050_cl = -0.403; TG carriers have 40.3% lower CL/F, giving a 1.68-fold higher d-methylphenidate AUC0-inf, Lyauk 2016 Table 1 theta 14 and Figure 4).
• Notes: Member of the auto-approving SNP_<GENE>_<RSID> family. Scoped specific rather than general, unlike its sibling SNP_CES1_RS71647871: the founding paper is the first and so far only report of an rs115629050 effect on any drug’s PK, its Discussion records that an in-vitro study of CES1-metabolised ACE inhibitors failed to detect an effect of the same variant, and 42% of the founding cohort could not be genotyped at this locus because the assay cannot resolve sequence downstream of CES1A1 exon 5 when CES1A2 is present. Promote to general only once a second paper ratifies the effect. Confirmed to be in linkage equilibrium with rs71647871 in the founding cohort (Lyauk 2016 Supplementary Figure S1), which is what permits both to enter one model. Pair with SNP_CES1_RS115629050_MISSING.

SNP_CES1_RS115629050_MISSING (canonical for CES1 rs115629050 genotype-missing indicator)

• Description: Binary indicator that a subject’s CES1 rs115629050 genotype could not be determined. 1 = genotype missing; 0 = genotype measured. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (rs115629050 genotype successfully measured). Mutually exclusive with SNP_CES1_RS115629050 = 1.
• Source aliases:
  • X7A sentinel value 999 – used in Lyauk_2016_methylphenidate.R (IF(X7A.EQ.999) CLX7A = ( 1 + THETA(14))). Convert on ingestion with SNP_CES1_RS115629050_MISSING = as.integer(X7A == 999).
• Example models: Lyauk_2016_methylphenidate.R (linear proportional effect on apparent oral clearance, e_snp_ces1_rs115629050_missing_cl = 0.090, Lyauk 2016 Table 1 theta 15).
• Notes: Member of the <COV>_MISSING family, used here for the EXTRA / EST missing-not-at-random method (see SNP_CES1_RS71647871_MISSING Notes). This particular flag covers an unusually large 42% of the founding cohort, for the structural assay reason recorded under SNP_CES1_RS115629050, and its estimated coefficient is the only one in that model whose bootstrap 95% CI spans zero (-0.0623 to 0.262) – i.e. ungenotyped subjects are statistically indistinguishable from wild-type. Do not read a large missing fraction as a data-quality defect that could be cleaned away: here missingness is a deterministic consequence of the subject’s CES1A2 copy number, which is precisely why the founding paper classified it as missing not at random rather than missing at random.

CES1_HAPA2_HET (canonical for the CES1A2-bearing haplotype heterozygote indicator)

• Description: Binary indicator that a subject carries exactly one copy of the hybrid gene CES1A2. A CES1 haplotype carries either the nonfunctional pseudogene CES1P1 or the duplicated CES1A2 segment in addition to CES1A1, so the diplotype is a three-level germline dosage; this column flags the heterozygous CES1A1-CES1P1 / CES1A1-CES1A2 group. 1 = one CES1A2 copy; 0 = otherwise (the union of the zero-copy and two-copy groups; the paired indicator CES1_HAPA2_HOM flags the two-copy group). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (CES1A1-CES1P1 / CES1A1-CES1P1, no CES1A2 copy – the most common diplotype – when both CES1_HAPA2_HET and CES1_HAPA2_HOM are 0).
• Source aliases:
  • NCOP – used in Lyauk_2016_methylphenidate.R, which carries the combined CES1A1 + CES1A2 copy number (2 = no CES1A2, 3 = one copy, 4 = two copies, 999 = missing). Convert on ingestion with CES1_HAPA2_HET = as.integer(NCOP == 3).
• Example models: Lyauk_2016_methylphenidate.R (linear proportional effect on apparent oral clearance, e_ces1_hapa2_het_cl = -0.182; one-copy carriers have 18.2% lower CL/F, giving a 1.22-fold higher d-methylphenidate AUC0-inf, Lyauk 2016 Table 1 theta 11 and Figure 4).
• Notes: Member of the <GENE>_HAP<haplotype>_HET / _HOM human-germline-haplotype-dosage family established by SLCO1B1_HAP15_HET / SLCO1B1_HAP15_HOM, and registered there rather than under a copy-number name because the underlying genetics is haplotype dosage: one CES1A2 copy means one of the subject’s two CES1 haplotypes is the CES1A2-bearing form. Deliberately not placed in the PFMDR1_CN2 / PFMDR1_CN3PLUS family, which is a specific parasite-genome amplification set for P. falciparum and carries no implication of diploid haplotype dosage. Not a SNP_<GENE>_<RSID> entry either: CES1A2 is a structural hybrid gene, not a point variant, and has no rsID. Scoped specific pending a second paper – the founding study’s Discussion records that its direction conflicts with an irinotecan study in which CES1A2 presence increased clearance, and with an oseltamivir study that found no diplotype effect. Pair with CES1_HAPA2_MISSING when the source handles undetermined copy number with an EXTRA / EST parameter.

CES1_HAPA2_HOM (canonical for the CES1A2-bearing haplotype homozygote indicator)

• Description: Binary indicator that a subject carries two copies of the hybrid gene CES1A2, i.e. the CES1A1-CES1A2 / CES1A1-CES1A2 diplotype in which both CES1 haplotypes are the CES1A2-bearing form. 1 = two CES1A2 copies; 0 = otherwise (the union of the zero-copy and one-copy groups; the paired indicator CES1_HAPA2_HET flags the one-copy group). Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (CES1A1-CES1P1 / CES1A1-CES1P1, no CES1A2 copy, when both CES1_HAPA2_HET and CES1_HAPA2_HOM are 0).
• Source aliases:
  • NCOP – used in Lyauk_2016_methylphenidate.R (combined CES1A1 + CES1A2 copy number). Convert on ingestion with CES1_HAPA2_HOM = as.integer(NCOP == 4).
• Example models: Lyauk_2016_methylphenidate.R (linear proportional effect on apparent oral clearance, e_ces1_hapa2_hom_cl = -0.410; two-copy carriers have 41.0% lower CL/F, giving a 1.70-fold higher d-methylphenidate AUC0-inf, Lyauk 2016 Table 1 theta 12 and Figure 4).
• Notes: Paired with CES1_HAPA2_HET to encode the three-level CES1A2 diplotype with the no-copy group as the implicit reference (both indicators 0). See CES1_HAPA2_HET Notes for the family rationale and the boundaries against PFMDR1_CN* and SNP_<GENE>_<RSID>. Rest on a small stratum: only 5 of 121 successfully genotyped subjects in the founding cohort carried two copies (Lyauk 2016 Supplementary Table S2), and the paper’s own large-population simulation assumes a 2.5% frequency of the two-copy diplotype against 21.5% for one copy in a Caucasian population.

CES1_HAPA2_MISSING (canonical for the CES1A2 diplotype missing indicator)

• Description: Binary indicator that a subject’s CES1A2 copy number (and hence CES1 diplotype) could not be determined. 1 = diplotype missing; 0 = diplotype measured. Time-fixed per subject.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (CES1A2 copy number successfully measured). Mutually exclusive with CES1_HAPA2_HET = 1 and CES1_HAPA2_HOM = 1.
• Source aliases:
  • NCOP sentinel value 999 – used in Lyauk_2016_methylphenidate.R (IF(NCOP.EQ.999) CLNCOP = ( 1 + THETA(11))). Convert on ingestion with CES1_HAPA2_MISSING = as.integer(NCOP == 999).
• Example models: Lyauk_2016_methylphenidate.R (linear proportional effect on apparent oral clearance, e_ces1_hapa2_missing_cl = -0.535, Lyauk 2016 Table 1 theta 13).
• Notes: Member of the <COV>_MISSING family, used for the EXTRA / EST missing-not-at-random method (see SNP_CES1_RS71647871_MISSING Notes). Unlike its two sibling missing-indicators in the same model, this one covers only 0.8% of the founding cohort – a single subject – yet carries a large (-53.5%) and nominally precise (RSE 9.0%) coefficient. Read it as a nuisance parameter absorbing that one subject rather than as a transferable effect, and do not carry the estimate across to another analysis; a source with no ungenotyped subjects should omit the column entirely rather than set it to 0 for everyone, since checkNamingRegisters() gates registered-but-unused entries.

STUDY_A105 (canonical for the aripiprazole lauroxil phase I study A105 cohort indicator)

• Description: 1 = subject enrolled in A105 (ALK9072-A105, NCT02320032), the open-label phase I study of aripiprazole lauroxil 441 mg q4wk / 882 mg q6wk / 1064 mg q8wk in stable schizophrenia; 0 = subject from whichever other studies the analysis pooled A105 with. Subject-level (time-fixed) binary indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0, but the reference set is analysis-specific and must be read from the model file: the four earlier pooled studies 002 / 101 / 102 / 003 in Hard_2017_aripiprazole_lauroxil.R, and the three nanocrystal-dispersion studies B101 / B102 / B103 in Hard_2018_aripiprazole_lauroxil_nanocrystal.R.
• Source aliases:
  • Study – Hard 2017 ESM Supplemental Table 7 labels the two residual-error rows “sigma2prop ARI A105” and “sigma2prop ARI Non-A105”; Hard 2018 Supplementary Table S3 labels them “sigma2 prop Studies 1, 2, and 3” and “sigma2 prop Study 4”, Study 4 being A105.
• Example models: Hard_2017_aripiprazole_lauroxil.R (selects the proportional residual SD: sqrt(0.0238) = 0.154 for A105 against sqrt(0.0565) = 0.238 for the earlier studies; applied inside model() as propSdStdyA105 * STUDY_A105 + propSdStdyPrior * (1 - STUDY_A105)), Hard_2018_aripiprazole_lauroxil_nanocrystal.R (same role, same study, a later analysis that pools A105 with the three nanocrystal-dispersion studies: sqrt(0.0207) = 0.144 for A105 against sqrt(0.0359) = 0.189 for B101 / B102 / B103).
• Notes: Well-formed member of the auto-approved STUDY_<id> canonical family. The between-study difference is confined to residual error, the same role STUDY_ODYSSEY plays in Chandasana_2024b_dolutegravir.R and STUDY_NOVA in Gaffney_2026_niraparib.R: A105 sampled up to 48 times per patient over 309 days whereas the pooled phase III study 003 sampled 13 times, so the denser study supports a smaller residual magnitude. No structural or covariate parameter differs by study in this model. A105 is the only one of the five pooled studies that gave no oral aripiprazole at all, so a data set carrying STUDY_A105 = 1 should have no cmt = "depot" records.

In-vitro signaling-network perturbation inputs

Logic-based / Boolean signaling-network models (CellNOptR, CNORode, MaBoSS and relatives) are trained on perturbation panels in which a cultured cell line is exposed to combinations of ligands and targeted kinase inhibitors. Two families of per-condition covariate come out of that design: a STIM_<ligand>_<units> member per applied ligand, and an INH_<target> member per pharmacologically inhibited network node. Neither is a PK-driven exposure – there is no absorption, distribution or elimination anywhere in these models; the covariate is the experimental condition itself, held constant for the whole solve.

STIM_EGF_NORM, STIM_IGF1_NORM, STIM_IL6_NORM, STIM_TNFA_NORM, STIM_DHT_NORM (canonical for the normalized activity of an applied ligand input node in a logic-based signaling-network model)

• Description: Activity assigned to the named ligand’s input node in a logic / logic-ODE signaling network, on the model’s own normalized [0, 1] activity scale. The reference level is the model’s normalized basal value (0.5 in the log2-fold-change normalization CellNOptR uses, where 0.5 corresponds to a log2 fold change of zero), and 1 means the ligand was applied to the culture. The five members cover epidermal growth factor (EGF), insulin-like growth factor 1 (IGF-1), interleukin-6 (IL-6), tumour necrosis factor alpha (TNF-alpha) and the androgen dihydrotestosterone (DHT).
• Units: fraction of maximal activity (unitless, [0, 1])
• Type: continuous
• Scope: specific
• Reference category: the model’s normalized basal activity – 0.5 for the CellNOptR log2-fold-change normalization, which is the value the founding model uses. Not 0: an unstimulated input node in that normalization sits at basal, not at zero activity. Reference values observed: Traynard 2017 used 0.5 / 1 only.
• Source aliases:
  • TR:EGF, TR:IGF_1, TR:IL6, TR:TNFa, TR:DHT – the MIDAS-format stimulus columns of the Traynard 2017 supplement perturbationData_LNCaP_MIDAS.csv, which carry 0 / 1 and are recoded to 0.5 / 1 by the supplement’s own fitting script (cnolist$valueStimuli[cnolist$valueStimuli==0]=0.5).
• Example models: Traynard_2017_prostateSignaling_qsp.R (the five ligand inputs of a 20-node logic-based ODE model of LNCaP prostate cancer signaling; STIM_IGF1_NORM and STIM_TNFA_NORM are inert in the fitted model because their receptor nodes have a zero lifetime parameter, but are retained because they are part of the published structure).
• Notes: Member of the in-vitro applied-stimulus STIM_<agent>_<units> family (siblings STIM_ARTESUNATE_NM, STIM_CHLOROQUINE_NM, STIM_LUMEFANTRINE_NM, STIM_MEFLOQUINE_NM, STIM_QUININE_MM), and shares that family’s defining property: the value is what contacts the target directly, with no PK layer. It differs from those siblings in its units token: the applied molar concentration is reported in the upstream data paper rather than in the modeling paper, and the model itself is written on a normalized activity scale, so _NORM records a unitless [0, 1] activity rather than a concentration. Use _NORM only when the model genuinely consumes a normalized activity; when an applied concentration is available and the model consumes it, register a concentration-unit member instead. Distinct from CAV and the CP_<drug> family (systemic plasma exposure) and from CONC_<drug>_MGL (in-vitro antibacterial bath concentration in mg/L).

DOSE_MGSO4_IV_G (canonical for the total intravenous magnesium sulfate dose over a treatment course)

• Description: TOTAL intravenous magnesium sulfate dose administered over a treatment course – loading plus maintenance combined – expressed in grams of the administered salt MgSO4-7H2O (magnesium sulfate heptahydrate, Epsom salt). A per-course scalar, not a per-administration amount. Zero for intramuscular-only regimens and for placebo.
• Units: g MgSO4-7H2O
• Type: continuous
• Scope: specific
• Reference category: n/a – enters a closed-form exposure metric AUC = (DOSE_MGSO4_IV_G + F * DOSE_MGSO4_IM_G) * 1000 * 0.098612 / CL, uncentred, after conversion to milligrams of elemental magnesium.
• Source aliases:
  • DOSE_IV – Du 2019 Methods, Pharmacokinetic Exposure (DOSE_IV,i, “the total intravenous dose for patient i received”); the source NMTRAN column name is not separately reported.
• Example models: Du_2019_magnesiumSulfate_eclampsia.R (landmark logistic exposure-response for eclampsia in preeclampsia; the standard Zuspan regimen of 4 g over 20 min then 1 g/h for 24 h is DOSE_MGSO4_IV_G = 28, DOSE_MGSO4_IM_G = 0).
• Notes: Member of the DOSE_<DRUG>_<ROUTE>_<UNITS> auto-approve family, route-qualified on the precedent of DOSE_BPN_SL_MG and for the same reason: the founding model’s exposure metric weights the intramuscular route by a bioavailability of 0.862 and the intravenous route by 1, so an unqualified dose column would be ambiguous. The unit token is the SALT, not elemental magnesium. Clinical magnesium sulfate regimens are universally prescribed in grams of MgSO4-7H2O, and the founding paper’s Table 1 footnote pins the relationship (“Dosage form in simulated dosing regimen was MgSO4-7H2O, which contains ~10% of magnesium”), so the column carries what a clinician writes and the model does the 0.098612 g/g conversion internally. A future extraction that reports elemental-magnesium doses, or anhydrous MgSO4, must register its own sibling rather than overload this name – the three differ by about a factor of 10 and 1.2 respectively. Distinct from the rxode2/nlmixr2 event column amt, which is not readable from model and in any case carries a per-administration rather than a per-course amount; the founding model has no dose events at all.

DOSE_MGSO4_IM_G (canonical for the total intramuscular magnesium sulfate dose over a treatment course)

• Description: TOTAL intramuscular magnesium sulfate dose administered over a treatment course – loading plus maintenance combined – expressed in grams of the administered salt MgSO4-7H2O. A per-course scalar, not a per-injection amount. Zero for intravenous-only regimens and for placebo.
• Units: g MgSO4-7H2O
• Type: continuous
• Scope: specific
• Reference category: n/a – enters the same closed-form exposure metric as DOSE_MGSO4_IV_G, scaled first by the intramuscular absolute bioavailability.
• Source aliases:
  • DOSE_IM – Du 2019 Methods, Pharmacokinetic Exposure (DOSE_IM,i, “the total intramuscular dose for patient i received”); the source NMTRAN column name is not separately reported.
• Example models: Du_2019_magnesiumSulfate_eclampsia.R (the standard Pritchard regimen – 4 g IV plus a 10 g IM loading dose, then 5 g IM every 4 h for 5 doses – is DOSE_MGSO4_IV_G = 4, DOSE_MGSO4_IM_G = 35, reproducing that row’s Table 1 total of 39 g).
• Notes: The route-qualified sibling of DOSE_MGSO4_IV_G; see that entry for the salt-versus-elemental unit rule, which applies identically here. Kept as a separate column rather than folded into a single total-dose column plus a route indicator because the founding regimens routinely combine both routes in one course (an intravenous loading dose followed by intramuscular maintenance), so a subject can carry a non-zero value in both columns simultaneously. In the founding model the intramuscular contribution is multiplied by an absolute bioavailability of 0.862, a literature value from Salinger 2013 (doi:10.1111/1471-0528.12222) rather than an estimate of the founding analysis.

PI3K_INHIBITED, MEK_INHIBITED, MTOR_INHIBITED, P38_INHIBITED, IKK_INHIBITED (canonical for a targeted-inhibitor indicator on a named node of a signaling-network model)

• Description: 1 = the culture was treated with a small-molecule inhibitor of the named signaling node, 0 = untreated. Members take the form <TARGET>_INHIBITED, where the target token is the network node’s own symbol, upper-cased (PI3K_INHIBITED inhibits the PI3K node, MTOR_INHIBITED the mTOR node, and so on), so the covariate and the state it acts on are unambiguously paired. Values on (0, 1) are meaningful where the model’s inhibition term is continuous – in CNORode the inhibitor multiplies the whole right-hand side of its target node by (1 - <TARGET>_INHIBITED), so an intermediate value is a partial inhibition and a value of 1 pins the node at its initial (basal) activity.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (no inhibitor of that node).
• Source aliases:
  • TR:PI3Ki, TR:MEKi, TR:mTORi, TR:p38i, TR:IKKai – the MIDAS-format inhibitor columns of the Traynard 2017 supplement perturbationData_LNCaP_MIDAS.csv.
• Example models: Traynard_2017_prostateSignaling_qsp.R (five inhibitor covariates gating the PI3K, MEK, mTOR, p38 and IKKa nodes of a 20-node logic-based ODE model; IKK_INHIBITED is inert in the fitted model because tau_IKKa is zero).
• Notes: Registered by target node rather than by compound because signaling-network perturbation panels routinely identify their inhibitors only by target class (“PI3K inhibitor”), with the compound identity and concentration living in the upstream data paper – which is exactly the situation in the founding model. The <TARGET>_INHIBITED agent-first / state-second word order follows the existing one-off in-vitro treatment flags M3M3FBS_PRESENT and MUCIN_PRESENT. When the compound IS named and the model treats it as a co-administered drug rather than as a node clamp, use CONMED_<INN> instead; when the model consumes an applied inhibitor concentration rather than a flag, use the STIM_<agent>_<units> or CONC_<drug>_<units> families. Distinct from INH_MCT_KM_RATIO / INH_MCT_CONC_RATIO, which are inhibitor-to-Ki concentration ratios modulating a saturable transport term rather than indicators that a node was clamped. A new member is added per inhibited node symbol; do not overload an existing member with a different target.

DOSE_LY2510924_MGD (canonical for the administered daily LY2510924 dose)

• Description: Patient’s assigned once-daily subcutaneous LY2510924 dose, in mg/day. This is the DOSE LEVEL as a covariate rather than the amount of the current dosing record: apparent clearance is a function of the daily dose the patient is assigned to, so the column is constant within a dosing regimen and changes only if the assigned daily dose changes. Set to 0 mg/day for the control arms of Studies CXAB and CXAC, which received standard of care only.
• Units: mg/day
• Type: continuous
• Scope: specific
• Reference category: n/a – enters the decreasing-sigmoid dose dependence of apparent elimination clearance, CL/F = (WT/80.1)^bCL * (CLmin/F + CLdelta/F - CLdelta/F*(dose-1)/((dose50-1)+(dose-1))), which is truncated at 1 mg so doses at or below 1 mg/day all take the maximum clearance CLmin/F + CLdelta/F.
• Source aliases:
  • dose – the daily-dose symbol of Bihorel 2017 Eq. 1 (“the LY2510924 dose at which CL/F reaches half of its possible range”). Studied levels are 1, 2.5, 5, 10, 20 and 30 mg/day in the Study CXAA dose-escalation phase, 2.5 or 20 mg/day in its dose-confirmation phase, and 20 mg/day in Studies CXAB and CXAC.
• Example models: Bihorel_2017_LY2510924.R (the only dose-on-clearance covariate in the library; drives lcl_dosemin / lcl_dosespan / lcl_dose50).
• Notes: Member of the DOSE_<DRUG>_<UNITS> auto-approve family. LY2510924 has no INN, so the drug token is the development code. Distinct from every other member of the family in what it does: the sibling DOSE_*_MGD columns feed an exposure metric that drives a PD effect, whereas this one modifies a PK disposition parameter directly. A dose-dependent clearance is a genuine structural nonlinearity that happens to be parameterised on the dose axis rather than the concentration axis; the authors chose it over a Michaelis-Menten elimination because the phase 1 noncompartmental analysis showed clearance falling with dose while distribution volumes stayed flat.

DOSE_AMOXICILLIN_UG (canonical for the administered single amoxicillin dose in micrograms)

• Description: Amount of the current single amoxicillin dose, in ug per animal (not per kg). Supplied as a covariate column, alongside the dosing event itself, for models in which a structural parameter depends on the dose level.
• Units: ug
• Type: continuous
• Scope: specific
• Reference category: n/a – enters an additive, MPLA-gated dose dependence of apparent clearance in the founding model, CL/F = 124 + (-0.145) * DOSE_AMOXICILLIN_UG * CONMED_MPLA mL/h (Franck 2021 Supplementary Eq. 1, Table S1).
• Source aliases:
  • DOSE – Franck 2021 Supplementary Eq. 1 (‘P = theta1 + theta2 * DOSE’). The paper states doses in mg/kg (0.2, 0.4, 1.2 and 14 mg/kg oral gavage); its reported clearances (73.3 mL/h at 14 mg/kg and 123 mL/h at 0.4 mg/kg with MPLA) reproduce exactly with the mg/kg dose times a 25 g body weight, i.e. 350 ug and 10 ug.
• Example models: Franck_2021_amoxicillin_mouse.R, Franck_2021_amoxicillin_mpla_mouse.R.
• Notes: Member of the DOSE_<DRUG>_<UNITS> auto-approve family. The ug unit follows the source’s clearance-slope unit (mL/h/ug) so the coefficient can be transcribed without rescaling; a mouse dose in mg/kg converts as mg/kg * body weight (g).

STUDY_CXAA (canonical for the LY2510924 phase 1 Study I2V-MC-CXAA indicator)

• Description: 1 = subject enrolled in Study I2V-MC-CXAA, the phase 1 dose-escalation and dose-confirmation study of LY2510924 in advanced metastatic cancer (richly sampled, doses 1-30 mg/day); 0 = subject enrolled in one of the two phase 2 studies, I2V-MC-CXAB or I2V-MC-CXAC. Because CXAA is the only phase 1 study of the pooled analysis, this indicator is exactly the phase 1 / phase 2 split.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (phase 2 Studies I2V-MC-CXAB and I2V-MC-CXAC).
• Source aliases:
  • “phase I studies” versus “phase II studies” – Bihorel 2017 Results (“RV was estimated using separate additive plus CCV error models for phase I and phase II studies”) and the Table 2 rows Phase 1 RV / Phase 2 RV.
• Example models: Bihorel_2017_LY2510924.R (selects between the two LY2510924 residual-error models: additive 0.496 ng/mL with proportional 0.222 for CXAA, against additive 1.458 ng/mL with proportional 0.406 for the phase 2 studies).
• Notes: Member of the STUDY_<id> auto-approve family. Purely a residual-error stratum: no structural PK or PD parameter depends on it. The magnitude difference is a sampling-scheme artefact rather than a population difference – CXAA used a rich scheme with eight samples on day 1 and day 28, while CXAB and CXAC collected two or three sparse samples per cycle with actual sampling times relative to dosing known less precisely. Paired with STUDY_CXAC in the same model, which is a structural covariate; a subject with both indicators 0 is in Study CXAB.

STUDY_CXAC (canonical for the LY2510924 phase 2 Study I2V-MC-CXAC indicator)

• Description: 1 = subject enrolled in Study I2V-MC-CXAC, the randomised phase 2 study in treatment-naive extensive-stage small cell lung carcinoma in which both arms received etoposide plus carboplatin standard of care; 0 = subject enrolled in Study I2V-MC-CXAA or I2V-MC-CXAB. Applies to BOTH arms of CXAC, including the control arm that received no LY2510924.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (Studies I2V-MC-CXAA and I2V-MC-CXAB).
• Source aliases:
  • FCXAC – Bihorel 2017 Eq. 5, “set to 1 after the first dose for patients enrolled in Study CXAC and zero otherwise”.
• Example models: Bihorel_2017_LY2510924.R (gates the empirical signal state, whose build-up drives an LY2510924-independent stimulation of CD34+ cell mobilisation).
• Notes: Member of the STUDY_<id> auto-approve family. Unusually for a study indicator this is a STRUCTURAL covariate, not a variance stratum: it switches on an entire ODE state. The authors are explicit that it is a composite and largely empirical stand-in – the CD34+ rise in the CXAC control arm could be due to the etoposide/carboplatin regimen, to concomitant G-CSF (used by 25-39% of CXAC patients versus none in CXAA or CXAB) or to erythropoietin, and “due to the differences in population and design across studies, the effects of SoC, cancer type, and study could not be distinguished” (Methods). Do not read it as a pure study effect, and do not reuse this canonical to carry a G-CSF or chemotherapy effect in another model – CONMED_<INN> is the family for a named co-medication. Paired with STUDY_CXAA in the same model. Because the indicator is described as taking effect “after the first dose” and the model’s signal state starts at zero at the time origin, the plain indicator reproduces FCXAC exactly when time is measured from the first dose.

STUDY_ATORVA_SUPPORT (canonical for the Stillemans 2022 atorvastatin support-dataset indicator)

• Description: 1 = subject from one of the two richly sampled support datasets pooled into the Stillemans 2022 atorvastatin popPK fit (Lemahieu 2005 healthy volunteers, control phase; Hermann 2006 patients with statin-induced myopathy and healthy controls); 0 = subject from the sparsely sampled real-life investigation cohort (ambulatory patients, NCT03604471). Subject-level (time-fixed).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (investigation cohort).
• Source aliases:
  • investigation / support – Stillemans 2022 Table 3 row labels (theta CL investigation, theta CL support).
• Example models: Stillemans_2022_atorvastatin.R (selects between two separately estimated CL/F typical values, 535 vs 400 L/h, AND two separate CL/F random effects; the SLCO1B1 c.521T>C effect applies only in the investigation cohort).
• Notes: Member of the STUDY_<id> auto-approve family. The authors split CL/F into one fixed/random-effect pair per cohort because only the investigation cohort had covariate data, so that covariate effects could be estimated without shifting the support-dataset estimates (Stillemans 2022 Results, Covariate analysis). The indicator therefore carries dataset-level differences (design, assay, population) rather than a single physiological factor; set it to 0 to simulate the real-life ambulatory population the paper is about.

STUDY_REACH (canonical for the REACH rituximab phase III study indicator)

• Description: 1 = record from REACH (BO17072, NCT00090051), the phase III trial of fludarabine and cyclophosphamide with or without intravenous rituximab in previously treated chronic lymphocytic leukemia; 0 = record from any other pooled study.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (SAWYER part 2 in Gibiansky 2021).
• Source aliases:
  • ST17072 – Gibiansky 2021 Supplementary Table S5, IF(STU.EQ.17072) ST17072 = 1.
• Example models: Gibiansky_2021_rituximab.R (founding example; multiplies the concentration-dependent log-scale residual SD by 1.42).
• Notes: Member of the STUDY_<id> auto-approve family. In the founding model it is a residual-error stratum only; no structural or covariate parameter differs by study. Set 0 when simulating new subjects.

STUDY_SAWYER_PART1 (canonical for the SAWYER rituximab phase Ib part 1 dose-finding cohort indicator)

• Description: 1 = record from part 1 (the dose-finding cohort) of SAWYER (BO25341, NCT01292603), in which previously untreated chronic lymphocytic leukemia patients received intravenous rituximab in cycles 1-5 and a single subcutaneous dose of 1400-2200 mg in cycle 6; 0 = record from SAWYER part 2 or from another pooled study.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (SAWYER part 2 in Gibiansky 2021).
• Source aliases:
  • COHA – Gibiansky 2021 Supplementary Table S5, IF(STU.EQ.25341.AND.ARM.EQ.1) COHA = 1; Table 3 labels its effect “effect of SAWYER Part 1 on residual error”.
• Example models: Gibiansky_2021_rituximab.R (founding example; multiplies the concentration-dependent log-scale residual SD by 0.568).
• Notes: Member of the STUDY_<id> auto-approve family. A residual-error stratum only in the founding model. Part 2 records (the randomised i.v.-versus-s.c. comparison) take 0 and are the reference. Set 0 when simulating new subjects.

STUDY_EXPAT (canonical for the EXPAT ciprofloxacin ICU study indicator)

• Description: 1 = record from the EXPAT study (Erasmus University Medical Center and Maasstad Hospital, Rotterdam; published as Abdulla 2020, doi:10.1007/s00228-020-02873-5), the “study 1” of the three-study pooled ciprofloxacin ICU analysis of Guo 2022; 0 = record from one of the other two pooled studies. Per-subject (study-fixed) binary indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 with STUDY_RDRN also 0, which selects study 2 of Guo 2022 (Radboud University Medical Center, Nijmegen; Gieling 2020), the reference stratum.
• Source aliases:
  • Study 1 – Guo 2022 Table 1 and Table 2 (‘Add Study1’, ‘Prop Study1’).
• Example models: Guo_2022_ciprofloxacin.R (founding example; selects the combined residual error, additive 0.151 mg/L + proportional 17.5%, in place of the study 2 proportional 13.7%).
• Notes: Member of the STUDY_<id> auto-approve family, named for the trial rather than the paper’s positional “study 1” because the positional label is meaningless outside that paper. A residual-error stratum only: Guo 2022 deliberately did not test study as a covariate on any structural parameter. Mutually exclusive with STUDY_RDRN. The same cohort is the sole data set of Abdulla_2020_ciprofloxacin.R, which does not need the indicator.

STUDY_RDRN (canonical for the Right Dose Right Now ICU study indicator)

• Description: 1 = record from the Right Dose Right Now study (Amsterdam UMC location VU University Medical Center; model-based bedside antibiotic dosing in severe sepsis and septic shock, Roggeveen 2019), the “study 3” of the three-study pooled ciprofloxacin ICU analysis of Guo 2022; 0 = record from one of the other two pooled studies. Per-subject (study-fixed) binary indicator.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 with STUDY_EXPAT also 0, which selects study 2 of Guo 2022 (Radboud University Medical Center, Nijmegen; Gieling 2020), the reference stratum.
• Source aliases:
  • Study 3 – Guo 2022 Table 1 and Table 2 (‘Prop Study3’).
• Example models: Guo_2022_ciprofloxacin.R (founding example; selects the proportional residual error 24.5% in place of the study 2 value 13.7%).
• Notes: Member of the STUDY_<id> auto-approve family. A residual-error stratum only, like STUDY_EXPAT. Mutually exclusive with STUDY_EXPAT.

FORM_OLZ_F0 (canonical for the olanzapine ‘formulation #0’ product indicator)

• Description: 1 = the dose was given as the olanzapine product Li 2018 labels “formulation #0”, 0 = any other olanzapine product. Per-dose-record indicator: the founding study’s healthy-volunteer cohort received formulations #0 and #1 in a two-period crossover with a 3-week washout, so one subject contributes records under both. The three products are oral tablets from three different manufacturers, anonymised by the paper as “formulation #0”, “#1” and “#2”.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0. FORM_OLZ_F0 = FORM_OLZ_F1 = 0 selects formulation #2, the model reference; there is deliberately no FORM_OLZ_F2 column, matching the paper, whose Equations (7) and (9) carry no formulation multiplier.
• Source aliases:
  • DRUG0 – Li 2018 Table 2 rows “DRUG0 ON Ka”, “DRUG0 ON Vc/F” and “DRUG0 ON CL/F”. The paper’s Table 1 “Drug type” row codes the three products as 0, 1 and 2; DRUG0 is the binarisation of that code.
• Example models: Li_2018_olanzapine.R (multiplicative effects on all three of the absorption rate constant, the apparent central volume and the apparent oral clearance: ka * 1.89, Vc/F * 0.216, CL/F * 0.610; founding example).
• Notes: Member of the FORM_<drug>_<formulation> family, and the closest structural sibling of the FORM_ELA_F1..FORM_ELA_F6 set – both register anonymised numbered clinical products with one indicator per non-reference formulation. The OLZ drug token is the standard olanzapine abbreviation; the library’s other olanzapine covariate, DOSE_TV44749_MG, is keyed to a development code rather than to the INN, so no abbreviation conflict arises. Interpretation caution, load-bearing: in the founding study formulation #2 was given exclusively to the 234 therapeutic-drug-monitoring patients and formulations #0/#1 exclusively to the 22 healthy male volunteers, so the fitted formulation effects on Vc/F and CL/F are completely confounded with cohort, study design and sampling density. Li 2018 says so explicitly (“the influences of formulation and population on PK parameters were mixed and could not be distinguished due to data limitations”, Discussion) and repeats the caveat in the Conclusion. Treat the column as a product-and-population stratum indicator; it is not evidence of a bioavailability difference between the tablets. The effect on ka is the one the authors consider comparatively well supported, because formulations #0 and #1 were both studied with rich absorption-phase sampling in the same subjects.

FORM_OLZ_F1 (canonical for the olanzapine ‘formulation #1’ product indicator)

• Description: 1 = the dose was given as the olanzapine product Li 2018 labels “formulation #1”, 0 = any other olanzapine product. Per-dose-record indicator, for the same crossover reason given under FORM_OLZ_F0.
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (formulation #2 when both FORM_OLZ_* indicators are 0).
• Source aliases:
  • DRUG1 – Li 2018 Table 2 rows “DRUG1 ON Vc/F” and “DRUG1 ON CL/F”.
• Example models: Li_2018_olanzapine.R (multiplicative effects on the apparent central volume and the apparent oral clearance only: Vc/F * 0.207, CL/F * 0.610).
• Notes: Member of the FORM_<drug>_<formulation> family; sibling of FORM_OLZ_F0, whose Notes carry the confounding caveat that applies equally here. Asymmetric with FORM_OLZ_F0 in one respect worth preserving when re-using the column: Table 2 has no “DRUG1 ON Ka” row, because the final model lets formulation #1 share the reference absorption rate constant while giving formulation #0 its own. The Discussion justifies that choice on an objective-function comparison (4126.82 vs 4133.16) and notes that the sparse, mostly post-absorption sampling of the patient cohort could not support a separate ka for formulation #2 either.

STUDY_FIH_MAD (canonical for the GLPG1690 first-in-human multiple-ascending-dose study-part indicator)

• Description: 1 = record from the multiple-ascending-dose (MAD) part of the GLPG1690 first-in-human study NCT02179502 (healthy male volunteers; 150 mg twice daily, 600 mg or 1000 mg once daily for 14 days), 0 = any other study or study part. Subject-level (time-fixed).
• Units: (binary)
• Type: binary
• Scope: specific
• Reference category: 0 (single-ascending-dose part, drug-drug-interaction study and proof-of-concept study in Taneja 2019).
• Source aliases:
  • FIH study, MAD part – Taneja 2019 Table 3 row ‘Imax at day 14 and later of the FIH study, MAD part’. Free-text label, not a column header.
• Example models: Taneja_2019_glpg1690.R (founding example; enters only in the product STUDY_FIH_MAD * DAY14, which adds 0.2043 to the probit of Imax from day 14 of the MAD part, 90.8% -> 93.7%).
• Notes: Member of the STUDY_<id> auto-approve family. The study gate is needed because the 12-week proof-of-concept study also samples past day 14 without carrying the effect.