Skip to contents

Gefapixant (Chawla 2023)

Source: vignettes/articles/Chawla_2023_gefapixant.Rmd
Chawla_2023_gefapixant.Rmd

Model and source

  • Citation: Chawla A, Largajolli A, Hussain A, et al. Population pharmacokinetic analysis of the P2X3-receptor antagonist gefapixant. CPT Pharmacometrics Syst Pharmacol. 2023;12(8):1107-1118. doi:10.1002/psp4.12978
  • Description: Two-compartment population PK model with first-order absorption and an absorption lag time for the P2X3-receptor antagonist gefapixant in healthy volunteers and adults with refractory or unexplained chronic cough (Chawla 2023)
  • Article: https://doi.org/10.1002/psp4.12978
  • Supplement (Tables S1-S6, Figures S1-S4): https://doi.org/10.1002/psp4.12978, Supporting Information

Gefapixant is an orally administered P2X3-receptor antagonist approved in Japan and Switzerland for refractory chronic cough (RCC) and unexplained chronic cough (UCC). Chawla 2023 developed a population PK model from pooled phase I, IIb, and III data in order to quantify between- and within-participant variability and to support dosage recommendations in renal impairment.

The structural model is two-compartment with first-order absorption and an absorption lag time. All disposition parameters are apparent (/F); no absolute bioavailability term is identifiable from oral-only data, so relative bioavailability is fixed at 1 across formulations.

Population

The analysis data set pooled nine studies: six intensively sampled phase I healthy-volunteer studies (including a dedicated renal-impairment study, a Japanese food-effect study, and a food-and-formulation study), one phase IIb study in RCC/UCC conducted in the United Kingdom and United States, and the two global phase III studies COUGH-1 and COUGH-2 (Chawla 2023 Tables S1 and S2).

Of 1677 participants included, 1618 had evaluable PK data (121 healthy volunteers and 1497 with RCC or UCC), contributing 8886 measurable gefapixant concentrations after exclusions for inconsistent dosing times, an extra dose within 3 days of sampling, measurable pre-first-dose concentrations, missing dose times, and |CWRES| > 5 outliers on the newly added phase II/III data (Chawla 2023 Figure 2).

Baseline demographics (Chawla 2023 Table S3): median age 59 years (range 18-89), median body weight 74 kg (range 35-159), median eGFR 86.9 mL/min/1.73 m^2 (range 13-243), 71.3% female, 78.5% White / 8.3% Asian / 3.6% Black / 9.5% Other, and 17.0% Hispanic. Doses spanned 7.5-150 mg b.i.d. in phase I, up to 50 mg b.i.d. in phase IIb, and 15 and 45 mg b.i.d. in phase III. Among the 1555 chronic-cough participants, 664 had normal renal function, 817 mild renal impairment (RI), and 74 moderate RI; no phase II/III participant had severe RI, and individuals with end-stage renal disease or on hemodialysis were excluded.

The same information is available programmatically via readModelDb("Chawla_2023_gefapixant")()$population.

Source trace

Every ini() entry in inst/modeldb/specificDrugs/Chawla_2023_gefapixant.R carries an in-file comment naming its source location. They are collected here for review. All point estimates come from Chawla 2023 Table 1; the centering values and categorical reference categories come from Table 1 footnote b.

Equation / parameter Value Source location
lka (Ka) 2.25 1/h Table 1, “Absorption rate constant (Ka)”; 95% CI 1.86-2.85, RSE 9.9%
lcl (CL/F) 10.3 L/h Table 1, “Apparent clearance (CL/F)”; 95% CI 10.1-10.5, RSE 1.1%
lvc (Vc/F) 101 L Table 1, “Apparent central volume of distribution (Vc/F)”; 95% CI 96.9-104, RSE 1.8%
lq (Q/F) 3.51 L/h Table 1, “Apparent intercompartmental clearance (Q/F)”; 95% CI 2.7-4.42, RSE 12.8%
lvp (Vp/F) 32.8 L Table 1, “Apparent peripheral volume (Vp/F)”; 95% CI 26.9-46.6, RSE 12.7%
ltlag (ALAG) 0.432 h Table 1, “Absorption lag time (ALAG)”; 95% CI 0.415-0.445, RSE 1.7%
e_crcl_cl 0.375 Table 1, “eGFR power relationship on clearance”; RSE 8.4%
e_age_cl -0.229 Table 1, “Age power relationship on clearance”; RSE 13.3%
e_wt_cl 0.35 Table 1, “Weight power relationship on clearance”; RSE 11.3%
e_sex_cl 0.0931 Table 1, “Sex relationship on clearance”; RSE 22.7%
e_age_vc 0.0911 Table 1, “Age power relationship on central volume”; RSE 37.1%
e_wt_vc 0.541 Table 1, “Weight power relationship on central volume”; RSE 8.4%
e_sex_vc 0.181 Table 1, “Sex relationship on central volume”; RSE 13.6%
e_fed_ka -0.594 Table 1, “Food effect on absorption rate for the F02 formulation”; RSE 7.7%
etalka variance 0.551 Table 1, “Interindividual variability on absorption rate (omega^2 Ka)”; 85.7% CV, RSE 18.7%
etalcl variance 0.0708 Table 1, “Interindividual variability on apparent clearance (omega^2 CL)”; 27.1% CV, RSE 6.7%
etalcl-etalvc covariance 0.0176 Table 1, “Covariance between CL/F and Vc/F”; RSE 22.9%
etalvc variance 0.0161 Table 1, “Interindividual variability on apparent central volume (omega^2 Vc)”; 12.7% CV, RSE 41.8%
propSd 0.303 Table 1, “Proportional residual error (sigma PROP)”; RSE 2.3%
addSd 3.04 ng/mL Table 1, “Additive residual error (sigma ADD)”; RSE 15.4%
Centering: eGFR 87.2, BW 74 kg, age 59 y n/a Table 1 footnote b
Reference categories: sex = female, fed = fasted n/a Table 1 footnote b
Continuous covariate equation P_i = P_TV * (cov_i / median(cov))^theta * exp(eta_i) n/a Methods, “Modeling strategy”, first displayed equation
Categorical covariate equation P_i = P_TV * (1 + theta_cov,c * I_cov,c,i) * exp(eta_i) n/a Methods, “Modeling strategy”, second displayed equation
Two-compartment structure with lag-time absorption n/a Results, “Concentration-time profiles” (biphasic elimination in Figure 3a) and Table 1 parameter set
Food effect restricted to the F02 formulation n/a Methods, “Modeling strategy”; Discussion, formulations paragraph

Reading the residual-error and IIV scales

Chawla 2023 Table 1 distinguishes the two scales by symbol, and both readings were checked against the table’s own footnotes before transcription:

  • The IIV entries are labelled omega^2 (variances). Footnote d gives %CV = sqrt(exp(omega^2) - 1) * 100; e.g. omega^2 Ka = 0.551 yields sqrt(exp(0.551) - 1) = 0.857, matching the reported 85.7% CV. They are therefore entered in ini() as variances, which is what nlmixr2 expects.
  • The residual-error entries are labelled sigma (standard deviations). The additive term carries ng/mL units (a variance would be ng2/mL2) and the proportional term 0.303 corresponds to 30.3% CV, consistent with the paper describing gefapixant as “a low- to moderate-variability drug”. They are therefore entered as propSd / addSd.

The CL/F-Vc/F term is labelled “Covariance between CL/F and Vc/F”. Read as a covariance, 0.0176 implies a correlation of 0.0176 / sqrt(0.0708 * 0.0161) = 0.52, consistent with the Results narrative that IIV was retained on both parameters because of their correlation. Read instead as a correlation it would imply a near-zero covariance, contradicting that narrative. The resulting 2x2 block is positive definite (0.0708 * 0.0161 = 1.14e-3 > 0.0176^2 = 3.10e-4).

mod <- readModelDb("Chawla_2023_gefapixant")
mod
#> function() {
#>   description <- "Two-compartment population PK model with first-order absorption and an absorption lag time for the P2X3-receptor antagonist gefapixant in healthy volunteers and adults with refractory or unexplained chronic cough (Chawla 2023)"
#>   reference <- "Chawla A, Largajolli A, Hussain A, et al. Population pharmacokinetic analysis of the P2X3-receptor antagonist gefapixant. CPT Pharmacometrics Syst Pharmacol. 2023;12(8):1107-1118. doi:10.1002/psp4.12978"
#>   vignette <- "Chawla_2023_gefapixant"
#>   units <- list(time = "h", dosing = "mg", concentration = "ng/mL")
#> 
#>   # Issue #482: what each ODE state holds, in what amount units, in what
#>   # biological matrix. Derived mechanically; verified = FALSE means it has
#>   # NOT been checked against the source paper.
#>   compartmentData <- list(
#>     depot       = list(analyte = "gefapixant", units = "mg", specimen = "administration site", verified = FALSE),
#>     central     = list(analyte = "gefapixant", units = "mg", specimen = "plasma", verified = FALSE),
#>     peripheral1 = list(analyte = "gefapixant", units = "mg", specimen = "plasma", verified = FALSE)
#>   )
#> 
#>   covariateData <- list(
#>     CRCL = list(
#>       description        = "Baseline estimated glomerular filtration rate (eGFR)",
#>       units              = "mL/min/1.73 m^2",
#>       type               = "continuous",
#>       reference_category = NULL,
#>       notes              = paste(
#>         "Creatinine-based estimated GFR, BSA-normalized. Power effect on CL/F centered on the",
#>         "data median of 87.2 mL/min/1.73 m^2 (Chawla 2023 Table 1 footnote b). Added to CL/F as",
#>         "part of base-model development (not via the stepwise covariate method) because gefapixant",
#>         "is primarily renally eliminated, and retained through the final model. A sensitivity",
#>         "analysis substituting a power relationship of creatinine clearance for eGFR slightly",
#>         "worsened the objective function value and was not retained. Cohort median 86.9",
#>         "(range 13-243) mL/min/1.73 m^2 per Table S3. Individuals with end-stage renal disease or",
#>         "on hemodialysis were excluded, so the eGFR-CL/F relationship is not validated below the",
#>         "severe-RI range."
#>       ),
#>       source_name        = "eGFR"
#>     ),
#>     WT = list(
#>       description        = "Baseline body weight",
#>       units              = "kg",
#>       type               = "continuous",
#>       reference_category = NULL,
#>       notes              = paste(
#>         "Power effects on both CL/F and Vc/F centered on the data median of 74 kg",
#>         "(Chawla 2023 Table 1 footnote b). Cohort median 74 kg (range 35-159) per Table S3.",
#>         "Body mass index was deliberately not evaluated because of its high correlation with",
#>         "body weight. Fixed allometric scaling on CL/F and the volume parameters was tested as a",
#>         "sensitivity analysis and gave no improvement over these estimated exponents."
#>       ),
#>       source_name        = "BW"
#>     ),
#>     AGE = list(
#>       description        = "Baseline age",
#>       units              = "years",
#>       type               = "continuous",
#>       reference_category = NULL,
#>       notes              = paste(
#>         "Power effects on both CL/F and Vc/F centered on the data median of 59 years",
#>         "(Chawla 2023 Table 1 footnote b). Cohort median 59 years (range 18-89) per Table S3.",
#>         "The age effect on Vc/F was the least precisely estimated fixed effect in the final",
#>         "model (RSE 37.1%)."
#>       ),
#>       source_name        = "AGE"
#>     ),
#>     SEXF = list(
#>       description        = "Biological sex indicator (1 = female, 0 = male)",
#>       units              = "(binary)",
#>       type               = "binary",
#>       reference_category = "1 (female) -- inverted relative to the canonical SEXF reference",
#>       notes              = paste(
#>         "Chawla 2023 uses a MALE indicator with FEMALE as the reference category (Table 1",
#>         "footnote b: 'Categorical covariate references: sex = female'), which is the inverse of",
#>         "the canonical SEXF orientation (1 = female, reference 0 = male). To store the column",
#>         "under the canonical SEXF name while reproducing the paper's published CL/F = 10.3 L/h",
#>         "and Vc/F = 101 L verbatim (both of which are female-reference typical values), the",
#>         "effects are applied in model() through sex_male <- 1 - SEXF, so SEXF = 1 gives a factor",
#>         "of exactly 1 and SEXF = 0 gives the paper's male-vs-female fractional change. Same",
#>         "pattern as Bajaj_2017_nivolumab.R. Cohort 71.3% female per Table S3."
#>       ),
#>       source_name        = "SEX (male indicator)"
#>     ),
#>     FED = list(
#>       description        = "Fed-vs-fasted state at the time of dosing (1 = fed, 0 = fasted)",
#>       units              = "(binary)",
#>       type               = "binary",
#>       reference_category = "0 (fasted)",
#>       notes              = paste(
#>         "General fed-vs-fasted dose-record flag, not a high-fat-meal challenge, so FED applies",
#>         "rather than FED_HIGHFAT. Fractional-change effect on Ka, and applied ONLY to records",
#>         "carrying the F02 formulation (see FORM_GEF_F02): dedicated phase I relative-bioavailability",
#>         "studies had already shown that fed status affects gefapixant exposure for F02 but not for",
#>         "F04, so food effects were tested for F02 only. Food effects on absorption lag time and on",
#>         "bioavailability were also tested during base-model development but were not retained",
#>         "(nonsignificant objective-function drop or poor precision)."
#>       ),
#>       source_name        = "FED"
#>     ),
#>     FORM_GEF_F02 = list(
#>       description        = "Gefapixant F02 (wet-granulation, citric-acid-containing, film-coated immediate-release tablet) formulation indicator",
#>       units              = "(binary)",
#>       type               = "binary",
#>       reference_category = "0 (F04 / F04A gefapixant-citrate formulations)",
#>       notes              = paste(
#>         "1 = F02, the earlier wet-granulation immediate-release tablet containing citric acid as",
#>         "an acidulant (7.5, 20, and 50 mg strengths), used in several phase I studies and the",
#>         "phase IIb chronic-cough study. 0 = F04 (20A film coating, two phase I studies) or F04A",
#>         "(03K film coating, COUGH-1 and COUGH-2), both of which contain gefapixant citrate as the",
#>         "active ingredient. This indicator carries NO main effect on any PK parameter in the final",
#>         "model -- no relative-bioavailability difference between F02 and F04/F04A was retained. It",
#>         "exists solely to gate the FED effect on Ka to F02 records. Formulation assignment by study",
#>         "is tabulated in Chawla 2023 Table S2. The marketed F04B formulation (F04A without citric",
#>         "acid) is bioequivalent to F04A, so F04B records take FORM_GEF_F02 = 0."
#>       ),
#>       source_name        = "FORM"
#>     )
#>   )
#> 
#>   covariatesDataExcluded <- list(
#>     CONMED_PPI = list(
#>       description = "Concomitant proton-pump inhibitor (omeprazole) use",
#>       units       = "(binary)",
#>       type        = "binary",
#>       notes       = paste(
#>         "Tested during base-model development on absorption rate constant, absorption lag time,",
#>         "and relative bioavailability for the F02 formulation only (dedicated phase I studies had",
#>         "shown no PPI effect on F04). No PPI relationship reached the retention criteria, so PPI",
#>         "use appears nowhere in the final model. Chawla 2023 Results, 'Model development'."
#>       )
#>     ),
#>     RACE_WHITE = list(
#>       description = "White race indicator",
#>       units       = "(binary)",
#>       type        = "binary",
#>       notes       = paste(
#>         "Race was screened on CL/F and Vc/F in the stepwise covariate method and was statistically",
#>         "significant on CL/F in the integrated phase I-III analysis, but the signal was driven",
#>         "entirely by the 'multiple' race category (effect size 0.19 vs ~0.05 for every other",
#>         "category) which comprised only ~5% of the target population. 'Multiple' was therefore",
#>         "merged into 'other' and the race effect was removed from the final model as not clinically",
#>         "relevant. Chawla 2023 Results and Discussion; no retained point estimate exists."
#>       )
#>     ),
#>     RACE_ASIAN = list(
#>       description = "Asian race indicator",
#>       units       = "(binary)",
#>       type        = "binary",
#>       notes       = "Screened with the other race indicators; race removed from the final model. See RACE_WHITE note."
#>     ),
#>     RACE_BLACK = list(
#>       description = "Black / African American race indicator",
#>       units       = "(binary)",
#>       type        = "binary",
#>       notes       = "Screened with the other race indicators; race removed from the final model. See RACE_WHITE note."
#>     ),
#>     RACE_OTHER = list(
#>       description = "Race-category 'Other' indicator (includes American Indian or Alaskan Native, multiple or other, and Native Hawaiian or Pacific Islander)",
#>       units       = "(binary)",
#>       type        = "binary",
#>       notes       = paste(
#>         "Screened with the other race indicators; the 'multiple' category was merged into 'other'",
#>         "before race was dropped from the final model. See RACE_WHITE note."
#>       )
#>     ),
#>     RACE_HISPANIC = list(
#>       description = "Hispanic / Latino ethnicity indicator",
#>       units       = "(binary)",
#>       type        = "binary",
#>       notes       = paste(
#>         "Ethnicity was screened on CL/F and Vc/F in the stepwise covariate method and was not",
#>         "retained in the final model. Cohort 17.0% Hispanic per Chawla 2023 Table S3."
#>       )
#>     )
#>   )
#> 
#>   population <- list(
#>     species        = "human",
#>     n_subjects     = 1618,
#>     n_studies      = 9,
#>     age_range      = "18-89 years",
#>     age_median     = "59 years",
#>     weight_range   = "35-159 kg",
#>     weight_median  = "74 kg",
#>     sex_female_pct = 71.3,
#>     race_ethnicity = c(White = 78.5, Asian = 8.3, Black = 3.6, Other = 9.5),
#>     disease_state  = "Healthy volunteers (n = 121) and adults with refractory or unexplained chronic cough (n = 1497)",
#>     dose_range     = "7.5-150 mg oral b.i.d. (phase I, single and multiple dose); up to 50 mg b.i.d. (phase IIb); 15 and 45 mg b.i.d. (phase III COUGH-1 and COUGH-2)",
#>     regions        = "Global; United Kingdom and United States (phase IIb), Japan (one phase I study), multinational (phase III)",
#>     renal_function = paste(
#>       "eGFR median 86.9 mL/min/1.73 m^2 (range 13-243). Among the 1555 chronic-cough participants,",
#>       "664 had normal renal function, 817 mild renal impairment, and 74 moderate renal impairment;",
#>       "no phase II/III participant had severe renal impairment, which is represented only by a",
#>       "dedicated phase I renal-impairment study. Individuals with end-stage renal disease and",
#>       "individuals on hemodialysis were excluded."
#>     ),
#>     formulations   = "F02, F04, and F04A; the earlier F01 formulation was excluded",
#>     notes          = paste(
#>       "Baseline demographics from Chawla 2023 Table S3; per-study participant counts and",
#>       "formulation / food / PPI assignments from Table S2; study designs from Table S1.",
#>       "1677 participants were included in the data set and 1618 had evaluable PK data",
#>       "(8886 measurable concentrations) after exclusions for inconsistent dosing times, an extra",
#>       "dose within 3 days before sampling, measurable pre-first-dose concentrations, missing dose",
#>       "times, and |CWRES| > 5 outliers on the newly added phase II/III data."
#>     )
#>   )
#> 
#>   ini({
#>     # Structural parameters. CL/F and Vc/F are typical values at the reference
#>     # covariate set: eGFR 87.2 mL/min/1.73 m^2, age 59 y, body weight 74 kg, female.
#>     # Ka is the typical value under the fasted-state reference.
#>     lka   <- log(2.25);  label("Absorption rate constant (Ka, 1/h), fasted reference")                              # Chawla 2023 Table 1: Ka = 2.25 (95% CI 1.86, 2.85), RSE 9.9%
#>     lcl   <- log(10.3);  label("Apparent clearance (CL/F, L/h) at reference eGFR, age, weight, and female sex")     # Chawla 2023 Table 1: CL/F = 10.3 (10.1, 10.5), RSE 1.1%
#>     lvc   <- log(101);   label("Apparent central volume of distribution (Vc/F, L) at reference age, weight, sex")   # Chawla 2023 Table 1: Vc/F = 101 (96.9, 104), RSE 1.8%
#>     lq    <- log(3.51);  label("Apparent intercompartmental clearance (Q/F, L/h)")                                  # Chawla 2023 Table 1: Q/F = 3.51 (2.7, 4.42), RSE 12.8%
#>     lvp   <- log(32.8);  label("Apparent peripheral volume of distribution (Vp/F, L)")                              # Chawla 2023 Table 1: Vp/F = 32.8 (26.9, 46.6), RSE 12.7%
#>     ltlag <- log(0.432); label("Absorption lag time (ALAG, h)")                                                     # Chawla 2023 Table 1: ALAG = 0.432 (0.415, 0.445), RSE 1.7%
#> 
#>     # Covariate effects on CL/F.
#>     # Continuous covariates enter as median-centered power functions and categorical
#>     # covariates as fractional changes, per the two Chawla 2023 Methods equations:
#>     #   P_i = P_TV * (cov_i / median(cov))^theta * exp(eta_i)
#>     #   P_i = P_TV * (1 + theta_cov,c * I_cov,c,i) * exp(eta_i)
#>     e_crcl_cl <- 0.375;  label("Power exponent of eGFR on CL/F, centered at 87.2 mL/min/1.73 m^2 (unitless)")       # Chawla 2023 Table 1: Cl_eGFR = 0.375 (0.317, 0.429), RSE 8.4%
#>     e_age_cl  <- -0.229; label("Power exponent of age on CL/F, centered at 59 years (unitless)")                    # Chawla 2023 Table 1: CL_AGE = -0.229 (-0.284, -0.171), RSE 13.3%
#>     e_wt_cl   <- 0.35;   label("Power exponent of body weight on CL/F, centered at 74 kg (unitless)")               # Chawla 2023 Table 1: CL_BW = 0.35 (0.27, 0.43), RSE 11.3%
#>     e_sex_cl  <- 0.0931; label("Fractional change in CL/F for male sex vs female reference (applied as (1 - SEXF))") # Chawla 2023 Table 1: CL_SEX = 0.0931 (0.0545, 0.133), RSE 22.7%
#> 
#>     # Covariate effects on Vc/F.
#>     e_age_vc  <- 0.0911; label("Power exponent of age on Vc/F, centered at 59 years (unitless)")                    # Chawla 2023 Table 1: Vc_AGE = 0.0911 (0.0239, 0.162), RSE 37.1%
#>     e_wt_vc   <- 0.541;  label("Power exponent of body weight on Vc/F, centered at 74 kg (unitless)")               # Chawla 2023 Table 1: Vc_BW = 0.541 (0.452, 0.627), RSE 8.4%
#>     e_sex_vc  <- 0.181;  label("Fractional change in Vc/F for male sex vs female reference (applied as (1 - SEXF))") # Chawla 2023 Table 1: Vc_SEX = 0.181 (0.138, 0.234), RSE 13.6%
#> 
#>     # Covariate effect on Ka. Applies only to F02-formulation records; dedicated
#>     # phase I studies showed no food effect for the F04 formulations.
#>     e_fed_ka  <- -0.594; label("Fractional change in Ka for the fed state, F02 formulation only (vs fasted reference)") # Chawla 2023 Table 1: Ka_FEED = -0.594 (-0.675, -0.496), RSE 7.7%
#> 
#>     # Interindividual variability. Chawla 2023 Table 1 reports these as variances
#>     # (omega^2) with footnote d giving %CV = sqrt(exp(omega^2) - 1) * 100.
#>     # CL/F and Vc/F IIV are correlated; the reported CL/F-Vc/F term is a covariance.
#>     etalcl + etalvc ~ c(0.0708,
#>                         0.0176, 0.0161)  # Chawla 2023 Table 1: omega^2_CL = 0.0708 (27.1% CV, RSE 6.7%); CL-Vc covariance = 0.0176 (RSE 22.9%); omega^2_Vc = 0.0161 (12.7% CV, RSE 41.8%)
#>     etalka          ~ 0.551              # Chawla 2023 Table 1: omega^2_Ka = 0.551 (85.7% CV, RSE 18.7%); informed by phase I data only
#> 
#>     # Residual error. Chawla 2023 Table 1 reports these as standard deviations
#>     # (sigma), not variances: the additive term carries ng/mL units and the
#>     # proportional term corresponds to 30.3% CV.
#>     propSd <- 0.303; label("Proportional residual error (fraction)")  # Chawla 2023 Table 1: sigma_PROP = 0.303 (0.29, 0.316), RSE 2.3%
#>     addSd  <- 3.04;  label("Additive residual error (ng/mL)")         # Chawla 2023 Table 1: sigma_ADD = 3.04 (2.3, 3.81) ng/mL, RSE 15.4%
#>   })
#>   model({
#>     # 1. Derived covariate terms.
#>     # Chawla 2023 uses a male indicator with female as the reference category, the
#>     # inverse of the canonical SEXF orientation; (1 - SEXF) reproduces the paper's
#>     # male = 1 column while keeping the published female-reference CL/F and Vc/F.
#>     sex_male <- 1 - SEXF
#> 
#>     # Food slows gefapixant absorption for the F02 formulation only.
#>     fed_ka <- 1 + e_fed_ka * FED * FORM_GEF_F02
#> 
#>     # 2. Individual parameters.
#>     ka   <- exp(lka + etalka) * fed_ka
#>     tlag <- exp(ltlag)
#>     cl   <- exp(lcl + etalcl) *
#>       (CRCL / 87.2)^e_crcl_cl *
#>       (AGE / 59)^e_age_cl *
#>       (WT / 74)^e_wt_cl *
#>       (1 + e_sex_cl * sex_male)
#>     vc   <- exp(lvc + etalvc) *
#>       (AGE / 59)^e_age_vc *
#>       (WT / 74)^e_wt_vc *
#>       (1 + e_sex_vc * sex_male)
#>     q    <- exp(lq)
#>     vp   <- exp(lvp)
#> 
#>     # 3. Micro-constants.
#>     kel <- cl / vc
#>     k12 <- q / vc
#>     k21 <- q / vp
#> 
#>     # 4. ODE system.
#>     d/dt(depot)       <- -ka * depot
#>     d/dt(central)     <-  ka * depot - kel * central - k12 * central + k21 * peripheral1
#>     d/dt(peripheral1) <-  k12 * central - k21 * peripheral1
#> 
#>     # 5. Absorption lag time.
#>     alag(depot) <- tlag
#> 
#>     # 6. Observation and error.
#>     # Dose in mg and volume in L give mg/L; * 1000 converts to ng/mL to match the
#>     # additive residual error units and the reported concentration scale.
#>     Cc <- 1000 * central / vc
#>     Cc ~ add(addSd) + prop(propSd)
#>   })
#> }
#> <environment: 0x559a46cf8ac8>

Virtual cohort

Original participant-level data are not publicly available. The cohorts below are virtual populations whose covariate distributions approximate the published demographics (Chawla 2023 Table S3). Each arm uses 100-200 participants.

set.seed(20230815)

gm <- function(x) exp(mean(log(x[is.finite(x) & x > 0])))

# Truncated-normal draw so sampled covariates stay inside the published ranges.
rtnorm <- function(n, mean, sd, lo, hi) {
  pmin(pmax(stats::rnorm(n, mean, sd), lo), hi)
}

# Chawla 2023 Table S3, RCC/UCC column.
make_cough_subjects <- function(n, id_offset = 0L, crcl = NULL) {
  tibble(
    id   = id_offset + seq_len(n),
    AGE  = rtnorm(n, 58.9, 11.8, 19, 89),
    WT   = rtnorm(n, 77.0, 18.0, 35, 159),
    SEXF = as.numeric(stats::runif(n) < 0.749),
    CRCL = if (is.null(crcl)) rtnorm(n, 90.7, 24.0, 34.5, 243) else crcl(n)
  )
}

# Chawla 2023 Table S3, healthy-volunteer column (phase I food/formulation study).
make_hv_subjects <- function(n, id_offset = 0L) {
  tibble(
    id   = id_offset + seq_len(n),
    AGE  = rtnorm(n, 38.6, 14.1, 18, 65),
    WT   = rtnorm(n, 74.0, 13.2, 50.9, 106),
    SEXF = as.numeric(stats::runif(n) < 0.262),
    CRCL = rtnorm(n, 99.4, 27.9, 60, 164)
  )
}

# Expand per-subject covariates into a dosing row (with ii/addl) plus an
# observation grid. Observation rows use cmt = "central", the ODE state, never
# the algebraic observable Cc.
make_events <- function(subjects, dose, ii, addl, obs_times,
                        fed = 0, f02 = 0, treatment) {
  dosing <- subjects |>
    mutate(time = 0, amt = dose, evid = 1L, cmt = "depot",
           ii = ii, addl = addl)
  obs <- subjects |>
    tidyr::crossing(time = obs_times) |>
    mutate(amt = NA_real_, evid = 0L, cmt = "central",
           ii = 0, addl = 0L)
  bind_rows(dosing, obs) |>
    mutate(FED = fed, FORM_GEF_F02 = f02, treatment = treatment) |>
    arrange(id, time, desc(evid))
}

Multiple-dose cohorts (45 mg b.i.d., F04A)

COUGH-1 and COUGH-2 used the F04A formulation, for which no food effect applies, so FORM_GEF_F02 = 0 and the FED covariate is inoperative for these arms. Dosing runs for 7 days (14 doses); with a terminal half-life near 12 h that is roughly 14 half-lives, so the final interval is at steady state.

tau      <- 12
n_days   <- 7
t_last   <- (n_days * 24) - tau        # 156 h: time of the final b.i.d. dose
obs_first <- seq(0, tau, by = 0.25)
obs_ss    <- seq(t_last, t_last + tau, by = 0.25)
obs_bid   <- unique(c(obs_first, obs_ss))

# Renal-function strata. The paper simulated these from a pooled Merck virtual
# RI database that is not published, so representative eGFR values are drawn
# uniformly within the conventional KDIGO/FDA bins (see Assumptions).
ri_bins <- tibble::tribble(
  ~treatment,        ~lo, ~hi,
  "Normal (>=90)",    90, 130,
  "Mild RI (60-89)",  60,  89,
  "Moderate RI (30-59)", 30, 59,
  "Severe RI (15-29)",   15, 29
)

# Renal-function strata: replicates the Figure 5a / Discussion AUC ratios.
make_ri_arm <- function(k) {
  b <- ri_bins[k, ]
  make_events(
    make_cough_subjects(
      100, id_offset = 1000L * k,
      crcl = function(n) stats::runif(n, b$lo, b$hi)
    ),
    dose = 45, ii = tau, addl = (n_days * 2L) - 1L,
    obs_times = obs_bid, treatment = b$treatment
  )
}

events_bid <- bind_rows(
  # Target phase II/III population, unstratified: replicates Table S6.
  make_events(make_cough_subjects(200, id_offset = 0L),
              dose = 45, ii = tau, addl = (n_days * 2L) - 1L,
              obs_times = obs_bid, treatment = "Phase II/III population"),
  bind_rows(lapply(seq_len(nrow(ri_bins)), make_ri_arm))
)

stopifnot(!anyDuplicated(unique(events_bid[, c("id", "time", "evid")])))

Severe renal impairment on 45 mg once daily

Chawla 2023 Figure 5b compares 45 mg q.d. in severe RI against 45 mg b.i.d. in normal renal function. This arm is simulated separately because its dosing interval (24 h) puts the steady-state window at a different time.

tau_qd    <- 24
t_last_qd <- (n_days - 1) * 24         # 144 h: time of the final q.d. dose
obs_qd    <- seq(t_last_qd, t_last_qd + tau_qd, by = 0.25)

events_qd <- make_events(
  make_cough_subjects(100, id_offset = 9000L,
                      crcl = function(n) stats::runif(n, 15, 29)),
  dose = 45, ii = tau_qd, addl = n_days - 1L,
  obs_times = obs_qd, treatment = "Severe RI, 45 mg q.d."
)

Single-dose F02 food-effect cohort

The phase I food-and-formulation study (MK-7264-025) gave healthy adults single 50 mg doses of F02 under fasted and fed conditions. This arm exercises the only extrinsic covariate retained in the final model, and its long washout also supports an NCA terminal-half-life estimate.

obs_sd <- c(seq(0, 12, by = 0.25), seq(13, 24, by = 1), seq(28, 72, by = 4))

events_sd <- bind_rows(
  make_events(make_hv_subjects(100, id_offset = 20000L),
              dose = 50, ii = 0, addl = 0L, obs_times = obs_sd,
              fed = 0, f02 = 1, treatment = "F02 fasted"),
  make_events(make_hv_subjects(100, id_offset = 21000L),
              dose = 50, ii = 0, addl = 0L, obs_times = obs_sd,
              fed = 1, f02 = 1, treatment = "F02 fed")
)

stopifnot(!anyDuplicated(unique(events_sd[, c("id", "time", "evid")])))

Simulation

Chawla 2023 generated its simulated exposures from the final model with interindividual variability but without residual error (Methods, “Simulations”: PK parameters were generated “by incorporating parameter uncertainty (only on fixed effects) and by accounting for IIV for each sampled participant”). The comparisons below therefore use the individual-prediction column Cc, which carries IIV but no residual error. Parameter uncertainty on the fixed effects is not propagated here; the published 95% CIs are narrow (for example CL/F 10.1-10.5 L/h), so its contribution to the geometric means is small.

keep_cols <- c("treatment", "AGE", "WT", "SEXF", "CRCL", "FED", "FORM_GEF_F02")

sim_bid <- rxode2::rxSolve(mod, events = events_bid, keep = keep_cols) |>
  as.data.frame()
#> ℹ parameter labels from comments will be replaced by 'label()'
sim_qd  <- rxode2::rxSolve(mod, events = events_qd,  keep = keep_cols) |>
  as.data.frame()
sim_sd  <- rxode2::rxSolve(mod, events = events_sd,  keep = keep_cols) |>
  as.data.frame()

nrow(sim_bid)
#> [1] 58800

Replicate published figures 

# Companion to Figure 3a of Chawla 2023: biphasic decline after a single oral
# dose, which is the paper's stated rationale for a two-compartment model.
sim_sd |>
  filter(treatment == "F02 fasted", !is.na(Cc), time > 0) |>
  group_by(time) |>
  summarise(Q05 = quantile(Cc, 0.05), Q50 = quantile(Cc, 0.50),
            Q95 = quantile(Cc, 0.95), .groups = "drop") |>
  ggplot(aes(time, Q50)) +
  geom_ribbon(aes(ymin = Q05, ymax = Q95), alpha = 0.25) +
  geom_line() +
  scale_y_log10() +
  labs(x = "Time (h)", y = "Gefapixant concentration (ng/mL)",
       title = "Single 50 mg F02 dose, fasted",
       caption = paste("Companion to Figure 3a of Chawla 2023: median with 5th-95th",
                       "percentiles; the change in slope is the biphasic",
                       "elimination motivating the two-compartment model."))
#> Warning in scale_y_log10(): log-10 transformation introduced infinite values.
#> log-10 transformation introduced infinite values.
#> log-10 transformation introduced infinite values.
#> log-10 transformation introduced infinite values.

# Replicates the Table 1 food effect on Ka (Ka_FEED = -0.594): food slows
# absorption for the F02 formulation, lowering and delaying the peak.
sim_sd |>
  filter(!is.na(Cc), time <= 12) |>
  group_by(treatment, time) |>
  summarise(Q50 = median(Cc), .groups = "drop") |>
  ggplot(aes(time, Q50, colour = treatment)) +
  geom_line(linewidth = 0.8) +
  labs(x = "Time (h)", y = "Median gefapixant concentration (ng/mL)",
       colour = NULL, title = "F02 food effect on absorption",
       caption = paste("Replicates the Chawla 2023 Table 1 food effect on Ka",
                       "(Ka_FEED = -0.594, a 59.4% reduction when fed)."))

# Replicates Figure 5b of Chawla 2023: severe RI on 45 mg q.d. versus normal
# renal function on 45 mg b.i.d., aligned on the final 24 h of dosing.
bind_rows(
  sim_bid |>
    filter(treatment == "Normal (>=90)", time >= 144, !is.na(Cc)) |>
    mutate(tad24 = time - 144),
  sim_qd |>
    filter(time >= 144, !is.na(Cc)) |>
    mutate(tad24 = time - 144)
) |>
  group_by(treatment, tad24) |>
  summarise(Q50 = median(Cc), .groups = "drop") |>
  ggplot(aes(tad24, Q50, colour = treatment)) +
  geom_line(linewidth = 0.8) +
  labs(x = "Time within final 24 h (h)", y = "Median concentration (ng/mL)",
       colour = NULL,
       title = "Severe RI 45 mg q.d. vs normal renal function 45 mg b.i.d.",
       caption = paste("Replicates Figure 5b of Chawla 2023: once-daily dosing in",
                       "severe RI avoids the second peak while delivering a",
                       "comparable total daily exposure."))

PKNCA validation

Steady state on 45 mg b.i.d.

Two intervals are requested: the first dosing interval (0-12 h) and the final one (156-168 h), which gives both AUCss,0-12 and the accumulation ratio.

sim_nca_bid <- sim_bid |>
  filter(!is.na(Cc)) |>
  select(id, time, Cc, treatment)

# Guarantee a time-zero record per subject (pre-dose Cc = 0 is correct for an
# extravascular model); existing rows win via .keep_all on the first occurrence.
sim_nca_bid <- bind_rows(
  sim_nca_bid,
  sim_nca_bid |> distinct(id, treatment) |> mutate(time = 0, Cc = 0)
) |>
  distinct(id, treatment, time, .keep_all = TRUE) |>
  arrange(id, treatment, time)

conc_bid <- PKNCA::PKNCAconc(sim_nca_bid, Cc ~ time | treatment + id,
                             concu = "ng/mL", timeu = "h")
dose_bid <- PKNCA::PKNCAdose(
  events_bid |> filter(evid == 1) |> select(id, time, amt, treatment),
  amt ~ time | treatment + id, doseu = "mg"
)

intervals_bid <- data.frame(
  start   = c(0, t_last),
  end     = c(tau, t_last + tau),
  cmax    = TRUE,
  tmax    = TRUE,
  cmin    = TRUE,
  auclast = TRUE,
  cav     = TRUE
)

nca_bid <- PKNCA::pk.nca(
  PKNCA::PKNCAdata(conc_bid, dose_bid, intervals = intervals_bid)
)

# PKNCA emits dependency rows alongside requested parameters, and both
# intervals share the same PPTESTCD codes, so filter on start/end as well.
res_bid <- as.data.frame(nca_bid$result) |>
  filter(PPTESTCD %in% c("cmax", "tmax", "cmin", "auclast", "cav"))

res_ss    <- res_bid |> filter(start == t_last)
res_first <- res_bid |> filter(start == 0)

Accumulation ratio 

acc <- inner_join(
  res_first |> filter(PPTESTCD == "auclast") |> select(id, treatment, auc_first = PPORRES),
  res_ss    |> filter(PPTESTCD == "auclast") |> select(id, treatment, auc_ss    = PPORRES),
  by = c("id", "treatment")
) |>
  mutate(ratio = auc_ss / auc_first)

acc_tbl <- tibble(
  Quantity = "Accumulation ratio (AUC0-12 at steady state / after first dose)",
  Published = "1.65 (95% CI 1.61-1.70)",
  Simulated = sprintf("%.2f", gm(acc$ratio[acc$treatment == "Phase II/III population"]))
)

knitr::kable(
  acc_tbl,
  caption = paste("Accumulation ratio, geometric mean. Published value from",
                  "Chawla 2023 Results, 'Exposures in target population'.")
)
Accumulation ratio, geometric mean. Published value from Chawla 2023 Results, ‘Exposures in target population’.
Quantity Published Simulated
Accumulation ratio (AUC0-12 at steady state / after first dose) 1.65 (95% CI 1.61-1.70) 1.66

Comparison against published steady-state exposures (Table S6)

Chawla 2023 Table S6 reports median geometric means across 200 simulation replicates for the phase III study population on 45 mg b.i.d.

sim_s6 <- res_ss |>
  filter(treatment == "Phase II/III population",
         PPTESTCD %in% c("auclast", "cmax", "cmin")) |>
  group_by(PPTESTCD) |>
  summarise(value = gm(PPORRES), .groups = "drop") |>
  tidyr::pivot_wider(names_from = PPTESTCD, values_from = value)

published_s6 <- tibble::tibble(auclast = 4144, cmax = 531, cmin = 201)

cmp_s6 <- nlmixr2lib::ncaComparisonTable(
  simulated = as.data.frame(sim_s6),
  reference = as.data.frame(published_s6),
  units = c(auclast = "ng*h/mL", cmax = "ng/mL", cmin = "ng/mL"),
  tolerance_pct = 20
)

knitr::kable(
  cmp_s6,
  caption = paste("Steady-state exposures on gefapixant 45 mg b.i.d. in the phase",
                  "III population: simulated geometric means vs Chawla 2023",
                  "Table S6. * differs from reference by >20%."),
  align = c("l", "r", "r", "r")
)
Steady-state exposures on gefapixant 45 mg b.i.d. in the phase III population: simulated geometric means vs Chawla 2023 Table S6. * differs from reference by >20%.
NCA parameter Reference Simulated % diff
Cmax (ng/mL) 531 543 +2.3%
Cmin (ng/mL) 201 202 +0.6%
AUClast (ng*h/mL) 4140 4260 +2.9% 
attr(cmp_s6, "footnote")
#> NULL

Derived parameters and half-life

Table S6 also reports CL/F, Vc/F, and a half-life. The clearance and volume are model parameters returned directly by rxSolve. The half-life needs care: the value Chawla 2023 reports is not the two-compartment terminal half-life but the one-compartment-style effective half-life computed from total apparent volume, ln(2) * (Vc/F + Vp/F) / (CL/F). Both readings are shown below.

# cl, vc, q, and vp are all assignments inside model(), so rxSolve returns them
# as columns -- no need to restate the Table 1 values here.
terminal_thalf <- function(cl, vc, q, vp) {
  k10 <- cl / vc; k12 <- q / vc; k21 <- q / vp
  s <- k10 + k12 + k21
  lambda2 <- (s - sqrt(s^2 - 4 * k10 * k21)) / 2
  log(2) / lambda2
}

subj_bid <- sim_bid |>
  filter(treatment == "Phase II/III population") |>
  distinct(id, cl, vc, q, vp) |>
  mutate(thalf_eff  = log(2) * (vc + vp) / cl,
         thalf_term = terminal_thalf(cl, vc, q, vp))

# NCA terminal half-life from the single-dose fasted arm (72 h washout).
conc_sd <- PKNCA::PKNCAconc(
  sim_sd |> filter(!is.na(Cc)) |> select(id, time, Cc, treatment),
  Cc ~ time | treatment + id, concu = "ng/mL", timeu = "h"
)
dose_sd <- PKNCA::PKNCAdose(
  events_sd |> filter(evid == 1) |> select(id, time, amt, treatment),
  amt ~ time | treatment + id, doseu = "mg"
)
nca_sd <- PKNCA::pk.nca(PKNCA::PKNCAdata(
  conc_sd, dose_sd,
  intervals = data.frame(start = 0, end = Inf, cmax = TRUE, tmax = TRUE,
                         aucinf.obs = TRUE, half.life = TRUE)
))
res_sd <- as.data.frame(nca_sd$result) |>
  filter(PPTESTCD %in% c("cmax", "tmax", "aucinf.obs", "half.life"))

derived_tbl <- tibble::tribble(
  ~Quantity, ~Published, ~Simulated,
  "CL/F (L/h)", "11",
  sprintf("%.1f", gm(subj_bid$cl)),
  "Vc/F (L)", "105",
  sprintf("%.0f", gm(subj_bid$vc)),
  "Effective half-life ln(2)*(Vc/F+Vp/F)/(CL/F) (h)", "9",
  sprintf("%.1f", gm(subj_bid$thalf_eff)),
  "Two-compartment terminal half-life (h), analytic", "not reported",
  sprintf("%.1f", gm(subj_bid$thalf_term)),
  "Terminal half-life (h), PKNCA on single 50 mg fasted dose", "6-10 (prior single-dose analyses)",
  sprintf("%.1f", gm(res_sd$PPORRES[res_sd$PPTESTCD == "half.life" &
                                      res_sd$treatment == "F02 fasted"]))
)

knitr::kable(
  derived_tbl,
  caption = paste("Derived parameters, geometric means. Published CL/F, Vc/F, and",
                  "half-life from Chawla 2023 Table S6; the 6-10 h single-dose",
                  "half-life range is from the Introduction (reference 22)."),
  align = c("l", "r", "r")
)
Derived parameters, geometric means. Published CL/F, Vc/F, and half-life from Chawla 2023 Table S6; the 6-10 h single-dose half-life range is from the Introduction (reference 22).
Quantity Published Simulated
CL/F (L/h) 11 10.6
Vc/F (L) 105 104
Effective half-life ln(2)*(Vc/F+Vp/F)/(CL/F) (h) 9 9.0
Two-compartment terminal half-life (h), analytic not reported 11.8
Terminal half-life (h), PKNCA on single 50 mg fasted dose 6-10 (prior single-dose analyses) 10.6

The effective half-life reproduces the Table S6 value, confirming that the paper’s half-life column uses the total-apparent-volume formula rather than a terminal-slope estimate. This also reconciles the paper’s per-RI-category half-lives, checked next.

Renal-impairment exposure ratios

Chawla 2023 reports AUC increases of 1.17-, 1.46-, and 1.89-fold for mild, moderate, and severe RI relative to normal renal function (Results, “Magnitude of covariate effects”; Figure 5a), and typical half-lives of 8.4, 9.5, 11.5, and 15.1 h across normal, mild, moderate, and severe RI (Results).

ri_auc <- res_ss |>
  filter(treatment %in% ri_bins$treatment, PPTESTCD == "auclast") |>
  group_by(treatment) |>
  summarise(auc_ss = gm(PPORRES), .groups = "drop")

ri_thalf <- sim_bid |>
  filter(treatment %in% ri_bins$treatment) |>
  distinct(id, treatment, cl, vc, vp) |>
  mutate(thalf_eff = log(2) * (vc + vp) / cl) |>
  group_by(treatment) |>
  summarise(thalf = gm(thalf_eff), .groups = "drop")

auc_norm <- ri_auc$auc_ss[ri_auc$treatment == "Normal (>=90)"]

published_ri <- tibble::tibble(
  treatment = ri_bins$treatment,
  auc_ratio_pub = c(1.00, 1.17, 1.46, 1.89),
  thalf_pub     = c(8.4, 9.5, 11.5, 15.1)
)

ri_tbl <- ri_auc |>
  left_join(ri_thalf, by = "treatment") |>
  left_join(published_ri, by = "treatment") |>
  mutate(
    auc_ratio_sim = auc_ss / auc_norm,
    across(c(auc_ratio_sim, auc_ratio_pub), \(x) round(x, 2)),
    across(c(thalf, thalf_pub), \(x) round(x, 1))
  ) |>
  arrange(match(treatment, ri_bins$treatment)) |>
  select(treatment, auc_ratio_pub, auc_ratio_sim, thalf_pub, thalf) |>
  rename(
    "Renal function"               = treatment,
    "AUC ratio, published"         = auc_ratio_pub,
    "AUC ratio, simulated"         = auc_ratio_sim,
    "Effective t1/2 (h), published" = thalf_pub,
    "Effective t1/2 (h), simulated" = thalf
  )

knitr::kable(
  ri_tbl,
  caption = paste("Steady-state AUC0-12 ratios relative to normal renal function",
                  "and effective half-lives by renal-function category.",
                  "Published values from Chawla 2023 Results and Figure 5a."),
  align = c("l", "r", "r", "r", "r")
)
Steady-state AUC0-12 ratios relative to normal renal function and effective half-lives by renal-function category. Published values from Chawla 2023 Results and Figure 5a.
Renal function AUC ratio, published AUC ratio, simulated Effective t1/2 (h), published Effective t1/2 (h), simulated
Normal (>=90) 1.00 1.00 8.4 8.7
Mild RI (60-89) 1.17 1.13 9.5 9.7
Moderate RI (30-59) 1.46 1.42 11.5 12.0
Severe RI (15-29) 1.89 1.73 15.1 15.1

Severe renal impairment: once-daily versus twice-daily 

conc_qd <- PKNCA::PKNCAconc(
  sim_qd |> filter(!is.na(Cc)) |> select(id, time, Cc, treatment),
  Cc ~ time | treatment + id, concu = "ng/mL", timeu = "h"
)
dose_qd <- PKNCA::PKNCAdose(
  events_qd |> filter(evid == 1) |> select(id, time, amt, treatment),
  amt ~ time | treatment + id, doseu = "mg"
)
nca_qd <- PKNCA::pk.nca(PKNCA::PKNCAdata(
  conc_qd, dose_qd,
  intervals = data.frame(start = t_last_qd, end = t_last_qd + tau_qd,
                         cmax = TRUE, cmin = TRUE, auclast = TRUE)
))
res_qd <- as.data.frame(nca_qd$result) |>
  filter(PPTESTCD %in% c("cmax", "cmin", "auclast"))

auc24_qd_severe <- gm(res_qd$PPORRES[res_qd$PPTESTCD == "auclast"])
auc24_bid_norm  <- 2 * auc_norm   # AUC0-24 = 2 x AUC0-12 at steady state

daily_tbl <- tibble::tibble(
  Regimen = c("Normal renal function, 45 mg b.i.d.",
              "Severe RI, 45 mg q.d.",
              "Severe RI, 45 mg b.i.d."),
  `AUC0-24 at steady state (ng*h/mL)` = round(c(
    auc24_bid_norm,
    auc24_qd_severe,
    2 * ri_auc$auc_ss[ri_auc$treatment == "Severe RI (15-29)"]
  )),
  `Ratio vs normal b.i.d.` = round(c(
    1,
    auc24_qd_severe / auc24_bid_norm,
    2 * ri_auc$auc_ss[ri_auc$treatment == "Severe RI (15-29)"] / auc24_bid_norm
  ), 2)
)

knitr::kable(
  daily_tbl,
  caption = paste("Total daily exposure supporting the once-daily recommendation",
                  "in severe RI (Chawla 2023 Figure 5b and Conclusions).",
                  "Cmax within the interval is also lower on q.d. dosing."),
  align = c("l", "r", "r")
)
Total daily exposure supporting the once-daily recommendation in severe RI (Chawla 2023 Figure 5b and Conclusions). Cmax within the interval is also lower on q.d. dosing.
Regimen AUC0-24 at steady state (ng*h/mL) Ratio vs normal b.i.d.
Normal renal function, 45 mg b.i.d. 8021 1.00
Severe RI, 45 mg q.d. 6812 0.85
Severe RI, 45 mg b.i.d. 13898 1.73

Single-dose food effect 

food_tbl <- res_sd |>
  filter(PPTESTCD %in% c("cmax", "tmax", "aucinf.obs")) |>
  group_by(treatment, PPTESTCD) |>
  summarise(value = gm(PPORRES), .groups = "drop") |>
  tidyr::pivot_wider(names_from = PPTESTCD, values_from = value) |>
  mutate(across(where(is.numeric), \(x) round(x, 2))) |>
  rename("Treatment"          = treatment,
         "Cmax (ng/mL)"       = cmax,
         "Tmax (h)"           = tmax,
         "AUC0-inf (ng*h/mL)" = aucinf.obs)

knitr::kable(
  food_tbl,
  caption = paste("Single 50 mg F02 dose, fasted vs fed, geometric means.",
                  "The paper reports no food effect on AUC (the effect is on Ka",
                  "only), so AUC0-inf should be unchanged while Tmax is later."),
  align = c("l", "r", "r", "r")
)
Single 50 mg F02 dose, fasted vs fed, geometric means. The paper reports no food effect on AUC (the effect is on Ka only), so AUC0-inf should be unchanged while Tmax is later.
Treatment AUC0-inf (ng*h/mL) Cmax (ng/mL) Tmax (h)
F02 fasted 3884.10 357.54 1.84
F02 fed 3976.18 313.66 2.85

Because the only retained food effect is on Ka, AUC is unaffected while Tmax is later and Cmax lower under fed conditions. This is the expected behaviour of a rate-only absorption covariate and matches the paper’s statement that “food had no effect on AUC”.

Assumptions and deviations

  • Sex reference-category inversion. Chawla 2023 uses a male indicator with female as the reference category (Table 1 footnote b), the inverse of the canonical SEXF orientation (1 = female, reference 0 = male). To keep the published female-reference CL/F = 10.3 L/h and Vc/F = 101 L verbatim, the model applies the effects through sex_male <- 1 - SEXF, so SEXF = 1 gives a factor of exactly 1 and SEXF = 0 gives the paper’s male-vs-female fractional change (+9.31% on CL/F, +18.1% on Vc/F). This is the same pattern used by Bajaj_2017_nivolumab.R. Reviewers should confirm the sign convention: a male in this model has higher CL/F and Vc/F than a female.
  • FORM_GEF_F02 is a gating covariate, newly registered. The F02-vs-F04 formulation indicator carries no main effect on any PK parameter – no relative bioavailability difference between formulations was retained – and exists only to restrict the FED effect on Ka to F02 records. It is registered as a new canonical entry in inst/references/covariate-columns.md alongside this extraction. For simulations of the approved 45 mg b.i.d. regimen set FORM_GEF_F02 = 0 (COUGH-1/COUGH-2 used F04A), which makes the food effect inoperative.
  • Renal-function strata eGFR values are assumed. The paper’s RI simulations drew demographics from a pooled Merck virtual renal-impairment database (N = 2611) that is not published, so per-category eGFR distributions are not recoverable. Representative values here are drawn uniformly within the conventional KDIGO/FDA bins (normal >= 90, mild 60-89, moderate 30-59, severe 15-29 mL/min/1.73 m^2), with age, weight, and sex held at the RCC/UCC distribution across all four strata. The paper’s ratios additionally embed whatever demographic differences existed between its RI strata, so the simulated ratios here isolate the eGFR effect. The result is that the mild and moderate ratios land within 3% of the published values (1.13 vs 1.17, 1.42 vs 1.46) while the severe-RI ratio runs about 8% low (1.73 vs 1.89) – all well inside the 20% review tolerance. Two known contributors: the normal-renal- function reference arm here averages about 110 mL/min/1.73 m^2 against the 104 mL/min/1.73 m^2 the paper reports for its normal cohort, which deflates every ratio slightly; and the paper’s higher-RI strata are older on average, so their published exposures also pick up the negative age effect on CL/F, which the fixed demographics used here deliberately exclude. No parameter was adjusted.
  • Half-life definition differs from an NCA terminal half-life. The half-life values Chawla 2023 reports (Table S6: 9 h; Results: 8.4 / 9.5 / 11.5 / 15.1 h by RI category) are reproduced by ln(2) * (Vc/F + Vp/F) / (CL/F), a one-compartment-style effective half-life on the total apparent volume, not by the two-compartment terminal slope (about 11.7 h for the typical individual) or by a PKNCA half.life estimate. Both are tabulated above so the difference is explicit; this is a definitional difference, not a transcription error, and no parameter was adjusted to reconcile them.
  • Parameter uncertainty is not propagated. The paper’s simulations sampled fixed-effect uncertainty in addition to IIV. Only IIV is simulated here. The published 95% CIs are narrow, so the effect on geometric means is small, but the simulated between-replicate spread is correspondingly narrower than the paper’s.
  • Residual error is excluded from the comparisons. The paper’s simulated exposures are individual predictions without residual error, so the comparisons use the Cc individual-prediction column. propSd and addSd are still part of the packaged model for estimation use.
  • Covariate distributions are independent draws. Age, weight, sex, and eGFR are sampled independently from the marginal distributions in Table S3. Chawla 2023 Figure S2 shows correlations among the continuous covariates that are not reproduced here; the marginal geometric-mean exposures are only mildly sensitive to this, but the simulated variability differs from the published CVs.
  • Body weight and eGFR are held constant over the dosing period. The source analysis used baseline values for both, so this matches the paper.
  • Race and ethnicity are omitted from the model. Race reached statistical significance on CL/F during covariate model building, but the signal came entirely from the ~5% “multiple” race category, which was merged into “other” before race was dropped from the final model as not clinically relevant. No retained point estimate exists, so race and ethnicity are documented in the model file’s covariatesDataExcluded rather than implemented. Proton-pump inhibitor use is documented there for the same reason (tested, not retained).
  • IIV on Ka was informed by phase I data only (Table 1 footnote c), whereas CL/F and Vc/F IIV used all data. The packaged model applies the single reported omega^2 Ka = 0.551 to every simulated individual regardless of study phase, which is the only behaviour the published parameter set supports.
  • Vp/F and Q/F carry no IIV and no covariate effects, matching Table 1. The peripheral compartment is therefore identical across simulated individuals.