Skip to contents

Atomoxetine (Tobin 2026)

Source: vignettes/articles/Tobin_2026_atomoxetine.Rmd
Tobin_2026_atomoxetine.Rmd

Model and source

  • Citation: Tobin KVT, Gobburu J, Leeder JS, Pritchett A, Dunn A. (2026). Understanding Atomoxetine Exposure Variability in Children and Adolescents With ADHD Through Population Pharmacokinetics. The Journal of Clinical Pharmacology 66(4):e70168. doi:10.1002/jcph.70168.
  • Description: One-compartment population pharmacokinetic model for oral atomoxetine in 86 children and adolescents (6-17 years) with ADHD across three studies (159 participant-occasions, 1946 plasma concentrations), pooling single-dose and steady-state occasions. Absorption is sequential zero-order input into a depot over D1 = 0.75 h followed by first-order transfer (ka = 27.35/h) into the central compartment, with linear elimination. Apparent volume (Vc/F = 175.94 L) and apparent clearance (CL/F = 39.03 L/h) are allometrically scaled on actual body weight centred at 70 kg with fixed exponents of 1.0 and 0.75. Both apparent parameters are divided by a relative-bioavailability factor Frel built from CYP2D6 poor (3.02) and intermediate (1.33) metabolizer status and CYP2C19 poor metabolizer status (2.32), all relative to normal metabolizers; apparent clearance additionally carries an 81% reduction in CYP2D6 poor metabolizers. CYP2D6 normal and ultrarapid metabolizers share the reference category because ultrarapid effects did not explain variability. Between-subject variability is large on the absorption parameters (127% CV on D1, 183% CV on ka) and moderate on disposition (64.5% CV on Vc/F, 75.5% CV on CL/F); residual error is combined additive (0.81 ng/mL) plus proportional (21.4% CV). Parameters come from a two-stage analysis in which each participant-occasion was treated as a separate individual, so the reported between-subject variability pools between-subject and between-occasion variability.
  • Article: https://doi.org/10.1002/jcph.70168

Atomoxetine is a non-stimulant selective norepinephrine reuptake inhibitor used in ADHD. Its exposure is strongly polymorphic: CYP2D6 converts roughly 95% of the parent drug to 4-hydroxyatomoxetine, and CYP2C19 carries the N-desmethylatomoxetine pathway that dominates when CYP2D6 activity is lost. Tobin 2026 pooled three paediatric studies to quantify that variability and to support phenotype-informed dosing toward a 400 ng/mL target peak.

Population

86 children and adolescents aged 6-17 years with ADHD contributed 1946 atomoxetine plasma concentrations over 159 participant-occasions, pooled from 3 studies run at Children’s Mercy Kansas City (Tobin 2026 Table 1). Mean age was 12.6 (SD 3.2) years, mean weight 54.5 (SD 26.4) kg, and mean BMI 21.8 (SD 6.4) kg/m^2; 26% met the paper’s obesity definition (BMI at or above the 95th percentile). The cohort was 81% male and 51% White / 33% Black. CYP2D6 phenotypes were unevenly represented – 41 intermediate, 35 normal, 6 poor and 4 ultrarapid metabolizers – and the CYP2C19 distribution was 31 normal, 26 rapid, 21 intermediate, 6 ultrarapid and 2 poor metabolizers.

A methodological point matters for interpreting the variance terms. Absorption was so erratic between and within subjects that a one-stage analysis would not converge, so the authors treated each participant-occasion as a separate individual in a two-stage analysis. The published between-subject variability therefore pools genuine between-subject with between-occasion variability, and the authors note that two-stage approaches tend to overestimate it.

The same information is available programmatically via readModelDb("Tobin_2026_atomoxetine")()$population.

Source trace

Per-parameter provenance is recorded as an in-file comment beside each ini() entry in inst/modeldb/specificDrugs/Tobin_2026_atomoxetine.R. Collected here:

Equation / parameter Value Source location
ld1 (zero-order depot duration) 0.75 h Table 3, row Dur; Results “Structural Model”
lka (first-order absorption) 27.35 1/h Table 3, row ka
lvc (Vc/F at 70 kg) 175.94 L Table 3, row Vc/F
lcl (CL/F at 70 kg) 39.03 L/h Table 3, row CL/F
e_wt_vc (allometric exponent, Vc/F) 1.0 (fixed) Equation (3), Equation (4); Results “Covariate Model”
e_wt_cl (allometric exponent, CL/F) 0.75 (fixed) Equation (3); Results “Covariate Model”
e_cyp2d6_pm_vc_cl (Frel, CYP2D6 PM) 3.02 Table 3, row Frel CYP2D6,PM; Results “Covariate Model”
e_cyp2d6_im_vc_cl (Frel, CYP2D6 IM) 1.33 Table 3, row labelled Frel CYP2C6,IM (description reads “CYP2D6 IMs”)
e_cyp2c19_pm_vc_cl (Frel, CYP2C19 PM) 2.32 Table 3, row Frel CYP2C19,PM
e_cyp2d6_pm_cl (CL/F effect, CYP2D6 PM) -0.81 Table 3, row E CYP2D6,PM; Discussion “Impact of Covariates”
etald1 / etalka / etalvc / etalcl 127.42 / 182.7 / 64.52 / 75.48 CV% Table 3, BSV block
addSd / propSd 0.81 ng/mL / 21.39 CV% Table 3, RUV block; Results “Structural Model”
d/dt(depot), d/dt(central), dur(depot) n/a Results “Structural Model” (one compartment, zero-order transit into depot, first-order absorption, first-order elimination)
vc <- ... / frel, cl <- ... / frel * (1 + E) n/a Equation (4), Equation (5), Table 3 covariate equations
Cc <- 1000 * central / vc n/a Unit conversion, mg/L to ng/mL (Tobin 2026 reports ng/mL throughout)

Structural checks at the published reference

These checks are deterministic (typical values, random effects zeroed), so the bounds below are exact rather than cohort-dependent.

mod <- readModelDb("Tobin_2026_atomoxetine")
mod_tv <- rxode2::zeroRe(mod)

# Table 1 phenotype-mean weights, used so each arm is compared at the weight
# the paper's own NCA stratum was observed at.
pheno <- tibble::tribble(
  ~phenotype, ~WT,  ~CYP2D6_PM, ~CYP2D6_IM, ~CYP2C19_PM,
  "PM",       61.2,  1,          0,          0,
  "IM",       57.2,  0,          1,          0,
  "NM",       50.2,  0,          0,          0,
  "UM",       54.2,  0,          0,          0
)

obs_grid <- sort(unique(c(seq(0, 6, by = 0.02), seq(6, 120, by = 0.25))))

make_sd_arm <- function(i, dose_mg_kg = 0.5, id_offset = 0L) {
  p <- pheno[i, ]
  dose_rows <- tibble::tibble(
    id = id_offset + i, time = 0, amt = dose_mg_kg * p$WT,
    evid = 1L, cmt = "depot",
    # rate = -2 selects the model-defined duration dur(depot) <- d1. Without
    # it rxode2 silently delivers the dose as a bolus and d1 has no effect.
    rate = -2
  )
  obs_rows <- tibble::tibble(
    id = id_offset + i, time = obs_grid, amt = NA_real_,
    evid = 0L, cmt = "central", rate = NA_real_
  )
  dplyr::bind_rows(dose_rows, obs_rows) |>
    dplyr::mutate(phenotype = p$phenotype, WT = p$WT,
                  CYP2D6_PM = p$CYP2D6_PM, CYP2D6_IM = p$CYP2D6_IM,
                  CYP2C19_PM = p$CYP2C19_PM)
}

ev_tv <- dplyr::bind_rows(lapply(seq_len(nrow(pheno)), make_sd_arm)) |>
  as.data.frame()
sim_tv <- rxode2::rxSolve(mod_tv, ev_tv, keep = c("phenotype", "WT")) |>
  as.data.frame()
#> ℹ omega/sigma items treated as zero: 'etald1', 'etalka', 'etalvc', 'etalcl'
#> Warning: multi-subject simulation without without 'omega'

par_tv <- sim_tv |>
  dplyr::filter(!is.na(Cc)) |>
  dplyr::group_by(phenotype) |>
  dplyr::summarise(WT = dplyr::first(WT), vc = dplyr::first(vc),
                   cl = dplyr::first(cl), .groups = "drop") |>
  dplyr::mutate(t_half = log(2) * vc / cl)

Typical values reproduce Table 3 at the 70 kg reference

At 70 kg with Frel = 1 and E_CYP2D6,PM = 0, the model must return the published Vc/F and CL/F exactly.

ref70 <- rxode2::rxSolve(
  mod_tv,
  data.frame(id = 1L, time = c(0, 1), amt = c(10, NA_real_),
             evid = c(1L, 0L), cmt = c("depot", "central"),
             rate = c(-2, NA_real_),
             WT = 70, CYP2D6_PM = 0, CYP2D6_IM = 0, CYP2C19_PM = 0)
) |> as.data.frame()
#> ℹ omega/sigma items treated as zero: 'etald1', 'etalka', 'etalvc', 'etalcl'

stopifnot(
  abs(ref70$vc[1] - 175.94) < 1e-6,   # Table 3, Vc/F
  abs(ref70$cl[1] -  39.03) < 1e-6    # Table 3, CL/F
)

The allometric centring weight is 70 kg, not the printed 10 kg

Tobin 2026 Equation (3) defines the size term as (wt_i / 70)^b with “70 is the typical adult body weight in kg”, and the Discussion repeats “centered around a 70-kg typical adult”. The printed Equation (5) and the Table 3 covariate-equation footer nevertheless show (wt_i / 10)^0.75 for CL/F.

The paper’s own NCA half-lives adjudicate this, and they do so without reference to any normalisation: because Frel divides Vc/F and CL/F identically, it cancels out of t1/2 = ln2 * (Vc/F) / (CL/F), leaving a quantity that depends only on weight and the CYP2D6-PM clearance effect.

t_half_at <- function(wt, centre, e_pm = 0) {
  log(2) * (175.94 * (wt / 70)^1.0) / (39.03 * (wt / centre)^0.75 * (1 + e_pm))
}
centring <- tibble::tibble(
  phenotype = pheno$phenotype,
  WT        = pheno$WT,
  observed  = c(17.3, 4.5, 2.5, 2.3),      # Table 2, "Half Life (hr)"
  obs_sd    = c( 4.9, 4.6, 0.6, 0.4),
  centre_70 = t_half_at(pheno$WT, 70, c(-0.81, 0, 0, 0)),
  centre_10 = t_half_at(pheno$WT, 10, c(-0.81, 0, 0, 0))
)
knitr::kable(centring, digits = 2,
             caption = "Half-life (h) under the two candidate centring weights.")
Half-life (h) under the two candidate centring weights.
phenotype WT observed obs_sd centre_70 centre_10
PM 61.2 17.3 4.9 15.90 3.70
IM 57.2 4.5 4.6 2.97 0.69
NM 50.2 2.5 0.6 2.88 0.67
UM 54.2 2.3 0.4 2.93 0.68 

# Discriminator 1: the 70 kg centring is closer to the observed half-life than
# the 10 kg centring in EVERY stratum. Both are deterministic, so this is exact.
stopifnot(all(abs(centring$centre_70 - centring$observed) <
                abs(centring$centre_10 - centring$observed)))

# Discriminator 2: the 10 kg centring falls outside the published mean +/- 1 SD
# wherever the observed SD is informative. The IM stratum is excluded because
# its SD (4.6 h) is as large as its mean (4.5 h) and cannot reject anything.
informative <- centring$phenotype != "IM"
stopifnot(all(abs(centring$centre_10 - centring$observed)[informative] >
                centring$obs_sd[informative]))

# Discriminator 3: the 70 kg centring lands inside the published SD for PM, IM
# and NM (UM is discussed under "Assumptions and deviations").
stopifnot(all(abs(centring$centre_70 - centring$observed)[1:3] <=
                centring$obs_sd[1:3]))

The model therefore encodes 70 kg on both parameters, and treats the 10 in Equation (5) as a typographical error. See “Assumptions and deviations”.

Relative bioavailability composes multiplicatively

Both CYP2C19 poor metabolizers in the cohort were also CYP2D6 intermediate metabolizers, so the two Frel factors multiply for those subjects (Table 3 covariate equations).

frel_case <- function(d6pm, d6im, c19pm) {
  rxode2::rxSolve(
    mod_tv,
    data.frame(id = 1L, time = c(0, 1), amt = c(10, NA_real_),
               evid = c(1L, 0L), cmt = c("depot", "central"),
               rate = c(-2, NA_real_),
               WT = 57.2, CYP2D6_PM = d6pm, CYP2D6_IM = d6im,
               CYP2C19_PM = c19pm)
  ) |> as.data.frame() |> dplyr::pull(vc) |> dplyr::first()
}
vc_ref  <- frel_case(0, 0, 0)
#> ℹ omega/sigma items treated as zero: 'etald1', 'etalka', 'etalvc', 'etalcl'
vc_im   <- frel_case(0, 1, 0)
#> ℹ omega/sigma items treated as zero: 'etald1', 'etalka', 'etalvc', 'etalcl'
vc_both <- frel_case(0, 1, 1)
#> ℹ omega/sigma items treated as zero: 'etald1', 'etalka', 'etalvc', 'etalcl'

stopifnot(
  abs(vc_ref / vc_im   - 1.33)        < 1e-6,  # CYP2D6 IM alone
  abs(vc_ref / vc_both - 1.33 * 2.32) < 1e-6   # IM and CYP2C19 PM together
)

Dose mass balance

CL/F * AUC(0-inf) must return the administered dose in every arm; this is the gate that catches a zero-order input that silently failed to deliver.

mass_balance <- sim_tv |>
  dplyr::filter(!is.na(Cc)) |>
  dplyr::group_by(phenotype) |>
  dplyr::summarise(
    dose  = 0.5 * dplyr::first(WT),
    # trapezoidal AUC over the grid, plus the analytic terminal tail
    auc   = sum(diff(time) * (head(Cc, -1) + tail(Cc, -1)) / 2) / 1000 +
              dplyr::last(Cc) / 1000 / (dplyr::first(cl) / dplyr::first(vc)),
    recovered = dplyr::first(cl) * auc,
    .groups = "drop"
  ) |>
  dplyr::mutate(pct = 100 * recovered / dose)
knitr::kable(mass_balance, digits = 3,
             caption = "CL/F * AUC(0-inf) against the administered dose (mg).")
CL/F * AUC(0-inf) against the administered dose (mg).
phenotype dose auc recovered pct
IM 28.6 1.134 28.602 100.008
NM 25.1 0.825 25.102 100.008
PM 30.6 13.783 30.600 100.001
UM 27.1 0.841 27.102 100.008 
stopifnot(max(abs(mass_balance$pct - 100)) < 0.5)

Steady-state accumulation occurs only in poor metabolizers

Tobin 2026 Results “Data”: “accumulation at steady state is evident for only CYP2D6 PMs at steady-state.” With once-daily dosing the accumulation ratio is 1 / (1 - exp(-kel * 24)).

accum <- par_tv |>
  dplyr::mutate(kel = cl / vc,
                ratio = 1 / (1 - exp(-kel * 24)))
knitr::kable(accum |> dplyr::select(phenotype, t_half, ratio), digits = 3,
             caption = "Predicted once-daily accumulation ratio by CYP2D6 phenotype.")
Predicted once-daily accumulation ratio by CYP2D6 phenotype.
phenotype t_half ratio
IM 2.971 1.004
NM 2.875 1.003
PM 15.902 1.542
UM 2.931 1.003 
stopifnot(
  accum$ratio[accum$phenotype == "PM"] > 1.3,                 # marked accumulation
  all(accum$ratio[accum$phenotype != "PM"] < 1.05)            # essentially none
)

Virtual cohort

Original observed data are not publicly available. The cohort below draws weights per CYP2D6 phenotype from log-normal distributions matched to the Table 1 stratum means and SDs, truncated to a plausible 6-17 year range, and gives every subject the Study A weight-based dose of 0.5 mg/kg.

# set.seed() seeds R's RNG only. rxode2's simulation RNG is partitioned per
# solver thread, so this cohort differs on a machine with a different thread
# count. Every assertion below is written to hold for any cohort the model
# can produce (pattern 12 of known-vignette-failure-patterns.md).
set.seed(20260912L)
n_per_arm <- 120L

# Coarser than the deterministic grid above: still 0.05 h through the
# absorption phase (D1 = 0.75 h, so Tmax is well resolved) but 2 h out in the
# terminal phase, which keeps the PKNCA half-life fit over ~500 subjects
# inside the vignette time budget.
cohort_grid <- sort(unique(c(seq(0, 6, by = 0.05),
                             seq(6, 24, by = 0.5),
                             seq(24, 120, by = 2))))

sample_wt <- function(n, m, s) {
  # log-normal matched on the arithmetic mean and SD, clipped to 20-120 kg
  sdlog  <- sqrt(log(1 + (s / m)^2))
  meanlog <- log(m) - sdlog^2 / 2
  pmin(pmax(stats::rlnorm(n, meanlog, sdlog), 20), 120)
}
wt_stats <- tibble::tribble(
  ~phenotype, ~m,   ~s,
  "PM",       61.2, 26.3,
  "IM",       57.2, 29.5,
  "NM",       50.2, 22.1,
  "UM",       54.2, 30.9
)

make_cohort <- function(i, id_offset) {
  p  <- pheno[i, ]
  ws <- wt_stats[wt_stats$phenotype == p$phenotype, ]
  subj <- tibble::tibble(
    id = id_offset + seq_len(n_per_arm),
    WT = sample_wt(n_per_arm, ws$m, ws$s)
  )
  dose_rows <- subj |>
    dplyr::mutate(time = 0, amt = 0.5 * WT, evid = 1L,
                  cmt = "depot", rate = -2)
  obs_rows <- subj |>
    tidyr::crossing(time = cohort_grid) |>
    dplyr::mutate(amt = NA_real_, evid = 0L,
                  cmt = "central", rate = NA_real_)
  dplyr::bind_rows(dose_rows, obs_rows) |>
    dplyr::mutate(phenotype = p$phenotype,
                  CYP2D6_PM = p$CYP2D6_PM, CYP2D6_IM = p$CYP2D6_IM,
                  CYP2C19_PM = p$CYP2C19_PM)
}

events <- dplyr::bind_rows(
  lapply(seq_len(nrow(pheno)),
         function(i) make_cohort(i, id_offset = (i - 1L) * 1000L))
) |>
  dplyr::arrange(id, time, dplyr::desc(evid)) |>
  as.data.frame()

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

Simulation 

sim <- rxode2::rxSolve(mod, events = events,
                       keep = c("phenotype", "WT")) |>
  as.data.frame()
sim$phenotype <- factor(sim$phenotype, levels = c("PM", "IM", "NM", "UM"))

Replicate published figures 

# Companion to Figure 1 of Tobin 2026 (individual atomoxetine plasma
# concentration over time, coloured by CYP2D6 phenotype) and to the Figure 4
# VPC: median with 10th-90th percentile band, by phenotype.
sim |>
  dplyr::filter(!is.na(Cc), time <= 48) |>
  dplyr::group_by(phenotype, time) |>
  dplyr::summarise(Q10 = quantile(Cc, 0.10), Q50 = quantile(Cc, 0.50),
                   Q90 = quantile(Cc, 0.90), .groups = "drop") |>
  ggplot(aes(time, Q50, colour = phenotype, fill = phenotype)) +
  geom_ribbon(aes(ymin = Q10, ymax = Q90), alpha = 0.2, colour = NA) +
  geom_line(linewidth = 0.8) +
  scale_y_log10() +
  labs(x = "Time (h)", y = "Atomoxetine (ng/mL)",
       colour = "CYP2D6", fill = "CYP2D6",
       title = "Single 0.5 mg/kg dose by CYP2D6 phenotype",
       caption = paste0("Median and 10th-90th percentiles, ", n_per_arm,
                        " subjects per arm. Companion to Figures 1 and 4 of",
                        " Tobin 2026.")) +
  theme_bw()
#> 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.

The rank order reproduces the paper’s central finding: at the label’s flat weight-based dose, exposure rises steeply as CYP2D6 activity falls, and poor metabolizers separate from every other phenotype.

# Reproduces the direction and rough magnitude of the Results statement that
# "CYP2D6 PMs show 130% and 624% higher dose-normalized Cmax and AUC0-24h,
# respectively, compared to NMs".
exposure <- sim |>
  dplyr::filter(!is.na(Cc)) |>
  dplyr::group_by(phenotype, id) |>
  dplyr::summarise(
    # rxSolve does not return `amt`; every subject received 0.5 mg/kg, so the
    # administered dose is recovered exactly from the kept WT column.
    dose  = 0.5 * dplyr::first(WT),
    cmax  = max(Cc),
    auc24 = sum(diff(time[time <= 24]) *
                  (head(Cc[time <= 24], -1) + tail(Cc[time <= 24], -1)) / 2),
    .groups = "drop"
  ) |>
  dplyr::group_by(phenotype) |>
  dplyr::summarise(cmax_dn = median(cmax / dose),
                   auc_dn  = median(auc24 / dose), .groups = "drop")

ratios <- exposure |>
  dplyr::mutate(
    cmax_vs_nm = cmax_dn / exposure$cmax_dn[exposure$phenotype == "NM"],
    auc_vs_nm  = auc_dn  / exposure$auc_dn[exposure$phenotype == "NM"]
  )
knitr::kable(ratios, digits = 2,
             caption = "Dose-normalized exposure relative to CYP2D6 normal metabolizers.")
Dose-normalized exposure relative to CYP2D6 normal metabolizers.
phenotype cmax_dn auc_dn cmax_vs_nm auc_vs_nm
PM 18.84 272.37 2.33 8.12
IM 7.92 45.54 0.98 1.36
NM 8.08 33.52 1.00 1.00
UM 6.73 29.56 0.83 0.88 

# The paper reports PM/NM ratios of 2.30 (Cmax) and 7.24 (AUC0-24) from its
# observed NCA. The model is built on typical values from a different (pooled,
# individualised-dose) dataset, so bound the comparison generously: what must
# hold is that PMs are several-fold higher on Cmax and an order of magnitude
# higher on AUC, and that IM sits between NM and PM.
pm <- ratios[ratios$phenotype == "PM", ]
im <- ratios[ratios$phenotype == "IM", ]
stopifnot(
  pm$cmax_vs_nm > 1.8, pm$cmax_vs_nm < 4.5,
  pm$auc_vs_nm  > 4.0, pm$auc_vs_nm  < 16,
  im$auc_vs_nm  > 1.0, im$auc_vs_nm  < pm$auc_vs_nm
)

PKNCA validation 

sim_nca <- sim |>
  dplyr::filter(!is.na(Cc)) |>
  dplyr::select(id, time, Cc, phenotype)

# Guarantee a time = 0 row per subject; pre-dose Cc = 0 for extravascular input.
sim_nca <- dplyr::bind_rows(
  sim_nca,
  sim_nca |> dplyr::distinct(id, phenotype) |>
    dplyr::mutate(time = 0, Cc = 0)
) |>
  dplyr::distinct(phenotype, id, time, .keep_all = TRUE) |>
  dplyr::arrange(id, time)

conc_obj <- PKNCA::PKNCAconc(as.data.frame(sim_nca), Cc ~ time | phenotype + id)
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found

dose_df <- events |>
  dplyr::filter(evid == 1) |>
  dplyr::select(id, time, amt, phenotype)
dose_obj <- PKNCA::PKNCAdose(dose_df, amt ~ time | phenotype + id)

intervals <- data.frame(
  start = 0, end = Inf,
  cmax = TRUE, tmax = TRUE, half.life = TRUE, aucinf.obs = TRUE
)
nca_res <- PKNCA::pk.nca(PKNCA::PKNCAdata(conc_obj, dose_obj,
                                          intervals = intervals))
#> Warning in assert_conc(conc = conc): Negative concentrations found
#> Warning in assert_conc(conc = conc): Negative concentrations found
#> Warning in assert_conc(conc = conc): Negative concentrations found
#> Warning in assert_conc(conc = conc): Negative concentrations found
#> Warning in assert_conc(conc = conc): Negative concentrations found
#> Warning in assert_conc(conc = conc): Negative concentrations found
#> Warning in log(data$conc): NaNs produced
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(conc.2/conc.1): NaNs produced
#> Warning in assert_conc(conc = conc): Negative concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(data$conc): NaNs produced
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(conc.2/conc.1): NaNs produced
#> Warning in assert_conc(conc = conc): Negative concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(data$conc): NaNs produced
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(conc.2/conc.1): NaNs produced
#> Warning in assert_conc(conc = conc): Negative concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(data$conc): NaNs produced
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(conc.2/conc.1): NaNs produced
#> Warning in assert_conc(conc = conc): Negative concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(data$conc): NaNs produced
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(conc.2/conc.1): NaNs produced
#> Warning in assert_conc(conc = conc): Negative concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(data$conc): NaNs produced
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(conc.2/conc.1): NaNs produced
#> Warning in assert_conc(conc = conc): Negative concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(data$conc): NaNs produced
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(conc.2/conc.1): NaNs produced
#> Warning in assert_conc(conc = conc): Negative concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(data$conc): NaNs produced
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(conc.2/conc.1): NaNs produced
#> Warning in assert_conc(conc = conc): Negative concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(data$conc): NaNs produced
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(conc.2/conc.1): NaNs produced
#> Warning in assert_conc(conc = conc): Negative concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(data$conc): NaNs produced
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(conc.2/conc.1): NaNs produced
#> Warning in assert_conc(conc = conc): Negative concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(data$conc): NaNs produced
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(conc.2/conc.1): NaNs produced
#> Warning in assert_conc(conc = conc): Negative concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(data$conc): NaNs produced
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(conc.2/conc.1): NaNs produced
#> Warning in assert_conc(conc = conc): Negative concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(data$conc): NaNs produced
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(conc.2/conc.1): NaNs produced
#> Warning in assert_conc(conc = conc): Negative concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(data$conc): NaNs produced
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(conc.2/conc.1): NaNs produced
#> Warning in assert_conc(conc = conc): Negative concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(data$conc): NaNs produced
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(conc.2/conc.1): NaNs produced
#> Warning in assert_conc(conc = conc): Negative concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(data$conc): NaNs produced
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(conc.2/conc.1): NaNs produced
#> Warning in assert_conc(conc = conc): Negative concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(data$conc): NaNs produced
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(conc.2/conc.1): NaNs produced
#> Warning in assert_conc(conc = conc): Negative concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(data$conc): NaNs produced
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(conc.2/conc.1): NaNs produced
#> Warning in assert_conc(conc = conc): Negative concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(data$conc): NaNs produced
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(conc.2/conc.1): NaNs produced
#> Warning in assert_conc(conc = conc): Negative concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(data$conc): NaNs produced
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(conc.2/conc.1): NaNs produced
#> Warning in assert_conc(conc = conc): Negative concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(data$conc): NaNs produced
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(conc.2/conc.1): NaNs produced
#> Warning in assert_conc(conc = conc): Negative concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(data$conc): NaNs produced
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(conc.2/conc.1): NaNs produced
#> Warning in assert_conc(conc = conc): Negative concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(data$conc): NaNs produced
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(conc.2/conc.1): NaNs produced
#> Warning in assert_conc(conc = conc): Negative concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(data$conc): NaNs produced
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(conc.2/conc.1): NaNs produced
#> Warning in assert_conc(conc = conc): Negative concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(data$conc): NaNs produced
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(conc.2/conc.1): NaNs produced
#> Warning in assert_conc(conc = conc): Negative concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(data$conc): NaNs produced
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(conc.2/conc.1): NaNs produced
#> Warning in assert_conc(conc = conc): Negative concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(data$conc): NaNs produced
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(conc.2/conc.1): NaNs produced
#> Warning in assert_conc(conc = conc): Negative concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(data$conc): NaNs produced
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(conc.2/conc.1): NaNs produced
#> Warning in assert_conc(conc = conc): Negative concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(data$conc): NaNs produced
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(conc.2/conc.1): NaNs produced
#> Warning in assert_conc(conc = conc): Negative concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(data$conc): NaNs produced
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found
#> Warning in log(conc.2/conc.1): NaNs produced

Comparison against published NCA

Tobin 2026 Table 2 reports four NCA rows per CYP2D6 phenotype. Only half-life and Tmax are comparable here; the two dose-normalized rows are not, for a reason internal to the paper (see below).

published <- tibble::tribble(
  ~phenotype, ~half.life, ~tmax,
  "PM",       17.3,       4.32,
  "IM",        4.5,       1.73,
  "NM",        2.5,       1.46,
  "UM",        2.3,       0.92
)

cmp <- nlmixr2lib::ncaComparisonTable(
  simulated = nca_res,
  reference = published,
  by        = "phenotype",
  units     = c(half.life = "h", tmax = "h"),
  tolerance_pct = 20
)
knitr::kable(cmp, caption = paste0(
  "Simulated (median of ", n_per_arm, " subjects per arm) vs Tobin 2026 ",
  "Table 2 (mean +/- SD). * differs from reference by >20%."
), align = c("l", "l", "r", "r", "r"))
Simulated (median of 120 subjects per arm) vs Tobin 2026 Table 2 (mean +/- SD). * differs from reference by >20%.
NCA parameter phenotype Reference Simulated % diff
Tmax (h) PM 4.32 1.25 -71.1%*
Tmax (h) IM 1.73 1.15 -33.5%*
Tmax (h) NM 1.46 1.05 -28.1%*
Tmax (h) UM 0.92 0.875 -4.9%
t½ (h) PM 17.3 18 +4.0%
t½ (h) IM 4.5 3.08 -31.5%*
t½ (h) NM 2.5 2.51 +0.2%
t½ (h) UM 2.3 2.6 +13.1% 
attr(cmp, "footnote")
#> [1] "* differs from reference by more than ±20%."

Half-life is reproduced for poor, normal and ultrarapid metabolizers (+4.0%, +0.2% and +13.1% against Table 2 on the render above) – a fifteen-hour PM half-life against a two-and-a-half-hour NM half-life is the quantity the covariate model is actually identified on, and it is the check that adjudicated the centring-weight typo above.

Intermediate-metabolizer half-life is flagged. The model predicts about 3.1 h against an observed mean of 4.5 h, because Frel cancels out of t1/2 = ln2 * (Vc/F) / (CL/F) and Tobin 2026 retained no clearance effect for CYP2D6 IMs (“the effect of CYP2D6 IM on CL/F was investigated, but did not explain the variability”). The model therefore predicts essentially the same half-life for IM, NM and UM by construction. The observed IM SD is 4.6 h on a mean of 4.5 h, so the published value cannot discriminate; this row is reported rather than gated.

Tmax is a known deviation. The model puts typical Tmax at roughly 0.8 h in every phenotype, because D1 and ka carry no phenotype effect: Tobin 2026 Results “Covariate Model” states that “no relationships in absorption parameter EBEs were observed for the potential covariates tested”. The observed 4.32 h mean Tmax in poor metabolizers therefore cannot be reproduced by construction. The model does generate a wide Tmax distribution from the very large absorption BSV (127% CV on D1, 183% CV on ka), but its median stays near the typical value. This is a faithful reproduction of a published limitation, not an encoding error, and is not gated.

hl <- as.data.frame(nca_res) |>
  dplyr::filter(PPTESTCD == "half.life") |>
  dplyr::group_by(phenotype) |>
  dplyr::summarise(median_hl = median(PPORRES, na.rm = TRUE), .groups = "drop") |>
  dplyr::left_join(published, by = "phenotype")

# Bound chosen well outside the spread seen across thread counts: the
# structural claim is that PM half-life is several-fold longer than every
# other phenotype, and that the non-PM phenotypes cluster near 3 h. A
# mis-transcribed volume, clearance or allometric exponent breaks this.
stopifnot(
  hl$median_hl[hl$phenotype == "PM"] > 10,
  hl$median_hl[hl$phenotype == "PM"] < 25,
  all(hl$median_hl[hl$phenotype != "PM"] > 1.5),
  all(hl$median_hl[hl$phenotype != "PM"] < 6)
)

Why Table 2’s dose-normalized rows are not a validation target

For any one-compartment drug, AUC(0-24) / Cmax has a hard lower bound of (1/kel) * (1 - exp(-24 * kel)), attained only for instantaneous absorption. Evaluating that bound with Table 2’s own half-lives shows each row’s printed ratio falling one to two orders of magnitude below its own floor:

t2 <- tibble::tibble(
  phenotype = c("PM", "IM", "NM", "UM"),
  cmax_dn   = c(275, 79, 40, 23),        # Table 2, dose-normalized Cmax
  auc24_dn  = c(21.8, 12.8, 9.5, 7.0),   # Table 2, dose-normalized AUC0-24
  half_life = c(17.3, 4.5, 2.5, 2.3)     # Table 2, Half Life
) |>
  dplyr::mutate(
    kel            = log(2) / half_life,
    ratio_printed  = auc24_dn / cmax_dn,
    ratio_floor    = (1 / kel) * (1 - exp(-24 * kel)),
    shortfall_fold = ratio_floor / ratio_printed
  )
knitr::kable(t2 |> dplyr::select(phenotype, ratio_printed, ratio_floor,
                                 shortfall_fold),
             digits = 3,
             caption = paste("Table 2's printed AUC0-24/Cmax against the",
                             "minimum implied by its own half-life column."))
Table 2’s printed AUC0-24/Cmax against the minimum implied by its own half-life column.
phenotype ratio_printed ratio_floor shortfall_fold
PM 0.079 15.417 194.485
IM 0.162 6.331 39.075
NM 0.238 3.602 15.167
UM 0.304 3.316 10.895 
stopifnot(all(t2$shortfall_fold > 5))

The two columns cannot therefore share a common dose normalization, so neither can be matched against a simulation without knowing which one to correct. The half-life and Tmax rows are unaffected and are used above.

Assumptions and deviations

  • Allometric centring weight (deviation from the printed Equation (5)). Equation (5) and the Table 3 covariate-equation footer print (wt_i / 10)^0.75 for CL/F. The model uses (wt_i / 70)^0.75, because Equation (3) defines the size term with “70 is the typical adult body weight in kg”, the Results and Discussion both state 70 kg, and a 10 kg centring is falsified by the paper’s own NCA half-lives by more than two published SDs in every stratum (see “Structural checks” above). Treated as a typographical error in the source.
  • Table 3 parameter-label typo. The relative-bioavailability row for CYP2D6 intermediate metabolizers is labelled Frel CYP2C6,IM; its Description column reads “Relative Bioavailability of CYP2D6 IMs” and the Results text confirms 1.33 for CYP2D6 IMs. Encoded as a CYP2D6 effect.
  • CYP2D6 PM clearance effect sign and scale. Results renders this as “-0.81%”, while Table 3 (-0.81, CI -0.84 to -0.78) and the Discussion (“an 81% reduction in clearance”) agree it is a proportion. Encoded as -81%.
  • Reference category pools normal and ultrarapid metabolizers. CYP2D6 ultrarapid effects on both Frel and CL/F were tested and did not explain variability, so CYP2D6_PM = 0, CYP2D6_IM = 0 denotes NM or UM. The model consequently predicts identical disposition for NM and UM, which is why the UM half-life row sits slightly outside the published SD.
  • CYP2C19 carries only a poor-metabolizer level. Intermediate, normal, rapid and ultrarapid CYP2C19 phenotypes are pooled into the reference, following the paper’s final model. The effect rests on 2 participants (4 occasions) and the authors advise interpreting it with caution.
  • Between-subject variability pools between-occasion variability. The two-stage analysis treated each participant-occasion as a separate individual, so etald1, etalka, etalvc and etalcl are wider than a conventional one-stage BSV would be. Simulated variability is correspondingly wide.
  • Bounded absorption parameters. The Discussion records that D1 and ka were bounded during estimation and that “the parameter distributions for zero-order duration and first-order rate constant show a large portion at the bounds”. The published point estimates are encoded as reported; simulated absorption is therefore smoother than the observed data, which frequently showed secondary peaks around 5 h that no covariate explained.
  • Residual-error scale. Table 3 labels the RUV rows with sigma-squared symbols but gives their units as ng/mL and CV%, and the Results text quotes “0.81 ng/mL and 21 CV%”. Both are encoded on the standard-deviation scale.
  • Virtual cohort weights are log-normal draws matched to the Table 1 stratum means and SDs and clipped to 20-120 kg; the paper reports means and SDs but no weight range. All subjects receive the Study A weight-based 0.5 mg/kg dose, so the figures show the label-style flat dosing rather than the phenotype-individualised dosing of Studies B and C.
  • No parameter value in this model came from outside Tobin 2026. No supplement was available on disk; every ini() entry traces to the main article’s Table 3 or its Results text, both of which report the final covariate model in full.