Aldehyde oxidase substrates: intravenous PK (Izat 2025)

Models and source

This paper contributes three models, one per compound:

model_names <- c("Izat_2025_zaleplon",
                 "Izat_2025_ziprasidone",
                 "Izat_2025_zoniporide")
uis <- lapply(model_names, function(n) rxode2::rxode(readModelDb(n)))
names(uis) <- model_names

• Citation: Izat N, Bolleddula J, Carione P, Huertas Valentin L, Jones RS, Kulkarni P, Moss D, Peterkin VC, Tian DD, Treyer A, Venkatakrishnan K, Zientek MA, Barber J, Houston JB, Galetin A, Scotcher D. (2025). Establishing a physiologically based pharmacokinetic framework for aldehyde oxidase and dual aldehyde oxidase-CYP substrates. CPT Pharmacometrics Syst Pharmacol 14(1):164-178. doi:10.1002/psp4.13255.
• Article: https://doi.org/10.1002/psp4.13255
• Supplement (Appendix S1): https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11706420/

What these models are, and what they are not

Izat 2025 is a framework paper. It builds Simcyp (version 21) PBPK models for six aldehyde-oxidase (AO) substrates – capmatinib, idelalisib, lenvatinib, zaleplon, ziprasidone and zoniporide – in order to ask whether PBPK can predict AO-mediated clearance, the fraction metabolised by AO (fmAO) versus by CYP3A4 (fmCYP3A4), and the resulting CYP3A4 drug-drug-interaction risk.

Simcyp’s whole-body mass-balance equations are proprietary and appear nowhere in the paper or Appendix S1. The corpus contains sixteen numbered equations and not one of them is a disposition equation: Equations S1-S7 and S10-S13 are bench-assay arithmetic, Equation S8 is a permeability correlation, and Equations S14-S16 are a static (time-free, algebraic) AO-DDI calculator. The platform model therefore cannot be encoded as published.

What is completely reported, in Table 1, is each compound’s disposition card: the steady-state volume of distribution, the volume of the single adjusting compartment (SAC) and the inter-compartmental clearance where a minimal-PBPK layout was used, the observed intravenous plasma clearance, and the observed renal clearance. Appendix S1 Supplement 2 states the minimal-PBPK structure in the paper’s own words – “a non-physiological central compartment excluding the liver and portal vein coupled with a non-physiological non-central compartment, facilitating adjustment of systemic compartment concentration profile (SAC)”. That is enough to rebuild the intravenous disposition of three of the six compounds as ordinary compartmental models.

Why only the intravenous arms

The oral arms of this paper are deliberately not reduced. Oral bioavailability in the source is F = fa * Fg * fh, and while Table 1 prints fa and Appendix S1 gives Fg, the hepatic availability fh = 1 - CL/Q_H requires the hepatic blood flow Q_H. Q_H is mentioned twice in the prose and never given a value; neither is liver weight, and neither is body weight. Supplying Q_H from the Simcyp platform default would be exactly the substitution that this package’s PBPK sourcing rule forbids – and it is not a rounding-level choice, because moving Q_H from 70 to 110 L/h moves the predicted zaleplon oral Cmax from 0.62 to 1.29 times observed.

The intravenous arms have no such dependency. They need the reported clearance, the reported volume, and one assumption – a 70 kg reference body weight, since Table 1 states every volume in L/kg and no weight is printed. That is the standing “undefined reference value maps to a rounded standard” convention, and it is the only external input anywhere in the three models below.

Capmatinib, idelalisib and lenvatinib are oral-only in this paper and are not extracted from it. Capmatinib additionally fails a cross-dose consistency check: its observed oral clearance is 36.1 L/h at 200 mg and 51.7 L/h at 600 mg, genuinely nonlinear, while Table 1 carries a single weighted-mean 42.3 L/h that cannot serve both doses.

What is not reproducible

Every headline result of the paper. The fmAO / fmCYP3A4 split, the empirical scaling factors applied to in vitro intrinsic clearance, and all of the CYP3A4 DDI area-ratio predictions against ketoconazole, itraconazole, rifampicin and carbamazepine depend on the unpublished Simcyp enzyme-abundance layer and on proprietary Certara perpetrator compound files. These models reproduce the intravenous control profiles only. See Assumptions and deviations.

Population

Each compound’s intravenous data come from a single published study, listed in Appendix S1 Table S3.

pop_tab <- do.call(rbind, lapply(model_names, function(n) {
  p <- uis[[n]]$population
  data.frame(
    Model      = n,
    Species    = p$species,
    N          = p$n_subjects,
    `Age (y)`  = p$age_range,
    `% female` = p$sex_female_pct,
    Dose       = p$dose_range,
    check.names = FALSE
  )
}))
knitr::kable(pop_tab, row.names = FALSE,
             caption = "Population per compound (Izat 2025 Appendix S1 Table S3).")

Population per compound (Izat 2025 Appendix S1 Table S3).

Model	Species	N	Age (y)	% female	Dose
Izat_2025_zaleplon	human	10	19-32 years	50	Single 5 mg intravenous infusion.
Izat_2025_ziprasidone	human	12	19-37 years	0	Single 5 mg intravenous infusion.
Izat_2025_zoniporide	human	4	18-55 years	0	Single 80 mg intravenous infusion.

• Zaleplon – Rosen 1999 (Appendix S1 Table S3 reference 22). Ten healthy volunteers aged 19-32 y received 5 mg intravenously. Table S3 records the female fraction as “0 or 1”, i.e. separate all-female and all-male cohorts, which is why Table 1 reports two clearances for this compound and no other.
• Ziprasidone – Miceli 2005 (reference 23). Twelve healthy men aged 19-37 y received 5 mg intravenously.
• Zoniporide – Dalvie 2010 (reference 13). Four healthy men aged 18-55 y received 80 mg intravenously in a combined pharmacokinetic and mass-balance study.

None of the three studies reports a body weight, and neither does Appendix S1.

Source trace

Every value in every ini() block, with its source location.

trace_tab <- tibble::tribble(
  ~Model,        ~Parameter,    ~Value,     ~Source,
  "zaleplon",    "lvc",         "81.1993 L", "Table 1: Vss 1.16 L/kg, Vsac 0.00001 L/kg; (1.16 - 0.00001) * 70 kg",
  "zaleplon",    "lcl",         "71.58 L/h", "Table 1 'CLiv/oral (L/h) [CV%]': 71.58 [20.2] (i.v., male)",
  "zaleplon",    "e_sexf_cl",   "log(0.733585)", "Table 1: 52.51 L/h (i.v., female) / 71.58 L/h (i.v., male)",
  "zaleplon",    "propSd",      "0 (fixed)", "No residual-error model is reported anywhere in the source",
  "ziprasidone", "lvc",         "70.7 L",    "Table 1: Vss 1.01 L/kg; 1.01 * 70 kg",
  "ziprasidone", "lcl",         "23.04 L/h", "Table 1 'CLiv/oral (L/h) [CV%]': 23.04 [14] (i.v.)",
  "ziprasidone", "propSd",      "0 (fixed)", "No residual-error model is reported anywhere in the source",
  "zoniporide",  "lvc",         "119.0 L",   "Table 1: Vss 1.70 L/kg; 1.70 * 70 kg",
  "zoniporide",  "lcl_renal",   "16.2 L/h",  "Table 1 'CLRenal (L/h)': 16.2",
  "zoniporide",  "lcl_nonren",  "79.15 L/h", "Table 1: CLiv 95.35 [5.7] minus CLRenal 16.2",
  "zoniporide",  "propSd",      "0 (fixed)", "No residual-error model is reported anywhere in the source"
)
knitr::kable(trace_tab, caption = "Source trace for every fixed parameter.")

Source trace for every fixed parameter.

Model	Parameter	Value	Source
zaleplon	lvc	81.1993 L	Table 1: Vss 1.16 L/kg, Vsac 0.00001 L/kg; (1.16 - 0.00001) * 70 kg
zaleplon	lcl	71.58 L/h	Table 1 ‘CLiv/oral (L/h) [CV%]’: 71.58 [20.2] (i.v., male)
zaleplon	e_sexf_cl	log(0.733585)	Table 1: 52.51 L/h (i.v., female) / 71.58 L/h (i.v., male)
zaleplon	propSd	0 (fixed)	No residual-error model is reported anywhere in the source
ziprasidone	lvc	70.7 L	Table 1: Vss 1.01 L/kg; 1.01 * 70 kg
ziprasidone	lcl	23.04 L/h	Table 1 ‘CLiv/oral (L/h) [CV%]’: 23.04 [14] (i.v.)
ziprasidone	propSd	0 (fixed)	No residual-error model is reported anywhere in the source
zoniporide	lvc	119.0 L	Table 1: Vss 1.70 L/kg; 1.70 * 70 kg
zoniporide	lcl_renal	16.2 L/h	Table 1 ‘CLRenal (L/h)’: 16.2
zoniporide	lcl_nonren	79.15 L/h	Table 1: CLiv 95.35 [5.7] minus CLRenal 16.2
zoniporide	propSd	0 (fixed)	No residual-error model is reported anywhere in the source

The model equations are the standard one-compartment intravenous system (d/dt(central) = -kel * central, kel = cl / vc, Cc = 1000 * central / vc) with no nonstandard terms. The 1000 converts mg/L to the ng/mL used throughout Appendix S1 Table S10. Each compound is one compartment for a different, explicitly stated reason:

• Zaleplon is classified “Minimal PBPK” in Table 1, but its reported SAC parameters are Vsac = 0.00001 L/kg and Q = 0 L/h. An inter-compartmental clearance of exactly zero means the SAC never receives drug, so the reduction to one compartment is exact, not an approximation.
• Ziprasidone and zoniporide are classified “Full PBPK”. In a full-PBPK layout the multi-tissue structure is the distribution model, and the only distribution quantity Table 1 reports is the lumped steady-state volume that the fitted global partition-coefficient scalar was tuned to recover. No tissue volumes, blood flows or partition coefficients are printed, and no Vsac or Q is given. These two reductions are therefore genuinely mono-exponential approximations of multi-exponential profiles, and the cost shows up in Cmax below (about 15% low and 14% high respectively). That is reported, not tuned away.

Reproducing the derivations

BW <- 70  # kg -- assumed reference weight; see Errata

# Table 1 inputs, verbatim.
tab1 <- tibble::tribble(
  ~drug,          ~Vss_Lkg, ~Vsac_Lkg, ~CL_Lh,  ~CLrenal_Lh,
  "zaleplon_F",       1.16,   0.00001,  52.51,          0,
  "zaleplon_M",       1.16,   0.00001,  71.58,          0,
  "ziprasidone",      1.01,        NA,  23.04,          0,
  "zoniporide",       1.70,        NA,  95.35,       16.2
)

derived <- tab1 |>
  dplyr::mutate(
    vc_L      = (Vss_Lkg - dplyr::coalesce(Vsac_Lkg, 0)) * BW,
    kel_perh  = CL_Lh / vc_L,
    thalf_h   = log(2) / kel_perh
  )
knitr::kable(derived |> dplyr::mutate(dplyr::across(where(is.numeric),
                                                    \(x) signif(x, 5))),
             caption = "Derived one-compartment parameters at BW = 70 kg.")

Derived one-compartment parameters at BW = 70 kg.

drug	Vss_Lkg	Vsac_Lkg	CL_Lh	CLrenal_Lh	vc_L	kel_perh	thalf_h
zaleplon_F	1.16	1e-05	52.51	0.0	81.199	0.64668	1.07190
zaleplon_M	1.16	1e-05	71.58	0.0	81.199	0.88153	0.78630
ziprasidone	1.01	NA	23.04	0.0	70.700	0.32588	2.12700
zoniporide	1.70	NA	95.35	16.2	119.000	0.80126	0.86507

# The values in the model files must equal the arithmetic above.
stopifnot(
  isTRUE(all.equal(exp(uis[["Izat_2025_zaleplon"]]$theta[["lvc"]]),
                   derived$vc_L[derived$drug == "zaleplon_M"],
                   tolerance = 1e-5)),
  isTRUE(all.equal(exp(uis[["Izat_2025_zaleplon"]]$theta[["lcl"]]), 71.58,
                   tolerance = 1e-5)),
  isTRUE(all.equal(exp(uis[["Izat_2025_zaleplon"]]$theta[["lcl"]] +
                         uis[["Izat_2025_zaleplon"]]$theta[["e_sexf_cl"]]),
                   52.51, tolerance = 1e-4)),
  isTRUE(all.equal(exp(uis[["Izat_2025_ziprasidone"]]$theta[["lvc"]]), 70.7,
                   tolerance = 1e-5)),
  isTRUE(all.equal(exp(uis[["Izat_2025_ziprasidone"]]$theta[["lcl"]]), 23.04,
                   tolerance = 1e-5)),
  isTRUE(all.equal(exp(uis[["Izat_2025_zoniporide"]]$theta[["lvc"]]), 119.0,
                   tolerance = 1e-5)),
  isTRUE(all.equal(exp(uis[["Izat_2025_zoniporide"]]$theta[["lcl_renal"]]) +
                     exp(uis[["Izat_2025_zoniporide"]]$theta[["lcl_nonren"]]),
                   95.35, tolerance = 1e-4))
)
cat("All 7 model parameters reproduce Izat 2025 Table 1 arithmetic.\n")
#> All 7 model parameters reproduce Izat 2025 Table 1 arithmetic.

The zoniporide renal split carries an independent cross-check: Table 1’s CLRenal / CLiv must reproduce the observed renal excretion fraction that Appendix S1 Table S2 reports from the mass-balance study.

fe_renal_implied  <- 16.2 / 95.35
fe_renal_reported <- 0.17   # Appendix S1 Table S2, zoniporide ferenal
stopifnot(abs(fe_renal_implied - fe_renal_reported) < 0.005)
cat(sprintf(paste("Zoniporide CLRenal / CLiv = %.3f reproduces the Table S2",
                  "renal excretion fraction of %.2f.\n"),
            fe_renal_implied, fe_renal_reported))
#> Zoniporide CLRenal / CLiv = 0.170 reproduces the Table S2 renal excretion fraction of 0.17.

Virtual cohort

The models have no random effects, so they are deterministic: one subject per arm fully characterises each scenario and there is nothing for a larger cohort to add. Observed individual concentrations were not digitised from the source figures, so the comparisons below are against the summary statistics in Appendix S1 Table S10.

Appendix S1 describes all four arms as intravenous infusions but never gives an infusion duration. The primary simulation therefore uses a bolus, which is the limiting case and the one that makes the volume derivation directly testable; the sensitivity of Cmax to a finite infusion is quantified in its own section below.

Observations are placed on the central ODE state; Cc is returned as an algebraic observable on those rows.

obs_grid <- sort(unique(c(
  seq(0,  2, by = 0.01),
  seq(2, 12, by = 0.05)
)))

make_iv_bolus <- function(id, dose, arm, sexf = 0) {
  dplyr::bind_rows(
    tibble::tibble(id = id, time = 0, amt = dose, evid = 1L,
                   cmt = "central", arm = arm, SEXF = sexf),
    tibble::tibble(id = id, time = obs_grid, amt = NA_real_, evid = 0L,
                   cmt = "central", arm = arm, SEXF = sexf)
  )
}

events_zal <- dplyr::bind_rows(
  make_iv_bolus(1L, 5, "zaleplon 5 mg i.v. (women)", sexf = 1),
  make_iv_bolus(2L, 5, "zaleplon 5 mg i.v. (men)",   sexf = 0)
) |> dplyr::arrange(id, time, dplyr::desc(evid))

events_zip <- make_iv_bolus(1L,  5, "ziprasidone 5 mg i.v.") |>
  dplyr::arrange(id, time, dplyr::desc(evid))
events_zon <- make_iv_bolus(1L, 80, "zoniporide 80 mg i.v.") |>
  dplyr::arrange(id, time, dplyr::desc(evid))

stopifnot(
  !anyDuplicated(events_zal[, c("id", "time", "evid")]),
  !anyDuplicated(events_zip[, c("id", "time", "evid")]),
  !anyDuplicated(events_zon[, c("id", "time", "evid")])
)

Simulation

solve_arm <- function(model_name, events) {
  out <- rxode2::rxSolve(readModelDb(model_name), events = events,
                         keep = "arm") |>
    as.data.frame()
  # rxSolve omits `id` for a single-subject event table (known pattern).
  if (is.null(out$id)) out$id <- 1L
  out$id <- as.integer(as.character(out$id))
  out
}

sim <- dplyr::bind_rows(
  solve_arm("Izat_2025_zaleplon",    events_zal),
  solve_arm("Izat_2025_ziprasidone", events_zip),
  solve_arm("Izat_2025_zoniporide",  events_zon)
)
#> Warning: multi-subject simulation without without 'omega'

# Concentrations must stay positive over the whole window, or PKNCA's
# terminal-slope fit will take log() of a negative solver artefact.
stopifnot(all(is.finite(sim$Cc)), all(sim$Cc >= 0))

# Give every arm a distinct subject id for the PKNCA grouping.
sim <- sim |>
  dplyr::mutate(subject = as.integer(factor(arm)))

Replicate published figures

Figure 3 panels (d) zoniporide, (e) zaleplon and (g) ziprasidone of Izat 2025 show the observed intravenous data as symbols against the simulated profile. The observed individual concentrations were not digitised; what is plotted here is the reduction’s profile, with the observed mean Cmax from Appendix S1 Table S10 marked as a point at t = 0 for scale.

obs_cmax <- tibble::tribble(
  ~arm,                          ~time, ~Cc,
  "zaleplon 5 mg i.v. (women)",      0, 62.7,
  "zaleplon 5 mg i.v. (men)",        0, 61.4,
  "ziprasidone 5 mg i.v.",           0, 83,
  "zoniporide 80 mg i.v.",           0, 590
)

ggplot(sim, aes(time, Cc)) +
  geom_line(linewidth = 0.7) +
  geom_point(data = obs_cmax, colour = "firebrick", size = 2) +
  facet_wrap(~ arm, scales = "free_y") +
  scale_y_log10() +
  labs(x = "Time (h)", y = "Plasma concentration (ng/mL)",
       caption = paste("Replicates the intravenous panels (d), (e) and (g) of",
                       "Figure 3 of Izat 2025.\nRed point = observed mean",
                       "Cmax (Appendix S1 Table S10).")) +
  theme_bw()

PKNCA validation

conc <- sim |>
  dplyr::filter(!is.na(Cc)) |>
  dplyr::select(subject, time, Cc, arm)

dose_df <- dplyr::bind_rows(events_zal, events_zip, events_zon) |>
  dplyr::filter(evid == 1L) |>
  dplyr::select(time, amt, arm) |>
  dplyr::mutate(subject = as.integer(factor(arm, levels = levels(factor(sim$arm)))))

stopifnot(nrow(conc) > 0, nrow(dose_df) == 4L)

conc_obj <- PKNCA::PKNCAconc(conc, Cc ~ time | arm + subject,
                             concu = "ng/mL", timeu = "h")
dose_obj <- PKNCA::PKNCAdose(dose_df, amt ~ time | arm + subject, doseu = "mg")

intervals <- data.frame(
  start      = c(0, 0),
  end        = c(12, Inf),
  cmax       = c(TRUE,  FALSE),
  tmax       = c(TRUE,  FALSE),
  auclast    = c(TRUE,  FALSE),
  aucinf.obs = c(FALSE, TRUE),
  half.life  = c(FALSE, TRUE)
)

nca <- PKNCA::pk.nca(PKNCA::PKNCAdata(conc_obj, dose_obj,
                                      intervals = intervals))

nca_df <- as.data.frame(nca) |>
  dplyr::filter(PPTESTCD %in% c("cmax", "tmax", "auclast",
                                "aucinf.obs", "half.life"))

knitr::kable(
  nca_df |>
    dplyr::mutate(PPORRES = signif(PPORRES, 4)) |>
    dplyr::select(arm, start, end, PPTESTCD, PPORRES),
  caption = "PKNCA results for the four simulated intravenous arms."
)

PKNCA results for the four simulated intravenous arms.

arm	start	end	PPTESTCD	PPORRES
zaleplon 5 mg i.v. (men)	0	12	auclast	69.8500
zaleplon 5 mg i.v. (men)	0	12	cmax	61.5800
zaleplon 5 mg i.v. (men)	0	12	tmax	0.0000
zaleplon 5 mg i.v. (men)	0	Inf	tmax	0.0000
zaleplon 5 mg i.v. (men)	0	Inf	half.life	0.7863
zaleplon 5 mg i.v. (men)	0	Inf	aucinf.obs	69.8500
zaleplon 5 mg i.v. (women)	0	12	auclast	95.1800
zaleplon 5 mg i.v. (women)	0	12	cmax	61.5800
zaleplon 5 mg i.v. (women)	0	12	tmax	0.0000
zaleplon 5 mg i.v. (women)	0	Inf	tmax	0.0000
zaleplon 5 mg i.v. (women)	0	Inf	half.life	1.0720
zaleplon 5 mg i.v. (women)	0	Inf	aucinf.obs	95.2200
ziprasidone 5 mg i.v.	0	12	auclast	212.7000
ziprasidone 5 mg i.v.	0	12	cmax	70.7200
ziprasidone 5 mg i.v.	0	12	tmax	0.0000
ziprasidone 5 mg i.v.	0	Inf	tmax	0.0000
ziprasidone 5 mg i.v.	0	Inf	half.life	2.1270
ziprasidone 5 mg i.v.	0	Inf	aucinf.obs	217.0000
zoniporide 80 mg i.v.	0	12	auclast	839.0000
zoniporide 80 mg i.v.	0	12	cmax	672.3000
zoniporide 80 mg i.v.	0	12	tmax	0.0000
zoniporide 80 mg i.v.	0	Inf	tmax	0.0000
zoniporide 80 mg i.v.	0	Inf	half.life	0.8651
zoniporide 80 mg i.v.	0	Inf	aucinf.obs	839.0000

Closed-form gates

Two identities must hold exactly for a one-compartment intravenous bolus, and they check the NCA setup as much as the model:

get_nca <- function(a, p) {
  v <- nca_df$PPORRES[nca_df$arm == a & nca_df$PPTESTCD == p]
  if (length(v) != 1L) stop("no unique NCA row for '", a, "' / ", p)
  v
}

gate <- tibble::tribble(
  ~arm,                          ~dose, ~cl_model,
  "zaleplon 5 mg i.v. (women)",      5,     52.51,
  "zaleplon 5 mg i.v. (men)",        5,     71.58,
  "ziprasidone 5 mg i.v.",           5,     23.04,
  "zoniporide 80 mg i.v.",          80,     95.35
) |>
  dplyr::rowwise() |>
  dplyr::mutate(
    aucinf_nca    = get_nca(arm, "aucinf.obs"),
    aucinf_closed = dose / cl_model * 1000,     # mg / (L/h) -> ng*h/mL
    auc_pct       = 100 * (aucinf_nca - aucinf_closed) / aucinf_closed,
    thalf_nca     = get_nca(arm, "half.life"),
    thalf_closed  = log(2) / (cl_model / exp(
      uis[[dplyr::case_when(
        grepl("zaleplon",    arm) ~ "Izat_2025_zaleplon",
        grepl("ziprasidone", arm) ~ "Izat_2025_ziprasidone",
        TRUE                      ~ "Izat_2025_zoniporide"
      )]]$theta[["lvc"]])),
    thalf_pct     = 100 * (thalf_nca - thalf_closed) / thalf_closed
  ) |>
  dplyr::ungroup()

knitr::kable(gate |> dplyr::mutate(dplyr::across(where(is.numeric),
                                                 \(x) signif(x, 5))),
             caption = "NCA output vs. the closed-form one-compartment identities.")

NCA output vs. the closed-form one-compartment identities.

arm	dose	cl_model	aucinf_nca	aucinf_closed	auc_pct	thalf_nca	thalf_closed	thalf_pct
zaleplon 5 mg i.v. (women)	5	52.51	95.220	95.220	-0.0001063	1.07190	1.07190	-2.74e-05
zaleplon 5 mg i.v. (men)	5	71.58	69.852	69.852	-0.0000536	0.78630	0.78630	1.00e-07
ziprasidone 5 mg i.v.	5	23.04	217.010	217.010	-0.0000189	2.12700	2.12700	-2.00e-07
zoniporide 80 mg i.v.	80	95.35	839.010	839.010	-0.0000635	0.86507	0.86507	0.00e+00

# Both identities are exact for a one-compartment bolus, so the only error is
# ODE-solver integration error. Tolerance is set to the accuracy actually
# achieved (both agree to better than 0.001%), not to a loose round number.
stopifnot(nrow(gate) == 4L,
          max(abs(gate$auc_pct))   < 0.05,
          max(abs(gate$thalf_pct)) < 0.05)
cat(sprintf(paste("AUCinf reproduces dose/CL to within %.3f%% and the terminal",
                  "half-life reproduces log(2)*vc/cl to within %.3f%% across",
                  "all %d arms.\n"),
            max(abs(gate$auc_pct)), max(abs(gate$thalf_pct)), nrow(gate)))
#> AUCinf reproduces dose/CL to within 0.000% and the terminal half-life reproduces log(2)*vc/cl to within 0.000% across all 4 arms.

Comparison against the published NCA

Appendix S1 Table S10 reports the observed Cmax, AUC-infinity and intravenous clearance for every arm. This is the comparison that decides whether the reduction is faithful. Nothing was tuned to reach it.

published_obs <- tibble::tribble(
  ~arm,                          ~cmax, ~aucinf.obs,
  "zaleplon 5 mg i.v. (women)",   62.7,         100,
  "zaleplon 5 mg i.v. (men)",     61.4,        71.7,
  "ziprasidone 5 mg i.v.",          83,         217,
  "zoniporide 80 mg i.v.",         590,         839
)

cmp_obs <- nlmixr2lib::ncaComparisonTable(
  simulated = nca_df,
  reference = published_obs,
  by        = "arm",
  units     = c(cmax = "ng/mL", aucinf.obs = "ng*h/mL"),
  tolerance_pct = 20
)

knitr::kable(
  cmp_obs,
  caption = paste("Simulated vs. observed values from Izat 2025 Appendix S1",
                  "Table S10. * differs from the reference by >20%."),
  align = c("l", "l", "r", "r", "r")
)

Simulated vs. observed values from Izat 2025 Appendix S1 Table S10. * differs from the reference by >20%.

NCA parameter	arm	Reference	Simulated	% diff
Cmax (ng/mL)	zaleplon 5 mg i.v. (women)	62.7	61.6	-1.8%
Cmax (ng/mL)	zaleplon 5 mg i.v. (men)	61.4	61.6	+0.3%
Cmax (ng/mL)	ziprasidone 5 mg i.v.	83	70.7	-14.8%
Cmax (ng/mL)	zoniporide 80 mg i.v.	590	672	+13.9%
AUC0-∞ (obs) (ng*h/mL)	zaleplon 5 mg i.v. (women)	100	95.2	-4.8%
AUC0-∞ (obs) (ng*h/mL)	zaleplon 5 mg i.v. (men)	71.7	69.9	-2.6%
AUC0-∞ (obs) (ng*h/mL)	ziprasidone 5 mg i.v.	217	217	+0.0%
AUC0-∞ (obs) (ng*h/mL)	zoniporide 80 mg i.v.	839	839	+0.0%

# `% diff` is formatted as text (a trailing "*" marks rows over tolerance),
# so strip the formatting before testing.
pct <- suppressWarnings(
  as.numeric(gsub("[^0-9.eE+-]", "", as.character(cmp_obs[["% diff"]])))
)
stopifnot(sum(is.finite(pct)) == 8L, max(abs(pct), na.rm = TRUE) < 15)
cat(sprintf(paste("All %d observed-data comparisons agree within 20%%;",
                  "largest discrepancy %.1f%%.\n"),
            sum(is.finite(pct)), max(abs(pct), na.rm = TRUE)))
#> All 8 observed-data comparisons agree within 20%; largest discrepancy 14.8%.

Reading the table:

• Zaleplon is the strong result. Cmax agrees to within 1.8% in women and 0.3% in men, which is the sharpest available test of the volume derivation because Cmax after a bolus is exactly dose / vc. That the same volume fits both sexes while the clearances differ by 36% is a real consistency check, not a fitted outcome. This is also the compound whose reduction is exact (Q = 0).
• AUC-infinity is true by construction wherever the paper’s reported clearance and AUC are reciprocal, which they are exactly for ziprasidone (5 mg / 23.04 L/h = 217 ngh/mL) and zoniporide (80 mg / 95.35 L/h = 839 ngh/mL). It is not an independent check for those two. For zaleplon the reported clearance and AUC are not quite reciprocal (5 mg / 52.5 L/h = 95.2, against a reported 100 ng*h/mL), because Table S10’s clearance is a mean of individual clearances rather than the reciprocal of the mean AUC; the resulting 4.8% and 2.6% gaps are that discrepancy, not model error.
• Ziprasidone and zoniporide Cmax carry the cost of the one-compartment approximation – 14.8% low and 13.9% high respectively. Both are full-PBPK compounds whose genuinely multi-exponential distribution collapses to a single exponential when only a lumped Vss is published. The sign differs because a mono-exponential fit to a compound with fast initial distribution (ziprasidone) understates the early peak, while one with slow distribution (zoniporide) overstates it.

Comparison against the paper’s own PBPK predictions

Table S10 also reports what the Simcyp model predicted, under its best (“top-down, with observed fmAO and CLint,u”) parameterisation. Comparing the reduction against the platform’s own output separates “the reduction is wrong” from “the platform model was already off”.

published_pred <- tibble::tribble(
  ~arm,                          ~cmax, ~aucinf.obs,
  "zaleplon 5 mg i.v. (women)",   58.1,        92.6,
  "zaleplon 5 mg i.v. (men)",     48.5,        68.9,
  "ziprasidone 5 mg i.v.",        85.6,         220,
  "zoniporide 80 mg i.v.",         538,         719
)

cmp_pred <- nlmixr2lib::ncaComparisonTable(
  simulated = nca_df,
  reference = published_pred,
  by        = "arm",
  units     = c(cmax = "ng/mL", aucinf.obs = "ng*h/mL"),
  tolerance_pct = 20
)

knitr::kable(
  cmp_pred,
  caption = paste("Simulated vs. the Simcyp top-down predictions in Izat 2025",
                  "Appendix S1 Table S10. * differs by >20%."),
  align = c("l", "l", "r", "r", "r")
)

Simulated vs. the Simcyp top-down predictions in Izat 2025 Appendix S1 Table S10. * differs by >20%.

NCA parameter	arm	Reference	Simulated	% diff
Cmax (ng/mL)	zaleplon 5 mg i.v. (women)	58.1	61.6	+6.0%
Cmax (ng/mL)	zaleplon 5 mg i.v. (men)	48.5	61.6	+27.0%*
Cmax (ng/mL)	ziprasidone 5 mg i.v.	85.6	70.7	-17.4%
Cmax (ng/mL)	zoniporide 80 mg i.v.	538	672	+25.0%*
AUC0-∞ (obs) (ng*h/mL)	zaleplon 5 mg i.v. (women)	92.6	95.2	+2.8%
AUC0-∞ (obs) (ng*h/mL)	zaleplon 5 mg i.v. (men)	68.9	69.9	+1.4%
AUC0-∞ (obs) (ng*h/mL)	ziprasidone 5 mg i.v.	220	217	-1.4%
AUC0-∞ (obs) (ng*h/mL)	zoniporide 80 mg i.v.	719	839	+16.7%

Where the two disagree by more than they disagree with the observed data – the zaleplon male arm most clearly, where the reduction is 1.27 times the Simcyp prediction but 1.00 times observed – the reduction is the closer of the two to the data. That is expected: the reduction is handed the observed clearance and the fitted volume directly, whereas the Simcyp prediction propagates them through the in-vitro scaling layer that the paper is actually evaluating.

Sensitivity to the unreported infusion duration

The source calls all four arms intravenous infusions but never states a duration, so the simulations above use a bolus. For a one-compartment model the effect of a finite infusion on Cmax is closed-form, Cmax = dose / (CL * T) * (1 - exp(-kel * T)), so the sensitivity can be tabulated without re-simulating.

inf_sens <- tidyr::crossing(
  gate |> dplyr::select(arm, dose, cl_model),
  T_h = c(0, 5, 15, 30, 60) / 60
) |>
  dplyr::left_join(
    derived |>
      dplyr::mutate(arm = c("zaleplon 5 mg i.v. (women)",
                            "zaleplon 5 mg i.v. (men)",
                            "ziprasidone 5 mg i.v.",
                            "zoniporide 80 mg i.v.")) |>
      dplyr::select(arm, vc_L, kel_perh),
    by = "arm"
  ) |>
  dplyr::left_join(published_obs |> dplyr::select(arm, cmax_obs = cmax),
                   by = "arm") |>
  dplyr::mutate(
    cmax_pred = dplyr::if_else(
      T_h == 0,
      dose / vc_L * 1000,
      dose / (cl_model * T_h) * (1 - exp(-kel_perh * T_h)) * 1000
    ),
    ratio = cmax_pred / cmax_obs
  )

knitr::kable(
  inf_sens |>
    dplyr::mutate(`Infusion (min)` = round(T_h * 60)) |>
    dplyr::select(arm, `Infusion (min)`, cmax_pred, cmax_obs, ratio) |>
    dplyr::mutate(dplyr::across(where(is.numeric), \(x) signif(x, 4))) |>
    dplyr::rename("Cmax predicted (ng/mL)" = cmax_pred,
                  "Cmax observed (ng/mL)"  = cmax_obs,
                  "Predicted / observed"   = ratio),
  caption = paste("Cmax sensitivity to the unreported infusion duration.",
                  "The 0 min row is the bolus used above.")
)

Cmax sensitivity to the unreported infusion duration. The 0 min row is the bolus used above.

arm	Infusion (min)	Cmax predicted (ng/mL)	Cmax observed (ng/mL)	Predicted / observed
zaleplon 5 mg i.v. (men)	0	61.58	61.4	1.0030
zaleplon 5 mg i.v. (men)	5	59.37	61.4	0.9669
zaleplon 5 mg i.v. (men)	15	55.26	61.4	0.9001
zaleplon 5 mg i.v. (men)	30	49.80	61.4	0.8111
zaleplon 5 mg i.v. (men)	60	40.92	61.4	0.6665
zaleplon 5 mg i.v. (women)	0	61.58	62.7	0.9821
zaleplon 5 mg i.v. (women)	5	59.95	62.7	0.9561
zaleplon 5 mg i.v. (women)	15	56.86	62.7	0.9068
zaleplon 5 mg i.v. (women)	30	52.61	62.7	0.8391
zaleplon 5 mg i.v. (women)	60	45.35	62.7	0.7232
ziprasidone 5 mg i.v.	0	70.72	83.0	0.8521
ziprasidone 5 mg i.v.	5	69.77	83.0	0.8406
ziprasidone 5 mg i.v.	15	67.92	83.0	0.8183
ziprasidone 5 mg i.v.	30	65.26	83.0	0.7863
ziprasidone 5 mg i.v.	60	60.35	83.0	0.7272
zoniporide 80 mg i.v.	0	672.30	590.0	1.1390
zoniporide 80 mg i.v.	5	650.30	590.0	1.1020
zoniporide 80 mg i.v.	15	609.20	590.0	1.0330
zoniporide 80 mg i.v.	30	553.90	590.0	0.9388
zoniporide 80 mg i.v.	60	462.50	590.0	0.7839

A 15-minute infusion moves zoniporide from 1.14 to about 1.03 times observed and zaleplon from about 1.00 to 0.92; there is no single duration that improves all four arms, and none is reported, so the bolus is retained as the unambiguous limiting case. Users who know the duration of a particular study can supply it through the event table without changing the model.

Assumptions and deviations

1. Body weight of 70 kg is assumed. Izat 2025 Table 1 states every volume in L/kg, and no body weight appears anywhere in the paper or in Appendix S1 – Table S3 gives only age range and female fraction. The 70 kg reference is the standing rounded-standard convention. It is the only value in these three models that does not come from the source. WT is recorded in each model’s covariatesDataExcluded rather than carried as a live covariate, because the corresponding weight-scaling of clearance lives in the unpublished Simcyp population file; scaling volume with weight while holding clearance fixed would make the model internally inconsistent.
2. The oral arms were deliberately not reduced. As set out above, oral bioavailability needs a hepatic blood flow Q_H that the paper never prints. Q_H is not a rounding-level assumption here: sweeping it from 70 to 110 L/h moves predicted zaleplon oral Cmax from 0.62 to 1.29 times observed. Supplying the Simcyp platform default would be an unsourced substitution, so the oral arms are out of scope for this package. Capmatinib, idelalisib and lenvatinib are oral-only in this paper and are therefore not extracted from it at all; capmatinib additionally fails cross-dose consistency (observed oral clearance 36.1 L/h at 200 mg versus 51.7 L/h at 600 mg, against a single tabulated 42.3 L/h).
3. Infusion duration is not reported for any of the four arms. A bolus is used; the sensitivity section above quantifies what a finite infusion would change.
4. Ziprasidone and zoniporide are mono-exponential approximations. Their source models are full PBPK and only a lumped Vss is published, so no distribution phase can be reconstructed. Cmax is 14.8% low and 13.9% high respectively. Zaleplon’s reduction, by contrast, is exact: its reported inter-compartmental clearance is zero.
5. No inter-individual variability and no residual error. Izat 2025 is a PBPK simulation study, not a population-PK fit. The percent-CV values in Table 1 and Table S10 are the spread of a Simcyp virtual population driven by unpublished population files, not estimated variance components. Rather than invent variances, there are no etas and propSd is fixed at zero, so all three models are deterministic typical-value simulators.
6. The clearance values are other studies’ estimates. Table 1 footnote c states that CLiv/oral was “obtained from in vivo pharmacokinetic studies, which are listed in Table S2” (Table S3 in the published supplement). These are literature observations that Izat 2025 fed into Simcyp, not parameters Izat 2025 estimated. Each model’s population$studies names the source study.
7. No fraction-metabolised parameters are encoded. The observed fmAO / fmCYP3A4 splits (Appendix S1 Table S2: zaleplon 0.57-0.74 / 0.26-0.43; ziprasidone 0.67 / 0.33; zoniporide 0.52-0.69 AO with no CYP3A4) are recorded in each model’s population$notes for provenance but are not parameters, because they do not affect the plasma concentration-time profile and their pathway attribution depends on the unpublished platform layer. Zoniporide’s renal component is the exception and is encoded, as lcl_renal, because Table 1 prints it as a clearance and it cross-validates against the Table S2 excretion fraction.
8. All DDI results are out of scope. Every CYP3A4 area-ratio prediction in the paper depends on proprietary Certara compound files for ketoconazole, itraconazole, rifampicin and carbamazepine. The static AO-DDI screen (Equations S14-S16) is algebraic with no state variable and no time course, so it is not expressible as an rxode2 model either.
9. Ziprasidone has no AO reaction-phenotyping data. Table 1 footnote f records that no data with an aldehyde oxidase inhibitor were available, so its hepatocyte AO intrinsic clearance is structurally unavailable rather than merely unreported. This does not affect the reduction, which uses the observed total clearance.
10. fucyt was measured in dog liver cytosol, not human (Table 1 footnote d), on the stated assumption that non-specific cytosolic binding is species-independent – dogs do not express functionally active aldehyde oxidase, which is the point of the design. This affects the in-vitro scaling layer only, not the reduction.