Skip to contents

Amodiaquine (Ali 2018)

Source: vignettes/articles/Ali_2018_amodiaquine.Rmd
Ali_2018_amodiaquine.Rmd

Model and source

  • Citation: Ali AM, Penny MA, Smith TA, Workman L, Sasi P, Adjei GO, Aweeka F, Kiechel J-R, Jullien V, Rijken MJ, McGready R, Mwesigwa J, Kristensen K, Stepniewska K, Tarning J, Barnes KI, Denti P; WWARN Amodiaquine PK Study Group (2018). Population pharmacokinetics of the antimalarial amodiaquine: a pooled analysis to optimize dosing. Antimicrobial Agents and Chemotherapy 62(10):e02193-17. doi:10.1128/AAC.02193-17.
  • Article: https://doi.org/10.1128/AAC.02193-17
  • PubMed: https://pubmed.ncbi.nlm.nih.gov/30038041/ (PMID 30038041; open access).
  • Supplemental material referenced by the publication (Fig S1, S2; basic GOF and age-stratified VPCs) was not on disk during this extraction and is not used; all model values are sourced from the main paper.

The package model can be loaded with:

mod_fn <- readModelDb("Ali_2018_amodiaquine")
mod    <- rxode2::rxode2(mod_fn())

Population

The Ali 2018 pooled analysis combined patient-level data from six studies (five cohorts) of artemisinin-based combination therapy of uncomplicated Plasmodium falciparum (and one P. vivax sub-cohort) malaria in five countries (Burkina Faso, Ghana, Kenya, Uganda, Thailand). A total of 261 patients with weights 6.5-93 kg and ages 1-60 years contributed 2,920 plasma samples. About 36% of patients were children under 5 years and 12.6% were under 2 years; 26 of the patients were pregnant women in their second or third trimester (Thailand sub-cohort). All subjects received daily oral amodiaquine 10 mg/kg of body weight (administered as the amodiaquine base) for three days, either alone or as artesunate + amodiaquine (loose tablets or fixed-dose combination). Baseline demographics are reported in Ali 2018 Table 1; per-study sampling schedules and assay LLOQs are in Ali 2018 Table 2.

str(attr(rxode2::rxode2(mod_fn()), "metadata")$population)
#>  NULL

Source trace

Every parameter and equation traces back to the main paper. Table 3 (page 7 of the publication) lists final parameter estimates and bootstrap 95% CIs; the Methods section (“Effect of body size and age”, Equation 1) gives the structural maturation formula.

Equation / parameter Value Source location
lka = log(0.589) (Ka, 1/h) 0.589 Ali 2018 Table 3 (Amodiaquine, Ka)
lmtt = log(0.236) (MTT, h) 0.236 Ali 2018 Table 3 (MTT)
lcl = log(2960) (CL_AQ, L/h, WT=50 kg) 2960 Ali 2018 Table 3 (CL_AQ)
lvc = log(13500) (Vc_AQ, L) 13,500 Ali 2018 Table 3 (Vc_AQ)
lq = log(2310) (Q_AQ, L/h) 2,310 Ali 2018 Table 3 (Q_AQ)
lvp = log(22700) (Vp_AQ, L) 22,700 Ali 2018 Table 3 (Vp_AQ)
lfdepot = fixed(log(1)) (F) 1 (fixed) Ali 2018 Table 3 (F = 1 Fixed)
lcl_deaq = log(32.6) 32.6 L/h Ali 2018 Table 3 (CL_DEAQ)
lvc_deaq = log(258) 258 L Ali 2018 Table 3 (Vc_DEAQ)
lq_deaq = log(154) 154 L/h Ali 2018 Table 3 (Q1_DEAQ)
lvp_deaq = log(2460) 2,460 L Ali 2018 Table 3 (Vp1_DEAQ)
lq2_deaq = log(31.3) 31.3 L/h Ali 2018 Table 3 (Q2_DEAQ)
lvp2_deaq = log(5580) 5,580 L Ali 2018 Table 3 (Vp2_DEAQ)
e_wt_cl = fixed(0.75), e_wt_vc = fixed(1.0) 0.75 / 1.0 Ali 2018 Methods, “Effect of body size and age” (allometric scaling)
pma50_aq = 11.8, hill_aq = 3.6 11.8 mo / 3.6 Ali 2018 Table 3 (PMA50 for AQ; Hill factor for AQ)
pma50_deaq = 12.9, hill_deaq = 3.22 12.9 mo / 3.22 Ali 2018 Table 3 (PMA50 for DEAQ; Hill factor for DEAQ)
e_cycle1_fdepot = -0.224 -22.4% Ali 2018 Table 3 (Effect of first dose on F)
BSV CL_AQ (etalcl) 32.2% CV Ali 2018 Table 3 (BSV CL_AQ)
BSV Vc_AQ (etalvc) 53.1% CV Ali 2018 Table 3 (BSV Vc_AQ)
BSV CL_DEAQ (etalcl_deaq) 20.0% CV Ali 2018 Table 3 (BSV CL_DEAQ)
BSV Vc_DEAQ (etalvc_deaq) 67.2% CV Ali 2018 Table 3 (BSV Vc_DEAQ)
2-compartment AQ + 3-compartment DEAQ with 2 transit compartments (NN = 2) – Ali 2018 Results, “Population pharmacokinetic model” (Figure 1 schematic)
KTR = (NN + 1) / MTT = 3 / MTT for the transit chain – Standard Savic 2007 transit-absorption convention used by the source paper
Complete in-vivo metabolism of AQ to DEAQ with molar correction MW_DEAQ / MW_AQ = 327.81 / 355.85 = 0.9212 – Ali 2018 Methods, “Structural model”
(WT / 50)^0.75 on CL and Q for both AQ and DEAQ; (WT / 50)^1 on Vc and Vp; reference 50 kg = pooled median – Ali 2018 Table 3 footnote c + Methods
Sigmoidal Hill maturation: CL_i = CL_TV * (BW/50)^0.75 * PMA^Hill / (PMA^Hill + PMA50^Hill) (Equation 1) – Ali 2018 Methods, “Effect of body size and age” (Equation 1)
First-day F effect: f(depot) = exp(lfdepot) * (1 + e_cycle1_fdepot * (CYCLE == 1)) – Ali 2018 Results, “Covariate effect” (paper attributes to a transient malaria disease effect on absorption)
Combined additive + proportional residual error per analyte – Ali 2018 Methods, “Stochastic model” + Table 3 footnote d

Virtual cohort

We construct typical-value patient profiles spanning the weight bands used in the published Figure 4 dosing-simulation: 8 kg infant (~7 mo postnatal age), 15 kg toddler (~3 yr), 21 kg cohort-median child, 30 kg older child (~9 yr), 50 kg adult, and 70 kg adult. Each is dosed once daily for three days at the WHO manufacturer recommendation reproduced in Ali 2018 Table 4 (current regimen): 67.5 mg/day (4-8 kg), 135 mg/day (9-17 kg), 270 mg/day (18-35 kg), 540 mg/day (>= 36 kg). The 50 kg / 540 mg/day arm is the paper’s reference for the day-7 exposure target.

set.seed(20260530L)

current_dose <- function(wt) {
  ifelse(wt <= 8,  67.5,
  ifelse(wt <= 17, 135,
  ifelse(wt <= 35, 270, 540)))
}

postnatal_to_page <- function(pna_months) pna_months + 9  # term gestation assumed

profiles <- tibble::tribble(
  ~label,          ~wt_kg, ~pna_months,
  "8 kg infant",        8,         7,
  "15 kg toddler",     15,        36,
  "21 kg child (cohort median)", 21, 72,
  "30 kg child",       30,       108,
  "50 kg adult",       50,       240,
  "70 kg adult",       70,       360
) |>
  mutate(
    page_months = postnatal_to_page(pna_months),
    dose_mg     = current_dose(wt_kg)
  )
knitr::kable(profiles, caption = "Typical-value profiles spanning Ali 2018 Figure 4 weight bands; PAGE assumes term gestation (PNA + 9 months); dose follows the current manufacturer recommendation reproduced in Ali 2018 Table 4.")
Typical-value profiles spanning Ali 2018 Figure 4 weight bands; PAGE assumes term gestation (PNA + 9 months); dose follows the current manufacturer recommendation reproduced in Ali 2018 Table 4.
label wt_kg pna_months page_months dose_mg
8 kg infant 8 7 16 67.5
15 kg toddler 15 36 45 135.0
21 kg child (cohort median) 21 72 81 270.0
30 kg child 30 108 117 270.0
50 kg adult 50 240 249 540.0
70 kg adult 70 360 369 540.0

Build the event table. Each subject receives three daily doses; observation times sample the first 28 days densely enough to capture both the AQ absorption peak and the slow DEAQ terminal decline.

obs_times <- c(seq(0, 12, by = 0.25),
               seq(12, 72, by = 1),
               seq(72, 168, by = 4),
               seq(168, 672, by = 12))
obs_times <- sort(unique(obs_times))

make_subject <- function(i, label, wt_kg, page_months, dose_mg) {
  doses <- data.frame(
    id    = i,
    time  = c(0, 24, 48),
    evid  = 1L,
    amt   = dose_mg,
    cmt   = 1L,         # depot
    CYCLE = c(1L, 2L, 3L)
  )
  obs_aq <- data.frame(
    id    = i, time = obs_times, evid = 0L, amt = 0,
    cmt   = 9L,         # Cc (AQ) output dvid
    CYCLE = 3L          # CYCLE carried forward; only matters on dose rows
  )
  obs_deaq <- data.frame(
    id    = i, time = obs_times, evid = 0L, amt = 0,
    cmt   = 10L,        # Cc_deaq output dvid
    CYCLE = 3L
  )
  out <- dplyr::bind_rows(doses, obs_aq, obs_deaq)
  out$label <- label
  out$WT    <- wt_kg
  out$PAGE  <- page_months
  out
}

events_typical <- profiles |>
  dplyr::mutate(i = dplyr::row_number()) |>
  purrr::pmap_dfr(\(i, label, wt_kg, pna_months, page_months, dose_mg)
                  make_subject(i, label, wt_kg, page_months, dose_mg))

events_typical <- events_typical[order(events_typical$id, events_typical$time, -events_typical$evid),]
stopifnot(!anyDuplicated(unique(events_typical[, c("id", "time", "evid", "cmt")])))

Simulation

Typical-value simulation (between-subject variability set to zero); the residual error is left in place but, because rxSolve returns Cc / Cc_deaq without sampling residual noise by default, the trajectories are deterministic typical-value predictions.

mod_typical <- rxode2::zeroRe(mod)

sim_typical <- rxode2::rxSolve(
  mod_typical,
  events = events_typical,
  keep   = c("label", "WT", "PAGE")
) |>
  as.data.frame()
#> ℹ omega/sigma items treated as zero: 'etalcl', 'etalvc', 'etalcl_deaq', 'etalvc_deaq'
#> Warning: multi-subject simulation without without 'omega'

For a 50-stochastic-subject VPC representing the reference 50 kg adult arm, draw between-subject random effects from the model’s omega covariance and re-simulate.

n_vpc <- 50L
vpc_profiles <- tibble::tibble(
  id = seq_len(n_vpc),
  WT = 50,
  PAGE = 240,
  dose_mg = 540
)

vpc_events <- vpc_profiles |>
  dplyr::mutate(i = dplyr::row_number()) |>
  purrr::pmap_dfr(\(id, WT, PAGE, dose_mg, i) {
    doses <- data.frame(
      id    = id, time = c(0, 24, 48), evid = 1L,
      amt   = dose_mg, cmt = 1L, CYCLE = c(1L, 2L, 3L)
    )
    obs_aq <- data.frame(
      id    = id, time = obs_times, evid = 0L, amt = 0,
      cmt   = 9L, CYCLE = 3L
    )
    obs_deaq <- data.frame(
      id    = id, time = obs_times, evid = 0L, amt = 0,
      cmt   = 10L, CYCLE = 3L
    )
    df <- dplyr::bind_rows(doses, obs_aq, obs_deaq)
    df$WT <- WT
    df$PAGE <- PAGE
    df
  })
vpc_events <- vpc_events[order(vpc_events$id, vpc_events$time, -vpc_events$evid),]

sim_vpc <- rxode2::rxSolve(mod, events = vpc_events) |>
  as.data.frame()

Replicate published figures

Day-7 desethylamodiaquine across weight bands (Ali 2018 Figure 4A)

Figure 4A of Ali 2018 plots Day-7 desethylamodiaquine plasma concentrations by 1-kg weight band under current manufacturer dosing and shows that patients at 8 kg, 15-17 kg, 33-35 kg, and >62 kg achieve about 25% lower exposure than the 50 kg reference. We reproduce the weight-band-by-weight-band typical-value Day-7 prediction below.

day7_typical <- sim_typical |>
  dplyr::filter(time == 168) |>
  dplyr::distinct(label, .keep_all = TRUE) |>
  dplyr::select(label, WT, Cc_deaq) |>
  dplyr::left_join(profiles |> dplyr::select(label, dose_mg), by = "label") |>
  dplyr::transmute(
    label,
    `Weight (kg)`   = WT,
    `Dose (mg/day)` = dose_mg,
    `Day-7 Cc_DEAQ (ng/mL)` = round(Cc_deaq, 1)
  )
knitr::kable(day7_typical,
             caption = "Typical-value Day-7 desethylamodiaquine plasma concentration by weight band under the current manufacturer dose regimen (Ali 2018 Table 4 left columns). The 50 kg arm is the published reference (target = 80% of median for efficacy, equivalent to 54 ng/mL per Ali 2018; ratio Day-7 / 50-kg reference reveals the >25% underdosed bands).")
Typical-value Day-7 desethylamodiaquine plasma concentration by weight band under the current manufacturer dose regimen (Ali 2018 Table 4 left columns). The 50 kg arm is the published reference (target = 80% of median for efficacy, equivalent to 54 ng/mL per Ali 2018; ratio Day-7 / 50-kg reference reveals the >25% underdosed bands).
label Weight (kg) Dose (mg/day) Day-7 Cc_DEAQ (ng/mL)
8 kg infant 8 67.5 51.0
15 kg toddler 15 135.0 41.7
21 kg child (cohort median) 21 270.0 63.5
30 kg child 30 270.0 48.5
50 kg adult 50 540.0 66.2
70 kg adult 70 540.0 51.6 
sim_typical |>
  dplyr::filter(time > 0, time <= 240, Cc_deaq > 0) |>
  ggplot(aes(time, Cc_deaq, colour = label)) +
  geom_line(linewidth = 0.7) +
  geom_vline(xintercept = 168, linetype = "dashed", colour = "grey50") +
  scale_y_log10() +
  labs(x = "Time after first dose (h)",
       y = "Cc_DEAQ (ng/mL, log)",
       colour = "Weight band",
       title = "Typical-value desethylamodiaquine trajectories by weight band",
       caption = "Replicates the body of Ali 2018 Figure 4A (typical-value layer). Dashed line marks Day 7 (168 h).")

Parent + metabolite trajectory for a typical 50-kg adult (visual companion to Ali 2018 Figure 2) 

ref_adult <- sim_typical |>
  dplyr::filter(label == "50 kg adult", time > 0)

ref_adult |>
  tidyr::pivot_longer(c(Cc, Cc_deaq), names_to = "analyte", values_to = "ng_mL") |>
  dplyr::mutate(analyte = dplyr::recode(analyte,
                                        Cc      = "Amodiaquine (AQ)",
                                        Cc_deaq = "Desethylamodiaquine (DEAQ)")) |>
  dplyr::filter(ng_mL > 0.05) |>
  ggplot(aes(time, ng_mL, colour = analyte)) +
  geom_line(linewidth = 0.7) +
  scale_y_log10() +
  labs(x = "Time after first dose (h)",
       y = "Plasma concentration (ng/mL, log)",
       colour = NULL,
       title = "Typical 50 kg adult on 540 mg/day for 3 days",
       caption = "Companion to Ali 2018 Figure 2 (study-stratified VPCs); typical-value layer only.")

50-subject VPC at 50 kg adult arm 

vpc_summary <- sim_vpc |>
  dplyr::filter(time > 0, time <= 720) |>
  dplyr::group_by(time) |>
  dplyr::summarise(
    Q05_aq = quantile(Cc, 0.05, na.rm = TRUE),
    Q50_aq = quantile(Cc, 0.50, na.rm = TRUE),
    Q95_aq = quantile(Cc, 0.95, na.rm = TRUE),
    Q05_deaq = quantile(Cc_deaq, 0.05, na.rm = TRUE),
    Q50_deaq = quantile(Cc_deaq, 0.50, na.rm = TRUE),
    Q95_deaq = quantile(Cc_deaq, 0.95, na.rm = TRUE),
    .groups = "drop"
  )

vpc_long <- vpc_summary |>
  tidyr::pivot_longer(-time, names_to = c("q", "analyte"), names_pattern = "(.*)_(.*)") |>
  tidyr::pivot_wider(names_from = q, values_from = value) |>
  dplyr::mutate(analyte = dplyr::recode(analyte,
                                        aq   = "Amodiaquine",
                                        deaq = "Desethylamodiaquine"))

vpc_long |>
  dplyr::filter(Q50 > 0.1) |>
  ggplot(aes(time, Q50)) +
  geom_ribbon(aes(ymin = Q05, ymax = Q95), alpha = 0.25, fill = "#3a7ca5") +
  geom_line(linewidth = 0.7, colour = "#3a7ca5") +
  facet_wrap(~analyte, scales = "free_y") +
  scale_y_log10() +
  labs(x = "Time after first dose (h)",
       y = "Plasma concentration (ng/mL, log)",
       title = "50-subject simulated VPC, 50 kg adult on 540 mg/day x 3 days",
       caption = "5th, 50th, 95th percentiles. BSV applied per Ali 2018 Table 3; BOV on Ka, MTT, F not modelled.")

PKNCA validation

We compute NCA parameters per weight band using the typical-value trajectory. The PKNCA call uses Cc_deaq ~ time | label/id so per-weight-band summaries can be compared against any published per-band exposure.

sim_nca_deaq <- sim_typical |>
  dplyr::filter(!is.na(Cc_deaq), time >= 0) |>
  dplyr::transmute(id, time, Cc_deaq, label) |>
  dplyr::distinct(id, time, label, .keep_all = TRUE)

conc_obj_deaq <- PKNCA::PKNCAconc(sim_nca_deaq, Cc_deaq ~ time | label + id)

dose_df <- events_typical |>
  dplyr::filter(evid == 1) |>
  dplyr::transmute(id, time, amt, label)

dose_obj <- PKNCA::PKNCAdose(dose_df, amt ~ time | label + id)

intervals_deaq <- data.frame(
  start = 0, end = 672,
  cmax        = TRUE,
  tmax        = TRUE,
  aucinf.obs  = TRUE,
  half.life   = TRUE
)

nca_data_deaq <- PKNCA::PKNCAdata(conc_obj_deaq, dose_obj, intervals = intervals_deaq)
nca_res_deaq  <- PKNCA::pk.nca(nca_data_deaq)

nca_summary_deaq <- as.data.frame(nca_res_deaq$result) |>
  dplyr::filter(PPTESTCD %in% c("cmax", "tmax", "aucinf.obs", "half.life")) |>
  dplyr::transmute(
    `Weight band` = label,
    `NCA parameter` = PPTESTCD,
    `Value (DEAQ)`  = signif(PPORRES, 3)
  )

knitr::kable(nca_summary_deaq,
             caption = "Typical-value NCA parameters for desethylamodiaquine (DEAQ) by weight band over 0-672 h (28 days). Cmax in ng/mL; Tmax in h; AUCinf.obs in ng*h/mL; half-life in h.")
Typical-value NCA parameters for desethylamodiaquine (DEAQ) by weight band over 0-672 h (28 days). Cmax in ng/mL; Tmax in h; AUCinf.obs in ng*h/mL; half-life in h.
Weight band NCA parameter Value (DEAQ)
15 kg toddler cmax 307
15 kg toddler tmax 50
15 kg toddler half.life 204
15 kg toddler aucinf.obs 26600
21 kg child (cohort median) cmax 450
21 kg child (cohort median) tmax 50
21 kg child (cohort median) half.life 220
21 kg child (cohort median) aucinf.obs 40700
30 kg child cmax 321
30 kg child tmax 50
30 kg child half.life 240
30 kg child aucinf.obs 31100
50 kg adult cmax 393
50 kg adult tmax 50
50 kg adult half.life 271
50 kg adult aucinf.obs 42300
70 kg adult cmax 283
70 kg adult tmax 50
70 kg adult half.life 295
70 kg adult aucinf.obs 32800
8 kg infant cmax 261
8 kg infant tmax 50
8 kg infant half.life 226
8 kg infant aucinf.obs 31400 
sim_nca_aq <- sim_typical |>
  dplyr::filter(!is.na(Cc), time >= 0) |>
  dplyr::transmute(id, time, Cc, label) |>
  dplyr::distinct(id, time, label, .keep_all = TRUE)

conc_obj_aq <- PKNCA::PKNCAconc(sim_nca_aq, Cc ~ time | label + id)
#> Warning in assert_conc(conc, any_missing_conc = any_missing_conc): Negative
#> concentrations found

intervals_aq <- data.frame(
  start = 0, end = 72,
  cmax        = TRUE,
  tmax        = TRUE,
  auclast     = TRUE
)

nca_data_aq <- PKNCA::PKNCAdata(conc_obj_aq, dose_obj, intervals = intervals_aq)
nca_res_aq  <- PKNCA::pk.nca(nca_data_aq)

nca_summary_aq <- as.data.frame(nca_res_aq$result) |>
  dplyr::filter(PPTESTCD %in% c("cmax", "tmax", "auclast")) |>
  dplyr::transmute(
    `Weight band` = label,
    `NCA parameter` = PPTESTCD,
    `Value (AQ)`    = signif(PPORRES, 3)
  )

knitr::kable(nca_summary_aq,
             caption = "Typical-value NCA parameters for amodiaquine (AQ) by weight band over 0-72 h (parent compound is rapidly cleared, so AUClast over the 3-day dosing window is reported in place of AUCinf).")
Typical-value NCA parameters for amodiaquine (AQ) by weight band over 0-72 h (parent compound is rapidly cleared, so AUClast over the 3-day dosing window is reported in place of AUCinf).
Weight band NCA parameter Value (AQ)
15 kg toddler auclast 280.0
15 kg toddler cmax 23.5
15 kg toddler tmax 49.0
21 kg child (cohort median) auclast 431.0
21 kg child (cohort median) cmax 34.8
21 kg child (cohort median) tmax 49.0
30 kg child auclast 328.0
30 kg child cmax 25.2
30 kg child tmax 49.0
50 kg adult auclast 442.0
50 kg adult cmax 31.7
50 kg adult tmax 49.0
70 kg adult auclast 340.0
70 kg adult cmax 23.3
70 kg adult tmax 49.0
8 kg infant auclast 297.0
8 kg infant cmax 22.3
8 kg infant tmax 49.0

Comparison against published values

Ali 2018 reports two anchor values that can be compared against simulated DEAQ exposures:

  1. The reference DEAQ Day-7 concentration in a typical 50 kg adult on 540 mg/day for 3 days, defined implicitly as 80% of which equals 54.0 ng/mL (per Ali 2018 “Simulations” paragraph); the published median Day-7 DEAQ in the 50 kg reference is therefore 54.0 / 0.80 = 67.5 ng/mL.
  2. The Cmax upper threshold for safety monitoring: 575 ng/mL.
ref_row <- sim_typical |>
  dplyr::filter(label == "50 kg adult", time == 168) |>
  dplyr::distinct(time, .keep_all = TRUE) |>
  dplyr::pull(Cc_deaq) |>
  utils::head(1)

cmax_50kg <- sim_typical |>
  dplyr::filter(label == "50 kg adult") |>
  dplyr::pull(Cc_deaq) |>
  max(na.rm = TRUE)

comparison <- tibble::tibble(
  Quantity            = c("Day-7 DEAQ in 50 kg adult on 540 mg/day x 3 (ng/mL)",
                          "Peak DEAQ in 50 kg adult on 540 mg/day x 3 (ng/mL)"),
  `Published median`  = c("67.5", "< 575 (upper threshold)"),
  `Simulated typical` = c(round(ref_row, 1), round(cmax_50kg, 1)),
  `Ratio sim / pub`   = c(round(ref_row / 67.5, 3), round(cmax_50kg / 575, 3))
)
knitr::kable(comparison,
             caption = "Comparison of simulated typical-value exposures against Ali 2018 published anchor values (50 kg adult arm, 540 mg/day x 3 days).")
Comparison of simulated typical-value exposures against Ali 2018 published anchor values (50 kg adult arm, 540 mg/day x 3 days).
Quantity Published median Simulated typical Ratio sim / pub
Day-7 DEAQ in 50 kg adult on 540 mg/day x 3 (ng/mL) 67.5 66.2 0.981
Peak DEAQ in 50 kg adult on 540 mg/day x 3 (ng/mL) < 575 (upper threshold) 393.2 0.684

The simulated typical-value Day-7 DEAQ matches the published reference within about 2%, confirming that the structural and covariate equations in the model file reproduce the source publication.

Assumptions and deviations

  • Postmenstrual age (PAGE). Ali 2018 did not record individual gestational ages and substituted a fixed 9-month gestation for every subject (Methods, “Effect of body size and age”). The packaged model follows this convention; subjects with a known GA can override the PAGE covariate using the canonical PAGE = GA_weeks / 4.35 + PNA_months derivation.
  • NN = 2 transit compartments. The paper estimated NN = 2.00 (95% CI 1.09-6.31). The implementation treats NN as fixed at 2 because the number of compartments in an rxode2 model must be a positive integer; the high RSE on NN (78%) makes the parameter weakly identified anyway.
  • Allometric exponents. Ali 2018 fixed the CL allometric exponent at 0.75 and the V exponent at 1.0 (canonical biological-prior values). The model file encodes both as fixed() to preserve provenance.
  • First-dose F effect (CYCLE). The packaged model uses the canonical CYCLE covariate (dose-number counter, 1 = first day, 2-3 = subsequent days) to encode the -22.4% Day-1 bioavailability reduction. The user must populate CYCLE correctly per dose row in the events table; the helper recipes in the cohort chunk above show one approach.
  • Between-occasion variability (BOV) not modelled. Ali 2018 reports moderate-to-large BOV on Ka (78.5% CV), MTT (93.4% CV), and F (30.9% CV). The packaged model file does not encode BOV; it captures only the four BSV terms (CL_AQ, Vc_AQ, CL_DEAQ, Vc_DEAQ) reported in Table 3. For typical-value simulations BOV has no effect (it averages to zero); for stochastic VPC-style use, the within-subject spread is therefore underestimated.
  • Residual additive error and study-specific LLOQ. Ali 2018 added 20% of the per-study LLOQ to the AQ additive error and held DEAQ additive error at LLOQ/5 (the extra estimated component was not significantly different from zero for DEAQ). The per-study LLOQ ranged from 1 to 10 ng/mL across the five contributing studies. The packaged model uses LLOQ = 10 ng/mL (the most well-represented study, Adjei 2008, n = 101 of 261) as the representative LLOQ; users targeting a different study’s assay can override addSd and addSd_deaq accordingly.
  • Molar conversion. The mass flux of AQ leaving the AQ central compartment is converted to DEAQ mass flux using MW_DEAQ / MW_AQ = 327.81 / 355.85 = 0.9212 (Ali 2018 Methods, “Structural model”). This assumes 100% in-vivo metabolic conversion of AQ to DEAQ, consistent with the paper’s assumption.
  • Supplemental material not on disk. The Ali 2018 supplement (Fig S1, S2; basic GOF plots and age-stratified VPCs) was not available during extraction. All parameter values are sourced from the main paper, which contains the complete final-parameter table; the supplement holds only graphical diagnostics.