Skip to contents

Flurbiprofen (Kumpulainen 2010)

Source: vignettes/articles/Kumpulainen_2010_flurbiprofen.Rmd
Kumpulainen_2010_flurbiprofen.Rmd

Model and source

  • Citation: Kumpulainen E, Valitalo P, Kokki M, Lehtonen M, Hooker A, Ranta V-P, Kokki H. Plasma and cerebrospinal fluid pharmacokinetics of flurbiprofen in children. Br J Clin Pharmacol. 2010;70(4):557-566. doi:10.1111/j.1365-2125.2010.03720.x.

  • Description: Three-compartment population PK model with a separate cerebrospinal-fluid (CSF) compartment for flurbiprofen in 64 healthy children aged 3 months to 13 years (Kumpulainen 2010). Two parallel absorption routes: oral syrup via an absorption compartment with lag time (K12) and a single first-order ka, and IV flurbiprofen axetil prodrug via a separate dosing compartment that converts to flurbiprofen with a first-order rate constant (K42). Plasma kinetics scaled allometrically by weight (exponents fixed at 0.75 for CL and 1 for all volumes, including the CSF volume held fixed at 0.15 L/70 kg per literature). The paper’s QCSF + UPTK parameterisation is encoded as canonical influx / efflux clearances clin = QCSF * UPTK and clef = QCSF, with fraction unbound (fu) gating only the central-to-CSF flux.

  • Article: https://doi.org/10.1111/j.1365-2125.2010.03720.x

Population

64 healthy children aged 3 months to 13 years (median 5.2 years; weight median 20 kg, range 7-76 kg) scheduled for elective lower-body surgery with spinal anaesthesia at Kuopio University Hospital, Finland (Kumpulainen 2010 Table 1). 37 received a single 1 mg/kg oral dose of flurbiprofen syrup (Froben, 5 mg/mL); 27 received a 10-min IV injection of 0.9 mg/kg flurbiprofen axetil (Ropion, 10 mg/mL) corresponding to approximately 0.65 mg/kg of flurbiprofen after in-vivo conversion by plasma esterases. 304 total plasma + 62 protein-free plasma + 60 total CSF flurbiprofen concentrations were modelled in NONMEM VI with first-order conditional estimation with interaction (FOCE-I).

Programmatic access to the structured population metadata is via readModelDb("Kumpulainen_2010_flurbiprofen")()$population.

Source trace

Every ini() parameter carries an in-file source-trace comment next to its value in inst/modeldb/specificDrugs/Kumpulainen_2010_flurbiprofen.R. The table below collects them in one place for review. All references are to the final-model column of Kumpulainen 2010 Table 2 unless noted.

Equation / parameter Value Source location
lka (K12, oral ka) 5.5 1/h Table 2: K12 = 5.5 (RSE 0.24)
lka2 (K42, axetil conversion) 29 1/h Table 2: K42 = 29 (RSE 0.32), half-life ~1.4 min
ltlag (oral lag time) 0.11 h Table 2: lag time = 0.11 (RSE 0.16) = 6 min
lcl (CL) 0.96 L/h / 70 kg Table 2: CL = 0.96 (RSE 0.057)
lvc (V_central) 3.6 L / 70 kg Table 2: V (central) = 3.6 (RSE 0.11)
lq (Q shallow peripheral) 1.5 L/h Table 2: Q (shallow peripheral, “Q2”) = 1.5 (RSE 0.39)
lvp (V_shallow) 1.8 L / 70 kg Table 2: V (shallow peripheral) = 1.8 (RSE 0.20)
lq2 (Q deep peripheral) 0.18 L/h Table 2: Q (deep peripheral) = 0.18 (RSE 0.30)
lvp2 (V_deep) 2.7 L / 70 kg Table 2: V (deep peripheral) = 2.7 (RSE 0.18)
lfdepot (oral F) 0.81 Table 2: bioavailability = 0.81 (RSE 0.055; bootstrap 0.71-0.91)
lclin (= QCSF * UPTK) 0.12 * 6.8 = 0.816 L/h Table 2: QCSF = 0.12; UPTK = 6.8; see reparameterisation note below
lclef (= QCSF) 0.12 L/h Table 2: QCSF = 0.12 (RSE 0.27)
lvcsf (V_CSF, fixed) 0.15 L / 70 kg Results, popPK model: “The volume of the CSF compartment (0.15 l) was retrieved from the literature [13] and also scaled by weight”
lfu (fraction unbound) 0.00031 Table 2: protein-free fraction = 0.00031 (RSE 0.043); = 0.031% per Results
e_wt_cl (allometric CL) 0.75 (fixed) Table 2 footnote; Discussion: “the exponents were fixed to 1 and 0.75 in this study”
e_wt_vc_vp (allometric volumes) 1.0 (fixed) Table 2 footnote; applied to V_central, V_shallow, V_deep, V_CSF
etalcl (IIV on CL) omega = 0.28 (SD) Table 2: omega_CL = 0.28 (RSE 0.20; bootstrap 0.22-0.34)
etalvc (IIV on V_central) omega = 0.28 (SD) Table 2: omega_Vd = 0.28 (RSE 0.25; bootstrap 0.19-0.35)
etalka (IIV on K12) omega = 0.81 (SD) Table 2: omega_K12 = 0.81 (RSE 0.40; bootstrap 0.42-1.4)
etalfu (IIV on fu) omega = 0.30 (SD) Table 2: omega_fu = 0.30 (RSE 0.28; bootstrap 0.19-0.36)
propSd (plasma residual SD) 0.13 Table 2: sigma plasma = 0.13 (RSE 0.065)
propSd_Cu (unbound residual) 0.13 Same as plasma per Methods: “Unbound observations were included in the same compartment as total plasma observations”
propSd_Ccsf (CSF residual SD) 0.50 Table 2: sigma CSF = 0.50 (RSE 0.091)
d/dt(central) (Figure 1 ODE) n/a Figure 1 schematic + Methods modelling strategy
d/dt(csf) (CSF kinetics) n/a Figure 1 + Methods: K_central->CSF = (QCSF * UPTK * fu) / V_central; K_CSF->central = QCSF / V_CSF

Virtual cohort

Original observed data are not publicly available. The figures below use a virtual paediatric cohort whose body-weight distribution approximates the paper’s reported median 20 kg with range 7-76 kg (Table 1). Half of the virtual subjects receive 1 mg/kg oral syrup; the other half receive 0.65 mg/kg of flurbiprofen-equivalent IV axetil (the depot2 compartment receives this mass directly; the in-vivo axetil-to-flurbiprofen conversion is captured by the model’s ka2 = K42 first-order kinetic).

set.seed(20100506L)  # Kumpulainen 2010 accepted date 6 May 2010 = 2010-05-06

n_per_arm <- 50L
ages_yr <- c(0.25, runif(n_per_arm - 2L, min = 0.5, max = 13), 13)
# Simple WHO-style weight-for-age approximation: kg ~ 8 + age*2 in the paediatric range,
# with mild log-normal noise to span the paper's observed 7-76 kg envelope.
wt_kg_typical <- 8 + 2 * ages_yr
weights_kg <- pmax(7, pmin(76,
  exp(log(wt_kg_typical) + rnorm(n_per_arm, sd = 0.20))
))

make_cohort <- function(arm, dose_mg_per_kg, cmt, weights_kg, id_offset = 0L) {
  ids <- id_offset + seq_along(weights_kg)
  doses <- tibble(
    id    = ids,
    time  = 0,
    amt   = dose_mg_per_kg * weights_kg,
    evid  = 1L,
    cmt   = cmt,
    WT    = weights_kg,
    arm   = arm
  )
  obs_times <- c(seq(0, 1, by = 0.05), seq(1.1, 6, by = 0.1), seq(6.5, 20, by = 0.5))
  obs <- tidyr::expand_grid(
    id   = ids,
    time = obs_times
  ) |>
    dplyr::left_join(doses |> dplyr::select(id, WT, arm), by = "id") |>
    dplyr::mutate(amt = NA_real_, evid = 0L, cmt = "Cc")
  dplyr::bind_rows(doses, obs) |>
    dplyr::arrange(id, time, dplyr::desc(evid))
}

events <- dplyr::bind_rows(
  make_cohort("oral 1 mg/kg",   1.00, "depot",  weights_kg, id_offset =   0L),
  make_cohort("IV 0.65 mg/kg",  0.65, "depot2", weights_kg, id_offset = 100L)
)
stopifnot(!anyDuplicated(unique(events[, c("id", "time", "evid")])))

Simulation

The figures below combine two simulations:

  1. A typical-value simulation for a 20 kg child to compare directly against the source paper’s “predicted Cmax / tmax for a typical 20 kg child” claims in Results, and
  2. A stochastic population simulation with the full random-effect structure for the VPC-style plots and the NCA validation.
typical_events <- dplyr::bind_rows(
  make_cohort("oral 1 mg/kg",   1.00, "depot",  rep(20, 1L), id_offset = 0L),
  make_cohort("IV 0.65 mg/kg",  0.65, "depot2", rep(20, 1L), id_offset = 100L)
)
mod_typical <- m |> rxode2::zeroRe()
#> ℹ parameter labels from comments will be replaced by 'label()'
sim_typical <- rxode2::rxSolve(mod_typical, events = typical_events,
                               keep = c("WT", "arm")) |>
  as.data.frame()
#> ℹ omega/sigma items treated as zero: 'etalcl', 'etalvc', 'etalka', 'etalfu'
#> Warning: multi-subject simulation without without 'omega'
sim <- rxode2::rxSolve(m, events = events, keep = c("WT", "arm")) |>
  as.data.frame()
#> ℹ parameter labels from comments will be replaced by 'label()'

Replicate published figures

Figure 3: total plasma, unbound plasma, and CSF concentration-time profiles

Kumpulainen 2010 Figure 3 shows raw observed concentration-time data in four panels: (A) IV total plasma, (B) oral total plasma, (C) both arms protein-free (unbound) plasma, (D) both arms total CSF. The panels below show the stochastic-simulation median + 90% interval per arm for each of the three modelled outputs (total plasma Cc, unbound plasma Cu, CSF Ccsf).

plasma_summary <- sim |>
  group_by(arm, time) |>
  summarise(
    Q05 = quantile(Cc, 0.05, na.rm = TRUE),
    Q50 = quantile(Cc, 0.50, na.rm = TRUE),
    Q95 = quantile(Cc, 0.95, na.rm = TRUE),
    .groups = "drop"
  ) |>
  filter(time <= 6)

ggplot(plasma_summary, aes(time, Q50)) +
  geom_ribbon(aes(ymin = Q05, ymax = Q95), alpha = 0.25) +
  geom_line() +
  facet_wrap(~ arm) +
  labs(x = "Time (h)", y = "Total plasma flurbiprofen Cc (mg/L)",
       title = "Figure 3A-B equivalent: total plasma simulated profiles",
       caption = "Replicates Figure 3A (IV arm) and 3B (oral arm) of Kumpulainen 2010. Ribbon = 5%-95% percentile band over the simulated cohort.") +
  theme_bw()

unbound_summary <- sim |>
  group_by(arm, time) |>
  summarise(
    Q05 = quantile(Cu * 1000, 0.05, na.rm = TRUE),
    Q50 = quantile(Cu * 1000, 0.50, na.rm = TRUE),
    Q95 = quantile(Cu * 1000, 0.95, na.rm = TRUE),
    .groups = "drop"
  ) |>
  filter(time <= 6)

ggplot(unbound_summary, aes(time, Q50)) +
  geom_ribbon(aes(ymin = Q05, ymax = Q95), alpha = 0.25) +
  geom_line() +
  facet_wrap(~ arm) +
  labs(x = "Time (h)", y = "Unbound plasma flurbiprofen Cu (ug/L)",
       title = "Figure 3C equivalent: unbound plasma simulated profiles",
       caption = "Replicates Figure 3C of Kumpulainen 2010. Cu = fu * Cc with fu = 0.00031.") +
  theme_bw()

csf_summary <- sim |>
  group_by(arm, time) |>
  summarise(
    Q05 = quantile(Ccsf * 1000, 0.05, na.rm = TRUE),
    Q50 = quantile(Ccsf * 1000, 0.50, na.rm = TRUE),
    Q95 = quantile(Ccsf * 1000, 0.95, na.rm = TRUE),
    .groups = "drop"
  ) |>
  filter(time <= 6)

ggplot(csf_summary, aes(time, Q50)) +
  geom_ribbon(aes(ymin = Q05, ymax = Q95), alpha = 0.25) +
  geom_line() +
  facet_wrap(~ arm) +
  labs(x = "Time (h)", y = "Total CSF flurbiprofen Ccsf (ug/L)",
       title = "Figure 3D equivalent: total CSF simulated profiles",
       caption = "Replicates Figure 3D of Kumpulainen 2010. Median CSF Cmax 1-2 h post-dose.") +
  theme_bw()

PKNCA validation

The source paper reports model-predicted Cmax and tmax for the three outputs in a typical 20 kg child receiving the standard 1 mg/kg oral dose (Cc and Ccsf) or 0.65 mg/kg IV (Ccsf only). PKNCA reproduces those summaries from the typical-value simulation. The remaining NCA parameters (AUC, t1/2) are reported here for completeness; the source paper does not publish reference values for them, so no Reference column is shown.

nca_for_output <- function(sim_df, conc_col, conc_unit, time_unit = "h") {
  conc <- sim_df |>
    dplyr::filter(!is.na(.data[[conc_col]])) |>
    dplyr::transmute(id = id, time = time, Cc = .data[[conc_col]], arm = arm)
  # Guarantee a t=0 row per (id, arm) so PKNCA can anchor AUC0-* (extravascular
  # dosing -> Cc(0) = 0). Without this, the "Requesting an AUC range starting
  # (0) before the first measurement" warning fires for every subject.
  conc <- dplyr::bind_rows(
    conc,
    conc |> dplyr::distinct(id, arm) |> dplyr::mutate(time = 0, Cc = 0)
  ) |>
    dplyr::distinct(id, arm, time, .keep_all = TRUE) |>
    dplyr::arrange(id, arm, time)
  conc_obj <- PKNCA::PKNCAconc(conc, Cc ~ time | arm + id,
                               concu = conc_unit, timeu = time_unit)

  doses <- sim_df |>
    dplyr::group_by(id, arm) |>
    dplyr::slice_head(n = 1) |>
    dplyr::ungroup() |>
    dplyr::transmute(id = id, time = 0, amt = WT * dplyr::if_else(arm == "oral 1 mg/kg", 1.00, 0.65),
                     arm = arm)
  dose_obj <- PKNCA::PKNCAdose(doses, amt ~ time | arm + id, doseu = "mg")

  intervals <- data.frame(
    start       = 0,
    end         = Inf,
    cmax        = TRUE,
    tmax        = TRUE,
    aucinf.obs  = TRUE,
    half.life   = TRUE
  )
  PKNCA::pk.nca(PKNCA::PKNCAdata(conc_obj, dose_obj, intervals = intervals))
}
# Typical-value (1 child per arm) NCA: directly comparable to paper's Cmax/tmax
nca_typ_Cc   <- nca_for_output(sim_typical, "Cc",   "mg/L")
nca_typ_Ccsf <- nca_for_output(sim_typical, "Ccsf", "mg/L")
# Population NCA: medians across the virtual cohort, for completeness
nca_pop_Cc   <- nca_for_output(sim, "Cc",   "mg/L")
nca_pop_Ccsf <- nca_for_output(sim, "Ccsf", "mg/L")

Comparison against published predicted Cmax / tmax (typical 20 kg child)

Kumpulainen 2010 Results state, for a 20 kg child:

  • Oral 1 mg/kg: “the predicted tmax was 27 min and Cmax 10 mg/L” (Cc).
  • Oral 1 mg/kg: “model-predicted CSF Cmax and tmax after oral dosing (1 mg/kg) were 17 ug/L and 60 min, respectively”.
  • IV 0.65 mg/kg: “model-predicted Cmax and tmax for CSF concentrations after i.v. injection (0.65 mg/kg) in a typical 20 kg child were 14 ug/L and 40 min, respectively”.
# Pull the typical-value Cmax/tmax for each (arm, output) into a long table
typ_results <- function(nca_res, output_label, scale = 1) {
  res <- as.data.frame(nca_res$result) |>
    dplyr::filter(PPTESTCD %in% c("cmax", "tmax")) |>
    dplyr::transmute(arm = arm, PPTESTCD = PPTESTCD,
                     PPORRES = PPORRES * dplyr::if_else(PPTESTCD == "cmax", scale, 1)) |>
    dplyr::mutate(output = output_label)
  res
}

simulated_long <- dplyr::bind_rows(
  typ_results(nca_typ_Cc,   "Cc (mg/L)"),
  typ_results(nca_typ_Ccsf, "Ccsf (ug/L)", scale = 1000)
)

# Paper-reported published values (Cmax in matching units, tmax in h)
published_typ <- tibble::tribble(
  ~arm,            ~output,        ~cmax, ~tmax,
  "oral 1 mg/kg",  "Cc (mg/L)",      10,   27/60,
  "oral 1 mg/kg",  "Ccsf (ug/L)",    17,   60/60,
  "IV 0.65 mg/kg", "Ccsf (ug/L)",    14,   40/60
)

# Build a comparison table per output
sim_table_long <- simulated_long |>
  dplyr::transmute(arm = arm, output = output, PPTESTCD = PPTESTCD, PPORRES = PPORRES)

comparison_tables <- lapply(unique(published_typ$output), function(out) {
  pub <- published_typ |> dplyr::filter(output == out) |>
    dplyr::select(arm, cmax, tmax)
  sim <- sim_table_long |> dplyr::filter(output == out) |>
    dplyr::select(arm, PPTESTCD, PPORRES)
  tbl <- nlmixr2lib::ncaComparisonTable(
    simulated = sim,
    reference = pub,
    by        = "arm",
    units     = c(cmax = if (out == "Cc (mg/L)") "mg/L" else "ug/L",
                  tmax = "h"),
    tolerance_pct = 20
  )
  cbind(`Output` = out, tbl)
})

comparison_tbl <- do.call(rbind, comparison_tables)

knitr::kable(
  comparison_tbl,
  caption = "Simulated (typical 20 kg child, zero random effects) vs Kumpulainen 2010 Results-section predicted Cmax / tmax. * differs from reference by more than +/- 20%.",
  align = c("l", "l", "l", "r", "r", "r")
)
Simulated (typical 20 kg child, zero random effects) vs Kumpulainen 2010 Results-section predicted Cmax / tmax. * differs from reference by more than +/- 20%.
Output NCA parameter arm Reference Simulated % diff
Cc (mg/L) Cmax (mg/L) oral 1 mg/kg 10 9.62 -3.8%
Cc (mg/L) Tmax (h) oral 1 mg/kg 0.45 0.5 +11.1%
Ccsf (ug/L) Cmax (ug/L) oral 1 mg/kg 17 16.4 -3.3%
Ccsf (ug/L) Cmax (ug/L) IV 0.65 mg/kg 14 13.7 -2.5%
Ccsf (ug/L) Tmax (h) oral 1 mg/kg 1 1 +0.0%
Ccsf (ug/L) Tmax (h) IV 0.65 mg/kg 0.667 0.7 +5.0%

Population-cohort NCA summary (no published reference values)

The remaining NCA parameters (AUC, t1/2) over the simulated cohort are reported here for context; the source paper does not publish reference values for them, so no Reference column is shown.

pop_summary <- function(nca_res, output_label, scale = 1) {
  as.data.frame(nca_res$result) |>
    dplyr::filter(PPTESTCD %in% c("aucinf.obs", "half.life")) |>
    dplyr::group_by(arm, PPTESTCD) |>
    dplyr::summarise(
      median = stats::median(PPORRES, na.rm = TRUE),
      q05    = stats::quantile(PPORRES, 0.05, na.rm = TRUE),
      q95    = stats::quantile(PPORRES, 0.95, na.rm = TRUE),
      .groups = "drop"
    ) |>
    dplyr::mutate(parameter = dplyr::recode(PPTESTCD,
                                            aucinf.obs = "AUC0-Inf (obs)",
                                            half.life  = "t1/2"),
                  output = output_label) |>
    dplyr::select(output, arm, parameter, median, q05, q95)
}

pop_table <- dplyr::bind_rows(
  pop_summary(nca_pop_Cc,   "Cc (mg/L)"),
  pop_summary(nca_pop_Ccsf, "Ccsf (ug/L)")
) |>
  dplyr::mutate(across(c(median, q05, q95), ~ signif(.x, 3)))

pop_table |>
  dplyr::rename(
    "Output"         = output,
    "Arm"            = arm,
    "NCA parameter"  = parameter,
    "Median"         = median,
    "5% percentile"  = q05,
    "95% percentile" = q95
  ) |>
  knitr::kable(
  caption = "Population-cohort median and 5%-95% range for AUC0-Inf and t1/2 (per arm and per output). No published reference values.",
  align = c("l", "l", "l", "r", "r", "r")
)
Population-cohort median and 5%-95% range for AUC0-Inf and t1/2 (per arm and per output). No published reference values.
Output Arm NCA parameter Median 5% percentile 95% percentile
Cc (mg/L) IV 0.65 mg/kg AUC0-Inf (obs) 33.9000 21.3000 52.200
Cc (mg/L) IV 0.65 mg/kg t1/2 5.4600 3.9000 8.120
Cc (mg/L) oral 1 mg/kg AUC0-Inf (obs) 42.1000 26.0000 76.000
Cc (mg/L) oral 1 mg/kg t1/2 5.6700 3.7100 8.980
Ccsf (ug/L) IV 0.65 mg/kg AUC0-Inf (obs) 0.0651 0.0358 0.121
Ccsf (ug/L) IV 0.65 mg/kg t1/2 5.4500 3.9000 8.050
Ccsf (ug/L) oral 1 mg/kg AUC0-Inf (obs) 0.0958 0.0362 0.163
Ccsf (ug/L) oral 1 mg/kg t1/2 5.6600 3.7200 8.970

Assumptions and deviations

  • QCSF + UPTK reparameterised as clin + clef. Kumpulainen 2010 Figure 1 / Methods write the central-to-CSF transfer rate as K_central_to_CSF = (QCSF * UPTK * fu) / V_central, with the CSF-to-central return rate K_CSF_to_central = QCSF / V_CSF. UPTK is dimensionless (estimated as 6.8 with bootstrap 95% CI 6.0-7.7). The packaged model preserves all numerical content but encodes this as the canonical Campagne 2019 clin / clef pair: clin = QCSF * UPTK (CSF influx clearance applied to the unbound plasma concentration fu * Cc) and clef = QCSF (CSF efflux clearance applied to Ccsf). The uptake ratio UPTK = clin / clef = 6.8 is recoverable post-hoc; in-file source-trace comments in inst/modeldb/specificDrugs/Kumpulainen_2010_flurbiprofen.R document both the paper’s UPTK = 6.8 and QCSF = 0.12 L/h values.
  • CSF volume held fixed at 0.15 L / 70 kg. Kumpulainen 2010 Results, popPK section: “The volume of the CSF compartment (0.15 l) was retrieved from the literature [13] and also scaled by weight.” lvcsf is wrapped in fixed() to preserve this provenance.
  • Allometric exponents fixed at 0.75 (CL) and 1 (volumes). The paper estimated both exponents (0.774 and 1.01) and noted close agreement with the canonical values “the exponents were fixed to 1 and 0.75 in this study”. The shared volume exponent applies to V_central, V_shallow, V_deep, and V_CSF; the intercompartmental clearances Q2 (shallow), Q (deep), and QCSF carry no allometric scaling in Table 2 (the (WT/70) annotation appears only on CL and the four V rows).
  • BSV omega values are SDs of eta on the log scale. The paper’s Table 2 reports omega values consistent with the Results-section claim of “roughly 30% CV” inter-individual variability; the omega for K12 (0.81) likewise aligns with “the inter-individual variability for the oral absorption rate constant was estimated at 80%”. The internal nlmixr2 variance is therefore (paper omega)^2 (0.28^2 = 0.0784 for CL, V, and Vd; 0.81^2 = 0.6561 for K12; 0.30^2 = 0.09 for fu).
  • IIV on volume applies to V_central only. The paper states “BSV was assigned to volume of distribution, clearance, oral absorption rate and fraction unbound” – a single omega is reported for “Vd” in Table 2. This is interpreted as IIV on the central volume V_central; the peripheral and CSF volumes carry no IIV.
  • Plasma sigma shared between total and unbound observations. The paper reports two residual variances (plasma and CSF), and writes Methods: “Unbound observations were included in the same compartment as total plasma observations. They were defined as special cases of total plasma observations”. The packaged model has three outputs (Cc, Cu, Ccsf) because nlmixr2 requires per-output residual-error parameters; propSd (Cc) and propSd_Cu (Cu) carry the same source value of 0.13 to preserve this constraint.
  • IV axetil dosing convention. The paper reports the IV dose as 0.9 mg/kg of flurbiprofen axetil prodrug, which corresponds to approximately 0.65 mg/kg of flurbiprofen on a mass-equivalent basis. The packaged model assumes the user supplies the flurbiprofen-equivalent mass directly to the depot2 compartment; the K42 = 29 1/h conversion-rate constant then describes the in-vivo axetil-to-flurbiprofen hydrolysis. The 10-min IV infusion duration is short relative to K42’s half-life (1.4 min) and is captured by modelling the IV dose as an instantaneous bolus into depot2.
  • Virtual cohort weight distribution. Original Table 1 weight values are not publicly available. The vignette uses a simple weight-for-age envelope (kg ~ 8 + 2 * age_years, with mild log-normal noise) to span the paper’s median 20 kg and range 7-76 kg envelope; this is good enough for typical-value Cmax / tmax replication but is not a high-fidelity match of the trial’s actual paediatric weight distribution.
  • No NCA reference values for AUC or t1/2. The paper does not report population NCA summaries for AUC or terminal half-life, so the population-cohort summary table omits a Reference column for those parameters; only the typical-value Cmax / tmax claims in the Results section are validated against the simulation.
  • Race / ethnicity not modelled. Single-centre Finnish paediatric cohort; race / ethnicity was not reported in the source paper, so the population metadata records “Not reported” and the model carries no race covariate effects.