Skip to contents

Model and source

Desflurane is a volatile inhalational anaesthetic with a low blood-gas partition coefficient. Lu 2024 gave seven mechanically ventilated piglets a fixed 6% inspiratory desflurane concentration for 2 h (uptake) and then 40% oxygen alone for 2 h (disposition), and followed the end-tidal gas, the femoral-arterial blood and the pulmonary-artery (mixed venous) blood concentrations. Blood concentrations were measured by headspace gas chromatography and, like the gas concentrations, are expressed as percent vapour (the partial pressure as a volume percent).

Each concentration-time curve was fitted per animal in WinNonlin with a zero-order (constant-infusion) input over the 120-min uptake phase and biexponential disposition, i.e. a two-compartment model, and Table 3 reports the mean (SD) of the individual estimates. The three sites were fitted independently, so the paper contributes three model files:

des_models <- c(
  "Lu_2024_desflurane_pig_endtidal",    # end-tidal gas
  "Lu_2024_desflurane_pig_arterial",    # femoral-arterial blood
  "Lu_2024_desflurane_pig_mixedvenous"  # pulmonary-artery (mixed venous) blood
)
site_label <- c(
  Lu_2024_desflurane_pig_endtidal = "End-tidal",
  Lu_2024_desflurane_pig_arterial = "Arterial blood",
  Lu_2024_desflurane_pig_mixedvenous = "Pulmonary artery blood"
)
mods <- lapply(stats::setNames(des_models, des_models), function(m) rxode2::rxode(readModelDb(m)))
for (m in des_models) {
  cat(m, "\n", sep = "")
  cat(strwrap(mods[[m]]$description, 76), sep = "\n")
  cat("\n\n")
}
#> Lu_2024_desflurane_pig_endtidal
#> Preclinical (pig). Two-compartment model with zero-order input for the
#> end-tidal desflurane concentration (percent vapour) in mechanically
#> ventilated piglets breathing 6 percent inspiratory desflurane for 2 h,
#> followed by a 2-h washout (Lu 2024). The compartments are those of a
#> descriptive curve fit observed in exhaled gas, not anatomical spaces. Lu
#> 2024 fitted each piglet individually in WinNonlin with constant-rate
#> (zero-order) input over the 120-min uptake phase and reported the mean of
#> the individual biexponential estimates (Table 3); the volumes and
#> clearances carried here are reconstructed from those macro-constants with
#> the clearance anchored to the printed mean, so no between-subject
#> variability or residual-error model is available and the residual SD is
#> fixed at zero. The end-tidal means are dominated by between-animal spread
#> (beta half-life 61.90 +/- 105.93 min), so this typical-value curve
#> over-predicts the observed mean end-tidal plateau by about half; prefer the
#> blood models for simulation. The input is the nominal amount of the
#> WinNonlin fit, in percent x litre: a 120-min zero-order infusion of 76.953
#> percent x litre into central corresponds to the 6 percent inspiratory
#> exposure (the amount scales only the clearance and volumes, not the
#> predicted concentrations). Lu 2024 fitted the arterial and pulmonary-artery
#> curves as separate models; see modellib('Lu_2024_desflurane_pig_arterial')
#> and modellib('Lu_2024_desflurane_pig_mixedvenous').
#> 
#> 
#> Lu_2024_desflurane_pig_arterial
#> Preclinical (pig). Two-compartment model with zero-order input for the
#> femoral-arterial blood desflurane concentration (percent vapour, i.e.
#> partial pressure as volume percent) in mechanically ventilated piglets
#> breathing 6 percent inspiratory desflurane for 2 h, followed by a 2-h
#> washout (Lu 2024). Lu 2024 fitted each piglet individually in WinNonlin
#> with constant-rate (zero-order) input over the 120-min uptake phase and
#> reported the mean of the individual biexponential estimates (Table 3); the
#> volumes and clearances carried here are reconstructed from those
#> macro-constants with the clearance anchored to the printed mean, so no
#> between-subject variability or residual-error model is available and the
#> residual SD is fixed at zero. The input is the nominal amount of the
#> WinNonlin fit, in percent x litre: a 120-min zero-order infusion of 53.443
#> percent x litre into central reproduces the 6 percent inspiratory exposure
#> (the amount scales only the clearance and volumes, not the predicted
#> concentrations). Lu 2024 fitted the end-tidal and pulmonary-artery curves
#> as separate models; see modellib('Lu_2024_desflurane_pig_endtidal') and
#> modellib('Lu_2024_desflurane_pig_mixedvenous').
#> 
#> 
#> Lu_2024_desflurane_pig_mixedvenous
#> Preclinical (pig). Two-compartment model with zero-order input for the
#> pulmonary-artery (mixed venous) blood desflurane concentration (percent
#> vapour, i.e. partial pressure as volume percent) in mechanically ventilated
#> piglets breathing 6 percent inspiratory desflurane for 2 h, followed by a
#> 2-h washout (Lu 2024). Lu 2024 fitted each piglet individually in WinNonlin
#> with constant-rate (zero-order) input over the 120-min uptake phase and
#> reported the mean of the individual biexponential estimates (Table 3); the
#> volumes and clearances carried here are reconstructed from those
#> macro-constants with the clearance anchored to the printed mean, so no
#> between-subject variability or residual-error model is available and the
#> residual SD is fixed at zero. The input is the nominal amount of the
#> WinNonlin fit, in percent x litre: a 120-min zero-order infusion of 52.291
#> percent x litre into central reproduces the 6 percent inspiratory exposure
#> (the amount scales only the clearance and volumes, not the predicted
#> concentrations). Lu 2024 fitted the end-tidal and femoral-arterial curves
#> as separate models; see modellib('Lu_2024_desflurane_pig_endtidal') and
#> modellib('Lu_2024_desflurane_pig_arterial').

Population

Seven male domestic piglets, 5 months old and weighing 28 +/- 5 kg, were anaesthetised with intramuscular tiletamine/zolazepam and atropine, maintained on fentanyl, propofol and pancuronium, tracheostomised and ventilated with 40% oxygen at a tidal volume of 8 mL/kg (Lu 2024 Sections 2.2-2.3). Desflurane was delivered through a vaporiser at a 5 L/min fresh-gas flow. Arterial and pulmonary-artery blood was drawn simultaneously at 1, 3, 5, 10, 20, 30, 40, 50, 60, 80, 100 and 120 min of each phase (Section 2.5). End-tidal values below 0.2% were discarded, which truncates the end-tidal curve after about 60 min of washout (Sections 2.4 and 4.5). No covariates were investigated.

str(mods[["Lu_2024_desflurane_pig_arterial"]]$population)
#> List of 10
#>  $ species       : chr "pig (domestic piglets)"
#>  $ n_subjects    : num 7
#>  $ n_studies     : num 1
#>  $ age_range     : chr "5 months"
#>  $ weight_mean   : chr "28 +/- 5 kg"
#>  $ sex_female_pct: num 0
#>  $ disease_state : chr "Healthy piglets under general anaesthesia (tiletamine/zolazepam and atropine induction; fentanyl, propofol and "| __truncated__
#>  $ dose_range    : chr "Desflurane 6 percent inspiratory via a 5 L/min fresh-gas flow for 120 min (uptake), then 40 percent oxygen alon"| __truncated__
#>  $ regions       : chr "Taiwan"
#>  $ notes         : chr "Lu 2024 Sections 2.2-2.5: seven male domestic piglets, fasted overnight. Femoral-arterial and pulmonary-artery "| __truncated__

Source trace

Every ini() value carries its origin as an in-file comment; the table below collects them.

Equation / parameter Value Source location
Disposition equation Y = A*exp(-alpha*t) + B*exp(-beta*t) n/a Table 3 footnote
Zero-order input over the 120-min uptake phase n/a Abstract; Section 4.2; Figure 3 caption
Lu_2024_desflurane_pig_endtidal
lcl 0.07 L/min Table 3, end-tidal, Clearance
lvc 0.10474 L derived CL/k10 from Table 3 end-tidal A, B, alpha and beta half-lives (Table 3 prints 0.10)
lq 0.026351 L/min derived k12*vc, Table 3 end-tidal macro-constants
lvp 1.7017 L derived q/k21, Table 3 end-tidal macro-constants
Lu_2024_desflurane_pig_arterial
lcl 0.08 L/min Table 3, arterial blood, Clearance
lvc 0.10349 L derived CL/k10 from Table 3 arterial A, B, alpha and beta half-lives (Table 3 prints 1.09; see below)
lq 0.020623 L/min derived k12*vc, Table 3 arterial macro-constants
lvp 1.0929 L derived q/k21, Table 3 arterial macro-constants
Lu_2024_desflurane_pig_mixedvenous
lcl 0.09 L/min Table 3, pulmonary artery blood, Clearance
lvc 0.34996 L derived CL/k10 from Table 3 pulmonary artery A, B, alpha and beta half-lives (Table 3 prints 0.35)
lq 0.038638 L/min derived k12*vc, Table 3 pulmonary artery macro-constants
lvp 2.3577 L derived q/k21, Table 3 pulmonary artery macro-constants
addSd (all three files) 0 (fixed) not reported; see Assumptions

Reconstructing the parameters from Table 3

Table 3 gives the biexponential macro-constants (A, B and the alpha and beta half-lives) together with secondary parameters (clearance, central volume, steady-state volume, mean residence time). The rate constants follow from the macro-constants alone; the clearance then fixes the scale of the volumes. The chunk below repeats the derivation from the Table 3 means and, in the other direction, recovers the macro-constants from the packaged parameters.

tab3 <- data.frame(
  model = des_models,
  A = c(731.26, 514.32, 147.32),       # %, Table 3
  B = c(3.45, 2.11, 2.10),             # %, Table 3
  thalpha = c(0.75, 0.71, 1.86),       # min, Table 3
  thbeta = c(61.90, 46.39, 61.29),     # min, Table 3
  CL = c(0.07, 0.08, 0.09),            # L/min, Table 3
  Vc = c(0.10, 1.09, 0.35),            # L, Table 3
  Vss = c(1.29, 1.20, 2.75),           # L, Table 3
  MRT = c(15.51, 15.00, 31.26),        # min, Table 3
  cmax = c(5.68, 4.85, 4.07),          # %, Table 3
  aucinf.obs = c(695.88, 603.03, 533.10) # % x min, Table 3
)

# Packaged typical values, read back from each model file
pk <- bind_rows(lapply(des_models, function(m) {
  th <- mods[[m]]$theta
  data.frame(
    model = m,
    cl = exp(th[["lcl"]]), vc = exp(th[["lvc"]]),
    q = exp(th[["lq"]]), vp = exp(th[["lvp"]])
  )
}))

recovered <- pk |>
  mutate(
    k10 = cl / vc, k12 = q / vc, k21 = q / vp,
    s = k10 + k12 + k21, p = k10 * k21,
    alpha = (s + sqrt(s^2 - 4 * p)) / 2,
    beta = (s - sqrt(s^2 - 4 * p)) / 2
  ) |>
  left_join(tab3, by = "model") |>
  mutate(
    # nominal input amount implied by the fit: A + B = dose / vc
    dose = vc * (A + B),
    A_rec = dose / vc * (alpha - k21) / (alpha - beta),
    B_rec = dose / vc * (k21 - beta) / (alpha - beta),
    thalpha_rec = log(2) / alpha,
    thbeta_rec = log(2) / beta,
    Vss_rec = vc + vp,
    MRT_rec = Vss_rec / cl,
    # central volume implied by the PRINTED Vss and MRT with the macro-constants
    Vc_check = Vss * (A / (log(2) / thalpha) + B / (log(2) / thbeta)) / ((A + B) * MRT)
  )

recovered |>
  transmute(
    Site = site_label[model],
    A = sprintf("%.2f / %.2f", A, A_rec),
    B = sprintf("%.2f / %.2f", B, B_rec),
    `alpha t1/2 (min)` = sprintf("%.2f / %.2f", thalpha, thalpha_rec),
    `beta t1/2 (min)` = sprintf("%.2f / %.2f", thbeta, thbeta_rec),
    `Input amount (% x L)` = sprintf("%.3f", dose)
  ) |>
  knitr::kable(caption = "Table 3 macro-constants (printed / recovered from the packaged model).")
Table 3 macro-constants (printed / recovered from the packaged model).
Site A B alpha t1/2 (min) beta t1/2 (min) Input amount (% x L)
End-tidal 731.26 / 731.26 3.45 / 3.45 0.75 / 0.75 61.90 / 61.90 76.953
Arterial blood 514.32 / 514.32 2.11 / 2.11 0.71 / 0.71 46.39 / 46.39 53.443
Pulmonary artery blood 147.32 / 147.32 2.10 / 2.10 1.86 / 1.86 61.29 / 61.29 52.291

stopifnot(
  # Closed-form round trip on the same numbers: only floating-point error.
  all(abs(recovered$A_rec / recovered$A - 1) < 1e-5),
  all(abs(recovered$B_rec / recovered$B - 1) < 1e-5),
  all(abs(recovered$thalpha_rec / recovered$thalpha - 1) < 1e-5),
  all(abs(recovered$thbeta_rec / recovered$thbeta - 1) < 1e-5)
)

The secondary parameters Table 3 prints act as an external check, because they were averaged across animals separately from the macro-constants:

recovered |>
  transmute(
    Site = site_label[model],
    `CL (L/min) printed / packaged` = sprintf("%.2f / %.3f", CL, cl),
    `Vc (L) printed / packaged` = sprintf("%.2f / %.3f", Vc, vc),
    `Vss (L) printed / packaged` = sprintf("%.2f / %.3f", Vss, Vss_rec),
    `MRT (min) printed / packaged` = sprintf("%.2f / %.2f", MRT, MRT_rec)
  ) |>
  knitr::kable(caption = "Table 3 secondary parameters against the packaged models.")
Table 3 secondary parameters against the packaged models.
Site CL (L/min) printed / packaged Vc (L) printed / packaged Vss (L) printed / packaged MRT (min) printed / packaged
End-tidal 0.07 / 0.070 0.10 / 0.105 1.29 / 1.806 15.51 / 25.81
Arterial blood 0.08 / 0.080 1.09 / 0.103 1.20 / 1.196 15.00 / 14.96
Pulmonary artery blood 0.09 / 0.090 0.35 / 0.350 2.75 / 2.708 31.26 / 30.09

For the two blood sites the packaged parameters regenerate the printed steady-state volumes and mean residence times to within 4%, and the pulmonary-artery central volume to rounding. The arterial central volume is printed as 1.09 L, but with the arterial macro-constants the central volume is fixed at Vss * AUC / ((A + B) * MRT) = 0.103 L from the printed Vss and MRT, whatever the input amount, and 1.09 L is instead the derived peripheral volume (1.093 L = Vss - Vc). The packaged arterial model therefore uses the central volume implied by the macro-constants; the printed 1.09 L is treated as a transcription slip.

The end-tidal means do not hang together as one curve: the beta half-life is 61.90 +/- 105.93 min, so the mean is pulled by one or more animals with a long terminal phase, and the reconstructed Vss and MRT exceed the printed means by 40-65%. The model reproduces the printed end-tidal macro-constants exactly, but, as the comparisons below show, a curve built from those mean macro-constants over-predicts the observed end-tidal plateau.

The product CL x AUC is close to 48 % x L for all three sites (0.07 x 695.88, 0.08 x 603.03, 0.09 x 533.10), which suggests the fits used one nominal input amount. The packaged models instead take the amount implied by each site’s printed clearance (the “Input amount” column above). Because the amount and the volumes scale together, this choice changes the clearance and volumes but not the predicted concentrations.

Simulation

The 6% inspiratory exposure is a zero-order infusion of each model’s nominal input amount into central over 120 min, followed by 120 min of washout. The models have no between-subject variability, so a single typical-value profile per site is the model prediction.

obs_times <- sort(unique(c(seq(0, 240, by = 0.5), 121, 123, 125, 130, 140, 180)))

sim <- bind_rows(lapply(des_models, function(m) {
  d <- recovered$dose[recovered$model == m]
  ev <- rxode2::et(amt = d, dur = 120, cmt = "central") |>
    rxode2::et(obs_times)
  out <- rxode2::rxSolve(mods[[m]], ev, returnType = "data.frame")
  out$model <- m
  out
}))
sim$site <- factor(site_label[sim$model], levels = site_label)

Replicate Figure 3

Table 1 reports the observed mean (SD) at 1, 3, 5, 10, 20, 60 and 120 min of each phase; Figure 3 shows the same means against the fitted curves.

t1_times <- c(1, 3, 5, 10, 20, 60, 120)
table1 <- data.frame(
  site = rep(site_label, each = 14),
  time = rep(c(t1_times, 120 + t1_times), 3),
  mean = c(
    # End-tidal (ET), %; '--' at 120 min of disposition (below the 0.2% cutoff)
    3.39, 4.75, 4.93, 5.35, 5.45, 5.61, 5.76, 2.18, 1.23, 1.04, 0.64, 0.38, 0.20, NA,
    # Arterial blood (A), %
    2.75, 3.65, 4.00, 4.17, 4.44, 4.58, 4.65, 2.23, 1.39, 1.24, 0.84, 0.54, 0.32, 0.23,
    # Pulmonary artery blood (PA), %
    1.15, 2.24, 2.61, 3.02, 3.27, 3.62, 4.10, 2.96, 2.08, 1.69, 1.22, 0.85, 0.49, 0.34
  ),
  sd = c(
    0.79, 0.62, 0.23, 0.28, 0.09, 0.09, 0.05, 0.88, 0.47, 0.55, 0.17, 0.10, 0.04, NA,
    0.67, 0.39, 0.53, 0.54, 0.39, 0.40, 0.26, 0.87, 0.43, 0.47, 0.17, 0.07, 0.05, 0.06,
    0.65, 0.45, 0.39, 0.26, 0.34, 0.47, 0.41, 0.54, 0.39, 0.33, 0.23, 0.13, 0.06, 0.05
  )
) |>
  filter(!is.na(mean))
table1$site <- factor(table1$site, levels = site_label)
ggplot(sim, aes(time, Cc)) +
  geom_line(colour = "steelblue") +
  geom_pointrange(
    data = table1,
    aes(time, mean, ymin = mean - sd, ymax = mean + sd),
    shape = 1, size = 0.3
  ) +
  facet_wrap(~site, ncol = 1) +
  scale_x_continuous(breaks = seq(0, 240, by = 60)) +
  labs(x = "Time (min)", y = "Desflurane concentration (%)")
Replicates Figure 3 of Lu 2024: typical-value prediction from the mean Table 3 parameters (line) against the observed means +/- SD from Table 1 (points).

Replicates Figure 3 of Lu 2024: typical-value prediction from the mean Table 3 parameters (line) against the observed means +/- SD from Table 1 (points).

cmp <- table1 |>
  left_join(sim |> select(site, time, Cc), by = c("site", "time")) |>
  mutate(pct_diff = 100 * (Cc - mean) / mean)

cmp |>
  mutate(
    Phase = if_else(time <= 120, "uptake", "disposition"),
    `Phase time (min)` = if_else(time <= 120, time, time - 120)
  ) |>
  select(site, Phase, `Phase time (min)`, mean, Cc, pct_diff) |>
  mutate(Cc = signif(Cc, 3), pct_diff = round(pct_diff, 1)) |>
  rename(
    "Site" = site,
    "Observed mean (%)" = mean,
    "Model (%)" = Cc,
    "Difference (%)" = pct_diff
  ) |>
  knitr::kable(caption = "Model prediction against the Table 1 means.")
Model prediction against the Table 1 means.
Site Phase Phase time (min) Observed mean (%) Model (%) Difference (%)
End-tidal uptake 1 3.39 4.010 18.2
End-tidal uptake 3 4.75 6.270 31.9
End-tidal uptake 5 4.93 6.670 35.3
End-tidal uptake 10 5.35 6.870 28.3
End-tidal uptake 20 5.45 7.110 30.4
End-tidal uptake 60 5.61 7.850 39.9
End-tidal uptake 120 5.76 8.490 47.4
End-tidal disposition 1 2.18 4.490 106.1
End-tidal disposition 3 1.23 2.250 82.7
End-tidal disposition 5 1.04 1.860 78.8
End-tidal disposition 10 0.64 1.700 165.2
End-tidal disposition 20 0.38 1.520 299.2
End-tidal disposition 60 0.20 0.969 384.6
Arterial blood uptake 1 2.75 2.750 0.1
Arterial blood uptake 3 3.65 4.210 15.3
Arterial blood uptake 5 4.00 4.440 11.0
Arterial blood uptake 10 4.17 4.550 9.2
Arterial blood uptake 20 4.44 4.690 5.7
Arterial blood uptake 60 4.58 5.090 11.1
Arterial blood uptake 120 4.65 5.370 15.5
Arterial blood disposition 1 2.23 2.620 17.5
Arterial blood disposition 3 1.39 1.170 -15.6
Arterial blood disposition 5 1.24 0.944 -23.9
Arterial blood disposition 10 0.84 0.845 0.6
Arterial blood disposition 20 0.54 0.728 34.7
Arterial blood disposition 60 0.32 0.400 25.1
Arterial blood disposition 120 0.23 0.163 -29.0
Pulmonary artery blood uptake 1 1.15 1.040 -9.4
Pulmonary artery blood uptake 3 2.24 2.270 1.3
Pulmonary artery blood uptake 5 2.61 2.870 9.9
Pulmonary artery blood uptake 10 3.02 3.380 11.9
Pulmonary artery blood uptake 20 3.27 3.610 10.3
Pulmonary artery blood uptake 60 3.62 4.060 12.1
Pulmonary artery blood uptake 120 4.10 4.440 8.4
Pulmonary artery blood disposition 1 2.96 3.410 15.1
Pulmonary artery blood disposition 3 2.08 2.190 5.2
Pulmonary artery blood disposition 5 1.69 1.600 -5.5
Pulmonary artery blood disposition 10 1.22 1.110 -9.4
Pulmonary artery blood disposition 20 0.85 0.918 8.0
Pulmonary artery blood disposition 60 0.49 0.583 19.0
Pulmonary artery blood disposition 120 0.34 0.296 -13.0

The pulmonary-artery curve tracks the observed means through both phases to within about 15%. The arterial curve runs 5-15% high on the uptake plateau and is up to about 30% off the observed means during washout, where the fast alpha phase (half-life 0.71 min) makes the early disposition points sensitive to averaging across animals. The end-tidal curve reaches a plateau far above the 6% inspired concentration, which is physically impossible for an end-tidal value. This is the averaging artefact described above, not a property of the measured data.

blood <- cmp |> filter(site != "End-tidal")
stopifnot(
  # Uptake plateau (20-120 min): the blood models sit within ~15% of the
  # observed means. The prediction is deterministic (no random effects).
  all(abs(blood$pct_diff[blood$time %in% c(20, 60, 120)]) < 20),
  # Across all tabulated blood timepoints the centre of the error is modest.
  abs(median(blood$pct_diff)) < 15
)

Validation against the published NCA (PKNCA)

Table 3 reports Cmax and AUC for each site. In a WinNonlin infusion fit the AUC is the area under the fitted curve to infinity, so it is compared with PKNCA’s aucinf.obs on the simulated profile.

conc <- sim |>
  filter(!is.na(Cc)) |>
  mutate(id = 1L, treatment = as.character(site)) |>
  select(id, treatment, time, Cc)

dose_df <- recovered |>
  transmute(id = 1L, treatment = site_label[model], time = 0, amt = dose, duration = 120)

conc_obj <- PKNCA::PKNCAconc(conc, Cc ~ time | treatment + id)
dose_obj <- PKNCA::PKNCAdose(
  dose_df, amt ~ time | treatment + id,
  duration = "duration", route = "intravascular"
)
intervals <- data.frame(start = 0, end = Inf, cmax = TRUE, tmax = TRUE, aucinf.obs = TRUE)
nca_res <- PKNCA::pk.nca(PKNCA::PKNCAdata(conc_obj, dose_obj, intervals = intervals))

reference <- tab3 |>
  transmute(treatment = site_label[model], cmax, aucinf.obs)

nca_tbl <- nlmixr2lib::ncaComparisonTable(
  nca_res, reference,
  by = "treatment",
  params = c("cmax", "aucinf.obs"),
  units = c(cmax = "%", aucinf.obs = "% x min")
)
nca_tbl |>
  dplyr::rename("Site" = treatment) |>
  knitr::kable(caption = "Simulated NCA against Lu 2024 Table 3.")
Simulated NCA against Lu 2024 Table 3.
NCA parameter Site Reference Simulated % diff
Cmax (%) End-tidal 5.68 8.49 +49.5%*
Cmax (%) Arterial blood 4.85 5.37 +10.7%
Cmax (%) Pulmonary artery blood 4.07 4.44 +9.2%
AUC0-∞ (obs) (% x min) End-tidal 696 1100 +58.0%*
AUC0-∞ (obs) (% x min) Arterial blood 603 668 +10.8%
AUC0-∞ (obs) (% x min) Pulmonary artery blood 533 581 +9.0%
  • differs from reference by more than ±20%.

The two blood models reproduce the printed Cmax and AUC within about 11%; the residual is the expected gap between a curve built from mean parameters and the mean of the individual Cmax and AUC values. The end-tidal row is starred for the reason given above.

nca_wide <- as.data.frame(nca_res) |>
  select(treatment, PPTESTCD, PPORRES) |>
  pivot_wider(names_from = PPTESTCD, values_from = PPORRES) |>
  left_join(reference, by = "treatment", suffix = c("_sim", "_ref")) |>
  filter(treatment != "End-tidal")
stopifnot(
  all(abs(nca_wide$cmax_sim / nca_wide$cmax_ref - 1) < 0.15),
  all(abs(nca_wide$aucinf.obs_sim / nca_wide$aucinf.obs_ref - 1) < 0.15)
)

Assumptions and deviations

  • Typical values only. Lu 2024 fitted each piglet separately and reported the mean and SD of the individual estimates; there is no population model, so the packaged models carry no between-subject variability and the residual SD is fixed at zero.
  • Parameterisation. The volumes and intercompartmental clearance are derived from the Table 3 macro-constants, with the clearance anchored to the printed Table 3 mean. Any other anchor would rescale the volumes and the input amount together without changing the predicted concentrations.
  • Input amount. The paper does not state the dose amount used in the WinNonlin fits. Each model’s description gives the amount, in % x L, that the packaged clearance implies for the 6%, 120-min exposure. Simulating other inspired concentrations by scaling the infusion rate in proportion assumes the linear zero-order input the authors fitted; the paper studied 6% only.
  • Arterial central volume. The printed 1.09 L is inconsistent with the arterial macro-constants and equals the derived peripheral volume; the model uses the 0.103 L the macro-constants imply.
  • End-tidal model. Shipped for completeness. Its mean macro-constants over-predict the observed end-tidal plateau by about 50% and should not be used for quantitative end-tidal simulation. The compartmentData specimen for its states is “not applicable”, because they are curve-fit spaces observed through exhaled gas, for which the specimen vocabulary has no member.
  • Concentration units. All concentrations are percent vapour (the partial pressure as a volume percent), as reported. Converting blood values to an amount requires the blood-gas partition coefficient (0.38 +/- 0.01 in these piglets).