Skip to contents

Model and source

  • Citation: Gebreyesus MS, Dresner A, Wiesner L, Coetzee E, Verschuuren T, Wasmann R, Denti P. Dose optimization of cefazolin in South African children undergoing cardiac surgery with cardiopulmonary bypass. CPT Pharmacometrics Syst Pharmacol 2024;13(9):1595-1605. doi:10.1002/psp4.13196.
  • Description: Two-compartment IV population PK model for total plasma cefazolin in South African children under 25 kg undergoing cardiac surgery with cardiopulmonary bypass (CPB), with a third compartment for the CPB circuit. The circuit compartment exchanges with the central compartment only while the patient is on bypass; its volume is the circuit priming volume and its flow a fixed adult cardiac output (336 L/h, allometric to 70 kg), both multiplied by an estimated scaling factor (0.679). Clearances and volumes scale allometrically on body weight (reference 10 kg) and clearance is proportional to bedside-Schwartz eGFR (reference 120 mL/min/1.73 m^2). A dose added to the CPB priming fluid is a bolus into the circuit compartment.
  • Article: https://doi.org/10.1002/psp4.13196 (open access; PMC11881764)

Gebreyesus and colleagues modelled total plasma cefazolin in 22 children under 25 kg undergoing cardiac surgery with cardiopulmonary bypass (CPB) in Cape Town, then used the model to compare the hospital’s 50 mg/kg every-4-hour bolus regimen against a continuous infusion whose loading and maintenance doses are computed from the model’s typical clearance and steady-state volume. The NONMEM control stream is published as Supplementary Material S3, and every value below is taken from it and from Table 3, which agree.

The CPB circuit is a third compartment. While the patient is connected, drug exchanges between the central compartment and the circuit at a flow fixed to an adult cardiac output (336 L/h, scaled allometrically from 70 kg), and the circuit volume is the priming volume; an estimated factor (0.679) multiplies both, which the authors note is close to the plasma fraction of whole blood. A dose added to the priming fluid is a bolus into the circuit compartment. When the patient comes off bypass the exchange stops and whatever drug is in the circuit stays there.

Population

Twenty-two children (13 male, 9 female) aged 1-94 months (median 17), weighing 2.0-22 kg (median 8.5) and with bedside-Schwartz eGFR 48-159 mL/min/1.73 m^2 (median 116) contributed 249 plasma concentrations (Table 2). They underwent surgery for congenital heart disease such as atrioventricular septal defect, pulmonary atresia and tetralogy of Fallot; surgical duration was 2.38 (0.567-5.63) h and priming volume 0.40 (0.25-0.55) L. Each child received 50 mg/kg cefazolin by 45-second IV injection after induction and again 4-6 h later (at most three doses); six also had 50 mg/kg added to the CPB priming fluid.

The same information is available programmatically via readModelDb("Gebreyesus_2024_cefazolin")()$population.

Source trace

Equation / parameter Value Source location
lcl log(0.856) L/h Table 3 ‘Clearance’; S3 THETA(1)
lvc log(1.07) L Table 3 ‘Central volume of distribution’; S3 THETA(2)
lq log(1.41) L/h Table 3 ‘Intercompartmental clearance’; S3 THETA(6)
lvp log(1.61) L Table 3 ‘Peripheral volume of distribution’; S3 THETA(5)
qbfr 336 L/h, fixed Table 3 ‘Intercompartmental clearance for CPB compartment’; S3 THETA(7)
fcircuit 0.679 Table 3 ‘Scaling factor for the CPB compartment’; S3 THETA(8)
e_wt_cl_q 0.75, fixed S1 ‘Population pharmacokinetic modelling’; S3 ALLMCL_WT, ALLMQ2
e_wt_vc_vp 1, fixed S1; S3 ALLMV_WT
etalcl 0.0523 S3 $OMEGA (Table 3: 23.2% CV)
etalvc 0.033 S3 $OMEGA (Table 3: 18.3% CV)
etalvp 0.0779 S3 $OMEGA (Table 3: 28.5% CV)
propSd 0.155 Table 3 ‘Proportional error’; S3 THETA(3)
addSd 0.1172 mg/L, fixed Table 3 footnote f (20% of LLOQ 0.586 mg/L); S3 ADD = THETA(4) + LLOQ*0.2
cl = TVCL (WT/10)^0.75 (CRCL/120) n/a Results equation; S3 TVCL
vc, vp = TV (WT/10) n/a Results equation; S3 TVV, TVV2
q = TVQ (WT/10)^0.75 (no BSV) n/a S3 TVQ, Q = TVQ
circuit flow 336 (WT/70)^0.75 x 0.679 n/a Results; S3 Q2 = TVQ2*THETA(8)
circuit volume CIRCUIT_PRIME_VOL x 0.679 n/a S3 V3 = VCPB0*THETA(8)
d/dt(...) with the circuit terms gated by the CPB switch n/a S3 $DES
eGFR 0.413 x HT / SCr n/a S1 Eq. 1; S3 CRCL_SCHW
Cc ~ add + prop, combined2 n/a S3 W = SQRT(ADD**2+PROP**2)

Dose calculator (Tables 1, S1 and S2)

The continuous-infusion regimen sets the loading dose to Vss x 100 mg/L and the hourly maintenance rate to CL x 100 mg/L, using the population-predicted parameters for the child’s weight and eGFR (Table 1). Supplementary Table S1 works the example for a 2 kg child with eGFR 50. The model’s typical values reproduce it exactly.

mod <- readModelDb("Gebreyesus_2024_cefazolin")
mod_typical <- rxode2::zeroRe(mod)
#> ℹ parameter labels from comments will be replaced by 'label()'

typical_params <- function(WT, CRCL) {
  ev <- data.frame(
    id = seq_along(WT), time = 0, evid = 0L, amt = 0, cmt = "central",
    WT = WT, CRCL = CRCL, CPB_ON = 0L, CPB_REWARM = 0L, CIRCUIT_PRIME_VOL = 0.4
  )
  sim <- as.data.frame(rxode2::rxSolve(mod_typical, ev, returnType = "data.frame"))
  # rxSolve drops the id column when there is a single subject
  if (!"id" %in% names(sim)) sim$id <- 1L
  dplyr::transmute(sim, id, cl, vss = vc + vp)
}

c_target <- 100 # mg/L (Methods)
s1 <- typical_params(WT = 2, CRCL = 50) |>
  dplyr::mutate(
    loading_mg = vss * c_target,
    maint_mg_h = cl * c_target,
    cpb_dose_mg = 0.5 * loading_mg
  )
#> ℹ omega/sigma items treated as zero: 'etalcl', 'etalvc', 'etalvp'
s1_published <- c(vss = 0.536, cl = 0.107, loading_mg = 53.6, maint_mg_h = 10.7, cpb_dose_mg = 26.8)
s1_model <- unlist(s1[names(s1_published)])
knitr::kable(
  data.frame(
    Quantity = c("Vss (L)", "CL (L/h)", "Loading dose (mg)", "Maintenance (mg/h)", "CPB dose (mg)"),
    Model = signif(s1_model, 3),
    `Table S1` = s1_published,
    check.names = FALSE
  ),
  caption = "Replicates Supplementary Table S1 (2 kg, eGFR 50 mL/min/1.73 m^2)."
)
Replicates Supplementary Table S1 (2 kg, eGFR 50 mL/min/1.73 m^2).
Quantity Model Table S1
vss Vss (L) 0.536 0.536
cl CL (L/h) 0.107 0.107
loading_mg Loading dose (mg) 53.600 53.600
maint_mg_h Maintenance (mg/h) 10.700 10.700
cpb_dose_mg CPB dose (mg) 26.800 26.800
# Same equations evaluated on both sides; only the table's 3-digit rounding differs.
stopifnot(all(abs(s1_model / s1_published - 1) < 0.005))

Supplementary Table S2 lists, for each of the 22 study patients, the total cefazolin given over 12 h by each regimen. The continuous-infusion totals are the loading dose plus the maintenance rate over the 11 h 40 min between the start of the infusion (20 min after induction) and the end of the 12-h window.

table_s2 <- tibble::tribble(
  ~patient, ~WT, ~CRCL, ~standard_mg, ~alternative_mg,
  1, 2, 48, 300, 174,
  2, 5.5, 81, 825, 580,
  3, 5.9, 85, 885, 635,
  4, 5.4, 91, 810, 623,
  5, 8.4, 97, 1260, 933,
  6, 6, 97, 900, 713,
  7, 5.3, 99, 795, 655,
  8, 8.9, 108, 1335, 1063,
  9, 11.8, 108, 1770, 1337,
  10, 10, 114, 1500, 1213,
  11, 8.5, 116, 1275, 1080,
  12, 9.4, 117, 1410, 1177,
  13, 18.7, 119, 2805, 2080,
  14, 10, 128, 1500, 1331,
  15, 7.8, 129, 1170, 1103,
  16, 12.5, 131, 1875, 1620,
  17, 6.3, 131, 945, 938,
  18, 8.1, 132, 1215, 1156,
  19, 17, 135, 2550, 2125,
  20, 21.6, 140, 3240, 2651,
  21, 14, 145, 2100, 1924,
  22, 12.6, 159, 1890, 1915
)
infusion_h <- 12 - 20 / 60
s2_check <- typical_params(table_s2$WT, table_s2$CRCL) |>
  dplyr::mutate(
    model_alternative_mg = vss * c_target + cl * c_target * infusion_h,
    model_standard_mg = 3 * 50 * table_s2$WT
  ) |>
  dplyr::bind_cols(table_s2 |> dplyr::select(-patient)) |>
  dplyr::mutate(pct_diff = 100 * (model_alternative_mg / alternative_mg - 1))
#> ℹ omega/sigma items treated as zero: 'etalcl', 'etalvc', 'etalvp'
#> Warning: multi-subject simulation without without 'omega'
s2_check |>
  dplyr::transmute(
    Patient = id, `Weight (kg)` = WT, `eGFR` = CRCL,
    `Standard, Table S2 (mg)` = standard_mg, `Standard, model (mg)` = model_standard_mg,
    `Infusion, Table S2 (mg)` = alternative_mg,
    `Infusion, model (mg)` = round(model_alternative_mg),
    `Difference (%)` = round(pct_diff, 2)
  ) |>
  knitr::kable(caption = "Replicates Supplementary Table S2 (totals over 12 h, no CPB dose).")
Replicates Supplementary Table S2 (totals over 12 h, no CPB dose).
Patient Weight (kg) eGFR Standard, Table S2 (mg) Standard, model (mg) Infusion, Table S2 (mg) Infusion, model (mg) Difference (%)
1 2.0 48 300 300 174 173 -0.54
2 5.5 81 825 825 580 578 -0.36
3 5.9 85 885 885 635 634 -0.11
4 5.4 91 810 810 623 622 -0.20
5 8.4 97 1260 1260 933 933 0.05
6 6.0 97 900 900 713 711 -0.26
7 5.3 99 795 795 655 654 -0.18
8 8.9 108 1335 1335 1063 1062 -0.08
9 11.8 108 1770 1770 1337 1334 -0.24
10 10.0 114 1500 1500 1213 1217 0.31
11 8.5 116 1275 1275 1080 1082 0.22
12 9.4 117 1410 1410 1177 1181 0.38
13 18.7 119 2805 2805 2080 2085 0.23
14 10.0 128 1500 1500 1331 1333 0.17
15 7.8 129 1170 1170 1103 1100 -0.26
16 12.5 131 1875 1875 1620 1624 0.24
17 6.3 131 945 945 938 940 0.19
18 8.1 132 1215 1215 1156 1155 -0.08
19 17.0 135 2550 2550 2125 2128 0.15
20 21.6 140 3240 3240 2651 2655 0.14
21 14.0 145 2100 2100 1924 1928 0.22
22 12.6 159 1890 1890 1915 1911 -0.19
# Table S2 prints eGFR rounded to an integer; the published totals were
# computed from the unrounded value, which moves the total by at most ~0.3%.
stopifnot(
  all(s2_check$model_standard_mg == s2_check$standard_mg),
  all(abs(s2_check$pct_diff) < 1)
)

Typical profiles of the smallest patient (Figures S2 and S3)

The authors simulated the smallest study patient (2 kg, eGFR 50) through a 12-h dosing period with the circuit connected from 1 to 6.5 h (the grey band in Figures S2 and S3) and the 0.4 L priming volume the hospital uses below 10 kg. The standard regimen is 100 mg at 0, 4 and 8 h, with an optional 100 mg CPB dose; the continuous infusion is a 53.6 mg loading dose and 10.7 mg/h from 20 min, with an optional 26.8 mg CPB dose (Table S1). The CPB dose is placed in the circuit at the moment of connection.

cpb_start <- 1
cpb_end <- 6.5
cpb_switch <- function(time) as.integer(time >= cpb_start & time < cpb_end)

make_events <- function(doses, WT, CRCL, prime, obs_times, id = 1L) {
  obs <- data.frame(time = obs_times, evid = 0L, amt = 0, rate = 0, cmt = "central")
  dplyr::bind_rows(doses, obs) |>
    dplyr::mutate(
      id = id, WT = WT, CRCL = CRCL, CIRCUIT_PRIME_VOL = prime,
      CPB_ON = cpb_switch(time), CPB_REWARM = 0L
    ) |>
    dplyr::arrange(time, dplyr::desc(evid))
}

regimen_doses <- function(regimen, WT, loading, maint) {
  bolus <- 50 * WT
  std <- data.frame(time = c(0, 4, 8), evid = 1L, amt = bolus, rate = 0, cmt = "central")
  ci <- data.frame(
    time = c(0, 20 / 60), evid = 1L,
    amt = c(loading, maint * infusion_h), rate = c(0, maint), cmt = "central"
  )
  cpb <- function(amt) data.frame(time = cpb_start, evid = 1L, amt = amt, rate = 0, cmt = "circuit_plasma")
  switch(regimen,
    "Standard" = std,
    "Standard + CPB dose" = dplyr::bind_rows(std, cpb(bolus)),
    "Infusion" = ci,
    "Infusion + CPB dose" = dplyr::bind_rows(ci, cpb(0.5 * loading))
  )
}

regimens <- c("Standard", "Standard + CPB dose", "Infusion", "Infusion + CPB dose")
fine_times <- sort(unique(c(seq(0, 12, by = 0.02), 0.999, 3.999, 6.499, 7.999)))
small <- lapply(seq_along(regimens), function(i) {
  ev <- make_events(
    regimen_doses(regimens[i], WT = 2, loading = s1$loading_mg, maint = s1$maint_mg_h),
    WT = 2, CRCL = 50, prime = 0.4, obs_times = fine_times, id = i
  )
  as.data.frame(rxode2::rxSolve(mod_typical, ev, returnType = "data.frame")) |>
    dplyr::mutate(regimen = regimens[i])
}) |>
  dplyr::bind_rows()
#> ℹ omega/sigma items treated as zero: 'etalcl', 'etalvc', 'etalvp'
#> ℹ omega/sigma items treated as zero: 'etalcl', 'etalvc', 'etalvp'
#> ℹ omega/sigma items treated as zero: 'etalcl', 'etalvc', 'etalvp'
#> ℹ omega/sigma items treated as zero: 'etalcl', 'etalvc', 'etalvp'

ggplot(small, aes(time, Cc)) +
  annotate("rect", xmin = cpb_start, xmax = cpb_end, ymin = 0, ymax = Inf, alpha = 0.2) +
  geom_line(colour = "purple") +
  geom_hline(yintercept = c(40, 360), colour = "red") +
  facet_wrap(~regimen) +
  labs(
    x = "Time after first dose (h)", y = "Cefazolin concentration (mg/L)",
    caption = "Replicates Figures S2 and S3 of Gebreyesus 2024 (grey: connected to CPB)."
  )

The maintainers read the concentrations at the troughs and band edges off Figures S2 and S3 (peaks are not compared: the published curves are drawn on a grid too coarse to resolve the instant after a bolus).

digitised <- tibble::tribble(
  ~regimen, ~time, ~published,
  "Standard", 0.999, 133,
  "Standard", 3.999, 60,
  "Standard", 6.499, 128,
  "Standard", 7.999, 95,
  "Standard", 12, 118,
  "Standard + CPB dose", 3.999, 140,
  "Standard + CPB dose", 6.499, 185,
  "Standard + CPB dose", 7.999, 138,
  "Standard + CPB dose", 12, 142,
  "Infusion", 0.999, 90,
  "Infusion", 6.499, 80,
  "Infusion", 12, 92,
  "Infusion + CPB dose", 6.499, 95,
  "Infusion + CPB dose", 12, 98
)
small_check <- digitised |>
  dplyr::left_join(small |> dplyr::select(regimen, time, Cc), by = c("regimen", "time")) |>
  dplyr::mutate(pct_diff = 100 * (Cc / published - 1))
knitr::kable(small_check, digits = 1, caption = "Typical-value concentrations (mg/L) vs Figures S2 and S3.")
Typical-value concentrations (mg/L) vs Figures S2 and S3.
regimen time published Cc pct_diff
Standard 1.0 133 137.4 3.3
Standard 4.0 60 62.8 4.6
Standard 6.5 128 128.8 0.6
Standard 8.0 95 94.7 -0.3
Standard 12.0 118 120.5 2.1
Standard + CPB dose 4.0 140 140.6 0.5
Standard + CPB dose 6.5 185 185.5 0.2
Standard + CPB dose 8.0 138 136.6 -1.0
Standard + CPB dose 12.0 142 140.7 -0.9
Infusion 1.0 90 92.4 2.6
Infusion 6.5 80 80.6 0.7
Infusion 12.0 92 93.1 1.2
Infusion + CPB dose 6.5 95 95.8 0.8
Infusion + CPB dose 12.0 98 98.5 0.5
# Typical-value solve against digitised curves: the only error is the
# read-off (about +/- 5 mg/L on these axes), so a 10% envelope is tight
# enough that a wrong circuit volume or flow fails it.
stopifnot(all(abs(small_check$pct_diff) < 10))

Probability of target attainment (Figure 3)

The efficacy target is a free concentration above 4 x MIC for the whole 12-h period; with a free fraction of 0.2 the total-concentration threshold is 20 x MIC (40 mg/L at the MSSA cutoff of 2 mg/L). Figure 3 shows the standard regimen failing worst in small children with high eGFR. The paper used 1000 virtual patients per weight and renal-function group; this vignette simulates the most discriminating row of the figure (eGFR 130 mL/min/1.73 m^2) with 200 children per weight band and regimen, weight uniform within each band, and the hospital’s priming volumes (0.4 L below 10 kg, 0.6 L for 10-20 kg, 0.8 L above).

rxode2::rxSetSeed(2024)
n_per_arm <- 200
weight_bands <- tibble::tribble(
  ~band, ~lo, ~hi,
  "<5 kg", 2, 5,
  "5-10 kg", 5, 10,
  "10-15 kg", 10, 15,
  "15-20 kg", 15, 20,
  "20-25 kg", 20, 25
)
# Troughs fall just before the 4 h and 8 h doses, at 12 h, and in the dip
# just after connection to the circuit at 1 h.
pta_times <- sort(unique(c(seq(0, 12, by = 0.05), seq(1, 1.3, by = 0.01), 3.999, 7.999)))
theta <- rxode2::rxode2(mod)$theta
#> ℹ parameter labels from comments will be replaced by 'label()'
arm_grid <- tidyr::expand_grid(band = weight_bands$band, regimen = regimens)

pta_events <- lapply(seq_len(nrow(arm_grid)), function(a) {
  wb <- weight_bands[weight_bands$band == arm_grid$band[a], ]
  wts <- stats::runif(n_per_arm, wb$lo, wb$hi)
  lapply(seq_len(n_per_arm), function(j) {
    WT <- wts[j]
    loading <- (exp(theta[["lvc"]]) + exp(theta[["lvp"]])) * (WT / 10) * c_target
    maint <- exp(theta[["lcl"]]) * (WT / 10)^0.75 * (130 / 120) * c_target
    prime <- if (WT < 10) 0.4 else if (WT < 20) 0.6 else 0.8
    make_events(
      regimen_doses(arm_grid$regimen[a], WT, loading, maint),
      WT = WT, CRCL = 130, prime = prime, obs_times = pta_times,
      id = (a - 1L) * n_per_arm + j
    ) |>
      dplyr::mutate(band = arm_grid$band[a], regimen = arm_grid$regimen[a])
  }) |>
    dplyr::bind_rows()
}) |>
  dplyr::bind_rows()
stopifnot(!anyDuplicated(unique(pta_events[, c("id", "time", "evid", "cmt")])))

pta_sim <- rxode2::rxSolve(mod, pta_events, keep = c("band", "regimen"), returnType = "data.frame") |>
  as.data.frame()
#> ℹ parameter labels from comments will be replaced by 'label()'

Cc is the individual prediction without residual error, which is what a target-attainment simulation evaluates.

mic <- c(0.125, 0.25, 0.5, 1, 2)
# The t = 0 row follows the first bolus, so it is a peak, not a trough.
cmin_by_id <- pta_sim |>
  dplyr::group_by(id, band, regimen) |>
  dplyr::summarise(cmin = min(Cc), .groups = "drop")
pta <- tidyr::expand_grid(cmin_by_id, mic = mic) |>
  dplyr::group_by(band, regimen, mic) |>
  dplyr::summarise(pta = 100 * mean(cmin > 20 * mic), .groups = "drop") |>
  dplyr::mutate(band = factor(band, levels = weight_bands$band))

ggplot(pta, aes(factor(mic), pta, colour = regimen, group = regimen)) +
  geom_line() +
  geom_point() +
  geom_hline(yintercept = 90, linetype = "dashed") +
  facet_wrap(~band, nrow = 1) +
  labs(
    x = "MIC for MSSA (mg/L)", y = "PTA (%)", colour = NULL,
    caption = "Replicates the eGFR 130 mL/min/1.73 m^2 row of Figure 3 of Gebreyesus 2024."
  ) +
  theme(legend.position = "bottom")

The maintainers read the MIC 2 mg/L points of that row off Figure 3.

pta_published <- tibble::tribble(
  ~band, ~regimen, ~published,
  "<5 kg", "Standard", 15,
  "5-10 kg", "Standard", 35,
  "10-15 kg", "Standard", 50,
  "15-20 kg", "Standard", 62,
  "20-25 kg", "Standard", 67,
  "<5 kg", "Standard + CPB dose", 52,
  "5-10 kg", "Standard + CPB dose", 75,
  "10-15 kg", "Standard + CPB dose", 86,
  "15-20 kg", "Standard + CPB dose", 91,
  "20-25 kg", "Standard + CPB dose", 94
)
pta_check <- pta |>
  dplyr::filter(mic == 2) |>
  dplyr::mutate(band = as.character(band)) |>
  dplyr::left_join(pta_published, by = c("band", "regimen")) |>
  dplyr::mutate(diff_pp = pta - published) |>
  dplyr::arrange(regimen, factor(band, levels = weight_bands$band))
pta_check |>
  dplyr::rename(`Weight band` = band, Regimen = regimen, `MIC (mg/L)` = mic,
                `PTA, simulated (%)` = pta, `PTA, Figure 3 (%)` = published,
                `Difference (pp)` = diff_pp) |>
  knitr::kable(digits = 1, caption = "PTA at MIC 2 mg/L, eGFR 130 mL/min/1.73 m^2.")
PTA at MIC 2 mg/L, eGFR 130 mL/min/1.73 m^2.
Weight band Regimen MIC (mg/L) PTA, simulated (%) PTA, Figure 3 (%) Difference (pp)
<5 kg Infusion 2 83.0 NA NA
5-10 kg Infusion 2 100.0 NA NA
10-15 kg Infusion 2 100.0 NA NA
15-20 kg Infusion 2 100.0 NA NA
20-25 kg Infusion 2 100.0 NA NA
<5 kg Infusion + CPB dose 2 100.0 NA NA
5-10 kg Infusion + CPB dose 2 100.0 NA NA
10-15 kg Infusion + CPB dose 2 100.0 NA NA
15-20 kg Infusion + CPB dose 2 100.0 NA NA
20-25 kg Infusion + CPB dose 2 100.0 NA NA
<5 kg Standard 2 14.5 15 -0.5
5-10 kg Standard 2 31.0 35 -4.0
10-15 kg Standard 2 42.0 50 -8.0
15-20 kg Standard 2 60.5 62 -1.5
20-25 kg Standard 2 65.0 67 -2.0
<5 kg Standard + CPB dose 2 42.5 52 -9.5
5-10 kg Standard + CPB dose 2 71.5 75 -3.5
10-15 kg Standard + CPB dose 2 88.0 86 2.0
15-20 kg Standard + CPB dose 2 94.0 91 3.0
20-25 kg Standard + CPB dose 2 96.0 94 2.0

bolus_rows <- pta_check |> dplyr::filter(!is.na(published))
infusion_rows <- pta_check |> dplyr::filter(grepl("Infusion", regimen))
stopifnot(
  # Centre: a mis-scaled clearance or circuit volume shifts every band by tens
  # of points; 200 per arm gives a Monte Carlo SE of at most 3.5 points.
  abs(median(bolus_rows$diff_pp)) < 10,
  # Envelope across the ten bolus arms, robust to which arm sits in a tail.
  quantile(abs(bolus_rows$diff_pp), 0.9) < 20,
  # Figure 3: the infusion regimens attain the target in essentially every child.
  median(infusion_rows$pta) > 95
)

One arm departs from the figure: the infusion without a CPB dose in children under 5 kg reaches about 85% here against about 97% in Figure 3. In these children the circuit volume (0.4 L x 0.679) is comparable to the central volume, so concentrations fall by roughly half the moment the circuit is connected and recover within minutes (the dip in Figure S3). The simulation above looks for the minimum on a 0.01-h grid around connection; on a coarser grid the dip is missed and the PTA rises to the published value, which suggests the paper evaluated attainment on a coarser grid. The paper does not state its grid, so the comparison is shown rather than asserted.

grid_pta <- lapply(c(0.01, 0.05, 0.1, 0.25), function(step) {
  pta_sim |>
    dplyr::filter(band == "<5 kg", regimen == "Infusion", abs(time / step - round(time / step)) < 1e-6) |>
    dplyr::group_by(id) |>
    dplyr::summarise(cmin = min(Cc), .groups = "drop") |>
    dplyr::summarise(`Grid step (h)` = step, `PTA at MIC 2 (%)` = 100 * mean(cmin > 40))
}) |>
  dplyr::bind_rows()
knitr::kable(grid_pta, digits = c(2, 1), caption = "Infusion without CPB dose, < 5 kg, eGFR 130: PTA by evaluation grid.")
Infusion without CPB dose, < 5 kg, eGFR 130: PTA by evaluation grid.
Grid step (h) PTA at MIC 2 (%)
0.01 83.0
0.05 86.0
0.10 90.5
0.25 98.0

PKNCA

The paper reports no NCA parameters; Figure S6 shows trough and average concentrations only as box plots. The block below computes the 12-h exposure of the 10-15 kg band at eGFR 130 per regimen, on a 0.25-h subsample of the simulation grid, so the regimens can be compared on peak, average and AUC. The time-zero row is the simulated concentration just after the first bolus.

nca_conc <- pta_sim |>
  dplyr::filter(band == "10-15 kg", !is.na(Cc), abs(time / 0.25 - round(time / 0.25)) < 1e-6) |>
  dplyr::select(id, time, Cc, regimen)
nca_conc <- dplyr::bind_rows(
  nca_conc,
  nca_conc |> dplyr::distinct(id, regimen) |> dplyr::mutate(time = 0, Cc = 0)
) |>
  dplyr::distinct(id, regimen, time, .keep_all = TRUE) |>
  dplyr::arrange(id, regimen, time)

nca_dose <- pta_events |>
  dplyr::filter(band == "10-15 kg", evid == 1, cmt == "central") |>
  dplyr::group_by(id, regimen) |>
  dplyr::slice(1) |>
  dplyr::ungroup() |>
  dplyr::select(id, time, amt, regimen)

conc_obj <- PKNCA::PKNCAconc(nca_conc, Cc ~ time | regimen + id)
dose_obj <- PKNCA::PKNCAdose(nca_dose, amt ~ time | regimen + id)
intervals <- data.frame(start = 0, end = 12, auclast = TRUE, cav = TRUE, cmax = TRUE)
nca_res <- PKNCA::pk.nca(PKNCA::PKNCAdata(conc_obj, dose_obj, intervals = intervals))

nca_summary <- as.data.frame(nca_res$result) |>
  dplyr::filter(PPTESTCD %in% c("auclast", "cav", "cmax")) |>
  dplyr::group_by(regimen, PPTESTCD) |>
  dplyr::summarise(median = stats::median(PPORRES), .groups = "drop") |>
  tidyr::pivot_wider(names_from = PPTESTCD, values_from = median)
nca_summary |>
  dplyr::select(regimen, cmax, cav, auclast) |>
  dplyr::rename(
    Regimen = regimen, `Cmax (mg/L)` = cmax,
    `Cav 0-12 (mg/L)` = cav, `AUC0-12 (mg*h/L)` = auclast
  ) |>
  knitr::kable(digits = 1, caption = "Median simulated 12-h exposure, 10-15 kg, eGFR 130.")
Median simulated 12-h exposure, 10-15 kg, eGFR 130.
Regimen Cmax (mg/L) Cav 0-12 (mg/L) AUC0-12 (mg*h/L)
Infusion 250.6 97.2 1166.0
Infusion + CPB dose 246.5 105.9 1270.7
Standard 521.9 127.9 1534.4
Standard + CPB dose 579.1 177.7 2132.0

# Cav of the infusion arm is set by design near the 100 mg/L target.
cav_infusion <- nca_summary$cav[nca_summary$regimen == "Infusion"]
stopifnot(cav_infusion > 70, cav_infusion < 130)

Assumptions and deviations

  • CPB switch. The control stream turns the circuit on at the start of bypass and off at its end with a single data item and does not split out rewarming. The model uses CPB_ON + CPB_REWARM as the switch, so the circuit is connected whenever the patient is on bypass in either phase; a user with no separate rewarming record sets CPB_ON = 1 for the whole bypass run.
  • eGFR as an input. The control stream computes eGFR from height and serum creatinine (bedside Schwartz, 0.413 x HT / SCr). The model takes that value as the CRCL column, which is how the paper’s own dose calculator and simulations specify renal function.
  • CPB dose timing. The control stream loads the CPB dose into the circuit compartment as its initial amount; the vignette gives it as a bolus into circuit_plasma at the moment of connection, which is equivalent because the circuit is isolated until then.
  • Figure S3 caption. The caption gives the infusion regimen’s CPB dose as 50 mg/kg, while the Methods and Table S1 give 0.5 x loading dose (26.8 mg for the 2 kg child); the vignette uses 26.8 mg, and the resulting curve matches the figure.
  • Virtual population. The paper’s 5000-child virtual population (from its reference 24) is not available. Weight is drawn uniformly within each band, with the smallest band starting at 2 kg (the smallest study patient), and only the eGFR 130 row of Figure 3 is simulated, at 200 children per arm.
  • Attainment grid. Attainment is judged on the minimum concentration over a 0.05-h grid refined to 0.01 h around connection to the circuit; the paper does not state its grid (see the infusion arm under 5 kg above).
  • Digitised values. The comparison values for Figures S2, S3 and 3 were read off the published figures by the maintainers.
  • Errata. No correction notice for this article was found in Europe PMC as of 2026-11-03.