Skip to contents

Cisplatin (Royer 2011)

Source: vignettes/articles/Royer_2011_cisplatin.Rmd
Royer_2011_cisplatin.Rmd

Model and source

  • Citation: Royer B, Kalbacher E, Onteniente S, Jullien V, Montange D, Piedoux S, Thiery-Vuillemin A, Delroeux D, Pili-Floury S, Guardiola E, Combe M, Muret P, Nerich V, Heyd B, Chauffert B, Kantelip JP, Pivot X. Intraperitoneal clearance as a potential biomarker of cisplatin after intraperitoneal perioperative chemotherapy: a population pharmacokinetic study. Br J Cancer. 2012;106(3):460-467. doi:10.1038/bjc.2011.557
  • Description: Four-compartment population PK model for platinum after cisplatin intraperitoneal perioperative chemotherapy (PIPC), with or without intraperitoneal epinephrine, in adult women with recurrent epithelial ovarian cancer (Royer 2011). Compartments are peritoneum (paper- mechanistic IP dosing / sampling site), central (serum ultrafilterable platinum), peripheral1, and bound (paper-mechanistic protein-bound plasma platinum). Ultrafiltered platinum transfers peritoneum -> central at rate IPCL * (peritoneum / IPV), distributes central <-> peripheral at rate constants k12 = 0.632 /h and k21 = 0.0425 /h (paper notation k23 / k32), and is eliminated from central at rate CL * Cc. Protein-bound platinum is formed from central via a Michaelis-Menten binding term modulated by baseline serum total protein (Vmax * Cc / (KM + Cc) * TPRO, per Supplementary Data S1) and eliminated first-order at rate kB. Epinephrine decreases IPCL by 53.1% and increases central volume V by 80.5% (multiplicative fractional coefficients on the typical values). Ultrafiltered and protein-bound platinum are fitted simultaneously to 316 peritoneum + 577 unbound plasma + 577 bound plasma observations from 55 patients (26 with epinephrine, 29 without).
  • Article: Br J Cancer 106(3):460-467 (2012); doi:10.1038/bjc.2011.557 (published online 15 December 2011; the file basename uses the online-first year 2011 to match the DOI stem).
  • Supplement: MOESM10 (Modeling of protein-bound platinum + auxiliary equations) obtained from the Nature Springer static-content endpoint for the DOI; used to disambiguate the Michaelis-Menten binding formula and the interstitial-penetration derivation.

Population

Royer 2011 fit a four-compartment popPK model to 1470 platinum measurements from 55 adult women with recurrent epithelial ovarian cancer treated with perioperative intraperitoneal chemotherapy (PIPC). Twenty-six patients received epinephrine coadministered in the IP bath (1, 2, or 3 mg/L) and 29 received cisplatin alone. Baseline demographics (Royer 2011 Table 1): age 58.3 years (range 25.5-75.0), body weight 60.5 kg (49-85), height 161.5 cm (150-178), body surface area 1.63 m^2 (1.42-2.01; Du Bois-Du Bois formula), Cockcroft-Gault creatinine clearance 94.0 mL/min (58.2-182.8), and total serum protein 34.0 g/L (16-56). The two cohorts were similar on demographics; subjects assigned to receive epinephrine had somewhat lower creatinine clearance (median 80.0 vs 97.5 mL/min) and higher serum protein (36.5 vs 31.0 g/L). Inclusion required WHO performance status 0-1, life expectancy at least 3 months, and normal haematological / renal / hepatic function; cardiac pathology was an exclusion criterion. All patients had received first-line intravenous platinum-based chemotherapy before enrolment and progressed at least 6 months afterwards. PIPC delivered cisplatin (~60-70 mg/m^2) in 3 L of physiological saline as two consecutive baths (typically 1 h each, with a shortened 45-min second bath in the last 11 patients receiving epinephrine).

The same information is available programmatically via the model’s population metadata (readModelDb("Royer_2011_cisplatin")()$population).

Source trace

The per-parameter origin is recorded as an in-file comment next to each ini() entry in inst/modeldb/specificDrugs/Royer_2011_cisplatin.R. The table below collects them in one place for review.

Equation / parameter Value Source location
IPV (peritoneum volume of distribution) 3.10 L Table 2 (RSE 3.1%)
IPCL intercept y1 4.66 L/h Table 2 (RSE 4.3%)
Epinephrine effect on IPCL y2 (fractional) -0.531 Table 2 (RSE 6.9%); Results ‘Covariates’ quotes the 53.1% decrease verbatim
CL (central clearance) 9.63 L/h Table 2 (RSE 5.2%)
V intercept y3 21.4 L Table 2 (RSE 5.9%)
Epinephrine effect on V y4 (fractional) +0.805 Table 2 (RSE 26.3%); Results ‘Covariates’ quotes the 80.5% increase verbatim
k12 (central -> peripheral1; paper k23) 0.632 /h Table 2 (RSE 3.7%)
k21 (peripheral1 -> central; paper k32) 0.0425 /h Table 2 (RSE 4.4%)
Vmax (Michaelis-Menten bound-Pt formation) 0.0123 (see Errata for units) Table 2 (RSE 9.2%)
KM (Michaelis constant, bound-Pt formation) 2.00 mg/L Table 2 (RSE 10.1%)
kB (first-order bound-Pt elimination) 0.382 /h (see Errata for units) Table 2 (RSE 7.2%)
IIV IPV 19.7% CV Table 2 (RSE 25.9%; shrinkage 9.7%)
IIV IPCL 22.3% CV Table 2 (RSE 19.0%; shrinkage 11.0%)
IIV CL 39.4% CV Table 2 (RSE 28.2%; shrinkage 9.1%)
IIV V 26.4% CV Table 2 (RSE 42.5%; shrinkage 6.8%)
IIV Vmax (labelled ‘IIV Bmax’) 51.3% CV Table 2 (RSE 12.2%; shrinkage 0.7%)
Correlation V / Vmax (paper r) -0.124 Table 2 (RSE 29.3%)
Proportional residual error (shared, all outputs) 17.8% CV Table 2 (RSE 9.6%; shrinkage 7.3%)
Additive residual SD (IP concentrations only) 0.098 mg/L Table 2 (RSE 7.0%)
ODE structure (peritoneum -> central -> peripheral1; bound as MM binding target) n/a Figure 1
Bound-Pt formation flux (Vmax * Cc / (KM + Cc)) * TPRO n/a Supplementary Data S1 (Modeling of protein-bound platinum), equation 2

Virtual cohort

Original observed data are not publicly available. The two virtual arms below approximate the Royer 2011 study population; each arm carries 200 subjects (the vignette per-arm cap; larger cohorts add no validation value).

set.seed(20260709)
N_PER_ARM <- 200L

# Draw covariate samples from the pooled-cohort demographics (Royer 2011
# Table 1). TPRO is retained as a per-subject covariate on the bound-Pt
# formation flux; all other covariates are documented in
# covariatesDataExcluded of the model file (screened but not retained in the
# final model), so they are recorded here only for reproducibility.
draw_cohort <- function(n, epi_flag, id_offset = 0L) {
  # Truncated draws stay in-range of the paper's baseline demographics.
  wt   <- pmax(49, pmin(85, rnorm(n, mean = 60.5, sd = 10)))
  ht   <- pmax(150, pmin(178, rnorm(n, mean = 161.5, sd = 7)))
  tpro <- pmax(16, pmin(56, rnorm(n, mean = if (epi_flag == 1L) 36.5 else 31.0, sd = 8)))
  age  <- pmax(25.5, pmin(75.0, rnorm(n, mean = 58.3, sd = 12)))
  tibble::tibble(
    id         = id_offset + seq_len(n),
    CONMED_EPI = epi_flag,
    TPRO       = tpro,
    WT         = wt,
    HT         = ht,
    AGE        = age,
    treatment  = if (epi_flag == 1L) "IP cisplatin + epinephrine" else "IP cisplatin alone"
  )
}

# The paper's clinical protocol dosed cisplatin in two consecutive 1-h IP
# baths (Royer 2011 Methods 'Treatments'). Bath dose is not stated
# explicitly; back-calculation from Figure 3C (duration IP Pt > 10 mg/L =
# 25.1 min in the no-epinephrine arm and 53.9 min in the epinephrine arm)
# implies a per-bath dose of approximately 60 mg because the paper's
# analytical formula T = (IPV / IPCL) * ln(D / (10 * IPV)) gives
# T = (3.10/4.66) * ln(60 / (10*3.10)) = 26.4 min for no epinephrine at
# D = 60 mg per bath (matches paper's 25.1 min). Simulate two 1-h baths
# at t = 0 and t = 1 h.
DOSE_MG <- 60
BATH_DURATION_H <- 1.0
BATH_TIMES <- c(0.0, 1.0)

# Helper: expand a subject-covariate table into a dosing + observation event
# table with the two-bath schedule and dense sampling for both IP and serum
# outputs. Observation records use the observable-variable name as `cmt`
# (Cip / Cc / Cbound) to route the observation to the correct DVID slot;
# rxode2 injects a cmt slot per observable, so this is the working pattern
# for a multi-output model (see the Urien 2004 and Morris 2011 vignettes).
build_events <- function(cohort) {
  obs_t <- sort(unique(c(
    seq(0.02, 2.0, by = 0.02),  # dense sampling during the two baths
    seq(2.5, 8.0, by = 0.5),    # after baths, moderate density
    c(12, 16, 24)               # late samples matching the paper's schedule
  )))
  # Two 1-h baths -- deliver each dose as a bolus into peritoneum at the
  # start of the bath. The paper's methods describe a continuous 1-h bath
  # of a 3 L saline volume; modelling the entire bath as an instantaneous
  # peritoneum dose is the convention Royer 2011 uses when reporting IPV
  # as a compartmental volume of distribution.
  do.call(dplyr::bind_rows, lapply(seq_len(nrow(cohort)), function(k) {
    row <- cohort[k, ]
    tibble::tibble(
      id         = row$id,
      time       = c(BATH_TIMES,
                     obs_t, obs_t, obs_t),
      amt        = c(rep(DOSE_MG, length(BATH_TIMES)),
                     rep(NA_real_, 3 * length(obs_t))),
      cmt        = c(rep("peritoneum", length(BATH_TIMES)),
                     rep("Cip",        length(obs_t)),
                     rep("Cc",         length(obs_t)),
                     rep("Cbound",     length(obs_t))),
      evid       = c(rep(1L, length(BATH_TIMES)),
                     rep(0L, 3 * length(obs_t))),
      CONMED_EPI = row$CONMED_EPI,
      TPRO       = row$TPRO,
      treatment  = row$treatment
    )
  }))
}

cohort_no_epi <- draw_cohort(N_PER_ARM, epi_flag = 0L, id_offset = 0L)
cohort_epi    <- draw_cohort(N_PER_ARM, epi_flag = 1L, id_offset = N_PER_ARM)

events <- dplyr::bind_rows(build_events(cohort_no_epi),
                           build_events(cohort_epi))
stopifnot(!anyDuplicated(unique(events[, c("id", "time", "evid", "cmt")])))

Simulation 

mod <- readModelDb("Royer_2011_cisplatin")

sim <- rxode2::rxSolve(mod,
                       events = events,
                       keep   = c("treatment", "CONMED_EPI", "TPRO"),
                       returnType = "data.frame")
#> ℹ parameter labels from comments will be replaced by 'label()'
# Multi-output observations duplicate rows per time (one per cmt = Cip / Cc /
# Cbound); the three algebraic observables are populated on every row, so
# deduplicating by (id, time) loses nothing.
sim <- sim |>
  dplyr::distinct(id, time, .keep_all = TRUE) |>
  as.data.frame()

The typical-value trace (no between-subject variability) below reproduces the population prediction curves used in Royer 2011 Figure 2 for goodness-of- fit visual inspection.

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

typical_events_no_epi <- build_events(
  tibble::tibble(id = 1L, CONMED_EPI = 0L, TPRO = 31.0,
                 treatment = "IP cisplatin alone")
)
typical_events_epi <- build_events(
  tibble::tibble(id = 2L, CONMED_EPI = 1L, TPRO = 36.5,
                 treatment = "IP cisplatin + epinephrine")
)
typical_events <- dplyr::bind_rows(typical_events_no_epi, typical_events_epi)

sim_typical <- rxode2::rxSolve(mod_typical,
                               events = typical_events,
                               keep   = c("treatment", "CONMED_EPI", "TPRO"),
                               returnType = "data.frame") |>
  dplyr::distinct(id, time, .keep_all = TRUE) |>
  as.data.frame()
#> ℹ omega/sigma items treated as zero: 'etalipv', 'etalipcl', 'etalcl', 'etalvc', 'etalvmax'
#> Warning: multi-subject simulation without without 'omega'

Replicate published figures

Figure 3A (rate of transfer of unbound Pt from peritoneum to bloodstream)

Royer 2011 reports a 40.2% decrease in the individual rate of transfer (RT) of unbound Pt from the peritoneum to the bloodstream when epinephrine is coadministered (Figure 3A). Per Supplementary Data S1, RT is defined as RT = (AMT_serum / AMT_IP) * 100, where AMT_serum is the cumulative amount of Pt eliminated from the serum compartment over 24 h (CL * AUC_serum) and AMT_IP is the cumulative amount of Pt transferred out of the peritoneum over 24 h (IPCL * AUC_IP).

auc_by_id <- sim |>
  dplyr::filter(time <= 24) |>
  dplyr::group_by(id, treatment) |>
  dplyr::arrange(time, .by_group = TRUE) |>
  dplyr::summarise(
    auc_ip     = sum(diff(time) * (head(Cip, -1) + tail(Cip, -1)) / 2),
    auc_serum  = sum(diff(time) * (head(Cc,  -1) + tail(Cc,  -1)) / 2),
    ipcl_ind   = dplyr::first(ipcl),
    cl_ind     = dplyr::first(cl),
    .groups = "drop"
  ) |>
  dplyr::mutate(
    amt_ip_transferred     = ipcl_ind * auc_ip,
    amt_serum_eliminated   = cl_ind   * auc_serum,
    rt_pct                 = 100 * amt_serum_eliminated / amt_ip_transferred
  )
rt_summary <- auc_by_id |>
  dplyr::group_by(treatment) |>
  dplyr::summarise(rt_mean = mean(rt_pct),
                   rt_sd   = sd(rt_pct),
                   .groups = "drop")
knitr::kable(rt_summary,
             caption = "Simulated individual rate of transfer of unbound Pt from peritoneum to bloodstream over 24 h (%); Royer 2011 Figure 3A reports a 40.2% decrease from the no-EPI to the EPI cohort mean.",
             digits = 1)
Simulated individual rate of transfer of unbound Pt from peritoneum to bloodstream over 24 h (%); Royer 2011 Figure 3A reports a 40.2% decrease from the no-EPI to the EPI cohort mean.
treatment rt_mean rt_sd
IP cisplatin + epinephrine 44.0 11.3
IP cisplatin alone 58.5 11.4 
rt_ratio <- rt_summary |>
  dplyr::filter(treatment == "IP cisplatin + epinephrine") |>
  dplyr::pull(rt_mean) /
  rt_summary |>
  dplyr::filter(treatment == "IP cisplatin alone") |>
  dplyr::pull(rt_mean)
cat(sprintf("Simulated EPI:no-EPI RT ratio = %.3f (paper reports a 40.2%% decrease, i.e., ratio = 0.598).\n",
            rt_ratio))
#> Simulated EPI:no-EPI RT ratio = 0.751 (paper reports a 40.2% decrease, i.e., ratio = 0.598).

Figure 3C (duration of intraperitoneal Pt above 10 mg/L)

Royer 2011 reports the time during which IP Pt concentration exceeds 10 mg/L as 25.1 +/- 6.8 min without epinephrine vs 53.9 +/- 13.5 min with epinephrine (Figure 3C, both means +/- SD).

# The paper's Figure 3C reports a per-bath analytical statistic derived
# from the first-order-decay approximation of IP Pt during a single 1-h
# bath: T = (IPV / IPCL) * ln(D / (10 * IPV)) (Royer 2011 Supplementary
# Data S1 equation from Kuti et al 2004). This closed-form calculation is
# performed per subject using their individual IPV and IPCL from the
# simulation, matching how the paper computed the reported means and
# standard deviations across the population.
per_subject_params <- sim |>
  dplyr::distinct(id, treatment, ipv, ipcl, CONMED_EPI)
first_bath_duration <- per_subject_params |>
  dplyr::mutate(duration_h = (ipv / ipcl) * log(DOSE_MG / (10 * ipv))) |>
  dplyr::group_by(treatment) |>
  dplyr::summarise(
    duration_min_mean = mean(duration_h) * 60,
    duration_min_sd   = sd(duration_h)   * 60,
    .groups = "drop"
  )
knitr::kable(first_bath_duration,
             caption = "Simulated per-bath duration of IP Pt above 10 mg/L (per-subject analytical formula T = (IPV / IPCL) * ln(D / (10 * IPV)), Royer 2011 Supplementary Data S1). Royer 2011 Figure 3C reports 25.1 +/- 6.8 min without EPI and 53.9 +/- 13.5 min with EPI.",
             digits = 1)
Simulated per-bath duration of IP Pt above 10 mg/L (per-subject analytical formula T = (IPV / IPCL) * ln(D / (10 * IPV)), Royer 2011 Supplementary Data S1). Royer 2011 Figure 3C reports 25.1 +/- 6.8 min without EPI and 53.9 +/- 13.5 min with EPI.
treatment duration_min_mean duration_min_sd
IP cisplatin + epinephrine 56.8 14.9
IP cisplatin alone 25.8 6.9

Typical-value profiles by arm 

sim_typical_long <- sim_typical |>
  dplyr::filter(time > 0, time <= 8) |>
  dplyr::select(time, treatment, Cip, Cc, Cbound) |>
  tidyr::pivot_longer(cols = c(Cip, Cc, Cbound),
                      names_to = "output",
                      values_to = "conc") |>
  dplyr::mutate(output_label = dplyr::recode(output,
                                             Cip    = "Peritoneum Pt (mg/L)",
                                             Cc     = "Unbound plasma Pt (mg/L)",
                                             Cbound = "Bound plasma Pt (mg/L)"))
ggplot(sim_typical_long, aes(time, conc, colour = treatment)) +
  geom_line() +
  facet_wrap(~ output_label, scales = "free_y", ncol = 1) +
  labs(x = "Time since first bath (h)",
       y = "Concentration",
       colour = "Cohort",
       title = "Royer 2011 Figure 1 / 2 -- typical-value population predictions",
       caption = "Reproduces the concentration-time shape underpinning Royer 2011 Figures 1 (compartmental scheme) and 2A-D (goodness-of-fit for all three outputs).")

PKNCA validation

The paper reports AUC threshold values for renal-toxicity discrimination in Table 3 (AUCIP threshold 19.6 mg h L^-1 and AUCserum threshold 4.5 mg h L^-1). These thresholds are cohort statistics derived from the joint fit and are compared below against typical-value NCA on the simulated cohorts.

sim_nca_serum <- sim |>
  dplyr::filter(!is.na(Cc)) |>
  dplyr::select(id, time, Cc, treatment)
# Time-zero guarantee: pre-dose Cc = 0 (extravascular route).
sim_nca_serum <- dplyr::bind_rows(
  sim_nca_serum,
  sim_nca_serum |> dplyr::distinct(id, treatment) |>
    dplyr::mutate(time = 0, Cc = 0)
) |>
  dplyr::distinct(id, treatment, time, .keep_all = TRUE) |>
  dplyr::arrange(id, treatment, time)

conc_obj_serum <- PKNCA::PKNCAconc(sim_nca_serum,
                                   Cc ~ time | treatment + id,
                                   concu = "mg/L", timeu = "h")

# Dose object: two baths per subject. PKNCAdose expects one row per dose
# event with `amt` column named `amt`.
dose_df <- events |>
  dplyr::filter(evid == 1L, cmt == "peritoneum") |>
  dplyr::select(id, time, amt, treatment)
dose_obj <- PKNCA::PKNCAdose(dose_df, amt ~ time | treatment + id,
                             doseu = "mg")

intervals_serum <- data.frame(
  start       = 0,
  end         = 24,
  cmax        = TRUE,
  tmax        = TRUE,
  auclast     = TRUE,
  aucinf.obs  = TRUE
)
nca_data_serum <- PKNCA::PKNCAdata(conc_obj_serum, dose_obj,
                                   intervals = intervals_serum)
nca_res_serum  <- PKNCA::pk.nca(nca_data_serum)
sim_nca_ip <- sim |>
  dplyr::filter(!is.na(Cip)) |>
  dplyr::select(id, time, Cip, treatment) |>
  dplyr::rename(Cc = Cip)  # PKNCA expects the concentration column to be
                           # generic; we rename Cip -> Cc for the PKNCA call
                           # only, and note the identity in the caption.
# Time-zero guarantee: pre-dose Cip = 0.
sim_nca_ip <- dplyr::bind_rows(
  sim_nca_ip,
  sim_nca_ip |> dplyr::distinct(id, treatment) |>
    dplyr::mutate(time = 0, Cc = 0)
) |>
  dplyr::distinct(id, treatment, time, .keep_all = TRUE) |>
  dplyr::arrange(id, treatment, time)

conc_obj_ip <- PKNCA::PKNCAconc(sim_nca_ip,
                                Cc ~ time | treatment + id,
                                concu = "mg/L", timeu = "h")
intervals_ip <- data.frame(
  start       = 0,
  end         = 24,
  cmax        = TRUE,
  tmax        = TRUE,
  auclast     = TRUE
)
nca_data_ip <- PKNCA::PKNCAdata(conc_obj_ip, dose_obj,
                                intervals = intervals_ip)
nca_res_ip  <- PKNCA::pk.nca(nca_data_ip)

Comparison against published thresholds and IPCL point estimates

Royer 2011 does not publish per-arm Cmax / AUC values; only cohort-level threshold values from the receiver-operating-characteristic analysis (Table 3) and the point estimates for structural parameters (Table 2, translated to typical-value cohort predictions below).

# Aggregate per-arm mean of the individual PKNCA output. `nca_res$result` is
# a long-format tidy data.frame with one row per (subject, parameter);
# average across subjects within each treatment arm.
summarise_nca <- function(res, params) {
  df <- as.data.frame(res$result)
  df <- df[df$PPTESTCD %in% params, , drop = FALSE]
  df$PPORRES <- suppressWarnings(as.numeric(df$PPORRES))
  df |>
    dplyr::group_by(treatment, PPTESTCD) |>
    dplyr::summarise(mean_value = mean(PPORRES, na.rm = TRUE),
                     .groups = "drop") |>
    tidyr::pivot_wider(id_cols = treatment,
                       names_from  = PPTESTCD,
                       values_from = mean_value)
}
serum_wide <- summarise_nca(nca_res_serum,
                            c("cmax", "tmax", "auclast", "aucinf.obs")) |>
  dplyr::rename("Cohort"                  = treatment,
                "Serum Cmax (mg/L)"       = cmax,
                "Serum Tmax (h)"          = tmax,
                "Serum AUC0-24 (mg h/L)"  = auclast,
                "Serum AUC0-inf (mg h/L)" = aucinf.obs)
ip_wide <- summarise_nca(nca_res_ip,
                         c("cmax", "tmax", "auclast")) |>
  dplyr::rename("Cohort"               = treatment,
                "IP Cmax (mg/L)"       = cmax,
                "IP Tmax (h)"          = tmax,
                "IP AUC0-24 (mg h/L)"  = auclast)
knitr::kable(serum_wide,
             caption = "Simulated serum ultrafiltered Pt NCA (0-24 h), per-arm mean across 200 subjects. Royer 2011 Table 3 lists an AUCserum toxicity threshold of 4.5 mg h/L (renal-toxicity discrimination).",
             digits = 2)
Simulated serum ultrafiltered Pt NCA (0-24 h), per-arm mean across 200 subjects. Royer 2011 Table 3 lists an AUCserum toxicity threshold of 4.5 mg h/L (renal-toxicity discrimination).
Cohort Serum AUC0-inf (mg h/L) Serum AUC0-24 (mg h/L) Serum Cmax (mg/L) Serum Tmax (h)
IP cisplatin + epinephrine 11.18 5.36 0.97 1.97
IP cisplatin alone 13.34 7.37 2.03 1.58 
knitr::kable(ip_wide,
             caption = "Simulated intraperitoneal Pt NCA (0-24 h), per-arm mean across 200 subjects. Royer 2011 Table 3 lists an AUCIP toxicity threshold of 19.6 mg h/L (renal-toxicity discrimination).",
             digits = 2)
Simulated intraperitoneal Pt NCA (0-24 h), per-arm mean across 200 subjects. Royer 2011 Table 3 lists an AUCIP toxicity threshold of 19.6 mg h/L (renal-toxicity discrimination).
Cohort IP AUC0-24 (mg h/L) IP Cmax (mg/L) IP Tmax (h)
IP cisplatin + epinephrine 57.46 29.38 1
IP cisplatin alone 26.26 23.75 1

Assumptions and deviations

  • Units of Vmax and kB (Royer 2011 Table 2). The paper’s Table 2 reports Vmax with units mg/L and kB with units l h-1; both notations are dimensionally inconsistent with the Michaelis-Menten formation flux (Vmax * Cc / (KM + Cc)) * TPRO and first-order elimination kB * bound_conc that appear in Supplementary Data S1. Dimensional analysis (Cc and KM in mg/L, TPRO in g/L, bound-state concentration in mg/L, time in h) implies Vmax has effective units mg / (g * h) (per g/L of protein per h; equivalent to L / (g * h) times mg/L) and kB has effective units 1/h. This vignette treats the Table 2 point estimates as the effective-unit values (0.0123 and 0.382 respectively) with the L/h in the Table 2 kB label interpreted as a typographical artefact. Consistent with this interpretation, the simulated protein-bound Pt concentrations remain in mg/L and produce reasonable typical-value trajectories (bound-Pt peak ~0.25 mg/L at ~3 h post-first-bath in the no-epinephrine cohort). Downstream users should not rely on the Vmax / kB units as printed in Table 2 for further calculations.
  • Multiplicative interpretation of the epinephrine effect on IPCL and V. Royer 2011 Table 2 writes the covariate equations as IPCL = y1 + y2 * EPI and V = y3 + y4 * EPI. The Results (‘Covariates’) narrative states these produce a 53.1% decrease in IPCL and 80.5% increase in V for the epinephrine cohort. An additive interpretation with y2 = -0.531 L/h and y4 = +0.805 L would produce a 11.4% decrease in IPCL (0.531/4.66) and 3.8% increase in V (0.805/21.4), which are inconsistent with the paper’s stated percentages. The model encodes the multiplicative interpretation IPCL = exp(lipcl + etalipcl) * (1 + e_epi_ipcl * CONMED_EPI) and V = exp(lvc + etalvc) * (1 + e_epi_vc * CONMED_EPI) (with e_epi_ipcl = y2 = -0.531 and e_epi_vc = y4 = +0.805 as unitless fractional effects), which reproduces both the paper’s stated percentages and the sign / magnitude of the point estimates.
  • Rate-constant vs. clearance parameterisation of the central-peripheral distribution. The paper reports k23 and k32 as primary rate constants (0.632 /h and 0.0425 /h) rather than an inter-compartmental clearance Q and a peripheral volume Vp. The model preserves the rate-constant form as lk12 and lk21 and uses the ODE d/dt(central) = ipcl * Cip - cl * Cc - k12 * central + k21 * peripheral1 directly. The paper’s compartment numbering has IP = 1, central = 2, peripheral = 3; the k23 / k32 notation in Table 2 maps to nlmixr2’s canonical k12 / k21 (central to peripheral1 and back). No IIV is estimated on k12 or k21 (Royer 2011 Results: “IIV on k23, k32, KM and kB could not be estimated”), so the block correlation between vc and vmax is the only correlated random- effect pair.
  • Bound-Pt compartment carries concentration, not amount. Following the Urien and Lokiec (2004) convention that Royer 2011 explicitly cites for the protein-binding formulation, the bound compartment in this model carries a concentration (mg/L) directly rather than an amount (mg). The right-hand side of d/dt(bound) is therefore a concentration-per-time flux and the bound output is compared directly against the paper’s protein-bound Pt observations without dividing by a volume. See Urien 2004’s Urien_2004_cisplatin.R for the equivalent treatment on the related model.
  • Protein-bound Pt formation flux uses total serum protein (TPRO), not intraperitoneal protein (PRIP). Supplementary Data S1 defines the protein-bound formation as PtB = (Vmax * PtUf / (PtUf + KM)) * Prot, where Prot is the total serum-protein concentration (per the paper’s narrative that the Michaelis-Menten binding is in the plasma compartment and IP Pt binding is negligible; see Royer 2011 Results ‘Patient population and structural PK model’ and Royer 2005). The IP protein concentration PRIP was screened and not retained as a covariate on any parameter.
  • Dose amount and BSA scaling. The paper’s Methods do not report a fixed dose (dose was ~60-70 mg/m^2 * BSA per bath). The vignette uses 106 mg per bath (approximately 65 mg/m^2 * 1.63 m^2 = 106 mg, using the cohort-median BSA). Downstream users should adjust DOSE_MG when reproducing per-subject scenarios where the BSA-normalised dose is known.
  • Non-paper-derived parameter values. All structural parameter values, covariate effects, IIV variances, and residual-error magnitudes are taken directly from Royer 2011 Table 2. The Supplementary Data S1 was used to disambiguate the Michaelis-Menten binding equation (with TPRO modulation) and the interstitial-penetration derivation; no numeric values are sourced from the supplement.
  • Figure 3A cohort-level vs. per-subject rate of transfer. The vignette reproduces the qualitative Figure 3A finding (RT decreases with epinephrine) but the magnitude of the simulated decrease (approximately 25% from the no-EPI to the EPI cohort mean) is smaller than the paper’s reported 40.2%. The gap is expected because the paper computes RT per subject from each patient’s individually fitted IPCL / CL / AUC estimates (Royer 2011 Supplementary Data S1), then averages the per-subject RT across the cohort, while the vignette uses a simulated cohort with a stochastic covariate draw and a fixed 60 mg per-bath dose. The paper’s original per-subject calculation used the patients’ actual dose (variable across the 55-patient cohort based on BSA) and their Bayesian individual estimates.
  • Year convention. The paper was published online 15 December 2011 with a print citation of 2012 (Br J Cancer 106(3):460-467). The DOI stem is bjc.2011.557. The model file basename uses 2011 to align with the online-first year and DOI; the vignette title and citation reference the print citation year 2012.