Skip to contents

Model and source

  • Citation: Jiang C, Ren F, Zhang M, Lu Q, Zeng S, Yang G, Zhu Y. Using pharmacokinetic and pharmacodynamic analysis to optimize the dosing regimens of fanastomig (EMB-02) in patients with advanced solid tumors. CPT Pharmacometrics Syst Pharmacol. 2025;14(5):975-986. doi:10.1002/psp4.70011.
  • Description: Two-compartment population PK model for fanastomig (EMB-02; anti-PD-1 x LAG-3 bispecific antibody) with sigmoid I_max time-dependent clearance and an albumin effect on central volume, plus a direct Emax model of peripheral-blood PD-1 receptor occupancy, in adults with advanced solid tumours (Jiang 2025)
  • Article: https://doi.org/10.1002/psp4.70011 (open access; PMC12072225)

Fanastomig (EMB-02) is a bispecific antibody against PD-1 and LAG-3. Jiang 2025 describes its serum pharmacokinetics in the first-in-human phase 1 study (NCT04618393) with a two-compartment model whose clearance falls over time following a sigmoid I_max function of time since the first dose, with serum albumin as the only retained covariate. A direct, hyperbolic Emax model links serum concentration to PD-1 receptor occupancy (RO) on peripheral-blood CD3+ T cells. The final Phoenix NLME model code is printed in the paper’s Data S1.

Population

Forty-seven adults with advanced solid tumours enrolled in China, the United States and Australia contributed PK samples. Twelve patients who had received a PD-1 inhibitor within the previous 4 months had elevated predose anti-PD-(L)1 concentrations and were excluded, so the final model was fitted to 625 concentrations from 35 patients. Table S1 summarises the 47 screened patients: median age 61 years (37-83), median weight 72.0 kg (42.1-107), 59.6% female, 72.3% White, 21.3% Asian, 4.3% Black; median albumin 38.0 g/L (23.0-49.6). Fanastomig was given intravenously every week at 6, 20, 60, 180, 360, 600 or 900 mg (dose escalation) and at 60, 180 or 600 mg (dose enrichment). ADA incidence was 95.7% (44/46); ADA titre did not change clearance materially.

The same information is available programmatically:

str(readModelDb("Jiang_2025_fanastomig")()$population)
#> List of 15
#>  $ species         : chr "human"
#>  $ n_subjects      : int 35
#>  $ n_studies       : int 1
#>  $ age_range       : chr "37-83 years"
#>  $ age_median      : chr "61 years"
#>  $ weight_range    : chr "42.1-107 kg"
#>  $ weight_median   : chr "72.0 kg"
#>  $ sex_female_pct  : num 59.6
#>  $ race_ethnicity  : Named num [1:4] 72.3 21.3 4.3 2.1
#>   ..- attr(*, "names")= chr [1:4] "White" "Asian" "Black" "Other"
#>  $ disease_state   : chr "Advanced solid tumours (first-in-human phase 1, NCT04618393)"
#>  $ dose_range      : chr "6, 20, 60, 180, 360, 600 and 900 mg IV QW (Part 1 dose escalation); 60, 180 and 600 mg IV QW (Part 2 dose enrichment)"
#>  $ regions         : chr "China, United States, Australia"
#>  $ albumin_median  : chr "38.0 g/L (range 23.0-49.6)"
#>  $ ada_positive_pct: num 95.7
#>  $ notes           : chr "Demographics are Jiang 2025 Table S1 (all 47 patients with PK data). Twelve patients pretreated with a PD-1 inh"| __truncated__

Source trace

Equation / parameter Value Source location
lcl (CL at t = 0) log(0.215 L/h) Table 1
lvc (Vc at ALB = 36 g/L) log(5.20 L) Table 1
lq (Q) log(0.0463 L/h) Table 1
lvp (Vp) log(3.15 L) Table 1
lcl_time_max (log abs(Imax)) log(0.954); Imax = -0.954 Table 1
lcl_t50 (T50) log(180 h) Table 1
lcl_time_hill (Hill) log(8.49) Table 1
e_alb_vc -0.924 Table 1 (‘Effect of ALB’); applied to the central volume per Data S1
etalvc, etalcl, etalcl_time_max, etalcl_t50 0.035, 0.068, 0.416, 0.394 Table 1 (variances); diagonal per Data S1 ranef(diag(...))
propSd 0.378 Table 1 (Prop sigma); Data S1 CEps = 0.378
lec50 log(0.084 mg/L) Abstract; Results, ‘Receptor Occupancy’
lemax log(92.5 %) Not printed; digitised from the Figure 3 median curve (see Assumptions)
CL(t) = CL0 exp(Imax t^Hill / (t^Hill + T50^Hill)) n/a Equation 1; Data S1 Cl = Cl0*exp(Imax*t^Hill/(t^Hill+T50^Hill))
Vc = tvV (ALB/36)^e_alb_vc exp(eta) n/a Data S1 stparm(V = tvV *(ALB/36)^dVdALB * exp(nV))
Two-compartment ODEs n/a Data S1 deriv(A1 = ...), deriv(A2 = ...)
Cobs = C (1 + eps) n/a Equation 4; Data S1 observe(CObs = C * (1 + CEps))
RO = Emax C / (EC50 + C) n/a Methods ‘Receptor Occupancy Assessment’; Results (0.756 ug/mL = 9 x EC50 gives 90 % of maximum)

Typical-value identities

These follow from the parameters alone and pin the published statements exactly.

mod <- readModelDb("Jiang_2025_fanastomig")
ui <- rxode2::rxode2(mod)
#> ℹ parameter labels from comments will be replaced by 'label()'
th <- ui$theta

imax <- -exp(th[["lcl_time_max"]])
ec50 <- exp(th[["lec50"]])
emax <- exp(th[["lemax"]])

# Results: 'a maximum reduction of 61.5%' in CL over time.
max_reduction <- 1 - exp(imax)
# Results: 0.756 ug/mL maintains 90 % of maximal peripheral RO; 3-fold for tumour.
c90 <- 9 * ec50
ro_c90_frac <- (emax * c90 / (ec50 + c90)) / emax

data.frame(
  quantity = c("Maximum CL reduction (%)", "C for 90 % of max RO (ug/mL)",
               "Tumour target, 3 x C90 (ug/mL)", "RO at C90 / Emax"),
  model = signif(c(100 * max_reduction, c90, 3 * c90, ro_c90_frac), 4),
  published = c(61.5, 0.756, 2.27, 0.90)
) |>
  knitr::kable()
quantity model published
Maximum CL reduction (%) 61.480 61.500
C for 90 % of max RO (ug/mL) 0.756 0.756
Tumour target, 3 x C90 (ug/mL) 2.268 2.270
RO at C90 / Emax 0.900 0.900

stopifnot(
  abs(100 * max_reduction - 61.5) < 0.05,
  abs(c90 - 0.756) < 1e-9,
  abs(3 * c90 - 2.27) < 0.005,
  abs(ro_c90_frac - 0.9) < 1e-12
)

Virtual cohort

The paper simulated 1000 patients per regimen from posthoc estimates; the observed data are not public. The cohort below draws the random effects from the model and albumin from a normal distribution centred on the Table S1 median (38 g/L, SD 5 g/L, truncated to the observed 23-49.6 g/L range).

Three regimens are simulated, 200 patients each: 360 mg QW (Figure 4), 180 mg QW for 8 doses followed by 360 mg Q2W, and 300 mg QW for 8 doses followed by 600 mg Q2W (Table S3, Figure S6). Infusions last 1 hour.

set.seed(20250311)
rxode2::rxSetSeed(20250311)

n_per_arm <- 200L
tau_end <- 2688 # h; end of the 12th dosing interval of the Q2W regimens

regimens <- list(
  "360 mg QW" = data.frame(time = (0:15) * 168, amt = 360),
  "180 mg QW x8 then 360 mg Q2W" = data.frame(
    time = c((0:7) * 168, 1344 + (0:3) * 336), amt = c(rep(180, 8), rep(360, 4))
  ),
  "300 mg QW x8 then 600 mg Q2W" = data.frame(
    time = c((0:7) * 168, 1344 + (0:3) * 336), amt = c(rep(300, 8), rep(600, 4))
  )
)

obs_times <- sort(unique(c(seq(0, tau_end, by = 6), 840, 2688)))

make_cohort <- function(n, doses, regimen, id_offset) {
  ids <- id_offset + seq_len(n)
  alb <- pmin(pmax(rnorm(n, 38, 5), 23), 49.6)
  dose_rows <- tidyr::expand_grid(id = ids, doses) |>
    dplyr::mutate(evid = 1L, cmt = "central", dur = 1)
  obs_rows <- tidyr::expand_grid(id = ids, time = obs_times) |>
    dplyr::mutate(amt = 0, evid = 0L, cmt = "central", dur = 0)
  dplyr::bind_rows(dose_rows, obs_rows) |>
    dplyr::left_join(data.frame(id = ids, ALB = alb), by = "id") |>
    dplyr::mutate(regimen = regimen) |>
    dplyr::arrange(id, time, dplyr::desc(evid))
}

events <- dplyr::bind_rows(lapply(seq_along(regimens), function(i) {
  make_cohort(n_per_arm, regimens[[i]], names(regimens)[i],
              id_offset = (i - 1L) * n_per_arm)
}))
stopifnot(!anyDuplicated(unique(events[, c("id", "time", "evid")])))

Simulation

sim <- rxode2::rxSolve(mod, events = events, keep = "regimen") |>
  as.data.frame() |>
  dplyr::mutate(regimen = factor(regimen, levels = names(regimens)))
#> ℹ parameter labels from comments will be replaced by 'label()'

Trough concentrations (Table S3)

Table S3 reports the mean and the 10th-50th percentiles of the simulated trough after the 5th and the 12th dose. “Trough after the k-th dose” is taken as the concentration at the end of the k-th dosing interval: 840 h for the 5th weekly dose, and 2688 h for the 12th dose of the Q2W regimens (by then the Q2W regimens are at steady state, so the troughs before and after the 12th dose agree).

Two Table S3 cells are misprinted: the 5th-dose P50 of the two loading regimens repeats their P10 (1.05 and 1.75 ug/mL), which cannot be a median that sits below the P30 printed beside it. They are excluded from the comparison.

published_s3 <- tibble::tribble(
  ~regimen,                         ~dose_no, ~mean, ~P10, ~P20, ~P30, ~P50,
  "360 mg QW",                      5,        19.46, 2.19, 3.53, 5.10, 9.03,
  "180 mg QW x8 then 360 mg Q2W",   5,         9.51, 1.05, 1.72, 2.37, NA,
  "180 mg QW x8 then 360 mg Q2W",   12,       11.46, 0.34, 0.60, 0.91, 2.06,
  "300 mg QW x8 then 600 mg Q2W",   5,        16.29, 1.75, 2.75, 4.02, NA,
  "300 mg QW x8 then 600 mg Q2W",   12,       19.99, 0.55, 0.98, 1.55, 3.67
)

trough_time <- c(`5` = 840, `12` = 2688)

sim_s3 <- published_s3 |>
  dplyr::select(regimen, dose_no) |>
  dplyr::rowwise() |>
  dplyr::mutate(
    ctr = list(sim$Cc[sim$regimen == regimen & sim$time == trough_time[[as.character(dose_no)]]])
  ) |>
  dplyr::ungroup()
stopifnot(all(lengths(sim_s3$ctr) == n_per_arm))

sim_s3 <- sim_s3 |>
  dplyr::mutate(
    mean = vapply(ctr, mean, numeric(1)),
    P10 = vapply(ctr, quantile, numeric(1), probs = 0.1),
    P20 = vapply(ctr, quantile, numeric(1), probs = 0.2),
    P30 = vapply(ctr, quantile, numeric(1), probs = 0.3),
    P50 = vapply(ctr, quantile, numeric(1), probs = 0.5)
  ) |>
  dplyr::select(-ctr)

cmp_s3 <- dplyr::inner_join(
  tidyr::pivot_longer(published_s3, mean:P50, names_to = "statistic", values_to = "published"),
  tidyr::pivot_longer(sim_s3, mean:P50, names_to = "statistic", values_to = "simulated"),
  by = c("regimen", "dose_no", "statistic")
) |>
  dplyr::filter(!is.na(published)) |>
  dplyr::mutate(pct_diff = 100 * (simulated / published - 1))

cmp_s3 |>
  dplyr::mutate(simulated = signif(simulated, 3), pct_diff = round(pct_diff)) |>
  dplyr::rename(
    "Regimen" = regimen, "Dose no." = dose_no, "Statistic" = statistic,
    "Published (ug/mL)" = published, "Simulated (ug/mL)" = simulated,
    "Difference (%)" = pct_diff
  ) |>
  knitr::kable(caption = "Simulated vs. Table S3 trough concentrations (200 patients per regimen vs. 1000 in the paper).")
Simulated vs. Table S3 trough concentrations (200 patients per regimen vs. 1000 in the paper).
Regimen Dose no. Statistic Published (ug/mL) Simulated (ug/mL) Difference (%)
360 mg QW 5 mean 19.46 17.100 -12
360 mg QW 5 P10 2.19 1.950 -11
360 mg QW 5 P20 3.53 3.060 -13
360 mg QW 5 P30 5.10 3.980 -22
360 mg QW 5 P50 9.03 7.040 -22
180 mg QW x8 then 360 mg Q2W 5 mean 9.51 8.790 -8
180 mg QW x8 then 360 mg Q2W 5 P10 1.05 0.962 -8
180 mg QW x8 then 360 mg Q2W 5 P20 1.72 1.450 -16
180 mg QW x8 then 360 mg Q2W 5 P30 2.37 2.070 -13
180 mg QW x8 then 360 mg Q2W 12 mean 11.46 9.560 -17
180 mg QW x8 then 360 mg Q2W 12 P10 0.34 0.292 -14
180 mg QW x8 then 360 mg Q2W 12 P20 0.60 0.484 -19
180 mg QW x8 then 360 mg Q2W 12 P30 0.91 0.775 -15
180 mg QW x8 then 360 mg Q2W 12 P50 2.06 1.960 -5
300 mg QW x8 then 600 mg Q2W 5 mean 16.29 14.800 -9
300 mg QW x8 then 600 mg Q2W 5 P10 1.75 1.730 -1
300 mg QW x8 then 600 mg Q2W 5 P20 2.75 3.290 19
300 mg QW x8 then 600 mg Q2W 5 P30 4.02 4.200 5
300 mg QW x8 then 600 mg Q2W 12 mean 19.99 13.200 -34
300 mg QW x8 then 600 mg Q2W 12 P10 0.55 0.540 -2
300 mg QW x8 then 600 mg Q2W 12 P20 0.98 1.180 20
300 mg QW x8 then 600 mg Q2W 12 P30 1.55 1.670 8
300 mg QW x8 then 600 mg Q2W 12 P50 3.67 4.720 29

The percentiles track Table S3 closely. The mean is dominated by the few patients with a small time-dependent CL reduction or a late T50, so it moves more from one cohort draw to the next than the percentiles do.

pct <- cmp_s3$pct_diff[cmp_s3$statistic != "mean"]
stopifnot(
  length(pct) == 18L,
  # Structural: a wrong T50 unit, Imax sign or volume moves every percentile
  # by a factor of several; this holds for any 200-patient draw.
  abs(median(pct)) < 20,
  # Envelope over the 18 percentile cells; 200-patient draws put the 90th
  # percentile at about 20-35 %.
  quantile(abs(pct), 0.9) < 45
)

Simulated concentration-time profiles (Figure 4 and Figure S6)

sim |>
  dplyr::filter(regimen == "360 mg QW", time <= 13 * 168) |>
  dplyr::group_by(time) |>
  dplyr::summarise(
    Q05 = quantile(Cc, 0.05), Q50 = quantile(Cc, 0.5), Q95 = quantile(Cc, 0.95),
    .groups = "drop"
  ) |>
  ggplot(aes(time / 24, Q50)) +
  geom_ribbon(aes(ymin = Q05, ymax = Q95), fill = "grey80") +
  geom_line(colour = "blue") +
  geom_hline(yintercept = 2.27, linetype = "dashed", colour = "red") +
  scale_y_log10() +
  labs(x = "Time (day)", y = "Fanastomig serum concentration (ug/mL)",
       caption = "Replicates Figure 4 of Jiang 2025: 360 mg QW, median and 5th-95th percentiles; dashed line 2.27 ug/mL tumour target.")
#> Warning in scale_y_log10(): log-10 transformation introduced infinite values.
#> log-10 transformation introduced infinite values.
#> log-10 transformation introduced infinite values.
#> log-10 transformation introduced infinite values.

sim |>
  dplyr::group_by(regimen, time) |>
  dplyr::summarise(Q50 = median(Cc), .groups = "drop") |>
  ggplot(aes(time / 24, Q50, colour = regimen)) +
  geom_line() +
  geom_hline(yintercept = 2.27, linetype = "dashed") +
  scale_y_log10() +
  labs(x = "Time (day)", y = "Median serum concentration (ug/mL)", colour = NULL,
       caption = "Replicates Figure S6 of Jiang 2025: median profiles of the three regimens.") +
  theme(legend.position = "bottom")
#> Warning in scale_y_log10(): log-10 transformation introduced infinite values.

The paper describes the share of patients whose trough exceeds the 2.27 ug/mL tumour target:

attain <- function(reg, t) mean(sim$Cc[sim$regimen == reg & sim$time == t] > 2.27)
data.frame(
  statement = c(
    "360 mg QW, after 5th dose: about 90 % (Table S3 P10 = 2.19)",
    "180 mg loading, after 5th dose: approximately 70 %",
    "180 mg loading, after 12th dose: approximately 50 %",
    "300 mg loading, after 12th dose: nearly 70 %"
  ),
  simulated_pct = round(100 * c(
    attain("360 mg QW", 840),
    attain("180 mg QW x8 then 360 mg Q2W", 840),
    attain("180 mg QW x8 then 360 mg Q2W", 2688),
    attain("300 mg QW x8 then 600 mg Q2W", 2688)
  ))
) |>
  knitr::kable()
statement simulated_pct
360 mg QW, after 5th dose: about 90 % (Table S3 P10 = 2.19) 86
180 mg loading, after 5th dose: approximately 70 % 68
180 mg loading, after 12th dose: approximately 50 % 46
300 mg loading, after 12th dose: nearly 70 % 66

Albumin effect (Figure 2)

Figure 2 shows Cmax and AUC at steady state for albumin 23 and 50 g/L relative to 38 g/L. The typical-value simulation below uses 360 mg QW and the 13th dosing interval.

alb_ev <- function(alb, id) {
  dplyr::bind_rows(
    data.frame(id = id, time = (0:12) * 168, amt = 360, evid = 1L, dur = 1),
    data.frame(id = id, time = seq(12 * 168, 13 * 168, by = 0.25), amt = 0, evid = 0L, dur = 0)
  ) |>
    dplyr::mutate(cmt = "central", ALB = alb) |>
    dplyr::arrange(time, dplyr::desc(evid))
}
ev_alb <- dplyr::bind_rows(alb_ev(23, 1L), alb_ev(38, 2L), alb_ev(50, 3L))
sim_alb <- rxode2::rxSolve(rxode2::zeroRe(mod), events = ev_alb) |>
  as.data.frame() |>
  dplyr::mutate(ALB = c(23, 38, 50)[id])
#> ℹ parameter labels from comments will be replaced by 'label()'
#> ℹ omega/sigma items treated as zero: 'etalvc', 'etalcl', 'etalcl_time_max', 'etalcl_t50'
#> Warning: multi-subject simulation without without 'omega'

expo <- sim_alb |>
  dplyr::group_by(ALB) |>
  dplyr::summarise(
    cmax = max(Cc),
    auc = sum(diff(time) * (head(Cc, -1) + tail(Cc, -1)) / 2),
    .groups = "drop"
  )
ref <- expo[expo$ALB == 38, ]
alb_tab <- expo |>
  dplyr::filter(ALB != 38) |>
  dplyr::transmute(
    ALB,
    cmax_ratio = cmax / ref$cmax,
    auc_ratio = auc / ref$auc,
    cmax_published = c(0.75, 1.12),
    auc_published = c(0.83, 1.14)
  )
alb_tab |>
  dplyr::mutate(dplyr::across(c(cmax_ratio, auc_ratio), \(x) round(x, 2))) |>
  dplyr::rename(
    "ALB (g/L)" = ALB, "Cmax ratio (model)" = cmax_ratio, "AUC ratio (model)" = auc_ratio,
    "Cmax ratio (Fig. 2)" = cmax_published, "AUC ratio (Fig. 2)" = auc_published
  ) |>
  knitr::kable(caption = "Exposure at albumin 23 and 50 g/L relative to 38 g/L (typical patient, 360 mg QW, 13th interval).")
Exposure at albumin 23 and 50 g/L relative to 38 g/L (typical patient, 360 mg QW, 13th interval).
ALB (g/L) Cmax ratio (model) AUC ratio (model) Cmax ratio (Fig. 2) AUC ratio (Fig. 2)
23 0.71 1 0.75 0.83
50 1.24 1 1.12 1.14

# Deterministic: lower albumin -> larger Vc -> lower Cmax, within the Figure 2
# 90 % CI at 23 g/L (0.65-0.85).
cmax23 <- alb_tab$cmax_ratio[alb_tab$ALB == 23]
stopifnot(cmax23 > 0.65, cmax23 < 0.85)

The Cmax ratio at 23 g/L (0.71) falls inside the Figure 2 interval (0.65-0.85); at 50 g/L the model gives 1.24 against a published 1.12 (0.99-1.32). A volume effect leaves the steady-state AUC over a dosing interval equal to dose / CL, so the model’s AUC ratio is 1, while Figure 2 shows 0.83 and 1.14. The paper does not define the AUC window used in the forest plot; an AUC over a window that ends before steady state, or one that includes the time-dependent CL change, would depend on Vc and could explain the difference.

PD-1 receptor occupancy (Figure 3)

ro_curve <- data.frame(Cc = 10^seq(log10(0.02), log10(40), length.out = 200)) |>
  dplyr::mutate(RO = emax * Cc / (ec50 + Cc))

ggplot(ro_curve, aes(Cc, RO)) +
  geom_line() +
  geom_vline(xintercept = c(ec50, c90, 3 * c90), linetype = c("dotted", "dashed", "dashed"),
             colour = c("grey40", "blue", "red")) +
  scale_x_log10() +
  coord_cartesian(ylim = c(0, 120)) +
  labs(x = "Fanastomig serum concentration (ug/mL)", y = "Receptor occupancy (%)",
       caption = paste("Replicates the median curve of Figure 3 of Jiang 2025.",
                       "Dotted: EC50; blue: 90 % of maximum (0.756); red: tumour target (2.27)."))

The RO column of a simulation applies the same curve at every time point. Its median at the trough times of the first, fifth and twelfth weekly dosing intervals:

sim |>
  dplyr::filter(time %in% c(168, 840, 2016)) |>
  dplyr::group_by(regimen, time) |>
  dplyr::summarise(median_RO = round(median(RO), 1), .groups = "drop") |>
  tidyr::pivot_wider(names_from = time, values_from = median_RO, names_prefix = "t = ") |>
  knitr::kable(caption = "Median peripheral PD-1 RO (%) at the trough times 168, 840 and 2016 h.")
Median peripheral PD-1 RO (%) at the trough times 168, 840 and 2016 h.
regimen t = 168 t = 840 t = 2016
360 mg QW 85.6 91.4 91.5
180 mg QW x8 then 360 mg Q2W 80.0 90.7 88.7
300 mg QW x8 then 600 mg Q2W 85.2 91.6 90.9

PKNCA validation

The paper reports no NCA table. PKNCA is run on the first dosing interval and on a late interval of each regimen to document exposure; the half-life row uses the derived two-compartment half-lives, which the paper summarises as a model-estimated median of 3.7 days.

sim_nca <- sim |>
  dplyr::filter(!is.na(Cc)) |>
  dplyr::select(id, time, Cc, regimen)
sim_nca <- dplyr::bind_rows(
  sim_nca,
  sim_nca |> dplyr::distinct(id, regimen) |> dplyr::mutate(time = 0, Cc = 0)
) |>
  dplyr::distinct(id, regimen, time, .keep_all = TRUE) |>
  dplyr::arrange(id, regimen, time)

dose_df <- events |>
  dplyr::filter(evid == 1) |>
  dplyr::select(id, time, amt, regimen)

conc_obj <- PKNCA::PKNCAconc(sim_nca, Cc ~ time | regimen + id)
dose_obj <- PKNCA::PKNCAdose(dose_df, amt ~ time | regimen + id)

# Trough: `ctrough` (end of interval) for the first interval, whose minimum is
# the predose zero; `cmin` for the late intervals, where the minimum is the
# end-of-interval trough (PKNCA returns `ctrough` as NA for an interval that
# does not start at time 0).
intervals <- data.frame(
  start = c(0, 2016, 2352),
  end = c(168, 2184, 2688),
  cmax = TRUE,
  tmax = TRUE,
  auclast = TRUE,
  ctrough = c(TRUE, FALSE, FALSE),
  cmin = c(FALSE, TRUE, TRUE)
)
nca_res <- PKNCA::pk.nca(PKNCA::PKNCAdata(conc_obj, dose_obj, intervals = intervals))

nca_tab <- as.data.frame(nca_res) |>
  dplyr::filter(PPTESTCD %in% c("cmax", "ctrough", "cmin", "auclast")) |>
  dplyr::mutate(PPTESTCD = dplyr::if_else(PPTESTCD %in% c("ctrough", "cmin"), "trough", PPTESTCD)) |>
  dplyr::group_by(regimen, start, end, PPTESTCD) |>
  dplyr::summarise(median = signif(median(PPORRES, na.rm = TRUE), 3), .groups = "drop") |>
  tidyr::pivot_wider(names_from = PPTESTCD, values_from = median)

# Late intervals: the 13th weekly interval for 360 mg QW, the 12th (Q2W)
# interval for the loading regimens.
nca_tab |>
  dplyr::filter(
    start == 0 |
      (regimen == "360 mg QW" & start == 2016) |
      (regimen != "360 mg QW" & start == 2352)
  ) |>
  dplyr::rename(
    "Regimen" = regimen, "Start (h)" = start, "End (h)" = end,
    "Cmax (ug/mL)" = cmax, "Trough (ug/mL)" = trough, "AUClast (ug*h/mL)" = auclast
  ) |>
  knitr::kable(caption = "Median simulated NCA parameters by regimen and interval.")
Median simulated NCA parameters by regimen and interval.
Regimen Start (h) End (h) AUClast (ug*h/mL) Cmax (ug/mL) Trough (ug/mL)
180 mg QW x8 then 360 mg Q2W 0 168 703 26.5 0.539
180 mg QW x8 then 360 mg Q2W 2352 2688 4070 66.9 1.960
300 mg QW x8 then 600 mg Q2W 0 168 1200 44.0 0.977
300 mg QW x8 then 600 mg Q2W 2352 2688 7860 111.0 4.720
360 mg QW 0 168 1450 53.7 1.040
360 mg QW 2016 2184 3920 75.2 7.540
# Individual two-compartment terminal half-lives at the first dose (CL0) and
# after the time-dependent decrease has completed (CL0 * exp(Imax)).
indiv <- sim |>
  dplyr::filter(time == 0) |>
  dplyr::select(id, cl0, cl_time_max_i, vc, vp, q)
beta_hl <- function(cl, vc, vp, q) {
  k10 <- cl / vc; k12 <- q / vc; k21 <- q / vp
  s <- k10 + k12 + k21
  beta <- (s - sqrt(s^2 - 4 * k10 * k21)) / 2
  log(2) / beta / 24
}
hl <- data.frame(
  phase = c("t = 0", "t >> T50"),
  median_half_life_day = round(c(
    median(beta_hl(indiv$cl0, indiv$vc, indiv$vp, indiv$q)),
    median(beta_hl(indiv$cl0 * exp(indiv$cl_time_max_i), indiv$vc, indiv$vp, indiv$q))
  ), 2)
)
knitr::kable(hl, caption = "Median terminal half-life (day); the paper reports a model-estimated median of 3.7 days.")
Median terminal half-life (day); the paper reports a model-estimated median of 3.7 days.
phase median_half_life_day
t = 0 2.53
t >> T50 4.03

The published 3.7 days falls between the median half-life at the first dose and after the clearance decrease has completed. The paper does not say at which time it evaluated the half-life.

Assumptions and deviations

  • Albumin acts on the central volume. Table 1 labels the covariate row “Effect of ALB on V p”, but the final model code in Data S1 applies it to the central volume (stparm(V = tvV *(ALB/36)^dVdALB * exp(nV)); V2 is the peripheral volume and carries no covariate). Figure 2 confirms this. Lower albumin cuts Cmax by 25%, which a central-volume effect reproduces (0.71 here) and a peripheral-volume effect cannot (Cmax ratio 1.00 in a simulation with the effect moved to Vp).
  • Albumin reference 36 g/L. The model code divides by 36 g/L. Table S1 reports a median of 38 g/L for all 47 screened patients, but the 35 patients in the final model are a subset of them.
  • Parameter values from Table 1. The Data S1 code carries slightly different digits from a re-run of the final model (for example T50 180.57 h, Imax -0.958, Hill 8.507, omega2 Imax 0.411). The published Table 1 estimates are used.
  • Emax digitised. The paper reports only EC50 (0.084 ug/mL) for the RO model. Its statement that 0.756 ug/mL (9 x EC50) maintains 90% of the maximal RO fixes a hyperbolic shape with no Hill coefficient. The maximal occupancy is not printed; the maintainers digitised the Figure 3 median curve (91-92.5% between 3 and 28.5 ug/mL) and set Emax to 92.5%. The paper reports no inter-individual or residual variability for RO, so RO is a deterministic output.
  • Time since first dose. CL(t) uses the model time t, the time since the patient’s first dose, matching Phoenix’s t in Data S1.
  • Infusion duration. The paper does not state the infusion length. The simulations use 1-hour infusions; Data S1 doses into A1 at a rate set in the data.
  • Albumin distribution. Normal, mean 38 g/L, SD 5 g/L, truncated to the Table S1 range; the paper gives only the median and range.
  • Exposure-response and ADA analyses not encoded. The logistic regressions of anaphylaxis on Cmax and AUC and the correlations of tumour-size change with Ctrough and AUC were not significant, and the paper prints no coefficients. The ADA-titre analyses are exploratory comparisons (Wilcoxon tests). None of these define a model to encode.
  • No erratum or correction notice was found for this article (EuropePMC, checked 2026-11-04).