Skip to contents

Model and source

Dimelow et al. (2024) extrapolated intravenous belimumab exposure from adults with active lupus nephritis (LN) to children aged 5-17 years with LN, to support a 10 mg/kg dose given on days 0, 14 and 28 and every 4 weeks thereafter. The pharmacokinetic engine is the adult LN population PK model fitted to the BLISS-LN trial, which Dimelow 2024 cites to Simeoni et al. (2024). Dimelow 2024 prints the model’s covariate structure (Electronic Supplementary Material Section S5) but not its estimates; those are in Simeoni 2024 Online Resource 7, so the packaged model is named after that paper and this vignette reproduces the Dimelow 2024 paediatric extrapolation with it.

  • Model: Simeoni_2024_belimumab
  • Citation: Simeoni M, Yang S, Tompson DJ, Dimelow R. Longitudinal modeling of efficacy response in patients with lupus nephritis receiving belimumab. J Pharmacokinet Pharmacodyn. 2024;51(3):289-301. doi:10.1007/s10928-024-09907-w (PK model parameters: Online Resource 7). Application: Dimelow R, Liefaard L, Green Y, Tomlinson R. Extrapolation of the Efficacy and Pharmacokinetics of Belimumab to Support its Use in Children with Lupus Nephritis. Clin Pharmacokinet. 2024;63(9):1313-1326. doi:10.1007/s40262-024-01422-y
  • Description: Linear two-compartment IV population PK model for belimumab in adults with active lupus nephritis (BLISS-LN), with baseline fat-free mass allometry on all disposition parameters and time-varying proteinuria (urine protein/creatinine ratio) and serum albumin on clearance (Simeoni 2024); used by Dimelow 2024 to extrapolate exposure to children aged 5-17 years with lupus nephritis
  • Extrapolation article (Dimelow 2024): https://doi.org/10.1007/s40262-024-01422-y
  • Parameter source (Simeoni 2024): https://doi.org/10.1007/s10928-024-09907-w

Population

The model was fitted to the 224 adults with active, biopsy-confirmed LN who received belimumab 10 mg/kg IV plus standard therapy in BLISS-LN (BEL114054). In that arm (Simeoni 2024 Table 1) the median age was 31 years (18-63), median weight 59.0 kg (36.9-136.9), median baseline fat-free mass (FFM) 38.1 kg (26.3-77.4), 88% were female and 51% Asian. Disease activity was high at baseline: median urine protein/creatinine ratio (uPCR) 2.61 g/g (0.18-16.57) and median serum albumin 31 g/L (16-42). PK samples were taken on days 0, 3, 14, 28, 56, 168, 171, 364 and 728.

Dimelow 2024 applied the model unchanged to children with LN. Body weight and body mass index (BMI) were sampled from NHANES for ages 5-17 years (90% female), scaled to 86% and 94% respectively so the distributions matched children with SLE in the PLUTO study, and FFM was computed with the Janmahasatian formula. Time courses of uPCR and albumin were sampled from the adult BLISS-LN patients, on the basis that baseline values and the response to standard therapy are similar in adults and children with LN (Dimelow 2024 Table 2).

str(rxode2::rxode(readModelDb("Simeoni_2024_belimumab"))$population)
#> ℹ parameter labels from comments will be replaced by 'label()'
#> List of 16
#>  $ species       : chr "human"
#>  $ n_subjects    : num 224
#>  $ n_studies     : num 1
#>  $ age_range     : chr "18-63 years"
#>  $ age_median    : chr "31 years"
#>  $ weight_range  : chr "36.9-136.9 kg"
#>  $ weight_median : chr "59.0 kg"
#>  $ ffm_median    : chr "38.1 kg (range 26.3-77.4)"
#>  $ sex_female_pct: num 88
#>  $ race_ethnicity: chr "Asian 51%, White 33%, Black African ancestry 13%, American Indian 2%, multiple 1%"
#>  $ disease_state : chr "Active, biopsy-confirmed lupus nephritis (adults), on standard therapy (mycophenolate mofetil or cyclophosphami"| __truncated__
#>  $ dose_range    : chr "Belimumab 10 mg/kg IV on days 0, 14, 28 and every 28 days thereafter, for 104 weeks"
#>  $ regions       : chr "Multinational"
#>  $ renal_function: chr "Baseline eGFR median 99 (range 14-208) mL/min/1.73 m^2; uPCR median 2.61 (0.18-16.57) g/g; serum albumin median 31 (16-42) g/L"
#>  $ studies       : chr "BEL114054 (BLISS-LN; NCT01639339), belimumab arm"
#>  $ notes         : chr "Belimumab-arm demographics from Simeoni 2024 Table 1. PK samples on days 0, 3, 14, 28, 56, 168, 171, 364 and 72"| __truncated__

Source trace

Equation / parameter Value Source location
lcl (CL at FFM 38.4 kg, ALB 42 g/L, uPCR 0) log(0.175 L/day) Simeoni 2024 Online Resource 7: CL = 175 mL/day
lvc (V1) log(2.728 L) Online Resource 7: V1 = 2728 mL
lq (Q) log(0.487 L/day) Online Resource 7: Q = 487 mL/day
lvp (V2) log(1.992 L) Online Resource 7: V2 = 1992 mL
e_ffm_cl (FFM exponent on CL and Q) 0.624 Online Resource 7 (Q row: ‘See FFMBL on CL’)
e_ffm_vc (FFM exponent on V1 and V2) 0.723 Online Resource 7 (V2 row: ‘See FFMBL on V1’)
e_alb_cl (theta1, albumin exponent) -1.74 Online Resource 7
e_upcr_alb_cl (theta2, attenuation by uPCR) 0.0832 Online Resource 7
e_upcr_cl (linear uPCR effect on CL) 0.0663 Online Resource 7
etalcl, etalvc, etalvp 0.0593, 0.0322, 0.133 Online Resource 7 (omega^2; CV 24.7%, 18.1%, 37.7%)
etapropSd (IIV on the proportional error) 0.346 Online Resource 7 (omega^2 PROP; CV 64.3%)
propSd 0.241 Online Resource 7 (PROP)
addSd 0.1 mg/L, fixed Online Resource 7 (ADD = 0.1 ug/mL, Fixed)
CL = CL_pop (FFM/38.4)^0.624 (ALB/42)^(theta1/(1 + theta2 uPCR)) (1 + theta uPCR) n/a Online Resource 7; Dimelow 2024 ESM Section S5
V1, V2 ~ (FFM/38.4)^0.723; Q ~ (FFM/38.4)^0.624 n/a Online Resource 7; Dimelow 2024 ESM Section S5
Two-compartment model, first-order distribution and elimination n/a Simeoni 2024 Online Resource 1; Dimelow 2024 Methods
Reference FFM 38.4 kg = BLISS-LN median over both arms (N = 448) n/a Simeoni 2024 Table 1, ‘Overall’ column

Virtual cohort

The individual BLISS-LN covariate records are not public, so the cohorts below approximate them from published summaries.

  • Children (5-11 and 12-17 years). Ages are uniform within each band. Median weight and BMI by year of age were digitised by the maintainers from the corrected NHANES sample in Dimelow 2024 Figure 2b; between-subject spread is log-normal (SD 0.22 on log weight, 0.18 on log BMI, correlation 0.8), chosen to cover the Figure 2b 95% prediction bands. 90% are female.
  • Adults. Weight is log-normal around the BLISS-LN belimumab-arm median of 59 kg; BMI around an assumed 23 kg/m^2 (BLISS-LN BMI is not published); 88% female. FFM from the Janmahasatian formula then lands near the reported 38.1 kg median, which is checked below.
  • Proteinuria and albumin, all cohorts. Baseline uPCR is log-normal around 2.6 g/g (Simeoni 2024 Table 1) and baseline albumin normal around 31 g/L, negatively correlated with uPCR. Each subject’s uPCR then falls as uPCR0 * (0.14 + 0.86 exp(-k t)) and albumin rises towards 40.5 g/L. The recovery rates were set so that the cohort medians pass through the BLISS-LN values of Dimelow 2024 Table 2 (uPCR medians 2.47, 0.73 and 0.44 g/g and albumin medians 31, 38 and 40 g/L at baseline, 6 and 12 months).
set.seed(20241001)
rxode2::rxSetSeed(20241001)

n_per_arm <- 200

# Dimelow 2024 Figure 2b: corrected NHANES median weight (kg) and BMI (kg/m^2)
# by age, digitised.
growth <- tibble::tribble(
  ~age, ~wt_med, ~bmi_med,
  5,    17.0,    15.3,
  6,    19.0,    15.3,
  7,    21.5,    15.8,
  8,    24.5,    16.5,
  9,    27.0,    17.0,
  10,   31.0,    18.0,
  11,   37.0,    19.0,
  12,   41.0,    20.0,
  13,   45.0,    20.5,
  14,   46.0,    21.0,
  15,   52.0,    22.0,
  16,   53.0,    22.0,
  17,   55.0,    22.5
)

# Janmahasatian et al. 2005 fat-free mass (Dimelow 2024 ref. 25), vectorised
# over sex.
ffm_janmahasatian <- function(wt, bmi, sexf) {
  ifelse(
    sexf == 1,
    9270 * wt / (8780 + 244 * bmi),
    9270 * wt / (6680 + 216 * bmi)
  )
}

make_subjects <- function(n, cohort, id_offset, age_lo = NA, age_hi = NA) {
  z_wt <- rnorm(n)
  z_bmi <- 0.8 * z_wt + 0.6 * rnorm(n)
  if (cohort == "Adults") {
    age <- NA_real_
    wt <- 59 * exp(0.22 * z_wt)
    bmi <- 23 * exp(0.18 * z_bmi)
    sexf <- rbinom(n, 1, 0.88)
  } else {
    age <- runif(n, age_lo, age_hi + 1)
    wt <- approx(growth$age, growth$wt_med, xout = age, rule = 2)$y *
      exp(0.22 * z_wt)
    bmi <- approx(growth$age, growth$bmi_med, xout = age, rule = 2)$y *
      exp(0.18 * z_bmi)
    sexf <- rbinom(n, 1, 0.9)
  }
  z_upcr <- rnorm(n)
  z_rec <- rnorm(n)
  tibble(
    id = id_offset + seq_len(n),
    cohort = cohort,
    age = age,
    WT = wt,
    BMI = bmi,
    SEXF = sexf,
    FFM = ffm_janmahasatian(wt, bmi, sexf),
    upcr0 = pmin(pmax(2.6 * exp(0.8 * z_upcr), 0.15), 17),
    alb0 = pmin(pmax(31 + 5 * (-0.5 * z_upcr + sqrt(0.75) * rnorm(n)), 14), 43),
    k_upcr = 0.0092 * exp(0.5 * z_rec),
    k_alb = 0.0073 * exp(0.5 * z_rec)
  )
}

subjects <- bind_rows(
  make_subjects(n_per_arm, "Adults", 0L),
  make_subjects(n_per_arm, "Children 5-11 y", n_per_arm, 5, 11),
  make_subjects(n_per_arm, "Children 12-17 y", 2L * n_per_arm, 12, 17)
) |>
  mutate(cohort = factor(cohort, levels = c("Adults", "Children 5-11 y", "Children 12-17 y")))

# 10 mg/kg as a 1-hour infusion on days 0, 14, 28 and every 28 days after.
dose_times <- c(0, 14, 28, seq(56, 168, by = 28))
inf_dur <- 1 / 24
obs_times <- sort(unique(c(seq(0, 168, by = 1), dose_times + inf_dur)))

doses <- subjects |>
  tidyr::expand_grid(time = dose_times) |>
  mutate(evid = 1L, amt = 10 * WT, rate = amt / inf_dur)
obs <- subjects |>
  tidyr::expand_grid(time = obs_times) |>
  mutate(evid = 0L, amt = 0, rate = 0)

events <- bind_rows(doses, obs) |>
  mutate(
    cmt = "central",
    UPCR = upcr0 * (0.14 + 0.86 * exp(-k_upcr * time)),
    ALB = 40.5 - (40.5 - alb0) * exp(-k_alb * time)
  ) |>
  arrange(id, time, desc(evid)) |>
  select(id, time, evid, amt, rate, cmt, cohort, WT, FFM, ALB, UPCR)

# One dose row and one observation row at most per subject and time.
stopifnot(!anyDuplicated(events[, c("id", "time", "evid")]))

The simulated cohorts against the published body-size and disease summaries:

subjects |>
  group_by(cohort) |>
  summarise(
    `Weight (kg)` = median(WT),
    `FFM (kg)` = median(FFM),
    `Baseline uPCR (g/g)` = median(upcr0),
    `Baseline albumin (g/L)` = median(alb0),
    .groups = "drop"
  ) |>
  knitr::kable(digits = 1, caption = "Median covariates by cohort.")
Median covariates by cohort.
cohort Weight (kg) FFM (kg) Baseline uPCR (g/g) Baseline albumin (g/L)
Adults 57.4 38.1 2.5 31.0
Children 5-11 y 26.0 19.0 2.8 30.9
Children 12-17 y 51.2 34.4 2.9 30.7

Days 182 and 365 are outside the 24-week simulation grid, so the medians at 6 and 12 months are evaluated directly from the recovery curves:

recovery <- subjects |>
  tidyr::expand_grid(time = c(0, 182, 365)) |>
  mutate(
    UPCR = upcr0 * (0.14 + 0.86 * exp(-k_upcr * time)),
    ALB = 40.5 - (40.5 - alb0) * exp(-k_alb * time)
  ) |>
  group_by(time) |>
  summarise(uPCR = median(UPCR), ALB = median(ALB), .groups = "drop")
knitr::kable(
  recovery,
  digits = 2,
  caption = "Cohort median uPCR (g/g) and albumin (g/L); BLISS-LN placebo medians in Dimelow 2024 Table 2 are 2.47 / 0.73 / 0.44 g/g and 31-32 / 38 / 40 g/L."
)
Cohort median uPCR (g/g) and albumin (g/L); BLISS-LN placebo medians in Dimelow 2024 Table 2 are 2.47 / 0.73 / 0.44 g/g and 31-32 / 38 / 40 g/L.
time uPCR ALB
0 2.70 30.94
182 0.85 38.16
365 0.52 39.91

adult_ffm <- median(subjects$FFM[subjects$cohort == "Adults"])
# The BLISS-LN belimumab arm median is 38.1 kg (Simeoni 2024 Table 1).
stopifnot(abs(adult_ffm / 38.1 - 1) < 0.10)

Simulation

mod <- readModelDb("Simeoni_2024_belimumab")
sim <- rxode2::rxSolve(mod, events = events, keep = c("cohort", "WT")) |>
  as.data.frame()
#> ℹ parameter labels from comments will be replaced by 'label()'

Replicate published figures

Figure 4: concentration-time profiles

prof <- sim |>
  group_by(cohort, time) |>
  summarise(
    Q025 = quantile(Cc, 0.025),
    Q25 = quantile(Cc, 0.25),
    Q50 = median(Cc),
    Q75 = quantile(Cc, 0.75),
    Q975 = quantile(Cc, 0.975),
    .groups = "drop"
  )
adult_prof <- prof |>
  filter(cohort == "Adults") |>
  select(time, A50 = Q50, A025 = Q025, A975 = Q975)

prof |>
  filter(cohort != "Adults") |>
  left_join(adult_prof, by = "time") |>
  ggplot(aes(time / 7)) +
  geom_ribbon(aes(ymin = Q025, ymax = Q975), fill = "grey80") +
  geom_ribbon(aes(ymin = Q25, ymax = Q75), fill = "grey55") +
  geom_line(aes(y = Q50)) +
  geom_line(aes(y = A50), colour = "red") +
  geom_line(aes(y = A025), colour = "red", linetype = "dashed") +
  geom_line(aes(y = A975), colour = "red", linetype = "dashed") +
  facet_wrap(~cohort) +
  labs(
    x = "Time (weeks)", y = "Belimumab serum concentration (ug/mL)",
    caption = paste(
      "Replicates Figure 4 of Dimelow 2024: children (black, IQR and 95% PI shaded)",
      "versus adults (red median and 95% PI)."
    )
  )

Figure 5: average concentration versus body weight

sim_nca <- sim |>
  filter(!is.na(Cc)) |>
  select(id, cohort, time, Cc)

conc_obj <- PKNCA::PKNCAconc(sim_nca, Cc ~ time | cohort + id)
dose_obj <- PKNCA::PKNCAdose(
  events |> filter(evid == 1) |> select(id, cohort, time, amt),
  amt ~ time | cohort + id
)
# Cmin (week N) is the pre-dose concentration at day 7N; its interval starts
# one day after the previous dose so the earlier trough is excluded.
intervals <- data.frame(
  start = c(0, 0, 56, 140, 57, 141),
  end = c(84, 168, 84, 168, 84, 168),
  cav = c(TRUE, TRUE, FALSE, FALSE, FALSE, FALSE),
  cmax = c(FALSE, FALSE, TRUE, TRUE, FALSE, FALSE),
  cmin = c(FALSE, FALSE, FALSE, FALSE, TRUE, TRUE)
)
nca_res <- PKNCA::pk.nca(PKNCA::PKNCAdata(conc_obj, dose_obj, intervals = intervals))

nca_long <- as.data.frame(nca_res) |>
  filter(PPTESTCD %in% c("cav", "cmax", "cmin")) |>
  mutate(
    window = case_when(
      PPTESTCD == "cav" & end == 84 ~ "Cavg (week 0-12)",
      PPTESTCD == "cav" & end == 168 ~ "Cavg (week 0-24)",
      PPTESTCD == "cmin" & end == 84 ~ "Cmin (week 12)",
      PPTESTCD == "cmin" & end == 168 ~ "Cmin (week 24)",
      PPTESTCD == "cmax" & end == 84 ~ "Cmax (week 8-12)",
      PPTESTCD == "cmax" & end == 168 ~ "Cmax (week 20-24)"
    )
  )
cavg12 <- nca_long |>
  filter(window == "Cavg (week 0-12)") |>
  select(id, cohort, cavg = PPORRES) |>
  left_join(distinct(events, id, WT), by = "id")
adult_lim <- quantile(cavg12$cavg[cavg12$cohort == "Adults"], c(0.025, 0.975))

ggplot(cavg12, aes(WT, cavg, colour = cohort)) +
  geom_point(alpha = 0.5) +
  geom_hline(yintercept = adult_lim, colour = "red", linetype = "dotted") +
  scale_x_log10() +
  labs(
    x = "Body weight (kg)", y = "Cavg, weeks 0-12 (ug/mL)", colour = NULL,
    caption = "Replicates Figure 5 of Dimelow 2024; dotted lines are the simulated adult 95% PI."
  )

Comparison against published exposure summaries

Dimelow 2024 Table 4 reports the median (95% prediction interval) of each exposure metric from 10,000 simulated adults and children; the comparison below uses 200 per cohort. ncaComparisonTable() compares medians.

published <- tibble::tribble(
  ~cohort,            ~window,             ~PPTESTCD, ~PPORRES,
  "Adults",           "Cavg (week 0-12)",  "cav",     89,
  "Adults",           "Cavg (week 0-24)",  "cav",     89,
  "Adults",           "Cmin (week 12)",    "cmin",    34,
  "Adults",           "Cmin (week 24)",    "cmin",    44,
  "Adults",           "Cmax (week 8-12)",  "cmax",    254,
  "Adults",           "Cmax (week 20-24)", "cmax",    259,
  "Children 5-11 y",  "Cavg (week 0-12)",  "cav",     62,
  "Children 5-11 y",  "Cavg (week 0-24)",  "cav",     62,
  "Children 5-11 y",  "Cmin (week 12)",    "cmin",    22,
  "Children 5-11 y",  "Cmin (week 24)",    "cmin",    28,
  "Children 5-11 y",  "Cmax (week 8-12)",  "cmax",    185,
  "Children 5-11 y",  "Cmax (week 20-24)", "cmax",    188,
  "Children 12-17 y", "Cavg (week 0-12)",  "cav",     85,
  "Children 12-17 y", "Cavg (week 0-24)",  "cav",     84,
  "Children 12-17 y", "Cmin (week 12)",    "cmin",    32,
  "Children 12-17 y", "Cmin (week 24)",    "cmin",    40,
  "Children 12-17 y", "Cmax (week 8-12)",  "cmax",    242,
  "Children 12-17 y", "Cmax (week 20-24)", "cmax",    247
)

cmp <- nlmixr2lib::ncaComparisonTable(
  simulated = nca_long |>
    mutate(cohort = as.character(cohort)) |>
    select(cohort, window, PPTESTCD, PPORRES),
  reference = published,
  by = c("cohort", "window"),
  units = c(cav = "ug/mL", cmin = "ug/mL", cmax = "ug/mL"),
  tolerance_pct = 20
)
knitr::kable(
  cmp,
  caption = "Simulated vs. Dimelow 2024 Table 4 medians. * differs by >20%."
)
Simulated vs. Dimelow 2024 Table 4 medians. * differs by >20%.
NCA parameter cohort window Reference Simulated % diff
Cmax (ug/mL) Adults Cmax (week 8-12) 254 241 -5.0%
Cmax (ug/mL) Adults Cmax (week 20-24) 259 242 -6.4%
Cmax (ug/mL) Children 5-11 y Cmax (week 8-12) 185 184 -0.5%
Cmax (ug/mL) Children 5-11 y Cmax (week 20-24) 188 185 -1.7%
Cmax (ug/mL) Children 12-17 y Cmax (week 8-12) 242 240 -0.7%
Cmax (ug/mL) Children 12-17 y Cmax (week 20-24) 247 243 -1.7%
Cmin (ug/mL) Adults Cmin (week 12) 34 33.8 -0.6%
Cmin (ug/mL) Adults Cmin (week 24) 44 39.7 -9.7%
Cmin (ug/mL) Children 5-11 y Cmin (week 12) 22 22.7 +3.2%
Cmin (ug/mL) Children 5-11 y Cmin (week 24) 28 27.9 -0.5%
Cmin (ug/mL) Children 12-17 y Cmin (week 12) 32 29.7 -7.2%
Cmin (ug/mL) Children 12-17 y Cmin (week 24) 40 37 -7.4%
Cavg (ug/mL) Adults Cavg (week 0-12) 89 86.4 -3.0%
Cavg (ug/mL) Adults Cavg (week 0-24) 89 84.3 -5.3%
Cavg (ug/mL) Children 5-11 y Cavg (week 0-12) 62 63.2 +1.9%
Cavg (ug/mL) Children 5-11 y Cavg (week 0-24) 62 61.5 -0.8%
Cavg (ug/mL) Children 12-17 y Cavg (week 0-12) 85 82.2 -3.2%
Cavg (ug/mL) Children 12-17 y Cavg (week 0-24) 84 82 -2.4%

Table 4 also gives the share of each population below the adult 2.5th percentile of Cavg (weeks 0-12), 41 ug/mL: 2.5% of adults, 15.3% of children aged 5-11 years and 3.6% of children aged 12-17 years.

below <- cavg12 |>
  group_by(cohort) |>
  summarise(
    `Median Cavg wk 0-12 (ug/mL)` = median(cavg),
    `2.5th pct` = quantile(cavg, 0.025),
    `97.5th pct` = quantile(cavg, 0.975),
    `% below 41 ug/mL` = 100 * mean(cavg < 41),
    .groups = "drop"
  ) |>
  mutate(`Published % below 41 ug/mL` = c(2.5, 15.3, 3.6))
knitr::kable(below, digits = 1)
cohort Median Cavg wk 0-12 (ug/mL) 2.5th pct 97.5th pct % below 41 ug/mL Published % below 41 ug/mL
Adults 86.4 46.2 140.8 1.0 2.5
Children 5-11 y 63.2 33.5 110.1 10.5 15.3
Children 12-17 y 82.2 38.5 129.4 3.0 3.6

med <- setNames(below$`Median Cavg wk 0-12 (ug/mL)`, as.character(below$cohort))
stopifnot(
  # Structural: a mis-transcribed clearance, exponent or unit moves the adult
  # median by tens of percent.
  abs(med[["Adults"]] / 89 - 1) < 0.20,
  abs(med[["Children 12-17 y"]] / 85 - 1) < 0.20,
  abs(med[["Children 5-11 y"]] / 62 - 1) < 0.25,
  # The paper's central finding: the weight-proportional dose under-exposes
  # the youngest children relative to adults (89 vs 62 ug/mL, about 30%).
  med[["Children 5-11 y"]] / med[["Adults"]] < 0.85
)

Dimelow 2024 Table 3 additionally reports the adult LN model simulation as a geometric mean Cavg (weeks 0-12) of 87 ug/mL (CV 36%; 95% PI 41-161).

adult_cavg <- cavg12$cavg[cavg12$cohort == "Adults"]
tibble(
  `Geometric mean (ug/mL)` = exp(mean(log(adult_cavg))),
  `CV (%)` = 100 * sqrt(exp(var(log(adult_cavg))) - 1),
  `2.5th pct` = quantile(adult_cavg, 0.025),
  `97.5th pct` = quantile(adult_cavg, 0.975)
) |>
  knitr::kable(digits = 1, caption = "Adult Cavg (weeks 0-12); published 87 (CV 36%), 41-161.")
Adult Cavg (weeks 0-12); published 87 (CV 36%), 41-161.
Geometric mean (ug/mL) CV (%) 2.5th pct 97.5th pct
85.4 29.8 46.2 140.8

The medians of every Table 4 metric are reproduced closely in all three cohorts, and so is the central finding: children aged 5-11 years have about 30% lower exposure than adults on the same 10 mg/kg dose, while adolescents match adults. The simulated distributions are somewhat narrower than the published ones (adult CV near 30% against 36%, and correspondingly fewer subjects below 41 ug/mL). The paper resampled the observed BLISS-LN proteinuria and albumin trajectories, whose between-patient spread (uPCR up to 17 g/g, albumin down to 14 g/L, flares and non-responders) is wider than the smooth recovery curves used here; that wider covariate spread is the most likely reason for the wider published tails.

Assumptions and deviations

  • Model source. Dimelow 2024 prints only the covariate structure of the adult LN model (ESM Section S5); every estimate is from Simeoni 2024 Online Resource 7, which Dimelow 2024 cites as the model’s origin.
  • Proteinuria units. Simeoni 2024 Online Resource 7 defines the proteinuria covariate as the urine protein/creatinine ratio in g/g, the measure BLISS-LN collected every 4 weeks. Dimelow 2024 ESM Section S5 labels it g/day. The primary source’s g/g is used.
  • Residual error form. Online Resource 7 lists a proportional (0.241) and a fixed additive (0.1 ug/mL) term, with between-patient variability on the proportional term, but does not print the combination rule. The conventional NONMEM form, SD = sqrt((prop * f)^2 + add^2), is used (combined2()), with only the proportional SD scaled by exp(etapropSd).
  • Infusion duration. Neither paper states it; the labelled 1-hour infusion is used. It affects only Cmax.
  • Covariate time courses. The individual uPCR and albumin profiles from BLISS-LN that Dimelow 2024 resampled are not public. Smooth exponential recovery curves were calibrated to the Table 2 placebo medians at baseline, 6 and 12 months; the BLISS-LN belimumab arm recovered somewhat faster, so the simulated exposures sit slightly low if anything. Early exposure (Cavg weeks 0-12) depends mostly on baseline uPCR and albumin, which match the trial.
  • Body size. Paediatric medians were digitised from Dimelow 2024 Figure 2b; the spread around them, the adult BMI (23 kg/m^2) and the weight-BMI correlation are assumptions. The Janmahasatian formula is applied to children as in Dimelow 2024, without a paediatric correction.
  • Cohort size. 200 subjects per cohort instead of 10,000, so the 2.5th and 97.5th percentiles are noisier than in the paper.
  • Figure 3 (VPC against the PLUTO paediatric SLE data) is not reproduced because the observed PLUTO concentrations are not public.
  • No correction notice was found for either paper as of 2026-10-30.