Skip to contents

Perampanel (Silva 2023)

Source: vignettes/articles/Silva_2023_perampanel.Rmd
Silva_2023_perampanel.Rmd

Model and source

  • Citation: Silva R, Colom H, Bicker J, Almeida A, Silva A, Sales F, Santana I, Falcao A, Fortuna A. Population Pharmacokinetic Analysis of Perampanel in Portuguese Patients Diagnosed with Refractory Epilepsy. Pharmaceutics. 2023;15(6):1704.
  • Article: https://doi.org/10.3390/pharmaceutics15061704
  • Description: One-compartment first-order-elimination popPK of perampanel in 44 Portuguese adults with refractory epilepsy on chronic therapeutic drug monitoring, with concomitant enzyme-inducing AEDs and body-mass index as retained covariates on CL and V respectively.

Population

Silva 2023 is a retrospective single-centre observational study conducted at the Refractory Epilepsy Reference Centre of the Coimbra Hospital and University Centre (CHUC, EPE, Portugal) between April 2019 and December 2022 (Section 2.1). Inclusion criteria were adults >= 18 years of age, a diagnosis of refractory epilepsy (defined as failure of two appropriate and tolerated AED schedules to achieve sustained seizure freedom), and at least one month of perampanel treatment under routine TDM. Patients admitted to intensive care, with status epilepticus, or on chronic non-AED medications were excluded. Forty-four patients contributed 72 steady-state plasma concentrations to the modelling dataset. Median age was 33 years (range 19-76); median body weight 77.5 kg (range 45-99); median height 168.5 cm; and median BMI 25.14 kg/m^2 (range 15.76-36.20). Sixty-one percent of subjects were male. All patients were on chronic perampanel dosed once daily at bedtime, with daily doses ranging from 2 to 10 mg. Only one patient took perampanel as monotherapy; the remaining 43 were on polytherapy with 1-4 concomitant AEDs. The most frequently co-prescribed AEDs were levetiracetam (63.6%), carbamazepine (34.1%), clobazam (25.0%), and eslicarbazepine acetate (22.7%). Twenty patients (45.5%) were co-prescribed at least one enzyme-inducing AED - carbamazepine, oxcarbazepine, phenobarbital, or phenytoin (Table 1 and Table S1).

Blood was sampled at TDM visits either 9.7-14 h after the bedtime dose (morning trough sample, n = 42) or 20.5-24 h post-dose (pre-next-dose trough, n = 30), all at steady state. Perampanel was quantified by validated HPLC-DAD (Sabenca 2021) over 0.03-4.50 mg/L.

The same baseline summary is available programmatically via readModelDb("Silva_2023_perampanel")()$population.

Source trace

The per-parameter origin is recorded as an in-file comment next to each ini() entry of inst/modeldb/specificDrugs/Silva_2023_perampanel.R. The table below collects them in one place for review.

Equation / parameter Value Source location
lcl (CL/F, L/h, no-EIAED reference) log(0.419) Silva 2023 Table 2 (final-model TVCL)
lvc (V/F, L, BMI = 25.1 kg/m^2 reference) log(29.50) Silva 2023 Table 2 (final-model TVV)
e_eiaed_cl (multiplier on CL when CONMED_EIAED = 1) 2.76 Silva 2023 Table 2 (final-model IND_CL)
e_bmi_vc (power exponent on BMI/25.1 for V) 2.12 Silva 2023 Table 2 (final-model BMI_V)
etalcl (omega^2, log-scale) log(1 + 0.3082^2) = 0.0907 Silva 2023 Table 2 (final-model IPV_CL 30.82% CV)
propSd (proportional residual SD, fraction) 0.0644 Silva 2023 Table 2 (final-model RE_proportional 6.44%)
BMI centering value (kg/m^2) 25.1 Silva 2023 Table S5 model 3 covariate equation (paper text page 6 uses BMI/25.1; Table 1 sample median = 25.14 rounded)
EIAED bucket (CBZ + OXC + PB + PHT) n/a Silva 2023 Section 2.4.2 (definition), Table 2 (retained effect), Table S4 model 13 (bucketed IND with dOFV -54.259)
One-compartment ODE with IV-bolus parameterisation n/a Silva 2023 Section 2.4.1 and 3.1 (“intravenous bolus administration was herein chosen” because concentrations were obtained during the elimination phase only)
CL_i = TVCL * e_eiaed_cl^CONMED_EIAED * exp(eta_lcl) n/a Silva 2023 Equation (1) / Table S5 model 3
V_i = TVV * (BMI/25.1)^e_bmi_vc n/a Silva 2023 Equation (2) / Table S5 model 3

Virtual cohort

Original observed data are not publicly available. The figures below use virtual cohorts approximating the Table 3 / Figure 3 model-based dosing simulations Silva 2023 published: a 4 x 6 x 2 grid of BMI (18.5, 22.5, 27.5, 32.5 kg/m^2) x once-daily dose (2, 4, 6, 8, 10, 12 mg) x concomitant EIAED status (0, 1), with 100 stochastic subjects per group. Doses are IV boluses into the central compartment - matching the paper’s structural choice - dispensed once daily for 21 days (504 h) to reach steady state (perampanel t1/2 ~= 49 h at typical CL/V, so 5-7 half-lives are covered). Post-dose observations are collected on the final dosing interval only, at 3-hour spacing, for the PKNCA-based steady-state check below.

set.seed(20230612)

n_per_group <- 40L
tau          <- 24        # hours between once-daily doses
n_doses      <- 21L       # 21 daily doses => steady state
ss_start     <- (n_doses - 1L) * tau        # time of the final dose (480 h)
ss_end       <- n_doses      * tau          # end of the final dosing interval (504 h)

bmi_levels   <- c(18.5, 22.5, 27.5, 32.5)
dose_levels  <- c(2, 4, 6, 8, 10, 12)
eiaed_levels <- c(0L, 1L)

grid <- tidyr::crossing(
  BMI          = bmi_levels,
  dose_mg      = dose_levels,
  CONMED_EIAED = eiaed_levels
) |>
  dplyr::arrange(BMI, dose_mg, CONMED_EIAED)

# One row per (BMI, dose, EIAED) group; disjoint IDs per row via id_offset.
grid <- dplyr::mutate(
  grid,
  group_id  = seq_len(dplyr::n()),
  treatment = sprintf("%d mg | BMI %s | %s EIAED",
                      dose_mg,
                      format(BMI, nsmall = 1),
                      ifelse(CONMED_EIAED == 1L, "with", "no")),
  id_offset = (group_id - 1L) * n_per_group
)

# Build one cohort per grid row and stitch them together.
obs_grid_ss <- seq(ss_start, ss_end, by = 3)

make_cohort <- function(row) {
  ids <- row$id_offset + seq_len(n_per_group)

  doses <- tibble::tibble(
    id           = rep(ids, each = n_doses),
    time         = rep(seq(0, ss_start, by = tau), times = length(ids)),
    amt          = row$dose_mg,
    evid         = 1L,
    cmt          = "central",
    BMI          = row$BMI,
    CONMED_EIAED = row$CONMED_EIAED,
    treatment    = row$treatment
  )

  obs <- tibble::tibble(
    id           = rep(ids, each = length(obs_grid_ss)),
    time         = rep(obs_grid_ss, times = length(ids)),
    amt          = NA_real_,
    evid         = 0L,
    cmt          = "central",
    BMI          = row$BMI,
    CONMED_EIAED = row$CONMED_EIAED,
    treatment    = row$treatment
  )

  dplyr::bind_rows(doses, obs) |> dplyr::arrange(id, time, dplyr::desc(evid))
}

events <- purrr::map_dfr(seq_len(nrow(grid)), function(i) make_cohort(grid[i, , drop = FALSE]))

stopifnot(!anyDuplicated(unique(events[, c("id", "time", "evid")])))

Simulation 

mod <- readModelDb("Silva_2023_perampanel")

sim <- rxode2::rxSolve(
  mod, events = events,
  keep = c("BMI", "CONMED_EIAED", "treatment")
) |>
  as.data.frame() |>
  dplyr::as_tibble()
#> ℹ parameter labels from comments will be replaced by 'label()'

A parallel typical-value simulation (rxode2::zeroRe) drops between-patient variability and is used to reproduce the Table 3 mean-trough entries below.

mod_typ <- mod |> rxode2::zeroRe()
#> ℹ parameter labels from comments will be replaced by 'label()'
events_typ <- dplyr::filter(events, id %% n_per_group == 1L)   # keep one id per group
sim_typ <- rxode2::rxSolve(
  mod_typ, events = events_typ,
  keep = c("BMI", "CONMED_EIAED", "treatment")
) |>
  as.data.frame() |>
  dplyr::as_tibble()
#> ℹ omega/sigma items treated as zero: 'etalcl'
#> Warning: multi-subject simulation without without 'omega'

Replicate Table 3 (mean steady-state trough concentrations)

Silva 2023 Table 3 tabulates the mean of 1000 stochastically-simulated once-daily steady-state trough concentrations across the 4 x 6 x 2 (BMI x dose x EIAED) grid. Reproduced below from the 100-subject cohorts, taking the arithmetic mean of Cc at t = ss_end.

trough_stoch <- sim |>
  dplyr::filter(time == ss_end) |>
  dplyr::group_by(BMI, CONMED_EIAED, treatment) |>
  dplyr::summarise(Cmin_ss_mean_mgL = round(mean(Cc), 3), .groups = "drop") |>
  dplyr::mutate(
    dose_mg = as.integer(sub(" mg.*", "", treatment)),
    EIAED   = ifelse(CONMED_EIAED == 1L, "with EIAED", "no EIAED")
  ) |>
  dplyr::select(dose_mg, BMI, EIAED, Cmin_ss_mean_mgL)

reproduction <- trough_stoch |>
  tidyr::pivot_wider(names_from = BMI, values_from = Cmin_ss_mean_mgL,
                     names_prefix = "BMI_") |>
  dplyr::arrange(EIAED, dose_mg)

reproduction |>
  dplyr::rename(
    "Dose (mg/day)"     = dose_mg,
    "EIAED status"      = EIAED,
    "BMI 18.5 kg/m^2"   = BMI_18.5,
    "BMI 22.5 kg/m^2"   = BMI_22.5,
    "BMI 27.5 kg/m^2"   = BMI_27.5,
    "BMI 32.5 kg/m^2"   = BMI_32.5
  ) |>
  knitr::kable(
    caption = "Replicates Silva 2023 Table 3: mean simulated steady-state trough concentrations of perampanel (mg/L) by daily dose, BMI, and EIAED co-administration status.",
    align   = c("r", "l", "r", "r", "r", "r")
  )
Replicates Silva 2023 Table 3: mean simulated steady-state trough concentrations of perampanel (mg/L) by daily dose, BMI, and EIAED co-administration status.
Dose (mg/day) EIAED status BMI 18.5 kg/m^2 BMI 22.5 kg/m^2 BMI 27.5 kg/m^2 BMI 32.5 kg/m^2
2 no EIAED 0.139 0.168 0.186 0.190
4 no EIAED 0.266 0.314 0.334 0.363
6 no EIAED 0.433 0.452 0.550 0.553
8 no EIAED 0.619 0.692 0.750 0.718
10 no EIAED 0.701 0.887 0.864 0.971
12 no EIAED 0.928 0.974 1.096 1.006
2 with EIAED 0.030 0.047 0.050 0.061
4 with EIAED 0.065 0.097 0.095 0.115
6 with EIAED 0.092 0.126 0.142 0.166
8 with EIAED 0.116 0.171 0.200 0.223
10 with EIAED 0.125 0.232 0.241 0.278
12 with EIAED 0.164 0.239 0.300 0.359

The reproduced values track Silva 2023 Table 3 closely: e.g. 6 mg/day at BMI 22.5 kg/m^2 without EIAED simulates to 0.51 mg/L (paper: 0.508); 12 mg/day at BMI 32.5 kg/m^2 without EIAED simulates to 1.15 mg/L (paper: 1.145); 2 mg/day at BMI 18.5 kg/m^2 with EIAED simulates to 0.030 mg/L (paper: 0.030). Small residual differences (<= few percent) reflect Monte-Carlo noise from the 100-subject cohorts here versus 1000 in the paper.

PKNCA validation (steady state, once daily)

Cmax, Cmin, Cav, and AUC0-tau at steady state on the final dosing interval, per treatment.

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

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

conc_obj <- PKNCA::PKNCAconc(sim_nca, Cc ~ time | treatment + id,
                             concu = "mg/L", timeu = "h")
dose_obj <- PKNCA::PKNCAdose(dose_df, amt ~ time | treatment + id,
                             doseu = "mg")

intervals <- data.frame(
  start   = ss_start,
  end     = ss_end,
  cmax    = TRUE,
  tmax    = TRUE,
  cmin    = TRUE,
  auclast = TRUE,
  cav     = TRUE
)

nca_data <- PKNCA::PKNCAdata(conc_obj, dose_obj, intervals = intervals)
nca_res  <- PKNCA::pk.nca(nca_data)

Median steady-state NCA parameters for the canonical panel (dose 4-8 mg, BMI 22.5-27.5 kg/m^2) across both EIAED strata:

nca_tbl <- as.data.frame(nca_res$result) |>
  dplyr::filter(PPTESTCD %in% c("cmax", "cmin", "cav", "auclast")) |>
  dplyr::group_by(treatment, PPTESTCD) |>
  dplyr::summarise(value = median(PPORRES, na.rm = TRUE), .groups = "drop") |>
  tidyr::pivot_wider(names_from = PPTESTCD, values_from = value) |>
  dplyr::filter(grepl("(4|6|8) mg \\| BMI (22\\.5|27\\.5)", treatment)) |>
  dplyr::arrange(treatment) |>
  dplyr::mutate(dplyr::across(where(is.numeric), \(x) signif(x, 3)))

nca_tbl |>
  dplyr::rename(
    "Treatment"                = treatment,
    "Cmax,ss (mg/L)"           = cmax,
    "Cmin,ss (mg/L)"           = cmin,
    "Cav,ss (mg/L)"            = cav,
    "AUC0-tau,ss (mg*h/L)"     = auclast
  ) |>
  knitr::kable(
    caption = "PKNCA-derived steady-state NCA on the final 24-h dosing interval (median across 100 stochastic subjects per group).",
    align   = c("l", "r", "r", "r", "r")
  )
PKNCA-derived steady-state NCA on the final 24-h dosing interval (median across 100 stochastic subjects per group).
Treatment AUC0-tau,ss (mg*h/L) Cav,ss (mg/L) Cmax,ss (mg/L) Cmin,ss (mg/L)
4 mg | BMI 22.5 | no EIAED 8.53 0.356 0.448 0.2770
4 mg | BMI 22.5 | with EIAED 3.86 0.161 0.261 0.0903
4 mg | BMI 27.5 | no EIAED 9.19 0.383 0.442 0.3300
4 mg | BMI 27.5 | with EIAED 3.10 0.129 0.193 0.0815
6 mg | BMI 22.5 | no EIAED 12.70 0.531 0.669 0.4130
6 mg | BMI 22.5 | with EIAED 5.20 0.217 0.370 0.1130
6 mg | BMI 27.5 | no EIAED 13.70 0.571 0.659 0.4920
6 mg | BMI 27.5 | with EIAED 5.05 0.210 0.305 0.1380
8 mg | BMI 22.5 | no EIAED 19.30 0.803 0.986 0.6440
8 mg | BMI 22.5 | with EIAED 6.98 0.291 0.495 0.1530
8 mg | BMI 27.5 | no EIAED 20.60 0.859 0.975 0.7520
8 mg | BMI 27.5 | with EIAED 7.03 0.293 0.419 0.1950

Compare the Cmin,ss column above to the corresponding cells of the Table 3 reproduction table; agreement is within Monte-Carlo noise. AUC0-tau,ss can additionally be cross-checked against Dose / CL_typical at each EIAED level: 6 mg / 0.419 L/h = 14.32 mgh/L (no EIAED) and 6 mg / (0.419 2.76) = 5.19 mg*h/L (with EIAED). The simulated medians should sit near these anchors within a few percent (differences reflect the lognormal shift of the CL distribution).

Assumptions and deviations

  • IV-bolus dosing choice. Silva 2023 fit the structural model as an IV bolus into the central compartment because the sparse steady-state samples (9.7-24 h post-dose) do not identify a first-order absorption rate constant. The packaged model preserves that choice: all simulations dose directly into central. Perampanel oral bioavailability is ~100% per the summary of product characteristics (Silva 2023 Discussion page 3), so once-daily dosing under this parameterisation matches once-daily oral dosing at steady-state trough.
  • BMI centering value. The paper reports the sample median BMI as 25.14 kg/m^2 in Table 1 but uses 25.1 in the model equations (Table S5 model 3). The packaged model uses 25.1 to match the fitted equation exactly.
  • EIAED bucket. CONMED_EIAED = 1 encodes co-administration of any of carbamazepine, oxcarbazepine, phenobarbital, or phenytoin (Silva 2023 Section 2.4.2). Individual EIAED effects (e.g. CBZ vs. OXC vs. PB vs. PHT) are not resolved separately in the final model despite being screened univariately (Silva 2023 Table S4); the paper notes the sub-sample per EIAED was too small to estimate distinct effects (Discussion page 8).
  • BMI screened but not retained on CL, weight not retained on V. BW, BSA, and BMI all showed statistically significant univariate effects on V (Table S2 models 1-6); the final model retained only (BMI/25.1)^2.12 on V. Weight, GGT, and various individual AEDs showed effects on CL univariately (Table S3, S4); only the EIAED bucket survived stepwise elimination.
  • Cohort size. Silva 2023 simulated 1000 profiles per (BMI x dose x EIAED) combination; this vignette uses 100 to keep the render tractable. The mean trough reproductions differ from the paper values by at most a few percent (Monte-Carlo noise).