Skip to contents

Model and source

Weiss 2023 studied nebulized niclosamide ethanolamine (NEN), an inhaled formulation of the oral anthelminthic niclosamide, in healthy ewes. Plasma concentrations after three nebulized doses and one intravenous infusion were fitted simultaneously with a three-compartment population PK model in NONMEM (FOCE+I), which estimates the absolute bioavailability of the nebulized dose. Epithelial lining fluid (ELF) concentrations and the oral arm were analysed by non-compartmental methods only (the paper reports that no ELF model could be established), so they are not part of the model.

mod <- readModelDb("Weiss_2023_niclosamide_sheep")
mod_ui <- rxode2::rxode2(mod)
#> ℹ parameter labels from comments will be replaced by 'label()'
mod_ui
#>  ── rxode2-based free-form 4-cmt ODE model ────────────────────────────────────── 
#>  ── Initalization: ──  
#> Fixed Effects ($theta): 
#>        lka    lfdepot        lcl        lvc        lvp         lq       lvp2 
#>  3.0397492 -0.1531512  5.1357984  2.7536607  2.3125354  4.0500443  4.9972123 
#>        lq2     propSd      addSd 
#>  2.0108950  0.2520000  0.2600000 
#> 
#> Omega ($omega): 
#>          etalka   etalcl   etalvc    etalvp
#> etalka 0.455905 0.000000 0.000000 0.0000000
#> etalcl 0.000000 0.109392 0.000000 0.0000000
#> etalvc 0.000000 0.000000 0.408195 0.0000000
#> etalvp 0.000000 0.000000 0.000000 0.0754785
#> attr(,"lotriLabels")
#> [1] "Table S2 IIV ka = 76% CV" "Table S2 IIV CL = 34% CV"
#> [3] "Table S2 IIV V1 = 71% CV" "Table S2 IIV V2 = 28% CV"
#> attr(,"lotriFix")
#>        etalka etalcl etalvc etalvp
#> etalka  FALSE  FALSE  FALSE  FALSE
#> etalcl  FALSE  FALSE  FALSE  FALSE
#> etalvc  FALSE  FALSE  FALSE  FALSE
#> etalvp  FALSE  FALSE  FALSE  FALSE
#> 
#> States ($state or $stateDf): 
#>   Compartment Number Compartment Name
#> 1                  1            depot
#> 2                  2          central
#> 3                  3      peripheral1
#> 4                  4      peripheral2
#>  ── μ-referencing ($muRefTable): ──  
#>   theta    eta level
#> 1   lka etalka    id
#> 2   lcl etalcl    id
#> 3   lvc etalvc    id
#> 4   lvp etalvp    id
#> 
#>  ── Model (Normalized Syntax): ── 
#> function() {
#>     compartmentData <- list(depot = list(analyte = "niclosamide", 
#>         units = "mg", specimen = "administration site", verified = TRUE), 
#>         central = list(analyte = "niclosamide", units = "mg", 
#>             specimen = "plasma", verified = TRUE), peripheral1 = list(analyte = "niclosamide", 
#>             units = "mg", specimen = "plasma", verified = TRUE), 
#>         peripheral2 = list(analyte = "niclosamide", units = "mg", 
#>             specimen = "plasma", verified = TRUE))
#>     covariateData <- list()
#>     description <- "Preclinical (sheep). Three-compartment population PK model for plasma niclosamide in ewes after intravenous infusion and nebulized (inhaled) niclosamide ethanolamine, with first-order absorption from a lung absorption compartment, estimated nebulized bioavailability, linear elimination, and combined proportional and additive residual error (Weiss 2023)."
#>     population <- list(species = "sheep (Merino cross-bred ewes)", 
#>         n_subjects = 10L, n_studies = 1L, age_range = "12 months (all animals)", 
#>         weight_range = "25-35 kg (nebulized cohort, Methods 'Animals'); 40.6 +/- 2.1 kg (intravenous / oral cohort, Table I footnote)", 
#>         sex_female_pct = 100, disease_state = "Healthy", dose_range = "Nebulized niclosamide ethanolamine (NEN) 10, 30 and 60 mg (6 mL of 0.167%, 0.5% and 1% NEN) delivered over 24-27 min; intravenous NEN 3 mg (3 mL of 0.1% NEN) infused over 30 min (Table I, Table S1).", 
#>         regions = "Monash Animal Research Platform, Monash University, Australia.", 
#>         notes = "Two cohorts of five ewes (Table I). Experiment 1 (sheep 1-5) received nebulized placebo and the three ascending nebulized doses with 7-day washouts; Experiment 2 (sheep 11-14, 18) received an oral 1132 mg niclosamide dose and the intravenous dose. The population PK model was fitted to the plasma concentrations after the intravenous and nebulized doses only; the oral arm and the epithelial lining fluid concentrations were analysed by non-compartmental methods and are not part of this model. Dose nebulized administration with the DELIVERED dose (Table S1: 0.271, 0.762 and 1.401 mg/kg NEN), not the nominal nebulizer charge: the paper's own Figure S3 simulations sit 0.6-0.75 times below a nominal-dose simulation, matching the delivery loss (see vignette). Doses are in mg NEN as the paper reports them; the paper does not state whether a salt correction to niclosamide base was applied before modelling.")
#>     reference <- "Weiss A, Bischof RJ, Landersdorfer CB, Nguyen TH, Davies A, Ibrahim J, Wynne P, Wright P, Ditzinger G, Montgomery AB, Meeusen E, McIntosh MP, Sommer MOA. Single-dose pharmacokinetics and lung function of nebulized niclosamide ethanolamine in sheep. Pharm Res. 2023;40(8):1915-1925. doi:10.1007/s11095-023-03559-0"
#>     units <- list(time = "h", dosing = "mg", concentration = "ng/mL")
#>     vignette <- "Weiss_2023_niclosamide_sheep"
#>     ini({
#>         lka <- 3.03974915897077
#>         label("First-order absorption rate constant from the lung absorption compartment (1/h)")
#>         lfdepot <- -0.153151179494175
#>         label("Bioavailability of nebulized dosing (fraction)")
#>         lcl <- 5.13579843705026
#>         label("Systemic elimination clearance (L/h)")
#>         lvc <- 2.75366071235426
#>         label("Central volume of distribution (L)")
#>         lvp <- 2.31253542384721
#>         label("Shallow peripheral volume of distribution (L)")
#>         lq <- 4.05004430332552
#>         label("Distribution clearance, central to shallow peripheral (L/h)")
#>         lvp2 <- 4.99721227376411
#>         label("Deep peripheral volume of distribution (L)")
#>         lq2 <- 2.01089499914473
#>         label("Distribution clearance, central to deep peripheral (L/h)")
#>         propSd <- c(0, 0.252)
#>         label("Proportional residual error (fraction)")
#>         addSd <- c(0, 0.26)
#>         label("Additive residual error (ng/mL)")
#>         etalka ~ 0.455905
#>         label("Table S2 IIV ka = 76% CV")
#>         etalcl ~ 0.109392
#>         label("Table S2 IIV CL = 34% CV")
#>         etalvc ~ 0.408195
#>         label("Table S2 IIV V1 = 71% CV")
#>         etalvp ~ 0.0754785
#>         label("Table S2 IIV V2 = 28% CV")
#>     })
#>     model({
#>         ka <- exp(lka + etalka)
#>         fdepot <- exp(lfdepot)
#>         cl <- exp(lcl + etalcl)
#>         vc <- exp(lvc + etalvc)
#>         vp <- exp(lvp + etalvp)
#>         q <- exp(lq)
#>         vp2 <- exp(lvp2)
#>         q2 <- exp(lq2)
#>         kel <- cl/vc
#>         k12 <- q/vc
#>         k21 <- q/vp
#>         k13 <- q2/vc
#>         k31 <- q2/vp2
#>         d/dt(depot) <- -ka * depot
#>         d/dt(central) <- ka * depot - kel * central - k12 * central + 
#>             k21 * peripheral1 - k13 * central + k31 * peripheral2
#>         d/dt(peripheral1) <- k12 * central - k21 * peripheral1
#>         d/dt(peripheral2) <- k13 * central - k31 * peripheral2
#>         f(depot) <- fdepot
#>         Cc <- 1000 * central/vc
#>         Cc ~ prop(propSd) + add(addSd)
#>     })
#> }

Population

Ten 12-month-old Merino cross-bred ewes in two cohorts of five (Table I). The nebulized cohort (sheep 1-5, 25-35 kg) received a placebo and then 10, 30 and 60 mg NEN by an in-line eFlow nebulizer through an endotracheal tube, over 24-27 min, with 7-day washouts. The second cohort (sheep 11-14 and 18, 40.6 +/- 2.1 kg) received 1132 mg oral niclosamide and a 3 mg intravenous NEN infusion over 30 min. Plasma was sampled at 14 time points over 12 h (Table I). Only the nebulized and intravenous plasma data enter the population PK model.

Source trace

Model element Value Source
Structure: absorption compartment, central, shallow and deep peripheral – Methods, “Non-compartmental and compartmental Pharmacokinetic analysis”, equations for dAabs/dt, dA1/dt, dA2/dt, dA3/dt
Nebulized input: zero-order into the absorption compartment over the nebulization time Methods, same section (“time-delimited zero-order … input”)
IV input: zero-order into central 3 mL at 0.1 mL/min (30 min) Methods, “Niclosamide administration procedures”; Table S1
lka 20.9 1/h Table S2; Table II
lfdepot (Fneb) 85.8% Table S2 (Table II rounds to 86%)
lcl 170 L/h Table S2
lvc (V1) 15.7 L Table S2
lvp (V2, shallow) 10.1 L Table S2
lq (CLd1) 57.4 L/h Table S2
lvp2 (V3, deep) 148 L Table S2
lq2 (CLd2) 7.47 L/h Table S2
etalka 76% CV Table S2
etalcl 34% CV Table S2
etalvc 71% CV Table S2
etalvp 28% CV Table S2
propSd 25.2% Table S2
addSd 0.26 ng/mL Table S2
Half-lives (validation target) 0.041, 0.2, 14 h Table II; Results text
Plasma NCA (validation target) Cmax, Tmax, AUC0-t Table II

Typical-value checks

Disposition half-lives

Table II reports half-lives of 0.041 h (alpha), 0.2 h (beta) and 14 h (terminal). These are the eigenvalues of the typical-value disposition matrix, so they follow from CL, V1, V2, V3, CLd1 and CLd2 alone.

th <- exp(mod_ui$theta)
kel <- th[["lcl"]] / th[["lvc"]]
k12 <- th[["lq"]] / th[["lvc"]]
k21 <- th[["lq"]] / th[["lvp"]]
k13 <- th[["lq2"]] / th[["lvc"]]
k31 <- th[["lq2"]] / th[["lvp2"]]
disp <- matrix(
  c(-(kel + k12 + k13), k21, k31,
    k12, -k21, 0,
    k13, 0, -k31),
  nrow = 3, byrow = TRUE
)
thalf <- sort(log(2) / -eigen(disp)$values)
data.frame(
  phase = c("alpha", "beta", "terminal"),
  model_h = signif(thalf, 3),
  published_h = c(0.041, 0.2, 14)
) |>
  knitr::kable(caption = "Typical-value half-lives versus Table II.")
Typical-value half-lives versus Table II.
phase model_h published_h
alpha 0.0412 0.041
beta 0.1810 0.200
terminal 14.3000 14.000

# The parameters are deterministic here, so the comparison is a rounding check.
stopifnot(
  abs(thalf[1] - 0.041) < 0.0005,
  abs(thalf[2] - 0.2) < 0.025,
  abs(thalf[3] - 14) < 0.5
)

Nebulized bioavailability and dose linearity

The dose-normalised typical-value AUC0-inf of a nebulized dose relative to the intravenous dose recovers Fneb, and the model is linear in dose (the paper reports dose-linear plasma PK across 10-60 mg NEN, Fig. S3).

mod_typ <- rxode2::zeroRe(mod_ui)
ev_typ <- rbind(
  data.frame(id = 1L, time = 0, amt = 10, dur = 27 / 60, evid = 1L, cmt = "depot"),
  data.frame(id = 2L, time = 0, amt = 60, dur = 25 / 60, evid = 1L, cmt = "depot"),
  data.frame(id = 3L, time = 0, amt = 3, dur = 0.5, evid = 1L, cmt = "central")
)
obs_typ <- expand.grid(id = 1:3, time = c(seq(0, 2, by = 0.005), seq(2.05, 400, by = 0.05)))
obs_typ <- transform(obs_typ, amt = 0, dur = 0, evid = 0L, cmt = "central")
sim_typ <- rxode2::rxSolve(mod_typ, rbind(ev_typ, obs_typ), returnType = "data.frame")
#> ℹ omega/sigma items treated as zero: 'etalka', 'etalcl', 'etalvc', 'etalvp'
#> Warning: multi-subject simulation without without 'omega'

auc_typ <- sim_typ |>
  dplyr::group_by(id) |>
  dplyr::summarise(
    cmax = max(Cc),
    auc = sum(diff(time) * (head(Cc, -1) + tail(Cc, -1)) / 2)
  ) |>
  dplyr::mutate(dose = c(10, 60, 3), auc_per_mg = auc / dose)
f_neb <- auc_typ$auc_per_mg[1] / auc_typ$auc_per_mg[3]
knitr::kable(auc_typ, digits = 3, caption = "Typical-value exposure (AUC to 400 h).")
Typical-value exposure (AUC to 400 h).
id cmax auc dose auc_per_mg
1 97.327 50.471 10 5.047
2 621.324 302.824 60 5.047
3 31.696 17.647 3 5.882
f_neb
#> [1] 0.8579998

stopifnot(
  # AUC(nebulized)/AUC(IV), dose-normalised, equals Fneb = 0.858.
  abs(f_neb - 0.858) < 0.01,
  # Linear model: 60 mg and 10 mg give the same dose-normalised AUC.
  abs(auc_typ$auc_per_mg[2] / auc_typ$auc_per_mg[1] - 1) < 0.01
)

Which dose to enter: nominal or delivered

The nebulizer was charged with 10, 30 or 60 mg NEN, but drug left in the nebulizer, lines and expiratory filters was rinsed out and assayed so that the actual delivered dose could be calculated (Methods). Table S1 reports the delivered doses as 0.271, 0.762 and 1.401 mg/kg; for ewes of 25-35 kg that is roughly 70-80% of the nominal charge. The intravenous dose (0.073 mg/kg in 40.6 kg ewes, 2.96 mg) has no such loss.

The paper does not say which of the two entered the modelling data set, but its own model simulations settle it. The median (P50) curves of Figure S3 were digitised by the maintainers at 2-5 h, where they are clear of the observed points, and compared with the typical-value prediction for the nominal doses:

# Figure S3 P50, digitised by the maintainers against the 0.5 ng/mL LLOQ line
# (which reads back as 0.495 ng/mL). Time is from the start of nebulization.
fig_s3 <- data.frame(
  nominal_mg = c(10, 10, 10, 10, 30, 30, 30, 60, 60, 60, 60),
  time = c(2, 3, 4, 4.9, 3, 4, 4.9, 2, 3, 4, 4.9),
  p50_fig = c(0.18, 0.075, 0.062, 0.062, 0.20, 0.16, 0.16, 0.95, 0.41, 0.35, 0.33)
)
neb_dur <- c(`10` = 27, `30` = 24, `60` = 25) / 60
ev_nom <- do.call(rbind, lapply(seq_along(neb_dur), function(i) {
  rbind(
    data.frame(id = i, time = 0, amt = as.numeric(names(neb_dur)[i]),
               dur = neb_dur[[i]], evid = 1L, cmt = "depot"),
    data.frame(id = i, time = c(2, 3, 4, 4.9), amt = 0, dur = 0, evid = 0L,
               cmt = "central")
  )
}))
typ_nom <- rxode2::rxSolve(mod_typ, ev_nom, returnType = "data.frame") |>
  dplyr::mutate(nominal_mg = as.numeric(names(neb_dur))[id]) |>
  dplyr::select(nominal_mg, time, Cc_nominal = Cc)
#> ℹ omega/sigma items treated as zero: 'etalka', 'etalcl', 'etalvc', 'etalvp'
#> Warning: multi-subject simulation without without 'omega'
dose_basis <- dplyr::inner_join(fig_s3, typ_nom, by = c("nominal_mg", "time")) |>
  dplyr::mutate(ratio = p50_fig / Cc_nominal)
knitr::kable(dose_basis, digits = 3,
             caption = "Figure S3 P50 versus the typical-value prediction for the nominal nebulizer charge.")
Figure S3 P50 versus the typical-value prediction for the nominal nebulizer charge.
nominal_mg time p50_fig Cc_nominal ratio
10 2.0 0.180 0.219 0.823
10 3.0 0.075 0.094 0.798
10 4.0 0.062 0.087 0.712
10 4.9 0.062 0.083 0.745
30 3.0 0.200 0.281 0.713
30 4.0 0.160 0.261 0.614
30 4.9 0.160 0.250 0.641
60 2.0 0.950 1.254 0.757
60 3.0 0.410 0.562 0.729
60 4.0 0.350 0.522 0.671
60 4.9 0.330 0.499 0.661
summary(dose_basis$ratio)
#>    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
#>  0.6136  0.6659  0.7125  0.7149  0.7509  0.8233

# The paper's simulations sit well below a nominal-dose simulation in every
# dose group, so the data set did not carry the nominal charge.
stopifnot(median(dose_basis$ratio) < 0.85)

Figure S3 lies 0.6-0.75 times below the nominal-dose prediction in all three groups and at every time, which is the size of the delivery loss. The model should therefore be dosed with the delivered nebulized dose. The simulations below use the Table S1 mean delivered dose in mg/kg multiplied by 30 kg, the middle of the 25-35 kg weight range of the nebulized cohort.

Virtual cohort and simulation

One hundred virtual ewes per arm receive each of the four modelled regimens. Nebulization times are the Table S1 means (27, 24 and 25 min for the 10, 30 and 60 mg doses); the IV dose is a 30 min infusion. Plasma is sampled on the Table I schedule: before dosing, at 7.5 and 15 min during dosing, and at 0, 5, 15 and 30 min and 1, 2, 3, 4, 6, 8 and 12 h after the end of dosing. A dense grid is added for the figure.

arms <- data.frame(
  treatment = c("0.167% NEN nebulized", "0.5% NEN nebulized",
                "1% NEN nebulized", "0.1% NEN intravenous"),
  nominal_mg = c(10, 30, 60, 3),
  # Table S1 delivered dose (mg/kg) x 30 kg for the nebulized cohort and
  # x 40.6 kg (Table I footnote) for the intravenous cohort.
  amt = c(0.271 * 30, 0.762 * 30, 1.401 * 30, 0.073 * 40.6),
  dur_min = c(27, 24, 25, 30),
  cmt = c("depot", "depot", "depot", "central")
)
knitr::kable(arms, digits = 2, caption = "Simulated regimens (amt = delivered dose, mg NEN).")
Simulated regimens (amt = delivered dose, mg NEN).
treatment nominal_mg amt dur_min cmt
0.167% NEN nebulized 10 8.13 27 depot
0.5% NEN nebulized 30 22.86 24 depot
1% NEN nebulized 60 42.03 25 depot
0.1% NEN intravenous 3 2.96 30 central
arms$treatment <- factor(arms$treatment, levels = arms$treatment)
n_per_arm <- 100L
post_end_h <- c(0, 5, 15, 30, 60, 120, 180, 240, 360, 480, 720) / 60

make_arm <- function(i, n, id_offset) {
  a <- arms[i, ]
  ids <- id_offset + seq_len(n)
  dur <- a$dur_min / 60
  sched <- sort(unique(c(0, 7.5 / 60, 15 / 60, dur + post_end_h)))
  grid <- seq(0, dur + 12, by = 1 / 60)
  sched <- round(sched, 6)
  obs_times <- sort(unique(round(c(sched, grid), 6)))
  dose <- data.frame(id = ids, time = 0, amt = a$amt, dur = dur, evid = 1L,
                     cmt = a$cmt, treatment = a$treatment, scheduled = FALSE)
  obs <- expand.grid(id = ids, time = obs_times)
  obs$amt <- 0
  obs$dur <- 0
  obs$evid <- 0L
  obs$cmt <- "central"
  obs$treatment <- a$treatment
  obs$scheduled <- obs$time %in% sched
  rbind(dose, obs)
}
events <- do.call(rbind, lapply(seq_len(nrow(arms)), function(i) {
  make_arm(i, n_per_arm, (i - 1L) * n_per_arm)
}))
events <- events[order(events$id, events$time, -events$evid), ]
stopifnot(!anyDuplicated(events[events$evid == 1, "id"]))

rxode2::rxSetSeed(20230727)
sim <- rxode2::rxSolve(mod_ui, events, keep = c("treatment", "scheduled"),
                       returnType = "data.frame")

Replicate Figure 2A

llq <- 0.5
sim |>
  dplyr::group_by(treatment, time) |>
  dplyr::summarise(
    p05 = quantile(Cc, 0.05), p50 = median(Cc), p95 = quantile(Cc, 0.95),
    .groups = "drop"
  ) |>
  dplyr::filter(time > 0, time <= 3) |>
  ggplot(aes(time, p50, colour = treatment, fill = treatment)) +
  geom_ribbon(aes(ymin = p05, ymax = p95), alpha = 0.15, colour = NA) +
  geom_line() +
  geom_hline(yintercept = llq, linetype = "dashed") +
  scale_y_log10() +
  labs(x = "Time after start of dosing (h)", y = "Plasma niclosamide (ng/mL)",
       colour = NULL, fill = NULL) +
  theme_bw() +
  theme(legend.position = "bottom")

Replicates Figure 2A (semilog panel) of Weiss 2023: median and 90% prediction interval of plasma niclosamide by dose group; dashed line is the 0.5 ng/mL LLOQ. As in the paper, concentrations peak at the end of nebulization, fall steeply over the first 1-2 h and approach the LLOQ by 1-2 h for the 10 mg dose, with the IV profile tracking the 10 mg nebulized profile.

Replicate Figure S3

sim_p50 <- sim |>
  dplyr::filter(treatment != "0.1% NEN intravenous") |>
  dplyr::mutate(nominal_mg = arms$nominal_mg[match(treatment, arms$treatment)]) |>
  dplyr::group_by(nominal_mg, treatment, time) |>
  dplyr::summarise(
    P10 = quantile(Cc, 0.1), P50 = median(Cc), P90 = quantile(Cc, 0.9),
    .groups = "drop"
  )
fig_s3_lab <- fig_s3 |>
  dplyr::mutate(treatment = arms$treatment[match(nominal_mg, arms$nominal_mg)])
sim_p50 |>
  dplyr::filter(time > 0, time <= 5) |>
  tidyr::pivot_longer(c(P10, P50, P90), names_to = "percentile", values_to = "Cc") |>
  ggplot(aes(time, Cc)) +
  geom_line(aes(linetype = percentile)) +
  geom_point(data = fig_s3_lab, aes(time, p50_fig), colour = "red") +
  geom_hline(yintercept = llq, linetype = "dotted") +
  facet_wrap(~treatment) +
  scale_y_log10() +
  labs(x = "Time (h)", y = "Plasma niclosamide (ng/mL)", linetype = NULL) +
  theme_bw() +
  theme(legend.position = "bottom")


s3_check <- dplyr::inner_join(
  fig_s3,
  sim_p50 |> dplyr::mutate(time = round(time, 4)),
  by = c("nominal_mg", "time")
) |>
  dplyr::mutate(ratio = P50 / p50_fig)
knitr::kable(dplyr::select(s3_check, nominal_mg, time, p50_fig, P50, ratio), digits = 3,
             caption = "Simulated P50 (delivered doses) versus the digitised Figure S3 P50.")
Simulated P50 (delivered doses) versus the digitised Figure S3 P50.
nominal_mg time p50_fig P50 ratio
10 2.0 0.180 0.217 1.206
10 3.0 0.075 0.084 1.124
10 4.0 0.062 0.070 1.136
10 4.9 0.062 0.065 1.050
30 3.0 0.200 0.255 1.274
30 4.0 0.160 0.226 1.412
30 4.9 0.160 0.215 1.346
60 2.0 0.950 1.108 1.166
60 3.0 0.410 0.476 1.161
60 4.0 0.350 0.414 1.182
60 4.9 0.330 0.393 1.192

stopifnot(
  nrow(s3_check) == nrow(fig_s3),
  abs(log(median(s3_check$ratio))) < log(1.3)
)

Replicates Figure S3 of Weiss 2023 (P10, P50 and P90 of the model simulations by dose group); red points are the digitised published P50. With the delivered doses the simulated median comes within about 20% of the published one (median ratio near 1.2). The remaining gap is the size of either a lighter cohort (about 25 kg rather than 30 kg) or a salt-to-base correction of the doses (factor 0.84), which this comparison cannot separate.

PKNCA validation

The paper’s NCA (PKSolver) used the scheduled samples above the LLOQ, so the simulated individual predictions are taken on the Table I schedule and samples below 0.5 ng/mL are dropped before the NCA. The pre-dose sample is kept as zero.

sim_nca <- sim |>
  dplyr::filter(scheduled) |>
  dplyr::mutate(Cc = ifelse(time > 0 & Cc < llq, NA_real_, Cc)) |>
  dplyr::filter(!is.na(Cc)) |>
  dplyr::select(id, time, Cc, treatment)
stopifnot(all(sim_nca$Cc[sim_nca$time == 0] == 0))
stopifnot(length(unique(sim_nca$id[sim_nca$time == 0])) == nrow(arms) * n_per_arm)

conc_obj <- PKNCA::PKNCAconc(sim_nca, Cc ~ time | treatment + id)
dose_df <- events |>
  dplyr::filter(evid == 1) |>
  dplyr::select(id, time, amt, dur, treatment)
dose_obj <- PKNCA::PKNCAdose(dose_df, amt ~ time | treatment + id, duration = "dur")

intervals <- data.frame(start = 0, end = Inf, cmax = TRUE, tmax = TRUE, auclast = TRUE)
nca_res <- PKNCA::pk.nca(PKNCA::PKNCAdata(conc_obj, dose_obj, intervals = intervals))

Comparison against published NCA

Table II reports the mean of the observed NCA parameters for the nebulized groups (Tmax converted from min to h below); no plasma NCA is tabulated for the intravenous arm. ncaComparisonTable() summarises the simulated cohort by its median.

published <- data.frame(
  treatment = c("0.167% NEN nebulized", "0.5% NEN nebulized", "1% NEN nebulized"),
  cmax = c(76.6, 239.8, 577.1),
  tmax = c(27.1, 24.3, 22.2) / 60,
  auclast = c(35.9, 96.1, 261.4)
)

cmp <- nlmixr2lib::ncaComparisonTable(
  simulated = nca_res,
  reference = published,
  by = "treatment",
  units = c(cmax = "ng/mL", tmax = "h", auclast = "ng*h/mL"),
  tolerance_pct = 20
)
knitr::kable(cmp, caption = "Simulated (median) vs. published (mean, Table II) plasma NCA. * differs from reference by >20%.")
Simulated (median) vs. published (mean, Table II) plasma NCA. * differs from reference by >20%.
NCA parameter treatment Reference Simulated % diff
Cmax (ng/mL) 0.167% NEN nebulized 76.6 70.8 -7.5%
Cmax (ng/mL) 0.5% NEN nebulized 240 225 -6.0%
Cmax (ng/mL) 1% NEN nebulized 577 433 -25.0%*
Tmax (h) 0.167% NEN nebulized 0.452 0.45 -0.4%
Tmax (h) 0.5% NEN nebulized 0.405 0.4 -1.2%
Tmax (h) 1% NEN nebulized 0.37 0.417 +12.6%
AUClast (ng*h/mL) 0.167% NEN nebulized 35.9 38 +5.8%
AUClast (ng*h/mL) 0.5% NEN nebulized 96.1 116 +20.5%*
AUClast (ng*h/mL) 1% NEN nebulized 261 212 -19.0%
sim_med <- as.data.frame(nca_res$result) |>
  dplyr::filter(PPTESTCD %in% c("cmax", "auclast"),
                treatment %in% published$treatment) |>
  dplyr::group_by(treatment, PPTESTCD) |>
  dplyr::summarise(med = median(PPORRES), .groups = "drop") |>
  tidyr::pivot_wider(names_from = PPTESTCD, values_from = med) |>
  dplyr::left_join(published, by = "treatment", suffix = c("_sim", "_pub"))
sim_med
#> # A tibble: 3 × 6
#>   treatment            auclast_sim cmax_sim cmax_pub  tmax auclast_pub
#>   <chr>                      <dbl>    <dbl>    <dbl> <dbl>       <dbl>
#> 1 0.167% NEN nebulized        38.0     70.8     76.6 0.452        35.9
#> 2 0.5% NEN nebulized         116.     225.     240.  0.405        96.1
#> 3 1% NEN nebulized           212.     433.     577.  0.37        261.

# Structural check on the centre of the distribution: a mis-transcribed CL,
# volume, Fneb or dose unit moves these ratios by far more than the band.
stopifnot(
  all(sim_med$cmax_sim / sim_med$cmax_pub > 0.67 & sim_med$cmax_sim / sim_med$cmax_pub < 1.5),
  all(sim_med$auclast_sim / sim_med$auclast_pub > 0.67 & sim_med$auclast_sim / sim_med$auclast_pub < 1.5)
)

Rows starred in the table (typically the 1% Cmax, about 25% low, and the 0.5% AUC0-t, about 20% high) go in opposite directions across groups: the observed means are not dose-proportional (Cmax per mg NEN 7.7, 8.0 and 9.6; AUC0-t per mg 3.6, 3.2 and 4.4), whereas the model is linear, as the paper concluded from its population analysis. The simulated medians are compared with observed arithmetic means of five sheep per group, whose standard deviations in Table II are 20-38% of the mean for Cmax and 29-32% for AUC0-t, and from which eight implausible data points were excluded before the NCA (Fig. 2 legend). The simulated doses also rest on an assumed 30 kg body weight. Differences of the order of 20-30% are therefore within the precision of the published values.

Assumptions and deviations

  • Delivered, not nominal, nebulized dose. The paper does not state which dose entered the modelling data set. Its own Figure S3 simulations sit 0.6-0.75 times below a nominal-dose simulation in every group, matching the delivery loss implied by Table S1, so the model is dosed with the delivered dose. The individual body weights of the nebulized cohort are not reported; the vignette uses 30 kg, the middle of the stated 25-35 kg range.
  • Salt basis. Doses are entered in mg of niclosamide ethanolamine (NEN), the unit the paper reports for every dose (Table I, Table S1). The paper does not state a salt correction before modelling. If the data set used niclosamide-equivalent doses instead (NEN is about 84% niclosamide by mass), CL and the volumes would be about 16% lower on a NEN-dose basis; Fneb is unaffected because both routes share the same basis. The Figure S3 comparison cannot separate the two, because the unknown body weights (25-35 kg) move the delivered dose by more than 16%.
  • IIV scale. Table S2 reports IIV as %CV for a log-normal model; the variances are omega^2 = log(CV^2 + 1). Table S2 lists no covariance, so the etas are independent (the paper tested a CL-V covariance during model building).
  • Bioavailability placement. The paper writes Fneb on the absorption-to-central flux (ka * Fneb * Aabs). The model applies it to the nebulized dose (f(depot)), which gives identical central, peripheral and plasma amounts.
  • Nebulization time. The zero-order nebulized input uses the Table S1 mean delivery time per dose group in this vignette; the paper used the individual recorded delivery times.
  • Out of scope. No ELF model was established and the oral arm was not modelled (Results; Discussion), so neither appears here. The deep peripheral parameters had RSE above 100% (Table S2 footnote); the paper cautions against extrapolating beyond 3 h for regimens outside those studied.
  • Errata. No correction notice was found for this article (Europe PMC, checked 2026-10-27).