Dihydroartemisinin + piperaquine in pregnant and non-pregnant women with uncomplicated malaria (Tarning 2012) • nlmixr2lib

Models and source

Tarning et al. (2012) simultaneously developed two independent population PK models – one for piperaquine (3-compartment disposition with 5 transit compartments) and one for dihydroartemisinin (1-compartment disposition with 7 transit compartments) – from a single matched cohort of 24 pregnant and 24 non-pregnant women with uncomplicated malaria on the Thai-Myanmar border who received the fixed-dose oral dihydroartemisinin-piperaquine combination once daily for 3 days.

Citation: Tarning J, Rijken MJ, McGready R, Phyo AP, Hanpithakpong W, Day NPJ, White NJ, Nosten F, Lindegardh N (2012). Population pharmacokinetics of dihydroartemisinin and piperaquine in pregnant and nonpregnant women with uncomplicated malaria. Antimicrobial Agents and Chemotherapy 56(4):1997-2007.
Article: https://doi.org/10.1128/AAC.05756-11
Models in nlmixr2lib: Tarning_2012_piperaquine and Tarning_2012_dihydroartemisinin.

mod_pip <- nlmixr2lib::readModelDb("Tarning_2012_piperaquine")
mod_dha <- nlmixr2lib::readModelDb("Tarning_2012_dihydroartemisinin")

Population

Tarning 2012 enrolled 24 pregnant women in the second or third trimester (estimated gestational age 13.1-33.4 weeks, median 25.3) plus 24 age-, smoking- and parasitaemia-matched non-pregnant women, all with uncomplicated Plasmodium falciparum malaria (or mixed P. falciparum + P. vivax) at the Wang Pha Clinic of the Shoklo Malaria Research Unit on the Thai-Myanmar border. Pregnant median weight 51 kg (range 36-58); non-pregnant median 48 kg (range 37-78). Admission parasitaemia: pregnant median 6,780 (range 96-177,000) parasites/uL; non-pregnant median 9,670 (range 32-136,000). All women received the standard fixed-dose oral dihydroartemisinin-piperaquine combination tablet (40 mg dihydroartemisinin + 320 mg piperaquine tetraphosphate, equivalent to 184 mg piperaquine base per tablet; Holleypharm, People’s Republic of China) once daily for 3 days at 0, 24, and 48 h, divided to the nearest quarter tablet based on body weight. The per-cohort total piperaquine base dose was 30.8 (28.3-33.8) mg/kg in pregnant and 29.0 (27.7-33.8) mg/kg in non-pregnant women; dihydroartemisinin total dose 6.67 (6.12-7.32) mg/kg in pregnant and 6.28 (6.00-7.32) mg/kg in non-pregnant. See Tarning 2012 Table 1.

Source trace

The per-parameter origin is recorded as an in-file comment next to each ini() entry in inst/modeldb/specificDrugs/Tarning_2012_piperaquine.R and inst/modeldb/specificDrugs/Tarning_2012_dihydroartemisinin.R. The table below collects the published structural quantities in one place for review.

Drug	Parameter	Value	Source location
Piperaquine	CL/F	60.2 L/h	Tarning 2012 Table 2
Piperaquine	Vc/F	3,070 L	Tarning 2012 Table 2
Piperaquine	Q1/F	427 L/h	Tarning 2012 Table 2
Piperaquine	Vp1/F	4,440 L	Tarning 2012 Table 2
Piperaquine	Q2/F	160 L/h	Tarning 2012 Table 2
Piperaquine	Vp2/F	31,400 L	Tarning 2012 Table 2
Piperaquine	MTT	2.04 h	Tarning 2012 Table 2
Piperaquine	n_transit	5 (fixed)	Tarning 2012 Table 2; Fig 1A
Piperaquine	F	1 (fixed)	Tarning 2012 Table 2
Piperaquine	sigma (log)	0.285	Tarning 2012 Table 2
Piperaquine	IIV CL CV	21.5%	Tarning 2012 Table 2
Piperaquine	IIV Vc CV	39.5%	Tarning 2012 Table 2
Piperaquine	BOV MTT CV	45.8%	Tarning 2012 Table 2
Piperaquine	BOV F CV	56.3%	Tarning 2012 Table 2
Piperaquine	Pregnancy on CL	+45.0%	Tarning 2012 Table 2; Results
Piperaquine	Pregnancy on F	+46.8%	Tarning 2012 Table 2; Results
Piperaquine	Structural ODEs	depot -> 5 transit -> central <-> peripheral1, central <-> peripheral2	Fig 1A; Methods
Dihydroartem.	CL/F (48.5 kg)	78.0 L/h	Tarning 2012 Table 4
Dihydroartem.	V/F (48.5 kg)	129 L	Tarning 2012 Table 4
Dihydroartem.	MTT	0.982 h	Tarning 2012 Table 4
Dihydroartem.	n_transit	7 (fixed)	Tarning 2012 Table 4; Fig 1B
Dihydroartem.	F	1 (fixed)	Tarning 2012 Table 4
Dihydroartem.	sigma (log)	0.580	Tarning 2012 Table 4
Dihydroartem.	Allometric exp CL	3/4 (fixed)	Methods + Results
Dihydroartem.	Allometric exp V	1 (fixed)	Methods + Results
Dihydroartem.	IIV V CV	12.8%	Tarning 2012 Table 4
Dihydroartem.	IIV F CV	30.3%	Tarning 2012 Table 4
Dihydroartem.	BOV MTT CV	50.9%	Tarning 2012 Table 4
Dihydroartem.	Pregnancy on F	-37.5%	Tarning 2012 Table 4; Results
Dihydroartem.	log10(PARA) on F	+27.8% per log10 (centered at 3.98)	Tarning 2012 Table 4; Results
Dihydroartem.	Structural ODEs	depot -> 7 transit -> central -> out	Fig 1B; Methods

Virtual cohort

Original observed data are not publicly available. Construct 100 virtual subjects per pregnancy group to roughly match Table 1 (median 51 kg pregnant and 48 kg non-pregnant women; admission parasitaemia 6,780 and 9,670 parasites/uL).

set.seed(2012)

n_per_arm <- 100L

build_one <- function(id, wt, preg, para, dose_mg_dha, dose_mg_pip,
                      tobs, cohort_label) {
  occ_for_time <- function(t) pmin(3L, as.integer(floor(t / 24)) + 1L)
  ev <- rxode2::et(amt = dose_mg_pip, ii = 24, until = 48, cmt = "depot") |>
    rxode2::et(tobs)
  ev <- as.data.frame(ev)
  ev$id    <- id
  ev$WT    <- wt
  ev$PREG  <- preg
  ev$PARA  <- para
  ev$OCC   <- occ_for_time(ev$time)
  ev$cohort <- cohort_label
  ev$dose_dha_mg <- dose_mg_dha
  ev$dose_pip_mg <- dose_mg_pip
  ev
}

# Build cohorts with disjoint id ranges (events list will be re-used for
# both models via the cohort_label column).
make_cohort <- function(n, wt_med, wt_sd, preg, para_med, para_gsd,
                        dose_pip_mgkg, dose_dha_mgkg, tobs,
                        cohort_label, id_offset = 0L) {
  ids <- id_offset + seq_len(n)
  wt  <- pmax(35, rnorm(n, wt_med, wt_sd))
  para <- pmax(50, round(exp(log(para_med) + log(para_gsd) * rnorm(n))))
  do.call(rbind, lapply(seq_len(n), function(i) {
    dose_pip_per <- dose_pip_mgkg * wt[i] / 3
    dose_dha_per <- dose_dha_mgkg * wt[i] / 3
    build_one(ids[i], wt[i], preg, para[i], dose_dha_per, dose_pip_per,
              tobs, cohort_label)
  }))
}

tobs_short <- c(0, 0.5, 1, 1.5, 2, 4, 8, 16, 24.5, 25.5, 28, 32,
                48.25, 48.5, 49, 50, 51, 52, 54, 56, 60, 72)
tobs_long  <- c(tobs_short, 24 * c(5, 7, 14, 21, 28, 35, 42, 49, 56, 63, 77, 84))

events_pip <- dplyr::bind_rows(
  make_cohort(n_per_arm, wt_med = 51, wt_sd = 5,
              preg = 1, para_med = 6780, para_gsd = 6,
              dose_pip_mgkg = 30.8, dose_dha_mgkg = 6.67,
              tobs = tobs_long, cohort_label = "Pregnant",     id_offset = 0L),
  make_cohort(n_per_arm, wt_med = 48, wt_sd = 7,
              preg = 0, para_med = 9670, para_gsd = 6,
              dose_pip_mgkg = 29.0, dose_dha_mgkg = 6.28,
              tobs = tobs_long, cohort_label = "Non-pregnant", id_offset = n_per_arm)
)
stopifnot(!anyDuplicated(unique(events_pip[, c("id", "time", "evid")])))

events_dha <- dplyr::bind_rows(
  make_cohort(n_per_arm, wt_med = 51, wt_sd = 5,
              preg = 1, para_med = 6780, para_gsd = 6,
              dose_pip_mgkg = 30.8, dose_dha_mgkg = 6.67,
              tobs = tobs_short, cohort_label = "Pregnant",     id_offset = 0L),
  make_cohort(n_per_arm, wt_med = 48, wt_sd = 7,
              preg = 0, para_med = 9670, para_gsd = 6,
              dose_pip_mgkg = 29.0, dose_dha_mgkg = 6.28,
              tobs = tobs_short, cohort_label = "Non-pregnant", id_offset = n_per_arm)
)
stopifnot(!anyDuplicated(unique(events_dha[, c("id", "time", "evid")])))

# For DHA, dose amounts must reflect the dihydroartemisinin dose, not the
# piperaquine base dose (the event tables above default amt to dose_mg_pip);
# overwrite the dose rows with DHA-specific amounts before passing to the
# DHA model.
events_dha$amt[events_dha$evid == 1L] <-
  events_dha$dose_dha_mg[events_dha$evid == 1L]

Simulation

Use the typical-value model (no random effects) for the figure overlays; keep a stochastic copy for the NCA comparison.

set.seed(2012)

mod_pip_typ <- rxode2::zeroRe(mod_pip())
sim_pip_typ <- rxode2::rxSolve(
  mod_pip_typ, events = events_pip,
  keep = c("cohort", "PREG", "WT", "PARA", "OCC")
) |> as.data.frame()

mod_dha_typ <- rxode2::zeroRe(mod_dha())
sim_dha_typ <- rxode2::rxSolve(
  mod_dha_typ, events = events_dha,
  keep = c("cohort", "PREG", "WT", "PARA", "OCC")
) |> as.data.frame()

# Stochastic simulation (random IIV / BOV / residual draws) for NCA comparison
sim_pip_sto <- rxode2::rxSolve(
  mod_pip(), events = events_pip,
  keep = c("cohort", "PREG", "WT", "PARA", "OCC")
) |> as.data.frame()

sim_dha_sto <- rxode2::rxSolve(
  mod_dha(), events = events_dha,
  keep = c("cohort", "PREG", "WT", "PARA", "OCC")
) |> as.data.frame()

Replicate published figures

Tarning 2012 Figure 5 shows simulated population mean concentration-time curves for piperaquine and dihydroartemisinin in pregnant and non-pregnant women. The typical-value simulations below reproduce the qualitative shape of those curves; small differences are expected because (a) the published figure pools 2,000 Monte-Carlo simulations rather than typical-value predictions and (b) the empirical between-occasion variability draws across the cohort drift positively over the three doses (see Errata).

sim_pip_typ |>
  dplyr::filter(time <= 24 * 60) |>
  ggplot(aes(time / 24, Cc, group = id, colour = cohort)) +
  geom_line(alpha = 0.5) +
  scale_y_log10(limits = c(1, 1e3)) +
  scale_colour_manual(values = c("Pregnant" = "#1b9e77",
                                  "Non-pregnant" = "#d95f02")) +
  labs(x = "Time (days)", y = "Piperaquine plasma concentration (ng/mL)",
       colour = NULL,
       title = "Figure 5A -- Piperaquine typical-value profiles",
       caption = "Replicates Figure 5A of Tarning 2012 AAC (typical-value approximation).")

sim_dha_typ |>
  dplyr::filter(time <= 72) |>
  ggplot(aes(time, Cc, group = id, colour = cohort)) +
  geom_line(alpha = 0.5) +
  scale_y_log10(limits = c(1, 1e3)) +
  scale_colour_manual(values = c("Pregnant" = "#1b9e77",
                                  "Non-pregnant" = "#d95f02")) +
  labs(x = "Time (h)", y = "Dihydroartemisinin plasma concentration (ng/mL)",
       colour = NULL,
       title = "Figure 5B -- Dihydroartemisinin typical-value profiles",
       caption = "Replicates Figure 5B of Tarning 2012 AAC (typical-value approximation).")

PKNCA validation

Use PKNCA to compute Cmax, Tmax, AUC, and – for dihydroartemisinin – AUC0-24 after a single dose; for piperaquine the paper reports terminal half-life and AUC0-92(days) on the post-hoc cohort, so we focus on the Day-7 and Day-28 trough concentrations and on Cmax over the full simulated window.

sim_nca_dha <- sim_dha_sto |>
  dplyr::filter(!is.na(Cc), time <= 24) |>
  dplyr::select(id, time, Cc, cohort)

dose_df_dha <- events_dha |>
  dplyr::filter(evid == 1L, time == 0) |>
  dplyr::select(id, time, amt, cohort)

conc_obj_dha <- PKNCA::PKNCAconc(sim_nca_dha,
                                 Cc ~ time | cohort + id,
                                 concu = "ng/mL", timeu = "h")
dose_obj_dha <- PKNCA::PKNCAdose(dose_df_dha,
                                 amt ~ time | cohort + id,
                                 doseu = "mg")

intervals_dha <- data.frame(start = 0, end = 24, cmax = TRUE,
                            tmax = TRUE, auclast = TRUE,
                            half.life = TRUE)
nca_dha <- PKNCA::pk.nca(PKNCA::PKNCAdata(conc_obj_dha, dose_obj_dha,
                                          intervals = intervals_dha))

nca_dha_tab <- as.data.frame(nca_dha$result) |>
  dplyr::group_by(cohort, PPTESTCD) |>
  dplyr::summarise(median = stats::median(PPORRES, na.rm = TRUE),
                   q25    = stats::quantile(PPORRES, 0.25, na.rm = TRUE),
                   q75    = stats::quantile(PPORRES, 0.75, na.rm = TRUE),
                   .groups = "drop")

knitr::kable(nca_dha_tab,
             digits = 2,
             caption = "Simulated dihydroartemisinin NCA after a single dose (0-24 h).")

Simulated dihydroartemisinin NCA after a single dose (0-24 h).
cohort	PPTESTCD	median	q25	q75
Non-pregnant	adj.r.squared	1.00	1.00	1.00
Non-pregnant	auclast	1219.00	950.22	1581.90
Non-pregnant	clast.pred	0.08	0.04	0.18
Non-pregnant	cmax	481.65	327.69	647.56
Non-pregnant	half.life	1.14	1.04	1.24
Non-pregnant	lambda.z	0.61	0.56	0.67
Non-pregnant	lambda.z.n.points	4.00	3.00	5.00
Non-pregnant	lambda.z.time.first	2.00	1.50	4.00
Non-pregnant	lambda.z.time.last	16.00	16.00	16.00
Non-pregnant	r.squared	1.00	1.00	1.00
Non-pregnant	span.ratio	11.86	10.56	13.55
Non-pregnant	tlast	16.00	16.00	16.00
Non-pregnant	tmax	1.50	1.00	2.00
Pregnant	adj.r.squared	1.00	1.00	1.00
Pregnant	auclast	818.28	562.49	981.32
Pregnant	clast.pred	0.06	0.02	0.13
Pregnant	cmax	313.34	209.10	400.40
Pregnant	half.life	1.16	1.06	1.26
Pregnant	lambda.z	0.60	0.55	0.65
Pregnant	lambda.z.n.points	4.00	3.00	5.00
Pregnant	lambda.z.time.first	2.00	1.50	4.00
Pregnant	lambda.z.time.last	16.00	16.00	16.00
Pregnant	r.squared	1.00	1.00	1.00
Pregnant	span.ratio	11.64	10.33	13.07
Pregnant	tlast	16.00	16.00	16.00
Pregnant	tmax	1.50	1.00	2.00

sim_nca_pip <- sim_pip_sto |>
  dplyr::filter(!is.na(Cc)) |>
  dplyr::select(id, time, Cc, cohort)

dose_df_pip <- events_pip |>
  dplyr::filter(evid == 1L) |>
  dplyr::select(id, time, amt, cohort)

conc_obj_pip <- PKNCA::PKNCAconc(sim_nca_pip,
                                 Cc ~ time | cohort + id,
                                 concu = "ng/mL", timeu = "h")
dose_obj_pip <- PKNCA::PKNCAdose(dose_df_pip,
                                 amt ~ time | cohort + id,
                                 doseu = "mg")

intervals_pip <- data.frame(start = 0, end = 24 * 92,
                            cmax = TRUE, tmax = TRUE,
                            auclast = TRUE, half.life = TRUE)
nca_pip <- PKNCA::pk.nca(PKNCA::PKNCAdata(conc_obj_pip, dose_obj_pip,
                                          intervals = intervals_pip))

nca_pip_tab <- as.data.frame(nca_pip$result) |>
  dplyr::group_by(cohort, PPTESTCD) |>
  dplyr::summarise(median = stats::median(PPORRES, na.rm = TRUE),
                   q25    = stats::quantile(PPORRES, 0.25, na.rm = TRUE),
                   q75    = stats::quantile(PPORRES, 0.75, na.rm = TRUE),
                   .groups = "drop")

knitr::kable(nca_pip_tab, digits = 2,
             caption = "Simulated piperaquine NCA over 0-92 days.")

Simulated piperaquine NCA over 0-92 days.
cohort	PPTESTCD	median	q25	q75
Non-pregnant	adj.r.squared	1.00	1.00	1.00
Non-pregnant	auclast	21140.83	16381.32	29312.52
Non-pregnant	clast.pred	1.92	1.39	2.90
Non-pregnant	cmax	145.98	112.74	215.03
Non-pregnant	half.life	573.13	512.37	640.18
Non-pregnant	lambda.z	0.00	0.00	0.00
Non-pregnant	lambda.z.n.points	10.00	10.00	10.00
Non-pregnant	lambda.z.time.first	336.00	336.00	336.00
Non-pregnant	lambda.z.time.last	2016.00	2016.00	2016.00
Non-pregnant	r.squared	1.00	1.00	1.00
Non-pregnant	span.ratio	2.93	2.62	3.28
Non-pregnant	tlast	2016.00	2016.00	2016.00
Non-pregnant	tmax	50.00	28.00	51.00
Pregnant	adj.r.squared	1.00	1.00	1.00
Pregnant	auclast	27064.15	19520.77	37510.72
Pregnant	clast.pred	1.22	0.78	2.03
Pregnant	cmax	261.79	186.45	356.58
Pregnant	half.life	431.23	388.73	477.01
Pregnant	lambda.z	0.00	0.00	0.00
Pregnant	lambda.z.n.points	10.00	10.00	10.00
Pregnant	lambda.z.time.first	336.00	336.00	336.00
Pregnant	lambda.z.time.last	2016.00	2016.00	2016.00
Pregnant	r.squared	1.00	1.00	1.00
Pregnant	span.ratio	3.90	3.52	4.32
Pregnant	tlast	2016.00	2016.00	2016.00
Pregnant	tmax	51.00	28.00	52.00

Day-7 and Day-28 piperaquine trough concentrations

Tarning 2012 Table 3 reports per-cohort medians and inter-quartile ranges of Day-7 (28.1 [20.5-34.2] ng/mL pregnant, 22.7 [17.6-32.8] non-pregnant) and Day-28 (10.3 [9.18-14.4] pregnant, 10.3 [8.06-14.9] non-pregnant) post-hoc piperaquine concentrations. Compute the simulated analogues.

nearest_time <- function(df, target_h) {
  df |>
    dplyr::filter(!is.na(Cc)) |>
    dplyr::group_by(id, cohort) |>
    dplyr::slice_min(abs(time - target_h), n = 1, with_ties = FALSE) |>
    dplyr::ungroup()
}

pip_d7  <- nearest_time(sim_pip_sto, 24 * 7) |>
  dplyr::transmute(id, cohort, Cc_d7 = Cc)
pip_d28 <- nearest_time(sim_pip_sto, 24 * 28) |>
  dplyr::transmute(id, cohort, Cc_d28 = Cc)

pip_tab <- pip_d7 |>
  dplyr::inner_join(pip_d28, by = c("id", "cohort")) |>
  dplyr::group_by(cohort) |>
  dplyr::summarise(
    Day7_median  = stats::median(Cc_d7),
    Day7_q25     = stats::quantile(Cc_d7, 0.25),
    Day7_q75     = stats::quantile(Cc_d7, 0.75),
    Day28_median = stats::median(Cc_d28),
    Day28_q25    = stats::quantile(Cc_d28, 0.25),
    Day28_q75    = stats::quantile(Cc_d28, 0.75),
    .groups = "drop"
  )
knitr::kable(pip_tab, digits = 2,
             caption = "Simulated piperaquine Day-7 and Day-28 plasma concentrations (ng/mL).")

Simulated piperaquine Day-7 and Day-28 plasma concentrations (ng/mL).
cohort	Day7_median	Day7_q25	Day7_q75	Day28_median	Day28_q25	Day28_q75
Non-pregnant	20.92	16.10	28.18	9.73	7.58	13.80
Pregnant	28.30	20.69	38.62	11.09	7.86	15.97

Comparison against published NCA

Tarning 2012 Tables 3 and 5 report post-hoc empirical-Bayes-estimate summary statistics for piperaquine and dihydroartemisinin in the total cohort and split by pregnancy. The table below pairs each simulated median with the published median.

Drug	Quantity	Cohort	Simulated (median)	Published median (IQR)	Comment
Dihydroartemisinin	Cmax (ng/mL)	Pregnant	see NCA table above	391 (241-545)	Table 5
Dihydroartemisinin	Cmax (ng/mL)	Non-pregnant	see NCA table above	500 (399-715)	Table 5
Dihydroartemisinin	AUC0-24 (ng*h/mL)	Pregnant	see NCA table above	956 (559-1,110)	Table 5
Dihydroartemisinin	AUC0-24 (ng*h/mL)	Non-pregnant	see NCA table above	1,210 (1,030-1,530)	Table 5
Piperaquine	Cmax (ng/mL)	Pregnant	see NCA table above	291 (194-362)	Table 3
Piperaquine	Cmax (ng/mL)	Non-pregnant	see NCA table above	216 (139-276)	Table 3
Piperaquine	Day-7 (ng/mL)	Pregnant	see Day7/Day28 table	28.8 (23.6-34.6)	Table 3
Piperaquine	Day-7 (ng/mL)	Non-pregnant	see Day7/Day28 table	22.7 (17.6-32.8)	Table 3
Piperaquine	Day-28 (ng/mL)	Pregnant	see Day7/Day28 table	10.3 (9.18-14.4)	Table 3
Piperaquine	Day-28 (ng/mL)	Non-pregnant	see Day7/Day28 table	10.3 (8.06-14.9)	Table 3

Differences between simulated and published medians for piperaquine Cmax and Day-7 trough are expected to fall in the 30-60% range because the published Table 3 medians reflect each subject’s empirical-Bayes draw of the per-occasion BOV on F, which the published cohort happened to draw with positive mean (the 101% / 130% / 170% F drift across doses 1-3 reported in Results), whereas the simulation above draws mean-zero BOV from the structural model. The Day-28 trough is dominated by the long terminal half-life and is therefore less sensitive to the per-occasion F drift – it agrees with the published median to within a few percent. See Assumptions and deviations below.

Assumptions and deviations

No structural per-occasion F drift. Tarning 2012 Results state that the empirical post-hoc piperaquine F values were 101%, 130% and 170% at doses 1, 2 and 3. Table 2 lists F as a single fixed value of 1 with between-occasion variability (BOV) 56.3% CV and does NOT report per-occasion fixed effects. The model is therefore encoded with F = 1 (fixed) plus a zero-mean log-normal BOV on F multiplexed by the OCC indicator; the empirical 101 / 130 / 170% drift is interpreted as a posterior summary of the published cohort’s BOV draws rather than a structural feature. Downstream typical-value simulations and Monte-Carlo VPCs will therefore systematically under-predict piperaquine Cmax for doses 2 and 3 (and any short-window exposure summary dominated by them) by a factor of approximately the cohort-mean F drift. A future extension could add three per-occasion structural fixed effects on F if a control-stream-level confirmation becomes available.
Residual error encoding. Tarning 2012 Methods state that the natural-log-transformed concentrations were fit with an additive residual (‘essentially equivalent to an exponential error model on a linear scale’). The model files encode this as proportional residual error on the linear-concentration scale (Cc ~ prop(propSd)) per the Kloprogge 2018 lumefantrine precedent. For small sigma (piperaquine sigma = 0.285) the approximation is tight; for the larger dihydroartemisinin sigma = 0.580 the proportional form slightly underestimates the upper tail of the residual distribution relative to a true log-normal (~ lnorm(expSd)) parameterisation.
PARA centering log-base. Tarning 2012 Table 4 footnote centres the typical patient at “logarithmic parasitaemia 3.98” without explicit log base. log10 of the pooled-cohort median (between log10(6,780) = 3.83 in pregnant women and log10(9,670) = 3.99 in non-pregnant women) brackets 3.98; natural log would require parasitaemia exp(3.98) ~= 54 parasites/uL, which is below the cohort range. The centering value 3.98 is therefore interpreted as log10, consistent with the Kloprogge 2014 / 2018 Mahidol-Oxford malaria popPK precedents. See the PARA covariate notes in the dihydroartemisinin model file.
Virtual cohort parasitaemia distribution. Tarning 2012 Table 1 reports admission parasitaemia medians of 6,780 and 9,670 parasites/uL with very wide ranges (32-177,000) but does not report the cohort geometric standard deviation. The virtual cohort uses a log-normal distribution with geometric SD 6 to approximate the observed range; downstream users with cohort-specific parasitaemia data should override the simulated values via the PARA covariate column.
No allometric scaling on piperaquine. Tarning 2012 tested body weight as a covariate on piperaquine CL and V but did not retain it (delta-OFV = +4.62 and R-matrix non-positive-semidefinite warning); the dihydroartemisinin model retained allometric scaling (delta-OFV = -9.08). The piperaquine model file therefore has no WT entry in covariateData and does not reference WT in model(); the dihydroartemisinin model uses fixed 3/4 and 1 allometric exponents on CL/F and V/F respectively, centered at the cohort reference 48.5 kg.
OCC as a time-varying covariate. The OCC column in this vignette is constructed from time (1 for 0 <= t < 24, 2 for 24 <= t < 48, 3 for t >= 48). Downstream users supplying multi-dose data must include the OCC column in their event table with the same per-time-window encoding.