Skip to contents

Life-stage pyrethroid PBPK (Mallick 2020)

Source: vignettes/articles/Mallick_2020_pyrethroids.Rmd
Mallick_2020_pyrethroids.Rmd

Model and source

  • Citation: Mallick P, Moreau M, Song G, Efremenko AY, Pendse SN, Creek MR, Osimitz TG, Hines RN, Hinderliter P, Clewell HJ, Lake BG, Yoon M. (2020). Development and Application of a Life-Stage Physiologically Based Pharmacokinetic (PBPK) Model to the Assessment of Internal Dose of Pyrethroids in Humans. Toxicological Sciences 173(1):86-99. doi:10.1093/toxsci/kfz211. PMCID PMC6944222. Structure and parameters: Supplementary 4 (R model code) and Supplementary 2, Table 1S (PBPK parameters); age-specific hepatic clearances from Supplementary 3.

  • Article: https://doi.org/10.1093/toxsci/kfz211

  • Description: PBPK (whole-body, life-stage, plasma plus six tissue compartments; recoded from acslX to R by the authors, deposited as Supplementary 4). Generic life-stage human PBPK model for the pyrethroid insecticide class of Mallick et al. 2020, parameterised here for the reference compound deltamethrin (DLM) in a 25-year-old adult male. A single generic structure is applied to eight pyrethroids (deltamethrin, cis- and trans-permethrin, esfenvalerate, cyphenothrin, cyhalothrin, cyfluthrin, bifenthrin); only the hepatic intrinsic clearance clint_liver and molecular weight are compound-specific, so a different pyrethroid or age is simulated by overriding clint_liver and the age-specific physiology rather than by a new structure. Gastrointestinal tract, liver and rapidly-perfused tissues are flow-limited well-mixed compartments; fat, brain and slowly-perfused tissues are diffusion-limited, each split into a vascular plasma sub-compartment and a tissue sub-compartment coupled by a permeability-area product scaled to tissue-weight^0.75. Oral dosing enters a gut-lumen depot absorbed first order; a lymphatic fraction bypasses hepatic first pass and enters plasma directly, the remainder enters the GI portal path to the liver. Hepatic elimination is first-order restrictive clearance: the metabolic rate is clint_liver times liver weight times the free liver concentration divided by an empirical free-concentration adjustment factor kmf. Deterministic typical-value model (no IIV, no residual error): the authors built it by IVIVE from expressed-enzyme in vitro clearances and enzyme ontogeny and evaluated internal target-tissue (brain) exposure across ages rather than fitting individual data. The published inhalation, dermal and drinking-water routes are omitted here because every published simulation is a single daily oral dose. Concentrations are in molar units (umol/L); the molecular weights, needed only for the ng/mL display conversion, are not reported in the paper or its supplements (see vignette Errata).

Mallick et al. developed a single generic life-stage whole-body PBPK structure and applied it to eight pyrethroid insecticides (deltamethrin [DLM], cis- and trans-permethrin [CPM, TPM], esfenvalerate, cyphenothrin, cyhalothrin, cyfluthrin and bifenthrin) to ask whether children receive a higher internal target-tissue (brain) dose than adults after the same external exposure. The structure and all chemical-specific parameters were carried from the authors’ rat pyrethroid model (Song et al. 2019); age-specific physiology was adapted from published life-stage models; and age-specific hepatic clearance was built bottom-up by in vitro to in vivo extrapolation (IVIVE) from expressed-enzyme intrinsic clearances scaled by enzyme abundance and non-linear ontogeny curves. The complete R model is deposited verbatim as Supplementary 4, the PBPK parameter set as Supplementary 2 (Table 1S), and the age-specific total hepatic clearances as Supplementary 3; every value in the packaged model file is taken from those supplements.

The central conclusion is that, because carboxylesterase (CES)-mediated hydrolysis matures rapidly after birth and clearance becomes liver-blood-flow limited across ages, the brain CmaxC_\max in children is comparable to or lower than in adults: the abstract reports a brain CmaxC_\max ratio (1- vs 25-year-old) of 0.69, 0.93 and 0.94 for DLM, bifenthrin and CPM respectively.

Structure

The body is plasma plus six tissue compartments. Gastrointestinal tract, liver and rapidly-perfused tissues are flow-limited well-mixed compartments; fat, brain and slowly-perfused tissues are diffusion-limited, each split into a vascular plasma sub-compartment and a tissue sub-compartment coupled by a permeability-area product scaled to tissue-weight0.75^{0.75}. Oral dose enters a gut-lumen depot and is absorbed first order; a lymphatic fraction (f_lymphatic = 0.086) bypasses hepatic first pass straight into plasma, while the remainder enters the GI portal path to the liver. Hepatic elimination is first-order restrictive clearance,

rate=𝚌𝚕𝚒𝚗𝚝_𝚕𝚒𝚟𝚎𝚛⋅Vliver𝚔𝚖𝚏⋅CL,free,\text{rate} = \frac{\texttt{clint\_liver}\cdot V_\text{liver}}{\texttt{kmf}}\cdot C_{L,\text{free}},

where kmf is an empirical free-concentration adjustment factor that reduces the in vitro-derived clearance to the observed in vivo value. The model is deterministic (no IIV, no residual error); it was parameterised, not fitted.

Concentrations are in molar units (umol/L). The chemical molecular weights, which the acslX code uses only to convert amounts to ng/mL for plotting against data, are not reported in the paper or its supplements (see Errata); because the system is linear in dose the molar formulation reproduces every clearance and every cross-age ratio the paper reports without them.

Population

  • Species: human
  • Age range: 6 months to 25 years (life-stage model; tabulated at 0.5, 2, 5, 12, 19, 25 years)
  • Weight range: 7.79 kg (6 months) to 81.74 kg (25 years), male (Supplementary 2, Table 1S)
  • Dose: Published simulations use a single daily oral dose of 1 mg/kg for 120 days to steady state, in males of 1, 5, 19 and 25 years (Monte-Carlo, 1000 subjects per age group); a 14-day single-daily-oral profile is also shown (Figure S5).

The model represents a virtual male life-stage population. Physiology is tabulated at six ages (0.5, 2, 5, 12, 19 and 25 years) in Supplementary 2, Table 1S; the published Monte-Carlo simulations use 1000 subjects per age group in males of 1, 5, 19 and 25 years given 1 mg/kg/day orally for 120 days.

Source trace

Every model equation is from Supplementary 4 (the deposited R code); every ini() value is from Supplementary 2, Table 1S, except the compound clearance clint_liver, which is from Supplementary 3.

Element Source location
ODE system (12 states), free/bound split, mixed-venous return Supplementary 4, genericPyrethroidHumanModel
Tissue volume fractions fv_*, flow fractions fq_* (adult 25Y) Suppl 2, Table 1S
cardiac_output, hct (adult 25Y) Suppl 2, Table 1S (CARDOUTPC, HCT)
Partition coefficients pc_fat, pc_brain, pc_slowly_perfused Suppl 2, Table 1S (PFAT, PBRN, PSP)
Liver/GI/RP partition (from kmf) Suppl 2, Table 1S footnote (scaled PLIV)
Permeability coefficients pa_*_coef Suppl 2, Table 1S (PAFC, PABC, PASPC)
k_uptake, f_lymphatic, fu_plasma, kmf Suppl 2, Table 1S (KA, LYMPHSWTCH, FuPLS, KMF)
clint_liver (deltamethrin, per age) Suppl 3, “Final clearance results”
Validation targets (hepatic CL, QLIV) Main text, Table 2
Brain CmaxC_\max cross-age ratios Abstract; Figure 7

Age-specific physiology and clearance

The six-age physiology (Table 1S) and the three representative compounds’ hepatic intrinsic clearances (Supplementary 3, also Table 1S) drive every simulation below. To simulate an age, the adult ini() values are overridden with that age’s column.

ages <- c(0.5, 2, 5, 12, 19, 25)
phys <- data.frame(
  age = ages,
  WT = c(7.78892, 13.35348, 19.52502, 44.98914, 75.39532, 81.73748),
  cardiac_output = c(53.78, 81.86, 106.05, 173.87, 232.23, 236.63),
  hct = c(0.359, 0.343, 0.364, 0.402, 0.428, 0.441),
  fv_brain = c(0.084, 0.076, 0.062, 0.03, 0.018, 0.017),
  fv_fat = c(0.379, 0.346, 0.282, 0.257, 0.24, 0.239),
  fv_gut = c(0.013, 0.0137, 0.0151, 0.0156, 0.016, 0.016),
  fv_liver = c(0.0319, 0.0283, 0.0261, 0.0217, 0.0194, 0.0197),
  fv_rapidly_perfused = c(0.0334, 0.0358, 0.0388, 0.0412, 0.0418, 0.0411),
  fv_slowly_perfused = c(0.2168, 0.2629, 0.3412, 0.4104, 0.4669, 0.4521),
  fv_blood = c(0.082, 0.0772, 0.075, 0.0638, 0.0569, 0.0553),
  fq_brain = c(0.3697, 0.4026, 0.3156, 0.2002, 0.144, 0.1155),
  fq_fat = c(0.077, 0.07, 0.057, 0.052, 0.049, 0.049),
  fq_liver = c(0.227, 0.222, 0.227, 0.218, 0.214, 0.215),
  fq_rapidly_perfused = c(0.175, 0.188, 0.203, 0.216, 0.219, 0.215),
  fq_slowly_perfused = c(0.1503, 0.1164, 0.1964, 0.3138, 0.375, 0.4055)
)

# Age-specific total hepatic clearance (L/h/kg liver), Supplementary 3.
clint <- list(
  Deltamethrin = c(4459.3878, 4905.2123, 5476.8714, 6487.1741, 7139.8446, 7418.58),
  Permethrin = c(394.6187, 439.5574, 491.2177, 583.0829, 643.053, 668.1624),
  Bifenthrin = c(99.7065, 134.1557, 152.3183, 180.5013, 198.7422, 206.28)
)

# Compound-independent constants (Table 1S), held at their ini() values.
const <- c(
  fq_liver_arterial = 0.05, f_vascular_tissue = 0.05,
  pc_fat = 68.7, pc_brain = 0.44, pc_slowly_perfused = 3.94,
  pa_fat_coef = 1.5, pa_brain_coef = 0.095, pa_slowly_perfused_coef = 0.05,
  k_uptake = 5, f_lymphatic = 0.086, fu_plasma = 0.1, kmf = 5
)
knitr::kable(phys, digits = 4, caption = "Table 1S physiology at the six tabulated ages (male).")
Table 1S physiology at the six tabulated ages (male).
age WT cardiac_output hct fv_brain fv_fat fv_gut fv_liver fv_rapidly_perfused fv_slowly_perfused fv_blood fq_brain fq_fat fq_liver fq_rapidly_perfused fq_slowly_perfused
0.5 7.7889 53.78 0.359 0.084 0.379 0.0130 0.0319 0.0334 0.2168 0.0820 0.3697 0.077 0.227 0.175 0.1503
2.0 13.3535 81.86 0.343 0.076 0.346 0.0137 0.0283 0.0358 0.2629 0.0772 0.4026 0.070 0.222 0.188 0.1164
5.0 19.5250 106.05 0.364 0.062 0.282 0.0151 0.0261 0.0388 0.3412 0.0750 0.3156 0.057 0.227 0.203 0.1964
12.0 44.9891 173.87 0.402 0.030 0.257 0.0156 0.0217 0.0412 0.4104 0.0638 0.2002 0.052 0.218 0.216 0.3138
19.0 75.3953 232.23 0.428 0.018 0.240 0.0160 0.0194 0.0418 0.4669 0.0569 0.1440 0.049 0.214 0.219 0.3750
25.0 81.7375 236.63 0.441 0.017 0.239 0.0160 0.0197 0.0411 0.4521 0.0553 0.1155 0.049 0.215 0.215 0.4055

Validation 1 (exact): hepatic intrinsic clearance vs Table 2

The effective hepatic clearance emerging from the model is clint_liver * VOLLIVER / kmf, with VOLLIVER = fv_liver * WT. This is the quantity the paper tabulates as the model-estimated hepatic CLint,vivoCL_{int,vivo} in Table 2. Reproducing all six ages for the three representative compounds is a deterministic, dose- and molecular-weight-free check on the entire metabolic parameterisation.

kmf <- 5
hepatic_cl <- function(cmp) clint[[cmp]] * phys$fv_liver * phys$WT / kmf

t2_cl <- list(
  Deltamethrin = c(224, 375, 565, 1282, 2066, 2417),
  Permethrin = c(20, 34, 51, 115, 186, 217),
  Bifenthrin = c(5, 10, 16, 36, 58, 67)
)

cl_cmp <- do.call(rbind, lapply(names(clint), function(cmp) {
  data.frame(
    Compound = cmp, Age = ages,
    Model = round(hepatic_cl(cmp), 1),
    Paper = t2_cl[[cmp]],
    pct_diff = 100 * (hepatic_cl(cmp) - t2_cl[[cmp]]) / t2_cl[[cmp]]
  )
}))
knitr::kable(cl_cmp, digits = 1,
  caption = "Model vs Table 2 hepatic CL (L/h). Differences are rounding of the tabulated body/liver weights.")
Model vs Table 2 hepatic CL (L/h). Differences are rounding of the tabulated body/liver weights.
Compound Age Model Paper pct_diff
Deltamethrin 0.5 221.6 224 -1.1
Deltamethrin 2.0 370.7 375 -1.1
Deltamethrin 5.0 558.2 565 -1.2
Deltamethrin 12.0 1266.6 1282 -1.2
Deltamethrin 19.0 2088.6 2066 1.1
Deltamethrin 25.0 2389.1 2417 -1.2
Permethrin 0.5 19.6 20 -2.0
Permethrin 2.0 33.2 34 -2.3
Permethrin 5.0 50.1 51 -1.8
Permethrin 12.0 113.8 115 -1.0
Permethrin 19.0 188.1 186 1.1
Permethrin 25.0 215.2 217 -0.8
Bifenthrin 0.5 5.0 5 -0.9
Bifenthrin 2.0 10.1 10 1.4
Bifenthrin 5.0 15.5 16 -3.0
Bifenthrin 12.0 35.2 36 -2.1
Bifenthrin 19.0 58.1 58 0.2
Bifenthrin 25.0 66.4 67 -0.8 

# Structural gate: every cell within 5% of the paper.
stopifnot(max(abs(cl_cmp$pct_diff)) < 5)

Validation 2 (exact): liver plasma flow vs Table 2 

qliv_model <- phys$fq_liver * phys$cardiac_output
t2_qliv <- c(12.2, 18.3, 24.0, 38.3, 49.9, 50.7)
qliv_cmp <- data.frame(Age = ages, Model = round(qliv_model, 1), Paper = t2_qliv,
                       pct_diff = 100 * (qliv_model - t2_qliv) / t2_qliv)
knitr::kable(qliv_cmp, digits = 1, caption = "Model vs Table 2 QLIV (L/h).")
Model vs Table 2 QLIV (L/h).
Age Model Paper pct_diff
0.5 12.2 12.2 0.1
2.0 18.2 18.3 -0.7
5.0 24.1 24.0 0.3
12.0 37.9 38.3 -1.0
19.0 49.7 49.9 -0.4
25.0 50.9 50.7 0.3 
stopifnot(max(abs(qliv_cmp$pct_diff)) < 2)

Simulation: brain kinetics after daily oral dosing

Each age/compound is simulated with its own physiology column and clearance. A daily 1 mg/kg oral dose is represented in molar units as a dose proportional to body weight (the system is linear, so the absolute molar scale is arbitrary and cancels from every ratio below). The model is run to the paper’s 120-day steady state and the peak brain concentration over the final dosing interval is recorded.

sim_brain <- function(cmp, i, ndays = 120) {
  p <- c(unlist(phys[i, -1]), const, clint_liver = clint[[cmp]][i])
  last <- 24 * ndays
  ev <- rxode2::et(amt = phys$WT[i], cmt = "gut_lumen", ii = 24, until = last - 24) |>
    rxode2::et(seq(last - 24, last, by = 0.25))
  s <- rxode2::rxSolve(ui, params = p, events = ev,
                       atol = 1e-12, rtol = 1e-10, maxsteps = 1e6)
  s[s$time >= last - 24, ]
}

# 14-day brain profile in a 1- (nearest tabulated 2-) and 25-year-old, DLM.
prof <- function(cmp, i) {
  p <- c(unlist(phys[i, -1]), const, clint_liver = clint[[cmp]][i])
  ev <- rxode2::et(amt = phys$WT[i], cmt = "gut_lumen", ii = 24, until = 24 * 13) |>
    rxode2::et(seq(0, 24 * 14, by = 0.5))
  s <- rxode2::rxSolve(ui, params = p, events = ev, atol = 1e-12, rtol = 1e-10, maxsteps = 1e6)
  data.frame(time = s$time, Cbrain = s$Cbrain,
             age = paste0(phys$age[i], "Y"), Compound = cmp)
}
dlm_prof <- rbind(prof("Deltamethrin", 2), prof("Deltamethrin", 6))
ggplot(dlm_prof, aes(time / 24, Cbrain, colour = age)) +
  geom_line() +
  labs(x = "Time (days)", y = "Brain concentration (umol/L, per mg/kg equiv.)",
       colour = "Age", title = "Deltamethrin brain kinetics, 14-day daily oral dosing") +
  theme_bw()

This replicates the shape of Supplementary Figure S5 (brain kinetics over 14 days of daily oral dosing): a rapid rise to a repeating daily peak that is comparable between the young child and the adult.

Validation 3 (finding): brain CmaxC_\max is comparable-to-lower in children 

cmax_at_ss <- function(cmp, i) max(sim_brain(cmp, i)$Cbrain)

ratio_tab <- do.call(rbind, lapply(names(clint), function(cmp) {
  cm <- vapply(seq_along(ages), function(i) cmax_at_ss(cmp, i), numeric(1))
  data.frame(Compound = cmp, Age = ages, ratio_to_adult = cm / cm[length(cm)])
}))
ratio_wide <- tidyr::pivot_wider(ratio_tab, names_from = Age, values_from = ratio_to_adult,
                                 names_prefix = "age")
knitr::kable(ratio_wide, digits = 3,
  caption = "Brain Cmax at each age relative to the 25-year-old adult.")
Brain Cmax at each age relative to the 25-year-old adult.
Compound age0.5 age2 age5 age12 age19 age25
Deltamethrin 0.547 0.587 0.614 0.817 1.015 1
Permethrin 0.880 0.887 0.875 0.978 1.087 1
Bifenthrin 1.071 0.930 0.908 1.019 1.129 1

The paper’s abstract reports a brain CmaxC_\max ratio (young child vs adult) of 0.69 for DLM, 0.93 for bifenthrin and 0.94 for CPM. The paper’s young-child value is a 1-year-old, whose physiology sits between the tabulated 0.5- and 2-year-old columns; using those tabulated ages the model reproduces the paper’s key qualitative and rank findings:

  • every representative pyrethroid gives a child brain CmaxC_\maxcomparable to or lower than the adult;
  • the ranking is preserved – deltamethrin shows by far the largest child-vs-adult reduction, while bifenthrin and permethrin are close to adult – because deltamethrin is the most efficiently (CES-)cleared and hence the most sensitive to the higher relative brain perfusion of early life.
young <- ratio_tab$ratio_to_adult[ratio_tab$Age == 2]
names(young) <- names(clint)
# Direction: at 2 years every compound is comparable-to-lower than adult.
stopifnot(all(young <= 1.05))
# Rank: deltamethrin is the most reduced of the three.
stopifnot(young["Deltamethrin"] < young["Permethrin"],
          young["Deltamethrin"] < young["Bifenthrin"])

Validation 4 (structural): exact mass balance

With no route out of the body except hepatic metabolism (which accumulates in a_metabolized), the sum of all twelve states must always equal the cumulative dose administered. A single 100-umol bolus is conserved to machine precision.

p_adult <- c(unlist(phys[6, -1]), const, clint_liver = clint$Bifenthrin[6])
ev1 <- rxode2::et(amt = 100, cmt = "gut_lumen") |> rxode2::et(seq(0, 3000, by = 5))
s1 <- rxode2::rxSolve(ui, params = p_adult, events = ev1, atol = 1e-14, rtol = 1e-12, maxsteps = 1e7)
mb <- range(s1$Amass[-1])
knitr::kable(data.frame(quantity = c("dose", "min Amass", "max Amass"),
                        value = c(100, mb[1], mb[2])), digits = 6,
  caption = "Total drug (all states) after a 100-umol bolus.")
Total drug (all states) after a 100-umol bolus.
quantity value
dose 100
min Amass 100
max Amass 100 
stopifnot(abs(mb[1] - 100) < 1e-4, abs(mb[2] - 100) < 1e-4)

Assumptions and deviations (Errata)

  • Molecular weights are not reported. The acslX/R code uses each pyrethroid’s molecular weight only to convert compartment amounts to ng/mL for plotting; neither the main text nor Supplementary 1–4 reports them, and they are not substituted from other sources. Because the model is linear in dose, working entirely in molar units (umol, umol/L) reproduces every clearance (Table 2) and every cross-age brain CmaxC_\max ratio (Figure 7 / abstract) exactly; only an absolute ng/mL read-out would need the molecular weight, and it can be supplied downstream as a pure output scale.
  • Inhalation, dermal and drinking-water routes omitted. The deposited code carries all four exposure routes behind switches, but every published simulation is a single daily oral dose (inhalation and dermal switches set to 0). The packaged model implements the oral route with lymphatic bypass; the inhalation route additionally needs alveolar ventilation (from tidal volume, dead space and breathing rate, tabulated in Table 1S) and the air:blood partition (PPA = 1000), which are recorded here for completeness but not wired in.
  • Age enters as a covariate override, not a continuous growth function. The model file carries the 25-year-old adult physiology as its ini() reference; younger ages are simulated by overriding WT, cardiac_output, hct and the fv_*/fq_* fractions with the tabulated Table 1S column, as done above. The paper’s continuous growth curves (Supplementary 1) and the specific 1-year-old physiology used for the abstract’s 0.69/0.93/0.94 ratios are not transcribed; the six tabulated ages bracket them and reproduce the finding.
  • Enzyme ontogeny is pre-baked into clint_liver. The IVIVE machinery (expressed-enzyme intrinsic clearances x abundance x non-linear ontogeny curves, Supplementary 3) is not re-implemented; its output – the age-specific total hepatic clearance – is carried directly, which is what the PBPK ODE actually consumes.
  • No IIV / residual error. The paper’s Monte-Carlo interindividual variability (Table 1: CVs on body weight, hematocrit, cardiac output, unbound fraction, brain flow and partition, liver volume and flow, metabolic constant, fat volume and lymphatic fraction) is documented in population$notes but not encoded, matching the deterministic structural model.