Skip to contents

The model

Tensil_2026_losartan_pbpk is a whole-body physiologically based pharmacokinetic/pharmacodynamic model of losartan, its active CYP2C9 metabolite E3174 (losartan carboxylic acid) and the E3174 glucuronide L158, built by Tensil, Myshkina and Koenig (2026) from 25 curated clinical studies and distributed by the authors as SBML under CC-BY.

mod <- rxode2::rxode(readModelDb("Tensil_2026_losartan_pbpk"))
length(mod$state)
#> [1] 49

The 49 ODEs are organised as a systemic circulation (venous, arterial, portal and hepatic-venous plasma pools, plus lung and a lumped rest-of-body) linking three pharmacokinetic submodels and one pharmacodynamic submodel:

  • Intestine. Oral losartan dissolves into the gut lumen at ka_dis_los. Of the flux leaving the lumen, 60% (f_los_abs) enters the enterocytes and 40% is excreted in feces. P-glycoprotein (ABCB1) effluxes losartan from the enterocytes back to the lumen; the remainder passes to gut plasma. E3174 and L158 that reach the lumen in bile are excreted in feces.
  • Liver. Losartan is taken up into liver tissue, converted to E3174 by CYP2C9 (Michaelis-Menten; the covariate CYP2C9 scales the Vmax), and E3174 to L158 (first order). E3174 and L158 are exported back to plasma, and all three species are secreted into bile, which empties into the gut lumen.
  • Kidney. First-order urinary excretion of all three species from kidney plasma, scaled by RENALFUNC_REL.
  • RAAS. One Emax term on venous E3174 raises renin secretion (up to six-fold) and suppresses aldosterone secretion. Renin converts a constant angiotensinogen pool to angiotensin I, which turns over to angiotensin II, and systolic, diastolic and mean arterial pressure move in proportion to the aldosterone deviation from baseline.

Hepatic impairment (HEPFUNC_REL) shrinks the metabolising liver volume and shunts portal and hepatic-arterial blood past the liver; body weight (WT) scales every organ volume and blood flow.

A note on units and dosing. Drug states are concentrations in umol/L (cumulative urine and feces pools are amounts in umol) and the RAAS hormones are in pmol/L; the outputs Cc, Cc_e3174 and Cc_l158 are venous plasma concentrations in nmol/L, matching the paper’s figures. Doses (mg) must go into depot (oral), depot_iv (intravenous losartan) or depot_iv_e3174 (intravenous E3174) and never directly into a plasma or tissue compartment. Model time is in minutes.

Population

pop <- mod$population
data.frame(Field = names(pop), Value = vapply(pop, as.character, character(1))) |>
  knitr::kable(row.names = FALSE)
Field Value
species human
n_subjects NA
n_studies 25
age_range adults; not tabulated per study in the source
weight_range study-specific mean bodyweight where reported (52-84.6 kg across Supplementary Table S5); 75 kg reference individual otherwise
sex_female_pct NA
disease_state Healthy volunteers, CYP2C9 and ABCB1 genotyped volunteers, and subjects with hepatic (cirrhosis) or renal impairment (Table 1). Animal, pediatric, single-patient, cocktail and combination studies were excluded (Methods Section 2.1).
dose_range Oral losartan 2-200 mg single and multiple dose (Table 1); intravenous losartan 10-30 mg and intravenous E3174 10-20 mg (FDA 1995 studies 60 and 67, Lo 1995).
regions not reported
notes Deterministic typical-individual model. Methods Section 2.3 states that ‘all simulations were performed deterministically using the optimized parameter set representing the typical (mean) individual’ and that inter-individual variability was NOT included, so no IIV and no residual-error model are reported and none are invented here. Data were curated into PK-DB (identifiers in Table 1) from a PubMed and PKPDAI search screening 787 records down to 25 studies. Parameters were fitted sequentially by weighted least squares with 100 multi-start local optimisation runs - pharmacokinetic first (12 parameters, Table S2), then pharmacodynamic (4 parameters, Table S3). Pharmacodynamic data (renin, angiotensin, aldosterone, blood pressure) came from a smaller subset of studies and showed high between-study variability (Discussion).

The model is deterministic. Methods Section 2.3 states that all simulations were “performed deterministically using the optimized parameter set representing the typical (mean) individual” and that inter-individual variability was not included, so the model carries no IIV and no residual-error estimate, and propSd is fixed at zero. Every simulation below is a single typical individual per scenario.

Source trace

Quantity Parameters Source
Fractional organ volumes (gut, kidney, liver, lung) fvgu, fvki, fvli, fvlu Methods 2.3 (1.71%, 0.44%, 2.10%, 0.76%); Table S7
Fractional blood-pool volumes fvar, fvve Table S7 (0.0257, 0.0514 L/kg)
Portal and hepatic-venous fractional volumes fvpo, fvhv Archived SBML FVpo, FVhv (not tabulated)
Fractional blood flows fqgu, fqh, fqki, fqlu Methods 2.3 (18.00%, 21.50%, 19.00%, 100%); Table S7
Cardiac output, hematocrit covbw, hct Table S7 COBW (1.55 mL/s/kg; SBML 1.548), HCT (0.51)
Blood fraction, cardiac function, enterocyte and RAAS volumes fblood, fcardio, v_entero, vplasma Archived SBML Fblood, f_cardiac_function, GU__Ventero, Vplasma
Molecular weights mr_los, mr_e3174 Archived SBML Mr_los, Mr_e3174
Partition coefficient, tissue distribution kp_los, ftissue_los Table S2 (3.262; 0.145 L/min); Methods 2.3
Dissolution, absorption, absorbed fraction ka_dis_los, losabs_k, f_los_abs Table S7 (2 /h); Table S2 (0.036 /min); Table S7 (0.6)
P-gp efflux, fecal metabolite excretion f_losefl_k, metexc_k Table S2 (1.020; 2.696e-5 /min)
Uptake and P-gp activity scalars f_oatp2b1, f_abcb1 Table S7 (1, 1)
Hepatic CYP2C9 metabolism los2e3174_vmax, los2e3174_km Table S2 (7.251e-4 mmol/min/L); Table S7 Km 0.0011 mmol/L (literature)
E3174 to L158, E3174 export, biliary export e3174l158_k, li_e3174ex_k, mbiex_k Table S2 (1.134e-3; 0.011; 0.066 /min)
Hepatic losartan uptake, L158 export losim_k, li_l158ex_k Archived SBML LI__LOSIM_k, LI__L158EX_k (1 /min; not fitted)
Renal excretion ki_losex_k, ki_e3174ex_k, ki_l158ex_k Table S2 (0.077; 0.029; 0.289 /min)
RAAS fitted parameters anggen2ang1_k, e50_e3174, aldsec_k, bp_ald_fe Table S3 (0.100 L/min; 2.911e-4 mmol/L; -; 0.312); aldsec_k from Table S7 (1.01e-6 mmol/min) – see Errata
RAAS fixed parameters rensec_k, rensec_fa_e3174 Table S7 (0.1 mmol/min; 5)
Baseline hormones ren_ref, ang1_ref, ang2_ref, ald_ref, anggen_ref Archived SBML *_ref (1, 10, 8, 250, 10 pmol/L)
Baseline blood pressure sbp_ref, dbp_ref Results 3.2 (120/80 mmHg)
Scenario covariates WT, RENALFUNC_REL, HEPFUNC_REL, CYP2C9 Table S7 (BW, KI__f_renal_function, LI__f_cyp2c9); f_cirrhosis from the archived SBML and Tables S5/S6; per-study values Table S5

None of the rate laws or ODEs are printed in the manuscript or the Supplementary Materials. They were taken from the authors’ archived SBML model, which Methods Section 2.3 cites as shared on Zenodo at v0.8.0 (reference 53: 10.5281/zenodo.18213663; file models/losartan_body_flat.md of the 0.8.0 release of github.com/matthiaskoenig/losartan-model). Every parameter value of that release agrees with Supplementary Tables S2, S3 and S7 to the printed precision, apart from the ALDSEC_k entry of Table S3 (see Errata); the full-precision SBML values are used in ini().

Simulation helper

# Time is in minutes. Observations are placed on the `venous` ODE state; rxode2
# returns every algebraic observable (Cc, Cc_e3174, SBP, ...) as a column on
# those rows.
sim_los <- function(dose_mg = 50, route = "depot", hours = 48, by_min = 5,
                    wt = 75, renal = 1, hep = 1, cyp2c9 = 1, model = mod,
                    times = NULL, tight = FALSE) {
  tt <- if (is.null(times)) seq(0, hours * 60, by = by_min) else times
  ev <- rbind(
    data.frame(time = 0, amt = dose_mg, evid = 1L, cmt = route),
    data.frame(time = tt, amt = NA_real_, evid = 0L, cmt = "venous")
  )
  ev$WT <- wt
  ev$RENALFUNC_REL <- renal
  ev$HEPFUNC_REL <- hep
  ev$CYP2C9 <- cyp2c9
  out <- if (tight) {
    rxode2::rxSolve(model, ev, returnType = "data.frame", atol = 1e-12, rtol = 1e-10)
  } else {
    rxode2::rxSolve(model, ev, returnType = "data.frame")
  }
  out <- as.data.frame(out)
  out$hour <- out$time / 60
  out
}

# Trapezoidal AUC on the dense simulation grid (summary use only; the PKNCA
# section below is the NCA of record).
trap <- function(t, y) sum(diff(t) * (head(y, -1) + tail(y, -1)) / 2)

Exact reproduction of the authors’ model

The strongest check available for a model distributed as SBML is to simulate that SBML directly and compare. The table below holds reference values computed with libroadrunner 2.11 from the authors’ archived losartan_body_flat.xml (release 0.8.0) at tolerances 1e-14 absolute / 1e-10 relative, for ten scenarios covering every route and every covariate: oral 10, 50 and 100 mg; 50 mg with cirrhosis (f_cirrhosis 0.67, the FDA 1995 study 67 value), renal impairment (KI__f_renal_function 0.14), CYP2C9 activity 0.17, ABCB1 activity 0.306 and body weight 60 kg; intravenous losartan 30 mg and intravenous E3174 20 mg. Concentrations are in nmol/L, renin and aldosterone in pmol/L, SBP in mmHg and cumulative urinary E3174 in umol.

rr <- tibble::tribble(
  ~scenario, ~time, ~Cc, ~Cc_e3174, ~Cc_l158, ~renin, ~aldosterone, ~SBP, ~urine_e3174,
  "po50", 60, 400.9947, 167.9322, 8.349906, 2.829307, 161.0193, 106.6743, 0.02697759,
  "po50", 180, 101.7525, 501.1334,  9.45803, 4.162898, 91.82095, 96.31116, 0.4765545,
  "po50", 720, 21.57749, 108.7675, 0.8303907, 2.360137, 181.8482, 109.7936, 1.789382,
  "po10", 60, 74.78837, 35.72772, 1.779888, 1.546629, 223.6971, 116.0609, 0.005694748,
  "po10", 180, 19.52231, 104.1864, 1.931643,  2.31795, 184.0634, 110.1254, 0.100279,
  "po10", 720, 4.075586, 21.46125, 0.1586794, 1.343343, 232.7842, 117.4218, 0.3676898,
  "po100", 60, 876.0789, 307.5659, 15.24276, 3.568878, 124.2178, 101.1629, 0.05008329,
  "po100", 180, 215.4605, 945.5142, 18.32171,  4.82308, 58.83488, 91.37118, 0.8834033,
  "po100", 720, 46.58667, 221.6223, 1.765669, 3.161346, 141.7627, 103.7904, 3.439145,
  "po50_cirr067", 60,  727.248, 73.25295, 4.417902, 2.005322, 201.5384, 112.7424, 0.01138136,
  "po50_cirr067", 180, 234.4563, 280.9034, 7.406594, 3.455552, 127.3435,  101.631, 0.2383445,
  "po50_cirr067", 720, 81.10768, 169.6968, 1.339973,  2.84145, 157.8574, 106.2008, 1.393155,
  "po50_renal014", 60,  411.142, 173.6355, 9.922634, 2.868221, 159.1443, 106.3935, 0.0038965,
  "po50_renal014", 180, 105.6414, 576.6098, 17.51916,   4.3227, 83.90742, 95.12603, 0.07318147,
  "po50_renal014", 720, 23.02296, 213.0373, 1.808677, 3.113004, 144.2097, 104.1569, 0.3432374,
  "po50_cyp017", 60, 671.0298, 49.42683, 2.460787, 1.725799, 215.0347, 114.7636, 0.007882573,
  "po50_cyp017", 180, 236.1442, 178.2883, 3.814194, 2.899266, 155.1204, 105.7909, 0.1542534,
  "po50_cyp017", 720, 73.15641, 73.88037, 0.7720712,  2.01219, 199.3283, 112.4114, 0.7614893,
  "po50_abcb1_0306", 60, 544.9191, 204.0262, 10.19727, 3.060459, 149.6927,  104.978, 0.03199825,
  "po50_abcb1_0306", 180, 122.6635, 635.2368, 11.93637, 4.428858, 78.51121,  94.3179, 0.6009125,
  "po50_abcb1_0306", 720, 27.97719, 137.2101, 1.071993, 2.601863, 169.7521, 107.9821, 2.237123,
  "po50_bw60", 60, 448.4444, 184.0536, 9.176471, 2.936891, 155.7647, 105.8874, 0.02335062,
  "po50_bw60", 180, 134.2359, 590.1041, 11.57109, 4.348386, 82.57712, 94.92681, 0.4357186,
  "po50_bw60", 720, 31.08124, 146.4356, 1.183895, 2.673518, 166.1722,  107.446, 1.746884,
  "iv30", 60, 131.9622,  482.696, 18.00487, 4.119123, 94.23357, 96.67248, 0.1608347,
  "iv30", 180, 84.67303, 400.3066, 4.034879, 3.894985, 105.0876, 98.29797, 0.6649228,
  "iv30", 720, 46.95715, 160.8498, 1.680699, 2.779617,  160.927, 106.6605, 1.928968,
  "iv20_e3174", 60,        0, 6713.611, 6.658412,  5.79223, 10.33153, 84.10734, 4.297814,
  "iv20_e3174", 180,        0, 3845.284, 5.915158, 5.648153, 17.49886, 85.18071, 10.04074,
  "iv20_e3174", 720,        0, 313.2995, 0.4934254, 3.591941, 120.0445, 100.5379, 17.11368
)

scenarios <- list(
  po10 = list(dose_mg = 10),
  po50 = list(dose_mg = 50),
  po100 = list(dose_mg = 100),
  po50_cirr067 = list(hep = 1 - 0.67),
  po50_renal014 = list(renal = 0.14),
  po50_cyp017 = list(cyp2c9 = 0.17),
  po50_abcb1_0306 = list(model = mod |> rxode2::ini(f_abcb1 = 0.306)),
  po50_bw60 = list(wt = 60),
  iv30 = list(dose_mg = 30, route = "depot_iv"),
  iv20_e3174 = list(dose_mg = 20, route = "depot_iv_e3174")
)
#> ℹ change initial estimate of `f_abcb1` to `0.306`

ours <- bind_rows(lapply(names(scenarios), function(nm) {
  args <- c(scenarios[[nm]], list(times = c(60, 180, 720), tight = TRUE))
  do.call(sim_los, args) |>
    mutate(scenario = nm) |>
    select(scenario, time, Cc, Cc_e3174, Cc_l158, renin, aldosterone, SBP, urine_e3174)
}))

cmp <- rr |>
  pivot_longer(-c(scenario, time), names_to = "output", values_to = "roadrunner") |>
  inner_join(
    ours |> pivot_longer(-c(scenario, time), names_to = "output", values_to = "rxode2"),
    by = c("scenario", "time", "output")
  ) |>
  filter(roadrunner != 0) |>
  mutate(rel_diff = abs(rxode2 / roadrunner - 1))

cmp |>
  group_by(output) |>
  summarise(n = n(), max_rel_diff = max(rel_diff)) |>
  rename("Output" = output, "Values compared" = n, "Max relative difference" = max_rel_diff) |>
  knitr::kable(digits = 10)
Output Values compared Max relative difference
Cc 27 3.918e-07
Cc_e3174 30 3.367e-07
Cc_l158 30 4.099e-07
SBP 30 4.492e-07
aldosterone 30 3.010e-07
renin 30 2.196e-07
urine_e3174 30 2.442e-07

stopifnot(
  nrow(cmp) == 7 * 30 - 3, # the three losartan values of the IV-E3174 arm are exactly 0
  max(cmp$rel_diff) < 1e-4 # measured ~4e-7 at these tolerances
)

All 207 non-zero values agree to better than 1e-4, so every flow, volume, rate law and scenario scalar of the SBML has been carried over faithfully.

Mass balance

Losartan, E3174 and L158 are interconverted one-to-one, so the total drug-related amount in the body plus what has been excreted must equal the dose at every time.

mb <- sim_los(50, hours = 24 * 7, by_min = 60, tight = TRUE)
dose_umol <- 50 / 422.911 * 1000
species_amount <- function(s, sfx) {
  with(s, get(paste0("gut_lumen", sfx)) * vgu +
    get(paste0("gut_plasma", sfx)) * vgu_plasma +
    get(paste0("portal", sfx)) * vpo +
    get(paste0("liver_plasma", sfx)) * vli_plasma +
    get(paste0("liver", sfx)) * vli_tissue +
    get(paste0("hepatic_vein", sfx)) * vhv +
    get(paste0("kidney_plasma", sfx)) * vki_plasma +
    get(paste0("lung_plasma", sfx)) * vlu_plasma +
    get(paste0("rest_plasma", sfx)) * vre_plasma +
    get(paste0("venous", sfx)) * vve +
    get(paste0("arterial", sfx)) * var_ +
    get(paste0("urine", sfx)) + get(paste0("feces", sfx)))
}
mb <- mb |>
  mutate(
    total = 1000 * depot / 422.911 + gut * 1 +
      lung * vlu_tissue + rest * vre_tissue +
      species_amount(mb, "") + species_amount(mb, "_e3174") + species_amount(mb, "_l158")
  )
stopifnot(max(abs(mb$total / dose_umol - 1)) < 1e-6)

last <- tail(mb, 1)
recovery <- tibble::tibble(
  Route = c("Urine", "Urine", "Urine", "Feces", "Feces", "Feces"),
  Species = rep(c("losartan", "E3174", "L158"), 2),
  `Percent of dose at 7 days` = 100 * c(
    last$urine, last$urine_e3174, last$urine_l158,
    last$feces, last$feces_e3174, last$feces_l158
  ) / dose_umol
)
knitr::kable(recovery, digits = 2)
Route Species Percent of dose at 7 days
Urine losartan 1.70
Urine E3174 2.10
Urine L158 0.26
Feces losartan 59.90
Feces E3174 8.14
Feces L158 0.08

The model conserves mass to integrator precision. It routes about 60% of the dose to feces as unchanged losartan – the 40% of the luminal flux that is never absorbed, plus losartan returned in bile – and another 8% as E3174, but only a few percent to urine. Figure 2 of the paper quotes clinical recovery of 49-58% of the dose in feces and 5-12% (losartan) and 6-7% (E3174) in urine, so the model somewhat over-predicts fecal and under-predicts urinary recovery; see the Errata.

Dose dependency (Figure 3)

doses <- c(10, 25, 50, 75, 100)
dd <- bind_rows(lapply(doses, function(d) sim_los(d, hours = 36) |> mutate(dose_mg = d)))

dd |>
  select(hour, dose_mg, Cc, Cc_e3174, Cc_l158, renin, ang1, aldosterone, SBP, DBP) |>
  pivot_longer(-c(hour, dose_mg)) |>
  mutate(name = factor(name, levels = c("Cc", "Cc_e3174", "Cc_l158", "renin", "ang1",
                                        "aldosterone", "SBP", "DBP"))) |>
  ggplot(aes(hour, value, colour = factor(dose_mg))) +
  geom_line() +
  facet_wrap(~name, scales = "free_y", ncol = 4) +
  labs(x = "Time (h)", y = "Concentration (nmol/L or pmol/L) / pressure (mmHg)",
       colour = "Dose (mg)",
       title = "Replicates Figure 3B of Tensil 2026")

dose_tab <- dd |>
  group_by(dose_mg) |>
  summarise(
    auc_los = trap(hour, Cc) / 1000,
    auc_e3174 = trap(hour, Cc_e3174) / 1000,
    cmax_los = max(Cc) / 1000,
    cmax_e3174 = max(Cc_e3174) / 1000,
    ratio = auc_e3174 / auc_los,
    min_sbp = min(SBP),
    min_dbp = min(DBP),
    max_renin = max(renin),
    min_ald = min(aldosterone)
  )
dose_tab |>
  rename(
    "Dose (mg)" = dose_mg, "Losartan AUC0-36 (umol*h/L)" = auc_los,
    "E3174 AUC0-36 (umol*h/L)" = auc_e3174, "Losartan Cmax (umol/L)" = cmax_los,
    "E3174 Cmax (umol/L)" = cmax_e3174, "E3174/losartan AUC ratio" = ratio,
    "Min SBP (mmHg)" = min_sbp, "Min DBP (mmHg)" = min_dbp,
    "Max renin (pmol/L)" = max_renin, "Min aldosterone (pmol/L)" = min_ald
  ) |>
  knitr::kable(digits = 3)
Dose (mg) Losartan AUC0-36 (umol*h/L) E3174 AUC0-36 (umol*h/L) Losartan Cmax (umol/L) E3174 Cmax (umol/L) E3174/losartan AUC ratio Min SBP (mmHg) Min DBP (mmHg) Max renin (pmol/L) Min aldosterone (pmol/L)
10 0.244 0.852 0.075 0.105 3.488 110.100 73.400 2.322 183.895
25 0.623 2.120 0.192 0.258 3.402 102.414 68.276 3.349 132.569
50 1.292 4.208 0.401 0.502 3.258 96.297 64.198 4.166 91.726
75 2.011 6.261 0.629 0.732 3.114 93.211 62.141 4.578 71.120
100 2.785 8.274 0.876 0.946 2.971 91.370 60.913 4.824 58.826
stopifnot(
  # "a decreasing E3174/losartan ratio at higher doses" (Section 3.3) --
  # saturable CYP2C9 metabolism.
  all(diff(dose_tab$ratio) < 0),
  # "nonlinear increases in AUC and Cmax": losartan AUC rises more than
  # dose-proportionally.
  dose_tab$auc_los[dose_tab$dose_mg == 100] / dose_tab$auc_los[dose_tab$dose_mg == 10] > 10,
  # Stronger suppression of aldosterone and blood pressure, and larger renin
  # rise, at higher doses.
  all(diff(dose_tab$min_sbp) < 0), all(diff(dose_tab$min_ald) < 0),
  all(diff(dose_tab$max_renin) > 0)
)

Figure 3D plots the minimum SBP and DBP against dose. The model gives the values read off that panel at 50 mg (about 96 and 64 mmHg) and the 10-100 mg span (about 110 to 91 mmHg systolic):

fig3d <- tibble::tribble(
  ~dose_mg, ~pressure, ~figure,
  10, "SBP", 110,
  50, "SBP", 96,
  100, "SBP", 91.5,
  10, "DBP", 73.5,
  50, "DBP", 64,
  100, "DBP", 61
)
fig3d_cmp <- fig3d |>
  left_join(
    dose_tab |>
      select(dose_mg, SBP = min_sbp, DBP = min_dbp) |>
      pivot_longer(-dose_mg, names_to = "pressure", values_to = "model"),
    by = c("dose_mg", "pressure")
  ) |>
  mutate(diff_mmHg = model - figure)
fig3d_cmp |>
  rename("Dose (mg)" = dose_mg, "Pressure" = pressure,
         "Figure 3D read-off (mmHg)" = figure, "Model (mmHg)" = model,
         "Difference (mmHg)" = diff_mmHg) |>
  knitr::kable(digits = 1)
Dose (mg) Pressure Figure 3D read-off (mmHg) Model (mmHg) Difference (mmHg)
10 SBP 110.0 110.1 0.1
50 SBP 96.0 96.3 0.3
100 SBP 91.5 91.4 -0.1
10 DBP 73.5 73.4 -0.1
50 DBP 64.0 64.2 0.2
100 DBP 61.0 60.9 -0.1
# Read-off precision on the published panel is about +/-1 mmHg.
stopifnot(all(abs(fig3d_cmp$diff_mmHg) < 2))

Hepatic impairment (Figure 4)

cirr <- c(0, 0.3, 0.5, 0.67, 0.8, 0.9)
hep <- bind_rows(lapply(cirr, function(f) sim_los(50, hours = 48, hep = 1 - f) |>
  mutate(f_cirrhosis = f)))

hep |>
  select(hour, f_cirrhosis, Cc, Cc_e3174, SBP) |>
  pivot_longer(-c(hour, f_cirrhosis)) |>
  ggplot(aes(hour, value, colour = factor(f_cirrhosis))) +
  geom_line() +
  facet_wrap(~name, scales = "free_y") +
  labs(x = "Time (h)", y = NULL, colour = "f_cirrhosis",
       title = "Replicates Figure 4B of Tensil 2026 (50 mg)")


hep_tab <- hep |>
  group_by(f_cirrhosis) |>
  summarise(auc_los = trap(hour, Cc) / 1000, auc_e3174 = trap(hour, Cc_e3174) / 1000,
            ratio = auc_e3174 / auc_los, min_sbp = min(SBP),
            feces_los = 100 * last(feces) / dose_umol)
hep_tab |>
  rename("f_cirrhosis" = f_cirrhosis, "Losartan AUC (umol*h/L)" = auc_los,
         "E3174 AUC (umol*h/L)" = auc_e3174, "E3174/losartan" = ratio,
         "Min SBP (mmHg)" = min_sbp, "Fecal losartan (% dose)" = feces_los) |>
  knitr::kable(digits = 3)
f_cirrhosis Losartan AUC (umol*h/L) E3174 AUC (umol*h/L) E3174/losartan Min SBP (mmHg) Fecal losartan (% dose)
0.00 1.332 4.337 3.257 96.297 59.856
0.30 2.036 5.034 2.473 97.479 59.753
0.50 2.847 5.473 1.922 99.008 59.598
0.67 4.003 5.571 1.392 101.384 59.333
0.80 5.473 5.016 0.916 104.745 58.939
0.90 7.277 3.571 0.491 109.552 58.392

stopifnot(
  # "As cirrhosis severity increases, losartan plasma ... concentrations rise,
  # while ... the E3174/losartan ratio decrease[s]" (Section 3.4).
  all(diff(hep_tab$auc_los) > 0), all(diff(hep_tab$ratio) < 0),
  # "Fecal excretion remains largely unchanged".
  diff(range(hep_tab$feces_los)) < 5,
  # Attenuated blood-pressure response in severe cirrhosis.
  hep_tab$min_sbp[hep_tab$f_cirrhosis == 0.9] > hep_tab$min_sbp[hep_tab$f_cirrhosis == 0]
)

Renal impairment (Figure 5)

renal_lv <- c(1, 0.95, 0.5, 0.14)
ren <- bind_rows(lapply(renal_lv, function(r) sim_los(50, hours = 48, renal = r) |>
  mutate(renal = r)))
ren_tab <- ren |>
  group_by(renal) |>
  summarise(auc_los = trap(hour, Cc) / 1000, auc_e3174 = trap(hour, Cc_e3174) / 1000,
            urine_los = last(urine), urine_e3174 = last(urine_e3174),
            min_sbp = min(SBP)) |>
  arrange(desc(renal))
ren_tab |>
  rename("Relative renal function" = renal, "Losartan AUC (umol*h/L)" = auc_los,
         "E3174 AUC (umol*h/L)" = auc_e3174, "Urinary losartan (umol)" = urine_los,
         "Urinary E3174 (umol)" = urine_e3174, "Min SBP (mmHg)" = min_sbp) |>
  knitr::kable(digits = 3)
Relative renal function Losartan AUC (umol*h/L) E3174 AUC (umol*h/L) Urinary losartan (umol) Urinary E3174 (umol) Min SBP (mmHg)
1.00 1.332 4.337 1.963 2.417 96.297
0.95 1.335 4.429 1.871 2.346 96.231
0.50 1.367 5.463 1.017 1.528 95.622
0.14 1.395 6.682 0.292 0.525 95.097

stopifnot(
  # "losartan plasma concentrations remain largely unchanged, whereas E3174
  # levels increase, accompanied by reduced urinary excretion of both"
  # (Section 3.5).
  diff(range(ren_tab$auc_los)) / ren_tab$auc_los[1] < 0.1,
  all(diff(ren_tab$auc_e3174) > 0),
  all(diff(ren_tab$urine_los) < 0), all(diff(ren_tab$urine_e3174) < 0),
  # "impaired renal function amplifies losartan's antihypertensive effects".
  all(diff(ren_tab$min_sbp) < 0)
)

ABCB1 activity (Figure 6)

ABCB1 activity is a scaling parameter (f_abcb1, 1 = reference) rather than a covariate. The Shin 2020 simulation uses 1.0, 0.653 and 0.306 for its three diplotype groups (Supplementary Table S5; Section 3.6 names the GG/CC, GT/CT and TT/TT diplotypes).

abcb1_lv <- c(1, 0.653, 0.306)
abc <- bind_rows(lapply(abcb1_lv, function(a) {
  sim_los(50, hours = 48, model = mod |> rxode2::ini(f_abcb1 = a)) |> mutate(abcb1 = a)
}))
#> ℹ change initial estimate of `f_abcb1` to `1`
#> ℹ change initial estimate of `f_abcb1` to `0.653`
#> ℹ change initial estimate of `f_abcb1` to `0.306`
abc_tab <- abc |>
  group_by(abcb1) |>
  summarise(auc_los = trap(hour, Cc) / 1000, auc_e3174 = trap(hour, Cc_e3174) / 1000,
            feces_los = 100 * last(feces) / dose_umol, min_sbp = min(SBP)) |>
  arrange(desc(abcb1))
abc_tab |>
  rename("ABCB1 activity" = abcb1, "Losartan AUC (umol*h/L)" = auc_los,
         "E3174 AUC (umol*h/L)" = auc_e3174, "Fecal losartan (% dose)" = feces_los,
         "Min SBP (mmHg)" = min_sbp) |>
  knitr::kable(digits = 3)
ABCB1 activity Losartan AUC (umol*h/L) E3174 AUC (umol*h/L) Fecal losartan (% dose) Min SBP (mmHg)
1.000 1.332 4.337 59.856 96.297
0.653 1.503 4.836 55.196 95.333
0.306 1.727 5.466 49.304 94.289

stopifnot(
  # "reduced ABCB1 activity leads to increased plasma ... concentrations of
  # both compounds, while fecal excretion of losartan declines" (Section 3.6).
  all(diff(abc_tab$auc_los) > 0), all(diff(abc_tab$auc_e3174) > 0),
  all(diff(abc_tab$feces_los) < 0),
  # Only a slight enhancement of the blood-pressure effect.
  abs(abc_tab$min_sbp[3] - abc_tab$min_sbp[1]) < 5
)

CYP2C9 activity (Figure 7)

The per-study simulations assign a CYP2C9 activity to each genotype group (Supplementary Table S5); for example Li 2009 uses 1.0, 0.525 and 0.17 for its three groups. The paper does not print which diplotype each value belongs to.

cyp_lv <- c(1, 0.8, 0.585, 0.525, 0.385, 0.17)
cyp <- bind_rows(lapply(cyp_lv, function(a) sim_los(50, hours = 48, cyp2c9 = a) |>
  mutate(cyp2c9 = a)))

cyp |>
  select(hour, cyp2c9, Cc, Cc_e3174, SBP) |>
  pivot_longer(-c(hour, cyp2c9)) |>
  ggplot(aes(hour, value, colour = factor(cyp2c9))) +
  geom_line() +
  facet_wrap(~name, scales = "free_y") +
  labs(x = "Time (h)", y = NULL, colour = "CYP2C9 activity",
       title = "Replicates Figure 7B of Tensil 2026 (50 mg)")


cyp_tab <- cyp |>
  group_by(cyp2c9) |>
  summarise(auc_los = trap(hour, Cc) / 1000, auc_e3174 = trap(hour, Cc_e3174) / 1000,
            min_sbp = min(SBP)) |>
  arrange(desc(cyp2c9)) |>
  mutate(auc_los_ratio = auc_los / auc_los[1])
cyp_tab |>
  rename("CYP2C9 activity" = cyp2c9, "Losartan AUC (umol*h/L)" = auc_los,
         "E3174 AUC (umol*h/L)" = auc_e3174, "Min SBP (mmHg)" = min_sbp,
         "Losartan AUC ratio vs activity 1" = auc_los_ratio) |>
  knitr::kable(digits = 3)
CYP2C9 activity Losartan AUC (umol*h/L) E3174 AUC (umol*h/L) Min SBP (mmHg) Losartan AUC ratio vs activity 1
1.000 1.332 4.337 96.297 1.000
0.800 1.520 4.214 96.991 1.142
0.585 1.842 3.998 98.215 1.383
0.525 1.968 3.911 98.710 1.478
0.385 2.372 3.625 100.330 1.781
0.170 3.631 2.655 105.692 2.726

stopifnot(
  # "reduced-function alleles lead to a 1.6- to 3-fold increase in losartan
  # AUC" (Abstract) -- the poorest-function group lands inside that range.
  cyp_tab$auc_los_ratio[cyp_tab$cyp2c9 == 0.17] > 1.6,
  cyp_tab$auc_los_ratio[cyp_tab$cyp2c9 == 0.17] < 3,
  all(diff(cyp_tab$auc_los) > 0), all(diff(cyp_tab$auc_e3174) < 0),
  # "diminished blood pressure reduction".
  all(diff(cyp_tab$min_sbp) > 0)
)

Multiple dosing

The model handles repeated doses through ordinary dosing records. This illustrative simulation gives 50 mg once daily for seven days at the 72 kg body weight used for the Fischer 2002 multiple-dose study (Table 1, Table S5); the paper does not print that study’s regimen.

ev_md <- rbind(
  data.frame(time = 0, amt = 50, evid = 1L, cmt = "depot", addl = 6L, ii = 1440),
  data.frame(time = seq(0, 7 * 1440, by = 30), amt = NA_real_, evid = 0L,
             cmt = "venous", addl = NA_integer_, ii = NA_real_)
)
ev_md$WT <- 72
ev_md$RENALFUNC_REL <- 1
ev_md$HEPFUNC_REL <- 1
ev_md$CYP2C9 <- 1
md <- as.data.frame(rxode2::rxSolve(mod, ev_md, returnType = "data.frame"))
md$hour <- md$time / 60

md |>
  select(hour, Cc, Cc_e3174) |>
  pivot_longer(-hour) |>
  ggplot(aes(hour, value, colour = name)) +
  geom_line() +
  labs(x = "Time (h)", y = "Venous plasma concentration (nmol/L)", colour = NULL)


day1 <- max(md$Cc_e3174[md$hour <= 24])
day7 <- max(md$Cc_e3174[md$hour > 144])
stopifnot(day7 / day1 > 1, day7 / day1 < 1.2) # little accumulation, t1/2 << 24 h

Non-compartmental analysis

PKNCA on single oral doses of 10, 25, 50 and 100 mg, for losartan and E3174.

nca_doses <- c(10, 25, 50, 100)
nca_sim <- bind_rows(lapply(seq_along(nca_doses), function(i) {
  sim_los(nca_doses[i], hours = 48, by_min = 10) |>
    mutate(id = i, treatment = paste(nca_doses[i], "mg"), dose = nca_doses[i])
}))

conc_long <- nca_sim |>
  select(id, treatment, hour, losartan = Cc, E3174 = Cc_e3174) |>
  pivot_longer(c(losartan, E3174), names_to = "analyte", values_to = "conc") |>
  filter(!is.na(conc)) |>
  mutate(conc = pmax(conc, 0) / 1000) # umol/L

dose_df <- nca_sim |>
  distinct(id, treatment, dose) |>
  mutate(hour = 0) |>
  as.data.frame()

o_conc <- PKNCA::PKNCAconc(as.data.frame(conc_long), conc ~ hour | treatment + id / analyte)
o_dose <- PKNCA::PKNCAdose(dose_df, dose ~ hour | treatment + id)
nca_res <- PKNCA::pk.nca(PKNCA::PKNCAdata(
  o_conc, o_dose,
  intervals = data.frame(start = 0, end = Inf, cmax = TRUE, tmax = TRUE,
                         aucinf.obs = TRUE, half.life = TRUE)
))

nca_tab <- as.data.frame(nca_res$result) |>
  filter(PPTESTCD %in% c("cmax", "tmax", "aucinf.obs", "half.life")) |>
  select(treatment, analyte, PPTESTCD, PPORRES) |>
  pivot_wider(names_from = PPTESTCD, values_from = PPORRES) |>
  arrange(analyte, as.numeric(sub(" mg", "", treatment)))
nca_tab |>
  rename("Dose" = treatment, "Analyte" = analyte, "Cmax (umol/L)" = cmax,
         "tmax (h)" = tmax, "AUCinf (umol*h/L)" = aucinf.obs,
         "Half-life (h)" = half.life) |>
  knitr::kable(digits = 3)
Dose Analyte Cmax (umol/L) tmax (h) Half-life (h) AUCinf (umol*h/L)
10 mg E3174 0.105 2.833 10.809 0.897
25 mg E3174 0.258 2.833 10.811 2.236
50 mg E3174 0.502 2.833 10.820 4.449
100 mg E3174 0.946 3.000 10.835 8.793
10 mg losartan 0.075 1.000 10.960 0.258
25 mg losartan 0.192 1.000 10.954 0.659
50 mg losartan 0.401 1.000 10.954 1.366
100 mg losartan 0.876 1.000 10.956 2.946

Over 10-100 mg losartan AUCinf rises 11.4-fold and E3174 AUCinf 9.8-fold for a 10-fold dose increase (saturable CYP2C9 conversion), and the terminal half-life of about 11 h is the slow tail of a multi-phasic profile; Figure 3C plots half-lives of about 6.6-8.3 h, which the authors computed with their own terminal-phase window, so half-life is not compared. Figure 3C also plots AUC and Cmax against dose; the values read off that panel at 50 mg are compared below.

fig3c <- tibble::tribble(
  ~treatment, ~analyte, ~cmax, ~aucinf.obs,
  "50 mg", "losartan", 0.40, 1.3,
  "50 mg", "E3174", 0.50, 4.4
)
sim_50 <- as.data.frame(nca_res$result) |> filter(treatment == "50 mg")
nca_cmp <- nlmixr2lib::ncaComparisonTable(
  simulated = sim_50,
  reference = fig3c,
  by = c("treatment", "analyte"),
  params = c("cmax", "aucinf.obs"),
  units = c(cmax = "umol/L", aucinf.obs = "umol*h/L"),
  tolerance_pct = 20
)
knitr::kable(nca_cmp, caption = "Simulated NCA vs. Figure 3C of Tensil 2026 (read off the plot). * differs by >20%.")
Simulated NCA vs. Figure 3C of Tensil 2026 (read off the plot). * differs by >20%.
NCA parameter treatment analyte Reference Simulated % diff
Cmax (umol/L) 50 mg losartan 0.4 0.401 +0.2%
Cmax (umol/L) 50 mg E3174 0.5 0.502 +0.5%
AUC0-∞ (obs) (umol*h/L) 50 mg losartan 1.3 1.37 +5.1%
AUC0-∞ (obs) (umol*h/L) 50 mg E3174 4.4 4.45 +1.1%

nca_50 <- nca_tab |> filter(treatment == "50 mg")
stopifnot(
  abs(nca_50$cmax[nca_50$analyte == "losartan"] / 0.40 - 1) < 0.1,
  abs(nca_50$cmax[nca_50$analyte == "E3174"] / 0.50 - 1) < 0.1,
  abs(nca_50$aucinf.obs[nca_50$analyte == "E3174"] / 4.4 - 1) < 0.1,
  # Losartan peaks within about 1 h and E3174 later (about 3 h), as in the
  # clinical data quoted in the Introduction (1-2 h and 3-4 h).
  nca_50$tmax[nca_50$analyte == "losartan"] <= 1.5,
  nca_50$tmax[nca_50$analyte == "E3174"] > 2, nca_50$tmax[nca_50$analyte == "E3174"] < 4
)

Oral bioavailability

po <- sim_los(50, hours = 96, by_min = 5)
iv <- sim_los(30, route = "depot_iv", hours = 96, by_min = 5)
f_oral <- (trap(po$hour, po$Cc) / 50) / (trap(iv$hour, iv$Cc) / 30)
f_oral
#> [1] 0.2935073
stopifnot(f_oral > 0.2, f_oral < 0.4)

The model’s oral bioavailability of losartan is about 29%, close to the approximately 33% quoted in the paper’s Introduction.

Assumptions and deviations / Errata

1. The model equations come from the authors’ SBML archive. The paper and its supplement give the parameter values but no rate law or ODE. The equations were taken from the authors’ archive at release 0.8.0, the release the paper cites; its parameters match Supplementary Tables S2, S3 and S7. A later release (0.9.0) re-fits several parameters (its release notes list, for example, Kp_los 3.262 to 1.167 and E50_e3174 2.91e-4 to 2.07e-4 mmol/L) and is not the published model.

2. Table S3 prints the aldosterone secretion rate in the wrong unit. Table S3 lists ALDSEC_k = 1.011 mmole/min, Table S7 and the SBML give 1.01e-6 mmol/min. Because the aldosterone degradation constant is ALDSEC_k / ald_ref, only the second value keeps aldosterone at its 250 pmol/L baseline with the published dynamics, so 1.0105e-6 mmol/min is used.

3. Aldosterone secretion does not depend on angiotensin II. Results 3.2 says “angiotensin II stimulates aldosterone secretion”, but in the archived model aldosterone secretion is ALDSEC_k * (1 - fe_e3174) and angiotensin II has no downstream effect. The model is carried over as archived, so angiotensin I and II are outputs only. Renin, too, responds only to E3174 (there is no angiotensin-II feedback), and the renin turnover rate constant (rensec_k / (ren_ref * vplasma), about 2e7 /min) makes renin follow E3174 instantaneously.

4. Omitted states. The SBML carries seven further states that never change the solution: three bile pools whose inflow and outflow are the same expression (so they stay at zero; the biliary flux goes straight to the gut lumen here), an unused stomach pool, the constant angiotensinogen pool (here the parameter anggen_ref), and two cumulative-infusion counters. An intravenous infusion (SBML Ri_los, Ri_e3174) is given here as an ordinary rate into depot_iv or depot_iv_e3174, which is the same input.

5. Units are rescaled. The SBML works in mmol and mmol/L. Drug states here are in umol/L, the RAAS hormones in pmol/L and the outputs Cc, Cc_e3174, Cc_l158 in nmol/L, with the Michaelis constant and E50 converted inside model(). The rescaling is exact (the roadrunner comparison above confirms it) and keeps baseline renin (1e-9 mmol/L) well above the solver’s absolute tolerance.

6. Study-specific baselines. The per-study simulations reset the baseline hormone levels and blood pressures to each study’s values (Supplementary Table S5). These are the parameters ren_ref, ang1_ref, ang2_ref, ald_ref, sbp_ref and dbp_ref (pmol/L and mmHg); override them with ini() to reproduce a given study. The defaults are the model’s reference values (renin 1, angiotensin I 10, angiotensin II 8, aldosterone 250 pmol/L; 120/80 mmHg).

7. Covariate encoding. Hepatic impairment is the complement of the source’s cirrhosis severity (HEPFUNC_REL = 1 - f_cirrhosis). The paper does not print a genotype-to-activity map for CYP2C9 or ABCB1, nor a KDIGO-to-renal-function map; the values used in each study simulation are listed in Supplementary Table S5 and quoted above. ABCB1 activity is kept as the parameter f_abcb1 because the library has no continuous P-glycoprotein activity covariate.

8. Known model limitations reported by the authors. The model under-predicts E3174 accumulation in cirrhosis and over-predicts it in renal impairment relative to clinical data (Sections 3.4 and 3.5), and Figure 2 reports higher urinary recovery than the model produces. These are properties of the published model and are reproduced, not corrected.