Skip to contents

Dexmedetomidine (Smuszkiewicz 2017)

Source: vignettes/articles/Smuszkiewicz_2017_dexmedetomidine.Rmd
Smuszkiewicz_2017_dexmedetomidine.Rmd

Model and source

  • Citation: Smuszkiewicz P, Wiczling P, Ber J, Warzybok J, Malkiewicz T, Matysiak J, Klupczynska A, Trojanowska I, Kokot Z, Grzeskowiak E, Krzyzanski W, Bienert A. Pharmacokinetics of dexmedetomidine during analgosedation in ICU patients. J Pharmacokinet Pharmacodyn. 2018;45(2):277-284. doi:10.1007/s10928-017-9564-7.
  • Description: Two-compartment population PK model for intravenous dexmedetomidine continuous infusion in adult ICU patients undergoing analgosedation (Smuszkiewicz 2017). 27 medical and surgical ICU patients (17 male, 10 female; median age 59.5 y, median weight 75 kg) received continuous infusions of 0.1-1.5 ug/kg/h for 23.7-102 h. Age, sex, body weight, infusion duration, pretreatment SOFA score, and inotrope use were screened as covariates but none reached statistical significance, so the final model contains no covariate effects. IIVs on Vc, CL, Vp, and Q are diagonal (no clear correlations). Proportional residual error.
  • Article: https://doi.org/10.1007/s10928-017-9564-7 (open access)

Population

Smuszkiewicz et al. (2017) studied 27 adult patients (17 male, 10 female) in a mixed medical and surgical intensive care unit at Poznan University of Medical Sciences. Median age was 59.5 years (range 19-84) and median body weight was 75 kg (range 45-100). All subjects required analgosedation and mechanical ventilation, or treatment of hyperactive delirium/agitation refractory to haloperidol. The pretreatment SOFA score median was 12 (range 5-16), 21 of 27 received inotropes, and the median infusion duration was 42.8 h (range 23.7-102). Dexmedetomidine (Dexdor, Orion Pharma) was infused continuously without a loading dose at 0.4-1.5 ug/kg/h, titrated to a modified Ramsay sedation score of 2-3. Baseline demographics are tabulated in the source paper Table 1; the same information is available programmatically via readModelDb("Smuszkiewicz_2017_dexmedetomidine")$population.

Age, sex, body weight, infusion duration, pretreatment SOFA, and inotrope use were screened as covariates on Vc, CL, Vp, and Q. None reached statistical significance, so the final model has no covariate effects. The model file’s covariateData slot is therefore empty.

Source trace

Per-parameter origin is recorded as an in-file comment next to each ini() entry in inst/modeldb/specificDrugs/Smuszkiewicz_2017_dexmedetomidine.R. The table below collects them in one place for review.

Equation / parameter Value Source location
lvc (Vc) log(27.0 L) Table 2 (theta_VC)
lcl (CL) log(38.5 L/h) Table 2 (theta_CL)
lvp (Vp) log(87.6 L) Table 2 (theta_VT)
lq (Q) log(46.4 L/h) Table 2 (theta_Q)
etalvc 0.9312 Table 2 (omega^2_VC = 124 %CV; log(1 + 1.24^2))
etalcl 0.3361 Table 2 (omega^2_CL = 63.2 %CV; log(1 + 0.632^2))
etalvp 0.5832 Table 2 (omega^2_VT = 89.0 %CV; log(1 + 0.89^2))
etalq 0.5034 Table 2 (omega^2_Q = 80.9 %CV; log(1 + 0.809^2))
propSd 0.24 Table 2 (sigma^2 = 24 %CV)
Two-compartment ODE n/a PK model equations (Methods / PK model section)
IIV diagonal n/a Results: “no clear correlations between the eta estimates”
No covariate effects n/a Results: covariates screened but none statistically significant

Virtual cohort

The original observed data are not publicly distributed. The cohort below uses 27 simulated subjects whose age, weight, and sex distributions approximate the published Table 1 demographics. Two infusion-duration groups (a short 24 h infusion and a long 72 h infusion at the mid-range rate of 0.7 ug/kg/h) bracket the observed range and let PKNCA produce per-group summaries.

set.seed(20180101)

n_per_group <- 27L
infusion_rate_ugkgh <- 0.7  # ug/kg/h, mid-range of 0.4-1.5

make_cohort <- function(n, infusion_h, group_label, id_offset = 0L) {
  # Demographics drawn to approximate Smuszkiewicz 2017 Table 1
  wt <- pmin(pmax(rnorm(n, mean = 72, sd = 14), 45), 100)
  age <- pmin(pmax(rnorm(n, mean = 58, sd = 17), 19), 84)
  sexf <- as.integer(runif(n) < 0.37)  # 37% female per Table 1

  ids <- id_offset + seq_len(n)
  # ug/h infusion amount per subject = rate(ug/kg/h) * WT(kg) * duration(h)
  amt_ug <- infusion_rate_ugkgh * wt * infusion_h
  # rxode2 expects a rate column (ug/h) for an infusion-duration dose
  rate_ugh <- infusion_rate_ugkgh * wt

  dose_rows <- data.frame(
    id   = ids,
    time = 0,
    amt  = amt_ug,
    rate = rate_ugh,
    evid = 1L,
    cmt  = "central",
    WT   = wt,
    AGE  = age,
    SEXF = sexf,
    group = group_label,
    stringsAsFactors = FALSE
  )

  during_inf <- c(0, 0.5, 1, 2, 4, 6, 8, 12, 16, 20, 24)
  during_inf <- during_inf[during_inf <= infusion_h]
  if (infusion_h > 28) {
    during_inf <- c(during_inf, seq(28, infusion_h, by = 8))
  }
  post_inf <- infusion_h + c(5/60, 10/60, 20/60, 60/60, 2, 4, 6, 12)
  obs_times <- sort(unique(c(during_inf, post_inf)))
  obs_times <- obs_times[obs_times >= 0]

  obs_rows <- expand.grid(id = ids, time = obs_times,
                          KEEP.OUT.ATTRS = FALSE, stringsAsFactors = FALSE)
  obs_rows$amt  <- NA_real_
  obs_rows$rate <- NA_real_
  obs_rows$evid <- 0L
  obs_rows$cmt  <- "central"
  # Carry per-subject covariates onto observation rows (one-per-subject)
  obs_rows <- merge(
    obs_rows,
    dose_rows[, c("id", "WT", "AGE", "SEXF", "group")],
    by = "id"
  )

  out <- rbind(
    dose_rows[, c("id","time","amt","rate","evid","cmt","WT","AGE","SEXF","group")],
    obs_rows [, c("id","time","amt","rate","evid","cmt","WT","AGE","SEXF","group")]
  )
  out[order(out$id, out$time, -out$evid), ]
}

events <- dplyr::bind_rows(
  make_cohort(n_per_group, infusion_h = 24, group_label = "24 h",
              id_offset =   0L),
  make_cohort(n_per_group, infusion_h = 72, group_label = "72 h",
              id_offset = 100L)
)

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

Simulation 

mod <- readModelDb("Smuszkiewicz_2017_dexmedetomidine")
sim <- rxode2::rxSolve(mod, events = events, keep = c("group", "WT")) |>
  as.data.frame() |>
  dplyr::filter(time > 0)  # drop the t=0 placeholder so log-y plots behave
#> ℹ parameter labels from comments will be replaced by 'label()'

Replicate published figure (prediction-corrected VPC region)

Smuszkiewicz 2017 Figure 2 is a prediction-corrected VPC of dexmedetomidine concentration vs. time. The plot below replicates the figure qualitatively as the 10th, 50th, and 90th percentile envelope of simulated concentrations from the 27-subject 72 h infusion cohort – the figure’s longest infusion arm. The band is bounded by the same percentiles used in the source paper.

# Replicates Figure 2 of Smuszkiewicz 2017: percentile envelope of Cc over time
# (72 h infusion arm). The VPC scale in the source is ng/mL, log10 y.
sim_72 <- sim |> dplyr::filter(group == "72 h")

vpc_72 <- sim_72 |>
  dplyr::filter(Cc > 0) |>
  dplyr::mutate(tbin = cut(time, breaks = seq(0, 96, by = 4))) |>
  dplyr::group_by(tbin) |>
  dplyr::summarise(
    t_mid = mean(time),
    Q10   = quantile(Cc, 0.10),
    Q50   = quantile(Cc, 0.50),
    Q90   = quantile(Cc, 0.90),
    .groups = "drop"
  )

ggplot(vpc_72, aes(t_mid, Q50)) +
  geom_ribbon(aes(ymin = Q10, ymax = Q90), alpha = 0.25, fill = "steelblue") +
  geom_line(colour = "steelblue", linewidth = 0.9) +
  scale_y_log10() +
  labs(
    x = "Time after infusion start (h)",
    y = "Dexmedetomidine concentration (ng/mL)",
    title = "Simulated VPC, 72 h infusion at 0.7 ug/kg/h",
    caption = paste(
      "Replicates Figure 2 of Smuszkiewicz 2017 (10th, 50th, 90th percentiles)."
    )
  )

PKNCA validation

PKNCA is run separately for each infusion-duration group. AUCinf is reported relative to the end of infusion (steady-state-like for a continuous infusion over many half-lives) and the half-life is the apparent terminal half-life from the post-infusion samples. Dexmedetomidine’s reported terminal half-life in adults is around 2-3 h; the simulated value should match.

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

dose_nca <- events |>
  dplyr::filter(evid == 1) |>
  dplyr::transmute(id, time, dose = amt, group)

conc_obj <- PKNCA::PKNCAconc(sim_nca, Cc ~ time | group + id)
dose_obj <- PKNCA::PKNCAdose(dose_nca, dose ~ time | group + id)

intervals <- data.frame(
  start      = 0,
  end        = Inf,
  cmax       = TRUE,
  tmax       = TRUE,
  aucinf.obs = TRUE,
  half.life  = TRUE
)

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

nca_summary <- as.data.frame(nca_res$result) |>
  dplyr::filter(PPTESTCD %in% c("cmax", "tmax", "aucinf.obs", "half.life")) |>
  dplyr::group_by(group, PPTESTCD) |>
  dplyr::summarise(
    median = round(median(PPORRES, na.rm = TRUE), 3),
    p10    = round(quantile(PPORRES, 0.10, na.rm = TRUE), 3),
    p90    = round(quantile(PPORRES, 0.90, na.rm = TRUE), 3),
    .groups = "drop"
  )
knitr::kable(
  nca_summary,
  caption = "Simulated NCA parameters by infusion-duration group (median, 10th and 90th percentiles)."
)
Simulated NCA parameters by infusion-duration group (median, 10th and 90th percentiles).
group PPTESTCD median p10 p90
24 h aucinf.obs NA NA NA
24 h cmax 1.253 0.396 2.730
24 h half.life 5.172 1.127 8.849
24 h tmax 24.000 24.000 24.000
72 h aucinf.obs NA NA NA
72 h cmax 1.280 0.679 2.820
72 h half.life 3.878 1.416 9.081
72 h tmax 68.000 48.800 68.000

Comparison against published values

Smuszkiewicz 2017 does not report a Cmax / AUC summary table for the trial cohort. Instead, the paper compares its structural parameters with literature ranges (Discussion). The comparison below uses those literature anchors and the model’s derived steady-state volume of distribution Vss = Vc + Vp.

Quantity Smuszkiewicz 2017 / literature Model (this vignette)
Systemic clearance CL (L/h) 38.5 (this study); literature 33.7-53.4 38.5 (typical), Table 2
Vss = Vc + Vp (L) 27.0 + 87.6 = 114.6 (this study); literature 79.3-161.3 114.6 (typical)
Terminal half-life (h) not tabulated; literature ~2-3 h for adults see half.life row above
Inter-individual CV of CL (%) 63.2 encoded as etalcl ~ 0.3361 (log(1.632) for CV 63.2%)

The simulated CL and Vss are exact reproductions of the Table 2 typical values, and the derived half-life from the PKNCA summary above should fall within the 2-3 h range typical for dexmedetomidine in adults.

Assumptions and deviations

  • Virtual covariate distributions were drawn to approximate the source Table 1 demographics (weight ~ N(72, 14) kg truncated to 45-100; age ~ N(58,
    1. y truncated to 19-84; 37% female). None of these covariates enters the structural model – the covariate search retained no effects – so the virtual values are purely descriptive metadata carried for traceability.
  • Infusion regimens are 24 h and 72 h at 0.7 ug/kg/h (mid-range of the observed 0.4-1.5 ug/kg/h). The paper itself uses heterogeneous, physician- titrated infusion histories; the two regimens here bracket the duration range and let PKNCA produce a stable terminal half-life and Vss summary.
  • No published per-cohort NCA table to compare against; the source paper reports structural and IIV parameters only. The “Comparison against published values” section therefore compares model output against the structural parameters from Table 2 and the literature ranges discussed in Smuszkiewicz 2017 (~33.7-53.4 L/h CL; ~79.3-161.3 L Vss in adults).
  • No covariate effects were retained in the source’s final model, so the packaged model has none either. Downstream users wanting a weight-on-CL sensitivity analysis should fork the model and add e_wt_cl themselves.
  • Units conversion is exact, not approximate: dosing in ug and volume in L produces concentrations in ug/L, and 1 ug/L = 1 ng/mL, so the model needs no scale factor between the dosing units and the concentration units. checkModelConventions() flags this for inspection (unitsdosinginug,unitsdosing in ug, unitsconcentration in ng/mL) but the conversion is correct as written.