Population Pharmacokinetics of Fentanyl in the Critically Ill

Abstract

Objective: To characterize fentanyl population pharmacokinetics in patients with critical illness and identify patient characteristics associated with altered fentanyl concentrations.

Design: Prospective cohort study.

Setting: Medical and surgical ICUs in a large tertiary care hospital in the United States.

Patients: Patients with acute respiratory failure and/or shock who received fentanyl during the first 5 days of their ICU stay.

Measurements and main results: We collected clinical and hourly drug administration data and measured fentanyl concentrations in plasma collected once daily for up to 5 days after enrollment. Among 337 patients, the mean duration of infusion was 58 hours at a median rate of 100 μg/hr. Using a nonlinear mixed-effects model implemented by NONMEM, we found that fentanyl pharmacokinetics were best described by a two-compartment model in which weight, severe liver disease, and congestive heart failure most affected fentanyl concentrations. For a patient population with a mean weight of 92 kg and no history of severe liver disease or congestive heart failure, the final model, which performed well in repeated 10-fold cross-validation, estimated total clearance, intercompartmental clearance (Q), and volumes of distribution for the central (V1) and peripheral compartments (V2) to be 35 L/hr (95% CI, 32-39 L/hr), 55 L/hr (95% CI, 42-68 L/hr), 203 L (95% CI, 140-266 L), and 523 L (95% CI, 428-618 L), respectively. Severity of illness was marginally associated with fentanyl pharmacokinetics but did not improve the model fit after liver and heart diseases were included.

Conclusions: In this study, fentanyl pharmacokinetics during critical illness were strongly influenced by severe liver disease, congestive heart failure, and weight, factors that should be considered when dosing fentanyl in the ICU. Future studies are needed to determine if data-driven fentanyl dosing algorithms can improve outcomes for ICU patients.

Figure 1

Observed plasma fentanyl concentrations versus population predicted plasma fentanyl concentrations (A) and individual predicted plasma fentanyl concentrations on log-log axes (B); Conditional weighted residuals versus time after dose (C); Histogram of conditional weighted residuals (D).

Figure 2

Observed and predicted fentanyl concentrations as determined by repeated 10-fold cross-validation. All observed fentanyl plasma concentrations that were measured during the 5-day study period are shown as black empty dots. Additionally, the orange line (and gray band) shows the median (and 95% prediction interval) fentanyl plasma concentration predicted in cross-validation models; i.e., the orange line shows the 50^th percentile of predicted concentrations at a given point in time, and the lower and upper bounds of the gray band show the 2.5^th and 97.5^th concentrations predicted at a given point in time.

Figure 3

Predicted fentanyl concentrations over time for patients weighing 59, 86, or 128 kg, which were the 10^th, 50^th, and 90^th percentiles in our study population: (A) no history of severe liver disease (SLD) or congestive heart failure (CHF); (B) any history of severe liver disease (SLD) or congestive heart failure (CHF). In these simulations, fentanyl is continuously infused at 100 µg/hr for 3 days, after which the infusion is stopped (gray arrow), and four 100 µg bolus doses (some of which are indicated by orange arrows) are given at random times.

Figure 4

Predicted fentanyl concentrations over time for patients weighing 59, 86, or 128 kg, which were the 10^th, 50^th, and 90^th percentiles in our study population, presented as percent of the predicted fentanyl concentrations for patients with the median weight: (A) no history of severe liver disease (SLD) or congestive heart failure (CHF); (B) any history of severe liver disease (SLD) or congestive heart failure (CHF). In these simulations, fentanyl is continuously infused at 100 µg/hr for 3 days, after which the infusion is stopped, and four 100 µg bolus doses are given at random times.