Population pharmacokinetics of sirolimus in adult renal transplant patients and Bayesian estimation of individual systemic exposure
Djebli, Nassim 1, Pierre Marquet1,2, Yann Le Meur1,3, Annick Rousseau1,4
EA3838, Laboratory of Pharmacology, Faculty of Medicine, Limoges University, Limoges, France
Introduction: The present study aimed at (1) analyzing the population pharmacokinetics (popPK) of sirolimus in renal transplant recipients and studying the influence of different potential covariates (including genetic polymorphisms of metabolic enzymes) on sirolimus pharmacokinetic (PK) parameters; and (2) developing a Bayesian estimator able to reliably estimate the individual PK parameters and exposure indices.
Methods: Sirolimus population pharmacokinetics was studied in 22 adult kidney transplant patients using a population approach performed in Nonmem. Four to five full concentration-time profiles (938 sirolimus whole blood samples) were collected in each patient, at days 7, 14 and months 1, 2 and 3 post-transplantation. The covariates collected included demographic characteristics (weight, age, height, sex, body mass index, body surface area), biological factors (hematocrit, red blood cells, hemoglobin level, leucocytes, ASAT and ALAT activities, cholesterol and triglycerides levels, platelets, proteins, serum albumin, urea and serum creatinine), and the CYP3A5, CYP3A4 and MDR1 genotypes. Plots were used to screen for potential covariate relationships. Taking into account the size of the dataset, only very few candidate covariates were investigated. The final popPK model was validated using 200 bootstrap runs. A Bayesian estimator was built using 3 sampling times. Therefore, the circular permutation method was applied. The database was divided into four groups including each 25% of the studied profiles (corresponding to 25% of the patients). Then, four different subsets of 3 groups (i.e., 3 * 25 %) were constituted. The population parameter estimates in these four different subsets were used as priors to compute the individual PK parameters of the 25% remaining profiles. In a further step, this estimator has been prospectively evaluated in 34 PK-profiles collected in 28 patients.
Results: A two-compartment open model with first order elimination and Erlang’s distribution to describe the absorption phase (with 3 sequential delay-compartments placed upstream to the central compartment and connected by an identical exiting rate constant, ktr) best fitted the data. The mean PK parameter estimates were 5.25 h-1, 218 L and 292 L for the Erlang rate constant (Ktr), the apparent volume of the central (Vc/F) and peripheral (Vp/F) compartments, respectively. The CYP3A5*1/*3 polymorphism significantly influenced sirolimus apparent clearance: CL/F = 14.1 L.h-1 for CYP3A5 non expressers (CYP3A5*3/*3 genotype) versus 28.3 L.h-1 for CYP3A5 expressers (CYP3A5*1/*3 and *1/*1 genotypes). This finding is in accordance with previous pharmacogenetic studies.The standard errors of the parameter estimates were less than 15% for all the estimated parameters. Bootstrap resampling achieved very good agreement with mean PK parameters and their variabilities. MAP Bayesian forecasting allowed accurate prediction of sirolimus AUC0-24h using a combination of three sampling times (0, 1 and 3 hours post-dose), with a low mean bias of -6.6 % (range: -28.8% to +19.1%), and a good precision (RMSE = 10.6 %); only 3/34 AUC biases were out of the range > 20%.
Conclusion: This study presents an accurate popPK model showing the significant influence of the CYP3A5*1/*3 polymorphism on sirolimus apparent clearance, and a Bayesian estimator accurately predicting sirolimus pharmacokinetics taking into account this polymorphism, for the first time.