Objectives: To explore the contribution of renal failure to nimodipine overall pharmacokinetic variability after single and multiple oral dosing and to develop a population pharmacokinetic model by means of the nonparametric expectation maximization (NPEM2) algorithm based on sampled individual drug concentrations close to the estimated patients' C(SS)avs (NPEM2-C(SS)av).
Patients, materials and methods: 24 hypertensive patients with normal and reduced renal function, without clinical and laboratory data for hepatic dysfunction, were enrolled in the study and their nimodipine plasma levels were analyzed by means of a parametric and nonparametric population pharmacokinetic modeling using a maximum a posteriori Bayesian (MAPB) estimator in an iterative two-stage Bayesian population modeling program and NPEM2-algorithm.
Results: Comparison of parameter dispersion revealed higher variability of nimodipine disposition after the first dose than at steady-state except for apparent volume of distribution at steady-state, V(SS)/F, whose variability increased from 98% to 223%. The most variable was mean residence time, MRT, whose coefficient of variation (CV) was 288% after the first dose and decreased by more than 2 times at steady-state, followed by terminal elimination half-life, t(1/2el), with CV = 171% after the first dosing and decreasing by more than 3 times at steady-state. Concerning the impact of renal failure on disposition parameters variability, patients with slightly to moderately reduced renal function, creatinine clearances between 51 to 80 and 25 to 50 ml/min, resp., stated higher variation than patients with more definitively altered renal function. The validation of NPEM2-C(SS)av population model was performed by using a set of 272 individual plasma drug concentrations, including trough levels as well as concentrations belonging to mono-exponential elimination phases after single and multiple dosing. Bayesian forecasting, using 4 trough levels per patient as Bayesian priors, revealed highly significant correlation between observed and population model predicted drug concentrations (r = 0.526, p < 0.0001). The predictive performance of NPEM2-C(SS)av population model was characterized by low bias (mean error = -0.48 microg/l, 95% CI = -0.99-0.04 microg/l), and good precision (root mean squared error = 4.32 microg/l, 95% CI = -2.53-11.17 microg/l).
Conclusions: As predicted for high hepatic clearance drugs [Rowland 1985], nimodipine parameters variability decreased after reaching steady-state. NPEM2-C(SS)av population model demonstrated high accuracy and precision in predicting drug levels from terminal exponential phase including trough levels at steady-state.