Characterization of the highly variable bioavailability of tiludronate in normal volunteers using population...

Currently, the use of classical bioequivalence criteria is being called into question for certain classes of drugs such as bisphosphonates. These compounds typically possess a wide therapeutic index but may be characterized by low and variable absorption. The purpose of this communication was to characterize the highly variable bioavailability of tiludronate using a population pharmacokinetic method (NONMEM program) and compare the results to a standard 2 way cross-over bioequivalence trial in healthy subjects. Over 3500 plasma samples from 153 healthy subjects, representing 12 different clinical trials were pooled for mixed effect modeling purposes (complete data set). These studies, conducted under single and multiple dose conditions, contained all the directly comparable data available in healthy subjects administered a 400 mg dose of tiludronate. A two compartment model with first order absorption was fit to the plasma concentration-time data and a term for relative bioavailability (BA) was included. Intersubject and residual variability were modeled using a constant coefficient of variation (CCV) model. A pilot model development data set was obtained from a 24 subject cross-over bioequivalence study. Population estimates of BA and its associated 90% confidence interval of 1.12 and 0.89-1.35 compared favorably to standard bioequivalence methodology (1.15 and 0.93-1.42, respectively). Since a good fit of predicted and observed plasma concentrations as well as estimates of BA were obtained, a two compartment model with a term for BA was then applied to the complete data set. Under these conditions, BA and its 90% confidence interval were found to be 1.17 and 0.98-1.36. Intersubject variability of 31%, compared with 38% in the pilot model development data set and residual variability of 38% were seen. No differences in absorption characteristics as measured by Ka were found. Good agreement between the population pharmacokinetic parameters were observed when the pilot data set was compared with the full data set. The proposed model was confirmed by creating 10 additional smaller data sets that were matched for the number of subjects given both formulations under single and multiple dose conditions. No change in the estimate of BA was observed under these study conditions. This study demonstrated that population pharmacokinetic methodology can be applied successfully to problematical bioequivalence issues that may occur during the development process. Increasing the number of subjects in the overall analysis did not alter the estimate of BA or its 90% confidence interval, when compared to the original cross-over bioequivalence study. Bayesian approaches can be of value in large clinical trials where typically relatively few plasma samples are obtained from individual subjects.