Prediction of Intra-individual Variability in Bioequivalence Studies of 278 Formulations: Comprehensive Analysis Using Physicochemical and Pharmacokinetic Parameters

Abstract

The purpose of the present study was to predict the intra-individual variability (%CV_intra) values of C_max using observed parameters of physicochemical and pharmacokinetic for a variety of formulations. A database was used to summarize the parameters of clinical bioequivalence (BE) studies of oral drugs, including the highest dose tablets, orally disintegrating tablets (ODT), and capsules (278 formulations [238 compounds]). As explanatory variables, %CV_intra, inter-individual variability (%CV_inter), absolute bioavailability (BA), T_max, t_1/2, dose number (Do), and dissolution rate (D%) were selected. Explanatory variables correlated with %CV_intra were identified by multivariate analysis and grouped quantitatively by K-means clustering analysis. The %CV_intra predictions compared three models of multiple regression, boosting tree, and neural network. In the neural network, the coefficient of determination (R²) and the root mean square error (RMSE) were the best, with good correlation between the predicted and observed values of the test data (R² = 0.69). The explanatory variables used in this study are readily available from the literature of reference formulation and in vitro measurement. Therefore, predicting %CV_intra for C_max without conducting pilot studies is useful for clinical planning in the early stages of generic drug development. We believe that we could further contribute to speeding up and reducing the cost of generic drug development.