Abstract
The objective of this study was to propose a new population pharmacokinetic analysis method that is applicable for groups where the distribution of each parameter is a normal mixture. In this study, using the extended least squares (ELS) technique, which is one of the population pharmacokinetic analysis methods, population pharmacokinetic parameters were estimated for plasma concentrations observed from many people. This method was composed of two steps. First, a set of population pharmacokinetic parameters with varying initial values was estimated. Second, the set of estimated parameters was classified using non-hierarchical cluster analysis, and then the representative value of each cluster was calculated. The representative values that were closest to the known pharmacokinetic values were judged to be the population pharmacokinetic parameters of the group where each parameter was a normal mixture distribution. In order to show the usefulness of the present method, simulation experiments were conducted. The dataset of plasma venlafaxine concentration from 100 subjects was created based on published reports. The 100 subjects were classified into the two genotypes; 50 subjects were CY P2D6 *1/*10, and 50 subjects were CY P2D6 *10/*10. The distributions of the pharmacokinetic parameters were bi-modal. Simulation results were only slightly different from the pharmacokinetic parameter distributions of the dataset, proving the usefulness of the method presented.
