A great deal of attention has been given to the problem of estimating the pharmacokinetic parameters of individual patients in order to optimize dosage choices. In the present study, the individual pharmacokinetic parameters of valproic acid were estimated using only 2 data points for each patients by a newly developed program, MULTI 2 (BAYES). This is a nonlinear multiple regression program based on the Bayesian method and written in BASIC for the microcomputer. From the information on the distribution of population pharmacokinetic parameters and plasma drug con centration measurements, MULTI 2 (BAYES) calculates the most likely set of para meters for the individual. The accuracy and precision of estimates were evaluated based on the difference between estimated and measured drug concentration. Selec tion of the population parameter set was found to affect the individual pharmacokine tic parameters estimated by MULTI 2 (BAYES). Using the population parameter set obtained from the data of patients in Kyoto University Hospital by NONMEM gave the most accurate estimates of individual patients by MULTI 2 (BAYES). Thus, use of a population parameter appropriate to the individual in question is necessary to obtain a good estimate of the individual parameter. This study provides significant evidence for the validity and usefulness of the Bayesian method in esti mating individual pharmacokinetic parameters from routine clinical data.
