We previously examined the methods for evaluating the population pharmacokinetics for Michaelis-Menten elimination, and estimated the population pharmacokinetic parameters of phenytoin in Japanese patients with epilepsy (Biol. Pharm. Bull., 19, 444, 1996). In the present study, we evaluated the predictive performance of the bayesian analysis based on this prior information. To estimate the individual pharmacokinetic parameters of phenytoin in 50 patients, 3 serum phenytoin concentration data at a stady-state after repetitive dosing were collected retrospectively for each subject. The bayesian analysis using from 1 to 2 feedback concentrations gave the accurate predictions for another dosage regimens. The bayesian regression-analysis method described here in thus appears to be useful for predicting the steady-state serum phenytoin concentration from the limited number of data obtained for therapeutic drug monitoring purposes.