We investigated the influence of the population pharmacokinetic parameters on the performance of the Bayesian feedback method in predicting the phenytoin (PHT) dosage needed to achieve a desired PHT serum concentration. Population pharmacokinetic parameters studied were taken from reports by Sheiner et al. (I), Grasela et al. (II), Miller et al. (III), and the authors (IV). The predictive abilities of the Bayesian feedback method were evaluated by using retrospective data from 70 patients. The mean prediction error, mean absolute prediction error (MAE), and root mean squared error (RMSE) served as measures of prediction bias and precision. The precision of the initial estimates based on the population parameter set IV was superior to those of other initial estimates studied. The performance of the Bayesian feedback method to predict PHT dosage at the steady state was relatively insensitive to bias in the estimates of the population parameters. However, the revised estimates derived from the Bayesian feedback method using the population parameter set IV with one feedback gave the lowest MAE and RMSE in predicting PHT dosage.