A nonlinear mixed-effect modeling (NONMEM) program was used to evaluate the effects of cytochrome P450 (CYP) 2C9 and CYP2C19 polymorphisms on the phenobarbital (PB) population clearance for Japanese epileptics. The pharmacokinetics of the 260 PB concentrations at a steady-state obtained from 79 patients was described with a one-compartment open pharmacokinetic model with first-order elimination. The covariates screened included the total body weight (BW), age, gender, PB daily dose, CYP2C9 and CYP2C19 genotypes, the coadministered antiepileptic drugs (AEDs), and complications. The final model of PB apparent clearance was as follows: CL = 0.23 x (BW/40)0.21 x 0.52CYP2C9*1/*3 x 0.68VPA x 0.85PHT x 0.85SMID x (1 + etaCL) where CL = the clearance of PB; CYP2C9*1/*3 = 1, otherwise 0; VPA = 1 if valproic acid is coadministered, otherwise 0; PHT = 1 if phenytoin is coadministered, otherwise 0; SMID = 1 if complications of severe or profound mental retardation with a significant behavior impairment are presented, otherwise 0; and etaCL = the independent random error distributed normally with the mean zero and variance equal to omegaP2. The total clearance of PB decreased by 48% in patients with CYP2C9*1/*3 genotype in comparison with those with CYP2C9*1/*1 genotype (P < 0.001). An effect of CYP2C19 polymorphisms was not detected. To our knowledge, this is the first report to demonstrate that the CYP2C9 genotype affects the PB metabolism in routine care, but the results should be further verified in other ethnic populations.