Validation of population pharmacokinetic parameters of phenytoin using the parallel Michaelis-Menten and first-order elimination model

This study was conducted to assess whether the parallel Michaelis-Menten and first-order elimination (MM+FO) model fitted the data better than the Miohaelis-Menten (MM) model, and to validate the MM+FO model and its parameter estimates. The models were fitted to 853 steady state dose : serum concentration pairs obtained in 332 adults with epilepsy using nonlinear mixed-effects modeling (NONMEM). The MM+FO model fitted the data better than the MM model. The validity of the pharmacokinetic models and the estimated population parameter values was tested using the naive prediction method. The estimation and validation of the pharmacokinetic parameters were undertaken in two separate patient groups (cross-validation) obtained by splitting the data set. Patients were randomly allocated to two equally matched groups (groups 1 and 2). The predictive performance was assessed using 770 paired predicted versus actual dose or measured serum concentrations. The population pharmacokinetic parameters estimated by NONMEM in group 1 were validated in group 2 and vice versa. When predicting steady state set-um concentration, the MM+FO model was clearly superior to the MM model (mean bias of 0.91 and 8.13 mg/L, respectively).