Objective To establish a population pharmacokinetic model of sodium valproate in patients with epilepsy in China, so as to promote individualized drug administration. Methods The plasma concentrations and clinical data were collected from 69 patients with epilepsy after receiving sodium valproate. Population pharmacokinetic model (n=101, showed by proportional residual model) was established using Phoenixs NLME software, and the effects of covariate (gender, age, body weight, drug combination and so on) on absorption rate constant (Ka), distribution volume (Vd) and clearance (CL) were estimated. The final model was validated and evaluated by Bootstraps (n=500). Results The final population pharmacokinetic model was:TvKa=2.38, V=TvV, CL=TvCL·exp(η_CL), and the robust rate was 100% as displayed by the Bootstrap. The average value and median of parameters obtained by Bootstrap were consistent with the model parameters, indicating that the model was relatively stable. Age, gender, body weight, drug combination and other variables were not introduced into the final model. Conclusion The population pharmacokinetic model established in this study is consistent with the basic model, providing reference for rational individualized drug administration in clinical practice.