RESUMEN
OBJECTIVE: This paper describes a developed pharmacokinetic model for the estimation of valproic acid (VPA) clearance (CL) calculated from routine clinical data taken from Egyptian epileptic patients. METHODS: Retrospective clinical data from 81 adult and paediatric epileptic patients with one trough VPA serum concentration per patient were analysed using NONMEM to estimate drug CL and determine the influence of different covariates. A qualification group of 20 epileptic children (3-13 years old) was used to evaluate the final model. RESULTS: The population CL as estimated by base model (no covariates) was 0.581 l h(-1) with inter-individual variability (C.V. %) of 17.4% and SD of residual error was 6.82 mg l(-1). Univariate selection and backward deletion of different covariates led to the development of the final regression model of CL as follows: CL(Lh-1) = 0.101 + 0.151 * CBZ + 0.000248 * VPADD + 0.0968 * age/20 + 0.0803 * INDI, in which CBZ indicates co-administration of carbamazepine, VPADD the daily dose of VPA and INDI uncontrolled epilepsy. The between-subject variability in CL was 23.6% while the standard deviation of the residual error was 5.24 mg l(-1). The model predictions in the qualification group were found to have no bias and satisfactory precision. CONCLUSION: The population pharmacokinetic model for VPA could be used for a priori recommendation and dose optimisation of that drug in the Egyptian population of epileptic patients.