External Validation of Population Pharmacokinetic Models of Lamotrigine in Patients with Epilepsy or Postneurosurgery

Background: This study aimed to evaluate the predictive performance of published lamotrigine (LTG) population pharmacokinetic (PPK) models using an external data set of Chinese patients with epilepsy or postneurosurgery. Methods: In total, 348 concentration measurements from 94 Chinese children and 254 Chinese adults with epilepsy or postneurosurgery were used for external validation. Data on published LTG PPK models were obtained from the literature. The predictability of the models was assessed using prediction-based diagnostics (eg, F20 and F30), simulation-based diagnostics, and Bayesian forecasting. Results: The results of prediction-based diagnostics for all 10 models were unsatisfactory. The best-performing models, characterized as one-compartment models with nonlinear pharmacokinetics, incorporated weight as a key covariate and included interindividual variability for both clearance and volume of distribution. These models achieved exceptional predictive performance in simulation-based diagnostics and Bayesian forecasting, with IF 30 values of 90.32%, 97.23%, and 99.61%, respectively, demonstrating superior precision and accuracy. Bayesian forecasting improved the predictive accuracy of 80% of the models, significantly enhancing model predictability. Conclusions: The published PPK models show extensive variation in predictive performance for extrapolation among Chinese patients with epilepsy or postneurosurgery. The lack of key covariates (such as concomitant medications, genetic polymorphisms, and age stratification) and fixed parameters of volume of distribution and absorption rate constant in the PPK modeling of LTG may explain its unsatisfactory predictive performance. Bayesian forecasting significantly improves the model predictability and may help individualize LTG dosing.