Comparative Predictive Performance of Serum and Dose-Normalized Valproate Levels for Seizure Control in Adults With Epilepsy

Background Valproate is a widely used antiepileptic drug, but its clinical response shows considerable interindividual variability. Conventional therapeutic ranges may not reliably predict seizure control, and the role of dose-normalized valproate levels in forecasting seizure as an outcome remains unclear. Objective To evaluate the predictive performance of serum valproate levels for seizure control and compare it with dose-normalized valproate levels, and also to assess whether multivariable models incorporating demographic factors improve the prediction of seizures in adult patients with epilepsy. Methods This study was a secondary analysis of prospectively collected data from adult patients with epilepsy receiving valproate monotherapy at a tertiary care centre. Demographic and clinical data were recorded, and patients were followed for one month to assess seizure occurrence. Serum valproate levels and dose-normalized levels were analysed at the end of the follow-up period. Logistic regression models were constructed, and predictive performance was assessed using ROC analysis. Results A total of 137 patients were included, of whom 30 (21.9%) experienced seizures during follow-up. Serum valproate levels demonstrated good predictive performance for seizure control (AUC: 0.891, 95% CI: 0.814-0.969), with an optimal cutoff of 56.1 µg/mL (sensitivity 86.7%, specificity 86.9%). Dose-normalized levels showed lower predictive performance (AUC: 0.829, 95% CI: 0.734-0.923). Logistic regression analysis confirmed serum valproate level as a significant predictor of seizure control, while dose-normalized levels showed weaker associations. Model-based predictions demonstrated comparable performance but did not significantly outperform serum levels. Conclusion Serum valproate levels provide a clinically meaningful prediction of seizure control and outperform dose-normalized metrics. These findings may support the use of concentration-based monitoring to guide individualized therapy in epilepsy.