Individualised prediction of seizure recurrence and long-term outcome after antiepileptic drug withdrawal

Abstract

Objective: When a child with epilepsy has been seizure free for some time, the question is whether antiepileptic drug (AED) treatment is still necessary. Stopping AEDs carries the risk of seizure recurrence, and although many publications studied predictors of seizure outcome it is difficult to apply this knowledge to the individual child. We created an individualised prediction model for seizure recurrence and long-term seizure outcome, through an Individual Participant Data meta-analysis. Methods: Systematic review of literature identified all candidate predictors and eligible publications, of which authors were contacted to provide individual participant data. Through regression analysis the strongest predictors were selected. Internal-external cross-validation was performed to ensure generalizability. Ultimately, nomograms were created to visually represent computed prediction models. Results: Ten cohorts with 1771 patients were gathered, with both children and adults, of whom 812 (46%) experienced seizure recurrence and 9% had seizures in the last year of follow-up (median 5.3 years, range 0–23 years). 1474 patients (83%) had childhood-onset epilepsy, 988 (56%) started AED withdrawal before the age of 18. Prediction models for seizure recurrence (c-statistic: 0.65; 95%CI 0.65–0.66) and chance of long-term seizure freedom (c-statistic: 0.71; 95%CI 0.70–0.71) were created with good calibration and stable validation across all ten populations. Conclusion: It is now possible to compute combined risks of seizure recurrence and the chance of long-term seizure freedom after AED withdrawal for an individual patient. The nomograms are based on a large cohort, validated in both children and adults, and will aid consultation of seizure free children with epilepsy. The nomograms may therefore guide the physician as well as the patient in person-tailored choices.