A shared frailty model for assessing time to seizure remission in adults with epilepsy

Abstract

Abstract Epilepsy remains a significant global health concern with increasing prevalence and incidence. This study aimed to model the time to first remission among epilepsy patients at Jimma University Medical Center, Ethiopia, using parametric shared frailty models. A retrospective study was conducted on epilepsy patients treated between 1st January 2018 and 30th December 2023. All patients received anti-seizure medications (ASMs) upon enrollment. Additionally, 12% of the cohort had received ASM treatment prior to enrollment outside JUMC. Log-logistic, log-normal, and Weibull baseline hazard functions were combined with gamma and inverse Gaussian frailty distributions to model time to first remission. Model selection was based on the Akaike Information Criterion (AIC). The median time for patients to achieve their first seizure remission was 38 months, with 45.5% (95% CI: 40.9–50.1%) of patients experiencing remission. The variability in remission times across different districts, as modelled by the log-normal-inverse Gaussian shared frailty model, was estimated to be θ = 0.454. Patients aged 25–44 years (acceleration factor 1.13 [95% CI 1.04–1.23], p = 0.005], those with more than five pre-treatment seizures (acceleration factor 1.08 [95% CI 1.02–1.15, p = 0.018]), and individuals with focal epilepsy (acceleration factor 1.15 [95% CI 1.07–1.25, p = 0.003]) were associated with significantly longer remission times compared to other patient groups while those with good treatment adherence (acceleration factor 0.88 [95% CI 0.81–0.96, p = 0.005]) were associated with significantly shorter remission times compared to poor treatment adherence. The log-normal-inverse Gaussian shared frailty model offers valuable insights into the variability of remission patterns among patients. Specifically, individuals aged 25–44 years, those with a history of more than five pre-treatment seizures, and patients with focal epilepsy experienced significantly longer remission times. In contrast, patients who adhered well to their treatment regimens achieved remission more quickly than other groups.