Meta-analysis of time to seizure relapse after a post-operative epilepsy surgery.
Epilepsy is a common disease world-wide whose suggested treatment vary from surgical to non-surgical. Although surgical treatments may be commonly recommended and successful in pharmacoresistant epilepsy for patients with refractory epilepsy, there are some patients who experience a seizure relapse. It is of interest to medical practitioners to determine how long patients survive epilepsy after a successful surgery. Survival analysis methods have been used to model time-to-event data. Hence we attempt to determine the time to a seizure relapse in epilepsy patients after a surgery. However, single study have some limitations, such as lack of accommodation of spatial factors, different research approaches to mention a few. Most of the time, single studies are under-powered to detect the factors of covariates. Meta-analysis methods have been developed to overcome this problem, where a number of studies are amalgamated and a common conclusion is drawn. This thesis aims is to determine the long-term seizure outcome after an epilepsy surgery of refractory epilepsy without focusing on the types of refractory but rather in resective surgery. In the current study a systematic review was done using Google Scholar, Medline, and PubMed. The event of interest is seizure relapse and our interest is to pool the time to first seizure relapse after surgical treatment. To measure the seizure freedom the clinical method call Engel class I was used. The univariate and metasurvival of fixed and random effect model were used to measure the proportion of seizure freedom. Our focus was only in single arm treatment (surgical treatment) . There were a total of 18 studies that satisfy the inclusion criteria with observations at 6 time points measured in months after post-operative (6, 12, 24, 36, 60 and 120 months). In the univariate analysis, the probabilities of seizure freedom of the fixed effects models were systematically larger than the random effect results. There was evidence of significance of heterogeneity between studies, and the true variation between studies test was large. The result that we got in univariate random effect model were for time-points 6, 12, 24, 36, 60 and 120 months were 0.74 95% confidence interval (CI)(0.66- 0.82), 0.69 95% CI (0.61- 0.77), 0.64 95% CI (0.56- 0.71), 0.60 95% CI (0.52- 0.68), 0.56 95% CI (0.48- 0.63) and 0.47 95% CI (0.38- 0.56) respectively. The meta-survival analysis also systematically showed that, the seizure free probability were larger in a fixed effects model than in a random effects model. The summary survival estimates of the random effect model that were pooled in the following time points 6, 12, 24, 36, 60 and 120 were 0.7655 95% CI (0.6808- 0.8613), 0.7140 95% CI (0.6246- 0.8163), 0.6462 95% (0.5614, 0.7438), 0.6105 95% (0.5225, 0.7133), 0.5700 95% (0.4892, 0.6641) and 0.4755 (0.4078, 0.5545) respectively. The median time to relapse was found using the meta-survival analysis in random effects model to be 104.46 months (8.87 years). We can conclude that the meta-survival analysis may be the method to pool the time-to-event data in one-arm treatment..