Purpose: To estimate the burden of lifetime epilepsy (LTE) and active epilepsy (AE) and examine the influence of study characteristics on prevalence estimates. Methods: We searched online databases and identified articles using prespecified criteria. Random-effects meta-analyses were used to estimate the median prevalence in developed countries and in urban and rural settings in developing countries. The impact of study characteristics on prevalence estimates was determined using meta-regression models. Results: The median LTE prevalence for developed countries was 5.8 per 1,000 (5th-95th percentile range 2.7-12.4) compared to 15.4 per 1,000 (4.8-49.6) for rural and 10.3 (2.8-37.7) for urban studies in developing countries. The median prevalence of AE was 4.9 per 1,000 (2.3-10.3) for developed countries and 12.7 per 1,000 (3.5-45.5) and 5.9 (3.4-10.2) in rural and urban studies in developing countries. The estimates of burden for LTE and AE in developed countries were 6.8 million (5th-95th percentile range 3.2-14.7) and 5.7 million (2.7-12.2), respectively. In developing countries these were 45 (14-145) million LTE and 17 (10-133) million AE in rural areas and 17 (5-61) million LTE and 10 (5-17) million AE in urban areas. Studies involving all ages or only adults showed higher estimates than pediatric studies. Higher prevalence estimates were also associated with rural location and small study size. Conclusions: This study estimates the global burden of epilepsy and the proportions with AE, which may benefit from treatment. There are systematic differences in reported prevalence estimates, which are only partially explained by study characteristics.