O-6-Methylguanine-DNA methyltransferase (MGMT) is a DNA repair gene. Epigenetic silencing of the MGMT promoter methylation compromises DNA repair and has been associated with longer survival in patients with glioblastoma (GBM) who receive alkylating agents. But the prognostic of MGMT promoter methylation in GBM patients of different race is still ambiguous. Based on an univariate or multivariate analysis between different race (Caucasian and Asian), a meta-analysis of the effects of MGMT promoter methylation on both progression-free survival (PFS) and overall survival (OS) among GBM patients was conducted. A total of 6,309 patients from 50 studies were involved in the analysis. Random effect models were applied to estimate the pooled hazard ratio (HR) with 95 % confidence intervals (CIs) for GBM patients of different race prognosis, the Chi square-based Q test was used to test heterogeneity. Begg's (funnel plot method) and Egger's linear regression tests were adopted to check publication bias (a bias with regard to what is likely to be published, among what is available to be published). The HR value estimated for OS was 0.524 (95 % CI 0.428-0.640) by univariate analysis and 0.427 (95 % CI 0.355-0.513) by multivariate analysis in Caucasian. The HR value estimated for OS was 0.892 (95 % CI 0.469-1.698) by univariate analysis and 0.562 (95 % CI 0.394-0.804) by multivariate analysis in Asian. The HR value estimated for PFS was 0.526 (95 % CI 0.372-0.743) by univariate analysis and 0.437 (95 % CI 0.356-0.537) by multivariate analysis in Caucasian. The HR value estimated for PFS was 0.132 (95 % CI 0.006-3.027) by multivariate analysis in Asian. This data revealed that GBM patients with MGMT promoter methylation had longer OS and PFS by univariate or multivariate analysis in Caucasian regardless of therapeutic intervention. However, GBM patients with MGMT promoter methylation only had longer OS by multivariate analysis in Asian.