Epigenetic silencing of the O(6)-methylguanine-DNA methyltransferase (MGMT) gene is associated with improved survival in patients with high-grade gliomas (HGG), with varying estimates of magnitude. The objective of this meta-analysis is to determine the prognostic value of MGMT silencing, and assess its predictive value by treatment type. MEDLINE and EMBASE databases were searched for studies relating to gliomas and MGMT. Studies reporting overall survival (OS) by MGMT status in patients with HGG were considered potentially eligible. We excluded studies that did not control for potential confounding variables. A meta-analysis of studies was performed via random-effects modelling. Subgroup meta-analyses by treatment were performed according to a priori hypotheses. Twenty studies were ultimately eligible, including 2,018 patients. In the pooled analysis, MGMT silencing was associated with improved OS (HR = 0.436; 95% CI: 0.333-0.571; P < 0.001). The prognostic utility of MGMT status varies significantly by treatment type (P = 0.001): the HR for OS for MGMT silenced tumors is 0.190 (0.047-0.770), 0.403 (0.282-0.576), 0.743 (0.579-0.954), and 1.070 (0.722-1.585) for studies using surgery plus the addition of either: chemotherapy (CT), chemoradiotherapy (CRT), radiotherapy (RT), and nothing (surgery alone), respectively. Epigenetic silencing of MGMT is associated with markedly improved survival in patients with HGG who receive adjuvant therapy. MGMT silencing serves as a predictive marker, with the largest benefit seen in patients receiving CT as a component of adjuvant treatment, an intermediate benefit in patients receiving adjuvant RT, and no evidence to support benefit in those receiving surgery alone.