Background: O6-methylguanine DNA-methyl transferase (MGMT) status is predictive for alkylating chemotherapy in most series, but there are non-benefitting subgroups. Despite multiple attempts, MGMT has not been unambiguously established as a predictive biomarker for patients with malignant gliomas. Further, these tumors are to be better classified according to global methylation profiles. Methods: Long-term efficacy data of the NOA-08 trial (NCT01502241) that compared efficacy and safety of radiotherapy (RT, n= 176) to temozolomide (TMZ, n= 193) in patients > 65 years with anaplastic astrocytoma (AA) or GB as well as genome-wide DNA methylation patterns and copy number variations assessed by methylation arrays in a biomarker subset (n= 104) and an independent cohort (n= 380) have been used to assess the interaction between MGMT status and methylation subgroups. Results: In the long-term update of NOA-08 patients with MGMT methylated tumors had longer OS and EFS when treated with TMZ (18.4 [13.9-24.4] months and 8.5 [6.9-13.3] months) versus RT (9.6 [6.4-13.7] months and 4.8 [4.3-6.2] months, HR 0.44 [0.27-0.70], p < 0.001 for OS and 0.46 [0.29-0.73], p = 0.001 for EFS). These data compared favorably with recently published data from patients treated with chemoradiation (Perry et al. NEJM 2017). Importantly, only patients with glioblastomas of the methylation class receptor tyrosine kinase II (RTKII) and mesenchymal but not RTK I demonstrated the predictive impact of MGMT in the NOA and the independent validation cohort. Conclusions:MGMT promoter methylation as a strong but methylation subclass-dependent predictive biomarker for the use of alkylating chemotherapy in malignant gliomas. The data call for embedding of MGMT tests into global methylation analyses for all patients with malignant gliomas potentially treated with alkylating chemotherapy.