The methylation of O(6)-methylguanine DNA methyltransferase (MGMT) gene promoter is a key biological marker in clinical neuro-oncology. Nevertheless, there is no consensus concerning the best technique for its assessment. In a recent study comparing five methods to analyze MGMT status, we found that the best prediction of survival was obtained with a pyrosequencing (PSQ) test assessing methylation of 5 CpGs (CpGs 74-78). In the present study we extended our PSQ analysis to 16 CpGs (CpGs 74-89) identified as critical for transcriptional control of the gene. The predictive value of the methylation levels at each CpG, as well as the mean methylation levels of selected sets of consecutive CpGs was tested in a cohort of 89 de novo glioblastoma patients who had received standard of care treatment (Stupp protocol). Using an optimal risk cut-off, each CpG or combination of CpGs, was associated with overall survival (OS) and progression free survival. The best predictive models for OS after stratification on performance score and age were obtained with CpG 89, CpG 84 and mean methylation of CpG 84-88 (Hazard ratio (HR), 0.31; p < 0.0001). The improvement compared to the predictive value of the test analyzing average methylation of CpG 74-78 (HR, 0.32; p < 0.0001) was however marginal. We recommend to test CpGs 74-78 when analyzing MGMT methylation status by PSQ because a commercial kit that has successfully been used in several studies is available, allowing reproducible and comparable results from one laboratory to another.