For glioblastoma (GBM), survival classification has primarily relied on clinical criteria, exemplified by the Radiation Therapy Oncology Group (RTOG) recursive partitioning analysis (RPA). We sought to improve tumor classification by combining tumor biomarkers with the clinical RPA data. To accomplish this, we first developed 4 molecular biomarkers derived from gene expression profiling, a glioma CpG island methylator phenotype, a novel MGMT promoter methylation assay, and IDH1 mutations. A molecular predictor (MP) model was created with these 4 biomarkers on a training set of 220 retrospectively collected archival GBMtumors. ThisMPwas further combined with RPA classification to develop a molecular-clinical predictor (MCP). The median survivals for the combined, 4-class MCP were 65 months, 31 months, 13 months, and 9 months, which was significantly improved when compared with the RPA alone. The MCP was then applied to 725 samples from the RTOG-0525 cohort, showing median survival for each risk group of NR, 26 months, 16 months, and 11 months. The MCP was significantly improved over the RPA at outcome prediction in the RTOG 0525 cohort with a 33%increase in explained variation with respect to survival, validating the result obtained in the training set. To illustrate the benefit of the MCP for patient stratification, we examined progression-free survival (PFS) for patients receiving standard-dose temozolomide (SD-TMZ) vs. dose-dense TMZ (DD-TMZ) in RPA and MCP risk groups. A significant difference between DD-TMZ and SD-TMZ was observed in the poorest surviving MCP risk group with a median PFS of 6 months vs. 3 months (p ¼ 0.048, log-rank test). This difference was not seen using the RPA classification alone. In summary, we have developed a combined molecular-clinical predictor that appears to improve outcome prediction when compared with clinical variables alone. This MCP may serve to better identify patients requiring intensive treatments beyond the standard of care.