A revised RTOG recursive partitioning analysis (RPA) model for glioblastoma based upon multiplatform biomarker profiles.

Abstract

2001 Background: The Radiation Therapy Oncology Group (RTOG) recursive partitioning analysis (RPA) model, which relies on clinical variables, has been used worldwide to establish distinct prognostic classes of patients (pts) with malignant glioma as well as eligibility criteria for clinical trials. In the present study, we have updated the RPA to include additional molecular variables, specifically for glioblastoma (GBM) patients treated in the temozolomide (TMZ)-era, to make the model more relevant, contemporary, and discriminatory. Methods: The dataset utilized was from RTOG 0525, a phase III study examining radiation (RT) with concurrent TMZ, followed by adjuvant standard dose vs. dose-dense TMZ in pts with newly-diagnosed GBM. 162 pts from RTOG 0525 had available tissues for profiling of key signaling molecules using the AQUA platform. Results: pAKT, c-met, and MGMT protein were each found to be significantly associated with adverse outcome on multivariate analysis. These variables were combined with clinical and genetic biomarkers (e.g., MGMT promoter methylation, IDH1 mutation, mRNA profiling) previously found to be of significance (MCP model, ASCO, 2011) to generate an even more robust, discriminatory RTOG RPA model. The explained variation for these three classification models was found to be 41.7 (Current RPA), 19 (MCP), and 14.9% (Clinical RPA), respectively, with higher values indicating better separation of prognostic groups (see table). Conclusions: The current RTOG RPA classification model, based upon incorporation of multi-platform biomarker analysis, holds promise for RT+TMZ-treated GBM patients; further validation of this model is planned. Financial Support: NCI grants U10 CA21661, U10 CA37422, U24 CA114734, 1RC2CA148190, 1RC2CA148190, RO1CA108633, and BTFC grant. [Table: see text]