BI-22 * CORRELATION OF MOLECULAR SUBTYPES WITH OVERALL SURVIVAL (OS) IN AVAGLIO, A RANDOMIZED, PLACEBO-CONTROLLED STUDY OF BEVACIZUMAB (BEV) PLUS RADIOTHERAPY (RT) AND TEMOZOLOMIDE (TMZ) FOR NEWLY DIAGNOSED GLIOBLASTOMA (GBM)

Abstract

BACKGROUND: Factors that correlate the extent of clinical benefit with anti-VEGF therapy are poorly understood. Two phase 3 trials in newly diagnosed GBM (AVAglio and RTOG-0825) reported that BEV + RT/TMZ prolonged progression-free survival but not OS, versus placebo + RT/TMZ. Specific GBM patient subgroups may derive OS benefit from first-line BEV; tumor profiling has uncovered intrinsic molecular subtypes. BEV efficacy in molecular subtypes was evaluated as an exploratory objective in AVAglio. METHODS: A total of 349 cases from AVAglio (BEV + RT/TMZ, n = 171; placebo + RT/TMZ, n = 178), for which formalin-fixed, paraffin-embedded tissue samples were available for biomarker analysis, were profiled and classified into known GBM molecular subtypes (pre-defined hypothesis, subtypes defined by Verhaak, Cancer Cell 2010 and Phillips, Cancer Cell 2006) using an 800-gene platform. GBM molecular subtypes were correlated with OS. RESULTS: Subtyping by expression of 30 markers described by Phillips (2006) are presented. As observed in recent data from The Cancer Genome Atlas (Brennan, Cell 2013), patients with Proneural isocitrate dehydrogenase 1 (IDH1) wild-type tumors had the worst prognosis among all GBM subtypes. BEV + RT/TMZ conferred a significantly longer OS for patients with Proneural IDH1 wild-type tumors, compared with placebo + RT/TMZ (17.1 vs 12.2 months, HR = 0.42, 95% CI: 0.24-0.72, p = 0.002). Multivariate analysis (accounting for prognostic covariates) revealed an interaction between the Proneural subtype and BEV (p = 0.012). In patients with Mesenchymal or Proliferative tumors, there was no evidence of a difference in OS between the treatment arms. CONCLUSIONS: Candidate molecular predictors were identified in AVAglio; GBM patients with Proneural IDH1 wild-type tumors may derive OS benefit from BEV; these data could impact patient stratification and therapy and give insights into BEV mode of action. The predictive value of the Proneural subtype should be validated in an independent dataset.