Biom-19. Gene Expression Biomarker Prediction of Benefit From Re-Irradiation in Recurrent Glioblastoma

INTRODUCTION Radiotherapy improves overall survival (OS) in patients with newly diagnosed glioblastoma (ndGBM). Re-irradiation (reRT) improves progression-free survival (PFS) when added to bevacizumab in recurrent GBM (rGBM), but an OS advantage in unselected populations has not been demonstrated. Here we test the hypothesis that gene expression profiling can predict OS benefit from reRT in rGBM. METHODS A retrospective cohort of 119 IDH-wildtype, CNS WHO grade 4 GBM tissue samples (n=61 ndGBM, n=58 rGBM) from 87 patients who underwent reRT at the time of any tumor recurrence was analyzed using a barcode-based RNA hybridization platform with a custom 291-gene panel representing pathways underlying GBM growth and therapeutic response. Univariate analyses identified 4 sets of 12-71 genes associated with OS. Ridge regression models (A-D) were trained using gene expression data to predict OS after reRT normalized to OS from diagnosis. Model outputs were mapped to reRT scores, which were dichotomized (high versus low) using maximally selected rank statistics. RESULTS Median time between RT courses was 19.9mo. Median OS after reRT was 10.2mo. Using gene expression from ndGBM, median OS after reRT for reRT score high versus low tumors was 11.9 versus 4.4mo (model A, p<0.0001), 68.2 versus 9.0mo (model B, p=0.0222), 11.7 versus 4.9mo (model C, p=0.0003), and 11.7 vs. 4.9mo (model D, p=0.0002). Using gene expression from rGBM, median OS after reRT for reRT score high versus low tumors was 15.0 versus 8.0mo (model A, p=0.0110), 16.7 versus 7.7mo (model B, p=0.0053), 16.8 versus 7.8mo (model C, p=0.0019), and 20.6 versus 10.2mo (model D, p=0.0242). ReRT score concordance across models was 0.939 (CI 0.887-0.97, p<0.001) and 0.97 (CI 0.949-0.983, p<0.001) for ndGBM versus rGBM. CONCLUSIONS Gene expression data from ndGBM or rGBM can stratify OS after reRT. The models reported here require validation in additional cohorts to determine predictive value.