GENO-17: AN 8-microRNA SIGNATURE PREDICTS RESPONSE TO BEVACIZUMAB IN GLIOBLASTOMA

Following the trials of bevacizumab for the treatment of glioblastoma multiforme, there has been considerable discussion over whether this drug should be recommended as a standard treatment for this disease. Trial data suggests prolongation of progression-free survival, but no effect on overall survival. Despite these results, tails of Kaplan Meier curves and observations on a case-by-case basis suggest bevacizumab improves overall survival in certain patient subsets. We used TCGA microRNA expression data from tumors resected at first diagnosis of glioblastoma in patients treated with bevacizumab and randomly separated these into training (n = 50) and test (n = 37) groups. Using a cross-validated penalized regression model we identified 8 microRNAs as potential predictors of overall survival in the training set. We created a response score using the total expression of these microRNAs, weighted by regression coefficient. This was dichotomized based on the most prognostic minimum of a density plot of the response scores (Log-rank HR = 0.16, p = 1.2e-5). This was then validated in the test cohort (one-sided log-rank HR = 0.34, p = 0.026). Analysis of the signature using all samples in the TCGA, regardless of treatment received, (n = 562) showed that the response prediction was much less powerful (HR = 0.73, p = 1.6e-3) in this cohort suggesting the signature is specific to bevacizumab. The definition of a bevacizumab responding group was validated using expression of microRNAs from RNA-seq data from patients treated with bevacizumab or CCNU at recurrence in the BELOB trial. This showed that ‘responders’ defined by the microRNA signature survived significantly longer than ‘non-responders’ after treatment with bevacizumab (n= 20, HR = 0.23, p = 0.020) but not after treatment with CCNU (n= 22, HR = 0.35, p = 0.052). In conclusion, we have identified an 8-microRNA signature that predicts overall survival in patients with glioblastoma treated with bevacizumab. This may be useful for identifying patients who are likely to benefit from this agent.