The CABARET trial (ACTRN12610000915055) reported no difference in overall survival (OS) between patients with recurrent glioblastoma (GBM) randomized to either bevacizumab monotherapy or bevacizumab plus carboplatin. However, a subset of patients showed durable responses and prolonged survival, with recorded survival times of over 30 months in five of 122 patients (4%). Patient selection for bevacizumab therapy would be enhanced if a predictive biomarker of response or survival could be identified; this biomarker sub-study attempted to identify novel biomarkers. Patients who opted to participate in this sub-study and who had adequate biospecimens for analysis (n = 54) were retrospectively evaluated for the expression of a series of tumor proteins. Immunohistochemistry (IHC) was used to measure the expression of 19 proteins previously implicated in cancer treatment response to bevacizumab. MGMT promoter methylation was also assessed. Tumor DNA from five patients with outlying survival duration (‘poor’ and ‘exceptional’ survivors) was subjected to whole genome sequencing (WGS). No single protein expression level, including VEGF-A, predicted OS in the cohort. WGS of poor and exceptional survivors identified a gain in Chromosome 19 that was exclusive to the exceptional survivors. Validation of this finding requires examination of a larger independent cohort.