Predictive Impact of Circulating Vascular Endothelial Growth Factor in Four Phase III Trials Evaluating Bevacizumab

We evaluated the prognostic and predictive use of circulating VEGF-A levels in phase III trials of bevacizumab in colorectal cancer, lung cancer, and renal cell carcinoma. Baseline plasma samples from 1,816 patients were analyzed for VEGF-A using an ELISA, which recognizes the major isoforms with equivalent sensitivity. HR and 95% confidence intervals (CI) for study end points were estimated using Cox regression analysis. A subset of matched archival tumor samples was analyzed for VEGF-A expression using in situ hybridization. Higher VEGF-A levels showed trends toward adverse prognostic significance in the control arms of multiple trials, reaching statistical significance for overall survival (OS) in AVF2107 (highest vs. lowest 50%: HR = 1.76; 95% CI, 1.28-2.41), AVAiL (HR = 1.52; 95% CI, 1.16-2.00), and AVOREN (HR = 1.67; 95% CI, 1.18-2.36). In predictive analyses, the HRs for progression-free survival were similar across low and high VEGF-A subgroups and favored bevacizumab-containing treatment. In the low VEGF-A subgroups, HRs (95% CIs) were 0.61 (0.43-0.87) in AVF2107, 0.71 (0.43-1.16) in E4599, 0.74 (0.59-0.94) in AVAiL (low-dose), 0.89 (0.70-1.13) in AVAiL (high-dose), and 0.56 (0.40-0.78) in AVOREN. Analyses of OS data have shown similar results. No correlation between primary tumor VEGF-A expression and plasma VEGF-A levels was observed. In this comprehensive evaluation, pretreatment total circulating VEGF-A was prognostic for outcome in metastatic colorectal, lung, and renal cell cancers, but it was not predictive for bevacizumab-based treatment benefit.