Abstract
PURPOSE: CALGB 90206 was a phase III trial of 732 patients with metastatic renal cell carcinoma (mRCC) comparing bevacizumab plus IFNα (BEV + IFN) with IFNα alone (IFN). No difference in overall survival (OS) was observed. Baseline samples were analyzed to identify predictive biomarkers for survival benefit. PATIENTS AND METHODS: A total of 32 biomarkers were assessed in 498 consenting patients randomly assigned into training (n = 279) and testing (n = 219) sets. The proportional hazards model was used to test for treatment arm and biomarker interactions of OS. The estimated coefficients from the training set were used to compute a risk score for each patient and to classify patients by risk in the testing set. The resulting model was assessed for predictive accuracy using the time-dependent area under the ROC curve (tAUROC). RESULTS: A statistically significant three-way interaction between IL6, hepatocyte growth factor (HGF), and bevacizumab treatment was observed in the training set and confirmed in the testing set (P < 0.0001). The model based on IL6, HGF, and bevacizumab treatment was predictive of OS (P < 0.001), with the high- and low-risk groups having a median OS of 10.2 [95% confidence interval (CI), 8.0-13.8] and 34.3 (95% CI, 28.5-40.5) months, respectively. The average tAUROC for the final model of OS based on 100 randomly split testing sets was 0.78 (first, third quartiles = 0.77, 0.79). CONCLUSIONS: IL6 and HGF are potential predictive biomarkers of OS benefit from BEV + IFN in patients with mRCC. The model based on key biological and clinical factors demonstrated predictive efficacy for OS. These markers warrant further validation in future anti-VEGF and immunotherapy in mRCC trials. See related commentaries by Mishkin and Kohn, p. 2722 and George and Bertagnolli, p. 2725.