Prognostic and predictive blood-based biomarkers in patients with advanced pancreatic cancer: Results from CALGB 80303

Abstract

BACKGROUND: CALGB 80303 was a phase III trial of 602 patients (pts) with locally advanced or metastatic pancreatic cancer (PC) comparing gemcitabine + bevacizumab (GB) vs. gemcitabine + placebo (GP). That study found no overall survival (OS) benefit from the addition of B. Blood samples were collected for prospective biomarker analyses. METHODS: Plasma samples were analyzed via a novel multiplex ELISA platform for >40 candidate factors related to tumor growth, angiogenesis, and inflammation. This platform was optimized for use in cancer patients. Baseline values were correlated with OS using univariate Cox proportional hazard regression models and multivariate Cox models with leave-one-out cross validation (LOOCV). Potential predictive markers were identified using a treatment by marker interaction term in the Cox model. RESULTS: 328 pts had baseline samples available. Univariate prognostic markers predicting OS identified within the GB and GP cohorts included: Ang2, CRP, ICAM-1, IGFBP-1, TSP-2 (all p<0.001). By multivariate analysis, these markers were significantly more prognostic than standard clinical factors, such as age, gender, extent of disease and performance status. LOOCV modeling yielded consistent multivariate prognostic signatures that differentiated patients with longer vs. shorter survival. The signature for the GB group consisted of IGFBP-1, IL-6, PDGF-AA, PDGF-BB, TSP-2; while the signature for the GP group consisted of CRP, IGFBP-1, PAI-1, PDGF-AA, PEDF (both p<0.0001). Three potential predictive markers of superior OS for GB vs. GP were identified: VEGF-D (p<0.01), SDF-1b (p<0.05), and Ang2 (p<0.05). Spearman's rank correlation coefficients suggested that many factors were co-regulated. CONCLUSIONS: This is one of the first phase III studies to report multiplex angiome profiling and it shows that tumor angiogenesis factors are highly prognostic in patients with PC. Several of these factors may define those pts more or less likely to benefit from the addition of B to G. This information should allow better stratification of patients with PC and may guide novel therapeutic interventions in PC and perhaps other cancers.