Background & Aims: Current methods of preoperative staging and predicting outcome following pancreatectomy for pancreatic cancer (PC) are inadequate. We evaluated the utility of multiple biomarkers from distinct biologic pathways as potential predictive markers of response to pancreatectomy and patient survival. Methods: We assessed the relationship of candidate biomarkers known, or suspected, to be aberrantly expressed in PC, with disease-specific survival and response to therapy in a cohort of 601 patients. Results: Of the 17 candidate biomarkers examined, only elevated expression of S100A2 was an independent predictor of survival in both the training (n = 162) and validation sets (n = 439; hazard ratio [HR], 2.19; 95% confidence interval [CI]: 1.48-3.25; P < .0001) when assessed in a multivariate model with clinical variables. Patients with high S100A2 expressing tumors had no survival benefit with pancreatectomy compared with those with locally advanced disease, whereas those without high S100A2 expression had a survival advantage of 10.6 months (19.4 vs 8.8 months, respectively) and a HR of 3.23 (95% CI: 2.39-4.33; P < .0001). Of significance, patients with S100A2-negative tumors had a significant survival benefit from pancreatectomy even in the presence of involved surgical margins (median, 15.7 months; P = .0007) or lymph node metastases (median, 17.4 months; P = .0002). Conclusions: S100A2 expression is a good predictor of response to pancreatectomy for PC and suggests that high S100A2 expression may be a marker of a metastatic phenotype. Prospective measurement of S100A2 expression in diagnostic biopsy samples has potential clinical utility as a predictive marker of response to pancreatectomy and other therapies that target locoregional disease.