Prognostic score predicting survival after resection of pancreatic neuroendocrine tumors: analysis of 3851 patients

BACKGROUND: Pancreatic neuroendocrine tumors (PNET) have a poorly defined natural history, and a staging system is not available. The objective of this study was to identify factors predicting survival after pancreatectomy for PNETs and to establish a postresection prognostic score. PATIENTS AND METHODS: From the National Cancer Data Base (1985-2004), patients were identified who underwent PNET resection. Multivariable Cox proportional hazards modeling was used to assess the impact of patient, tumor, treatment, and hospital factors on survival. A prognostic score based on the predictive factors from the Cox model was developed. RESULTS: Three thousand eight hundred fifty-one patients underwent resection for PNETs. Five-year overall survival was 59.3%, and the 10-year survival was 37.7%. On multivariable analysis, age, grade, distant metastases, tumor functionality, and type of resection were independent predictors of survival after resection of PNETs (P < 0.0001). Gender, race, socioeconomic status, tumor size, nodal status, margins, adjuvant chemotherapy, and hospital volume were not associated with survival. Age, grade, and distant metastases were the most significant predictors of survival and were incorporated into a PNET postresection prognostic score. The prognostic score correlated with outcomes and offered excellent survival discrimination by each of the 3 score subgroups: 76.7%, 50.9%, and 35.7% (P < 0.0001). The concordance index was 0.63 (95% CI 0.59-0.67), indicating reasonable agreement between actual outcomes and that predicted by the prognostic score. CONCLUSIONS: The prognostic score can be used to predict outcomes, guide adjuvant treatment, and stratify patients for clinical trials.