Establishment and validation of a clinicopathological prognosis model of gastroenteropancreatic neuroendocrine carcinomas

BACKGROUND: Gastroenteropancreatic neuroendocrine carcinomas (GEP-NECs) are a rare, highly malignant subset of gastroenteropancreatic neuroendocrine neoplasms (GEP-NENs). However, how to predict the prognosis of GEP-NECs by clinical features is still under study. This study aims to establish and validate a nomogram model of overall survival (OS) in patients with GEP-NECs for predicting their prognosis. METHODS: We selected patients diagnosed with GEP-NECs from the Surveillance, Epidemiology, and End Results (SEER) database and two Chinese hospitals. After randomization, we divided the data in the SEER database into the train cohort and the test cohort at a ratio of 7:3 and used the Chinese cohort as the validation cohort. The Cox univariate and multivariate analyses were performed to incorporate statistically significant variables into the nomogram model. We then established a nomogram and validated it by concordance index (C-index), calibration curve, receiver operating characteristic (ROC) curve, the area under the curve (AUC), and the decision curve analysis (DCA) curve. RESULTS: We calculated the nomogram C-index as 0.797 with a 95% confidence interval (95% CI) of 0.783–0.815 in the train cohort, 0.816 (95% CI: 0.794–0.833) in the test cohort and 0.801 (95% CI: 0.784–0.827) in the validation cohort. Then, we plotted the calibration curves and ROC curves, and AUCs were obtained to verify the specificity and sensitivity of the model, with 1-, 3- and 5-year AUCs of 0.776, 0.768, and 0.770, respectively, in the train cohort; 0.794, 0.808, and 0.799 in the test cohort; 0.922, 0.925, and 0.947 in the validation cohort. The calibration curve and DCA curves also indicated that this nomogram model had good clinical benefits. CONCLUSIONS: We established the OS nomogram model of GEP-NEC patients, including variables of age, race, sex, tumor site, tumor grade, and TNM stage. This model has good fitting, high sensitivity and specificity, and good clinical benefits.