Abstract
BACKGROUND: A novel nomogram based on the Surveillance, Epidemiology, and End Results (SEER) database has been developed to predict the survival of patients with esophageal carcinoma who received neoadjuvant therapy followed by surgery. We aimed to evaluate the accuracy and value of the nomogram with an external validation cohort. METHODS: A total of 2,224 patients in SEER database were divided into the training cohort (n = 1556) and the internal validation cohort (n = 668), while 77 patients in our institute were enrolled in the external validation cohort. A Cox proportional hazards regression model was used to develop a nomogram based on the training cohort, while the C-indexes, the calibration curves, receiver operating characteristics curve (ROC), and Kaplan-Meier survival curve were applied in the internal and external validation cohort. RESULTS: Five independent risk factors were identified and integrated into the nomogram (C-index = 0.645, 95%CI 0.627-0.663). The nomogram exhibited good prognostic value in the internal validation cohort (C-index = 0.648 95%CI 0.622-0.674). However, the C-index, calibration plot, receiver operating characteristics curve (ROC) analysis, Kaplan-Meier survival curve of the nomogram in the external validation cohort were not as good as the training and internal validation cohort (C-index = 0.584 95%CI 0.445-0.723). Further analysis demonstrated that the resection margin involvement (R0, R1, or R2 resection) was an independent risk factor for the patients, which was not included in the SEER cohort. CONCLUSIONS: the nomogram based on the SEER database fails to accurately predict the prognosis of the patients in the external validation cohort, which can be caused by the absence of essential information from the SEER database.