Nomogram based on the log odds of negative lymph node/T stage can predict the prognosis of patients with colorectal cancer: a retrospective study based on SEER database and external validation in China.

Abstract

This study investigated the prognostic role of log odds of negative lymph node/T stage (LONT) and established a nomogram based on LONT to predict the prognosis in colorectal cancer (CRC) patients. A retrospective cohort study. We enrolled 80 518 CRC patients from the Surveillance, Epidemiology and End Results database between 2010 and 2015. The dataset was split into a training cohort (56 364 patients) and a validation cohort (24 154 patients) at a ratio of 7:3. Furthermore, 500 CRC patients who underwent surgery in the Tenth Affiliated Hospital of Southern Medical University between 1 January 2017 and 20 December 2018, were recruited as the external validation set. 1-, 3- and 5-year cancer-specific survival (CSS). The univariate and multivariate Cox regression analyses were carried out to identify the significant independent prognostic factors of CSS. A nomogram was established based on LONT to predict the prognosis. The performance of the nomogram was comprehensively assessed via the time-dependent receiver operating characteristic curve, concordance index (C-index), calibration curve and decision curve analysis (DCA) comprehensively. Moreover, Kaplan-Meier curves were performed to assess the CSS of the three risk subgroups. LONT was a significant independent prognostic factor for CSS (LONT1 vs LONT2, HR=0.670, 95% CI 0.642 to 0.698, p<0.001; LONT1 vs LONT3, HR=0.443, 95% CI 0.420 to 0.467, p<0.001). LONT, age, sex, race, subsite, differentiation, histology, tumour size, T stage, N stage, M stage and chemotherapy were included in the nomogram. The 1-, 3- and 5-year survival area under the curve were 0.856, 0.862 and 0.852, respectively. The C-index of the model was 0.809 (95% CI 0.825 to 0.839) in the model. The calibration curve and DCA verified the favourable predictive performance and clinical application of the nomogram. CRC patients with a high LONT had a low incidence of CSS. The nomogram based on LONT could effectively predict the CSS of CRC.