Construction and external validation of a nomogram for predicting 10-year recurrence risk after radical resection in Chinese patients with stage III colorectal cancer.

Background: Patients with stage III colorectal cancer (CRC) face a high risk of postoperative recurrence, which significantly impacts long-term survival. We aimed to develop a prediction model for 5- and 10-year recurrence risks based on the Surveillance, Epidemiology, and End Results (SEER) database and validate it in an independent institutional cohort. Methods: Using the SEER Research Plus database (2000-2012, Chinese population subset), we identified 1,114 eligible patients with stage III CRC who underwent radical resection as the training cohort. Demographic, pathological, and therapeutic variables were collected. Independent prognostic factors were identified through univariate and multivariate Cox proportional hazards regression, and a nomogram was constructed to predict 5- and 10-year recurrence probabilities. Model performance was assessed using receiver operating characteristic (ROC) curves, calibration plots, and decision curve analysis (DCA). External validation was conducted in a cohort of 92 patients with stage III CRC who underwent radical resection at The Ninth Hospital of Hangzhou between 2012 and 2014, with 10-year follow-up data confirmed via medical records and follow-up visits. Results: The 10-year cumulative recurrence rate in the SEER cohort was 32.6%. Multivariate analysis identified age ≥80 years [hazard ratio (HR) =3.89, P<0.001], unfavorable histology (HR =1.92, P=0.03), T3-T4 stage (HR =2.67 and 5.03, P<0.05), and N2 stage (HR =1.82, P<0.001) as independent risk factors. A nomogram incorporating six variables [age, histological type, T stage, N stage, carcinoembryonic antigen (CEA) status, and tumor size] yielded area under the curve (AUC) of 0.74 and 0.73 for predicting 5- and 10-year recurrence in the training cohort, and 0.77 for both timepoints in the validation cohort, demonstrating good discrimination and robustness. Calibration plots indicated excellent agreement between predicted and observed outcomes, while DCA showed superior net clinical benefit. The cumulative 10-year recurrence risk curves showed consistent trends between the SEER and validation cohorts, with subgroup analyses confirming significantly increased risks in elderly patients, CEA-positive cases, mucinous adenocarcinoma (MAC), T3-T4, and N2 subgroups. Conclusions: We developed and externally validated a nomogram based on SEER data to predict recurrence risk after radical resection in stage III CRC. The model demonstrated favorable discrimination, calibration, and clinical utility. This tool may facilitate postoperative risk stratification, individualized follow-up, and adjuvant treatment decision-making for stage III CRC patients. Further validation in larger prospective, multicenter cohorts is warranted.