Construction and Validation of a Nomogram Model for Predicting Cancer-Specific Survival in Colorectal Cancer Based on...

Colorectal cancer (CRC) remains one of the most prevalent malignancies worldwide, and accurate prognostic prediction is essential for individualized prognostic assessment and patient risk communication. The conventional TNM staging system has limited capacity for personalized survival estimation. This study aimed to construct and validate a registry-based nomogram model for predicting 3-year and 5-year cancer-specific survival (CSS) in CRC patients using data from the Surveillance, Epidemiology, and End Results (SEER) database.

A retrospective cohort study was conducted using data from the SEER database (2010-2015), an interval selected to ensure consistent AJCC 7th edition staging and adequate follow-up for 5-year CSS assessment. A total of 45,218 CRC patients who met the inclusion criteria were randomly divided into a training cohort (n = 31,653) and a validation cohort (n = 13,565) at a 7:3 ratio. Univariate and multivariate Cox proportional hazards regression analyses were performed to identify independent prognostic factors for CSS. A nomogram was constructed based on these factors. Model performance was assessed by concordance index (C-index), time-dependent receiver operating characteristic (ROC) curves, calibration curves, and decision curve analysis (DCA), with DCA interpreted as an evaluation of prognostic risk-classification net benefit rather than as evidence for treatment selection.

Eight independent prognostic factors were identified: age at diagnosis, race, marital status, tumor grade, T stage, N stage, M stage, and carcinoembryonic antigen (CEA) level. The C-index of the nomogram was 0.783 (95% CI: 0.776-0.790) in the training cohort and 0.771 (95% CI: 0.761-0.781) in the validation cohort, both higher than the AJCC TNM staging system (training: 0.724; validation: 0.716). The absolute C-index improvements over TNM staging were 0.059 and 0.055 in the training and validation cohorts, respectively. The area under the ROC curve (AUC) for 3-year and 5-year CSS prediction was 0.812 and 0.794 in the training cohort, respectively, and 0.798 and 0.781 in the validation cohort. Calibration curves demonstrated close agreement between predicted and observed survival probabilities. DCA indicated that the nomogram provided higher net benefit for prognostic risk classification than the traditional TNM staging system across a wide range of threshold probabilities.

The nomogram model incorporating clinicopathological and demographic variables showed higher predictive accuracy than conventional TNM staging for CSS in CRC patients. Because the incremental gains over TNM staging were moderate and the model was developed from registry data without detailed chemotherapy information or molecular biomarkers, including MSI status, this tool should be interpreted only as an adjunct for individualized prognostic assessment and risk communication. It should not be used to determine treatment intensity or to replace guideline-based therapeutic decision-making, and it requires external validation in contemporary cohorts.