To predict cancer-specific survival (CSS), we developed a novel nomogram model and a risk classification system for classifying risk levels of metastatic colorectal cancer (mCRC) patients.

The data was extracted from the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) database between 2010 to 2015. All eligible cases were randomly divided into training and validating cohorts. Cox proportional hazards model was used to explore the independent risk factors for CSS. A novel nomogram model was developed to predict the CSS and evaluated via internal and external validations.

Using the multivariate Cox proportional hazards model, the independent risk factors were identified for CSS. Then a novel nomogram was developed for CSS based on such factors. The concordance indexes (C-index) were 0.718 (95%CI: 0.712~0.725) for this nomogram and 0.722 (95%CI: 0.711~0.732) for the validating cohort, indicating a good discrimination. The calibration plots and decision curve analysis (DCA) showed good consistency and nice potential clinical validity. A risk classification system divided all patients into three groups and Kaplan-Meier curves indicated good stratification and discrimination for CSS among different groups. In the total cohorts, the median CSS of patients in the low-risk, intermediate-risk, and high-risk groups was 36 months (95%CI: 34.987~37.013), 18 months (95%CI: 17.273~18.727), and 5 months (95%CI: 4.503~5.497), respectively.

We developed a novel nomogram model to predict the CSS for synchronous mCRC patients. Furthermore, a risk classification system could contribute to accurately assessing the prognosis and guiding treatment.