Development and validation of a survival prediction model for 113,239 patients with colon cancer: a retrospective...

Colon cancer (CC) is the third most commonly diagnosed malignant tumor and remains the second leading cause of cancer-related deaths worldwide. However, the risk assessment of poor prognosis of CC is limited in previous studies. This study aimed to develop a predictive nomogram for the survival of CC patients.

In this retrospective cohort study, 113,239 CC patients from the Surveillance, Epidemiology, and End Results (SEER) database were randomly divided into training (n=56,619) and testing (n=56,620) sets with a ratio of 1:1. Demographic, clinical data and survival status of patients were extracted. The outcomes were 3- and 5-year survival of CC. Univariate and multivariate Cox regression analyses were used to screen the predictors to develop the predictive nomogram. Internal validation and stratified analyses were further assessed the nomogram. The C-index and area under the curve (AUC) were calculated to estimate the model's predictive capacity, and calibration curves were adopted to estimate the model fit.

Totally 38,522 (34.02%) patients died during the 5-year follow-up. The nomogram incorporated variables associated with the prognosis of CC patients, including age, gender, marital status, insurance status, tumor grade, stage (T/N/M), surgery, and number of nodes examined, with a C-index of 0.775 in the training set and 0.774 in the testing set. The AUCs of the nomogram for the 3- and 5-year survival prediction in the training set were 0.817 and 0.808, with the sensitivity of 0.688 and 0.716, and the specificity of 0.785 and 0.740, respectively. Similar results were found in the testing set. The C-index of the predictive nomogram for male, female, White, Black, and other races was 0.769, 0.779, 0.773, 0.770, and 0.770, respectively. The calibration curves for the nomogram in the above five cohorts showed a good agreement between actual and predicted values.

The nomogram may exhibit a certain predictive performance based on the SEER database, which may provide individual survival predictions for CC patients.