Predicting the probability of death within five years after surgery in patients with colorectal cancer

Abstract

Background: This research aims to investigate the factors associated with mortality in colorectal cancer (CRC) patients within five years after surgery and to develop nomograms to predict the probability of death within five years after surgery. Methods: Patients diagnosed with CRC eligible for this retrospective cohort study between 2010 and 2015 were collected from the Surveillance, Epidemiology, and End Results (SEER) database. They were randomly divided intothe training set and the validation set. Random forest method and multivariate logistic regression analysis were used to identify factors associated with death within five years (survival time ≤ 60 months). Nomograms were developed to estimate the probability of cancer-specific death (CSD) and overall death (OD) in 5 years. The concordance index (C-Index), the receiver operating characteristic (ROC) curve, and the calibration plot were utilized to evaluate the nomograms' discrimination and calibration capabilities. Decision curve analysis (DCA) was used to assess the clinical value of the nomograms. Results: A total of 26459 eligible CRC patients were enrolled in this study. Multivariate logistic regression analysis showed that age, sex, histology, T stage, N stage, M stage, carcinoembryonic antigen (CEA), radiotherapy, and chemotherapy were associated with death in CRC patients. The C-index of the nomogram for predicting CDS was 0.824 (0.818-0.829), and the C-index for predicting OD was 0.807 (0.801-0.814). The ROC curve demonstrated that the nomograms had a high capacity for discrimination. The calibration curve revealed that the predicted curve was exceptionally near the reference line, showing that the nomogram's calibration capacity was excellent. DCA demonstrated a net clinical advantage over TNM staging. Conclusion: The nomograms can be used to predict the probability of death within five years after surgery for stage I-IV CRC. It is helpful for clinicians to identify high-risk populations, evaluate the prognosis of CRC patients, and customize targeted treatment plans for patients.