Nomogram model for predicting cause-specific mortality in patients with Early-Onset Colorectal Cancer: a competing risk analysis Insight from the SEER Database and a Chinese Cohort.

Abstract

Abstract Objective This study aims to analyze the risk factors for Cancer-Specific Mortality (CSM) and Other-Cause Mortality (OCM) in early-onset colorectal cancer (EOCRC) patients,and to construct a nomogram for predicting CSM based on a competitive risk model and validate it using training, internal, and external cohorts. Methods EOCRC patients from the SEER database(2008–2017). Furthermore, EOCRC patients treated at a Northeast China tertiary hospital were included(2014–2020). The SEER data were randomly divided into training and validation sets at a 7:3 ratio. Univariate COX regression model was used to screen for prognostic correlates. Multivariate Cox regression models were then employed to identify independent risk factors. A nomogram visualized results, assessed by C-index,AUC and calibration curves. DCA evaluated clinical utility. Results A total of 8,813 patients were collected from the SEER database, divided into training (N = 6,610) and validation (N = 2,203) sets. 76 patients were included from the Chinese cohort(N = 76). Multivariable Cox regression models revealed that race, tumor differentiation, carcinoembryonic antigen (CEA), marital status, histological type, AJCC stage, and surgical status were independent risk factors for CSM in EOCRC patients. The nomogram constructed based on those independent risk factors had good performance with C-index of 0.806 ,0.801and 0.810 for the training, internal validation and external validation cohorts, respectively.Calibration curves and AUC also indicated the nomogram's accuracy and discriminative ability. Also DCA reflects the good clinical value of the model. Conclusion This study successfully established a competing risk model for CSM in EOCRC patients, demonstrating good predictive value, which may help clinicians to make better treatment decision making.