Development and validation of prognostic nomograms for early-onset colon cancer in different tumor locations: a population-based study

Objective: The prevalence of early-onset colon cancer (EOCC) among individuals below the age of 50 has shown a marked upward trend in recent years. The embryology, clinical symptoms, incidence, molecular pathways, and oncologic outcomes differ between right-sided and left-sided colon cancers. However, the differences have not been fully researched in EOCC. Our study aims to develop and validate prognostic nomograms predicting overall survival (OS) for EOCC in different tumor locations based on the Surveillance, Epidemiology, and End Results (SEER) database. Methods: Using the SEER database, a total of 5,588 patients with EOCC were extracted and divided into development and validation cohorts in a random allocation ratio of 7:3 across three groups. Univariate and multivariate Cox regression analyses were performed to identify independent prognostic factors influencing OS outcomes. These factors were then utilized to construct nomogram models. The prognostic capabilities of the three models were assessed through various evaluation metrics, including the concordance index (C-index), receiver operating characteristic (ROC) curves, calibration curves, decision curve analysis (DCA), and validation cohorts respectively. Additionally, survival curves of the low- and high-risk groups were calculated using the Kaplan–Meier method together with the log-rank test. Results: Significant differences in clinical features were observed between right-sided and left-sided EOCCs, particularly in terms of OS (52 months vs 54 months) as demonstrated by Kaplan-Meier curves. Transverse-sided EOCCs exhibited clinical characteristics similar to right-sided EOCCs, suggesting a potential shared tumor microenvironment and therapeutic considerations. Advanced stage, liver metastasis, poor grade, pretreatment carcinoembryonic antigen (CEA) level, chemotherapy, and perineural invasion were identified as independent prognostic factors across all three tumor locations and were incorporated into the nomogram model. The C-index for the right-sided EOCC model was 0.820 in the development cohort and 0.818 in the validation cohort. Similarly, the left-sided EOCC model achieved a C-index of 0.788 in the development cohort and 0.787 in the validation cohort, while the transverse-sided EOCC model obtained a C-index of 0.845 in the training set and 0.835 in the validation set. ROC curves with calculated area under the curve (AUC) values exceeded 0.8 for all three groups in both the development and validation cohorts, indicating robust predictive performance. Calibration plots for 3-year and 5-year OS in both cohorts demonstrated excellent agreement between predicted outcomes and actual clinical observations. Furthermore, decision curve analysis (DCA) plots revealed a threshold probability range of 0.1 to 0.9, within which the nomogram model exhibited maximum benefit. Kaplan-Meier curves exhibited significant differences between the low- and high-risk groups in EOCC for all three tumor locations, further validating the prognostic value of the nomogram model. Conclusions: We successfully developed three precise nomogram models for EOCCs in different tumor locations, providing valuable support for clinicians in guiding clinical treatments and facilitating further prospective follow-up studies.