Developing a novel model for predicting overall survival in late-onset colon adenocarcinoma patients based on LODDS: a study based on the SEER database and external validation.

Abstract

Aim: To construct a predictive model based on the LODDS stage established for patients with late-onset colon adenocarcinoma to enhance survival stratification.

Methods: Late-onset colon adenocarcinoma data were obtained from the public database. After determining the optimal LODDS truncation value for the training set via X-tile software, we created a new staging system by integrating the T stage and M stage. Nomograms of the prognostic model were created after Cox analyses identified independent risk factors for overall survival (OS) and cause-specific survival (CSS) and were validated internally and externally. The efficacy of the nomograms was assessed by calibration, time-dependent area under the curve (AUC) and decision curve analysis (DCA). Finally, the prognoses of the patients were compared by plotting survival curves on the basis of risk scores.

Results: A total of 103,291 and 100 patients with late-onset colon adenocarcinoma (50-80 years old) were screened from the Surveillance, Epidemiology, and End Results (SEER) and The Cancer Genome Atlas (TCGA) databases, respectively. Cox regression analysis revealed independent risk factors for OS and CSS, including age, gender, race, size, LODDS stage, PLN stage, LNR stage, and TNM stage. A comparison of the four models constructed on the basis of different stages revealed that the model constructed with the LODDS stage had the minimum AIC (Akaike information criterion), maximum C-index (concordance index) and time-dependent AUC. Nomograms based on the LODDS stage were constructed and successfully validated for accuracy and clinical utility.

Conclusion: For patients with late-onset colon adenocarcinoma, LODDS may achieve optimal predictive performance. Furthermore, compared to the 8th edition of the AJCC classification system, the nomogram based on LODDS stage may demonstrate superior survival prediction capabilities.