Abstract
BACKGROUND: The presence of competing risks suggests that the classic Cox proportional hazards model may yield biased results when assessing prognostic factors for uveal melanoma (UM) patients. OBJECTIVE: The aim of this research is to utilize a competing risk model using the Surveillance, Epidemiology, and End Results (SEER) database in order to discover predictive factors for individuals with UM and contrast them with the conventional Cox proportional hazards model. METHODS: We collected information on individuals who were diagnosed with UM and registered in the SEER database from 2010 to 2015. The univariate analysis involved the application of the cumulative incidence function (CIF) and Gray's test, while a multivariate analysis was conducted using the Fine-Gray, cause-specific (CS) and Cox proportional hazards models. RESULTS: Among the 1,712 eligible patients diagnosed with UM, 631 individuals passed away: 400 due to UM and 231 from other causes. One-way Gray's test indicated that six variables significantly influenced the survival prognosis of UM patients (P < 0.05). The multivariate competing risk model indicated that age, race, histologic type, AJCC stage and surgery were all independent predictors in both the Fine-Gray model and the CS model, with differences only at the level of precise estimation. In contrast, the Cox model showed several deviations in the statistical significance of prognostic factors and the estimation of hazard ratios. CONCLUSIONS: This research developed competing risk models to evaluate prognostic factors for UM, demonstrating higher accuracy compared to conventional methods that do not account for competing risk factors.