Abstract
BACKGROUND Transitional cell bladder carcinoma (tcBC) is the predominant form of bladder cancer, making up around 95% of reported cases. Prognostic factors for older individuals with tcBC differ from those affecting younger patients. The main purpose of this study was to establish a prognostic competing risk model for elderly patients with tcBC. MATERIAL AND METHODS We conducted a retrospective analysis using data from the SEER database, randomly assigning patients to training and validation groups. We applied proportional subdistribution hazard (SH) to assess risk factors for cancer-related mortality (CSM). A competitive risk model was created to predict cancer-specific survival in elderly patients with tcBC. Model validation involved evaluating the area under the receiver operating curve, the consistency index, and a calibration curve. The Kaplan-Meier (K-M) curve was then used to compare mortality risk between high-risk and low-risk groups identified by the model. RESULTS This study randomly assigned 61 293 patients from the SEER database into training (42 905 patients) and validation (18 388 patients) groups in a 7: 3 ratio. Using a proportional subdistribution hazards model, we identified prognostic risk factors such as age, race, sex, marital status, TNM staging, grade, and metastatic status in brain, bone, liver, and lung. We developed a competitive risk model to predict 5-year cancer-specific survival (CSS) in elderly tcBC patients, achieving consistency index (C-index) values of 0.814 and 0.815 for the training and validation groups, respectively. Kaplan-Meier (K-M) analysis revealed 5-year survival probabilities of 35.1% (high-risk) and 42.2% (low-risk) in the training group, with similar rates of 35.7% and 42.0% in the validation group, both showing statistically significant differences (log-rank P<0.01). CONCLUSIONS We successfully established a competitive risk model for forecasting cancer-specific survival in elderly tcBC patients, primarily relying on these identified risk factors. The validation outcomes indicate the model's accuracy and dependability, making it a highly efficient predictive instrument. This tool enables making personalized clinical decisions for both medical professionals and patients.