Purpose: In this study we identified homogeneous risk groups, with no survival overlap among the subgroups that make up each risk group, in patients with transitional cell carcinoma of the bladder treated with radical cystectomy alone.
Materials and methods: Predictive factors for tumor death were analyzed with univariate and multivariate analysis among a group of 298 patients with transitional cell carcinoma of the bladder treated with radical cystectomy alone. Independent variables were progressively incorporated according to their statistical power in a stepwise process identifying a model with independent subgroups. The risk groups were identified according to different survival cutoff points including subgroups with similar survival. To search a clinical application and to check the strength of this model a new model was also set up using the weight score based on the size of hazard ratio from multivariate analysis.
Results: Univariate analysis demonstrated that lymphatic invasion status, pathological stage (P), lymph node status (N) and prostatic stroma status (St) were predictive variables for tumor death, and the latter 3 were independent variables in the multivariate analysis. By taking the most powerful, N, as the reference variable, and progressively incorporating additional variables, a model was found including 7 independent subgroups. In this model only 2 subgroups, N1 and N2-3, included more than 1 category and their survival was also calculated. Three risk groups were identified establishing different survival cutoffs. The 5-year cancer specific survival rate was 86.4% for low risk (P1-2N0St-), 64.4% (range 60.9% to 65.3%) for intermediate risk (P1-2N1St-, P3N0St-, HR = 2.7) and 28.1% (range 0% to 47.7%) for high risk (N2-3, P4, St+, N1P3, HR = 8.7). This model was also reproduced using the weight score based on the size of the hazard ratio from the multivariate analysis
Conclusions: Three homogeneous risk groups were identified with high statistically significant survival differences among them and no survival overlap among subgroups that make up the risk groups.