Postoperative nomogram for invasive bladder cancer: Does it really work? A multicenter cohort study

OBJECTIVES: Nomograms are statistical models designed to maximize predictive accuracy. We have tested the statistical correlation between the predictions of International Bladder Cancer Nomogram Consortium and the clinical outcomes in a multicenter Italian cohort of patients treated with radical cystectomy (RC) and pelvic lymph-nodes dissection. METHODS AND MATERIALS: Two hundred four patients who underwent RC were selected for multiple variable and then enrolled in the study. Patients were tested by the "online tool" based on the nomogram, then stratified and risk grouped for 5-year predicted disease-free survival (pDFS): low risk (67%-100%), intermediate risk (34%-66%), and high risk group (0%-33%). Receiving operating characteristic curve (ROC) quantified the area under curve (AUC) as predictive accuracy. Actual overall survival (aOS) and actual disease-free survival (aDFS) were calculated with Kaplan-Meyer analysis. Median of pDFS was compared with 5-year aDFS. RESULTS: AUC was 0.69 (P < 0.001). The aOS is 50% (95% confidence interval (95% CI, -7.68/+8.23) at 5-years after RC, the aDFS is 65.5% (95% CI, -7.56/+8.98). The 5-year aDFS is 75.3% (-8.82/+12.53) in low risk group; 67.3% (-12/+16.4) in intermediate risk group; 28.3% (-20/+17.2) in high risk group. The 5-year aDFS was well calculated by the nomogram but in all groups the nomogram slightly underestimated the prediction. In intermediate risk group, 5-year aDFS overscored both the nomogram pDFS and the pDFS range of this group (34%-66%). In all patients investigated, the pDFS was 65%, a value close to the 5-year aDFS found (65.6%). CONCLUSION: Statistical correlation between postoperative nomogram prediction and the clinical reality was observed