Prediction of cancer-specific survival after radical cystectomy in pT4a urothelial carcinoma of the bladder: development of a tool for clinical decision-making.

Objective: To externally validate the pT4a-specific risk model for cancer-specific survival (CSS) proposed by May et al. (Urol Oncol 2013; 31: 1141-1147) and to develop a new pT4a-specific nomogram predicting CSS in an international multicentre cohort of patients undergoing radical cystectomy (RC) for urothelial carcinoma of the bladder (UCB) PATIENTS AND METHODS: Data from 856 patients with pT4a UCB treated with RC at 21 centres in Europe and North-America were assessed. The risk model proposed by May et al., which includes female gender, presence of positive lymphovascular invasion (LVI) and lack of adjuvant chemotherapy administration as adverse predictors for CSS, was applied to our cohort. For the purpose of external validation, model discrimination was measured using the receiver-operating characteristic-derived area under the curve. A nomogram for predicting CSS in pT4a UCB after RC was developed after internal validation based on multivariable Cox proportional hazards regression analysis evaluating the impact of clinicopathological variables on CSS. Decision-curve analyses were applied to determine the net benefit derived from the two models.

Results: The estimated 5-year-CSS after RC was 34% in our cohort. The risk model devised by May et al. predicted individual 5-year-CSS with an accuracy of 60.1%. In multivariable Cox proportional hazards regression analysis, female gender (hazard ratio [HR] 1.45), LVI (HR 1.37), lymph node metastases (HR 2.54), positive soft tissue surgical margins (HR 1.39), neoadjuvant (HR 2.24) and lack of adjuvant chemotherapy (HR 1.67, all P < 0.05) were independent predictors of an adverse CSS rate and formed the features of our nomogram with a predictive accuracy of 67.1%. Decision-curve analyses showed higher net benefits for the use of the newly developed nomogram in our cohort over all thresholds.

Conclusions: The risk model devised by May et al. was validated with moderate discrimination and was outperformed by our newly developed pT4a-specific nomogram in the present study population. Our nomogram might be particularly suitable for postoperative patient counselling in the heterogeneous cohort of patients with pT4a UCB.