Purpose
To create a preoperative prediction model for estimating the risk of non-organ-confined (pT3-4 or pN+) bladder urothelial cancer (NOC–BUC) in patients with clinically OC–BUC (cT1-2N0M0).
Methods
The study involved 248 consecutive patients who had undergone radical surgery for clinically OC–BUC at a tertiary cancer center between 2003 and 2011. Logistic regression analysis was used to develop a prediction model for estimating the risk of pathological NOC disease. Prespecified predictors included age, gender, recurrent frequency, tumor size and number, hydronephrosis, and pathological characteristics at transurethral resection (T-stage, tumor grade, lymphovascular invasion (LVI), and carcinoma in situ). Discrimination ability was measured by the area under the receiver operating characteristic curve (AUC).
Results
Overall, 39.1 % of the patients with clinically OC–BUC had NOC disease at the time of radical surgery. In multivariate analysis, recurrent frequency, tumor size, hydronephrosis, and three pathological features at transurethral resection (T-stage, tumor grade, and LVI) were significantly associated with disease extent. The final prediction model included seven variables after backward elimination and achieved a bootstrap-corrected AUC of 0.79. Internal validation showed good calibration and clinical usefulness of the nomogram.
Conclusions
Based on readily available clinicopathological parameters, we developed a nomogram for predicting NOC tumor in clinically OC–BUC. Despite reasonable performance in internal validation, the prediction model should be assessed in external dataset before applied in clinical setting.