Abstract: PD34-06
Sources of Funding: none

Introduction

A prediction model that accurately predicts risk of stricture recurrence after anterior urethroplasty is lacking. We hypothesized that such a model would be dependent on both patient and stricture characteristics and could predict recurrence with high sensitivity and specificity.

Methods

We created models based on clinical outcomes of consecutive men undergoing anterior urethroplasty by one of 6 surgeons in the Trauma and Urologic Reconstruction Network of Surgeons (TURNS) from 2010 to 2015. The outcome variable of interest was stricture recurrence, defined as the need for a secondary procedure. All available pre-operative and operative variables, including interaction variables when deemed appropriate, were initial candidates and the final model was chosen by stepwise selection. Receiver operating characteristic (ROC) curves were created and the area under the curve (AUC) was reviewed. Colinearity with stricture length and location variables dictated that separate models for standard excisional (EPA) and substitutional (flap and/or graft; SUB) repairs be created. Surgeon effects were accounted for through an exchangeable structure of the working correlation matrix.

Results

There were 547 EPA and 706 SUB repairs used for model creation, of which recurrence was noted in 20 (3.7%) and 67 (9.5%) respectively. AUC was marginally higher for the SUB model (0.7777) versus the EPA model (0.7601). Significant variables in the SUB model included number of prior DVIUs (OR 1.14; 95% CI 1.07-1.21), stricture etiologies of prior TURP (OR 3.77; 2.07-6.89), ureteroscopy (OR 4.07; 1.66-10.02), infection (OR 5.77; 2.57-12.05) and/or lichen sclerosus (OR 3.33; 1.06-10.42) and ventral (OR 3.53; 1.55-8.03) or sandwich (OR 3.70; 2.23-6.11) graft placement. Stricture length was not an independent predictor (OR 1.03; 0.99-1.07) and smoking was protective (OR 0.58; 0.36-0.95). In the EPA model, only stricture length (OR 1.31; 1.12-1.53) and pre-operative urine residual (OR 1.00; 1.00-1.00) were significant variables.

Conclusions

Traditional pre- and intraoperative variables used to create these prediction models led to AUCs well below 0.8, indicating only moderate clinical usefulness. The low recurrence rate affected robust EPA model creation, though longer EPA repairs were more likely to fail as predicted. The SUB model was heavily dependent on stricture etiology and location of graft placement. Improved collection of pathologic, morphologic and surgical characteristics appears necessary if improvement in the prediction capabilities of these models is desired.

Funding

none