Preoperative predictive model of early urinary continence recovery after laparoscopic radical prostatectomy.

PURPOSE: To develop and validate a predictive model include magnetic resonance imaging (MRI) parameters preoperatively which can assess the risk of incontinence after laparoscopic radical prostatectomy (LRP) accurately.

METHODS

We retrospectively reviewed and included 170 patients with prostate cancer who underwent LRP between July 2015 and June 2018 in our institution. All 170 patients were randomly resampled and divided into training set (n = 124) and verification set (n = 46) according to the ratio of 7:3. The Nomogram prediction model of the risk of incontinence after LRP was established through the training set and verified by the verification set. Baseline patient characteristics were obtained, including age, body mass index, and prostate volume. Perioperative characteristics such as pre-biopsy prostate specific antigen, biopsy Gleason score, clinical staging, and NVB sparing status were also collected. MRI parameters preoperatively including membranous urethral length (MUL), prostate apex depth ratio (PADR), and intravesical prostatic protrusion length (IPPL) were obtained. The C index and visual inspection of calibration curve were used to evaluate the discrimination and calibration of the model.

RESULTS

According to the urinary incontinence (UI) at 3 months postoperatively, the patients were divided into 104 cases (61.2%) in the group with no incontinence and 66 patients (38.8%) in the group with incontinence. Multivariate logistic regression analysis of training set showed that cT3a (OR = 0.427, 95% CI 0.142-1.281, P = 0.1288), MUL (OR = 0.237, 95% CI 0.102-0.551, P < 0.01), PADR (OR = 0.276, 95% CI 0.116-0.655, P < 0.01), and IPPL (OR = 0.073, 95% CI 0.030-0.179, P < 0.01) were independent predictors of urinary incontinence at 3 months postoperatively. The model showed good discrimination with an area under the receiver operating characteristic (ROC) curve of 0.880, with the sensitivity and specificity 0.800 and 0.816, respectively, and good calibration (Hosmer-Lemeshow test result of 5.57, P = 0.695). Decision curve analysis demonstrated that the model was clinically useful.

CONCLUSION

This study developed and validated a preoperative model in the form of a nomogram to predict the risk of UI after LRP at 3 months. MUL, PADR, and IPPL were significant independent predictive factors of the postoperative early urinary continence.