Predicting incontinence and erectile function after prostate cancer surgery: International validation of models

Our objective is to externally validate the most accurate, published tools predicting urinary incontinence and erectile dysfunction following prostatectomy. Several models have been developed to predict the risks of adverse events, though most have not been externally validated.

Data were obtained from the Prostate Cancer Outcomes Registry of Australia and New Zealand (PCOR-ANZ). Self-reported urinary incontinence and erectile dysfunction were measured using EPIC-26 at 12 months after radical prostatectomy. Four predictive models were selected for external validation, being the top performing models from a systematic literature review. Two models related to urinary incontinence (Matsushita and Jeong) and two related to sexual function (Alemozaffar and Novara), were examined. Model discrimination was assessed by the Area Under the Received Operator Curve (AUC) and calibration was assessed.

We constructed a cohort of 590 patients resident in either New Zealand or South Australia who had received a radical prostatectomy 2007-2019. The average age at diagnosis was 65 years, with most men having few comorbidities (97.1 % Charlson comorbidity index 0) and treated with robotic surgery (93.6 %). In our external validation cohort, the Almozaffar model demonstrated the highest discrimination when predicting erectile dysfunction (AUC 0.73, 95%CI 0.67-0.78). The highest discrimination achieved by a model predicting urinary incontinence was developed by Jeong (AUC 0.69, 95%CI 0.61-0.76).

Models predicting erectile dysfunction performed well in external validation and may be suitable for clinical use. Models predicting post-prostatectomy urinary incontinence did not perform as well on validation.