Prostate cancer is one of the most common malignancies in men. It can be treated with prostatectomy, a surgical procedure that may result in postoperative sexual dysfunction. To effectively counsel patients and to anticipate potential complications, this work adopts traditional statistical methods and machine learning to predict patients’ sexual satisfaction one year post-surgery. A retrospective, single-center cohort study was conducted involving 409 patients, with a median age of 72 years. This cohort was derived from an initial population of 524 individuals, following the exclusion of cases with incomplete EPIC-26 Q12 data at the 12-month follow-up. We studied medical patients from the University Hospital of Naples “Federico II”, focusing on evaluations before and one year after surgery. Patients’ satisfaction with their sexual function was measured using question 12 from the Expanded Prostate Index Composite (EPIC-26). Statistical results showed a positive link between the International Prostate Symptom Score (IPSS) and sexual dissatisfaction after surgery. Pre-operative higher IPSS scores were associated with lower sexual satisfaction. However, most patients did not experience a significant drop in sexual satisfaction one year after surgery. Scores on the International Index of Erectile Function (IIEF) and EPIC-26-Q12 exhibit, in fact, an interesting inverse relationship, indicating that higher pre-operative scores on one correspond to lower scores on the other. Machine learning techniques on the other hand, allow a prediction accuracy of 81% for post-operative sexual impairment, with the Extra-Trees classifier yielding the best performance. The achieved prediction accuracy of 81% suggests that machine learning models may serve as valuable tools for preoperative risk stratification and for providing personalized patient counseling regarding postoperative sexual function. This study shows how the combination of statistical approaches and machine learning can improve sexual health management after prostate radical surgery, providing insights for doctors and patients, while further validation involving larger and multicenter datasets is necessary, these findings underscore the significant clinical potential of incorporating statistical and machine learning methodologies into the management of postoperative sexual health.
Rotbei et al. (Thu,) studied this question.