Radiation-induced erectile dysfunction in prostate cancer: A systematic review of pathophysiology, clinical radiobiology, and predictive modeling.

BACKGROUND: Radiation-induced erectile dysfunction (RIED) remains one of the most prevalent and distressing late toxicities among prostate cancer survivors. Despite advances in radiation therapy techniques, its underlying mechanisms, dose-response relationships, and predictive modeling strategies are not yet fully understood. METHODS: We searched PubMed/MEDLINE, Web of Science, Scopus, Embase and Cochrane Library from inception to November 10, 2025. Eligible English-language studies included adult prostate cancer patients treated with radiotherapy and reported evidence on erectile outcomes, pathophysiology, radiobiology, dose-response relationships or predictive modeling. Data were extracted using a standardized form, quality was assessed with the Newcastle-Ottawa Scale, and findings were synthesized narratively according to PRISMA 2020. RESULTS: Thirty-nine peer-reviewed studies were included in the primary synthesis. RIED was associated with interacting neurovascular, smooth-muscle and endocrine mechanisms. Age, baseline erectile function, comorbidities, androgen deprivation therapy and radiation dose to erectile-relevant structures were key determinants of risk. Dose-response findings, particularly for the penile bulb, were inconsistent across studies because of heterogeneous contouring, endpoints, treatment techniques and confounder adjustment. No externally validated RIED-specific NTCP model was identified. Machine-learning, deep-learning and radiogenomic approaches remain exploratory and are limited by small or heterogeneous cohorts and insufficient external validation. CONCLUSION: RIED is a multifactorial toxicity for which robust radiobiological parameters and clinically validated prediction tools are still lacking. Standardized contouring, validated patient-reported outcomes and adequately powered multi-institutional studies integrating clinical, dosimetric and biological variables are needed to refine dose constraints and improve individualized risk prediction.