Predicting the indication for adjuvant radiation therapy according to EAU guidelines among patients with high-risk prostate cancer: a novel multivariable model

Abstract

Background To develop a novel model for preoperatively predicting the indication for adjuvant radiation therapy after radical prostatectomy according to current guideline recommendations of the European Association of Urology (EAU) based on patient and clinical tumor characteristics in high-risk prostate cancer patients. Methods Within a high-volume center database (01/2010-08/2024), we identified high-risk prostate cancer patients. Univariable logistic regression models addressed indication for adjuvant radiation therapy. Multivariable logistic regression models included the most informative, statistically significant preoperative predictors. Harrell's concordance index (c-index) quantified accuracy after 2000 bootstrap resamples for internal validation. Results Of 5691 patients, 2137 (38%) had indication for adjuvant radiation therapy according to current EAU guidelines (2025). Indication for adjuvant radiation therapy was associated with higher prostate volume (> 45 cm3 and 25-45 cm3) and advanced tumor characteristics, namely higher prostate-specific antigen value (>20 ng/ml and 10-20 ng/ml), advanced clinical tumor stage (cT3/4 and cT2), lower number of sampled biopsy cores (<= 12), higher proportion of positive cores (continuous), and higher Gleason Grade Group in biopsy (5, 4, and 3). No association was observed for age and body-mass index and indication for adjuvant radiation therapy. Multivariable model c-index for the prediction of the indication for adjuvant radiation therapy was 0.761 (95% confidence interval 0.749-0.776). Conclusions Clinical tumor characteristics can be used for preoperatively predicting the indication for adjuvant radiation therapy after radical prostatectomy according to current EAU guideline recommendations in high-risk prostate cancer patients. Prior to clinical application, the present multivariable model should be externally validated within an independent cohort.