Predicting Patient-Reported Outcomes Following Treatment for Localized Prostate Cancer: Model Development and Evaluation in an International Cohort Study

BACKGROUND AND OBJECTIVE: Shared decision-making for localized prostate cancer (PCa) requires understanding how baseline factors predict patient-reported outcomes (PROs) after treatment. Most prior predictive models are limited by small, single-region cohorts. We sought to develop and evaluate models predicting 12-month PROs using the large, multi-national Movember True North Global Registry (TNGR). METHODS: We identified 27,499 men with localized PCa across 15 countries (2016-2022). Patients were randomly split into training (n = 18,332) and validation (n = 9,167) cohorts. Multivariable linear regressions incorporated baseline PROs, demographics, country, primary treatment (active surveillance, radical prostatectomy, external-beam radiotherapy ± androgen-deprivation therapy, or brachytherapy), and tumor characteristics (grade, stage, percent positive biopsy cores, PSA). Outcomes were changes from baseline to 12 months across 5 EPIC-26 domains. RESULTS: Baseline function, treatment modality, and tumor features were associated with 12-month changes across domains. When applied to the validation cohort, model performance explained 15%, 14%, 19%, 32%, and 28% of variance in urinary incontinence, irritative urinary, bowel, sexual, and hormonal function, respectively. CONCLUSIONS: In this first global analysis of functional outcomes after PCa treatment, predictive accuracy was modest, reflecting substantial regional and practice heterogeneity. These findings underscore both the limits of universal prediction models and the value of large-scale international data for benchmarking outcomes. Future TNGR work will focus on region-specific and locally calibrated models to support more personalized counseling and quality-improvement initiatives.