Background: mCSPC pts experience adverse outcomes (AOs) with intensified androgen deprivation, resulting in treatment discontinuation; however, there is limited data to identify those at risk. The IRONMAN registry is a global initiative to enroll advanced prostate cancer patients worldwide, and survey data on AOs, patient-reported outcomes (PROs), and clinico-epidemiologic (demographic, cancer, and treatment) details. We report the prevalence and predictors of AOs from baseline demographic, clinical, and PRO information with the onset of AOs and present a risk prediction model for clinical use. Methods: We analyzed data from 553 mCSPC patients in the IRONMAN registry from inception to March 2023 who received ADT+ alone (Abiraterone 47.6%, Enzalutamide 26.9%, Apalutamide 22.6%, Darolutamide 2.9%) and had completed baseline PROs. We defined AOs, reported by 60 global sites, as non-progression-related clinically significant serious adverse events (csSAE), treatment discontinuation, and dose reductions (DR/DC), delineated as a composite outcome based on first event. We processed baseline variable groups: Demographics (race, ethnicity, marital status, study site), Clinical Factors (age, ECOG score, metastatic sites, ADT+, concurrent medications, labs [ALP, Hb, LDH, and PSA]), Concurrent therapies (steroid use within 30 days [d] & steroid dose, radiation/surgery prior or planned in 6months) and PROs (EORTC QLQ-C30 and EPIC-23). We developed a multifactorial logistic regression model to predict adverse outcomes (AOs) within two years of initiating ADT+ therapy. All variables with less than 25-30% missing data were included. Variable selection was conducted in two layers: intra-group stepwise selection followed by an inter-group selection of variables retained from the first layer. The model's discriminatory ability was evaluated using the area under curve (AUC). Results: 553 pts met inclusion (51.6% North America, 72% White, mean age: 71 years [range, 40 to 96 years], and 62.9% confirmed bone-metastatic mCSPC). 119 pts (21.5%) developed AOs (61 csSAEs [18 dead; Median: 140 d; IQR 81-360 d], 58 DR/DC [Median: 122 d; IQR 43-262 d]). <5% went off study. The multifactorial model (n=314 pts) performed well at predicting AOs (AUC 0.77 [0.70-0.84]), with significant factors being more self-reported pain, greater urinary frequency, prednisone>5mg within 30d, more self-reported weakness and poor sexual function within 3 months of ADT+ initiation. Conclusions: 1 in 5 mCSPC patients on ADT+ experienced non-progression-related AOs at 2 years. Significant, easily assessed predictive factors are identified and may be utilized in their clinical management. A novel multifactorial model is presented to predict AOs at 2 years. Further validation, assessment of missingness bias, and imputation are planned.