PURPOSE: Radioembolization (RE) using (90)Y-microspheres is an effective and safe treatment for patients with unresectable liver malignancies. Radiation-induced liver disease (RILD) is rare after RE; however, greater understanding of radiation-related factors leading to serious liver toxicity is needed. METHODS AND MATERIALS: Retrospective review of radiation parameters was performed. All data pertaining to demographics, tumor, radiation, and outcomes were analyzed for significance and dependencies to develop a predictive model for RILD. Toxicity was scored using the National Cancer Institute Common Toxicity Criteria Adverse Events Version 3.0 scale. RESULTS: A total of 515 patients (287 men; 228 women) from 14 US and 2 EU centers underwent 680 separate RE treatments with resin (90)Y-microspheres in 2003-2006. Multifactorial analyses identified factors related to toxicity, including activity (GBq) Selective Internal Radiation Therapy delivered (p < 0.0001), prescribed (GBq) activity (p < 0.0001), percentage of empiric activity (GBq) delivered (p < 0.0001), number of prior liver treatments (p < 0.0008), and medical center (p < 0.0001). The RILD was diagnosed in 28 of 680 treatments (4%), with 21 of 28 cases (75%) from one center, which used the empiric method. CONCLUSIONS: There was an association between the empiric method, percentage of calculated activity delivered to the patient, and the most severe toxicity, RILD. A predictive model for RILD is not yet possible given the large variance in these data.