Radioembolization (RE)-induced liver disease (REILD) has been defined as jaundice and ascites appearing 1 to 2 months after RE in the absence of tumor progression or bile duct occlusion. Our aims were to study the incidence of REILD in a large cohort of patients and the impact of a series of changes introduced in the processes of treatment design, activity calculation, and the routine use of ursodeoxycholic acid and low-dose steroids (modified protocol). Between 2003 and 2011, 260 patients with liver tumors treated by RE were studied (standard protocol: 75, modified protocol: 185). REILD appeared only in patients with cirrhosis or in noncirrhosis patients exposed to systemic chemotherapy prior to RE. Globally, the incidence of REILD was reduced in the modified protocol group from 22.7% to 5.4% and the incidence of severe REILD from 13.3% to 2.2% (P < 0.0001). Treatment efficacy was not jeopardized since 3-month disease control rates were virtually identical in both groups (66.7% and 67.2%, P = 0.93). Exposure to chemotherapy in the 2-month period following RE and being treated by the standard protocol were independent predictors of REILD among noncirrhosis patients. In cirrhosis, the presence of a small liver (total volume <1.5 L), an abnormal bilirubin (>1.2 mg/dL), and treatment in a selective fashion were independently associated with REILD. Conclusion: REILD is an uncommon but relevant complication that appears when liver tissue primed by cirrhosis or prior and subsequent chemotherapy is exposed to the radiation delivered by radioactive microspheres. We designed a comprehensive treatment protocol that reduces the frequency and the severity of REILD. (HEPATOLOGY 2013)