This study was performed to compare the assessments of drug-induced liver injury obtained with 2 methods, the Council for International Organizations of Medical Sciences (CIOMS) scale and the recently validated Maria & Victorino (M&V) clinical scale, in cases submitted to a registry of hepatotoxicity. A total of 215 cases of hepatotoxicity reported with a structured reporting form were evaluated by 3 independent experts. Because of the use of multiple drugs, 228 ratings were generated. The probability of the diagnosis was classified as definitive, probable, possible, unlikely, or excluded, and evaluated for consistency with a weighted kappa statistical test. Absolute agreement between the 2 scales was observed in 42 cases (18%, weighted kappa 0.28) with disagreement of 1 level in 108 cases (47%), and of 2 levels in 70 cases (31%). The best correlation between the 2 scales was obtained for drug-induced liver injury involving a suggested immunoallergic mechanism: the disagreement was 1 level or less in 72% of the cases (34 of 48), compared with 60% of the cases (85 of 141) that involved a presumed idiosyncratic metabolic mechanism. The lowest agreement (6%) was observed in cases with evidence of cholestasis. No agreement was found in cases of fulminant hepatitis or death. The CIOMS scale showed better discriminative power and produced assessments closer to those of specialists. The performance of the M&V scale was poor in reactions with long latency periods (i.e., amoxycillin/clavulanic acid), evolution to chronicity after withdrawal (cholestatic pattern), or death.