In idiosyncratic drug reactions, the occurrence of hepatitis is usually infrequent (1 in 103–105 patients) and unpredictable; the response is not as clearly dose-dependent as is injury associated with direct hepatotoxins, and liver injury may occur at any time during or shortly after exposure to the drug. Adding to the difficulty of predicting or identifying idiosyncratic drug hepatotoxicity is the occurrence of mild, transient, nonprogressive serum aminotransferase elevations that resolve with continued drug use. Such “adaptation,” the mechanism of which is unknown, occurs in such drugs as isoniazid, valproate, phenytoin, and HMG-CoA reductase inhibitors (statins). Extrahepatic manifestations of hypersensitivity, such as rash, arthralgias, fever, leukocytosis, and eosinophilia, occur in about one-quarter of patients with idiosyncratic hepatotoxic drug reactions; this observation and the unpredictability of idiosyncratic drug hepatotoxicity contributed to the hypothesis that this category of drug reactions is immunologically mediated. More recent evidence, however, suggests that, in most cases, even idiosyncratic reactions represent direct hepatotoxicity but are caused by drug metabolites rather than by the intact compound. Even the prototypes of idiosyncratic hepatotoxicity reactions, halothane hepatitis and isoniazid hepatotoxicity, associated frequently with hypersensitivity manifestations, are now recognized to be mediated by toxic metabolites that damage liver cells directly. Currently, most idiosyncratic reactions are thought to result from differences in metabolic reactivity to specific agents; host susceptibility is mediated by the kinetics of toxic metabolite generation, which differs among individuals, probably mediated by genetic polymorphisms in drug-metabolizing pathways (e.g., differences in cytochrome C450 enzyme isotypes or in acetylation). Associations between certain HLAhaplotypes have been drawn with hepatotoxicity of such drugs as amoxicillin/clavulanate, statins, halothane, nitrofurantoin, chlorpromazine, and flucloxacillin. Occasionally, however, the clinical features of an allergic reaction (prominent tissue eosinophilia, autoantibodies, etc.) are difficult to ignore. In vitro models have been described in which lymphocyte cytotoxicity can be demonstrated against rabbit hepatocytes altered by incubation with the potential offending drug. Furthermore, several instances of drug hepatotoxicity are associated with the appearance of autoantibodies, including a class of antibodies to liver-kidney microsomes, anti-LKM2, directed against a cytochrome P450 enzyme. Similarly, in selected cases, a drug or its metabolite has been shown to bind to a host cellular component forming a hapten; the immune response to this “neoantigen” is postulated to play a role in the pathogenesis of liver injury. Therefore, some authorities subdivide idiosyncratic drug hepatotoxicity into hypersensitivity (allergic) and “metabolic” categories. Several unusual exceptions notwithstanding, true drug allergy is difficult to support in most cases of idiosyncratic drug-induced liver injury.