Clinical characteristics and circumstantial evidence suggest that idiosyncratic drug reactions are caused by reactive metabolites and are immune-mediated; however, there are few definitive data and there are likely exceptions. There are three principal hypotheses for how reactive metabolites might induce an immune-mediated idiosyncratic reaction: the hapten hypothesis, the danger hypothesis, and the PI hypothesis. It has been proposed that some idiosyncratic reactions, especially those involving the liver, represent metabolic idiosyncrasy; however, there are even less data to support this hypothesis. The unpredictable nature of these reactions makes mechanistic studies difficult. There is a very strong association with specific human leukocyte antigen (HLA) genes for certain reactions, but this has only been demonstrated for very few drugs. Animal models represent a very powerful tool for mechanistic studies, but the number of valid models is also limited. There may be biomarkers of risk; however, much more work needs to be done.