The decline in approval of new drugs during the past decade has led to a close analysis of the drug discovery process. One of the main reasons for attrition is preclinical toxicity, frequently attributed to the generation of protein-reactive drug metabolites. In this review, we present a critique of such reactive metabolites and evaluate the evidence linking them to observed toxic effects. Methodology for the characterization of reactive metabolites has advanced greatly in recent years, and is summarized first. Next, we consider the inhibition of key metabolic enzymes by electrophilic metabolites, as well as unfavorable drug-drug interactions that may ensue. One important class of protein-reactive metabolites, not linked conclusively to a toxic event, is acyl glucuronides. Their properties are discussed in light of the safety characteristics of carboxylic acid containing drugs. Many adverse drug reactions (ADRs) are known collectively as idiosyncratic events, that is, not predictable from knowledge of the pharmacology and pharmacokinetics of the parent compound. Observed ADRs may take various forms. Specific organ injury, particularly of the liver, is the most direct: we examine this in some detail. Moving to the cellular level, we also consider the upregulation of induced cellular processes. The related, but distinct, issue of hypersensitivity or allergic reactions to drugs and their metabolites, possibly via the immune system, is considered next. Finally, we discuss the impact of such data on the drug discovery process, both through early detection of reactive metabolites and informed synthetic design, which eliminates unfavorable functionality from drug candidates.