Few drug development surprises can be as devastating as toxicity problems that only show up under a combination of conditions as idiosyncratic toxicity. Because of the role of variations in human drug metabolizing enzymes there may only be subtle (or no) evidence of such problems during pre-clinical safety studies. Such problems are also unlikely to show up in all but the largest clinical trials, but if the side effects are serious, it can result in product withdrawal. What are the prospects for different approaches to screen for, or predict such problems, and when are such efforts justified? Particular attention will be paid to the problems of reactive metabolites, and ways to anticipate problems with such metabolism.