Number needed to treat: caveat emptor.

Number needed to treat (NNT)-the inverse of the absolute risk reduction resulting from an intervention-was introduced as a yardstick to describe the harm as well as the benefit of therapeutic maneuvers. Analysis using NNT works well when comparing two or more interventions that have their impact over the same period of time in similar populations or patients. Under other conditions, however, analysis based on NNT can produce results that diverge widely from the impact that the interventions can be expected to have on risk of death. This can happen either for entire populations or for an individual when comparing NNTs for interventions which have their effects on different subsets of the population or when comparing interventions which have their effects over different periods of time. We demonstrate how this can occur by comparing the NNTs and effect of intervention on deaths in a population for automatic implantable cardioverter defibrillators (AICDs), heart transplantation, and cholesterol lowering through nutritional intervention with plant stanol ester.