Number needed to treat (NNT) is one of several effect measures that does not capture the time dimension because it is based on observations at a given, random point in time. This aspect makes NNT a problematical measure of the effect of interventions against chronic disease processes in which time to event is an essential issue. Additionally, the statistical properties of NNT are not favorable. It may leave the impression that adverse disease events are completely avoided by intervention when in fact they are postponed. We suggest that NNT should be used and interpreted with caution.