Binary outcome measures such as death or myocardial infarction are common to clinical trials. In contrast to continuous outcomes, there is no natural best measure for the difference in treatment effects between the test and the placebo groups. The risk difference, the risk ratio, and the odds ratio are popular metrics. Unfortunately, each has some negative as well as positive characteristics. There is an additional problem when the outcomes are rare, which is often the case in safety studies in which we do not expect many side effects. The problem surfaces when there are zero occurrences in both the treatment and the control groups. Indeed, some meta-analyses have deleted studies in which there are zero cells in both groups. We here provide a survey of the characteristics of alternative metrics, of how they affect the analysis, and discuss some alternatives.