Many methods are available to deal with missing data in randomized clinical trials, and active statistical research in the area continues. When, however, a high proportion of outcome data is missing, the methods can produce inaccurate estimates of the true effect size. This article argues that trialists should aim to minimize the proportion of missing data. To that end, the article suggests training investigators and study participants about the importance of completing the trial. It proposes language for informed consent documents, protocols, and case report forms that will distinguish between stopping study medication and removal from the trial itself.