Time until all-cause treatment discontinuation was the primary outcome of the NIMH CATIE trial of antipsychotics. Three quarters of the patients discontinued prior to 18 months. We discuss advantages and disadvantages of time to discontinuation as an effectiveness outcome. Graphical displays of subjects’ response profiles illustrate the relationship between discontinuation or discontinuation for a specific reason and clinical outcomes such as PANSS scores, quality of life and treatment compliance. We present characteristics of subjects who discontinue for patient decision, and evaluate the extent to which outcome data for these subjects contains signal for treatment group comparisons versus non-informative noise. We consider a post-hoc reclassification of discontinuation reason, and further contrast all cause discontinuation with an outcome based on treatment failure.