Selective attrition causes serious threats to the validity of experimental trials. Experimental studies in behavior therapy typically include only data from those who complete an experiment. In this paper, we examine the probability of dropping out of a study contingent upon failure to achieve desired benefits. The data are derived from an experimental trial evaluating the effects of weight loss for patients with Diabetes Mellitus. Seventy-six percent of the original participants completed the intervention and the follow-up assessments. However, the probability of failing to appear for follow-ups over an 18 month period was examined. The chances of dropping out of the study were significantly higher for those who did not achieve a goal of weight reduction. This selective loss to follow-up results in an overestimate of treatment effectiveness. By failing to acknowledge drop-outs, treatment failures may be systematically eliminated from the analyses.