An example from the Cardiac Arrhythmia Suppression Trial (CAST) illustrates what we term the 'healthy responder' phenomenon. The hypothesis is that patients who respond to a given treatment are healthier than patients who do not respond. In observational studies this results in an apparent but not real benefit for the treatment. The unique design of CAST, namely, titration before randomization, allows illustration of this phenomenon since we can view the study as both a non-randomized as well as a randomized trial of therapy. We conclude that, in demonstration of a drug effect, randomized trials are essential.