Antidepressant continuation studies have used 2 different designs. In the placebo substitution design, all patients are initially treated with active medication in an open-label fashion, and then treatment responders are randomized to continue with medication or switch to placebo in a double-blind manner. In the extension design, patients are randomized to a double-blind placebo-controlled acute study at the outset, and responders to active treatment and placebo are continued on the treatment to which they initially responded. We hypothesized that the design of antidepressant continuation studies would impact on the likelihood of relapse. In the extension design, there is no change in treatment. Whether patients responded to placebo or medication, the treatment that produced the response is continued. In contrast, in the placebo substitution design, there is an obvious change in treatment protocol upon initiation of the continuation phase. Patients are aware that they initially received active medication, and there is now a chance that they will be switched to placebo. We speculated that the expectation of a continued positive response is lower in patients treated using the placebo substitution design than the extension design and therefore predicted that relapse rates would be higher. We conducted a meta-analysis of antidepressant continuation studies and compared the relapse rates in continuation studies using these 2 different designs. As predicted, for both the active medication and placebo groups, the frequency of relapse was lower in studies using an extension design. We also found that the difference in relapse risk between antidepressants and placebo was greater with the extension design. Thus, the design of continuation studies of antidepressants was associated with the absolute percentage of patients who relapse on both active medication and placebo, as well as estimates of differential relapse risk between antidepressants and placebo.