Do antidepressants really beget more depressive episodes?

To the Editor: Ghaemi et al 1 describe a randomized study as part of the Systematic Treatment Enhancement Program for Bipolar Disorder (STEP-BD) initiative in which euthymic patients were randomly assigned to continuation versus discontinuation of their antidepressant, but not " mood-stabilizing, " medications and were followed up to 3 years. The rather " eye-popping " result and conclusion, respectively, presented in the abstract are that " rapid-cycling course predicted 3 times more depressive episodes with antidepressant continuation " and " rapid-cycling patients had worsened outcomes with modern antidepressant continuation. " 1(p372) On first blush, this would seem an important finding. However, closer inspection of the data reveals these to be unsupported conclusions. To summarize: 32 patients were randomized to antidepres-sant continuation, while 38 were randomized to discontinuation. Of the 32 in the continuation group, 7 (by my calculations extrapolated from Table 1) had rapid cycling, while 25 were non–rapid cycling. Of the 38 who discontinued antidepressants, 10 were rapid cycling and 28 were non–rapid cycling. There was no signal whatsoever, in either rapid-cycling or non–rapid cycling subjects, of a tendency for those who continued antidepressants to have more manic states. Not to be dismayed, the authors, in an apparent attempt to inflate the presumably horrific, rapid cycling–inducing effects of antide-pressants, compare the number of depressive episodes per year in the resultant 4 groups. Among the continuation group, those who were rapid cycling had about 3 times as many such episodes per year as those who were non–rapid cycling. However, this is a virtually meaningless comparison, especially if taken in isolation as the authors do. A much better comparison to look at the rapid-cycling enhancement tendencies of antide-pressants would be to compare the number of episodes per year in rapid-cycling antidepressant continuation subjects (1.29) versus