To the Editor.—
A recent article by Lewis and Winokur (Archives 1982; 39:303-306) concluded with the suggestion that the "so-called switch effect due to TCA [tricyclic antidepressant] reported in the past probably represents random manifestations of bipolar illness," thus contradicting the common clinical observation of switch phenomena induced by TCAs in this population. Although the authors correctly pointed out the important clinical and theoretic implications of their finding, their design seems to suffer from methodologic flaws that severely limit interpretation of their results.An appropriate control group for this study would have been a sample of patients with bipolar illness at equal risk for switch, who did not receive treatment and were followed up for an equivalent time period. Factors likely to contribute to variance in risk might include current clinical state, prior frequency of cycling, and duration of the observation period after treatment assignment.Unfortunately, the Lewis and