To the Editor: Susan G. Kornstein, M.D., et al. (1) concluded that men and women with chronic depression showed different response to and tolerance of a selective serotonin reuptake inhibitor (SSRI) (sertraline) and a tricyclic antidepressant (imipramine) and also noted that menopausal status affected outcome. They recommended that "both gender and menopausal status should be considered when choosing an appropriate antidepressant for a depressed patient" (p. 1445). This conclusion is based on the fact that according to the method of "last observation carried forward," 45.9% (61 of 133) of the women responded to imipramine, and 56.5% (147 of 260) responded to sertraline; 62.3% (43 of 69) of the men responded to imipramine, and 45.3% (73 of 161) responded to sertraline. We think accurate interpretation of these data requires more careful assessment of the effects of differential attrition.

The authors reported that women taking imipramine were more likely to drop out than women taking sertraline: 26.5% (36 of 136) compared to 14.0% (37 of 264), respectively (Mantel-Haenzel χ²=9.27, df=1, p=0.002, with Yates’s correction). (For men, 19.2% [14 of 73] taking imipramine dropped out, and 24.1% [39 of 162] taking sertraline dropped out.) Thus, when all who began treatment were included in the analysis, the apparent differences in response rates may have been artifacts of the differences in dropout rates. An examination of the completers’ response rates illustrates the problem. Although the authors stated that "a response analysis of subjects who completed the 12-week trial also showed a statistically significant interaction of gender and treatment…with women more likely to respond to sertraline than to imipramine and men more likely to respond to imipramine than to sertraline" (pp. 1447–1448), the exact numbers are not given.

We calculated the proportion of responders for the women who completed the study by using data presented by Dr. Kornstein et al. and found that 62.9% (61 of 97) of the women responded to imipramine and 65.9% (147 of 223) responded to an SSRI. (The denominator for each percentage was calculated by subtracting the number of dropouts for each drug, as reported by Dr. Kornstein et al., from the number who had been taking each drug in the last observation carried forward. For imipramine, 133 in the last observation carried forward minus 36 dropouts equals 97 completers.) These results were not significantly different (χ²=0.16, df=1, p=0.70, with Yates’s correction). The authors’ reference to the interaction of gender and treatment rather than to the difference in the proportion of women responding to imipramine versus sertraline is potentially misleading.

Interpretation of data regarding the men also presents pitfalls. According to the last observation carried foward, the men were more likely to respond to imipramine than sertraline—62.3% (43 of 69) versus 45.3% (73 of 161). The completer analysis for the men suggests that the response rates were 78.2% (43 of 55) for imipramine and 59.8% (73 of 122) for sertraline (χ²=4.9, df=1, p=0.03, with Yates’s correction). Although the results of the completer analysis are consistent with the possibility that men have a superior response to tricyclic antidepressants than to SSRIs, dozens of contrasts between tricyclic antidepressants and SSRIs have failed to show the superiority of tricyclic antidepressants for men. Therefore, one result favoring imipramine out of dozens of studies is consistent with a type I error (i.e., assuming an observed difference is valid when it is, in fact, a function of the anticipated error rate). Furthermore, the completer analysis suggests little difference between the sexes in response to SSRIs, i.e., 65.9% of women and 59.8% of men (χ²=0.28, df=1, p=0.59, with Yates’s correction). The greatest outlier in the completer analysis was the men’s high rate of response to the tricyclic antidepressant (78.2%); the other proportions of response were nearly equal (women: 62.9% for imipramine, 65.9% for sertraline; men: 59.8% for sertraline) (χ²=1.3, df=2, p=0.50).

Regarding menopausal status, the same problem exists in differentiating the relevance of response from that of dropouts. Although postmenopausal women had similar rates of response to the two drugs, those taking sertraline had a higher dropout rate.

We suggest it is impossible to confidently interpret these data because the differences in response rates may result from different dropout rates, therapeutic differences, or a combination of the two. There is no way to assess the precise relevance of each factor (therapeutic effect versus dropout rate) in determining the difference in the proportion of responders noted in the analysis of the group in the last observation carried forward. Interpreting the men’s high rate of response to imipramine is complicated by the possibility of a type I error. We respectfully suggest that these data do not demonstrate an advantage of either drug in either sex or age group.