Dr. Robinson Replies

To the Editor: Drs. Franco and Malhotra appropriately point out that additional treatment trials are needed before firm conclusions about the most effective treatments for poststroke depression can be drawn. Given the medical vulnerabilities of this population, studies such as ours must always confront a number of challenges in achieving optimal randomization and truly blind placebo control. Some of the criticisms from Drs. Franco and Malhotra may be due to a lack of understanding of the medical constraints that apply to the whole population, not just the depressed subgroup. For example, they state that only 61% of the depressed fluoxetine group and 50% of the depressed nortriptyline group were randomized. Actually, the patients who were not randomized because of contraindications for either nortriptyline or fluoxetine were not all depressed patients but, rather, were divided between depressed and nondepressed groups. In fact, 87% of the depressed fluoxetine patients and 81% of the depressed nortriptyline patients were randomly assigned. In addition, although some patients were not randomly assigned to fluoxetine or nortriptyline, they were all randomly assigned to active or placebo treatment by using double-blind methodology. With the exception of sex distribution, there were no significant differences between the fluoxetine, nortriptyline, and placebo groups.

Drs. Franco and Malhotra next question our use of fluoxetine at a dose of 40 mg/day. It is possible that some of the side effects experienced with fluoxetine could have been dose related; however, this does not necessarily detract from its comparison to nortriptyline in terms of antidepressant efficacy. The authors may wish to note that we analyzed our data by looking at the comparison between 20 mg/day of fluoxetine and 75 mg/day of nortriptyline at 6 weeks and continued to find a significantly greater improvement in the nortriptyline than in the fluoxetine group. In addition, if we had used only the 20-mg/day dose of fluoxetine and had not tried to increase the dose to see if some further improvement would have occurred, we may have been criticized for not trying a higher dose of fluoxetine.

Furthermore, there is increasing evidence that even subsyndromal depressive symptoms may have important clinical and functional implications (1), which are likely to be particularly evident in elderly patients. It is erroneous to assume that reference ranges for rating scales developed in nonelderly populations necessarily apply to geriatric and special populations such as elderly patients in the poststroke condition. Therefore, we relied on clinical diagnosis first, which we then quantified through use of the Hamilton depression scale.

The next criticism was that we used patients with Hamilton depression scale scores of 12 or greater as the criterion for depression. In fact, the criterion for entering the study was meeting DSM-IV diagnostic criteria for either major or minor depressive disorder. A Hamilton depression scale score of at least 12 was used only to ensure at least a mild to moderate level of severity. Since fluoxetine has been demonstrated to be effective in treating milder forms of depression (2), and minor depressive disorder has been shown in our previous treatment trial to respond to treatment with nortriptyline (unpublished data), it was appropriate to include patients with both major and minor depressive disorders in the study group. Drs. Franco and Malhotra comment that the range of weeks poststroke for the fluoxetine group was greater than for the nortriptyline group. Perhaps they failed to note that we actually reanalyzed the data, including only patients who were treated during the first 12 weeks of the study. In that analysis, there was no difference in the time since stroke—i.e., a mean of 5 weeks poststroke for both the fluoxetine and the nortriptyline groups. We still showed that the effect of nortriptyline was greater than the effect of fluoxetine on Hamilton depression scale scores over 12 weeks. Only five of the 23 fluoxetine-treated patients were treated during the second 12-week period. The majority (18 patients) had been treated during the first 12-week period.

Drs. Franco and Malhotra next point out that patients taking SSRIs may gain weight after 12 weeks. Although this may be true, it was beyond the scope of the study presented here and may be an interesting project for Drs. Franco and Malhotra themselves to pursue. In our estimation, 12 weeks is now becoming the standard period for treatment trials, and 12 weeks is an adequate period of time for response to SSRIs. As our outcome of interest was depression response, it would have been outside our hypothesis testing to extend the study for the purposes of measuring weight. We were concerned that weight loss was presumably related to the anorexic effect of the medication (not a symptom of depression) and that the feeling of nausea associated with fluoxetine treatment led to the weight loss. This also happened, by the way, among patients who were nondepressed, demonstrating that this was not simply an interaction among a relatively small group of patients who were depressed but a side effect of fluoxetine treatment in elderly patients after a stroke.

Finally, Drs. Franco and Malhotra point out that SSRIs have been found to be efficacious in the treatment of poststroke depression. We are familiar with the studies they are referring to, one of which did not actually study poststroke depression. That is, the Dam et al. study examined recovery from physical impairment and did not require patients to be depressed to be included in the study nor was any depressive diagnosis made. The Andersen et al. study, however, was a double-blind randomized treatment study showing a significantly greater improvement among patients treated with citalopram than placebo. It may be worthwhile to note that in the Andersen et al. study, there was a 41% decline in Hamilton depression scale score during the 6 weeks of treatment, whereas our original double-blind study of nortriptyline over 6 weeks showed a 77% decline in Hamilton depression scale score; we found a 60% decline in our current study.

As we readily acknowledged in our article, there were methodological limitations in our comparison of fluoxetine with nortriptyline in poststroke depression, and we tried our best to reanalyze the data by looking at each of those potentially confounding methodological limitations. It is certainly possible, as we acknowledged in the article, that lower doses of fluoxetine (i.e., 10 mg/day) may be effective in the treatment of poststroke depression, and we certainly encourage other studies to be undertaken. Given the public health importance of research in this area, we feel it is critical to find ways to effectively treat individuals suffering from poststroke syndromes. Building on the findings of others in a mutually supportive way will help us achieve this end, as opposed to giving in to the distraction of searching for the perfect efficacy study.