The Gijsman Report

Randomized Trials of Antidepressants in Bipolar Disorder

If you arrived here from somewhere other than the Antidepressant Controversies page of this website, or the “More” page from Controversy 2, you might want to start there first to understand the context of the following essay.


Since this article by Gijsman and colleagues is likely to be cited frequently, and since the contents turn out to be so different than the title and abstract imply, a closer look is warranted. See the link above for the full abstract. Here are the key elements in that abstract:

RESULTS: Twelve randomized trials were included, with a total of 1,088 randomly assigned patients. Five trials compared one or more antidepressants with placebo….Antidepressants did not induce more switching to mania (the event rate for antidepressants was 3.8% and for placebo, it was 4.7%)….
CONCLUSIONS: Antidepressants are effective in the short-term treatment of bipolar depression. The trial data do not suggest that switching is a common early complication of treatment with antidepressants.

Sounds like 12 randomized trials find low rates of “switching”, right? Let’s take a look. (Mind you, when I read this article I tried to keep an open mind to the possibility that my previous impression was wrong and that here might be the data to change my thinking. It certainly would make life a lot easier to use antidepressants in bipolar depression and not worry about it so much.) There are three main problems, in my view, with this article:

  1. The title and abstract could easily lead you to believe that 12 randomized trails, taken together, indicate a much lower switch rate (switching patients from bipolar depression to some degree of bipolar mania or hypomania) than published elsewhere(e.g. Goldberg).
  2. However, there were only three randomized trials used to arrive at their conclusion, not twelve as in first sentence of the Results section of the abstract, shown above; or five as in next sentence of the abstract.
  3. Of the three trials, one study accounts for over 50% of the weight of data, and that study used olanzapine/Zyprexa as the “mood stabilizer” (in all three of these trials, patients were taking mood stabilizers as well as antidepressants).

Let’s look at the olanzapine issue first then return to the implications of the title and abstract.

Why the results were so good: olanzapine?

Olanzapine/Zyprexa is a superb mood stabilizer. The large (many patients) and lengthy trials (as long as 18 monthsTohen) of olanzapine versus placebo are some of the strongest data available demonstrating the effectiveness of a medication for bipolar disorder. Olanzapine appears to be better than divalproex/Depakote against manic recurrences, in a 47-week head-to-head trial Tohen(b) and a subsequent Cochrane database review.Rendell

If you could pick any “mood stabilizer” for protection against mania and hypomania while adding an antidepressant, and didn’t have to worry about any of the side effects, olanzapine would probably be your choice. But it’s not really a long-term medication option, if one can avoid it, because of the risks of causing weight gain and diabetes.

Thus my concern with the Gijsman review: more than half of the weight of data suggesting that antidepressants have low risk of causing “switching” comes from patients who were taking olanzapine.Tohen(c) If doctors could use olanzapine freely, if we could stop having to worry about weight gain, diabetes,other aspects of metabolic syndrome, and tardive dyskinesia, then indeed we could probably worry a great deal less about antidepressants in bipolar disorder!

Because of this weighting effect, the question remains: what is the risk of combining an antidepressant with any other mood stabilizer? In the larger of the two remaining studies reviewed by Gijsman and colleagues, there were 117 patients on lithium; 15% were also taking valproate or carbamazepine.Nemeroff Only 4 patients had mania during the 10 week study, 3 on imipramine and one on placebo (none on paroxetine). The authors observed that higher lithium levels seemed to be associated with greater protection against antidepressant-induced mania, but because so few patients had mania, conclusions could only be tentative.

For now, we might conclude that the stronger the mood stabilizer treatment, the greater the protection against anti-depressant induced hypomania or mania. This is obviously intuitively appealing.

Misleading title and abstract?

For the authors to imply that 12 studies, or even 5, are being combined for this review, when in truth only 3 met their criteria for inclusion in the analysis for one of their primary conclusions, seems strikingly misleading. Only on close reading, and arriving at their third graph, does one discover how few studies really were analyzed regarding antidepressant switch rates. For a reader who accesses only the abstract, it is easy to conclude that all twelve studies contributed to the authors’ conclusion that switching is not a “common early complication” . It would not be possible from the abstract to know that only three studies were involved.

Given the controversies surrounding antidepressants use in bipolar disorder, one might expect more deliberate attention to how the study is represented. Watch how this article gets quoted.

For example, the first article citing Gijsman et al has already appeared. In the American Psychiatric Association’s own newsletter, the October 1 issue includes an article entitled ‘Switching’ Risks Minimal in Bipolar Treatment, with the subtitle “Antidepressants can be both safe and effective in the treatment of bipolar depression, without risk of “switching” patients to mania“. Gijsman’s co-author Dr. G. Goodwin is quoted thus: “there is no strong reason to avoid antidepressants” [for patients already on a mood stabilizer]. The implication is that this conclusion derives from the review of 12 randomized trials. In the article, it is not possible to recognize that only three trials were involved.

Update August 2005: in this month’s issue of the same journal in which the Gijsman article appeared, three letters to the editor appear, followed by a response from Gijsman and his colleagues. One letter is somewhat neutral; the others criticize the article. One objects to the conclusions on statistical grounds,Ghaemi which are refuted in the response. The otherFetter emphasizes that the studies analyzed by Gijsman and colleagues are all short term trials of antidepressants, with the longest only 10 weeks. By contrast, they note that there are 3 placebo-controlled trials; two others without a control group but which followed patients for a year or more; and one study which examined patient outcomes after the fact. All three found that antidepressants were associated with worse long-term outcomes in bipolar disorder.

Am I biased? only reporting the negative point of view about antidepressants? Are there more pro-antidepressant articles out there I’m missing?  I’d love to report a study that showed antidepressants were safe and effective over the long term in this illness.