Everybody has won and all must have prizes

Fiona Godlee, BMJ editor, in her latest Editor's Choice, is pleased to see what she calls evidence that counters claims that antidepressants have little or no effect. She finds this evidence in an editorial that commented on a meta-analytic study, which found no substantial difference in efficacy between the second generation antidepressants.

The work that Godlee thinks "may well have dissuaded patients and some clinicians from considering or continuing antidepressants" is that of  Irving Kirsch, which I have discussed in a previous post. What the authors of the editorial say, which Godlee thinks counters Kirsch's work, is that the meta-analysis found that "[j]ust under two thirds of the patients responded to treatment by 12 weeks and just under half achieved full remission". They don't say compared to what or define response or remission. 

Before it can be concluded that this finding means that antidepressants are effective, it is necessary to ask what the controls were. People improve naturally from depression. There is a large placebo response to antidepressants. 

There is also a question about the definitions of response to treatment and remission in clinical trials. Fifty percent reduction in score severity is most commonly used as the measure of response. Many of these patients who are classified as responders nonetheless remain highly symptomatic. A reasonably low cutoff score on rating scales is usually proposed as the definition of remission. This is not the same as being completely asymptomatic.  So, the results noted by Godlee may not be as good as they seem. (See my previous blog entry, How easy is to treat depression?)

Moreover, most head-to-head comparisons of antidepressants do not include a placebo arm. Antidepressant response rates are higher in comparator trials compared to placebo-controlled trials, because of the greater expectancy effect as patients know they will definitely be given an antidepressant rather than have a chance of getting placebo. If all that is being suggested is that the effect size seems larger for trials comparing one drug with another rather than with placebo, this is not a new finding. 

What people find difficult, including the BMJ editor apparently, is accepting that randomised controlled trials of antidepressants (and probably drugs in the rest of medicine - see previous blog entry) have not eliminated expectancy effects because they are not really double-blind (see a critical exploration of the evidence). The throw-away remark in the BMJ editorial (the article is after all more about the comparative efficacy of second generation antidepressants than about the effectiveness of antidepressants as such) should have been subject to more scrutiny before publication. It then wouldn't have been taken up by the editor to undermine the challenge to the wish-fulfilling nature of antidepressant medication.