Don't Believe the CPA's Hype About Not Believing the Hype

The Canadian Psychiatric Association has cast its lot with the SSRIs. So sayeth Dr. Patrick White, CPA President, in an "Important Message to Physicians." The title of the article reads, "Don't believe the media hype surrounding the inefficacy of SSRIs" and it's a doozy. It critiques the Kirsch et al study in PLoS Medicine which concluded that antidepressant benefits over placebo were generally small. Quotes from his Important Message follow along with my commentary.

It is unfortunate the media coverage obscured the fact that the article does reinforce that antidepressants are in fact effective for persons with severe depression.

OK, fair enough. I'm too lazy to track down all the media coverage on the study, but my recollection was that a few outlets mentioned that Kirsch et al. found that antidepressants work better than placebo for severe depression. But for mild and moderate depression, what was the score? Not mentioned in the CPA piece, but you can see in my prior post, for mild depression, meds were not looking good.

The review combines data from all submissions received by the (US) FDA before drugs are introduced into the US market. Authors do not discriminate between studies which include doses (in dose-finding studies) below the anticipated therapeutic threshold, and studies with more conventional dosing levels. Combining studies in this manner ignores elementary pharmacology, and reduces the ability to discriminatebetween the active ingredient and the placebo. This criticism has also been voiced about their previous publications.

This critique would hold water if another meta-analysis that was published in the New England Journal of Medicine with a bit of media splash (so I assume Dr. White might have read it), noted that, including only approved doses, the impact of antidepressants over placebo was small. So this critique is lame at best.

The main thesis of the article—that there are many failed clinical trials of antidepressants in the FDA database that are not reported in publications—has been known for many years. Such trials are conducted for a variety of regulatory reasons, including dose finding. To show whether antidepressants work in clinical practice requires different studies, which are not included in this article.

I like this. It could be read as: "Sure, pharma doesn't publish a lot of their results -- who cares?" Again, going back to the Turner et al. study from NEJM, the average effect size for antidepressants vs. placebo was d = .15 (meager) in unpublished studies and d = .37 in published studies (small). Selective publication is not just because of dose-finding; it keeps negative information buried. This is not mentioned in the Important Message anywhere. The average effect size for antidepressants over placebo in unpublished studies was very small, which is likely why the studies were not published.

As we know in clinical practice, a substantial number of people do not respond to an antidepressant either:

• at the first dosage they are given, or
  
• within the usual six-week time frame of many of these studies, or indeed
• to the first antidepressant prescribed.
  

Therefore, testing any single antidepressant for a short space of time will bias the results towards diminished clinical efficacy. This point, highlighted by many of those who have commented on the report, has been ignored by the authors and any subsequent media coverage.

Wait a second -- we should just assume that drugs work better in the long-term than they do in short-term studies. No data are cited to support this point. Actually, not a single citation is offered in the entire piece -- apparently this Message was too Important to bother with data. The STAR-D research on antidepressants in clinical practice did not exactly give cause for celebration regarding antidepressant efficacy. If someone did not respond to an initial course of medication, then switching/augmenting was not particularly helpful for most.

Our national mortality from suicide is greater than that from motor vehicle accidents and HIV combined.

Sad. And where's the evidence suggesting that antidepressants reduce suicide more than a placebo? Could it be that providing any sort of intervention that matches up to a placebo (i.e., is credible, delivered by a caring professional, etc.) might possibly reduce suicide? And isn't it also possible that our treatments don't do much to reduce suicide? Intervention in the time of crisis may save lives. But overall, I have not been convinced that we are saving lives in droves through the massive prescription of antidepressants. I know, that is heresy, but if we are going to claim that we are White Knights riding in to save lives, we should have a little bit more solid data on our side.

Thanks to the anonymous reader who passed along this Important Message regarding why antidepressants really work tremendously well and that any research which dares to challenge this point is, by fiat, invalid.