Do the drugs do what now?

I was a little startled to read this editorial in Nature this week, which chides the media for reporting that antidepressive medications simply don't work after the publication of a meta-analytical study finding little-to-no differences in relief from depression when drugs were compared to placebo.

What is special about this paper is that the authors had to file Freedom of Information Act litigation to get the relevant data for the study. A meta-analysis is a study of studies, so to speak (hence meta-), that uses the data from many different experiments and compares all of the results to get a very good idea of whether an effect is really there or not. They literally had to subpoena drug companies and/or the FDA for the results of clinical trials so they could perform their analyses.

What they found wasn't particularly encouraging. They looked at all 47 clinical trials submitted to the FDA for the 6 most "popular" antidepressants from 1987-99. Over 10% (5 trials) didn't even report the improvement scores on psychological tests (Hamilton Depression Index, for example) after drug treatment. Another 7 were incomplete, and knocked 2 drugs out of their study. Happily, in 35 trials covering 4 drugs (including Prozac and Paxil), all of the drug-treated patients showed improvement. However, that improvement was not so much better than the placebo effect that they could be recommended as clinically efficient drugs.

The Nature writers berate the media for "...miss[ing] that the placebo effect can be remarkably strong in psychological and neurological disorders, especially in mild depression." This startled me because it completely misses the point, and I generally expect more from the Nature editorial board. The placebo effect is, in fact, remarkably strong in psychiatric disorders. It may likely be why the concepts of both the double-blind clinical trial and the placebo exist in modern medicine. But if a drug's effects aren't better than a placebo, then the drug doesn't work. It is as simple as that.

If I wrote for Nature I would have attacked from a more different angle, and you can bet the drug companies will: there is nothing simple about these drugs. There is nothing simple about neuropsychopharmacology. Sometimes the drugs work, sometimes they don't. The problem is that depression comes from a wide variety of neuropsychiatric dysfunctions, not just one. Some drugs are going to be efficacious in some of those situations, some will not. Drugs do help a large number of patients suffering from depression, but those who have recovered with pharmacological help will tell you it takes time to find the right drug and dosage. So hoping to respond properly on your first course of a given antidepressive therapy is perhaps ambitious.

But I'll try taking it a step further - the more we talk about "depression," the longer it's going to take us to solve the variety of neurobiological problems that underlie this term. We are talking about of myriad of things when we talk about depression, and until we start being able to devise better correlations between particular symptoms and the biology underlying those particular symptoms - to further develop a knowledge base and a language to talk about depression in a meaningful way - we won't ever have medications that work better than placebo.

I find this most interesting from a population health policy perspective. The study shows that placebos work almost as well as the drugs in all but the most severely depressed patients, and only because those patients get much worse on placebo but improve on medication. What this seems to suggest is that every patient initially presenting with depression should get a placebo (almost the same effect at a fraction of the cost), and those that are severely depressed will get worse rather than better after several weeks. Once you've identified them, they can have the "real" antidepressants. Our health care system would save tens of millions of dollars, and we wouldn't be pulling any more scientific fraud than the FDA has been since 1987...

UPDATE: I spoke with a psychiatrist who runs antidepressant trials, who tells me the 6-week follow-up point used in this study corresponds to the maximal response in the placebo group - scores taken at 10 weeks of treatment would have been much different. Which, of course, continues to beg for the following question: Why did the FDA license the drug on that data?