Antidepressants Ineffective? Don't Believe It! (Part 2)

Last week, I posted part 1 of my reaction to the JAMA study showing that antidepressants don't beat placebo for mild to moderate depression. The many comments I received were intelligent and informed--reading them will give anybody an instant education in the limitations of antidepressant research.

It has been clear for years that antidepressants barely outperform sugar pills, at least in the artificial environment of the double blind, placebo controlled study. It is also clear that most psychiatrists in the trenches are perplexed by these results, because in our practices we appear to see people responding very robustly to antidepressants--including those with only mild to moderate symptoms. So what gives?

I presented a hypothetical patient who became depressed because of financial woes, who saw a psychiatrist, received a prescription for Zoloft and Klonopin, and got better. Some commenters wondered why the patient was given medication rather than therapy. My answer is simply that in modern psychiatry, such a patient is more likely to be given meds, even though therapy may be just as effective. The reasons for this are complex, but have to do with how psychiatrists are trained (primarily, they are taught how to prescribe medications), how insurance companies reimburse (they pay twice as much for an hour's worth of medication visits than for an hour of therapy), and what patients want (many prefer the quick fix of medications over multiple therapy visits). I personally believe that psychiatrists should incorporate more therapy into their practices, but the current state of the field is as outlined above.

Anyway, I then considered what would happen if the same patient got enrolled in a randomized, placebo-controlled trial of an antidepressant. Once accepted, he is told that he will be randomly assigned to either active medication or placebo.

Let's assume that our patient lucks out and gets assigned to the latest SSRI, "Newexa," and let's further assume that Newexa works as well as all other SSRIs (like Zoloft or Lexapro). One would assume that since our patient is assigned to a drug that works as well as Zoloft, he stands the same chance of improvement. But not so.

When patients are enrolled in preliminary "open label" studies, in which everybody is given the active drug, and there is no blinding, the response rates are in the 70-80% range. But when patients are given the same drug in the context of a placebo-controlled trial, the response rates plummet to the 40-50% range. The reason for this is that if you strongly believe that you are taking an effective treatment, this confidence in itself boosts the effects of the drug. On the other hand, if you know that there is a 50% chance that you are taking a sugar pill, you have less confidence in the chances of a recovery. Essentially, by not telling patients what they are taking, researchers are turning down the volume of the placebo effect. But when I see a patient in my office, I do just the opposite. I try to maximize the placebo effect. I say things like, "I've had great success with Zoloft in patients like you....I would say you have an 80-90% chance of feeling much better within the next 2 weeks....These drugs are remarkably effective...."

So this is the crux of the issue. Response rates are always lower in randomized controlled trials than they are in the real world of the office, where our patients know exactly what they are taking. Thus, when I read about the measly outcome results of antidepressants in these trials, I take it with a big grain of salt. I interpret it in light of many other studies that have indeed shown antidepressants to be more effective than placebo in conditions as diverse as severe depression, anxiety disorders, and eating disorders. Rather than concluding that these meds don't work, I conclude that they do work, but that there is a large placebo component.

In the best of all worlds, we would be allowed to prescribe sugar pills to patients, and we would presumably get a robust response without many side effects. This is considered unethical by the AMA. Nonetheless, according to this recent survey of 679 U.S. physicians, about half reported prescribing a placebo on a regular basis, most of which were "active" placebos--things like vitamins and over the counter pain relievers such as Tylenol.

The bottom line is that while medicine could not survive without placebo-controlled trials, they are rarely generalizable to the real world. I view their results more as signals of efficacy rather than as definitive answers.