Everybody Expects the Placebo Inquisition

An unexpected gem from last year's Journal of the American Psychoanalytic Association: Mind over medicine.

Surprisingly, it has nothing to do with psychoanalysis. Rutherford and colleagues performed a meta-analysis of lots of clinical trials of antidepressants. Neuroskeptic readers will be all too familiar with these. But they did an interesting thing with the data: they compared the benefits of antidepressants in trials with a placebo condition, vs. trials with no placebo arm, such as trials comparing one drug to another drug.

Why do that comparison? Because the placebo effect is likely to be stronger in trials with no placebo condition. If you volunteer for a placebo controlled trial, you'll know that you've got (say) a 50-50 chance of getting inactive sugar pills. You'll probably be uncertain whether or not you'll get better, maybe even quite worried. On the other hand if you're in a trial where you definitely will get a real drug, you can rest assured that you'll feel better - and that in itself might make your depression improve.

The paper only presents very preliminary results, but they say that:

Our group at Columbia has completed preliminary work involving metaanalyses of randomized controlled trials comparing antidepressant medications to a placebo or active comparator in geriatric outpatients with Major Depressive Disorder (Sneed et al. 2006). In placebo controlled trials, the medication response rate was 48% and the remission rate 33%, compared to a response rate of 62% and remission rate of 43% in the comparator trials (p < .05). The effect size for the comparison of response rate to medications in the comparator and placebo controlled trials was large (Cohen’s d = 1.2).

They only looked at trials of old age patients, but the same probably applies to everyone else.

Why does this matter? The authors suggest one very important implication. There are quite a few trials nowadays comparing the effects of psychotherapy, medication, neither, or both. How it works is that everyone gets pills, 50% of them real drugs and 50% placebos; also, half the people get psychotherapy while the others remain on the waiting list.

These trials often find that medication plus psychotherapy is better than just medication alone. This has led to the idea that therapy and drugs should be combined in clinical practice, a message which goes down really well, because it gives both psychopharmacologists and therapists the feeling that they have an important job to do. An example of this kind of trial is the influential TADS from 2004, finding that Prozac and therapy both work in depressed teens, and combining them is best. Everyone's a winner.

But as Rutherford et al. point out, there's a problem with this reasoning. The people who only get antidepressants don't know that they're getting any treatment, because they might be getting placebo. But the people who get antidepressants and therapy know that they're getting at least one real treatment (therapy). This is likely to improve their outcome through an expectation effect. (In fact, for some reason, in TADS, the people on combination treatment were told that they were getting both - they specifically knew they would never get dummy pills - which will have made this even worse.)

Now you could say that this doesn't matter: TADS and similar studies show that therapy and medication is better than just medication, and it's purely academic whether that's "just a placebo effect". But the key point is that in real life people always get medication knowing that it's real - so, like the therapy plus medication people in the trials, they get the benefit of the certainty that they are getting a real treatment. In the trials the medication-only group don't know that, but in real life they do - so the benefits of adding psychotherapy might be less, or even zero, in real life.

The authors of the TADS study did acknowledge this in their original paper, but only very briefly - here's all they say about it:

Blinding patients in the placebo and fluoxetine alone groups but not in the CBT alone group (participants knew they would not be receiving fluoxetine) and the fluoxetine combined with CBT group (participants knew that they would be receiving fluoxetine) may have interacted with expectancy effects regarding improvement and acceptability of treatment assignment.

Yet this limitation means they, strictly speaking, all TADS showed is that Prozac works in this group. It doesn't prove that adding (very expensive) therapy benefits anyone, in the real world. This is not to say that psychotherapy doesn't work of course, maybe it does, but the point is that therapy + medication trials may be best without a placebo.

Rutherford, B., Roose, S., & Sneed, J. (2009). Mind Over Medicine: the Influence of Expectations on Antidepressant Response Journal of the American Psychoanalytic Association, 57 (2), 456-460 DOI: 10.1177/00030651090570020909