Listening to Prozac but hearing placebo: A meta-analysis of antidepressant medication

In paper citation (Kirsch & Sapirstein, 1998)

This paper specifically looks at the effects of the placebo in placebo-controlled drug studies treating depression. This analysis only includes drug studies before 1995 and therefore misses most of the modern drugs prescribed today. However, Kirsch and Sapirstein's results are really shocking. They find that in drug studies, the placebo group improves in proportion to the effectiveness of the drug. The correlation between the placebo response and the drug response is r=.90, p<.001 and the placebo response is 75% of the drug response. So if the drug response in a sertraline drug trial was 2.0, the placebo response in that trial would be around 1.5. These numbers were slightly higher for active placebos... side effects actually may convince people that they are receiving full treatment, and therefore cause them to expect a better response. This is also interesting in light of the more modern Turner and colleagues meta-analysis, which found that 49% of FDA approved clinical trials show no significant difference between placebo and antidepressant trials. Perhaps both the experimental and control patients are improving in proportion to the expected efficacy of the drug, and so only the most effective drugs will come out ahead of their placebos.

Furthermore, the authors compared the effect of taking a placebo pill to receiving no treatment while on a wait-list for psychotherapy trials. For the people on wait-lists, or receiving no placebo treatment at all, their depression scores worsened overall during the course of the study. This shows that the placebo pill is more effective than no treatment.

My remaining questions actually involve the rating scales of depression. It could be that different research centers treat their patients more holistically, and that is why the scores are so well correlated, but it could also be that people respond to the depression interviews in a rote manner. For example, if I know the symptoms of depression, and I want to be a good research subject, I may describe my pre-treatment symptoms in a way so that the doctor will give me the quick fix and treat me with an antidepressant. Then when the doctor asks me follow ups, I also know what improvement should look like, and I've been looking out for that in my daily life, and so I focus on the improvement for my follow up questionnaire. Whether the drug has actually improved my mood is hard to tell, because I started by looking for symptoms to describe to my doctor, and then I started looking for improvements to tell to my doctor. Thus the placebo effect.

Does anyone else think that is totally plausible? I guess Kirsch & Sapirstein would, but they do not mention it in their discussion. In cases of mild to moderate depression, I would like to see a study that looks at various types of placebos to see what works best. Inert substances should be better for our bodies in the long run, and especially useful for pregnant mothers who desire treatment for mood disorders.