Several recent meta-analysis agree that the effect size of antidepressant efficacy vs. placebo is around 0.32. This seems to be stable, irrespective of the data set used or the method of meta-analysis utilized. Some authors argue that this is because antidepressants have a preferential efficacy against the most severe cases of depression but their efficacy in the milder forms is negligible. This is rather an artifact, caused by the ‘mathematical coupling’ phenomenon which strongly binds baseline scores to change with a coefficient of 0.70 even when values are produced randomly. Even if one does not take into account the ‘mathematical coupling’, still the medication effect is superior to the response in the placebo group across all severity levels.
We argue that in our opinion the data suggest that placebo and drug effects are non-additive: antidepressants act independently of depression severity, while the placebo effect if present it is visible only in milder cases. While the response in the placebo group is due to unstable ‘noise’ and ‘artefacts’, the medication effect is reliable, valid and stable. The above should be viewed also under the light of recent meta-analysis suggesting the effect size of the difference between medication and psychotherapy is also around 0.32 and that the response in the psychotherapy arms also seems to be higher in more severe cases.