How to do an antidepressant trial properly

This study:

A Systematic Review of Comparative Efficacy of Treatments and Controls for Depression

Is fantastically considered. A Metastudy incorporating 13000 patients, the authors have gone to considerable length to seek out the truth. DSM classification of depression may have obfuscated the answer given that the trials used spread over DSM II, III and IV, but the authors considered this and retrospectively classified the patients acoording to various narrow classification systems to remove error, I only mention this to point out that these guys know what they're doing.

It begins sleazily enough, unblinded trials show more treatment effect than blinded ones. Published trials likewise perform better than unpublished ones.  Also, placebos had greater effects in unpublished trials than in the published ones, likewise blinded and unblinded trials respectively.

The study establishes the importance of rigorous methodology. Doubting unblinded trials is therefore very much justified. Furthermore, tests of publication bias are worth our attention. Also, the enhanced placebo effect of active engagement with patients is shown to be the most effective part of treatment, so we must always, therefore, include the counselling of the patient by the doctor in the consultation as a confounding factor. As they write, the data shows that the "type of treatment offered is less important than getting depressed patients involved in an active therapeutic program". Finally, very clinically, antidepressants alone do not perform better than various non medical interventions, like sham acupuncture.

Good Learning, and Happy New Year,

Rohan