Five years ago, Maryland psychologist Ed Pigott read the first published results of the NIMH's large STAR*D study of antidepressants and depression. However, even as he read that first article, he got the sense that "significant researcher trickery was afoot." Since then he has systematically exposed the trickery, piece by piece.

His latest article on the study, "STAR*D: A Tale and Trail of Bias," has just been published in Ethical Human Psychology and Psychiatry. He is also now blogging about his findings on, and has posted documents there that he relied upon in his "deconstruction" of the $35 million study.
Here is the (scientific) review:

I've seen this same sort of thing in other recently published psychiatry studies. It seems that protocols are worthless if one is going to not follow it and fudge ones findings as much as possible (eg accept much looser clinical thresholds, not publish half of your measures and use looser p values than stated in your published protocol).