Med-Psych.com

One of the topics that I mention in psychopharm lectures to students is the poor evidence base we really have when making decisions on medications. One of the big problems is publication bias. Essentially, negative trials don't get published while positive studies get published. The result is that if there were 10 studies done on a certain drug and 9 of the studies were negative with 1 being positive, the only one that gets published is the positive trial. This has been well known in academic circles, but there is finally a nice publication on this issue in the New England Journal of Medicine in regards to antidepressant trials:

Selective publication of antidepressant trials and its influence on apparent efficacy.

Briefly, after reviewing 74 FDA registered studies, 31% of the studies were not published. Of the 37 positive studies, only 1 was not published. In contrast, of those studies with negative or questionable results, 22 were not published, and 11 were published in a way that appeared to suggest a positive outcome. It appears only 3 negative studies were published as such. Therefore, 94% of the published trials appeared positive while in reality only 51% of the trials done were positive.

So - what does this mean? It means we have to get better about registering all trials and ensuring that the results get published whether the outcomes are positive or not.