Wednesday, February 22, 2012

Antidepressants - Publication Bias

In 2008 the FDA decided to look at all the published data about antidepressants.  They found that 51 studies had been published.  They looked back at their own records about how many studies had been done on antidepressants, and found the number to be much higher - 74.

Where were the missing 23 studies?  (Those 23 studies had 3,449 patients who went through full drug trials, and their results were never published)

Why weren't they published?

The FDA looked at the 51 studies that were published and the looked at the publication itself to see how the drugs were represented.  The graph below shows the result of that analysis:

(White Boxes above the red line are studies that showed that the drug worked.  Black Boxes below the red line are studies that had negative or mixed results)


Basically this says that out of the 51 studies, 48 showed that the drug had a positive effect.  That's 94% of all studies showing a positive result!
There were only 3 studies that showed that the drug had negative or mixed results. 
THEN, the FDA looked all 74 the trials that had been done, without looking at the publications.  They analyzed the data themselves to see what they would find.  Were the unpublished studies mostly positive, or negative?  Were all the published studies portrayed accurately?
The graph below is their result when they FDA analyzed the data from all 74 trials themselves:


Yep.  It turns out that 36 studies showed mixed or negative results.  Only 38 studies were positive.  The positive results dropped from 94% to 51%
All 23 missing studies showed negative or mixed results.  13 of the published studies were misrepresented in the publication to make them look like they had a positive result.

What does this say about "double-blinded, placebo controlled, randomized control trials?"  What does is it matter that the studies are done correctly if they are misrepresented or selectively published?

This is only part of the depressing news about anti-depressants.  More to come...

Source: New England Journal of Medicine Article

2 comments:

Kristen Bowen said...

I have always wondered about the socalled studies. Great article on how they do it.

Heidi Higley said...

I still want to try a placebo. Just for fun.