Tuesday, January 22, 2008

Epi 101: Publishing bias

Why are we now finding out that antidepressant-making drug companies only publish (the better) 2/3rds of their trials? Is it because drug companies are evil, money-grubbing, and soulless? Not exactly -- it's probably also a form of publication bias. Journals don't like to publish papers that report no differences (although this is a valid, and important, scientific finding). I've recently had that complaint from a reviewer -- we didn't find a significant difference (actually, we did, just not an easy one to distinguish, but I digress), so what's the point in publishing the results?

The point is that science consists of two types of trials:
  1. trials that work, and
  2. trials that don't work.
Number 1 shows us what to do next. Number 2 shows us what not to do next. Both very useful to know. Never publishing number 2 leads to repetition of useless trials. This is one of my pet peeves.

Oh, and the drug companies probably didn't want to report less than stellar findings.

No comments: