How complete is the peer-reviewed medical literature?

Many studies have suggested that there's a "positive effect" bias in publication -- if your study proves what you expected, it gets published. Otherwise, you set it aside. (This is sometimes called "the bottom drawer" effect, as well -- the unpublished studies get put into the bottom drawer of your desk.) One key question: is the bias unintentional (scientists don't publish negative results because they're not interesting), or is it actively harmful (because other scientists think that the research area is more successful than it really is)?

Another common concern is that the bias is actually deliberate, especially when it involves pharmaceutical-company-sponsored studies of drug effectiveness. That's what a new study claims:

Antidepressant Studies Unpublished - New York Times