An example of publication bias in systematic reviews

In a recent systematic review it is suggested that antidepressants, more specific the selective serotonin reuptake inhibitors (SSRIs) begin to have observable beneficial effects in depression during the first week of treatment. This conclusion is debated by the authors of a recent letter in the same Journal: Archives of General Psychiatry.

We argue that the criteria of including only studies reporting outcomes for at least 2 points in the first 4 weeks of treatment might have systematically selected positive trials, that is trials showing significant differences between placebo and active drugs during the early phases of the trial. In other words, it is possible that the authors of primary studies reported more often early outcome data suitable for reanalysis when differences emerged.

They also argue that probably most of these trials with the positive outcome were sponsored by the drug companies. And it is not unusual for these companies to not publish negative trials comparing SSRIs with placebo. The authors plead that scientific journals should require, before accepting reports of clinical trials, a statement in which authors agree that data may be accessed by organizations involved in research synthesis. Reports of clinical trials should not be accepted without such a statement. Only in this way would the damage of publication bias be mitigated.