The file drawer problem (described here) is a publication bias in which results that are significant are published more that results that are not statistically significant. Thus, non-significant findings end up in the file drawer.

A research team led by Erick Turner of the Oregon Health & Science University published a report in the New England Journal of Medicine, which describe this problem in some studies on the efficacy of SSRI medications. Here is an excerpt of some of the discussion in this article:

“Selective publication can lead doctors to make inappropriate prescribing decisions that may not be in the best interest of their patients and, thus, the public health,” they wrote.

The idea that unfavorable test results get quietly tucked away so nobody will see them — sometimes call the “file drawer effect” — has been around for years.

The Turner team was able to study the question because the U.S. Food and Drug Administration has a registry in which companies are supposed to log details of their drug tests before the experiments are begun.

They could see which experiments approved by the FDA between 1987 and 2004 were ultimately publicized in the medical literature and the main criteria the researchers planned to measure success.

“It tells you where they placed their bets before they saw the data,” Turner said in a telephone interview.

Of the 74 studies that started for the 12 antidepressants, 38 produced positive results for the drug. All but one of those studies were published.

However, when it came to the 36 studies with negative or questionable results, as assessed by the FDA, only three were published and another 11 were turned around and written as if the drug had worked.

“Not only were positive results more likely to be published, but studies that were not positive, in our opinion, were often published in a way that conveyed a positive outcome,” said the team.

For example, of the seven negative studies done on GlaxoSmithKline’s Paxil, five were never published. The researchers found three studies for GSK’s Wellbutrin SR, but the two negative ones never reached print.

There were five studies for Pfizer’s Zoloft, but the three showing the drug to be ineffective were not published and a fourth study, ruled as questionable by the FDA, was written and published to make it appear that the drug worked.


The figure below comes from the Wall Street Journal, which also has a good write-up of the study.