Study focuses on publication bias in journals

Washington -- Physicians trying to keep up with the latest journal articles have their work cut out for them in more ways than one. First, there is the massive number of articles to read. Second, the possibility of publication bias must be considered. Both pose significant challenges when weighing the risks and benefits of therapeutic treatments.

"Evidence shows that even if physicians were to do their darndest to keep up with all of the literature ... and even if they were just keeping up with the quality journals, they would still have to read 19 articles a day, 365 days a year. They just can't do it," said Kay Dickersin, PhD, professor of epidemiology and director of the Center for Clinical Trials at Johns Hopkins University's Bloomberg School of Public Health in Baltimore.

Plus, a study in the Jan. 17 New England Journal of Medicine, which quantified the previously recognized problem of publication bias, shows that it can be difficult to gather all the information needed to properly assess a medication by reviewing journal articles, since negative data are not always accessible.

The researchers determined that 94% of published studies on 12 antidepressants were positive. In contrast, an FDA analysis found that only 51% of all studies on the antidepressants, including those that were not published, were positive. "Selective reporting of clinical trial results may have adverse consequences for researchers, study participants, health care professionals and patients," the authors concluded.

The message is "Don't believe everything you don't read," said lead author Erick Turner, MD, assistant professor of psychiatry and assistant professor of pharmacology at the Oregon Health & Science University in Portland. He said he would welcome other researchers' efforts to apply his study technique to other drug classes.

When asked about the study's findings, the NEJM editors responded in a statement: "We published this paper because the authors went to such extremes to find all the previously unpublished data. It is critically important that physicians have all the available data when deciding which drugs to prescribe for their patients."

Pharmaceutical firms contested the findings. "We are definitely committed to full transparency," said Gwen Fisher, a spokeswoman for Wyeth Pharmaceutical, manufacturer of the antidepressant Effexor. "We make them available in a number of ways, through abstracts and presentations as well as through publication."

Eli Lilly and Co., manufacturer of Prozac, objected to implications in news stories about the study that seemed to suggest the company had suppressed negative trial results. The firm said in a statement, "We clearly have been transparent. The data is publicly available online; we've presented it -- more than once -- in peer-reviewed medical journals."

David Fassler, MD, clinical professor of psychiatry at the University of Vermont in Burlington, called the study useful and interesting, noting that publication bias presents "a significant problem for physicians, researchers and the general public."

"Published literature may overstate the efficacy or understate the risks of specific medications and/or other interventions," said Dr. Fassler, who also represents the American Academy of Child and Adolescent Psychiatry in the AMA House of Delegates.

David Shern, PhD, president of Mental Health America, formerly the National Mental Health Assn., also pointed out the value of the study but cautioned that people with depression should not be discouraged from seeking appropriate care. "It is critical that people understand that antidepressants do work and continue to help millions of Americans recover from depression and other mental health conditions."

Organized medicine has been in the forefront of addressing this issue, Dr. Fassler added. The AMA responded to concerns about the possibility of publication bias in pharmaceutical research in a 2004 report calling for a centralized, publicly accessible registry of all clinical trials. The recommendation received a boost from Congress and the president last fall. A new law now requires the expansion of an existing federal registry.

Also, major journals, including the Journal of the American Medical Association and the NEJM, require that clinical trials that serve as the basis for journal articles and evaluate therapeutic interventions be registered in a central, government-run database.

But a problem remains, Dr. Turner said. Databases do not include the drugs approved before this requirement was put in place. "What about all [those] drugs?" he asks. He cited top-selling Lipitor as an example of a drug in this category. "Are we just going to grandfather that in? We won't know the real data."

Instead, Dr. Turner suggests that an existing cache of data -- that assembled by the Food and Drug Administration for its review of new drug applications -- be made more readily available to the public. This solution would complement proposals by the AMA and others for clinical trial registries, wrote Dr. Turner in a December 2004 essay in the Public Library of Science Medicine, an open-access journal published online by a nonprofit organization of scientists.

Dr. Fassler suggests other steps be taken, too. "Journal editors need to ensure that well-designed studies with negative results are accepted for publication at the same rate as comparable studies with positive findings."

In addition, he urged physicians, the media and the general public to remain vigilant. "We need to read and interpret new studies with appropriate caution."