Published online 11 September 2009 | Nature | doi:10.1038/news.2009.902

News

Publication bias continues despite clinical-trial registration

Fewer than half of registered trials publish their results.

journalsSelective publication of clinical trial data is still skewing the literature.Getty

Selective publication of data from clinical trials is still going on, despite the efforts of a US governmental repository to prevent it, say two new studies. Fewer than half of published trials are adequately registered, and, on the other hand, fewer than half of registered trials are ever published in peer-reviewed journals.

The online, publicly available registry ClinicalTrials.gov was established in 2000 by the US National Institutes of Health to address the problem that sponsors of drug trials often fail to publicly document studies with negative results.

Since then, the International Committee of Medical Journal Editors has instructed researchers to deposit data in such registries as a precondition for publication. And the US Food and Drug Administration Amendments Act of 2007 mandates that clinical trials of FDA-regulated products must be registered at ClinicalTrials.gov.

Nonetheless, there were still clear gaps in the system, so two research teams set out to identify the registry's shortcomings.

Moving goalposts

In the first study, a team led by Philippe Ravaud, an epidemiologist at Paris Diderot University, reviewed 323 studies relating to three medical areas — cardiology, rheumatology and gastroenterology — published in high-impact journals last year. The team found that just 46% of the trials had been correctly registered with clearly stated goals before publication, they report in the Journal of the American Medical Association1.

“It's not enough just to conduct your research: you have to disseminate your findings.”

Joseph Ross
Mount Sinai School of Medicine in New York

Even among the articles that were properly registered, nearly 1 in 3 studies switched the stated goals in the final publication, say Ravaud's team.

The study "exposes the apparent failure of the editorial process to pick up on [authors changing the stated aims of clinical trials]", says Alastair Wood, a clinical pharmacologist at Weill Cornell Medical College in New York and the managing director of Symphony Capital, a private-equity company that invests in the biopharmaceutical sector. He proposes that journals should create an automated system to send ClinicalTrials.gov reports to reviewers along with the authors' manuscripts to allow for easier scrutiny during the peer-review process.

However, many manuscripts don't even acknowledge their linked records on the registry, according to the second study, published in PLoS Medicine2 by Joseph Ross, a health policy researcher at the Mount Sinai School of Medicine in New York and his colleagues.

Missing trials

Ross's team came at the issue of publication bias from a different angle than Ravaud and his coauthors. They started with the entries on ClinicalTrials.gov and asked how many trials were eventually published in the scientific literature. The researchers analysed 677 trials at phases II–IV completed before 2006 and found that only 46% had been published. Of those, fewer than a third had cited their ClinicalTrials.gov record of the trial.

Studies primarily sponsored by industry had the worst publication record, with only 40% appearing in medical journals, and NIH-sponsored trials were not much better, at 47%.

"It's not enough just to conduct your research: you have to disseminate your findings," says Ross. "We have a ways to go before we can truly convince ourselves that the medical literature is comprehensive."

Deborah Zarin, director of ClinicalTrials.gov, finds a silver lining in the two reports. "Registration alone cannot improve the quality of research," she says. "It can only allow it to be systematically examined and monitored." In that way, the two studies show that registry is succeeding at pinpointing the deficiencies in trial reporting.

Ravaud agrees, noting: "Without registries, such a study would be impossible to do."

ADVERTISEMENT

Ida Sim, a medical informatician at the University of California, San Francisco, notes that ClinicalTrials.gov was never intended to completely solve the publication bias problem. Rather, it was established simply to let people know that a particular trial had taken place.

"Now comes the more mundane part," Sim says. "Once we know that a study exists we have to focus on full reporting." She says that researchers should be compelled to deposit the results of clinical trials in an online database, but not necessarily in traditional journals.

Zarin also points out that the 2007 FDA regulations will phase in additional reporting of summary results to ClinicalTrials.gov. "These studies may provide a baseline against which we could judge the impact of [the FDA rules] in the future," she says. 

  • References

    1. Mathieu, S., Boutron, I., Moher, D., Altman, D. G. & Ravaud, P. J. Am. Med. Assoc. 302, 977–984 (2009). | Article | ChemPort |
    2. Ross, J. S., Mulvey, G. K., Hines, E. M., Nissen, S. E. & Krumholz, H. M. PLoS Med. 6, e1000144 (2009). | Article

Comments

If you find something abusive or inappropriate or which does not otherwise comply with our Terms or Community Guidelines, please select the relevant 'Report this comment' link.

Comments on this thread are vetted after posting.

  • #60454

    I think publication bias of clinical trials can lead to increased harms and costs for patients and society.

Commenting is now closed.