Author: Eva Pastor
Date: 09-08-09 17:14
Source: The Scientist.com
URL: http://www.the-scientist.com/blog/display/55953/
Date published: September 1st 2009
Trial registries not working
Posted by Alla Katsnelson
Clinical trial registries -- set up in the last few years to ensure trial data see the light of day -- are a long ways from correcting the problem, says a report to be published in the Journal of the American Medical Association (JAMA) tomorrow (September 2).
"This sort of evidence is disappointing, I think," said Ian Roberts, a professor of epidemiology and public health at the London School of Hygiene and Tropical Medicine who was not involved in the study. "It seems to me that there is a lack of implementation and enforcement" in registering trials in such databases.
Clinical trial registries were one solution proposed to fight what's known as publication bias, or when trial sponsors fail to publish negative data. For instance, sometimes an original outcome measure will fail to show an effect of a treatment, so the trial sponsors publish the trial using another measure, which did show a positive effect, as the primary outcome -- raising concerns that data from that original outcome may not even make it into the published report.
"We bring all the evidence of randomized controlled trials together and if there are [such] biases, we could get the answer wrong," Parker said. "People could really suffer."
In 2005, the International Committee of Medical Journal Editors, whose members work at prominent journals such as the British Medical Journal, The Lancet, and JAMA, began requiring trial investigators to register trials -- including the primary outcome measures -- in any of several existing databases before publishing the data. Registering trials is now the norm, but it doesn't seem to have fixed the problem of publication bias.
Philippe Ravaud, a professor of epidemiology at Paris Diderot University in France, and his colleagues chose 323 randomized controlled trials (RCTs) relating to three disease areas -- cardiology, rheumatology, and gastroenterology. The trials were indexed in MEDLINE last year after publication in 10 general medical journals and 10 specialized journals with the highest impact factors. The researchers then pulled the clinical trial registry entry for each of the trials to compare the information in both.
Just 45% of the published trials were properly registered, they found -- that is, registered before publication and spelling out the trial's primary outcomes. The others listed vague primary outcomes, or none at all. But even within those registered properly, 31% switched the primary outcome. In some cases, the article's primary outcome was the secondary outcome in the registry, or not listed in the registry at all; in others, the registered primary outcome was not included in the article.
It's not surprising, Ravaud wrote in an email to The Scientist, that registration was not perfect, considering how recently the process was instituted. "Our point is that even for registered trials there is discrepancies between what is registered and what is published," he wrote, adding that if the problem is so prevalent in highly cited journals, it may be worse still in less prominent publications.
Last year, Roberts and his colleagues published a similar study on 75 clinical trials submitted to The Lancet, which requires researchers to submit the original trial protocols before publication. Even in that journal, said Roberts, 11 of the 37 trials they examined showed major discrepancies in the primary outcomes stated in the protocols and the published studies. It's a problem, said Roberts, "but what's less clear is whose responsibility it is to fix it."
|
|
