By Frederik Joelving
Mon Nov 22, 2010 5:19pm EST
NEW YORK (Reuters Health) - The 'man bites dog' principle doesn't just apply to news-hungry journalists, researchers said Monday.
Experts who keep the gates at medical journals also have a penchant for reports that show one treatment works better than another, they found, potentially impacting patient care.
Given similar, fabricated reports that compared two treatments, reviewers from a pair of top-line orthopedic journals recommended publishing the fake results in 97 percent of cases when there was a difference between treatments, but only in 80 percent of cases when there was no difference.
"No-difference studies affect practice just as much as positive ones, but they aren't as sexy," said Dr. Seth Leopold of the University of Washington in Seattle, who led the new research. "Something splashy, something new, is more exciting to everybody."
The problem is that favoring studies that identify, say, a new drug as superior to an older one will make the newer drug seem better than it really is.
As a consequence, when doctors scour the scientific literature about a given treatment, they don't see the whole picture.
"The literature should be a fair representation of all the high-quality studies out there, so a practicing physician can make good treatment decisions," said Leopold.
"We felt that these results really confirm beyond any reasonable doubt that positive bias occurs at the level of peer review. That is critically important for the integrity of the medical literature," he added.
Earlier research has shown so-called publication bias is much more common when drugmakers fund studies of their own products than when the government picks up the bill.
But it hadn't been clear if academics - who don't stand to gain financially from a given report - would also favor positive outcomes.
To investigate that, Leopold and colleagues fabricated a pair of fake study reports on whether giving antibiotics after a surgery would help prevent infections.
The reports were nearly identical, but one showed the treatment worked and the other showed it didn't.
More than 200 regular reviewers from The Journal of Bone and Joint Surgery and Clinical Orthopedics and Related Research received one of the two studies.
During peer review, such experts make recommendations to a journal's editors as to whether or not they should publish a given paper.
Apart from being partial to the positive report, reviewers were also twice as likely to pick up errors in the no-difference study and generally rated the research quality lower - despite the fact that the two reports had the same intentional errors and the same methods.
Leopold said the findings suggest reviewers were looking for reasons to endorse positive findings and reject negative ones.
The editors-in-chief of both journals told Reuters Health they were concerned, but not surprised, by the results.
"For practicing orthopedic surgeons, if you see something that has the potential to change your practice, you're more likely to want to recommend that it go into print," said Dr. Vernon Tolo of The Journal of Bone and Joint Surgery.
He said he would circulate the new study among his reviewers to raise awareness about the issue.
Dr. Richard Brand, of Clinical Orthopedics and Related Research, said he pays special attention to publication bias when considering the recommendations from reviewers.
"Realizing the potential for positive outcome bias," he said, "I carefully consider the comments on any recommendation to reject a well-designed study that shows no difference when I receive other recommendations to accept."
SOURCE: link.reuters.com/fev96m Archives of Internal Medicine, November 22, 2010.