Tuesday, October 10, 2006

Who’s Conducting Your Meta-Analysis?

The BMJ just published a rather fascinating comparison of meta-analyses published through the Cochrane Collaboration (which seems generally free of industry influence) and those published by authors who reportedly received funding from drug companies. The authors limited their comparison to those that compared the same two drugs in the treatment of any disease to assure a close apples-to-apples comparison. For example, a Cochrane review of Drug X versus Drug Y for Disease Z would be compared to drug company supported review of Drug X versus Drug Y for Disease Z.

8 pairs of reviews compared Cochrane reviews to industry-supported reviews. The findings: Cochrane reviews were of higher quality than industry supported reviews (p < .01), They also more often stated the search methods used to find studies (p = .06), searched comprehensively (p = .06), avoided bias in the selection of studies (p = .03), reported criteria for assessing the validity of the studies (p = .03), used appropriate criteria in assessing the studies (p < .01), described methods of concealment of allocation (p = .02), and described excluded patients (p = .03), and they used more sources to identify studies (p = .02)... One of the industry supported reviews had no conclusion, as it referred to physiological characteristics of the drug. The other seven reviews supported by industry all recommended the experimental drug without reservations, compared with none of the Cochrane reviews (p < .01).

There is more in the article worth reading. Check out the full text here. It is currently an online only article that I imagine will be published soon in the print version. My respect for the BMJ continues to grow, as it appears to publish many more articles that probe into some of the unseemly practices in medicine than its counterparts, though the Lancet, JAMA, and New England Journal of Medicine all publish good work on these topics on occasion.

Bottom line: Industry engages in selective publication of trials, then they summarize the research by selectively including trials and making recommendations for treatment that are not supported by evidence. Not surprising, but it is more evidence of corruption.

No comments: