BMJ analysis reveals widespread publication/selection bias in research


Reporting on a study released by BMJ and characterized as an almost existential threat to the medical research system by Dr. Harlan Krumholz, the Milwaukee Journal Sentinel‘s John Fauber writes that “Drug research, even from clinical trials sponsored by the federal government, routinely is suppressed, harming patients and increasing health care costs.” The emphasis is mine, the strong language Fauber’s.

The conclusions are based on a survey of meta-analyses of individual participant data, which the authors broke down by data source characteristics and publication status. The work is heavy on statistical analysis, but even lay readers can understand the broad strokes of what appears to be a widespread issue.

Steve Nissen, the lead author of the analysis, said 35 of the 42 studies he looked at were unpublished and were obtained only because a court case required the drug’s maker, GlaxoSmithKline, to turn over the data.

And it isn’t just pharmaceutical companies’ financial concerns driving the suppression, Nissen and his coauthors found. At that point, it may more of an issue of confirmation bias and other problems which have always lurked within academic research.

A surprising finding in the BMJ analysis was that serious lapses occurred even in clinical trials funded by the National Institutes of Health.

That research showed that less than half of NIH-funded clinical trials were published in a medical journal within 30 months of the completion of the trial and after 51 months, one-third of trials remained unpublished.

Creative Commons License

Republish our articles for free, online or in print, under a Creative Commons license.