STATS at George Mason University how to evaluate health risks
9. Is there publication bias?


A related issue to watch out for is “publication bias.” Journals are three times more likely to publish studies of treatments showing positive results than those which show no effect.* This reflects a natural human tendency to look for “news,” which gets further exaggerated by the media — where studies that say something works are almost always given more coverage than those studies which show it doesn’t.

Naturally the drug companies want to show their products in the best possible light too. The corollary, warns Dickersin, is that “some research has shown that we need to be particularly concerned about industry studies which are negative, because these are even more likely to remain unpublished.”

This means that when you see a news story about a study showing a new drug to be effective, neither you nor the reporter who wrote it may be aware that there could be a pile of studies sitting in a drug company file cabinet showing that it didn’t work. Furthermore, if one of those negative studies does somehow get published, reporters are less likely to find it newsworthy and cover it.

To address this problem, Dickersin supports the creation of a centralized register of all trials conducted by industry, government and private funders. With access to such a registry, those trying to review the research, like the Cochrane Collaboration, could determine which studies were started but never published and then try to find out why. (A good source for review articles:

*Dickersin K. How important is publication bias? A synthesis of available data. AIDS Educ Prev 1997;9(1 Suppl):15-21.


risk home

STATS home

email this page to a friend

next risk topic

about staff publications get e-mail updates contact us search contribute