PUBLICATION BIAS AND PUBLIC-HEALTH POLICY ON ENVIRONMENTAL TOBACCO-SMOKE

Objective.-To examine the tobacco industry's claim that publication bias against negative studies invalidates the risk assessment of environmental tobacco smoke (ETS) conducted by the US Environmental Protection Agency and other reviews of the health effects of ETS. Design.-Determination of the number of published original research articles that tested the hypothesis that ETS exposure is associated with adverse health effects and that reported statistically significant (''positive'') or nonsignificant (''negative'') results; the number of articles that concluded that ETS is a health risk; and unpublished studies on the effects of ETS on health. Participants.-Articles identified by a computerized search of the medical literature supplemented with material obtained from the tobacco industry and hand searching. Articles were classified as peer-reviewed journal articles or articles from sponsored symposia. Main Outcome Measure.-The statistical significance of results reported in the article and whether or not the article concluded that ETS exposure is a health risk. Results.-More symposium articles than journal articles were reviews (46% vs 6%; P=.0001). More original journal articles than original symposium articles reported the use of statistical tests (96% vs 54%; P=.0001). Of articles with statistical analyses, similar proportions of journal articles and symposium articles reported statistically significant results (57% vs 47%; P=.329). The conclusions of 80% of the original journal articles were positive, compared with 51% of the original symposium articles (P=.006). Conclusions.-There is no publication bias against statistically nonsignificant results on ETS in the peer-reviewed literature. The high proportion of articles in symposia that reach the conclusion that ETS is not harmful primarily results from the inclusion of review articles.