Invited commentary: confounding, measurement error, and publication bias in studies of passive smoking.

In December 1992, the US Environmental Protection Agency (EPA) declared that environmental tobacco smoke is a class A human carcinogen, responsible for approximately 3,000 lung cancer deaths per year in nonsmokers (1). Although evidence is still accumulating on the role of passive smoking in coronary heart disease among nonsmokers (2, 3), if this association is real, then the total burden of deaths attributable to passive smoking could be much greater—up to 10-20 times higher—than the risk assessment provided by the EPA. After the EPA risk assessment, several tobacco companies filed a lawsuit against the Agency, claiming that "various sources of bias, including publication bias... could explain any association claimed by EPA between ETS [environmental tobacco smoke] and lung cancer" (4, p. 133). In the present commentary, we focus on three problems—publication bias, confounding, and measurement error—that could potentially threaten the validity of epidemiologic studies of passive smoking.