Passive smoking and cancer risk: the nature and uses of epidemiological evidence.

The apparent effect of passive smoking on cancer risk has become an important social and political issue. For this reason alone the strength of the epidemiological evidence warrants close examination. The research published to date indicates a positive association of passive smoking with lung cancer, but there is no consistent evidence of associations with cancer at other sites. We have summarised the epidemiological evidence, and examined the major criticisms raised against these studies. These criticisms include alleged bias arising from misclassification of exposure to environmental tobacco smoke (ETS) or of personal smoking history, and from differential publication of positive findings. In their strongest form, these critiques challenge the ability of epidemiology to establish causation on any issue. We argue that epidemiology is not inherently different from other branches of science--in each of which scientific "proof" of cause and effect involves judgement based on measurement and logical interference. We also describe the application of epidemiological data to establishing proof, in courts of law, of the lung cancer risk of passive smoking.