The use of statistical correlations to establish etiological relationships experienced its greatest challenge with cigarette smoking. Biomedical scientists and physicians who were committed to laboratory investigation refused to accept statistical correlations in epidemiological studies as compelling evidence that smoking caused disease. Their views were adopted by the cigarette industry and shared by many health-related government agencies and voluntary associations. The general public and government agencies not concerned with health were much more receptive to evidence provided by statistical correlations.

Statistical Relationships Between Smoking and Disease

Popular wisdom has long associated tobacco consumption with ill health and premature mortality. The popular American term for a cigarette, “coffin nail,” dates back to the nineteenth century. Yet smoking has also been associated with alcohol consumption, debauchery, riotous living, and moral debasement. In order to prove that smoking itself caused disease, its effects had to be separated from other behaviors of smokers.

Before mid-century most physicians did not believe that smoking adversely affected overall health. In his pioneering 1914 text on occupational diseases, W. Gilman Thompson observed that smoking (cigars, pipes, and chewing tobacco) was harmful when used to excess but it was “a great solace to many a workman whose pleasures are few, and, used in proper moderation, it neither shortens life nor impairs health, as is often claimed for it, provided it is not used in boyhood or early youth.” In 1950 an editorial in JAMA, the flagship journal of the American Medical Association, claimed that “more can be said in behalf of smoking as a form of escape from tension than against it… . [T]here does not seem to be any preponderance of evidence that would indicate the abolition of the use of tobacco as a substance contrary to the public health.”