The surgeon general's "epidemiologic criteria for causality." A critique

The methodology of the 1982 Report of the Surgeon General is examined with special reference to smoking and lung cancer. Part II of the Report describes the five criteria for causality that have guided the judgment of committees since 1964. I show that not one of the criteria, plausibly interpreted, is satisfied by the epidemiologic evidence for lung cancer. A weakness underlying all the Reports is a prior failure to recognize all the logical possibilities inherent in an association between smoking and a disease. The five criteria and the subjective method of "judgment" are inappropriate to a scientific analysis; they should be replaced by the objective testing of hypotheses. Limitations in the evidence and in concepts about tobacco carcinogenesis preclude definitive conclusions. Nevertheless, the entire association between cigarette smoking and lung cancer--at least in male Caucasoid populations--is unlikely to be explained by causation.