How Strong is the Association Between Smoking and Disease?

Abstract

So far I have taken for granted the existence of a strong statistical relationship between cigarette smoking and disease, particularly cancer and CHD. Such a relationship can be demonstrated along two rather different lines. The first of these is the cross-sectional method. Here one studies a group of patients suffering from a given disease and compares them with some form of control group that may be suffering from a different type of disease, or no disease at all, with respect to smoking habits. This method is obviously weak methodologically. The choice of a control group is clearly crucial, but any particular choice may be faulted for a variety of reasons. The method depends a good deal on memory (When did you take up smoking? How many cigarettes did you smoke? What types of cigarettes did you smoke?), and memory is known to be faulty and subject to falsification. The method has many other weaknesses, which are discussed later. Few epidemiologists would doubt that the second method to be discussed, namely, the prospective study, is superior because it is not subject to these weaknesses.