3 - Analytical epidemiology: techniques to determine causal relationships

Summary

In an analytic study, information on exposure to risk factors and on disease outcome are known for each individual in the investigation. Two analytical strategies are available to the epidemiologist for exploring causal cancer relationships. First, is the experimental or interventional approach, in which the investigator can determine the randomized entrance of an individual into the exposed (treatment) and nonexposed (control) groups. This type of study can most accurately address the question of causation.

The second type of study is the non-experimental or observational approach, in which the investigator has no active control over risk factors, but must assess such factors and disease outcome in the natural setting of the study. Study groups to be compared are based either on the presence or absence of disease (case-control) or of risk factors (cohort). On their own, these studies are generally less persuasive in addressing causation than experimental studies. Nevertheless, non-experimental studies do permit examination both of multiple etiological causes for a single disease (case-control) and of multiple diseases for a single cause (cohort).

A considerable literature exists on the techniques of analytical epidemiology which can only be outlined here. Reference should be made to the standard texts for further information (Breslow & Day, 1980, 1987; Lilienfeld & Lilienfeld, 1980; Schlesselman, 1982; Kelsey et al., 1986; Meinert, 1986; Rothman, 1986; Weiss, 1986; Hennekens & Burning, 1987; Checkoway et al., 1989; Kahn & Sempos, 1989). Associated laboratory studies are described in Chapter 5.