The Challenge of Interpreting Weak Epidemiologic Associations in Building the Evidence in Occupational Health

In occupational epidemiology, as in all epidemiology, there are historic issues in assessing whether an observed weak association is real or artifactual.  The main concern with weak associations is that they may be due to uncontrolled or residual confounding.  Therefore, distinctions are made between an “association” and a “causal association.”  The strength of the association is one criterion that helps in making that distinction.  Relative risks below 2.0, or in some cases 3.0, have been criticized as weak associations and not indicative of causality.  However, comprehensive use of Hill’s criteria can be employed to establish a case for causality.  In addition to the strength of the association, it needs to be biologically plausible, consistent with similar results, precede the effect, and in accord with knowledge about the outcome or with experimental evidence.  Associations meeting the criteria can be considered likely to be causal.  However, epidemiology also has a function of generating new hypotheses and stimulating research.  It also may serve as a “call to action.”  Is a weak association to be taken as no association or as a signal that “if it were true” should stimulate some protective actions to be taken as a precautionary measure?  For example, epidemiological studies of air pollution, cellular phones, or endocrine disruptors, where there might be weak associations, still might be useful to trigger precautionary action.  Meanwhile, there are ways of building stronger evidence for occupational safety and health by conducting molecular epidemiologic research, and developing new approaches to identifying nonlinear mechanisms.  When these types of data are available along with weak epidemiologic associations, a stronger evidence base for action can be built.