Box 9-3
Guidelines for Causal Inference
1. Strength of association: A strong association between a potential risk factor and an outcome
supports a causal hypothesis (e.g. a relative risk of 7 provides
stronger evidence of a casual association that a relative risk of 1.7
2. Consistency of findings Repeated findings of a an association with different study design
and different populations strengthen a casual inference
3. Biological plausibility Demonstration of a physiological mechanism by which the risk
factor acts to cause disease enhances the casual hypothesis.
Conversely, an association that does not initially seem biologically
defensible may later to discovered to be so
4. Demonstration of correct temporal sequence For a risk factor to cause an outcome, it must precede the onset of
the outcome
5. Dose – response relationship The risk for developing an outcome should increase with increasing
exposure (either in duration or quantity) to the risk factor of
interest. For example, studies have shown that the more a woman
smokes during pregnancy, the greater is the risk for delivering a
low birth weight infant
6. Specificity of the association The presence of a one to one relationship between an agent and a
disease (i.e., the idea that a disease caused by only one agent and
that agent results in only one disease lends support to a casual
hypothesis, but its absence does not rule out causality) This
criterion grows out of the infectious disease model in which it is
more often though not always satisfied and is less applicable in
chronic diseases
7. Experimental evidence Experimental design provide the strongest epidemiologic evidence
for causal associations, but they are not feasible or ethical to
conduct for many risk factor- disease associations