Confounding

Abstract

An apparent relationship between a disease and a risk factor may be explained (confounded) by their joint association with an intermediate third ‘true’ risk factor. The issue here is not one of impaired validity; if the study had been carefully conducted then the relationship observed was correct and the problem of confounding is therefore one of interpretation. Confounding can only be proved after appropriate analysis. Further confounding as an explanation of an association is not (or is only extremely rarely) an all-or-nothing phenomenon. The effect of confounding will be, more usually, to alter the strength of an apparent relationship between two variables (e.g. risk factor and disease).