Evaluating the role of confounding in pharmacoeconomic studies

Confounding may be present in nonrandomized etiological research involving human populations. It can result in erroneous conclusions about the effect of exposure on a disease outcome or about any form of causality between predictors and outcomes. Confounding can wholly or partially account for the apparent effect of the risk factor under consideration or mask the underlying, true association. Not controlling for the effects of confounding can lead to biased results, thus compromising the validity of study conclusions. The three goals of this article are: (1) to define a confounder or a confounding variable, (2) to discuss strategies for controlling the effects of confounding, and (3) to illustrate the perverse effects of confounding with the help of an example.