Matching, an appealing method to avoid confounding?

Abstract

Matching is a technique used to avoid confounding in a study design. In a cohort study this is done by ensuring an equal distribution among exposed and unexposed of the variables believed to be confounding. In a matched case-control study, a case, affected by the disease, is matched with one or more individuals not affected by the disease, the controls. Because in a matched case-control study case and control group become too similar not only in the distribution of the confounder but also in the distribution of the exposure, one finds a lower effect estimate (odds ratio closer to 1). A matched case-control study requires statistical analysis to correct for this phenomenon. Nonetheless, a matched case-control study is suitable for confounders that are difficult to measure.