What is the risk of overmatching and when should matching be avoided?

• A. Overmatching can obscure true associations; avoid matching on variables that are consequences of exposure or on variables in the causal pathway.
• B. Overmatching increases the likelihood of bias toward the null.
• C. Overmatching is only a concern in cross-sectional studies.
• D. Overmatching eliminates confounding completely.

Answer: (A)

The key idea here is that matching should target true confounders, not variables that lie on the causal path between exposure and outcome. When you overmatch by pairing or restricting on a variable that is a consequence of the exposure or sits on the causal pathway, you block part of the very mechanism you’re trying to study. This can make the exposure’s effect appear weaker or even vanish because the comparison is effectively held constant for processes the exposure would normally influence. In practice, this means you reduce the study’s ability to detect a real association and may bias the estimated effect toward the null.

So you should avoid matching on mediators or downstream consequences of the exposure, and focus on true confounders that precede both exposure and outcome. Matching on such confounders can help control bias without distorting the causal effect. It's not true that overmatching eliminates confounding completely or that it’s only a concern in one study design; and while matching can reduce bias, it does not guarantee complete control.