What is the difference between effect modification and confounding?

• A. Effect modification occurs when the effect of exposure varies by a third variable; confounding biases estimates.
• B. Both describe biases in estimates equally; there is no distinction.
• C. Confounding is when the exposure varies by a third variable; effect modification biases estimates.
• D. There is no difference; they are the same.

Answer: (D)

Confounding is a bias that distorts the apparent relationship between exposure and outcome. It happens when a third variable is linked to both the exposure and the outcome and is not on the causal path between them. This can make an association look stronger, weaker, or even in the opposite direction. You can address confounding by design (randomization, restriction, matching) or by analysis (adjusting for the confounder in a multivariable model).

Effect modification, or interaction, is not a bias. It reflects a real difference in how the exposure affects the outcome across levels of another variable. In other words, the effect of the exposure changes depending on the level of the modifier. This is shown by different effect estimates in different strata and is often explored by stratified analyses or by including an interaction term in a model.

So they are different: confounding distorts the true effect, while effect modification reveals that the effect itself varies across subgroups.