What is effect modification in a cohort study?

• A. When the exposure effect is the same across levels of a third variable.
• B. When the exposure effect differs by levels of a third variable.
• C. When there is random variation in the data.
• D. When the exposure is confounded by a measured variable.

Answer: (B)

Effect modification, also called interaction, happens when the association between an exposure and an outcome changes across levels of another variable. In a cohort study, this means the effect size (like the risk or risk ratio) of the exposure on the outcome is different for different subgroups defined by the third variable. For example, the exposure might increase risk more in older individuals than in younger ones, or differ between men and women. This is a real difference in effect across strata, not just a bias.

That’s why the best choice says the exposure effect differs by levels of a third variable. The other ideas describe no modification, random variation, or confounding, which are distinct concepts: random variation is sampling noise, and confounding is a variable associated with both exposure and outcome that distorts the observed association, not a subdivision where the effect itself changes.