Which statement best defines effect modification (interaction) in a study?

Which statement best defines effect modification (interaction) in a study?

• A. The effect of the exposure on the outcome changes across levels of a third variable.
• B. The third variable biases the observed effect and should be controlled.
• C. There is an error in measurement of the exposure.
• D. The sampling frame excludes certain groups, biasing the results.

Answer: (A)

Effect modification, also called interaction, is when the association between an exposure and an outcome changes across different levels of another variable. In the scenario, the statement that the effect of the exposure on the outcome varies depending on the level of a third variable captures this idea directly—your observed impact is not uniform but shifts as that third factor changes. That’s what interaction looks like in analysis: the effect you observe depends on another condition or category. The other ideas describe different problems. When a third variable biases the observed effect because it’s linked to both exposure and outcome, that’s confounding—you’d want to control for it to recover the true association, but it isn’t about the effect changing across levels. Measurement error in the exposure is misclassification, a reliability issue, not about modifying the effect by another variable. Excluding groups in the sampling frame is a form of selection bias, altering who is studied rather than how the exposure effect varies by a modifier.

Effect modification, also called interaction, is when the association between an exposure and an outcome changes across different levels of another variable. In the scenario, the statement that the effect of the exposure on the outcome varies depending on the level of a third variable captures this idea directly—your observed impact is not uniform but shifts as that third factor changes. That’s what interaction looks like in analysis: the effect you observe depends on another condition or category.

The other ideas describe different problems. When a third variable biases the observed effect because it’s linked to both exposure and outcome, that’s confounding—you’d want to control for it to recover the true association, but it isn’t about the effect changing across levels. Measurement error in the exposure is misclassification, a reliability issue, not about modifying the effect by another variable. Excluding groups in the sampling frame is a form of selection bias, altering who is studied rather than how the exposure effect varies by a modifier.