When effect modification is present, how is it typically reported?

• A. As a single pooled estimate only
• B. As stratum-specific estimates or an interaction p-value
• C. Only in descriptive analyses
• D. Only when there is no confounding

Answer: (B)

Effect modification means the effect of the exposure on the outcome changes across levels of another variable. When this happens, results are typically reported by showing the effect separately within each stratum of the modifier and by testing for interaction. In practice you’d present stratum-specific estimates (for example, separate relative risks or odds ratios for men and women) and/or provide a p-value for the interaction term from a regression model to show whether the differences between strata are statistically significant. This highlights how the association varies across groups, rather than masking it with a single pooled estimate. A pooled overall estimate can be misleading when there is genuine variation by the modifier. An interaction p-value helps quantify whether the variation is unlikely to be due to chance. It’s not required that there be no confounding, and effect modification is a separate consideration from confounding.