Evidence-appraisal glossary
Effect modification
Effect modification happens when the size or direction of a treatment or exposure's effect on an outcome genuinely differs across subgroups, defined by a third variable such as age or sex. Unlike a bias, it is a real finding to report, not remove, and analysts detect it by stratifying or testing interaction.
Also called: Effect measure modification, Statistical interaction, Treatment effect heterogeneity.
What it is. Effect modification (often called effect measure modification, or statistical interaction) means the effect of an exposure or treatment is not the same in everyone. A drug might cut risk sharply in younger patients but barely move it in older ones. The variable that splits the picture, here age, is the effect modifier.
Why it differs from confounding. Confounding is a distortion you want to remove; effect modification is a genuine biological or clinical signal you want to describe. With confounding, adjusting collapses the subgroups toward one truth. With effect modification, the subgroups really are different, and averaging them hides that.
How to use it when reading a study. Look for subgroup analyses, stratified estimates, or a reported interaction test. Ask whether modification was pre-specified and whether the interaction test (not just separate p-values per subgroup) supports it. Note that it is scale-dependent: an effect can be uniform on the relative scale yet vary on the absolute scale.
This is a plain-language methodology definition for reading research. It is general education, not medical advice.