A situation in which the magnitude or direction of the effect of an exposure on an outcome differs across levels or categories of a third variable (the effect modifier); this heterogeneity reflects true variation in effect rather than bias.

Principle

When effect modification is present, reporting a single averaged effect can obscure meaningful differences; effect estimates should be presented stratified by modifier levels to convey heterogeneity relevant for interpretation and decision‑making.

Demonstration

Illustrative scenario → A treatment reduces disease risk by 50% in younger adults but has no benefit in older adults. Recognition → Age modifies the treatment effect. Action → Present stratified effect estimates and consider age‑specific recommendations. Consequence → Enables targeted clinical guidance and informs external validity assessments.

Misapplication

Treating effect modification as confounding and simply adjusting for the modifier to obtain a single summary estimate. That obscures real heterogeneity and can mislead about whom the exposure benefits or harms.

Consequence

Identifying effect modification guides subgroup decision‑making, personalized interventions and external validity; failure to detect it can lead to inappropriate uniform recommendations.

Reversal

Apparent modification can arise from scale choice (additive vs multiplicative): an interaction on the multiplicative scale may not exist on the additive scale and vice versa. Measurement error and sparse data can mask or create spurious effect modification.

Boundary

Clearly within: a biological pathway that produces different exposure effects by genotype. Boundary case: a subgroup difference with wide confidence intervals where evidence is inconclusive. Clearly outside: a variable that is a confounder causing bias rather than genuine effect heterogeneity.

Semantic Tension

Tension between summarizing effects for simplicity and reporting subgroup‑specific effects for clinical or policy relevance; also between statistical interaction and biological interpretation.

Synthesis

Effect modification identifies heterogeneity of causal effect across contexts or populations; it is an empirical feature to be described, not a bias to be removed, and its interpretation depends on the causal question and chosen scale.