Using mediators to understand effect modification and interaction

Effect modification and interaction are key concepts within epidemiologic research. They refer to situations where the magnitude and/or direction of the causal effect of some exposure variable on an outcome depends on the level of a second variable (effect modification) or on the effect of a second variable (interaction). Interest in case of effect modification is primarily in one exposure variable, with its effects varying across subgroups, whereas primary interest in case of interaction is in the interplay of effects of two exposure variables. Distinctions between the concepts of effect modification and interaction are subtle. The goal of this article is to clarify these distinctions by using the concept of mediation, which focuses on elucidating mechanisms of causal effects.