[Effect modification and interaction]

Abstract

Effect modification and interaction are terms often used in medical research. However, the analysis of effect modification and interaction has differing objectives. These two phenomena require different methods of analysing and presenting. Effect modification means that the association under study differs between certain subgroups, e.g. men and women. The aim is to identify clinically relevant subgroups, without the need to know the underlying cause of the modification. The presence and direction of effect modification may depend on the effect measure, e.g. risk difference or risk ratio. Interaction means that the combination of two determinants has more or less effect than each of the determinants separately. The aim is to estimate the causal relationship between these two determinants and the outcome as accurately as possible. Interaction can be calculated on an additive or multiplicative scale.