TABLE 2—

Types of Biological Effect Modification Affecting the Generalizability of Findings From Randomized Controlled Trials

Category of Effect Modification	Description	Examples
A. Presence of other factors reduces the dose–response slope (antagonism)	Other factors are present in the target population that reduce the extent to which the intervention affects the outcome.	Iron and zinc supplementation will be less effective in places where the local diet contains substances that reduce their absorption (e.g., phytates and polyphenols).
B. Presence of other factors increases the dose–response slope (synergism)	Other factors are present in the target population that enhance the extent to which the intervention affects the outcome.	Iron supplementation will be more effective if the local diet is rich in meat and ascorbic acid, which enhance iron absorption.
C. Curvilinear dose–response association	Many biological responses are curvilinear; the same dose will have less effect if there is less need for it.	Iron supplementation will have different effects on hemoglobin according to baseline iron stores. Also, iron absorption is inversely related to iron status.
D. Limited scope for improvement in the impact (outcome) indicator because other interventions already provide protection	The intervention that is already in place acts on another link in the causal chain.	Use of insecticide-treated bed nets will have a limited effect on malaria mortality if case-management is already appropriate.
	The intervention acts on the same causal link.	Improved breastfeeding will have less effect if water supply and sanitation are adequate.
E. Intervention is inappropriate because a critical cofactor is missing	The intervention only works in the presence of another factor that is absent in the population in question.	Improving water quality will have an impact on diarrheal diseases only if water quantity is adequate.
F. Intervention is addressing a determinant that is not important	The intervention is being applied in a setting where it is not needed because the outcome it addresses has other causes.	Energy supplementation in pregnancy will have limited impact on low birthweight if the latter is mostly due to maternal smoking and to preterm deliveries caused by infections.
		The impact of improved breastfeeding on infant mortality will be lower in populations where infectious diseases account for a small proportion of deaths.