HIV infection and breast-feeding: policy implications through a decision analysis model

(1) To develop a comprehensive decision analysis model to compare mortality associated with HIV transmission from breast-feeding with the mortality from not breast-feeding in different populations and (2) to perform sensitivity analyses to illustrate critical boundaries for guiding research and policy.

Using a decision tree, mortality rates were estimated for all children, children born to mothers infected during pregnancy, and children born to mothers who were uninfected at delivery. Given various assumptions about child mortality rates, relative risks of mortality among children who are not breast-fed compared with those who are (R), rates of HIV transmission from breast-feeding, HIV prevalence, and HIV incidence, scenarios were created and sensitivity analysis used to delineate critical boundaries.

Our model shows that only in situations where R is approximately < or = 1.5 and HIV incidence/prevalence is high (prevalence > 10%, incidence > 5%) would universal breast-feeding result in equal or higher mortality compared with non-breast-feeding. Among populations in many developing countries, where there is a high relative risk of mortality if breast-feeding is not practiced, if R > 3, overall mortality is almost always lower among children who are breast-fed, even by HIV-infected mothers. In situations where maternal HIV status is known, the decision whether to breast-feed is largely dependent on the magnitude of additional mortality risk if the child is not breast-fed. The model illustrates the importance of distinguishing between population and individual recommendations.

Based on available data, the model supports current World Health Organization and Centers for Disease Control recommendations on HIV infection and breast-feeding. Given the importance of breast-feeding and the global impact of HIV infection, more research is needed, especially to clarify the range of HIV transmission rates from breast-feeding and to expand specific assessments of relative risks for different areas of the world.