In the HIV era, is breast always best? Response to ‘Long-term infant health outcomes critical in feeding decisions by HIV-positive mothers’ by Crowe et al.

Abstract

We thank Drs Crowe and colleagues for their comments. The authors state that our paper claims that ‘formula feeding is better than breastfeeding for infants of HIV-positive mothers in less developed countries’ because ‘babies of mothers who chose to formula feed because of preceding morbidity in either mother or infant had lower mortality rates than babies of mothers who chose to formula feed for other reasons'. This is incorrect, we clearly showed that mortality associated with formula feeding as a result of preceding morbidity is higher than formula feeding for voluntary reasons. We repeatedly reported higher child mortality rates in children who were never breastfed, compared with children who had been weaned. Our fundamental point is that the reason for not breastfeeding or weaning is important because mortality was consistently higher among children who were never breastfed or were weaned because of a preceding illness. There is clear evidence of reverse causality, whereby pre-existing morbidity may cause mothers not to breastfeed or to wean, and the resultant higher mortality risk arises from the preceding illness, rather than the mode of feeding per se. Such self-selection and reverse causality is also evident in other studies. The authors also suggested confusion as a result of conflation between previous infant with maternal illness, and a combination of data across countries at different levels of development and healthcare access. The majority of reasons for not breastfeeding or weaning were a result of the child's preceding illness. Omitting cases of maternal illness did not change our findings. The country-specific estimates demonstrated similar patterns, and were therefore combined to get an aggregate measure. They suggest that we disaggregate the excess mortality into components such as health access, as well as formula-related risks. Unfortunately, the DHHS survey data do not provide this information and we have stated this as a limitation of the study. However, it is striking that the findings were consistent in all countries examined, which indicates a consistent effect, irrespective of the level of development or healthcare. They also suggest that maternal immune status, micronutrient deficiencies and maternal symptomatology could increase the risk to the infant via maternal morbidity. We cannot speculate on this question because no data are available. They also recommend that we differentiate between exclusive and partial breastfeeding, but again, DHHS data are not available to address this issue. The authors also question whether our findings can be applied to children of HIV-infected mothers who voluntarily decide not to breastfeed. We suggest that counselling HIV-positive mothers on methods of feeding should recognize that current child mortality estimates for non-breastfeeding are biased overestimates, because they do not adjust for the reverse causality described above. Although breastfeeding results in lower infant and child mortality rates than formula feeding in non-HIV-infected women, this risk/benefit is reversed with HIV because of the high rates of mother-to-child HIV transmission via breast milk. Although we agree with the authors that a more specific definition of breastfeeding (i.e. mixed or exclusive) is important, this information is not available in cross-sectional surveys such as the DHHS and we suggest that prospective studies are needed on methods of feeding and reasons for weaning or the non-initiation of breastfeeding. Finally, it is argued that it is invalid to compare rates of HIV infection, which have long-term consequences for survival with the possible adverse effects of formula feeding, which are likely to be short term. We profoundly disagree with this view. The median survival time of HIV-infected children in Rakai, Uganda, is 2 years [1]. In the absence of antiretroviral therapy, all pediatric HIV infections are likely to be fatal. Given the high rates of mother-to-child transmission via breast milk, and the unconfirmed efficacy of exclusive breastfeeding for the prevention of mother-to-child transmission, based on one observational study [2], we believe that it is only reasonable to offer women the alternative of not breastfeeding in order to select the safest method for feeding infants.