HIV and infant feeding: a complex issue in resource-limited settings

Abstract

In a recent issue of AIDS, Magoni et al. [1] investigated the impact of different infant feeding strategies on mother-to-child transmission of HIV within a study conducted in an urban Ugandan site. They reported a lower postnatal risk of HIV transmission at the age of 6 months in formula-fed children compared with breastfed children, and concluded that formula feeding was the safest way of feeding children born to HIV-infected mothers. We acknowledge the importance of assessing such a question within a large prospective cohort, because conducting a clinical trial allocating the infant feeding option at random was not ethically acceptable in this context. However, we would like to raise several concerns about the quality of the definition and documentation of the outcomes, the statistical methods used and the interpretation of these findings. First, to compare the efficacy of reducing HIV postnatal transmission and the safety of different infant feeding strategies, these practices have to be properly defined, precisely documented using validated tools such as recall histories, and prospectively and frequently recorded to minimize the maternal recall bias [2–4]. To interpret the results of the study fully and accurately, all of these issues need to be clarified. The infant feeding definitions used in the report were not in accordance with those recommended by the World Health Organization [5,6], the method of collection of infant feeding practices was not specified, and the long-term postnatal compliance with the infant feeding strategies chosen at delivery was not presented in detail. Infant feeding counselling strategies should also have been described and discussed. Did counselling begin antenatally? How were the infant feeding interventions proposed to the women? Were the counselling sessions organized individually or collectively, and how frequently? Second, the simple cumulative proportion used to calculate HIV transmission rates was acceptable at the age of 6 weeks but was inappropriate at the age of 6 months. Consensus approaches for the statistical methods that should be considered for analysing long-term postnatal HIV transmission rates have already been developed, such as the extension of the Kaplan–Meier procedure to interval-censored data or competing risks analysis [7,8]. Such sophisticated methods account for the fact that the exact age at infection is unknown, that the risk of postnatal infection ends at weaning, or that censoring because of death may be informative. Moreover, these methods have already been used within several studies, which allows direct comparison of the efficacy of specific interventions between studies [9,10]. Third, the evaluation of two other judgement criteria relative to infant health, i.e. infant growth and morbidity, was too cursory to be interpretable. Such a strong conclusion as ‘formula feeding was the safest way to feed children’ would thus require a much more detailed evaluation of these events. Finally, we strongly believe that choosing the endpoint for both clinical and laboratory evaluation at 6 months postnatally only was of scientific concern. In Uganda, the median duration of breastfeeding was reported to be 21 months in urban settings [11]. Proposing that women breastfeed for not more than 6 months was thus unusual in this context, and the repercussions of this strategy on infant health should be investigated in the long term, especially after the early cessation of breastfeeding. It is recommended that clinical follow-up should continue until 2 years of age to assess health outcomes fully in studies evaluating alternatives to breastfeeding in the context of HIV [12]. Moreover, it is important to diagnose HIV postnatal transmission that might occur beyond 6 months to improve the clinical care of HIV-infected children. In high HIV prevalence resource-constrained countries, HIV-infected pregnant women face a real dilemma regarding the way of feeding their forthcoming infant. The evaluation of infant feeding strategies on the reduction of postnatal mother-to-child transmission of HIV is complex with multiple judgement criteria. To be fully assessed, the balance of the risks and benefits of each feeding practice has to include the reduction of HIV postnatal transmission on one hand, and all the potential risks for the mother (stigmatization, incidence of new pregnancies) and for infant health (growth characteristics, severe morbidity, mortality) on the other hand. These assessments require long-term follow-up until 2 years of age before providing a definitive picture. As a conclusion, we feel that the study provides useful knowledge on HIV postnatal transmission according to infant feeding practices at 6 months. However, it is probably too premature to solve the breastfeeding dilemma, as some judgement criteria that need to be taken into account are not fully measured. In this sensible context, inaccurate evaluations could be prejudicial to the debate.