Background: Data comparing survival of formula-fed to breast-fed infants in programmatic settings are limited. We compared mortality and HIV-free of breast and formula-fed infants born to HIV-positive mothers in a program in rural, Rakai District Uganda. Methodology/Principal Findings: One hundred eighty two infants born to HIV-positive mothers were followed at one, six and twelve months postpartum. Mothers were given infant-feeding counseling and allowed to make informed choices as to whether to formula-feed or breast-feed. Eligible mothers and infants received antiretroviral therapy (ART) if indicated. Mothers and their newborns received prophylaxis for prevention of mother-to-child HIV transmission (pMTCT) if they were not receiving ART. Infant HIV infection was detected by PCR (Roche Amplicor 1.5) during the follow-up visits. Kaplan Meier time-to-event methods were used to compare mortality and HIV-free survival. The adjusted hazard ratio (Adjusted HR) of infant HIV-free survival was estimated by Cox regression. Seventy-five infants (41%) were formula-fed while 107 (59%) were breast-fed. Exclusive breast-feeding was practiced by only 25% of breast-feeding women at one month postpartum. The cumulative 12-month probability of infant mortality was 18% (95% CI = 11%-29%) among the formula-fed compared to 3% (95% CI = 1%-9%) among the breast-fed infants (unadjusted hazard ratio (HR) = 6.1(95% CI = 1.7-21.4, P-value < 0.01). There were no statistically significant differentials in HIV-free survival by feeding choice (86% in the formula-fed compared to 96% in breast-fed group (Adjusted RH = 2.8 [95%CI = 0.67-11.7, P-value = 0.16] Conclusions/Significance: Formula-feeding was associated with a higher risk of infant mortality than breastfeeding in this rural population. Our findings suggest that formula-feeding should be discouraged in similar African settings.
The objective of this programmatic evaluation was to compare mortality and HIV-free survival among formula-fed and breast fed infants born to HIV-infected mothers, who self-selected their preferred method of infant feeding following counseling on prevention of breast milk HIV transmission. The data were derived from the ARV-related Maternal-Infant Study (ARMIS)-an offshoot of the Rakai Community Cohort Study (RCCS). The RCCS maintains annual HIV surveillance in a cohort of over 12,000 adults aged 15–49 in 50 villages in rural Rakai district. The cohort monitored HIV/STD incidence and prevalence as well associated behavioral and demographic determinants. Household demographic data were collected though an annual census and included information on household possessions (used as a proxy for social-economic status), access to water and electricity, and possession of sanitary facilities such as toilets/latrines. During the annual surveys women in this cohort were also screened for pregnancy by menstrual history, physical examination, and urinary hCG pregnancy tests. Consenting cohort participants were screened for HIV infection at each annual visit. From 2005, all consenting pregnant HIV-infected mothers from RCCS, and their newly born infants, were enrolled in a separate cohort study called “ARV-related Maternal-Infant Study (ARMIS)”, which assessed the effects of availability of antiretroviral therapy (ART) on health outcomes of pregnant women and their infants. Pregnant HIV-positive women identified through RCCS were visited at home by midwives who were also trained HIV-counselors (“midwife counselors or MWCs”). Pregnant women were asked to notify the MWCs (by cell phone or through a messenger) as soon as they had delivered. Mother-infant pairs were followed-up at home by the MWCs soon after notification of birth (usually within forty eight hours), then at one, six and twelve months after birth. Structured questionnaires were used to collect information on maternal and infant-morbidity and mortality as well as infant feeding. Heel-prick infant blood was collected for HIV PCR at all follow-up visits starting at one month. Infant HIV was not determined at birth. Mothers were offered their infant's HIV results and counseling on care and infant feeding. Pregnant women were offered prenatal voluntary HIV counseling and testing (VCT), antenatal care, hematinics, multivitamins, and presumptive malaria prophylaxis using Fansidar (Sulphamethoxazole + pyrimethamine) as recommended by the Ugandan Ministry of Health (MOH). They were also offered the single dose Nevirapine (sdNVP) for pMTCT using procedures described previously [7]. After September 2007, prophylaxis was changed to a combination antiretroviral regimen of AZT starting at 28 weeks gestation, in addition to sdNVP and 3TC given at the start of labor with a 7-day tail of 3TC and AZT postpartum for the mothers, and AZT syrup for the infant as per WHO recommendation [8]. Women who had WHO clinical stage 4 disease or CD4 counts less than or equal to 250 cells/ul were offered free antiretroviral therapy (ART) through a community-based program. They also received basic HIV care that included routine cotrimoxazole prophylaxis, treatment of opportunistic infections, insecticide treated bed nets for prevention of malaria and safe water vessels with hypochlorite for prevention of intestinal infections. HIV-infected infants were offered cotrimoxazole prophylaxis starting at six weeks and ART if they met the WHO eligibility criteria for initiation of antiretroviral therapy. Mothers and their Infants were symptomatically treated at home for simple health problems or referred to the government health units for follow-up care and for treatment that could not be provided by the MWCs at home. Government and private health care facilities are available in Rakai, but services are limited and most clinics accessible to rural communities (<5 Km) are staffed by non-physicians. Services are frequently interrupted by absence of personnel and stock outages of essential medicines. As part of the pMTCT service program, mothers were given infant feeding counseling for prevention of breast milk HIV transmission by the MWCs at the pre-natal visit and then provided follow-up counseling at other scheduled study visits. They were then allowed to make an informed choice between breast-feeding or to use the free formula (NAN from Nestle) provided by the pMTCT service program. Mothers who chose to breast-feed, were encouraged to breast-feed exclusively for six months and then wean their infants thereafter. Exclusive breast feeding was defined as breast feeding with no added supplements, and mixed breast-feeding was defined as breast-feeding with additional supplementary feeds (excluding medications). Mothers who chose to use formula were trained in hygienic preparation of feeds, measurement of the correct amounts of formula appropriate for the child's age and were provided with a free cup, spoon as well as a vacuum thermos flask for storage of night feeds. Mothers were discouraged from using infant feeding bottles as these were deemed difficult to keep clean. Mothers were encouraged to wash utensils with soap and water and to boil the utensils after washing. Re-use of feeding utensils without washing was strongly discouraged. Mothers were asked to only use freshly prepared formula and to discard any leftover. During the consent process for ARMIS, women were informed that their service-based data could be linked to data they provided through ARMIS. Therefore all women used for this evaluation provided consent for use of their data from the pMTCT service program. The RCCS and ARMIS were approved by the Science and Ethics Committee of Uganda Virus Research Institute, the Uganda National Council of Science and Technology and US-based Western IRB Only infants alive at birth were included in this analysis. In the case of multiple births, only the first born twin was included. The feeding option practiced by mothers at the first postpartum visit (usually within 48 hours after birth) was used to classify infants as either breast-feeding or formula-feeding. Maternal and Infant baseline characteristics were compared between the feeding groups. The Mann-Whitney U test was used for continuous variables while Fischer's exact test was used for categorical variable. Infant mortality and the composite outcome of mortality or HIV infection (i.e., the complement of HIV-free survival) in the two feeding groups were compared using Kaplan-Meier time-to-event methods. Multivariable Cox proportional hazards regression was used to estimate adjusted hazard ratios (adj HR) and 95% confidence intervals (95%CI) of infant mortality or HIV-infection, adjusting for maternal age, and use of ART as a time-varying covariate. CD4 counts were available for 84% of formula feeding mothers and only 54% of breast feeding mothers, so adjustment for maternal CD4 counts omitted many observations and resulted in poor model fit. However, receipt of ART was contingent on a CD4 count <250 so low CD4 counts were strongly negatively correlated with receipt of ART, and models adjusting for ART provided a better model fit. Therefore, in the final model, receipt of ART was used for adjustment instead of CD4 counts. Censoring for the mortality outcome occurred due to loss to follow-up at the visit when this was first noted. For the composite outcome of mortality and HIV-infection, censoring occurred due to loss to follow-up and absence of an HIV-result. Because we could not distinguish between in utero/peripartum HIV-infection and early breast-milk transmission at one month, all mother-child pairs in which transmission was detected at one month were left censored. One observation with a missing HIV test at one month where the first HIV-positive result was detected at six was also censored. The proportional hazards assumption was tested using Schoenfeld and scaled Schoenfeld residuals[9]. Model goodness-of-fit was tested using graphical methods based on Cox-Snell residuals[10]. Statistical analysis used STATA software (Release 9.2. Stata Corporation, College Station, Texas, USA).
N/A