Postnatal transmission of HIV, predominantly through breastfeeding, accounts for approximately half of all mother-to-child transmission in situations where breastfeeding is prolonged . Elimination of the risk of postnatal transmission is possible through exclusive replacement feeding. The risks of transmission through breastmilk need to be balanced, however, against the risks associated with replacement feeding in resource-poor contexts. There are few operational data on the risks of replacement feeding in the context of prevention of mother-to-child transmission (PMTCT) but it is understood that these risks vary depending on individual and environmental circumstances.
One randomized controlled trial conducted in an urban PMTCT setting in Kenya found that infants in the formula-feeding group, whose mothers had access to clean water, free formula and frequent support by health workers, had a 40% lower risk of HIV transmission but their 24-month mortality was similar to that in the breastfed group . The Mashi study  undertaken in Botswana also found increased mortality amongst infants who were formula fed in the context of provision of free formula milk. The 18-month HIV-free survival was similar in the breastfeeding and formula-feeding groups.
The DITRAME PLUS study  conducted in Côte d'Ivoire found a more positive outcome with formula feeding in an urban setting with high levels of support. Women could choose to formula feed their infants or exclusively breastfeed with early cessation. Formula-feeding women were provided with free replacement feeds and supplies for 9 months. This study found no significant difference in rates of infant illness and death at 24 months between breastfed and formula-fed infants suggesting that safe formula feeding can be achieved in settings where women have access to clean water, free healthcare, free transport to health facilities, free formula milk and formula-feeding supplies.
The World Health Organization (WHO)/UNICEF  recommendations for HIV-positive women is avoidance of all breastfeeding only if replacement feeding is acceptable, feasible, affordable, sustainable and safe. Otherwise exclusive breastfeeding for the first 6 months of life is recommended.
The translation of the WHO/UNICEF recommendation into operational settings is a challenge for health workers and counsellors as there is little guidance on what the terms ‘acceptable’, ‘sustainable’, ‘safe’ and ‘feasible’ mean in practice. PMTCT programmes are being scaled up across many countries in Africa and the responsibility of assisting HIV-positive women to make an infant-feeding choice rests with nurses and lay counsellors . Evidence suggests that the quality of infant-feeding counselling is poor, especially assessments of home circumstances . Little is known about the long-term risks of using commercial infant formula to feed HIV-exposed infants in PMTCT programmes in resource-poor contexts; similarly the excess risk of breastfeeding under conditions where formula feeding would be an appropriate choice have not been described.
Very little research has focused on the implementation of the WHO/UNICEF guidelines in operational settings and the consequences for infant outcomes. This report is of a prospective cohort study that followed HIV-positive mothers and their infants in order to assess vertical HIV transmission. The purpose was to identify key individual and environmental criteria that could be used to guide appropriate infant-feeding choices in operational settings, and to assess the effect of inappropriate feeding choices on infant HIV-free survival.
This prospective cohort study followed mother–infant pairs in three national PMTCT sites in South Africa: Paarl (Western Cape province), a periurban area; Rietvlei (Eastern Cape province), a rural area; and Umlazi (KwaZulu-Natal province), a periurban township. The antenatal prevalence rates were 15%, 29% and 39%, respectively .
The study recruited women from October 2002 to September 2003 from those attending one of the selected antenatal clinics in the study sites. Women were eligible if they were aged at least 16 years, agreed to HIV counselling and testing and had a positive HIV test result. Women were recruited from 28 weeks of pregnancy following an explanation about the study. Signed informed consent was obtained at enrolment, with verbal informed consent at each home visit. Ethical approval for this study was granted from the University of KwaZulu-Natal.
The PMTCT programme at these sites consisted of antenatal counselling and HIV testing, infant-feeding counselling once or twice antenatally, free infant formula to women who chose not to breastfeed and single-dose nevirapine for mother and infant. The infant-feeding counselling was based on the WHO/UNICEF recommendations that women should be given information about the risks of different feeding options and then helped to make an appropriate choice.
Recruitment over a period of 12 months yielded a total sample size of 635 HIV-positive mother–infant pairs: 149 from Paarl, 164 from Rietvlei and 322 from Umlazi. Loss to follow up at 36 weeks was 19% and was not significantly different between women who intended to breastfeed and women who intended to formula feed, yielding a final sample size of 514. Exclusion criteria were multiple births, congenital malformations and newborn illness requiring admission to intensive care.
Data were collected by trained field researchers in the participants' homes antenatally, and then postnatally every 2 weeks until 12 weeks and then monthly until 36 weeks through semistructured interviews with the infant's mother or primary caregiver. Infant-feeding intentions were captured prior to discharge from the postnatal ward. At each scheduled visit, infant-feeding practices were recorded through recall of the previous 24 h and 3 days. Infants were classified at each scheduled visit as exclusively or predominantly breastfed, mixed fed, or replacement fed using these recall histories. The following WHO definitions for infant-feeding practices  were used. Exclusive breastfeeding was defined as giving a child no other food or drink, including water, in addition to breastfeeding with the exception of medicines, vitamin drops or syrups, and mineral supplements. Predominant breastfeeding was defined as breastfeeding a child but also giving small amounts of water or water-based drinks. Neither food-based fluid nor solid food was allowed under this definition. Replacement feeding was defined as feeding a child on artificial feeds (including infant formula or powdered animal milk) and not breastfeeding at all. Mixed feeding was defined as breastfeeding while giving nonhuman milk such as infant formula or food-based fluid or solid food.
The field researchers collecting this information were not involved in infant-feeding counselling. Data on infant mortality was captured through a verbal autopsy questionnaire with the family.
Criteria deemed to constitute appropriateness of formula use based on the WHO/UNICEF guidelines (acceptable, feasible, affordable, sustainable and safe) were identified as piped water in the house or yard (safety); electricity, gas or paraffin for cooking fuel (feasibility); disclosure of HIV status by 3 weeks after birth (acceptability); having someone in the household employed (affordability and sustainability); and access to a fridge for storage of prepared formula (safety). This information was collected during the first postnatal visit at 3 weeks.
Infant HIV status was determined by HIV-1 DNA polymerase chain reaction (Amplicor version 1.5; Roche Molecular Systems, Branchburg, New Jersey, USA) from blood spots collected at 3, 24 and 36 weeks . Maternal HIV status was obtained from routine PMTCT medical records and maternal viral load was determined by quantitative HIV-1 RNA NASBA (Nuclisens ECL, bioMérieux, Boxtel, the Netherlands) at 3 weeks postpartum. Results of infant tests were offered to mothers following the 36 week visit.
HIV-free survival was chosen as the primary outcome for this analysis (the composite end-point of HIV transmission or death was used) as recommended in the literature for assessment of PMTCT effectiveness . Infants who were already HIV infected at 3 weeks were excluded from the analysis in order to assess the effect of infant-feeding choices on postnatal HIV transmission. Adherence to formula feeding was a secondary outcome and was defined as never introducing breastmilk amongst women who intended to formula feed.
Baseline characteristics were compared between women intending to formula feed and women intending to breastfeed using χ2 tests for categorical variables and one-way analysis of variance for continuous variables with the exception of infant mortality and postnatal transmission rates. These were estimated using the Kaplan–Meier method (thus adjusting for loss to follow-up and early death) with infection time taken to be the midpoint of the last negative and first positive tests. Log rank was used for testing statistical significance.
In order to determine which criteria should be included in a measure of appropriateness for infant-feeding choices according to the WHO/UNICEF guideline, Cox's proportional hazards regression models were run to determine which of the following individually predicted infant HIV-free survival amongst women who intended to formula feed: availability of piped water, availability of fuel, HIV status disclosure to anyone by 3 weeks' postpartum, access to a fridge and household employment. Efron's method was used to adjust for tied survival times (since there were a high proportion of tied survival times given only three testing points).
These selected criteria were subsequently examined in scores in a stepwise manner beginning with piped water alone (score A), and adding one criteria at a time to reach a score containing all five criteria (score E). In order to determine predictors of formula adherence, using ‘ever breastfeeding’ as a marker of nonadherence, a second analysis was done using logistic regression to determine which of the above criteria were predictors of adherence to exclusive formula feeding. A balance between protective effect (i.e., which criteria best predicted HIV-free survival and formula adherence) and not being too restrictive (i.e., having criteria that few women would meet) was sought in the selection of appropriate criteria for formula feeding.
Potential confounding factors that were examined included maternal age, parity, maternal education, maternal viral load, site and household asset score using a data-based approach [i.e., greater than 10% change in between-group hazard ratios (HR)] . Maternal viral load was found to play a confounding role for both outcomes, and adjusted HR and adjusted odds ratios (OR) are presented. Stata 9.2 (Stata Corp, College Station, Texas, USA) was used for data management and analysis.
Baseline study population characteristics
Among the 635 HIV-positive women who were included in the study and delivered a live infant, 289 (45.5%) were intending to formula feed exclusively, 311 (48.9%) were intending to breastfeed exclusively and 35 (5.7%) were intending to mix feed or were still undecided.
Characteristics of women intending to formula feed exclusively and those intending to breastfeed exclusively were similar except that women intending to formula feed had a significantly higher mean viral load, fewer antenatal clinic visits and a greater percentage were married (Table 1). With regard to actual feeding practices, only 13% of women who intended to breastfeed were exclusively breastfeeding at 12 weeks; the rest were mixed feeding (42%), predominantly breastfeeding (11%) or exclusively formula feeding (33.5%). Amongst women who intended to formula feed, 29% gave breastmilk at some point between birth and 36 weeks. Reliability of formula supplies was found to be poor, with 73% of women who intended to formula feed reporting that they had been to the clinic at least once between birth and 36 weeks and found that the clinic did not have formula stock. The most commonly given fluids during these times were breastmilk, purchased formula and sugar water.
These suboptimal feeding practices led to similar rates of postnatal transmission between women intending to formula feed exclusively (12%) and women intending to breastfeed exclusively (14%). Mortality rates were also not significantly different between the two feeding groups. The most common cause of infant death was infectious diseases.
With regard to the criteria considered for appropriate formula choice, despite the high percentage of mothers receiving counselling and frequent antenatal care attendance, infant-feeding intentions did not appear to be influenced by socioeconomic and environmental conditions such as household assets, water source and fuel type. There was, however, a significantly higher rate of early disclosure amongst women intending to formula feed (Table 1).
Defining a measure of appropriate choice
The proportion of women intending to formula feed who met the potential criteria for appropriate formula feeding decreased as the scores of appropriateness became stricter: from 53% with one criterion (score A) to only 15% with five criteria (score E) (Table 2). The outcomes for the group with three criteria of appropriateness was better than the group with one criterion: having piped water, fuel and disclosure of HIV status (score C) was substantially more protective in terms of HIV transmission or death compared with having just piped water (score A): HR 0.26 [95% confidence interval (CI), 0.09–0.75] and 0.68 (95% CI, 0.32–1.43), respectively.
Adherence to exclusive formula feeding was also better with three criteria; the odds of ever breastfeeding if the mother had piped water alone (score A) was 0.52 compared with 0.49 if the mother had three criteria (score C).
Examining the percentage of women who had the criteria, infants with HIV-free survival, and feeding choice adherence, score C (piped water in the house or yard, electricity, gas or paraffin for cooking fuel; and HIV status disclosure) appeared to be the best measure of appropriateness. Adding a fridge (score D) and employment (score E) did not result in any gains in terms of predicting HIV-free survival or adherence. The OR values for ever breastfeeding actually increased in these stricter scores, possibly through instability of estimates based on small numbers. One-third of women met the criteria for score C, decreasing to 24% and 15% for the more restrictive scores D and E, respectively.
Appropriateness of choice and infant HIV-free survival
Women were next classified into four groups based on the presence or absence of all three criteria in score C: (a) met criteria and chose to formula feed, (b) did not meet criteria and chose to breastfeed, (c) did not meet criteria and chose to formula feed, and (d) met criteria and chose to breastfeed. Based on this classification, amongst women who intended to formula feed, 195/289 (67.4%) women did not meet the criteria and amongst women who intended to breastfeed 95/311 (30.5%) women met the criteria for appropriate formula feeding.
There were significantly different risks of HIV transmission or death between these four groups (Table 3). Infants of women who chose to formula feed and did not meet the criteria had the highest risk of HIV transmission or death (adjusted HR, 3.63; 95% CI, 1.48–8.89). Similarly, infants of women who met the criteria but chose to breastfeed also had a greater risk of HIV transmission or death (adjusted HR, 3.35; 95% CI, 1.25–8.96) compared with appropriate formula feeders. The risk of HIV transmission or death compared with appropriate formula feeding was similar across the two breastfeeding groups (adjusted HR, 3.35 and 3.33, respectively).
To our knowledge, this study is the first attempt to provide a practical definition for the WHO/UNICEF infant-feeding guidelines for HIV-positive women. Using three criteria of appropriateness (piped water, electricity, gas or paraffin for cooking fuel, and early disclosure of HIV status), we found inappropriate choices being made in both directions. Almost a third of women who chose to breastfeed did have these three basic conditions suitable for formula use; and two thirds of women who chose to formula feed did not have the selected conditions for safe formula use. Both of these groups had a more than three-fold increased risk of HIV transmission or death compared with appropriate formula feeders. Infants of mothers who met the three criteria and chose to formula feed had the best outcome in terms of HIV-free survival. The risk of HIV transmission or death in both breastfeeding groups was high and reflected the low rate of exclusive breastfeeding.
Inappropriate infant-feeding choices amongst HIV-positive women are fairly widespread, especially in contexts where formula milk is provided free and the quality of counselling is poor. An evaluation  of the PMTCT pilot programme in South Africa found high levels of intentions to formula feed in rural areas (67%) despite the majority of households not having access to the required basic conditions. A UNICEF review  of experiences with free formula milk found that there was very high uptake of formula amongst HIV-positive women, with rates around 60% in Uganda and Zambia and 87% in Rwanda.
Excess mortality risks from formula feeding have been described in research settings in Africa. In a cohort study of HIV-infected infants in Nairobi , formula feeding was found to be a predictor of early infant mortality. The recently completed Mashi study  undertaken in Botswana also found increased mortality amongst infants who were formula fed in the context of free formula milk being provided. The 18-month HIV-free survival was similar in the breastfeeding and formula-feeding groups, suggesting that the gains in preventing postnatal transmission from formula feeding were lost owing to increased mortality.
Hildebrand et al.  reported encouraging findings from a large-scale PMTCT programme in Cape Town where many women were managing to formula feed safely. We also found better outcomes amongst women intending to formula feed who had appropriate conditions.
South Africa has a gross domestic product of over US$4000 per head and it should be a human rights priority that all women have access to clean water, fuel and conditions amenable to disclosure of HIV status. However, where this is not possible, decisions should be guided by assessment of personal and environmental circumstances for each mother, especially the availability of piped water, fuel and HIV status disclosure. These results suggest that where conditions for appropriate formula feeding cannot be met, women should be supported to breastfeed, as inappropriate formula feeding appears to carry the greatest risk of HIV transmission or death. Amongst those women who breastfeed, research suggests that exclusive breastfeeding should be encouraged and supported [17–19]. It is clear from this research that sustaining exclusive breastfeeding is extremely difficult without a supportive infrastructure, and previous research in these sites has described the quality of the infant-feeding counselling to be poor . Several studies, including one in South Africa, have found that community-based support can lead to increased rates of exclusive breastfeeding [19–21] and such a strategy could be considered as a complement to facility-based counselling to sustain exclusive breastfeeding.
Assisting HIV-positive women to make an appropriate infant-feeding choice is an onerous task that requires complex information to be given regarding competing risks. The currently available guidelines need to be simplified and combined with practical tools and simple assessment algorithms that counsellors can use to assess individual and environmental characteristics known to alter the risks of feeding choices.
Our study had several potential limitations. Women's intended feeding practices were used to determine the outcomes of infants based on certain personal and environmental criteria. The actual feeding practices of these women did change over time from their intentions and the rates of adherence to the chosen feeding method were low in both groups, but particularly amongst women intending to breastfeed exclusively. The sites where this research was undertaken were purposely selected and the infrastructural conditions differed greatly between them. They do, however, reflect the most common settings in South Africa, namely urban, periurban and rural areas. Participants were recruited from amongst all attendees at the selected antenatal facilities with no other selection criteria except a positive HIV test and a singleton birth without severe newborn illness. Infants were tested for HIV infection at 3–4 weeks rather than at 6 weeks, as is more commonly done. A concern regarding testing at this earlier age is that some cases of intrapartum HIV transmission may be wrongly attributed to postnatal transmission. However, a meta-analysis of published data from 271 infected children indicated that the HIV DNA assay was highly sensitive and that by 14 days after infection 93% (95% CI, 76–97) of infected children were positive by this method . By 28 days after delivery, this had risen to 96% sensitivity and 99% specificity to identify HIV proviral DNA in peripheral blood mononuclear cells. We are confident, therefore, that we did not misclassify more than a maximum of 5% of intrapartum infections as postnatal infections.
This study suggests that infant HIV-free survival could be improved if HIV-positive women who choose to formula feed fulfilled at least three criteria (piped water, fuel and have disclosed their HIV status) and that without these criteria a choice to breastfeed exclusively would be more appropriate. Public health policy needs to balance the desire to reduce HIV infection in infants with the need to ensure child survival. Given the increasing burden of HIV, with concomitant rising infant mortality rates, governments in countries with a high HIV burden need to strive to improve HIV-positive women's access to basic resources that would allow for appropriate formula feeding. Greater investment is also needed in supporting the role of infant-feeding counsellors through simplified, structured counselling tools combined with the provision of ongoing infant-feeding support in the postnatal period.
We acknowledge the field staff who worked tirelessly to track mothers and infants in this study.
Sponsorship: This study was supported by the South African National Department of Health, the US Centers for Disease Control and Prevention South Africa, UNICEF and the Swedish–South Africa Cooperation (SIDA/NRF).
1. Newell ML. Current issues in the prevention of mother-to-child transmission of HIV-1 infection. Trans R Soc Trop Med Hyg 2006; 100:1–5.
2. Mbori-Ngacha D, Nduati R, John G, Reilly M, Richardson B, Mwatha A, et al
. Morbidity and mortality in breastfed and formula-fed infants of HIV-1-infected women: A randomized clinical trial. JAMA 2001; 286:2413–2420.
3. Thior I, Lockman S, Smeaton LM, Shapiro RL, Wester C, Heymann SJ, et al
. Breastfeeding plus infant zidovudine prophylaxis for 6 months vs formula feeding plus infant zidovudine for 1 month to reduce mother-to-child HIV transmission in Botswana: a randomized trial: the Mashi Study. JAMA 2006; 296:794–805.
4. Becquet R, Ekouevi DK, Sakarovitch C, Viho I, Bequet L, Fassinou P, et al. Two-year morbidity and mortality in breastfed and formula fed children born to HIV-infected mothers, ANRS 1201/1202 ditrame plus, Abidjan, Cote d'Ivoire
. XVI International AIDS Conference
. Toronto, August 2006 [abstract TUPE0350].
5. World Health Organization. New Data on the Prevention of Mother-to-child Transmission of HIV and Their Policy Implications: Conclusions and Recommendations
. Geneva: UNAIDS/UNFPA/UNICEF/WHO; 000.
6. Doherty T, Chopra M, Nkonki L, Jackson D, Greiner T. Effect of the HIV epidemic on infant feeding in South Africa: ‘When they see me coming with the tins they laugh at me’. Bull World Health Organ 2006; 84:90–96.
7. Chopra M, Doherty T, Jackson D, Ashworth A. Preventing HIV transmission to children: quality of counselling of mothers in South Africa. Acta Paediatr 2005; 94:357–363.
8. National Department of Health. National HIV and Syphilis Antenatal Sero-prevalence Survey 2005
. Pretoria: South African Department of Health; 2006.
9. World Health Organization. Indicators for Assessing Breast-feeding Practices: Report of an Informal Meeting
. Geneva: World Health Organization; 1991.
10. Sherman GG, Stevens G, Jones SA, Horsfield P, Stevens WS. Dried blood spots improve access to HIV diagnosis and care for infants in low-resource settings. J Acquir Immune Defic Syndr 2005; 38:615–617.
11. Alioum A, Dabis F, Dequae-Merchadou L, Haverkamp G, Hudgens M, Hughes J, et al
. Estimating the efficacy of interventions to prevent mother-to-child transmission of HIV in breast-feeding populations: development of a consensus methodology. Stat Med 2001; 20:3539–3556.
12. Rothman KJ, Greenland S. Modern Epidemiology. 2nd edn. Philadelphia, PA: Lippincott-Raven; 1998.
13. Doherty TM, McCoy D, Donohue S. Health system constraints to optimal coverage of the prevention of mother-to-child HIV transmission programme in South Africa: lessons from the implementation of the national pilot programme. Afr Health Sci 2005; 5:213–218.
14. AED/Linkages/UNICEF. Review of UNICEF Experience with the Distribution of Free Infant Formula for Infants of HIV Infected Mothers in Africa
. Washington, DC: AED/Linkages/UNICEF; 2004.
15. Obimbo EM, Mbori-Ngacha DA, Ochieng JO, Richardson BA, Otieno PA, Bosire R, et al
. Predictors of early mortality in a cohort of human immunodeficiency virus type 1-infected african children. Pediatr Infect Dis J 2004; 23:536–543.
16. Hilderbrand K, Goemaere E, Coetzee D. The prevention of mother-to-child HIV transmission programme and infant feeding practices. S Afr Med J 2003; 93:779–781.
17. Coutsoudis A, Pillay K, Spooner E, Kuhn L, Coovadia HM. Influence of infant-feeding patterns on early mother-to-child transmission of HIV-1 in Durban. South Africa: a prospective cohort study. South African Vitamin A Study Group. Lancet 1999; 354:471–476.
18. Iliff PJ, Piwoz EG, Tavengwa NV, Zunguza CD, Marinda ET, Nathoo KJ, et al
. Early exclusive breastfeeding reduces the risk of postnatal HIV-1 transmission and increases HIV-free survival. AIDS 2005; 19:699–708.
19. Coovadia HM, Rollins NC, Bland RM, Little K, Coutsoudis A, Bennish ML, et al
. Mother-to-child transmission of HIV-1 infection during exclusive breastfeeding in the first 6 months of life: an intervention cohort study. Lancet 2007; 369:1107–1116.
20. Bhandari N, Bahl R, Mazumdar S, Martines J, Black RE, Bhan MK. Effect of community-based promotion of exclusive breastfeeding on diarrhoeal illness and growth: a cluster randomised controlled trial. Lancet 2003; 361:1418–1423.
21. Haider R, Ashworth A, Kabir I, Huttly SR. Effect of community-based peer counsellors on exclusive breastfeeding practices in Dhaka, Bangladesh: a randomised controlled trial. Lancet 2000; 356:1643–1647.
22. Dunn DT, Brandt CD, Krivine A, Cassol SA, Roques P, Borkowsky W, et al
. The sensitivity of HIV-1 DNA polymerase chain reaction in the neonatal period and the relative contributions of intra-uterine and intra-partum transmission. AIDS 1995; 9:F7–F11.