Pediatric HIV/AIDS is a major contributor to infant and under-5 mortality in resource-limited settings.1 HIV can be transmitted from mother-to-child during pregnancy, labor, and delivery or postnatally through breast-feeding (BF).2-5 In resource-limited settings, breast milk transmission contributes significantly to mother-to-child transmission (MTCT) of HIV.6-8 Avoiding all BF is still the only way to eliminate breast milk MTCT. However, in settings where this is not acceptable, feasible, affordable, sustainable, and safe (AFASS), exclusive breast-feeding (EBF), followed by stopping BF when this is AFASS, is recommended.9 Use of antiretrovirals to reduce breast milk MTCT has been investigated: A trial in Malawi shows that extended dose (for 14 weeks) nevirapine or nevirapine and zidovudine lowered the rate of HIV infection at 9 months (5.2% in the extended nevirapine group versus 6.4% in the extended dual prophylaxis group versus 10.6% in the single-dose nevirapine and 1-week zidovudine group).10 Most of the benefits of extended postexposure prophylaxis were gained during the period of prophylaxis, and the rate of additional HIV transmission after 14 weeks was similar in all 3 study arms. Similarly, the six week extended-dose neivrapine (SWEN) trial in Ethiopia, India, and Uganda showed that 6 weeks of nevirapine reduced HIV transmission at 6 weeks (compared with single-dose nevirapine), but this reduction was not sustained to 6 months.11 Although these postexposure prophylaxis trials are promising, questions around the feasibility of scaling up to programmatic level and how to improve HIV-free survival beyond the postexposure prophylaxis period arise, as most infants continue to breast-feed.10 Thus, investigating the feasibility of BF cessation (amongst women for whom replacement feeding was not AFASS at delivery) in operational (rather than research) settings is still urgently needed.
We describe adherence to and predictors of complete breast-feeding cessation (CBC) amongst HIV-positive BF women enrolled in prevention of mother-to-child Transmission (PMTCT) programs implemented through the South African national health care system. At the time of this study, national recommendations stated that HIV-positive women who do not meet the AFASS criteria, and thus choose to practice EBF, should stop BF between 12 and 24 weeks.12
Data are from a prospective observational cohort study, which sought to investigate HIV transmission at 9 months in an operational PMTCT setting. Results on early transmission, appropriateness of formula use, and HIV-free survival are reported elsewhere.13-15
The study was conducted at 3 government PMTCT sites (Paarl, Umlazi, and Rietvlei), purposively selected to reflect different socioeconomic contexts, rural-urban locations, and HIV prevalence rates. These sites have been described in detail by Jackson et al.15 In summary, HIV prevalence was 9% in Paarl, 28% in Rietvlei, and 47% in Umlazi, and infant mortality rates (per 1000 live births) were 40, 99, and 60, respectively.16
Study Procedures and Data Collection
Six hundred sixty-five HIV-positive and 218 HIV-negative women (and their infants) were recruited (September 2002-August 2003) from local antenatal clinics before or at the time of delivery. Maternal HIV status was determined from routine PMTCT medical records. HIV-positive women received antenatal infant feeding counseling from local clinics in accordance with national guidelines. Infant HIV status was only determined at 9 months, and the 3-week and 24-week specimens were tested, retrospectively.
Trained data collectors conducted the initial antenatal interview (32-36 weeks gestation), reviewed medical records postdelivery, and conducted structured interviews in the mother's home at 3, 5, 7, 9, 12, 16, 20, 24, and 36 weeks. At each time point, feeding practices over the previous 4 days were documented. At 24 weeks-the age by which women were recommended to stop BF completely-data on breast health were collected, and mothers were asked whether they had “completely stopped breast-feeding.” These questions were repeated at 36 weeks. The study team did not provide any interventions or advice on PMTCT or infant feeding-all information and interventions were provided by routine health services.
Ethical approval was obtained from the Biomedical Research Ethics Committee of the University of KwaZulu Natal. Buy-in was also obtained from participating institutions and provincial and national Departments of Health. Signed informed consent was obtained at enrollment, with verbal informed consent at each home visit. Compensation for interview time was based on local norms-food vouchers (Umlazi), cash (Paarl), or food parcels (Rietvlei).
Quantitative data from each visit were entered into MS ACCESS using double data entry at a central site (Medical Research Council Durban). After validation, the databases were exported to SAS version 9.1 (SAS Institute Inc, Cary, NC) for management and analysis. Our main outcome of interest was CBC which was first asked about at 24 weeks and not during previous visits. CBC was defined as reporting complete cessation of BF. Our second outcome of interest was not breast-feeding (NBF), defined as not giving breast milk during the 4 days before the last follow-up visit at or before 24 weeks. We were interested in examining BF cessation after BF initiation as per national recommendations. Thus data analysis was restricted to the subsample of HIV-positive women who practiced EBF or predominant breast-feeding (PBF) at 3 weeks. EBF was defined as giving the infant only breast milk with or without prescribed medicines during all the 4 days before the home visit. PBF was defined as giving breast milk (with or without prescribed medicines) during all the 4 days before the home visit and other nonnutritive liquids, during 1 or all the 4 days, but no nutritive liquids or solids during any of these 4 days. Although we expanded analysis to the subsample who gave any breast milk at 3 weeks, detailed analysis of this group is not reported due to its heterogeneity-it comprised women who initiated EBF and women who chose to avoid all BF; this would have conflicted with our overall aim of examining CBC in women who initiated EBF/PBF at delivery.
For the analysis of CBC as the main outcome of interest, characteristics of women who reported CBC were compared with those who reported continuing BF, using χ2 tests for categorical variables (Fisher exact test, if expected cell count <5; Cochran-Armitage test, for ordinal data with 1 binary variable) and t tests or Wilcoxon 2-sample test for continuous parametric or nonparametric data, respectively. Logistic regression was used to assess predictors of CBC. We hypothesized that site is an independent predictor of CBC as purposively selected sites differed in quality and type of infant feeding counseling.17 Explanatory variables that were clinically important and variables with a P < 0.25 were included in the preliminary main effects model. All explanatory variables were grouped into 6 key themes (Fig. 1). In an attempt to tease out the effects of each theme on CBC, we entered variables from the same theme systematically and consecutively into the model. Variables with a P value <0.05 or those that changed the odds ratio of the key exposure variable (site) by 10% or more were included in subsequent models. We checked the scale of 2 continuous variables-counseling score and number of postnatal clinic visits-and transformed them into categorical variables to check their linear distribution. We examined whether a critical counseling score or number of postnatal clinic visits facilitates CBC. We then also examined advice given for CBC, time taken for CBC, and breast health post-CBC. These data were only available for HIV-positive women who achieved CBC and neither for HIV-positive women who did not achieve CBC nor for HIV-positive women whose last visit was before 24 weeks.
Cumulative probabilities of NBF were assessed using Kaplan-Meier analysis, and associations between sociodemographic, maternal, and infant characteristics and NBF, infant death, and infant HIV-free survival were quantified using Cox regression. The validity of the proportional hazards (PH) assumption was assessed using correlation between Schoenfield residuals for a particular explanatory variable and the ranking of individual failure times. The extended Cox model was used when variables did not meet the PH assumption.
At week 3, 261 HIV-positive women reported practicing EBF or PBF (4-day recall data).
Data on CBC were only available for 201 women (77.3%) due to missed 24-week study visits. Women from this 3-week EBF/PBF cohort who remained in the study at 24 weeks were 2.9 times (95% CI 1.5 to 5.6) more likely to have ever disclosed their HIV status than women lost to follow-up. There was a tendency for more women to be lost to follow-up from Rietvlei (the rural site)-25% versus 10% from Paarl, and 13.9% from Umlazi-but this did not reach significance (P = 0.07). Of these 201, 88 (43.6%) reported CBC by 24 weeks.
Data on feeding pattern at 24 weeks or last follow-up visit was available for all 261 women. Of these, 102 (39.1%) reported NBF at week 24 or for 4 days before their last follow-up visit. The probability of NBF at 5, 7, 9, 12, 16, 20, and 24 weeks was 2.8% (95% CI 1.8% to 3.85), 4.3% (3.0% to 5.6%), 5.9% (4.4% to 7.4%), 9.8% (7.9% to 11.7%), 16.1 (13.8% to 18.4%), 23.1% (20.5% to 25.7%), and 37.6% (34.6% to 40.6%), respectively. By week 36, 123 (70.3%), of the 175 women who were EBF/PBF at week 3 and remained in the study at 36 weeks, reported completely stopping breastfeeding.
When we expanded our analysis to include the 309 HIV-positive women who initiated BF at delivery and were still BF at 3 weeks (though not necessarily EBF or PBF), 120 (38.8%) were NBF at last visit; 106 of 228 followed up to 24 weeks (46.49%) and 143 of 204 followed up to 36 weeks (70.10%) reported CBC by 24 and 36 weeks, respectively.
Comparing Women Who Were BF With Those Who Were Not Breastfeeding
There were significant differences in the select population who reported CBC and those who did not (Table 1): CBC was most common in Paarl followed by Rietvlei and Umlazi (69.6% versus 55.6% versus 35.1%, respectively, P = 0.0016). A significantly greater proportion of women who stopped BF were single and had been advised to introduce other foods or to formula feed. However, paradoxically, these women had significantly fewer antenatal visits and lower antenatal counseling scores. Quality of infant care, measured as whether the infant ever received cotrimoxazole prophylaxis, may have been better amongst women reporting CBC (P = 0.07). However, none of the other indicators of postpartum care (number of postpartum clinic visits, 9-month immunization, and having ever received multivitamins) were significantly different between the 2 groups.
On univariate analysis site, maternal knowledge, AFASS criteria, antenatal care, postnatal care and support, and infant health status were not significantly different amongst women practicing NBF compared with women who were still BF at the last visit (Table 1).
Advice Given, Time Taken to Achieve CBC, and Breast Health Post-CBC
Of the 196 women for whom data on CBC were available, 67 (34.2%) reported being told how to completely stop BF but not all of these provided details: 22 women (42.31%) reported being advised to introduce commercial infant formula, 25 (41.67%) reported being told to stop over weeks or months, and 2 (4.76%) reported being told to stop over days. Three women reported applying something on the breast to make the child refuse breast milk, and 2 women reported offering the child the breast for comfort but not for feeding. Amongst the 88 women who practiced CBC, 34 (38.64%) reported being criticized for CBC, mainly by their mother, mother-in law, and the father of the child. The median time taken to achieve CBC, from first trying, was 7 days (n = 78, range 1-90 days). Amongst 77 women, respondents who were EBF/PBF at 3 weeks and practiced CBC, 47 (61%) experienced self-reported breast health problems-21 (45%) reported engorged and inflamed (tender, red) breasts, 21 (45%) reported engorged breasts only, and 5 (10%) reported inflamed breasts only.
Predictors of CBC at 24 Weeks
Table 2 shows how we built the most parsimonious model to determine predictors of CBC at 24 weeks. In the final model, “health staff suggesting formula use” and infant hospitalization' were independent predictors of CBC. However, controlling for water quality (model 2 versus model 1) decreased the odds ratio and narrowed the CI of site Rietvlei versus Paarl. Furthermore, when antenatal counseling about stopping BF was controlled for (model 3 versus model 2), then site Rietvlei and Umlazi were protective against stopping BF. However, the addition of a postnatal care indicator (ever received cotrimoxazole) to model 3 removed all differences between Paarl and the other 2 sites (Final model) and site Rietvlei and Umlazi were no longer protective against stopping BF. Our data thus suggest that CBC at 24 weeks, although not independently predicted by many factors, is explained by a combination of factors, including water quality, antenatal counseling to stop BF, and postnatal infant care.
NBF at Last Visit at or Before 24 Weeks and Association With Infant HIV or Death
Between 3 and 24 weeks, 13 infants born to the 261 HIV-positive women died-all 13 were still breastfeeding before their death. Thirty-three infants either died or contracted HIV infection between 3 and 24 weeks. In a model that controlled for site, antenatal AFASS criteria, antenatal feeding intention, advice on formula feeding (ever), disclosure (ever), infant HIV status, antenatal knowledge that breast milk transmits HIV, infant hospitalization ever, and postnatal care, there were no independent predictors of NBF. Controlling for all these factors and for NBF, the only independent predictor of infant death was infant HIV status [hazard ratio (HR) 5.5 95% CI 2.4 to 12.5]. NBF was not protective against infant HIV or death in univariate (HR 0.89; 95% CI 0.66 to 1.18) and multivariable analysis, controlling for all the above-mentioned factors (HR 1.0 95% CI 0.7 to 1.4).
To our knowledge, this is the first study to examine the feasibility, effectiveness, and impact of national guidelines on BF cessation amongst HIV-positive women enrolled in routine PMTCT programs. We show that despite national guidelines which recommend that BF HIV-positive women stop BF by 24 weeks, only 39%-44% of women practicing EBF/PBF at week 3 adhered to these recommendations. The only other study we know of that examined CBC after EBF was conducted in a research setting in Abidjan18 and found that only 14% of women succeeded in practicing EBF with cessation by month 4. Failure was associated with having fewer children and a lower viral load. In our operational setting, 88 (43.8%) and 143 (70.10%) reported CBC by 24 and 36 weeks, respectively. It is encouraging that a larger percentage of women achieved CBC in our setting compared with the percentage documented in the Abidjan research setting, and that only 30% of the HIV-positive women in our analysis were still BF at 9 months-which is less than the estimated 60% of South African women in the general population still BF at 12 months.19 This suggests that PMTCT-related counseling on cessation is somewhat effective. However, our data also show that there is room for improvement: family support for CBC was suboptimal-amongst women who stopped BF, 34 (38.64%) were criticized for CBC. Furthermore, few women received advice on how to practice CBC, and duration of cessation varied tremendously. This further substantiates existing data which show that patterns of feeding are complex and dynamic.20 Our logistic regression analysis highlights the complex interplay of various factors when implementing CBC. The only independent predictors of CBC were health staff having suggested formula use postnatally and infant hospitalization. The latter, however, is difficult to interpret as reverse causation-as described among non-BF HIV-positive women in Kenya, Botswana, and Malawi-cannot be excluded.21-23 We expected that stopping BF early would be strongly associated with and independently predicted by infant feeding counseling, disclosure, number of antenatal visits, socioeconomic score or income, availability of water, maternal education level, and availability of fuel, as these factors contribute toward the AFASS criteria. Thus, BF HIV-positive women who subsequently meet these criteria should stop BF. However, our analyses in this article, and a previous article,14 found that in an operational setting, AFASS criteria were not being used to guide feeding choice and practice. This finding is different from a South African research setting where the AFASS criteria guided feeding choice and practice.24
Our limited sample suggests that breast problems immediately post-CBC were common. This differs with findings from a South African research setting which shows that with appropriate support, breast health problems during and after EBF are rare.25 These data suggest that in operational settings, health staff need to provide focused counseling and support to ease the transition from EBF to CBC so that breast problems are minimized.
Although our data do not identify the steps needed to facilitate CBC amongst HIV-positive women, they do highlight several issues: in our operational setting, CBC was more feasible at 36 than 24 weeks; BF cessation does not simply occur after antenatal infant feeding counseling; cessation may not always be easy to implement for personal, infant, and family reasons; although antenatal counseling, safety, and feasibility factors and postnatal counseling contribute toward cessation, cessation is not dependent on them; and lastly, early cessation of BF may carry no significant benefit for HIV-free survival-a finding similar to that of the Zambian Exclusive Breast-Feeding Study26 and the Hlabisa vertical transmission study27-but which we state with reservation due to inadequate statistical power in our analysis.
Our data has several limitations: first, data on CBC was only documented at 24 and 36 weeks; second, loss to follow-up at 24 weeks was high (22.7%); third, we only assessed explanatory variables associated with the AFASS criteria antenatally; fourth, quality of postnatal care was based on number of postnatal clinic visits and availability of cotrimoxazole; fifth, we relied on recall to document CBC and NBF; and sixth, we did not assess degree of infant illness and reason for hospitalization. All this information needs to be gathered at more than 1 time point in future studies examining the effectiveness and impact of stopping BF. However, our data do contribute to the body of literature-which is currently small-on effectiveness of stopping BF amongst women enrolled in routine PMTCT programs.
In conclusion, we show that completely stopping BF by 24 weeks was not feasible amongst most EBF/PBF HIV-positive women-only 39%-44% reported NBF or complete cessation of BF. After examining our data in the context of the literature, we suggest that if public health programs continue to recommend that women who initiate EBF stop BF early to reduce postnatal HIV transmission, then activities and interventions (within the health system and at individual, household, and family level) to facilitate CBC-with minimal adverse effects on maternal and infant health-need to be identified and tested in routine settings.
The Good Start Study Group would like to thank the study sponsors: 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. Black R, Morris S, Bryce J. Where and why are 10 million children dying every year? Lancet
2. European Collaborative study. Risk factors for mother-to-child transmission of HIV-1. Lancet
3. Nduati R, John G, Kreiss J. Postnatal transmission of HIV-1 through pooled breastmilk. Lancet
4. Nduati R, John G, Richardson B, et al. Human Immunodeficiency virus type 1 infected cells in breastmilk: association with immunosuppression and vitamin A deficiency. J Infect Dis
5. Palasanthiran P, Ziegler J, Stewart G, et al. Breastfeeding during primary maternal immunodeficiency virus infection and risk of transmission from mother to infant. J Infect Dis
6. Breastfeeding and HIV Inernational Transmission Study Group, Coutsoudis A, Dabis F, et al. Late postnatal transmission of HIV-1 in breast-fed children: an individual patient data meta-analysis. J Infect Dis
7. Miotti PG, Taha TE, Kumwenda NI, et al. HIV transmission through breastfeeding: a study in Malawi. JAMA
8. Nduati R, John G, Mbori-Ngacha D, et al. Effect of breastfeeding and formula feeding on transmission of HIV-1: a randomized clinical trial. JAMA
9. World Health Organisation on behalf of the Inter-Agency Task Team. WHO HIV and Infant Feeding Technical Consultation Held on behalf of the Inter-agency Task Team (IATT) on Prevention of HIV Infections in Pregnant Women, Mothers and their Infants Geneva, October 25-27, 2006
. World Health Organisation; Geneva, Switzerland. Available at: www.who.int/reproductive-health/stis/mtct/infantfeedingcinsensusstatement.pdf
. Accessed April 13, 2008.
10. Kumwenda N, Hoover DR, Mofenson L, et al. Extended antiretroviral prophylaxis to reduce breastmilk HIV-1 transmission. N Engl J Med
11. Six Week Extended Dose Nevirapine (SWEN) Study Team. Extended-dose nevirapine to 6 weeks of age for infants to prevent HIV transmission via breastfeeding in Ethiopia, India and Uganda: an analysis of three randomised controlled trials. Lancet
12. National Department of Health SA. National Department of Health Protocol for providing a comprehensive package of care for the prevention of mother to child transmission of HIV (PMTCT) in South Africa. Available at: www.doh.gov.za
. Accessed March 15, 2008.
13. Colvin M, Chopra M, Doherty T, et al. Operational effectiveness of single dose nevirapine in the South African programme to prevent mother-to-child transmission of HIV. Bull World Health Organ
14. Doherty T, Chopra M, Jackson D, et al. Infant feeding choices of HIV-positive women: do the WHO/UNICEF guidelines improve infant HIV-free survival. AIDS
15. Jackson D, Chopra M, Doherty T, et al. Operational effectiveness and 36 week HIV-free survival in the South African programme to prevent mother-to-child transmission of HIV-1. AIDS
16. National Department of Health SA. National HIV and syphilis antenatal sero-prevalence survey in South Africa. 2003; Available at: http://www.doh.gov.za/docs/reports/2003
. Acccessed October 9, 2006.
17. Chopra M, Jackson D, Ashworth A, et al. Preventing HIV transmission to children: an evaluation of the quality of counselling provided to mothers in three PMTCT pilot sites in South Africa. Acta Paediatr
18. Becquet R, Ekouevi D, Viho I, et al. Acceptability of exclusive breast-feeding with early cessation to prevent HIV transmission through breastmilk, ARNS 1201/1202 Ditrame Pluc, Abidjan, Cote d'Ivoire. J Acquir Immune Defic Syndr
20. Goga A, Jackson D, Doherty T, et al. Infant feeding patterns in the context of HIV. Presented at: International AIDS Conference; August 13-18, 2006; Toronto, Canada. Poster display.
21. Creek T, Arvelo W, Kim A, et al. A large outbreak of diarrhoea among non-breastfed children in Botswana, 2006-implications for HIV prevention strategies and child health. Presented at: Fourteenth conference on retroviruses and opportunistic infections (CROI); February 27, 2007; Los Angeles, CA.
22. Kafulafula G, Thigpen M, Hoover DR, et al. Post-weaning gastroenteritis and mortality in HIV uninfected African infants receiving antiretroviral prophylaxis to prevent MTCT-1 of HIV. Presented at: Fourteenth conference on retroviruses and opportunistic infections (CROI); February 25, 2007; Los Angeles, CA.
23. Kourtis A, Fitzgerald D, Hyde L, et al. Diarrhoea in uninfected infants of HIV-infected mothers who stop beastfeeding at 6 months: the BAN experience. Presented at:Fourteenth conference on retroviruses and opportunistic infections (CROI); February 27, 2007; Los Angeles, CA.
24. Bland R, Rollins N, Coovadia H, et al. Infant feeding counselling for HIV-infected and uninfected women: appropriateness of choice and practice. Bull World Health Organ
25. Bland R, Becquet R, Rollins N, et al. Breast health problems are rare in both HIV-infected and HIV-uninfected women who receive counseling and support for breast-feeding in South Africa. Clin Infect Dis
26. Kuhn L, Aldrovandi G, Sinkala M, et al. Effects of early weaning for HIV-free survival of children in Zambia. N Eng J Med
27. Rollins N, Becquet R, Bland R, et al. Infant feeding, HIV transmission and mortality at 18-months: the need for appropriate choices by mothers and prioritisation within programmes. AIDS
Keywords:© 2009 Lippincott Williams & Wilkins, Inc.
breast-feeding cessation; exclusive breast-feeding; formula feeding; prevention of mother-to-child Transmission; predominant breast-feeding; stopping breast-feeding early; stopping breast-feeding; vertical transmission