CROI: Four African studies suggest health risks associated with abrupt and early weaning of HIV-exposed infants

This article is more than 17 years old. Click here for more recent articles on this topic

Infant feeding in sub-Saharan Africa

Traditionally, most mothers in sub-Saharan Africa breastfeed their infants, introducing water and other foods gradually after a few months; but often they don’t totally wean their babies until they are between one to two years of age. Diarrhoea is common once mothers start giving their infants complementary foods, but breastfeeding protects against more serious cases, particularly early in life.

But when the mother is HIV-infected, there is risk of mother-to-child transmission (MTCT) of HIV to her infant that is cumulative over the course of breastfeeding.

Evidence that prolonged mixed breastfeeding especially erodes the benefits of antiretroviral prophylaxis in preventing MTCT led to WHO/UNICEF guidance that mothers with HIV avoid breastfeeding whenever replacement feeding options are acceptable, feasible, affordable, sustainable and safe (AFASS). However, when complete avoidance of breastfeeding isn’t possible from birth, it was recommended that mothers exclusively breastfeed, and wean their infants as quickly as possible, as soon as replacement-feeding options do become AFASS.

However, since the guidelines were issued there have been few data to show exactly how and when to wean HIV-exposed infants. However, the guidelines have been modified to recommend that if the circumstances for replacement feeding are not AFASS sooner, the mother should continue breastfeeding for six months.



Abnormal bowel movements, characterised by loose, watery or frequent stools, three or more times a day.

mother-to-child transmission (MTCT)

Transmission of HIV from a mother to her unborn child in the womb or during birth, or to infants via breast milk. Also known as vertical transmission.

exclusive breastfeeding

Feeding an infant only breast milk, with no other liquids or solids, for the first six months of life.

vertical transmission

Transmission of an infection from mother-to-baby, during pregnancy, childbirth, or breastfeeding.



The result of a statistical test which tells us whether the results of a study are likely to be due to chance and would not be confirmed if the study was repeated. All p-values are between 0 and 1; the most reliable studies have p-values very close to 0. A p-value of 0.001 means that there is a 1 in 1000 probability that the results are due to chance and do not reflect a real difference. A p-value of 0.05 means there is a 1 in 20 probability that the results are due to chance. When a p-value is 0.05 or below, the result is considered to be ‘statistically significant’. Confidence intervals give similar information to p-values but are easier to interpret. 

But over the last couple of years, there have been reports of operational difficulties putting this guidance into practice in many resource-limited settings. Mothers often have difficulty obtaining reliable replacement foods — even when formula feed is freely provided it may not be culturally acceptable and often puts the mother at risk of having her HIV status disclosed involuntarily to her family and community and of being stigmatised as a result (see here and here).

Furthermore, a recent outbreak of diarrhoea and a related increase in mortality among formula fed infants after a very heavy rainy season in Botswana has demonstrated that even in the best resourced sub-Saharan Africa settings it can be difficult to formula feed safely (see ) or at least keep the baby’s environment hygienic (see here ).

In the light of such reports, as well as observations that after implementing early and abrupt weaning guidance the rates of diarrhoea and death seemed to increase among HIV-exposed infants in their own sites, several teams of researchers started examining the effect of altered infant feeding practices in their local settings.

Three of these studies, conducted in Malawi, Kenya and Uganda compared health outcomes in HIV-exposed but uninfected infants before and after WHO/UNICEF guidance was introduced directing HIV-infected mothers to abruptly wean their infants after the first months of life.

Abrupt and early weaning of HIV-exposed but negative infants in sub-Saharan Africa is associated with high rates of severe diarrhoea and gastroenteritis which in some cases led to death according to four different poster presentations made at the Fourteenth Conference on Retroviruses and Opportunistic Infections in Los Angeles last week. For three of these studies, these rates were significantly higher than seen in historical controls, despite the fact that virtually all the children in the more recent studies are getting cotrimoxazole prophylaxis, which should have made the outcomes better.

“Following the WHO feeding guidelines for HIV-exposed infants… was associated with an increased rate of diarrhoea at the time of weaning,” said Dr Rose Masaba of the Kenya Medical Research Institute. She and other presenters suggested that there needs to be improved instruction on hygiene, sanitation, and water purification during pre-weaning counselling — and that infants being weaned early be more closely monitored, and, in most settings, offered quality nutritional support.

However, another presenter, Dr Carolyne Onyango, from the Makerere University and Johns Hopkins University Research Collaboration in Kampala, Uganda, went a step further and suggested that the relatively consistent data from the infant feeding studies presented at the conference “indicate the need to carefully reassess current WHO and MOH recommendations which encourage early breastfeeding cessation among HIV infected women in resource-limited settings.”

Diarrhoea in KiBS vs VTS

Likewise, the presentation by Dr Masaba compared the rates of diarrhoea and hospitalisation related to diarrhoea in the Kisumu Breastfeeding Study (KiBS) versus the Vertical Transmission Study (VTS) conducted at the same site in Kenya.

KiBS is an ongoing study examining the effect on HIV transmission and infant health of giving HIV-positive mothers combination ART from week 34 of gestation until six months after birth. Women in this study are instructed to exclusively breastfeed and then wean rapidly over two weeks before ART is discontinued (switching to locally available foods). Cotrimoxazole was given to both mothers and infants (after six weeks of age).

The VTS, which was conducted between 1996 and 2001, was not a PMTCT study, but rather examined the effects of malaria on HIV transmission. There were no nutritional inventions in this study, and mothers fed their infants according to traditional practices. No cotrimoxazole was provided to mothers in this study.

Over the course of twelve months of follow-up the incidence of diarrhoea was significantly lower in KiBS (n= 144, 5.7 episodes per 100 infant-months of observation) than in the vertical transmission group (n=347, 8.7 episodes/100 months of observation; RR = 0.66; 0.55 to 0.79). Rates of diarrhoea increased after the third month in the VTS study and remained elevated, but peaked dramatically in the KiBS study during months six and seven, and then decreased.

But as in the Malawian setting, rates of severe diarrhoea requiring hospitalisation were much higher in the more recent study (1.8 versus 0.5 per 100 infant months). This was consistently the case over one year of observation, however, rates peaked at six months (3.8 vs. 0.4 episodes per 100 infant months) and were significantly higher overall than in the VTS (RR = 3.60; 2.31 to 5.63). There were two deaths among HIV-infected infants due to diarrhoea (ages seven and twelve months) in KiBS.

Diarrhoea in PEPI vs. NVAZ

In the first report, the frequency of diarrhoea among 1,792 infants in an ongoing clinical trial, the Extended Infant Post-Exposure Prophylaxis study (PEPI), conducted in Blantyre, Malawi, was compared to the frequency of diarrhoea in 1,810 infants in a single-dose nevirapine (sdNVP) plus short-course AZT PMTCT study (NVAZ) conducted several years earlier at the same site.

According to Dr Michael Thigpen of the US Centers for Disease Control who presented the analysis, in PEPI “mothers received active education, counselling and guidance to encourage them to exclusively breastfeed and then wean abruptly at six months as per the WHO guidelines,” (for an median duration of overall breastfeeding of 183 days). Mothers were educated on how to properly breastfeed and on replacement feeding options after weaning — and both mothers and infants in the study received the antibiotic cotrimoxazole (which has been shown to markedly reduce infant mortality).

Early weaning had not been recommended to mothers in the NVAZ study who mixed breastfed according to traditional methods as described above (with a median overall breastfeeding duration of 732 days in that study) and cotrimoxazole was not widely used by infants in this study.

Over the course of a year, rates of diarrhoea were similar in both studies, however, the diarrhoea in the infants who were weaned early was 10-30 times more likely to be severe, leading to hospitalisation in the months immediately after abrupt weaning (3.1% at seven to nine months of age, compared to 0.1% in those who were weaned in the traditional manner).

The rate of death due to diarrhoea was also twice as high among the infants weaned abruptly (six vs. three deaths per 1,000 infants respectively) and remained higher through the end of the first year of life (28 vs. 12 deaths per 1000 infants, respectively). The Log-rank test for overall mortality was p = 0.04 and for gastroenteritis-related mortality, p = 0.0003.

Diarrhoea in HIVIGLOB vs. HIVNET 012

A third presentation examined the rates of gastroenteritis and mortality among HIV-negative infants in the HIVIGLOB study (an ongoing immunoglobulin vs. sdNVP antiretroviral trial for PMTCT) and in HIVNET 012 (the PMTCT study in the late 1990s which first demonstrated the benefit of sdNVP). Both studies were performed at the same site in Uganda, but by the time of the HIVIGLOB study, the Ugandan Ministry of Health had begun recommending that women exclusively breastfeed and then abruptly cease breastfeeding sometime between three to six months.

There are 593 HIV-negative infants in the HIVIGLOB trial, and 499 were in the HIVNET 012 trial. Demographic characteristics were quite similar for the mothers in each study, except that in HIVIGLOB, the infants were weaned at a median age of three months (interquartile range two to six months). Dr Onyango, who presented these results, noted that many mothers tended to wean as soon as they got their babies’ HIV-negative PCR test results after the first weeks of life. Infants in HIVNET 012 were weaned much later, at a median age of 8.7 months.

In the comparative analyses, the rate of serious gastroenteritis (dysentery and diarrhoea leading to hospitalisation) during the first year of life was about twice as high in HIVIGLOB compared to HIVNET 012.

In both trials, the frequency of serious diarrhoea jumped significantly during the three months post-weaning compared to the 3 months prior to weaning: rising from 0.3% prior weaning to 2.2% post-weaning in HIVNET 012 (Rate Ratio (RR) 7.5 (0.9-61.1), p=0.031), and from 1.7% to 3.9% in HIVIGLOB (RR 2.3 (1.1-4.8), p=0.033).

The cumulative mortality was also higher in HIVIGLOB, though only through month nine. By month ten, in HIVNET 012, the rate of death per 1000 infants was higher. However, follow-up of infants HIVIGLOB (which only completed enrolment in July 2006) is still ongoing.


In a session where these posters were discussed, a leading expert on perinatal transmission, Dr Taha Taha of Johns Hopkins University, and a co-author of some of the studies, stressed that although these studies all reached very consistent conclusions, they have to be interpreted cautiously for several reasons.

For instance, the studies are still ongoing, their data are preliminary, and for the purposes of these analyses, they all used historic controls. Nutritional assessments have not yet been reported for the infants in these studies (though such data are being gathered). The studies have not documented the local hygienic patterns and practices that could clearly impact results. Secular trends and undefined temporal changes, such as an unusually heavy rainy season (as was associated with the diarrhoea outbreak in Botswana in 2005-2006) or other biases could have accounted for the differences seen in rates of serious diarrhoea and infant mortality.

However, many of changes that have occurred in the quality of care over time should have biased the outcomes in favour of the more recent studies. Nutritional interventions, counselling and support programmes for the mothers should be better — and then there should be substantial benefits from the recent introduction of cotrimoxazole prophylaxis.

“If anything, these kids should be doing better,” said Dr Lynne Mofenson, of the National Institute of Child Health and Human Development who co-chaired the poster discussion session.

Furthermore, on the following day of the conference, complementary results were presented from a prospective randomised trial, the Zambia Exclusive Breastfeeding Study (ZEBS) that compared HIV-free survival among infants exclusively breastfed for over 6 months to a similar cohort of infants abruptly weaned at four months (though including both HIV-positive and negative infants). Those results, which suggest that early weaning is not associated with improved HIV-free survival, will be covered in more detail in a future article. More of the debates about the implications of the infant feeding studies presented at the conference will be covered in a subsequent article.


The fourth poster presentation investigated the occurrence of diarrhoea before and after abrupt weaning at 28 weeks among HIV-negative infants of HIV-infected mothers enrolled in the Breastfeeding Antiretrovirals Nutrition (BAN) study currently ongoing in Lilongwe, Malawi.

The BAN study is comparing the outcomes in over 1,000 infants of the standard PMTCT regimen (sdNVP plus one week of Combivir, and sdNVP for the infant) versus maternal ART or nevirapine for the infants over the course of exclusive breastfeeding. Women in this study are healthier than in KiBS, with CD4 cell counts above 200 and half of the mothers were also randomised to receive a nutritional supplement during breastfeeding. In addition, all infants in this study are provided with a six-month supply of highly nutritious food at the time of weaning (using a locally manufactured nutrient-fortified milk/peanut butter-based product which has been successfully used for the rehabilitation of malnourished children).

As of June 2006, 771 HIV-negative infants had been enrolled in BAN, of whom 225 uninfected infants had reached 28 weeks of age. As in the other two studies, cases of diarrhoea jumped dramatically around the time of weaning and remained elevated through the end of the first year of life (ranging from 21.4% to 63.4%). Hospitalisations due to diarrhoea also peaked in the months after weaning, in infants aged six to eight months old (2.3% at 8 months).

However, it is not clear that these rates are significantly different from the usual pattern of diarrhoea seen in infants in Malawi, which also peaks between six and eleven months of age (rising to around 41.2% in the 2004 Malawi Demographic and Health Survey).

The overall HIV-negative infant mortality (43/1000) was much lower than that reported in the DHS survey (76/1000 live births). However, that survey would have included both HIV-infected and uninfected infants — and few children in the Malawian survey would have been receiving the level of care and nutritional support given to those the BAN study.

Dr Athena Kourtis of the CDC, who presented these results, also noted that there was a significantly higher probability of infant diarrhoea during the rainy season, compared with the non-rainy season (p


Kourtis A et al. Diarrhea in uninfected infants of HIV-infected mothers who stop breastfeeding at 6 months: the BAN study experience. Fourteenth Conference on Retroviruses and Opportunistic Infections, Los Angeles, abstract 772, 2007.

This poster can be downloaded directly from the conference website.

Kafulafula G et al. Post-weaning gastroenteritis and mortality in HIV-uninfected African infants receiving antiretroviral prophylaxis to prevent MTCT of HIV-1. Fourteenth Conference on Retroviruses and Opportunistic Infections, Los Angeles, abstract 773, 2007.

Onyango C et al. Early breastfeeding cessation among HIV-exposed negative infants and risk of serious gastroenteritis: findings from a perinatal prevention trial in Kampala, Uganda. Fourteenth Conference on Retroviruses and Opportunistic Infections, Los Angeles, abstract 775, 2007.

This poster can also be downloaded from the conference website.

Thomas T et al. Rates of diarrhea associated with early weaning among infants in Kisumu, Kenya. Fourteenth Conference on Retroviruses and Opportunistic Infections, Los Angeles, abstract 774, 2007.