Corresponding Author Suzanne Filteau, Nutrition and Public Health Interventions Research Unit, London School of Hygiene and Tropical Medicine, Keppel Street, London WC1E 7HT, UK. Tel.: +44 20 7958 8108; Fax: +44 20 7958 8111; E-mail: email@example.com
Objective Parental HIV infection may affect even those exposed children who remain uninfected. We investigated early growth, an indicator of overall health, of infants born to Zambian mothers recruited for a study of breastfeeding and postpartum health.
Methods HIV-infected and uninfected women in Lusaka were followed regularly from late pregnancy to 16 weeks postpartum. Infant weight and length were measured at birth, 6 and 16 weeks. Infant HIV status could not be specifically determined in this cohort so comparisons were between all infants of HIV-uninfected mothers (n = 184) and those infants of HIV-infected mothers who were known to be alive and showed no clinical evidence of HIV infection at age 2–4 years (n = 85).
Results Most infants were exclusively or predominantly breastfed until 16 weeks. At all time points infants of HIV-infected mothers tended to have lower weight and length standard deviation (Z) scores (significant for weight at 6 weeks; P = 0.04), even after adjustment for their lower gestational age at birth, compared with infants of uninfected mothers. In multivariate analyses the major factors affecting weight or length at 6 or 16 weeks of age were birth weight or length, and maternal subclinical mastitis, primiparity and weight during pregnancy.
Conclusions Early growth of infants of HIV-infected mothers is less than that of uninfected mothers, in part associated with subclinical mastitis, and this effect cannot be overcome with intensive support of mothers to follow international recommendations regarding exclusive breastfeeding.
Objectif L'infection parentale par le HIV peut affecter même les enfants exposés qui restent non infectés. Nous avons étudié la croissance initiale, un indicateur de santé générale, des enfants en bas âge nés de mères zambiennes recrutées pour une étude sur l'allaitement et sur la santé post-partum. Méthodes: Des femmes de Lusaka infectées et non infectées par le VIH ont été suivies régulièrement à partir de la grossesse avancée jusqu’à 16 semaines après l'accouchement. Le poids et la taille des enfants ont été mesurés à la naissance, à 6 et à 16 semaines. Le statut VIH des enfants ne pouvant pas être spécifiquement déterminé dans cette cohorte, les comparaisons ont alors été réalisées entre tous les enfants en bas âge de mères non infectées par le VIH (n = 184) et les enfants en bas âge de mères infectées qui étaient vivants et ne présentaient aucune évidence clinique de l'infection VIH à l’âge 2 à 4 ans (n = 85).
Résultats La plupart des enfants en bas âge étaient exclusivement ou principalement allaité jusqu’à 16 semaines. A tout moment, les enfants en bas âge de mères infectées tendaient à avoir un écart type des ‘scores Z’ inférieurs pour le poids et la taille (significatifs pour le poids à 6 semaines, P = 0,04), même après ajustement pour leur âge gestationnel inférieur à la naissance par rapport aux enfants en bas âge de mères non infectées. Dans des analyses multivariées les facteurs principaux affectant le poids ou la taille à 6 ou 16 semaines d’âge étaient le poids ou la taille de naissance et, la mastite maternelle subclinique, la primiparité et le poids durant la grossesse.
Conclusions La croissance initiale des enfants en bas âge de mères infectées par le VIH est inférieure à celle des enfants de mères non infectées, en partie à cause de la mastite subclinique et cet effet ne peut pas être contrecarrer à cause du soutien intensif offert aux mères pour respecter les recommandations internationales concernant l'allaitement exclusif.
Objetivo La infección parental por VIH puede afectar incluso aquellos niños expuestos que continúan sin infectarse. Hemos investigado el crecimiento temprano, un indicador de salud total, en niños nacidos de madres en Zambia, reclutadas para un estudio sobre lactancia materna y salud postparto.
Métodos Se siguió de forma regular a mujeres infectadas y no infectadas por VIH en Lusaka, desde finales del embarazo y hasta 16 semanas después del parto. Se tomó el peso y la medida de los bebes al nacer, y 6 y 16 semanas después. El seroestatus para VIH de los bebés no pudo determinarse específicamente en esta cohorte, por lo cual las comparaciones fueron entre todos los bebes nacidos de madres no infectadas por VIH (n = 184) y aquellos de madres VIH positivas que se sabía estaban vivos y no tenían evidencia clínica de infección por VIH a los 2–4 años (n = 85).
Resultados La mayoría de los neonatos fueron exclusiva o predominantemente lactados hasta la semana 16. En todo momento los neonatos nacidos de madres seropositivas tendieron a tener un menor tamaño (Z-score) y peso (significativo para el peso en la 6 semana; P = 0.04), incluso después de ajustar por su menor edad gestacional al momento de nacer, comparados con neonatos de madres no infectadas. En un análisis multivariado los mayores factores que afectan el peso y el tamaño a las 6 o 16 semanas son: el peso y la medida al nacer, una mastitis materna subclínica, la primiparidad y el peso durante el embarazo.
Conclusiones El crecimiento temprano de neonatos con madres infectadas por VIH es menor que el de madres no infectadas, en parte asociado con una mastitis subclínica, y este efecto no puede ser superado con el apoyo intensivo de las madres a las recomendaciones internacionales sobre la exclusividad de la lactancia materna.
The HIV epidemic is having devastating effects on infant health and survival in Sub-Saharan Africa (Newell et al. 2004a). It is well recognized that infants who themselves become HIV-infected in utero, at birth, or postnatally through breastfeeding are at high risk of poor growth, frequent infections and early death (European Collaborative Study 2003; Newell et al. 2004a,b). Even infants of HIV-infected mothers who themselves remain HIV-uninfected also may be at increased risk of early mortality (Newell et al. 2004b). Maternal death is an important contributor to infant death (Nakiyingi et al. 2003; Ng'weshemi et al. 2003). Even surviving parents, if they are ill, may have difficulty caring optimally for their infant or working to maintain a sufficient family income. HIV-uninfected infants of infected Zairean mothers were at increased risk of diarrhoea and this was attributed to maternal illness or death (Thea et al. 1993). Two separate studies in Lusaka have shown that illness of the mother is a major reason for her to stop exclusively breastfeeding (Chisenga et al. 2005) or any breastfeeding (Kuhn et al. 2005) before the recommended time (WHO 2001). Postpartum illness, which is more common among HIV-infected women, is associated with subclinical mastitis (Kasonka et al. 2006) which in turn is associated with poor infant growth, probably due to decreased milk production as is seen in dairy cows with subclinical mastitis (Shuster et al. 1995).
Other factors than poor parental health may also contribute to poor growth among HIV-exposed but uninfected infants. There is evidence for immune abnormalities among HIV-exposed, uninfected infants (Ota et al. 1999; Chougnet et al. 2000; Clerici et al. 2000; Nielsen et al. 2001). These abnormalities may be along a causal pathway to poor infant growth or may be simply markers of fetal or early neonatal problems associated with poor growth.
Nevertheless, most published information suggests that in spite of these potential contributors to poor growth, the growth of HIV-exposed, uninfected children is not slower than that of appropriate controls (Lepage et al. 1996; Bailey et al. 1999; European Collaborative Study 2003). The choice of controls is important because in Africa even relatively healthy children may grow falter compared with international standards (Bobat et al. 2001) and in Europe and North America HIV-infected mothers often come from socially disadvantaged sections of the population and their children may grow poorly for reasons unassociated with HIV infection (Agostoni et al. 1998; Miller et al. 2001).
As programmes to prevent mother-to-child HIV transmission, including short course antiretroviral drugs around delivery, become more widespread in Sub-Saharan Africa, more and more infants of HIV-infected mothers will remain uninfected. Even in populations where prolonged breastfeeding is the norm, perinatal nevirapine treatment permits over 80% of infants to escape HIV infection (Jackson et al. 2003). Zambia has an estimated adult HIV prevalence of 16.5% overall (WHO 2004) with a considerably higher antenatal prevalence of about 27% in Lusaka (Fylkesnes et al. 2001). Thus HIV-exposed, uninfected infants represent about 13% of all infants born in the country and about 22% of all those born in Lusaka, making their problems of major public health importance. Poor infant growth is associated with both immediate and long-term poor health. Therefore it is important to investigate the factors contributing to poor growth of infants of HIV-infected mothers to design programmes to avoid or mitigate the problem. We investigated infant growth using detailed maternal health and infant feeding data from the Breastfeeding and Postpartum Health Study.
The Breastfeeding and Postpartum Health longitudinal cohort study was conducted in Chilenje, Lusaka, Zambia with recruitment from June 2001 to July 2003. The study received ethical approval from the ethics committees of the University of Zambia and the Institute of Child Health/Great Ormond Street Hospital, London, UK. The primary aim of the project was to determine how maternal health and breastfeeding practice were associated with subclinical mastitis, defined as raised milk sodium/potassium (Na/K) ratio (Filteau et al. 1999). Detailed methods and some results have been published (Chisenga et al. 2005; Gitau et al. 2005; Kasonka et al. 2006; Phiri et al. 2006). The population in Chilenje is middle income by Zambian standards and maternal height, weight and measures of vitamins A and E status indicate a fairly well-nourished population (Chisenga et al. 2005; Gitau et al. 2005).
Most women in Lusaka attend their local clinic for antenatal care and deliver either in the clinic or at the local University Teaching Hospital (UTH). In Lusaka women are offered antenatal HIV counselling and same day testing at the clinics. Nevirapine is given to the mother and infant around the time of delivery to decrease mother-to-child HIV transmission (Stringer et al. 2003). Clinics in Lusaka are all Baby Friendly with promotion of 6 months exclusive breastfeeding (EBF). As breastfeeding was the primary focus of the study, midwives in Chilenje additionally were trained in the WHO course for HIV and infant feeding. Zambia is a very poor country so that infant formula is prohibitively expensive and most HIV-infected women still elect to breastfeed their infant. Women choosing to breastfeed were given intensive counselling and support by the project midwives for doing so exclusively.
Inclusion criteria for the study were residence within the Chilenje clinic catchment area, known HIV status, intention to breastfeed, and written informed consent. HIV status could be either positive or negative but, as, for cost and logistical reasons, we aimed to recruit equal numbers of infected and uninfected women, there were more eligible uninfected women than could be recruited; the decision for recruitment was therefore based on the need to recruit similar numbers of infected and uninfected women in parallel.
Recruitment of eligible women was at 32–34 weeks gestation. A senior midwife administered a standardized questionnaire investigating personal and socioeconomic status and reproductive history and measured subjects’ height, weight and haemoglobin (by Hemocue, Dronfield, UK). Gestational age was determined from the last menstrual period which generally agreed with the estimate from midwife's examination. Most deliveries took place at Chilenje clinic unless the senior midwife or study obstetrician referred them to UTH. Infant weight and length were measured at delivery in Chilenje clinic but only birth weight for infants born at UTH or elsewhere as birth length is not routinely measured in Lusaka.
Women had 11 scheduled home or clinic visits by the study midwives or obstetrician within the first 16 weeks postpartum. At all visits women were administered questionnaires concerning her own and her infant's health, her breast health and details of the infant's diet. At days 3 and 7 and week 6 blood samples were taken to measure maternal haemoglobin. At weeks 6 and 16 infant weight and length were measured. At all postnatal visits spot milk samples were hand expressed separately from each breast into 20 ml plastic universal jars. Milk samples were frozen at −80 °C until assay for sodium and potassium as described previously (Filteau et al. 1999).
The duration of EBF was defined as the time since birth with nothing except breast milk given to the infant (Chisenga et al. 2005). As project visits were very frequent, especially early in lactation and up to the median time, 6 weeks, of stopping EBF, the duration could be estimated quite precisely.
We did not have ethical approval to assess infant's HIV status in this study since (i) at the time there were no antiretroviral drugs available so knowing the status could not be justified clinically and (ii) the study was not planned with a sample size sufficient to assess factors affecting mother-to-child transmission, so testing could not be justified epidemiologically. Although originally we planned to follow participants only to 16 weeks postpartum, we were later able to follow up about half of them to assess vital status of mothers and children between October 2004 and November 2005. At this point antiretroviral drugs were available for symptomatic children. At this later follow-up children were a mean of 2.74 years old (SD 0.43 years; range 1.7–4.2 years; no difference between children of HIV-infected and uninfected women). To restrict the analyses to only HIV-free surviving children, we compared early growth of the 85 infants of HIV-infected mothers who were available and appeared themselves uninfected at the later follow-up with the 184 infants of uninfected mothers with growth data until at least 6 weeks.
All data were double-entered into EpiInfo 6 (CDC, Atlanta, GA, USA), cross-checked, and imported into spss 12.0 (Chicago, IL, USA) for analysis. Infant anthropometric Z scores, both with and without correction for gestational age at delivery, were calculated from the British growth standards (Cole et al. 1998).
An overall rating of maternal or infant postpartum health in the 16 weeks was developed by inspection of data from the planned visits as well as any additional visits in case of serious illness. Women and infants were categorized as (i) ‘healthy’, i.e. no reported problems; (ii) ‘mild illness’, i.e. mild problems which could be solved by advice from the midwife or medications, such as painkillers or iron supplements; (iii) ‘moderate illness’, which referred to the doctor's prescription of antibiotics or antimalarials and (iv) ‘severe illness’ implying illness requiring greater medical intervention, usually hospitalization.
The primary outcome of the main BFPH study on which sample size was based was prevalence of subclinical mastitis, that is, raised milk sodium/potassium (Na/K) ratio. Na/K ratios for individual women were calculated as described previously (Kasonka et al. 2006) by comparison with the mean ratio at the same time point of a group of 32 ‘healthy’ women, defined as HIV-uninfected multiparous women who reported no illnesses during the 16 week follow-up. Individual differences from these mean values were averaged for each woman. Two sets of mean Na/K ratios were defined: early lactation (days 3 and 7) and established lactation (day 10 to week 16).
Univariate analyses were conducted with available independent continuous or categorical variables which might be related to infant growth. Maternal health variables were HIV status; parity (primiparous or multiparous); overall health rating as above; height, weight and haemoglobin. Infant data included overall health and duration of EBF. Socioeconomic variables were not associated with infant growth and are not shown. Variables showing significant associations with weight or length in univariate analyses were included in multivariate regressions.
Figure 1 shows the flow of mothers and infants through the main study and the follow-up in 2005. Of the HIV-uninfected women recruited, 184 of 218 (84%) of their infants’ early growth data was analysed. The major reason for dropout was early loss to follow-up, largely because of moving from the study area, and five infants died before 16 weeks. At follow-up 2–4 years later, two more infants and two mothers were known to have died but the major loss was due to moving away and there were few differences between mothers who were and were not available at follow-up (Table 1). For HIV-infected women, growth of 85 of 211 (40%) of infants was analysed. Fifteen singleton infants and one pair of twins died before 16 weeks and there were known to be 13 additional infant deaths and three maternal deaths by the later follow-up. Five more children were excluded from the analyses of HIV-free survivors because they were on antiretroviral treatment at follow-up after having been tested for HIV following presentation with AIDS-defining illnesses. We also excluded three infants of HIV-infected mothers who stopped breastfeeding before 16 weeks (and thus were removed from the main study which was about breastfeeding) and one infant of an HIV-infected mother who was an extreme outlier in terms of growth with Z scores consistently around −6.
Table 1. Characteristics at recruitment during pregnancy of mothers in original cohort and subset available and alive 2–4 years later
All with early growth data (n = 184)
Mothers and children alive 2–4 years later (n = 102)
All with early growth data (n = 180)
Mothers and children alive 2–4 years later (n = 85)
†Different from HIV-uninfected in the same group, P < 0.05.
‡Variable differed between those known alive at 2–4 years and those known to be dead or the child to be HIV-infected or the subjects not found, P < 0.05
23.2 (SD 5.0)
23.1 (SD 4.9)
25.4 (SD 5.3)†
25.5 (SD 5.5)†
64.0 (SD 9.6)
64.2 (SD 10.8)
63.8 (SD 8.7)
65.8 (SD 8.7)‡
160.1 (SD 5.8)
159.7 (SD 5.4)
160.1 (SD 5.8)
161.1 (SD 6.0)‡
108 (SD 15)
111 (SD 15)‡
107 (SD 16)
109 (SD 16)‡
Primiparous (n, %)
Married (n, %)
Table 1 presents descriptive data for the women at recruitment. Socioeconomic data did not differ between HIV-infected and uninfected women or between those found alive or not after 2–4 years and is not shown. HIV-infected women were older and fewer were primiparous than uninfected women. The HIV-infected women followed up 2–4 years later tended to have better nutritional status at recruitment during pregnancy than those in the full cohort.
For infants included in the analyses, 16 of 85 (19%) of those of HIV-infected mothers and 22 of 183 (11%) of those of uninfected mothers were born before 37 completed weeks gestation (P = 0.14 for HIV status; infant's gestational age missing for one HIV-uninfected mother). Eleven (31%) of the pre-term infants and 16 (7%) of the term infants with available data weighed <2.5 kg at birth.
Very similar results were obtained when weight and length Z scores were analysed with or without correction for gestational age. As pre-term delivery tended to be more common for HIV-infected women, further analyses of early infant growth avoided the complication of lower gestational age by using gestational age-adjusted Z scores (Figure 2). Occasional anthropometric data was missing because of missed visits and, in addition, birth lengths are missing from infants born at home or in other health facilities than the project clinic since only birth weight is measured routinely in Lusaka health facilities. There was a trend at both 6 and 16 weeks for infants of HIV-infected mothers to have lower weight/age and length/age Z scores than infants of HIV-infected mothers; this was significant (P = 0.04) only for weight at 6 weeks. When infant birth weight Z score was included in the analysis, the effect of maternal HIV status on weight at 6 weeks was no longer significant (P = 0.13).
Maternal height and weight during pregnancy were both positively correlated with infant Z scores (data not shown) which could represent both genetic and nutritional effects. Infants of primiparous mothers had slightly lower Z scores at birth and 6 weeks, but not 16 weeks, than infants of multiparous mothers; the differences were significant only for length/age Z scores at birth [−0.54 (SD 1.35) for 102 primiparous; 0.00 (SD 1.44) for 126 multiparous; P = 0.004] and 6 weeks [−0.35 (SD 1.38) for 116 primiparous; 0.05 (SD 1.55) for 142 multiparous; P = 0.03]. The four maternal haemoglobin measurements were highly intercorrelated. Weight/age at week 6 was positively correlated with all three postpartum haemoglobin measures (r = 0.20, P = 0.003, n = 218 for day 3; r = 0.17, P = 0.016, n = 211 for day 7; r = 0.13, P = 0.044, n = 231 for week 6). Week 6 length/age was correlated with week 6 haemoglobin (r = 0.13, P = 0.045, n = 231). Mean breast milk Na/K ratios from early and established lactation were intercorrelated and were previously shown in the full cohort to be significantly negatively correlated with weight at both 6 and 16 weeks and with height at 6 weeks (Kasonka et al. 2006). The associations of overall maternal or infant health with growth were not significant (data not shown).
The duration of EBF was associated with infant anthropometry not adjusted for gestational age (Chisenga et al. 2005) but poor growth preceded stopping EBF, rather than feeding practice affecting growth. Similar results were seen for gestational age-adjusted anthropometry (Table 2). Infants exclusively breastfed between 6 and 12 weeks had consistently the lowest Z scores, compared with infants exclusively breastfed for <6 weeks or for at least 16 weeks; differences were significant for weight at 6 and 16 weeks.
Table 2. Association of the gestational age-adjusted infant anthropometric Z scores with the duration of exclusive breastfeeding†
Duration of exclusive breastfeeding
†Values are mean (SD) Z scores, n. P-values are from the anova and superscripts from a Duncan's multiple range test where values followed by different superscripts are significantly different, P < 0.05.
−0.62 (1.25), 103
−0.88 (1.26), 58
−0.80 (1.07), 94
−0.14 (1.26), 90
−0.49 (1.56), 51
−0.28 (1.47), 79
6 week weight
0.79 (1.08),a 98
0.38 (1.35),b 57
0.81 (1.21),a 94
6 week length
−0.15 (1.47), 98
−0.38 (1.66), 58
−0.03 (1.38), 94
16 week weight
0.52 (1.01),a 97
−0.04 (1.36),b 56
0.51 (1.03),a 93
16 week length
−0.03 (1.36), 97
−0.18 (1.72), 57
−0.08 (1.17), 92
Multivariate linear regressions for gestational age-adjusted weight and length at 6 and 16 weeks were run including variables found significant in univariate analyses, i.e. maternal HIV status, parity, height, weight and haemoglobin at week 6, duration of EBF in weeks, birth weight or length and milk Na/K ratio in established lactation. Only birth anthropometry was included for analyses of anthropometry at 16 weeks because of high correlations with week 6 anthropometry and only Na/K ratio in established lactation was included because of its high correlation with Na/K ratio in early lactation. Results for the different analyses (Table 3) were broadly similar with birth anthropometry being the major determinant of week 6 and 16 anthropometry and high milk Na/K ratio also being strongly associated with poor growth. Maternal HIV status was significantly associated only with length at 6 weeks. Maternal anthropometry showed modest positive associations with infant anthropometry. Unlike in univariate analyses, primiparity was significantly associated with lower week 16, not week 6, anthropometry. Maternal haemoglobin and duration of EBF were not significant factors in any of the multivariate analyses.
Table 3. Multivariate linear regressions for infant weight and length gestational age-adjusted Z scores at 6 and 16 weeks of age*
6 weeks (n = 219)
16 weeks (n = 213)
6 weeks (n = 190)
16 weeks (n = 183)
B (95% CI)
B (95% CI)
B (95% CI)
B (95% CI)
*Values are regression coefficients (95% confidence intervals) from multiple linear regressions with backwards elimination of variables. Only variables with P < 0.1 in the final models are presented. Duration of exclusive breastfeeding and maternal haemoglobin at week 6 were entered into all regressions but were never significant. R2 was 53% for week 6 weight, 28% for week 16 weight, 38% for week 6 length and 32% for week 16 length.
†Birth weight was entered for weight analyses and birth length for length analyses.
‡HIV-negative = 0, HIV-positive = 1.
Birth weight or length Z†
0.75 (0.66, 0.85)
0.49 (0.37, 0.60)
0.61 (0.49, 0.73)
0.54 (0.42, 0.67)
Milk Na/K ratio
−0.37 (−0.62, −0.11)
−0.44 (−0.76, −0.13)
−0.48 (−0.90, −0.06)
−0.38 (−0.82, 0.05)
Maternal HIV status‡
−0.37 (−0.74, −0.01)
Maternal weight (kg)
0.02 (0.00, 0.03)
0.02 (0.01, 0.04)
Maternal height (cm)
−0.02 (−0.05, 0.003)
0.04 (0.01, 0.06)
−0.26 (−0.53, 0.01)
−0.37 (−0.74, −0.01)
Growth faltering is an early indication that infants of HIV-infected mothers have become HIV-infected themselves (Lepage et al. 1996; Agostoni et al. 1998; Bailey et al. 1999; Bobat et al. 2001; Miller et al. 2001). The health problems of these children require active medical intervention, for example, with prophylactic antibiotics (Coutsoudis et al. 2005) or antiretroviral drugs. However, a much larger proportion of infants of HIV-infected mothers are not themselves infected and their health problems will require different interventions. Our results suggest that HIV-exposed, uninfected Zambian infants have poorer early growth than their HIV-unexposed counterparts. The major contributors to this were smaller size at birth, even after controlling for gestational age, and maternal subclinical mastitis. Subclinical mastitis is associated with decreased milk production in dairy cows (Shuster et al. 1995) but milk production was not measured in the present study. It is notable that growth of HIV-exposed infants in Europe was not impaired compared with the British growth standards (European Collaborative Study 2003); these infants were mainly formula-fed so potentially decreased breast milk production by HIV-infected women would not have been an issue. Poor maternal health is a major determinant of subclinical mastitis (Kasonka et al. 2006). Thus, our results are similar to work in Africa showing the importance of maternal health for infant growth, health and survival (Thea et al. 1993; Nakiyingi et al. 2003; Ng'weshemi et al. 2003; Kuhn et al. 2005). Poor infant health was not significantly associated with poor early growth. Thus, although exposure to HIV in utero, at delivery or through breastmilk may alter the infant's immune system (Ota et al. 1999; Chougnet et al. 2000; Clerici et al. 2000; Nielsen et al. 2001), in this population this did not appear to be an important factor along the causal pathway for poor infant growth.
Our results differ from studies showing normal growth of uninfected infants of HIV-infected African mothers (Lepage et al. 1996; Bailey et al. 1999). The reasons for this are unknown but it may be relevant that overall the infants of the HIV-uninfected women in our study had higher anthropometric Z scores than infants in the other two African studies. Furthermore, in both those studies there appeared to be small, occasionally statistically significant, differences between HIV-exposed, uninfected infants, compared with HIV-unexposed infants, in early life which had resolved by about 1 year of age (Lepage et al. 1996; Bailey et al. 1999). Whether these early transient growth deficiencies have long-term effects will require further research.
An important limitation of our study is that we did not have ethical approval for HIV testing of infants in the cohort. However, our restricted analysis of a subgroup of infants alive and apparently HIV-free at age 2–4 years provides some information about HIV-exposed, uninfected children. More than half of HIV-infected African infants will have died before age 2 years compared with 7.6% of HIV-uninfected infants of infected mothers (Newell et al. 2004a). We also excluded infants of HIV-infected mothers who had died by the later follow-up to exclude women with late stage HIV who may have been too ill to care for their infant. The only bias we detected between those mother-infant pairs who were included in the later follow-up and those who were not was the intended one, i.e. especially among the HIV-infected group, those followed up were healthier, as indicated by nutritional status markers, than those not followed up.
The prenatal and early postpartum periods may be key for improving maternal and, consequently, infant health. Subclinical mastitis was seen from early lactation and was associated with slower infant growth (present study) and poor maternal health (Kasonka et al. 2006). In univariate analyses low weight at 6 weeks was associated with cessation of EBF shortly thereafter. It is unlikely that poor infant growth directly induced a woman to give additional foods or liquids for two main reasons: support for EBF to 6 months was intensive in this study, and overall infants gained Z scores from birth to 6 weeks so staff and mothers would not have been alerted to growth faltering. As advanced maternal HIV illness is a risk factor for mother-to-child transmission (John et al. 2001; Breastfeeding and HIV International Transmission Study Group 2004), and the severity of maternal illness is associated with poorer health and growth of her infant (Kuhn et al. 2005), a focus on improving maternal postpartum health may be able, in addition to improving maternal health, to improve infant growth and health. This could occur through both decreased HIV transmission as well as increased ability of the woman to care for her infant, for example, exclusively breastfeeding for a longer time (Chisenga et al. 2005). Multiple micronutrient supplementation (Fawzi et al. 2004) and prophylactic antibiotics may be low cost interventions to improve health of lactating HIV-infected women. In some cases, according to CD4 count, antiretroviral drugs may also be needed.
We are grateful to Dr Rachel Gitau, Dr Christine Kaseba, Dr Susan Murray and Dr Francis Kasolo for contributions to supervision of different parts of the work; to all the midwives in Chilenje Clinic, especially Tamara Kavimba, Esther Simwanza, Loveness Chilala, Catherine Katongo, Annie Kanunga, Mutinta Muzyamba, Catherine Msiska and Mulenga Nubule for follow-up and care of study subjects; to Hildah Mabuda for administrative assistance, and to John Chobo for home visits. We thank Professor Tim Cole, Institute of Child Health, for advice on analysis of infant growth data and Dr Tom Marshall, London School of Hygiene and Tropical Medicine, for advice on analysis of milk Na/K data. We are grateful to the Wellcome Trust for funding the main study and the Hestia Foundation for funding the later follow-up. This work was undertaken in collaboration with Great Ormond Street Hospital for Children NHS Trust that receives a proportion of its funding from the NHS Executive.