Growth failure in children is a global health problem in developing countries. Data on predictors of growth failure in infants born to women infected with HIV are limited in this setting(Reference Webb, Manji and Fawzi1). In vertically HIV-infected children, poor growth may be an early marker of infection or progression to disease(Reference Newell, Borja and Peckham2). HIV-infected children normally grow considerably more slowly than their HIV-exposed, uninfected counterparts, and the gap between the two groups tends to widen with age. Children born to HIV-positive women are particularly susceptible to malnutrition. Growth failure is frequently found in this population, especially in areas where a high prevalence of HIV/AIDS coexists with high rates of food insecurity(Reference Newell, Borja and Peckham2, Reference Arpadi, Fawzy and Aldrovandi3).
HIV exposure in utero without subsequent infection may affect growth in infancy and early childhood(Reference Newell, Borja and Peckham2). Newborns whose mothers are infected with HIV may have higher rates of fetal malnutrition than newborns of HIV-seronegative mothers(Reference Gangar4), and this disadvantage is likely to extend beyond infancy. A study in Zambia found that HIV-exposed, uninfected infants had poorer growth than their HIV-unexposed counterparts(Reference Makasa, Kasonka and Chisenga5). According to a WHO working group, uninfected infants born to HIV-positive mothers grow in an unfavourable intra-uterine environment where, depending upon the setting, they may suffer from greater exposure to maternal infection and maternal malnutrition, or to drug addiction, smoking and poor social conditions(Reference Agostoni, Zuccotti and Giovannini6). Furthermore, maternal HIV severity is associated with higher morbidity and mortality, and lower weight at age 4 months, among HIV-exposed but uninfected children(Reference Kuhn, Kasonde and Sinkala7). More studies have been recommended to elucidate the relationship between maternal immune status markers including viral load and CD4 cell count and child growth(Reference Ezeaka, Iroha and Akinsulie8).
Many other determinants of nutritional status of children have been reported. Parental education, particularly mother's education, is reported to have a strong influence on the nutritional status of the child(Reference Cattaneo, Timmer and Bomestar9–Reference Hong and Mishra11). Household socio-economic status and access to clean water may also have a significant impact on children's nutritional status(Reference Hong and Mishra11, Reference Kikafunda, Walker and Collett12). Some studies have found that rural children are more likely to be stunted, wasted or underweight than those living in urban settings(Reference Cattaneo, Timmer and Bomestar9, Reference Hien and Hoa13). Maternal health is also important, for example maternal depression may interfere with the quality of care given to the infant(Reference Patel, Rahman and Jacob14). In addition birth order of the child has been associated with poor growth; the higher the birth order, the higher the risk of malnutrition(Reference Kikafunda and Namusoke15).
The main aims of the present analysis were to: (i) assess the associations between maternal HIV infection and growth outcomes of HIV-exposed but uninfected infants; and (ii) identify other potential determinants for poor growth among infants in Entebbe municipality and Katabi sub-county of Uganda.
Materials and methods
Study area and participants
The study was conducted in Entebbe municipality and Katabi sub-county as previously described(Reference Elliott, Kizza and Quigley16). It was a randomized, double-blind, placebo-controlled trial of anthelminthics in pregnancy, using albendazole v. placebo and praziquantel v. placebo in a 2 × 2 factorial design. The study was originally designed to examine effects of anthelminthic treatment during pregnancy on infant responses to immunization and infectious disease incidence. Results of the trial have been reported elsewhere and no effect of the trial intervention on growth parameters was observed(Reference Ndibazza, Muhangi and Akishule17, Reference Webb, Mawa and Ndibazza18). The study results reported herein represent a retrospective observational analysis of differences in growth at 1 year of age between HIV-exposed but uninfected infants and infants born to women who were HIV-negative. Socio-economic, maternal and infant characteristics were also evaluated for associations with growth outcomes at 1 year of age.
Women were recruited at the antenatal clinic at Entebbe Hospital between April 2003 and November 2005. They were eligible if they were well, resident in the study area, planning to deliver their baby at the hospital, willing to participate and willing to know their HIV status; and excluded if they had Hb <8 g/dl, clinically apparent severe liver disease, diarrhoea with blood in the stool, abnormal pregnancy, history of adverse reaction to anthelminthic drugs or had participated in the study during an earlier pregnancy.
Following delivery, children were followed up according to the study protocol. Morbidity data were collected prospectively at the research clinic. Malaria was diagnosed as fever (≥37·5°C) with Plasmodium falciparum parasitaemia. Data on asymptomatic parasitaemia were collected annually. Study nurses were trained on how to take and record the anthropometric measurements. Weight measurements for 1-year-olds were taken using CMS Weighing Equipment, model MP25 (Chasmors Ltd, London, UK). The children were lightly dressed wearing undergarments only, and wore a measuring trouser which was then suspended on a weighing scale. Measurement of recumbent length at age 1 year was done using an adjustable child-length measuring board (Seca; Vogel & Halke, Hamburg, Germany).
In 2002, prior to the start of the study, the research team in collaboration with Entebbe Hospital established a programme for prevention of mother-to-child HIV transmission at the hospital. In accordance with guidelines current at the time(Reference Guay, Musoke and Fleming19), women identified as HIV-positive were offered a single dose of nevirapine for themselves and their infants, to be taken during labour and after delivery, respectively. HIV-positive women were provided with cotrimoxazole daily for prophylaxis of opportunistic infections; HIV-exposed infants were provided with cotrimoxazole syrup from age 6 weeks until their HIV status was assessed at age 18 months; at this time, cotrimoxazole was discontinued for HIV-negative children. HIV-positive infants were referred for care at nearby specialist centres. Highly active antiretroviral therapy (HAART) for treatment was not widely available in Uganda at the time when women were recruited to the study; however, five HIV-positive women were recorded as taking HAART during pregnancy.
HIV diagnosis
Mothers’ HIV status was assessed by a rapid test algorithm before delivery as previously described(Reference Muhangi, Woodburn and Omara20). Vertical HIV transmission was diagnosed by RNA and DNA PCR at age 6 weeks as previously described(Reference Webb, Mawa and Ndibazza18) and by rapid test at age 18 months. Infants were regarded as being HIV-positive if the 6-week sample had a positive DNA PCR for any of the viral regions examined and a viral load of 1000 copies per millilitre or more, or if the rapid test at 18 months of age was positive.
Blood samples were obtained from HIV-infected pregnant women. CD4 cell counts were ordered as part of the baseline measurement and done using a BD FACScount™ flow cytometer (Becton Dickinson, San Jose, CA, USA). Maternal immunological status was dichotomized based upon results of CD4 testing performed at enrolment, with women having a CD4 count of ≤350 cells/mm3 being categorized as having poor immunological status.
Outcome measures
The outcome variables for the present study were the three continuous anthropometric measures, length-for-age Z-score (LAZ), weight-for-age Z-score (WAZ) and weight-for-length Z-score (WLZ), derived by importing growth parameters into WHO Anthro software version 3 (April 2009). The continuous value of each of the three anthropometric measures was also used to derive three binary variables reflecting the presence or absence of a Z-score <−2, i.e. stunting (LAZ <−2), underweight (WAZ <−2) and wasting (WLZ <−2), which were then also used as dependent variables for the study objectives.
Statistical methods
Data management was done using the Microsoft® Access database designed for the anthelminthic trial. Statistical analyses were performed using the STATA statistical software package version 10. The second(Reference Ndibazza, Muhangi and Akishule17) or third infant(Reference Webb, Manji and Fawzi1) in a twin or triplet pregnancy was excluded from the analysis but five women on HAART were retained.
In order to evaluate the associations between HIV exposure, other factors and growth outcomes, we used as response variables the continuous Z-scores and the binary variables for stunting, underweight and wasting defined above.
Univariable analysis was first performed. For the continuous outcomes linear regression was used to investigate whether HIV exposure and other risk factors were associated with the mean Z-score; mean differences in Z-score and 95 % confidence intervals were used to quantify the association between each risk factor and growth. For the binary outcomes, logistic regression was used to investigate whether HIV exposure and other risk factors were associated with the prevalence of stunting, underweight and wasting; odds ratios were used to quantify the association between each risk factor and poor growth. Multivariable analyses were then conducted investigating factors considered likely to be potential confounders, including maternal age, maternal education, maternal income, household socio-economic status (a score based on building materials, number of rooms and items owned) and episodes of malaria during infancy; variables assumed to be potential confounders and variables that presented a P value of <0·05 in the univariable analysis were included in multivariable regression models. Significance levels in the final model were determined using likelihood ratio tests. A similar approach was used to investigate other risk factors for stunting, underweight and wasting in this population. The following factors were considered: sex of the child, low birth weight, low maternal CD4 cell count, number of living children in the family and early weaning (defined as introducing cow's milk at or before the age of 6 weeks).
Results
There were 2507 mothers enrolled and 2345 live births, with 2092 children under follow-up throughout the first year of life(Reference Webb, Mawa and Ndibazza18); 1701 children were seen at age 1 year, of whom 1554 had data on HIV exposure status and growth parameters. Retaining only one infant from twin or triplet pregnancies reduced this number to 1536. Mothers of children who did not have growth data available were, on average, younger than mothers whose children did have growth data (P < 0·001), more likely to be single mothers (P = 0·001), more likely to be HIV-positive (P < 0·001) and more likely to have malaria at enrolment (P < 0·001) There were no other differences between mothers whose children provided growth data at 1 year and mothers whose children did not contribute to growth data (data not shown).
Of the 1536 children, 1380 (89·8 %) were not exposed to HIV, 122 (8·0 %) were exposed but not infected, nineteen (1·2 %) were exposed and infected, fifteen (1·0 %) were exposed but their HIV infection status was not available. Of the nineteen exposed and infected, twelve infants tested positive for HIV at 6 weeks and at 18 months, five infants tested negative at 6 weeks but were HIV-positive at 18 months, two infants did not have a 6-week test result but tested positive for HIV at 18 months. The infected children (nineteen) and those whose status was not available (fifteen) were excluded, leaving a total of 1502 children for the present analysis.
Characteristics of the study participants, stratified by HIV exposure status, are shown in Table 1. HIV-positive mothers were more likely to be older, to have primary or no formal education, to be widowed/divorced/separated and to have had five or more pregnancies than HIV-negative mothers in this population. The CD4 count at enrolment ranged from 19 to 1473 cells/mm3, with a median of 538 (interquartile range 345–710) cells/mm3. HIV-exposed uninfected children had a significantly lower mean birth weight than unexposed children (P = 0·028).
Table 1 Maternal and child characteristics for HIV-unexposed and HIV-exposed but uninfected infants, Entebbe municipality and Katabi sub-county, Uganda, 2003–2005
MV, missing values.
*Values are presented as mean and standard deviation.
†Values are presented as number and percentage.
‡Values are presented as median and interquartile range (IQR).
Associations between maternal HIV and growth outcomes
The overall mean LAZ was −0·82, WAZ was −0·31 and WLZ was 0·12. Prevalence of stunting, underweight and wasting was 14·2 %, 8·0 % and 3·9 %, respectively. At age 1 year, HIV-exposed uninfected babies had a lower mean LAZ (adjusted mean difference −0·16; 95 % CI −0·39, 0·06; P = 0·16; Table 2) and a somewhat lower mean WAZ than HIV-unexposed babies (adjusted mean difference −0·13; 95 % CI −0·35, 0·09; P = 0·23; Table 3). There was also little difference between mean WLZ in the two groups (adjusted mean difference −0·04; 95 % CI −0·27, 0·19; P = 0·74; Table 4). Analysis of the binary variables showed that HIV exposure was significantly associated with underweight (P = 0·006; Table 3), but not with stunting or wasting. In other words, HIV-exposed but uninfected children tended to be shorter than HIV-unexposed infants, but their weight was proportionate to their length.
Table 2 Association of HIV exposure and other factors with stunting* in 1-year-olds, Entebbe municipality and Katabi sub-county, Uganda, 2003–2005
LAZ, length-for-age Z-score; Ush, Uganda shillings.
* Stunting: a chronic restriction of growth in length/height indicated by a low length-for-age (LAZ <−2).
†Adjusted differences and odds ratios estimated from multivariate linear and logistic regression models that included maternal age, maternal education, income, household socio-economic status, episodes of malaria and HIV exposure.
‡Early weaning: introducing cow's milk at or before the age of 6 weeks.
§Household socio-economic status is a score based on building materials, number of rooms and items owned (‘1’ representing lowest and ‘6’ representing highest status).
Table 3 Association of HIV exposure and other factors with underweight* in 1-year olds, Entebbe municipality and Katabi sub-county, Uganda, 2003–2005
WAZ, weight-for-age Z-score; Ush, Uganda shillings.
* Underweight: a reflection of low weight-for-age (WAZ <−2).
†Adjusted differences and odds ratios estimated from multivariate linear and logistic regression models that included maternal age, maternal education, income, household socio-economic status, episodes of malaria and HIV exposure.
‡Early weaning: introducing cow's milk at or before the age of 6 weeks.
§Household socio-economic status is a score based on building materials, number of rooms and items owned (‘1’ representing lowest and ‘6’ representing highest status).
Table 4 Association of HIV exposure and other factors with wasting* in 1-year olds, Entebbe municipality and Katabi sub-county, Uganda, 2003–2005
WLZ, weight-for-length Z-score.
* Wasting: an acute weight loss indicated by a low weight-for-length (WLZ <−2).
†Adjusted differences and odds ratios estimated from multivariate linear and logistic regression models that included maternal age, maternal education, income, household socio-economic status, episodes of malaria and HIV exposure.
‡Early weaning: introducing cow's milk at or before the age of 6 weeks.
Associations between other factors and growth outcomes
Young maternal age, low maternal education, low birth weight, male sex and repeated episodes of malaria in infancy were associated with an increased odds of stunting (Table 2), while being a male child, low birth weight, early weaning and having malaria during infancy were associated with being underweight (Table 3). Higher number of living children in the family, low birth weight and low maternal income were associated with wasting (Table 4). Maternal helminth infection was not associated with poor growth in this population (data not shown); as previously reported, anthelminthic treatment showed no effect on growth during infancy(Reference Webb, Mawa and Ndibazza18).
Associations between growth outcomes and maternal CD4 cell count were explored among infants of HIV-positive mothers by categorizing enrolment CD4 count into two categories: ≤350 cells/mm3 and >350 cells/mm3 (Tables 2, 3 and 4). Low maternal CD4 cell count increased the odds of both stunting and underweight but the association was not significant (Tables 2 and 3). The average breast-feeding duration for HIV-unexposed infants (mean 16·73 (sd 5·48) months) was significantly longer than that for HIV-exposed uninfected infants (mean 7·80 (sd 6·07) months; P < 0·001). Results from subgroup analysis showed that breast-feeding duration was not associated with growth outcomes.
Discussion
We have observed that maternal HIV infection is strongly associated with increased odds of being underweight, but not with stunting or wasting, among HIV-exposed uninfected infants in this population. The underlying mechanism is not fully understood and may be multifactorial. The association with underweight but not stunting or wasting indicates chronic nutritional deficiency, as opposed to a sudden effect of acute illness or deprivation. Our study finding that HIV exposure is associated with underweight among HIV-exposed uninfected children corroborates the findings of other studies from Sub-Saharan Africa(Reference Makasa, Kasonka and Chisenga5).
Growth faltering among HIV-exposed but uninfected infants could be caused indirectly by HIV-related symptoms in the mother: HIV may incapacitate the mother and this might translate into less provision of care to her child. HIV-positive women with low CD4 cell counts may be at greater risk of postnatal depression(Reference Kapetanovic, Christensen and Karim21, Reference Stewart22), and this could also influence the care that they are able to provide for their children at a critical stage in their development. Our analysis on measures of maternal immune status showed that 28 % of HIV-positive mothers had a CD4 count of ≤350 cells/mm3, representing a category of mothers who qualify for antiretroviral therapy (ART) according to the most recent recommendations(23); low maternal CD4 cell count increased the odds of both stunting and being underweight (although weakly), but was not associated with wasting.
It has been suggested that infections may be more common among HIV-exposed uninfected children; if so, such infections could contribute to poor growth(Reference Filteau24), but this was not the case in our study. The study clinic provided cotrimoxazole prophylaxis to all HIV-exposed uninfected infants and this could have helped to diminish infection risks. In multivariable models, following our modelling strategy we chose to adjust for malaria episodes in infancy as it was strongly associated with growth outcomes and hence may have improved the precision of estimates from linear regression. It is also likely to be a proxy for maternal malaria (which we considered likely to be a confounder of the association between HIV exposure and growth outcomes). Theoretically it is possible that malaria in infancy could be on the causal pathway between HIV exposure and growth outcomes, but this was not the case in our study due to the provision of cotrimoxazole for all HIV-exposed infants, so that HIV exposure was not in fact associated with infant malaria. Therefore inclusion of malaria episodes in infancy in the multivariable model had little impact on the adjusted associations between HIV exposure and growth outcomes.
Contrary to some of our study findings, researchers in Europe(Reference Newell, Borja and Peckham2) reported that exposure to maternal HIV infection is not associated with child's poor growth. This difference could be due to the fact that in Europe the HIV-infected population is more like the general population and less socio-economically disadvantaged than that in Uganda. This suggests that the associations that we and researchers in Zambia(Reference Makasa, Kasonka and Chisenga5) have observed in HIV-exposed but uninfected children may not be due to the infection itself, but rather to concomitant social or behavioural factors. It is also possible that the association we have observed between HIV exposure and underweight is a chance finding; this possibility is supported by the fact that analysis of the corresponding continuous WHZ outcome did not show an association with HIV exposure.
Uninfected children born to HIV-infected women are exposed antenatally to ART and this might affect their growth in early life. A study conducted in Thailand found that, although neither the total duration of exposure to zidovudine nor the duration of postnatal exposure was associated with infant growth from 6 weeks to 18 months, a longer in utero exposure to zidovudine had a negative impact on birth weight(Reference Briand, Le Coeur and Traisathit25). A study in Europe found a minimal effect of combination ART on growth up to 18 months of age(Reference Hankin, Thorne and Newell26). Other findings in Europe showed that ART exposure was not significantly associated with prevalence of congenital abnormalities or low birth weight(27). But ART exposure was associated with anaemia in infants, with Hb levels reverting to normal by 3 months of age(27). However, in our setting, single-dose nevirapine for mother and child was standard of care at the time of the study and prolonged exposure of the HIV-exposed but uninfected infants to ART in infancy or in utero was rare.
Strengths of the study are its large sample size and cohort design, meaning that data on exposures of interest and potential confounders were collected prospectively. Possible limitations of our study include the fact that the mothers and infants included in the present analysis were not representative of the study population, with children of HIV-infected women being less likely to provide growth data than children of HIV-uninfected women. Moreover, since HAART regimens were not standard of care during pregnancy in the study setting at the time, we are unable to comment on the impact of triple HAART regimens on infant growth.
On the other hand, in our study (as recommended in this setting), HIV-exposed but uninfected infants routinely received cotrimoxazole from 6 weeks to 18 months of age: a folate antagonist which can cause anaemia and might possibly interfere directly with growth. We found that HIV-exposed uninfected babies were more anaemic than HIV-unexposed, but the difference was not significant (data not shown).
As regards determinants of stunting or underweight other than HIV exposure, our results complement findings of previous studies(Reference Hien and Hoa13, Reference Kabubo-Mariara, Ndenge and Mwabu28) that being a male child is associated with a higher risk of both stunting and underweight. Other studies have also found that young maternal age is a strong predictor for stunting and being underweight(Reference Linnemayr, Alderman and Ka29) although results are not consistent(Reference Hien and Hoa13). Low maternal education, low birth weight and early weaning were associated with increased odds of stunting and underweight in our study, consistent with findings of earlier studies(Reference Cattaneo, Timmer and Bomestar9, Reference Bloss, Wainaina and Bailey30, Reference Mamiro, Kolsteren and Roberfroid31). As expected, the odds of stunting and being underweight increased significantly with the number of malaria episodes in infancy. This is in agreement with earlier reports that episodes of illness are associated with appetite loss, vomiting and poor nutrient utilization, which exacerbates malnutrition and perpetuates the cycle(Reference Devereux, Baulch and Phiri32). Number of living children was not associated with stunting or underweight but significantly increased the odds of wasting, consistent with the hypothesis that the oldest child is likely to be the least malnourished and that subsequent children are increasingly poorly provided for.
Conclusions
Maternal HIV infection was associated with being underweight in HIV-uninfected infants in this population. The underlying mechanism is not fully understood and may be multifactorial. Additional independent predictors of stunting or underweight in this population included young maternal age, low maternal education, low birth weight and malaria in infancy. In addition to effective interventions for prevention of mother-to-child HIV transmission, children born to HIV-infected mothers also require special care to ensure that their growth and development are not impaired. Prevention and treatment of undernutrition in children remain critical interventions in settings with high HIV prevalence.
Acknowledgements
Sources of funding: The study was funded by the Wellcome Trust (grant number 064693). Conflicts of interest: The authors have no conflict of interest to declare. Ethics: Ethical approval for the Entebbe trial of de-worming during pregnancy was given by the Science and Ethics Committee, Uganda Virus Research Institute; the Uganda National Council for Science and Technology; and the London School of Hygiene & Tropical Medicine. Authors’ contributions: S.A.L., H.M., J.N. and M.N. carried out the interviews, collected the data and contributed to revising the paper. M. Kizza and E.N. facilitated data collection and supported the fieldwork. M. Kihembo supported laboratory work especially the measurement of maternal CD4 cell counts. L.M. planned and undertook the analysis, undertook the interpretation of the data and drafted the manuscript. E.L.W. contributed to analysis, interpretation of data and revising the paper. A.M.E. designed the study, provided overall research supervision, and contributed to analysis, interpretation of data and revising the paper. Acknowledgements: The authors thank the study participants, the fieldwork team, Entebbe Hospital and the MRC/UVRI Uganda Research Unit on AIDS for their support. They also thank Professor Jonathan Levin who gave useful technical comments and advice.
Open access
