Corresponding Author Meera K. Chhagan, Department of Pediatrics and Child Health, University of KwaZulu-Natal, P/Bag 7, Congella 4013, South Africa. Tel.: +27 31 2604352; Fax: +27 31 2604388; E-mail: email@example.com
Objectives To describe baseline nutritional anthropometry and its determinants in a cohort of children commencing HAART, and subsequent longitudinal anthropometric trajectories over 2 years.
Methods Prospective observational study in a prepubertal cohort of children commencing HAART in Durban, South Africa.
Results Among 151 children with a median baseline age of 61.3 months (IQR 29.6, 90.1), prevalence of stunting was 54% (95% CI 46, 62) and of underweight, 37% (95% CI 29, 45). There was high prevalence of preceding respiratory and diarrhoeal comorbidities, which were associated with poorer anthropometry. There were significant improvements in height, weight and mid-upper-arm circumference z-scores after initiation of HAART regardless of preceding comorbidities. Stunted children remained shorter on average after 24 months, but younger children had better catch-up. Children who eventually died had persistently worse anthropometry. Children who were exposed to improved programs for prevention of mother-to-child transmission (PMTCT) were younger and had more severe growth impairments at baseline.
Conclusion Anthropometric status of children on HAART is influenced by age, preceding comorbidities, and by programmatic factors. With improved PMTCT programs, infants who would previously have died in infancy are now surviving to commence HAART. Poor outcomes are preceded by persistent anthropometric failure on HAART reaffirming the need for growth velocity monitoring.
Objectifs: Décrire l’anthropométrie nutritionnelle de base et ses déterminants dans une cohorte d’enfants commençant la thérapie antirétrovirale hautement active (HAART) et les trajectoires longitudinales anthropométriques subséquentes sur deux ans.
Méthodes: Etude observationnelle prospective d’une cohorte d’enfants pré-pubères commençant l’HAART à Durban, en Afrique du Sud.
Résultats: Parmi 151 enfants avec un âge médian de base de 61,3 mois (IQR: 29,6–90,1), la prévalence d’un retard de croissance était de 54% (IC95%: 46–62) et celle de l’insuffisance pondérale, 37% (IC95%: 29–45). Il y avait une forte prévalence de co-morbidités précédentes respiratoires et diarrhéiques, qui étaient associées à une mauvaise anthropométrie. Il y avait des améliorations significatives dans les z-scores pour la taille, le poids et le périmètre brachial après l’initiation de l’HAART, indépendamment des co-morbidités précédentes. Les enfants avec un retard de croissance sont restés de taille plus courte en moyenne après 24 mois, mais les plus jeunes enfants ont effectué un meilleur rattrapage. Les enfants qui sont décédés par la suite avaient une anthropométrie persistante pire. Les enfants qui étaient exposés à de meilleurs programmes de prévention de la transmission mère-enfant (PTME) étaient plus jeunes et avaient des problèmes de croissance plus graves au départ.
Conclusion: Le statut anthropométrique des enfants sous HAART est influencé par l’âge, les co-morbidités précédentes et les facteurs programmatiques. Avec de meilleurs programmes PTME, les nourrissons qui auparavant seraient décédés à bas âge survivent maintenant pour commencer l’HAART. Les mauvais résultats sont précédés par un échec anthropométrique persistant sous HAART, ce qui réaffirme la nécessité d’une surveillance de la vitesse de croissance.
Objetivos: Describir la antropometría nutricional de base y sus determinantes en una cohorte de niños que comienzan TARGA, y la subsecuente trayectoria antropométrica longitudinal a lo largo de dos años.
Métodos: Estudio observacional prospectivo de una cohorte de niños prepúberes que comienzan TARGA, en Durban, Sudáfrica.
Resultados: Entre 151 niños, con una mediana de edad de 61.3 meses al comienzo del estudio (IQR 29.6, 90.1), la prevalencia de retraso en el crecimiento era del 54% (IC 95% 46, 62) y del bajo peso del 37% (IC 95% 29, 45). Había una alta prevalencia de co-morbilidades respiratorias y diarreicas precedentes, lo cual estaba asociado a una antropometría más pobre. Había mejoras significativas en los Z-scores para altura, peso y la circunferencia superior del brazo tras haber iniciado la TARGA, independientemente de las comorbilidades precedentes. Los niños con retraso en el crecimiento se mantuvieron con una menor talla durante un promedio de 24 meses, pero los más jóvenes se reponían más rápidamente. Los niños que eventualmente murieron tenían una mala antropometría persistente. Los niños expuestos a programas mejorados de prevención de la transmisión vertical eran más jóvenes, y tenían retrasos en el crecimiento más severos en el momento de iniciar el estudio.
Conclusión: El estatus antropométrico de los niños recibiendo TARGA está influenciado por la edad, las comorbilidades precedentes y los factores programáticos. Con programas mejorados para la prevención de la transmisión vertical, los lactantes que previamente habrían muerto en sus primeros años, ahora sobreviven y comienzan TARGA. Unos malos resultados están precedidos por fallos antropométricos persistentes recibiendo TARGA, reafirmando la necesidad de monitorizar la velocidad del crecimiento.
Growth failure among HIV-infected children (Venkatesh et al. 2010) and anthropometric recovery subsequent to antiretroviral treatment initiation (HAART) have been described in African cohorts (Sutcliffe et al. 2011; Van Dijk et al. 2011). Findings differ across studies, perhaps because of program factors and individual factors. Program factors, for example, prevention of mother-to-child transmission (PMTCT) and early initiation of HAART in infancy, are anticipated to result in a healthier and younger cohort commencing HAART and hence in a more favourable subsequent course (Violari et al. 2008). Younger children have better potential for catch-up growth than older children. Specific drug regimens used in different age categories may also influence subsequent growth. Study findings and interpretation are influenced by design, with many arising from retrospective analyses of routine clinical data extracted for surviving members of treatment cohorts. With analyses of routine clinical data, validity may be compromised by issues such as measurement error, which is difficult to control in clinical settings with high staff turnover. Longitudinal growth data are especially susceptible to combined influences of measurement error at multiple occasions leading to cumulative error variances. We therefore conducted a prospective study of the growth and metabolic responses to HAART, using well-standardized anthropometric data to better describe the nutritional changes and explore factors that may determine this response.
The aims of this report are (i) to describe baseline characteristics and attained nutritional status of a cohort of South African children at the time of initiation of HAART; (ii) to describe the association of pre-baseline factors with baseline nutritional status; (iii) to describe longitudinal anthropometric trajectories during the first 2 years on HAART; and (iv) to describe associations between preceding comorbidities and subsequent growth trajectories.
In 2004, the South African Department of Health implemented the Comprehensive Plan for HIV Care. HAART was offered to adults and children in the public sector as per National Guidelines (National Department of Health, 2004). As experience in treatment of children accumulated, younger children and more of those with concurrent morbidities were being initiated rather than the older stable cohort of surviving children. This report arises from a study that recruited participants between September 2006 and July 2009, a period when the profile of HIV-infected children commencing HAART had already changed from that of an older stable cohort to one where infants were being identified earlier for treatment.
Recruitment and enrolment
We conducted a prospective observational study of a cohort of children commencing HAART at the King Edward VIII Hospital (KEH), a regional/semi-tertiary urban hospital in Durban, South Africa. It was among the first pilot sites implementing the HIV treatment program in 2004. Entry was either through the PMTCT program or via referral from paediatric services of various hospitals. Being a pilot site meant that patients who were initiated on treatment were from a wide catchment area beyond the geographic boundaries of the district.
Research staff identified eligible participants daily through consultation with routine clinical staff. Patients screened to commence treatment were simultaneously informed about the study and their willingness to participate ascertained. Inclusion criteria were: children commencing HAART who did not have any secondary sexual characteristics plus were aged below 11 years if girls, and below 12 years if boys. Attending clinicians determined whether children met criteria to commence HAART according to national guidelines. Only children who were treatment naïve at time of initiation, other than exposure to nevirapine in the PMTCT program, and who had a parent or guardian able and willing to provide informed consent/assent and comply with study schedule were enrolled. None of the children were participating in any therapeutic trials while enrolled in this observational study.
Follow-up and measurements
Besides routine clinical assessments, each child underwent a detailed clinical assessment by a research doctor prior to and during treatment. The assessment included a standardized measurement of weight, height or recumbent length (if below 2 years old or unable to stand), and mid-upper-arm circumference (MUAC). Intra-observer and inter-observer reliability (between the research doctor and principal investigator) was established at the start of the study with reliability statistics within the expected range for growth studies. All measurements were taken in duplicate and the average used in analyses. At baseline, details on current and past morbidities and hospitalizations were recorded based on a combination of caregiver recall and hospital record review. Children were followed up at 1, 3, 6, 9, 12, 18 and 24 months after initiation of HAART. Children were initiated on one of two regimens: nucleoside reverse transcriptase inhibitors (lamivudine plus stavudine) and non-nucleoside reverse transcriptase inhibitor (efavirenz) for those over 3 years and weighing above 10 kg; or lamivudine, stavudine and ritonavir-boosted lopinavir (LPV/r) for those aged 3 years or younger. According to clinical protocols, all children received micronutrient supplements at approximately 1RDA, and those who were underweight for age received additional food supplements with follow-up by a dietician. The locally formulated supplement was a fortified, amylase-enriched maize product with energy density of 1848 kJ/100 g. It could be re-constituted to drinkable or porridge consistency or added to other foods. Adherence to HAART was assessed through caregiver self-report, paper diary and pill counts at each visit. Virologic suppression was defined as achieving a plasma viral load below 1000 copies/ml after 6–12 months of HAART.
We used EpiData version 3.1 for data entry and verification, and SAS V9.2 (SAS Institute, Inc., Cary, NC, USA) for analyses.
Baseline characteristics of the study sample were graphically displayed then summarized using a combination of simple frequency distributions for categorical variables, and means and standard deviations (SD) or medians with interquartile ranges (IQR) for continuous variables. Normalizing transformation was applied where appropriate to continuous variables. Proportions were compared using the chi square or Fisher’s exact test. Point estimates are reported with 95% confidence intervals.
Anthropometric z-scores were computed using the WHO international Growth Standards for children up to 5 years age and the 2007 reference for weight, height and BMI for children 5–19 years old (De Onis et al. 2007; WHO, 2007). Children with z-scores below −2 for height/length-for-age (HAZ) or weight-for-age (WAZ) were classified as being stunted or underweight, respectively (WHO, 2007). MUAC for height z-scores was computed using the reference for international use recommended by a WHO Expert Committee (Mei et al. 1997). The main outcome for these analyses was attained mean HAZ or WAZ over 24 months post-initiation of HAART. Longitudinal growth trajectories were constructed using mixed effects regression models. Initial models included time since HAART initiation as the main predictor. We included current age in all models to account for differing growth velocity at different ages because children entered the cohort at different ages. Subsequently, models that included baseline and pre-baseline morbidity, and indicators for other potential determinants of longitudinal growth were explored. A spatial covariance structure was used to model correlations within children. A significant overall F-test and a significant effect of time were used to assess whether there was longitudinal change in anthropometry z-scores over time. In subsequent models, a significant effect of predictor, such as pre-baseline morbidity, and predictor–by-time interaction at type I error level of 0.05 were used to assess effect of predictor on longitudinal growth. We also explored, using higher order interaction terms, whether factors such as baseline CD4 percentage modified the relationship between various predictors and longitudinal growth. Growth trajectories were displayed graphically by fitting smoothed cubic splines. We corrected for regression to the mean (RTM) by adjusting for baseline value of anthropometry and using the first post-baseline observation in our outcomes rather than the baseline itself. The first post-baseline observation was measured within 2–4 weeks of the baseline and had a strong linear correlation (rho) of 0.9 with the baseline observation.
Based on anticipated baseline weight z-scores, we estimated that over a 2-year period and with a sample size of 120, we would have 90% power to detect a mean weight z-score change of 0.5 (SD of change of 1.7) at a significance level (alpha) of 0.05 using a two-sided one-sample t-test (Bobat et al. 2001). Similarly, we anticipated 80% power to detect a mean height z-score change of 0.3 (SD of change of 1.2). We had anticipated 20% attrition because of death, loss to follow-up or discontinuation of treatment.
The Ethics Review Committee of the University of KwaZulu-Natal approved the study. The KwaZulu-Natal Department of Health and medical management of King Edward VIII Hospital granted permission for conduct of the study.
Pre-baseline and baseline characteristics
From September 2006 to July 2009, we screened 165 patients, of whom 154 were eligible. A total of 151 children were eventually recruited (response rate of 98.1%) (Figure 1). The cohort consisted of 74 (49%) girls and 77 (51%) boys. Baseline age ranged from 6.9 to 159.3 months (Table 1). The primary caregiver was the biological mother in 73 (49%) and a grandmother in the same proportion. 35 children (23%) were maternal orphans. Comorbidities were highly prevalent among mothers with 13 (9%) currently on TB treatment and 34 (23%) completing TB treatment in the past 2 years.
Table 1. Baseline clinical characteristics of cohort (n = 151) at time of initiation of HAART
Mean (SD) or median (IQR)
Data are expressed as median (IQR), except for log viral load (mean ± SD).
Birth weight (kg)
3.1 (2.5, 3.4)
61.3 (29.6, 90.1)
−1.26 (−2.24, −0.42)
Length- or height-for-age z-score
−2.05 (−3.10, −1.39)
0.11 (−1.18, 1.01)
Mid-upper-arm circumference for height z-score
−0.94 (−1.94, 0.07)
12.6 (8.0, 19.1)
Log10 viral load
Median birth weight of the children was 3.1 kg (range 1.3–4.2 kg). 43 (28%) were being treated for tuberculosis at the time of initiating HAART, while 58 (38%) had a history of tuberculosis at any time in the past. Within the preceding 3 months, 42 children (28%) had been hospitalized, and 34 (22.5%) had a clinically relevant pathogenic organism identified from sites that included blood, urine, cerebrospinal fluid, stool or sputum. Preceding 3-month morbidities in these children were pneumonia (53%), diarrhoea (66%) and confirmed meningitis (4%). There were moderate to severe derangements in baseline weight, height, MUAC z-scores and laboratory parameters, but not in BMI z-scores (Table 1). Stunting was evident in 54% (95% CI 46%, 62%) and underweight in 37% (95% CI 29%, 45%). The majority were classified as having WHO HIV clinical stage 3 (70%) and stage 4 (19%) at baseline. Anthropometric z-scores were lower in those classified as having stage 4 disease even though the level of the z-score was not directly used for staging purposes (Table 2). There was a trend towards lower baseline anthropometric z-scores in those with comorbidities in the preceding 3 months (Table 2). We were unable to find associations between maternal factors (mother alive or ill) and baseline anthropometry. Children who were stunted at baseline had slightly lower mean birth weight (2.91 kg, 95% CI 2.76, 3.07) than non-stunted children (3.14 kg, 95% CI 2.98, 3.29).
Table 2. Distribution of baseline anthropometric z-scores by HIV disease stage and by preceding morbidities
Maternal single-dose nevirapine received intrapartum
−2.76 (−3.22, −2.30)
−1.61 (−2.10, −1.11)
−0.54 (−1.22, 0.13)
−1.94 (−2.26, −1.62)
−1.32 (−1.63, −1.01)
−1.07 (−1.35, −0.80)
Among children with available history of maternal single-dose nevirapine for prevention of perinatal HIV transmission, those with a positive history had significantly lower mean HAZ than those with a negative history (Table 2). Nevirapine-exposed children were also younger (26 months, 95% CI 20, 31 months) than those with a negative history (72 months, 95% CI 64, 79 months). All 30 mothers who received nevirapine were still alive, while 20% (19/97) of those who had not received nevirapine had died by the time the child was initiated on HAART (P < 0.01). Based on age- and weight-based criteria, 37 children (24%) commenced a LPV/r-based and 117 (76%) commenced an efavirenz-based first-line regimen.
Six children died during follow-up and 13 were lost to follow-up (Figure 1). Eleven children were transferred to other treatment sites for residential proximity. Of those with adequate data on virologic suppression in the first year of treatment, 17/127 had incomplete suppression. Information on adherence to HAART was available for 928 visits with adequate compliance documented in 94% of visits.
Anthropometric trajectories subsequent to HAART initiation
On average, mean WAZimproved by a 0.86 z-score and height-for-age by 1.37 z-score over 24 months (P < 0.001 for effect of time since treatment initiation). There was consistent increase in mean WAZ of the cohort over 24 months, with the steepest increase observed in the initial 6 months (Figure 2). In contrast, mean HAZ showed more gradual recovery, with greatest increase observed in the second year (Figure 2). These changes over time remained significant even after adjusting for age at initiation and baseline anthropometric z-scores. BMI and MUAC-for-height z-score (Figure 3) trajectories showed a similar pattern over 2 years, with increases in the initial 6 months reflecting the trend in WAZ, and the relative decline in the second year reflecting increasing HAZ. While there was a good correlation between MUAC-for-height and BMI z-scores (r = 0.57, P < 0.01), MUAC scores were generally one z-score lower than BMI.
Baseline stunting and comorbidities were associated with subsequent trajectories even after adjusting for age at initiation. Children who were stunted at baseline showed greater improvements than non-stunted children after commencing HAART but remained significantly shorter after 2 years of treatment (P < 0.001 for stunting by time interaction). Stunted children gained 1.4 z-scores in HAZ prior to correction for RTM, and 1.1 after correction. Non-stunted children gained 0.59 in HAZ prior to correction for RTM, and 0.29 after correction. Children who were underweight for age gained 1.45 z-scores in WAZ before correction for RTM, and 1.70 z-scores after correction. Underweight and non-underweight children had similar gains in WAZ.
In exploratory analyses, baseline CD4 percentage modified the relationship between stunting and longitudinal growth (P < 0.01 for 3-way interaction term of CD4 category, stunting and time). Among more immunocompromised children (CD4 percentage below 20%), both stunted and non-stunted children experienced marked improvements in z-scores with the difference between the two groups persisting up to 24 months. Among children who were less immunocompromised (CD4 percentage above 20%), the difference in mean HAZ between stunted and non-stunted groups diminished over 24 months of HAART with only stunted children showing increase in HAZ (Figure 4).
Children who experienced diarrhoeal illness in the 3 months preceding HAART initiation had a slower recovery of HAZ trajectory than those without diarrhoeal comorbidities (P = 0.040 for diarrhoea by time interaction), but caught up over 2 years on treatment if they survived. Similar trends were seen for those with pneumonia or those hospitalized in the 3 months preceding HAART initiation. Among children with preceding diarrhoeal illness, both stunted and non-stunted children experienced marked improvements in HAZ of almost 2 z-scores each. Among those without diarrhoeal illness, only stunted children experienced an increase in HAZ of almost one z-score. Children whose mothers received nevirapine started off shorter but caught up with the others by 2 years on treatment.
Age at HAART initiation influenced subsequent trajectories among stunted children (P = 0.002 for 3-way interaction between time, stunting and age-group). Stunted children aged below 3 years at baseline gained on average 1.3 (95% CI 0.7, 1.9) HAZ, while stunted children aged above 3 years gained on average 0.8 (95% CI 0.6, 0.9) HAZ over 24 months. Because age and HAART regimen were intricately linked, we did not perform separate analyses by regimen.
Children who eventually died had no detectable improvements in WAZ and HAZ trajectories over time and remained consistently below that of survivors. This was more marked for WAZ (Figure 5).
Discussion and conclusions
This cohort, like others, demonstrates that severe malnutrition, and more so stunting, is highly prevalent among HIV-infected children (Arpadi 2000; Bobat et al. 2001). General prevalence of stunting and underweight among children attending primary health facilities in this province have been reported as 24% and 12%, respectively (Schoeman et al. 2010). Longitudinal trends in anthropometry after commencing treatment are similar to those described recently in a Zambian cohort (Sutcliffe et al. 2011). In our study, the magnitude of linear growth seen among stunted children commencing HAART displayed inadequate catch-up over a 24-month period after correction for RTM effect (RTM). The Zambian cohort also identified persistent stunting in a fair proportion of children on HAART.
Our study expands on previous descriptions of longitudinal growth by investigating the contribution of pre-baseline factors. The relationship between preceding comorbidities and baseline anthropometry is not surprising, and partly reflects programmatic factors, in that children with more advanced stages of infection were being initiated rather than asymptomatic infants identified through early screening and detection. The association between HAZ and pre-baseline morbidities reinforces at clinical level some of the epidemiologic relationships between stunting and diarrhoeal disease and other child health outcomes. Although stunting is a major problem in HIV-infected children, it has not received as much attention in clinical algorithms as weight-based classifications. Similarly, MUAC measurements have not been widely used to monitor children prior to or during HAART.
Children with chronic malnutrition, as manifested by stunting at baseline, had some improvement on average, however, continued to have significantly lower HAZ by the end of 24 months of treatment. There may be several reasons for this observation, including persistence or consequence of comorbidities, or persistence of underlying social factors that contribute to poor nutrition in this group. Although nutritional supplementation and social support were part of routine care, it may be insufficient to address social poverty or food insecurity. We were unable to reliably measure consumption or sharing within households of food supplements, so did not attempt to measure their independent contribution to growth trajectories. Children with pre-baseline morbidities evidenced by history of hospitalization, diarrhoeal illness or confirmed infectious pathogens managed to catch-up anthropometrically with their peers. This suggests that HAZ trajectory may not be entirely explained by these prior morbidities. An alternative hypothesis is that chronic malnutrition and co-infections are associated with an altered hormonal milieu that delays increments in height. An interesting observation is that non-stunted children demonstrated clinically significant improvements in HAZ if they were severely immunocompromised (CD4 < 20%) or had experienced prior diarrhoeal illness. This gives some indication of the magnitude of growth impairment already present with more advanced HIV or diarrhoeal disease, even among those who do not appear stunted at a given cross-section in time. The implication is that in this environment, besides low CD4 percentage and attained nutritional status, linear growth velocity trends and recent diarrhoeal illness also warrant careful evaluation for HAART initiation.
The relationship between nevirapine exposure and baseline anthropometry is most likely related to a cohort effect. The cohort born in the period prior to strengthened PMTCT programs was more likely to be older survivors of their original cohort by the time they reached access to HAART. At least half of their cohort would have died within the first 2 years of life from intrauterine and rapidly progressive intrapartum infections (Newell et al. 2004). The nevirapine-exposed group was by contrast reaching treatment initiation at a younger age, albeit symptomatic. This younger group most likely has characteristics of those that were censored from the former group, leaving behind mainly those with later postpartum infections. PMTCT programs thus modified the trajectory of a group that would in previous years have progressed to death within the first 2 years of life. This demonstrates relevance of contextual factors for children commencing HAART.
Intuitively one may anticipate the younger cohort from the improved PMTCT period to be ‘healthier’ at the time of initiation of HAART, but we have shown that this is not necessarily true, and may be related to stage of the PMTCT program itself with interventions not initiated sufficiently early in pregnancy. Progress in local PMTCT programs is anticipated to change the profile of HIV-infected infant cohorts yet again, so that even fewer intrauterine and intrapartum infections remain.
Closely linked to better PMTCT programmatically is younger age of HAART initiation. Like the Zambian cohort, we show that younger age (<3 years) at initiation was associated with better gains in HAZ. We had insufficient numbers to stratify by younger age groups. Our findings among HIV-infected children are consistent with some of the international observations on reversibility of childhood stunting among HIV-uninfected cohorts. This nevertheless demonstrates how programmatic contexts may influence baseline characteristics and subsequent responses to HAART. Unlike older children those aged below 3 years received a protease-inhibitor containing treatment regimen. Because of study design, we are unable to make any inferences about effects of specific drugs on longitudinal growth.
While the current report does not include comprehensive analysis of mortality outcomes, it demonstrates, similar to previous reports, that death was preceded by persistent failure to recover weight z-scores (Yotebieng et al. 2010). This was not merely a manifestation of timing of death because WAZ improvements occurred largely in the first 6 months following HAART among survivors. Growth velocity monitoring remains important in monitoring progress even after HAART initiation.
In summary, this study demonstrates the inextricable link between anthropometry and infectious comorbidities. Significant improvements in anthropometry are achieved once HAART is initiated, even among those with co-infections and advanced disease stages. However, persistent deficits in stunted children need further investigation from both biomedical and programmatic perspectives.
This study was funded by the National Institutes of Health/Fogarty International Center. Opinions represented herein are the sole responsibility of the authors and do not necessarily represent views of the NIH. We thank all participating children and caregivers for their time. We acknowledge support of management and staff at King Edward VIII Hospital, KwaZulu-Natal Department of Health, research assistants and research doctor, Dr Ramji.