Growth trajectories of breastfed HIV-exposed uninfected and HIV-unexposed children under conditions of universal maternal antiretroviral therapy: a prospective study

Growth trajectories of breastfed HIV-exposed uninfected and HIV-unexposed children under conditions of universal maternal antiretroviral therapy: a prospective study

Articles Growth trajectories of breastfed HIV-exposed uninfected and HIV-unexposed children under conditions of universal maternal antiretroviral the...

405KB Sizes 0 Downloads 2 Views

Articles

Growth trajectories of breastfed HIV-exposed uninfected and HIV-unexposed children under conditions of universal maternal antiretroviral therapy: a prospective study Stanzi M le Roux, Elaine J Abrams, Kirsten A Donald, Kirsty Brittain, Tamsin K Phillips, Kelly K Nguyen, Allison Zerbe, Max Kroon, Landon Myer

Summary Lancet Child Adolesc Health 2019; 3: 234–44 Published Online February 14, 2019 http://dx.doi.org/10.1016/ S2352-4642(19)30007-0 See Comment page 201 Division of Epidemiology and Biostatistics (S M le Roux MBChB, K Brittain MPH, T K Phillips MPH, K K Nguyen MPH, Prof L Myer PhD) and Centre for Infectious Disease Epidemiology and Research (CIDER), School of Public Health & Family Medicine (K Brittain, T K Phillips, L Myer), Department of Paediatrics and Child Health (K A Donald PhD, M Kroon FRCPaed(SA)), and Neuroscience Institute, Faculty of Health Sciences (K A Donald), University of Cape Town, Cape Town, South Africa; ICAP at Columbia University, Mailman School of Public Health (Prof E J Abrams MD, A Zerbe MPH) and Vagelos College of Physicians and Surgeons (Prof E J Abrams), Columbia University, New York, NY, USA; Division of Developmental Paediatrics, Red Cross War Memorial Children’s Hospital, Cape Town, South Africa (K A Donald); and Neonatal Medicine, University of Cape Town and Mowbray Maternity Hospital, Cape Town, South Africa (M Kroon) Correspondence to: Dr Stanzi le Roux, Division of Epidemiology and Biostatistics, School of Public Health and Family Medicine, Health Sciences Faculty, University of Cape Town, Cape Town 7701, South Africa [email protected]

Background Over 1 million HIV-exposed uninfected (HEU) children are born in sub-Saharan Africa annually. Little data exist on the risk of impaired growth in this population under current policies of universal maternal antiretroviral therapy (ART) with breastfeeding. We aimed to study the growth of breastfed HEU children born to women who initiated ART during pregnancy and compare their growth with that of breastfed HIV-unexposed (HU) children drawn from the same community. Methods A prospective cohort of HIV-uninfected and HIV-infected pregnant women, who were initiating ART, were enrolled at their first antenatal care visit in a primary care centre in Gugulethu, Cape Town, South Africa. HIV infected women were participants of the Maternal Child Health Antiretroviral Therapy (MCH-ART) study, and HIV-uninfected pregnant women were participants in the HIV-Unexposed-Uninfected (HU2) study. All women were followed up during pregnancy, through delivery, to the early postnatal visit, which was scheduled for the first week after birth. At this visit, eligible breastfeeding mother–child pairs were recruited for continuation of postnatal follow-up until approximately age 12 months. Child anthropometry was measured at around 6 weeks, and every 3 months from month 3 to month 12. Weight-for-age (WAZ), length-for-age (LAZ), weight-for-length (WLZ), head circumference-for-age, and body-mass index-for-age Z scores were compared between HEU and HU children longitudinally using mixed effects linear regression. At 12 months, proportions of HEU and HU children with moderate or severe malnutrition were compared cross-sectionally using logistic regression. MCH-ART is registered with ClinicalTrials.gov, number NCT01933477. Findings Between June, 2013, and April, 2016, 884 breastfeeding mothers and their newborn babies (HEU, n=471; HU, n=413) were enrolled into postnatal follow-up. Excluding 12 children who tested HIV positive during follow-up, 461 HEU and 411 HU children attended 4511 study visits in total, with a median of 6 visits (IQR 5–6) per child. Birth characteristics were similar (overall, 94 [11%] of 872 preterm [<37 weeks] and 90 [10%] small-for-gestational age [birthweight <10th percentile]). Median duration of breastfeeding was shorter among HEU than HU children (3·9 months [IQR 1∙4–12∙0] vs 9·0 months [IQR 3∙0–12∙0]). Although WAZ scores increased over time in both groups, HEU children had consistently lower mean WAZ scores than HU children (overall β –0∙34, 95% CI –0∙47 to –0∙21). LAZ scores decreased in both groups after 9 months. At 12 months, HEU children had lower mean LAZ scores than HU children (β –0∙43, –0∙61 to –0∙25), with a higher proportion of children stunted (LAZ score <–2: 35 [10%] of 342 HEU vs 14 [4%] of 342 HU children; odds ratio [OR] 2∙67, 95% CI 1∙41 to 5∙06). Simultaneously, overweight (WLZ score >2) was common in both groups of children at 12 months (54 [16%] of 342 HEU vs 60 [18%] of 340 HU children; OR 0∙87, 95% CI 0∙58 to 1∙31). Interpretation Compared with HU children, HEU children have small deficits in early growth trajectories under policies of universal maternal ART and breastfeeding. Large proportions of both HEU and HU children were overweight by 12 months, indicating substantial risks for early onset obesity among South African children. Although the longer-term metabolic effects of ART exposure in the context of childhood obesity warrants further investigation, addressing childhood obesity should be an urgent public health priority in this setting. Funding Eunice Kennedy Shriver National Institute of Child Health and Human Development, Elizabeth Glaser Pediatric AIDS Foundation, South African Medical Research Council, and the Fogarty Foundation. Copyright © 2019 Elsevier Ltd. All rights reserved.

Introduction Most of the 1·1 million HIV-exposed uninfected (HEU) children born annually reside in sub-Saharan Africa.1,2 Understanding the health risks and needs of this large and growing population of children is a crucial step towards optimising child health in the region. HEU children might be at increased risk of suboptimal growth compared 234

with HIV-un­exposed (HU) children.3,4 However, findings from sub-Saharan Africa have been inconsistent, and interpretation has been complicated by the rapidly changing landscape of HIV treatment and infant feeding recommendations over the past two decades. Before 2010, avoidance of breastfeeding was commonly promoted as a strategy for prevention of mother-to-childwww.thelancet.com/child-adolescent Vol 3 April 2019

Articles

Research in context Evidence before this study We previously did a systematic review of health outcomes for HIV-exposed uninfected (HEU) children, which included publications up to October, 2015. Using the same search strategy, we did a supplemental search of PubMed for original papers reporting growth outcomes among HEU children compared with HIV-unexposed (HU) children in sub-Saharan Africa, from Oct 1, 2015, to Sept 1, 2018, without language restrictions, using the search terms “HIV”, “mother”, “uninfected”, “child” and “growth”. This supplemental search identified nine additional articles for consideration. From the previous systematic review and recent publications, maternal health (including disease severity and access to antiretroviral drugs) and infant feeding emerged as crucial determinants of growth among HEU children. Most studies reported on HEU child growth outcomes before guidelines endorsed triple-drug antiretroviral therapy (ART) for all HIV-infected pregnant women irrespective of CD4 cell counts, and reported poor growth outcomes among HEU children in this context. Since 2010, WHO have moved to endorse breastfeeding with ART as the optimal feeding strategy for HIV-exposed infants in most resource-limited settings. Reports from prospective studies comparing growth of breastfed HU and HEU children born under universal maternal ART in sub-Saharan Africa were limited to a single publication from Nigeria. Added value of this study To our knowledge, this is the largest longitudinal comparison of early HEU and HU child growth in the context of universal maternal ART and breastfeeding. Our data show a very early onset of obesity among both HEU and HU children, alongside

transmission of HIV (PMTCT);3 currently, the recom­ mended feeding strategy for mothers with HIV in most African countries is breastfeeding while the mother is on triple-drug antiretroviral therapy (ART).5 In 2013, the WHO PMTCT guidelines changed, shifting to universal ART (treatment for all, irrespective of CD4 cell thresholds) for pregnant and breastfeeding women.6 Previously, ART was unavailable in Africa or limited to those with severe disease stages. In eastern and southern Africa, antiretroviral coverage for pregnant women is now around 93%, and 66% of all people living with HIV received ART in 2017.1 Although reports of HEU child growth outcomes have varied over the past 20 years, two important determinants of child growth can be identified in this context. Generally, longer breast­feeding has been associated with improved growth,7,8 whereas more advanced maternal HIV and absence of maternal ART have been associated with sub­optimal growth.3,4,9–11 For example, before the widespread availability of ART in Africa, HEU children were reported to have increased risks of stunting, underweight, wasting, or microcephaly, or a combination of these.3,4,7,12 Subsequently, a large study from Botswana13 showed an increased risk of stunting www.thelancet.com/child-adolescent Vol 3 April 2019

suboptimal linear growth in late infancy. These findings highlight the double burden of malnutrition that is increasingly prevalent across sub-Saharan Africa, with both obesity and stunting occurring in the same populations, including in our study population of young HEU children. We also found somewhat higher risks of stunting among HEU than otherwise similar HU children, with the highest risk observed among children who were both HEU and small for gestational age (ie, with evidence of intrauterine growth restriction). Implications of all the available evidence These are reassuring findings for comparative growth outcomes among breastfed HEU infants born appropriate-forgestational age under conditions of universal maternal ART. The increased risk of stunting observed among HEU children in our study appears to be predominantly among those with evidence of intrauterine growth restriction (ie, small for gestational age). This observation raises important questions regarding the aetiology of intrauterine growth restriction in the context of maternal HIV and ART exposure, which warrant further investigation. Concerningly, we also report a remarkably high occurrence of overweight and obesity among very young children in this setting. To date, the potential public health crisis of childhood obesity has not received much attention in the context of maternal HIV infection. Although the longer-term metabolic effects of ART exposure in the context of childhood obesity clearly requires further investigation, addressing infant and child obesity should be an urgent public health priority for South Africa, for both HIV-affected and HIV-unaffected families.

among HEU children exposed to ART in utero compared with HEU children exposed to only zidovudine, raising some concern about the long-term growth consequences of fetal ART exposure. By contrast, higher than expected weight and length velocities have been also been reported in subgroups of HEU children who are formula fed, raising concerns for risk of childhood obesity.9,14 Notably, the current evidence base for growth outcomes of HEU children in sub-Saharan Africa is largely limited to data from settings where ART was not universally available during pregnancy, or from predominantly formula-fed children. Few data exist on the growth of HEU children in the context of universal maternal ART with breastfeeding, which is the currently recommended standard of care in sub-Saharan Africa.5 We aimed to address this knowledge gap by studying the growth of breastfed HEU children born to women who initiated ART during pregnancy and comparing their growth with that of breastfed HU children drawn from the same community. Specifically, we hypothesised that, under conditions of universal maternal ART and breastfeeding, HEU children would have similar growth trajectories to those of HU children from the same community. 235

Articles

Methods

Study design and participants

See Online for appendix

In this prospective cohort study, we followed HIVuninfected, and HIV-infected pregnant women initiating ART, from first antenatal clinic visit through pregnancy to delivery, and with their breastfed children until approximately 12 months post partum. All HIV-infected women were participants of the Maternal Child Health Antiretroviral Therapy (MCH-ART) study, a multiphase study investigating strategies to optimise post-partum retention in care.15 All HIV-uninfected women were participants of the HIV-unexposed-uninfected (HU2) cohort study.16 Run in parallel with MCH-ART—using the same study staff, facilities, and procedures—the HU2 study was designed specifically to complement MCH-ART by providing a community-control comparison group of HIV-uninfected mothers and children.15 Study procedures have been published previously.15,16 Briefly, women were recruited from the antenatal clinic of a large primary care centre, the Gugulethu Midwife and Obstetric Unit, in the peri-urban township of Gugulethu, Cape Town, South Africa. The centre serves a population of roughly 350 000 people (30% antenatal HIV prevalence) with colocated antiretroviral services.17,18 Eligibility criteria for antenatal enrolment was largely similar for both studies, and included written, informed consent; maternal age of 18 years or older with a confirmed intrauterine pregnancy and HIV test result from the day of enrolment. HIV-infected women were limited to those initiating ART (tenofovir + emtricitabine + efavirenz) in pregnancy. Gestational age was estimated at first antenatal care visit with ultrasound, last menstrual period, and palpation. After delivery, breastfeeding mother-child pairs returning for the first postnatal visit (coinciding with the routine socalled cord-care visit at the obstetric unit) were eligible for further postnatal follow-up if they presented within 28 days (MCH-ART) or 3 months (HU2) after birth and could provide additional written, informed consent. For full inclusion and exclusion criteria see the appendix. Children who were diagnosed with HIV infection during the study period continued study follow-up, but are excluded from the growth analysis. Both studies used the same study staff and, apart from HIV-specific measures, the same study procedures. An abridged protocol is in the appendix. University of Cape Town Faculty of Health Sciences Research Ethics Committee approved the MCH-ART and HU2 studies (UCT-HREC: 567/2014, 451/2012).

Procedures Pregnancy and birth anthropometry were abstracted from patient records for all mother-child pairs enrolled into the postnatal follow-up phase of the study. From 6 weeks onwards, maternal weight and height, and child weight, length, and head circumference were measured by trained study staff, according to a standardised protocol based on WHO guidelines. Further study visits were scheduled every 3 months from age 3 months to 12 months. The 236

study clinician repeated anthropometric training for all staff at regular intervals, with structured, supervised competency assessments.The same staff did measure­ ments for both groups. Infant weight was measured to the nearest 10 g using a calibrated digital infant scale (MTB 20 Baby Scale, Adam Equipment, Milton Keynes, UK); length to the nearest 0∙25 cm using a rigid recumbent length board (ADE MZ 1013 Baby Length Measuring Rod, ADE, Hamburg, Germany); and head circumference to the nearest 0∙1 cm using a non-stretch tape measure. Study nurses tested all HIV-exposed children at age 6 weeks and 48 weeks using HIV-PCR testing (Roche COBAS AmpliPrep/COBAS TaqMan HIV-1 qualitative assay; Roche Molecular systems, Branchburg, NJ, USA).15 Women who had tested HIV negative at first antenatal visit were retested for HIV in the third trimester where possible, and approximately every 3 months during breast­ feeding, by local, community-based PMTCT services. In the event of maternal seroconversion, previously HU children were urgently tested for HIV. Women who seroconverted were immediately referred to the co-located ART clinic with their children for further management. Trained field workers administered standardised quest­ ion­­naires addressing maternal health, and psychosocial and behavioural factors at all study visits, and after delivery, child feeding and health. At first antenatal visit, risky drinking was defined as an Alcohol Use Disorders Identification Test for consumption score of 3 or higher; intimate partner violence as any physical, sex­ual, or psychological violence reported using the WHO violence against women question­ naire; and depression as a score of 13 or higher on the Edinburgh Postnatal Depression Scale.15,16 Feeding practices were assessed at all postnatal visits with 24 h maternal recall; the last study visit at which breastfeeding was reported was used as the date of breastfeeding cessation.15,16 Cumulative breastfeeding practices in the first 6 months of life were classified into exclusive (only breastmilk and pre­ scribed medication), predominant, or partial breast­feeding using WHO guidelines.18 At age 6 months and 12 months, maternal smoking was assessed using maternal recall of any cigarette smoking in the preceding month. Household food security was assessed at 12 months using a quest­ion­ naire adapted from the Household Food Insecurity Access Scale, Food and Nutrition Technical Assistance Project, and the Community Childhood Hunger Identification Project Index.19

Statistical analysis Allowing for 15% early censoring and 10–15% exclusive formula feeding, we estimated a priori that a sample size of 1100–1150 pregnant women would provide a minimum of 880 eligible children for follow-up, achieving 84% power to detect a 0·2 or greater absolute difference in Z scores between HEU and HU children (two-sided α=0∙05; SD of 1·0). We generated age-adjusted, sex-adjusted, and gestationadjusted weight-for-age (WAZ), length-for-age (LAZ), www.thelancet.com/child-adolescent Vol 3 April 2019

Articles

weight-for-length (WLZ), head circumference-for-age (HCAZ), and body-mass index-for-age (BMIZ) Z scores using anthropometric software from the Intergrowth21st project20 (birth anthropometry and postnatal growth of preterm infants until 64 weeks postmenstrual age) and WHO Multicentre Growth Reference Study (all other postnatal measures).21 As per WHO guidelines, Z scores higher than 3 or lower than –3 were reviewed and corrected in the event of data capture errors, and unexplained values higher than 5 (WAZ, HCAZ, BMIZ, and WLZ) or 6 (LAZ); or lower than –5 (HCAZ, BMIZ, and WLZ) or –6 (WAZ and LAZ) were excluded from analysis (<1% of all measurements).21 Rapid weight gain in infancy was defined as a 0∙67 increase in WAZ score during the first year of life.22 We tested growth trajectories using mixed effects linear regression models including an interaction term for time (ie, study visit), with absolute differences in Z scores expressed as β coefficients with 95% CIs. At 12 months, we cross-sectionally compared the proportion of HEU and HU children who had moderate or severe malnutrition (underweight, defined as WAZ score <–2; stunting, defined as LAZ score <–2; microcephaly, defined as HCAZ score <–2; or overweight, defined as WLZ score or BMIZ score >2), using logistic regression models. We developed a directed acyclic graph including factors known to be associated with maternal HIV acquisition and child growth, and chose covariates for regression models accordingly (appendix). If multiple measures were available for a single construct (eg, socioeconomic status), our choice of variable (eg, availability of household amenities) was guided by improvement in model fit (using Aikaike’s Information Criterion). When indicated, we categorised continuous and discrete variables accord­ ing to published boundaries. We did several sensitivity analyses. Firstly, as birth outcomes (preterm birth [gestational age <37 weeks] and small-for-gestational age [SGA; birthweight <10th percentile]) might be on the causal pathway between maternal HIV status and child growth (as illustrated on the directed acyclic graph), we repeated the final linear regression models (which adjusted for socioeconomic factors, risky drinking, intimate partner violence, recent childhood illness, and infant feeding) with and without adjustment for birth outcomes. We also did four restricted analyses to assess the robustness of our findings, in the following subgroups: limited to children who attended at least five postnatal visits (ie, complete-case analysis); limited to children who were born appropriate-forgestational age (AGA), at full term; limited to children whose mothers did not report risky drinking at first antenatal visit; and limited to children with similar feeding patterns (defined for this analysis as having had at least 6 months of breastfeeding, of which the first 3 months were predominant or exclusive). Finally, we evaluated potential effect measure modification by infant sex (male vs female), breastfeeding duration (for ≥6 months compared www.thelancet.com/child-adolescent Vol 3 April 2019

1087 mother-child pairs screened at first postnatal visit 587 HIV-positive mothers and their HIV-exposed children (MCH-ART study) 500 HIV-negative mothers and their HIV-unexposed children (HU2 study)

203 excluded 116 HIV-positive mothers and their HIV-exposed children 78 not breastfeeding 23 missed enrolment window 12 unable to continue follow-up 3 other reasons 87 HIV-negative mothers and their HIV-unexposed children 14 not breastfeeding 63 missed enrolment window 9 unable to continue follow up 1 other reasons

884 mother–child pairs enrolled for postnatal follow-up 471 HIV-positive mothers and their HIV-exposed children 413 HIV-negative mothers and their HIV-unexposed children

12 children acquired HIV infection during postnatal follow-up 10 children exposed to HIV in utero (MCH-ART study) 2 children not exposed to HIV in utero (HU2 study)

872 mothers and HIV-uninfected children included in analysis 461 HIV-exposed uninfected children 411 HIV-unexposed children

155 mothers and HIV-uninfected children did not complete 12 months’ follow-up 97 HIV-exposed uninfected children 10 maternal or child death within 12 months postpartum 13 lost to follow-up 1 withdrew 57 relocations 16 other reasons 58 HIV-unexposed children 7 maternal or child death within 12 months postpartum 5 lost to follow-up 1 withdrew 33 relocations 12 other reasons

717 mother-child pairs completed 12 months follow-up 364 HIV-exposed uninfected children 343 with anthropometry at 12 months’ visit 353 HIV-unexposed children 343 with anthropometry at 12 months’ visit

Figure 1: Study profile

237

Articles

Total (n=872)

HIV-infected women and HEU children (n=461)

HIV-uninfected women p value and HU children (n=411)

Maternal and household characteristics at first antenatal visit Age, years

28 (6)

28 (5)

27 (6)

0·22

Married or cohabiting

373 (43%)

189 (41%)

184 (45%)

0·26

Completed secondary education

298 (34%)

114 (25%)

184 (45%)

<0·0001

Employed

376 (43%)

182 (40%)

194 (47%)

0·022

Formal housing

434 (50%)

219 (48%)

215 (52%)

0·16

Toilet inside home

292 (33%)

126 (27%)

166 (40%)

<0·0001

Running water inside home

404 (46%)

189 (41%)

215 (52%)

0·001

Lives in formal housing with inside toilet and running water 280 (32%)

125 (27%)

155 (38%)

0·001

40 (5%)

28 (6%)

12 (3%)

0·026

Risky drinking*

Household crowding (ie, ≥10 people)

147 (17%)

117 (26%)

30 (7%)

<0·0001

Intimate partner violence†

133 (15%)

101 (22%)

32 (8%)

<0·0001

75 (9%)

46 (10%)

29 (7%)

0·12

Depression‡ Maternal body-mass index ≥30 kg/m²§ Log10 HIV viral load at ART initiation, copies per mL CD4 count at ART initiation, cells per μL Gestational age at ART initiation, weeks

381/781 (49%)

171/386 (44%)

4·0 (3·3–4·5)

4·0 (3·3–4·5)

NA

··

354 (249–527)

354 (249–527)

NA

··

NA

··

22 (17–27)

22 (17–27)

210/395 (53%)

0·013

Birth and infant characteristics Gestational age at delivery, weeks

39 (38–40)

39 (38–40)

39 (38–40)

0·42

Preterm (<37 weeks)

94 (11%)

56 (12%)

38 (9%)

0·17

Small for gestational age¶

90 (10%)

51 (11%)

39 (10%)

0·45

81 (90%)

46 (90%)

35 (90%)

0·94

9 (10%)

5 (10%)

4 (10%)

Born at full term Born preterm Sex

··

··

··

·· 0·44

Female

444 (51%)

229 (50%)

215 (52%)

··

Male

428 (49%)

232 (50%)

196 (48%)

··

754 (86%)

421 (91%)

333 (81%)

<0·0001

Breastfeeding Ever exclusively breastfed|| Duration, months Duration of any breastfeeding, months Cumulative infant feeding from birth to 6 months||

1·4 (0·1–3·0)

1·4 (0∙2–3∙5)

1·3 (0∙1–3∙0)

6·0 (1∙5–12∙0)

3·9 (1∙4–12∙0)

9·0 (3∙0–12∙0)

··

··

··

<0·0001 0·0001 <0·0001

Exclusive breastfeeding

343 (39%)

206 (45%)

137 (33%)

··

Predominant breastfeeding

137 (16%)

42 (9%)

95 (23%)

··

Partial breastfeeding

392 (45%)

213 (46%)

179 (44%)

··

Post-partum household and environmental factors Maternal smoking**

··

··

··

At 6 months’ visit

21 (3%)

19 (5%)

2 (1%)

At 12 months’ visit

20 (3%)

20 (6%)

0

Household food security at 12 months’ visit††

··

··

··

·· <0·0001 <0·0001 <0·0001

No food insecurity

433 (63%)

190 (56%)

243 (71%)

··

At risk for food insecurity

177 (26%)

88 (26%)

89 (26%)

··

75 (11%)

64 (19%)

11 (3%)

··

Has food insecurity

Data are mean (SD), n (%), or median (IQR). p values for nominal data were calculated using χ² test. p values for means were calculated using Student’s t test and p values for medians using the Kruskal-Wallis test. ART=antiretroviral therapy. HEU=HIV-exposed uninfected. HU=HIV-unexposed. NA=not applicable. *Reported at first antenatal visit (missing data n=2). †Any physical, sexual, or psychological violence as measured with WHO violence against women questionnaire at first antenatal visit (missing data n=4). ‡Recorded at first antenatal visit (missing data n=2). §Missing data n=91. ¶Birthweight less than the 10th percentile, with percentile based on Intergrowth-21st reference standards. ||Maternal report (24 h recall), exclusive breastfeeding was defined as only breastmilk and prescribed medicine. **Based on maternal recall: history of any cigarette smoking in preceding month (missing data at 6 months n=2; missing data at 12 months n=5). ††Restricted to mother-infant pairs attending 12 month visit (n=686; missing data n=1).

Table 1: Maternal and infant characteristics by HIV infection and exposure status

with <6 months), and birth outcomes (preterm vs term delivery; and separately, SGA vs AGA) on the putative association between maternal HIV status and child 238

growth, using interaction terms and stratified analyses. During analysis, our data revealed a substantially higher prevalence of antenatal maternal risky drinking among www.thelancet.com/child-adolescent Vol 3 April 2019

Articles

WAZ score

LAZ score

WLZ score

HCAZ score

BMIZ score

–0·34 (–0·47 to –0·21)

–0·20 (–0·34 to –0·06)

–0·35 (–0·49 to –0·21)

–0·30 (–0·43 to –0·16)

–0·35 (–0·50 to –0·21)

Crude regression model Excluding HIV-exposure time interaction Including HIV-exposure time interaction

··

Birth

–0∙24 (–0∙40 to –0∙09)

6 week visit

–0∙30 (–0∙45 to –0∙14)

3 month visit

–0∙38 (–0∙54 to –0∙22)

6 month visit

–0∙35 (–0∙51 to –0∙19)

9 month visit

–0∙39 (–0∙55 to –0∙22)

12 month visit

–0∙39 (–0∙56 to –0∙23)

·· ·· 0∙01 (–0∙15 to 0∙18)

·· ··

·· ··

·· ··

–0∙54 (–0∙72 to –0∙35)

–0∙40 (–0∙57 to –0∙23)

–0∙53 (–0∙71 to –0∙35)

–0∙02 (–0∙20 to 0∙15)

–0∙44 (–0∙62 to –0∙25)

–0∙30 (–0∙47 to –0∙13)

–0∙50 (–0∙68 to –0∙32)

–0∙23 (–0∙40 to –0∙06)

–0∙30 (–0∙48 to –0∙11)

–0∙15 (–0∙32 to 0∙02)

–0∙30 (–0∙49 to –0∙12)

–0∙43 (–0∙60 to –0∙25)

–0∙22 (–0∙38 to –0∙03)

–0∙28 (–0∙45 to –0∙11)

–0∙20 (–0∙38 to –0∙01)

–0∙43 (–0∙61 to –0∙25)

–0∙22 (–0∙39 to –0∙09)

–0∙30 (–0∙47 to –0·13)

–0∙15 (–0∙34 to 0∙03)

Confounder-adjusted (multivariable) model Birth outcomes not included* –0·27 (–0·41 to –0·13)

–0·13 (–0·27 to 0·02)

–0·32 (–0·47 to –0·17)

–0·24 (–0·39 to –0·10)

–0·31 (–0·46 to –0·16)

Birth outcomes included†

–0·12 (–0·26 to 0·02)

–0·32 (–0·46 to –0·17)

–0·24 (–0·38 to –0·10)

–0·31 (–0·46 to –0·16)

–0·27 (–0·39 to –0·14)

Birth outcomes and HIV exposure time interaction included‡ Birth

–0·14 (–0·30 to 0·02)

6 week visit

–0·23 (–0·30 to –0·07)

3 month visit 6 month visit

··

··

··

··

0·11 (–0·06 to 0·29)

–0·51 (–0·70 to –0·32)

–0·33 (–0·50 to –0·16)

–0·53 (–0·72 to –0·35)

–0·30 (–0·46 to –0·14)

0·07 (–0·10 to 0·25)

–0·40 (–0·59 to –0·21)

–0·27 (–0·44 to –0·10)

–0·46 (–0·64 to –0·27)

–0·29 (–0·45 to –0·13)

–0·14 (–0·32 to 0·03)

–0·26 (–0·45 to –0·07)

–0·11 (–0·28 to 0·06)

–0·26 (–0·45 to –0·08)

9 month visit

–0·33 (–0·49 to –0·16)

–0·34 (–0·52 to –0·17)

–0·19 (–0·38 to 0·01)

–0·24 (–0·42 to –0·07)

–0·16 (–0·35 to 0·03)

12 month visit

–0·34 (–0·50 to –0·18)

–0·34 (–0·52 to –0·16)

–0·20 (–0·39 to 0)

–0·26 (–0·43 to 0)

–0·13 (–0·31 to 0·06)

Data are β coefficient for the crude regression model or adjusted β coefficient for the confounder-adjusted model, with 95% CIs in parentheses. β coefficients are the crude absolute difference in Z score comparing HIV-exposed uninfected infants with HIV-unexposed infants. Adjusted β coefficients are the adjusted absolute difference in Z score comparing HIV-exposed uninfected infants with HIV-unexposed infants. Mixed-effect linear regression with random intercept for child. BMIZ score=body-mass index-for-age Z score. HCAZ score=head circumference-for-age Z score. HEU=HIV-exposed uninefected. HU=HIV-unexposed. LAZ score=length-for-age Z score. SGA=small-for-gestational age. WLZ score=weight-for-length Z score. WAZ score=weight-for-age Z score. *Adjusted for time (study visit), socioeconomic factors (household crowding, marital status, and lack of amenities [<3 of running water, flush toilet, and electricity in home], maternal alcohol use, intimate partner violence, recent history of any childhood illness, and infant feeding [current milk feeding and weaning practices]). †Further adjusted for SGA (birthweight <10th percentile) and preterm birth (delivery <37 weeks’ gestation). ‡Contrasted predictive margins based on multivariable model as above (including SGA, preterm birth), with interaction term for HIV-exposure status and time (WAZ score pinternation=0·11; LAZ score pinternation<0·0001; HCAZ score pinternation=0·070; WLZ score pinternation =0·0041; BMIZ score pinternation<0·0001).

Table 2: Differences in growth between HEU and HU children over time by mixed-effects linear regression analysis

HIV-infected than HIV-uninfected women. Simul­ taneously, we noted substantial interaction effects between birth size and HIV exposure across growth outcomes. Given the known associations between antenatal alcohol use and birth outcomes, we repeated the HIV exposure and birth size interaction analysis after excluding women who reported antenatal risky drinking. We analysed all data using Stata 14.0 (Statacorp, College Station, TX, USA). MCH-ART is registered on ClinicalTrials.gov, number NCT01933477.

Role of the funding source The funders of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all data in the study and had final responsibility for the decision to submit for publication.

Results 1087 mothers and newborn babies (median age at first postnatal visit, 5 days IQR 4–8) were screened for enrolment into the postnatal phases (ie, for continuation of postnatal follow-up through to 12 months) of MCH-ART www.thelancet.com/child-adolescent Vol 3 April 2019

(HIV-positive women and their babies, n=587; recruitment from June 5, 2013, to Dec 10, 2014) and HU2 (HIV-negative women and their babies, n=500; recruitment from Oct 1, 2014, to April 5, 2016). After excluding 203 mother-child pairs (reasons for exclusion shown in figure 1), 884 were enrolled for further postnatal follow-up, which was concluded for all participants by March 28, 2017. After excluding the data from 12 children who tested HIV-positive during follow-up, 872 mother-child pairs contributed to this analysis (figure 1), with 4511 study visits (median of 6 visits per child; IQR 5–6). Details of mother-to-child HIV transmission during the study period will be published elsewhere. Maternal, child, and household characteristics are shown in table 1. Of the 872 mother-child pairs contributing to this analysis, 717 (82%) completed follow-up through to 12 months, 686 (96%) of whom had anthropometric data available at this study visit (figure 1; appendix). Motherchild pairs who did complete 12 months of follow-up and and those who did not had minimal differences in baseline characteristics (appendix). Most notably, those who were censored before 12 months had shorter durations of breastfeeding, partly a reflection of our definition of 239

Articles

A

B

2·0

HEU children (n=461) HU children (n=411)

Predicted Z scores

1·5

HEU children (n=409) HU children (n=385)

1·0 0·5 0 –0·5 –1·0 –1·5 –2·0

Birth

6 weeks

3 months

6 months

9 months

12 months

C

6 weeks

3 months

6 months

9 months

12 months

D

2·0

Predicted Z scores

1·5 1·0 0·5 0 –0·5 –1·0 HEU children (n=397) HU children (n=382)

–1·5 –2·0 6 weeks

3 months

6 months

9 months

HEU children (n=416) HU children (n=383)

12 months

Approximate age at study visit

6 weeks

3 months

6 months

9 months

12 months

Approximate age at study visit

Figure 2: Adjusted weight-for-age (A), length-for-age (B), weight-for-length (C), and head circumference-for-age (D) Z score growth trajectories of HEU and HU children Data points are Z scores with 95% CIs, calculated by use of adjusted mixed-effects linear regression models. HEU=HIV-exposed uninfected. HU=HIV-unexposed.

breastfeeding cessation date among those censored before 12 months. At first antenatal visit, HIV-infected women reported greater socioeconomic, psychosocial, and behavioural adversity than HIV-uninfected women, includ­ ing risky drinking and intimate partner violence. Maternal obesity was common in both groups (table 1). Overall, median birthweight was 3180 g (IQR 2820–3460), with 98 (11%) of 872 babies born with low birthweight (<2500 g) and ten (1%) born with very low birthweight (<1500 g). Prevalence of preterm delivery (<37 weeks) and SGA birthweight did not differ by maternal HIV status: 56 (12%) of 461 HEU children were preterm, compared with 38 (9%) of 411 HU children (p=0∙17); 51 (11%) of 461 HEU children were SGA, compared with 39 (10%) of 411 HU children (p=0∙45). Overall, 81 (90%) of 90 SGA children were born at full term. Although HEU children were more likely to ever be exclusively breastfed than HU children were, they had a shorter overall duration of any breast­ feeding (table 1). Complementary feeding (solid food or nutritive liquids other than breastmilk or formula milk) had been introduced to 38 (11%) of 363 HEU children versus 64 (17%) of 368 HU children by 3 months; 191 (50%) of 382 versus 244 (70%) of 345 by 6 months; and 351 (98%) of 360 versus 314 (95%) of 330 by 9 months (appendix). A greater proportion of HIVaffected house­holds than HIV-unaffected households had food insecurity at 12 months (table 1). At birth, the mean WAZ score was close to zero in both groups (HEU children, –0∙13 [SD 0∙92]; and HU children, 240

0∙12 [SD 1∙00]; crude β –0·24, 95% CI –0·40 to –0·09; table 2). After 6 weeks, WAZ scores gradually increased in both groups, with consistently lower scores among HEU than HU children throughout follow-up (crude overall β –0∙34 (95% CI –0∙47 to –0∙21; table 2). Results were similar in adjusted analyses, both when including or excluding birth outcomes in the final model (table 2; figure 2A). By 12 months, the mean WAZ score was above zero in both groups (HEU children 0∙24 [SD 1∙32], and HU children 0∙66 [SD 1∙24]; appendix). At the same study visit, only 16 (2%) of 684 children with WAZ scores available were underweight (HEU, n=12 vs HU children, n=4; odds ratio [OR] 3·09, 95% CI 0·99–9·68; appendix). By contrast, 78 (11%) of all 684 children had a WAZ score of more than 2 (data not shown). Overall, 299 (44%) of 684 children had rapid weight gain between birth and age 12 months. In the first 6 months of life, mean LAZ scores were similar, and below zero, in both groups (eg, at 6 weeks, mean LAZ score among HEU children was –0·34 [SD 1·36] compared with –0·37 [SD 1·12] among HU children; figure 2B; appendix). Between the 6 month and 9 month study visits, mean LAZ scores for HEU children remained below zero, but steadily increased to above zero for HU children (figure 2B; appendix). From 9 months onwards, LAZ scores decreased in both groups, signalling cohort-wide linear growth faltering. By 12 months, HEU children had significantly lower mean LAZ scores than HU children did (–0·51 [SD 1·27] vs –0·02 [SD 1·13]; β –0·43, 95% CI –0·61 www.thelancet.com/child-adolescent Vol 3 April 2019

Articles

www.thelancet.com/child-adolescent Vol 3 April 2019

Wasted

Expected

Overweight

100

80 Proportion of children per visit (%)

60

40

20

th s on

s th on

m 12

s m

th on

6

m

on

HEU children

9

s th

s ek 6

we

th on m

12

3m

s

s th

s th

on m 9

s th

on m 6

on 3m

we

ek

s

0

6

to –0·25; table 2, appendix). Inferences were unchanged in adjusted analysis, both when including and excluding birth outcomes in the final model (table 2). At 12 months, HEU children were also twice as likely to be stunted (LAZ score <–2) compared with HU children (35 [10%] of 342 HEU vs 14 [4%] of 342 HU children; OR 2·67, 95% CI 1·41–5·06). Results were similar after adjusting for confounders including SGA and preterm birth (appendix). At 6 weeks, mean WLZ scores were lower among HEU children than HU children (0·03 [SD 1·35] vs 0·55 [SD 1·37]; at 6 weeks, adjusted β coefficient –0∙51, 95% CI –0∙70 to –0∙32; figure 2C, table 2; appendix). Over time, WLZ scores steadily increased in both groups, with a more rapid increase among HEU children, until differ­ ences became negligible (figure 2C). By 12 months, a large proportion of both HEU and HU children were overweight (WLZ score >2: 54 [16%] of 342 HEU children and 60 [18%] of 340 HU children; OR 0·87, 95% CI 0·58–1·31; figure 3; appendix), or obese (WLZ score >3: 16 [5%] of 342 HEU children vs 20 [6%] of 342 HU children; appendix). By 12 months, only 11 (2%) of 682 children with WLZ scores available were wasted (WLZ score <–2); figure 3; appendix). The proportion of children who were both overweight and stunted by 12 months was very small (four [1%] of 682), with no difference between HEU and HU children (n=3 vs n=1; appendix). Findings were similar in analyses using body-mass index (BMI) instead of weightfor-length to compare proportional size between HEU and HU children, in both crude and adjusted linear and logistic regression models (table 2; appendix). Between 6 weeks and 12 months, mean HCAZ scores increased slightly in both groups (overall HCAZ score 0∙74 [SD 1∙29] at 6 weeks vs 1∙06 [SD 1∙12] at 12 months; appendix), with slightly lower HCAZ scores among HEU children than among HU children throughout follow-up in both crude and adjusted analysis (overall crude β –0∙30, 95% CI –0∙43 to –0∙16; figure 2D, table 2). Adjusted results were similar for final models both including and excluding birth outcomes (table 2). By 12 months, only five (1%) of 680 children with HCAZ measures had microcephaly (HCAZ score <–2: three HEU children vs two HU children), whereas 132 (19%) had macrocephaly (HCAZ score >2: 56 (17%) of 340 HEU children vs 76 (22%) of 340 HU children. Overall, over­weight children (ie, WLZ score >2) were substantially more likely to be macro­cephalic than children who were not overweight (at 12 months, OR 3·84, 95% CI 2·48–5·98; data not shown). We found no evidence of interaction effects between preterm birth and maternal HIV exposure or child sex and maternal HIV exposure on any child growth outcomes (data not shown). However, across all measures, inter­action effects were seen between size at birth (SGA compared with AGA) and maternal HIV exposure (HEU compared with HU) on child growth outcomes (figure 4; appendix). SGA was associated with lower mean WAZ, LAZ, HCAZ, and WLZ scores in both crude and adjusted analyses, with stronger associations noted among HEU children than HU

HU children

Figure 3: Distribution of overweight and wasting by study visit, comparing HEU and HU children Overweight is defined as a weight-for-length Z score greater than 2, and wasted is defined as a weight-for-length Z score of less than –2. HEU=HIV-exposed uninfected. HU=HIV-unexposed.

children in stratified analysis (appendix). For example, the crude overall mean difference in LAZ score comparing SGA HEU children with AGA HEU children was larger than when com­paring SGA HU children with AGA HU children (SGA vs AGA: β –1·24, 95% CI –1·52 to –0·95 among HEU children, and β –0·77, –1·10 to –0·44, among HU children). In turn, associations between HIV-exposure status and child growth outcomes differed within strata of birth size (appendix). For example, the crude overall mean difference in LAZ score comparing SGA HEU children with SGA HU children was greater than when comparing AGA HEU children with AGA HU children (HEU vs HU children: β –0·59, 95% CI –1·01 to –0·18 among children born SGwA, and β –0·13, –0·27 to 0·01 among children born AGA; appendix). Although these interaction effects were noticeable across growth measures (figure 4), the most striking effects graphically were for trajectories of linear growth (LAZ score over time; figure 4B). Although the mean linear growth trajectory of AGA HEU children was similar to that of AGA HU children, the trajectories of SGA HEU and SGA HU children diverged substantially (figure 4B). Between 6 weeks and 9 months, LAZ scores steadily increased among SGA HU children, but not among SGA HEU children. Thereafter, all four groups had group-level linear growth faltering (ie, decreasing mean LAZ score), but HEU SGA children had earlier onset and sharper decline than the other groups (figure 4B). At age 12 months, differences between AGA HEU and AGA HU children were relatively small (β –0·34, 95% CI –0·52 to –0·16), whereas differences between SGA HEU and SGA HU children were substantially larger (β –1·03, 95% CI –1·56 to –0·50), in both crude and adjusted analyses (appendix). To explore the contribution of antenatal alcohol 241

Articles

Predicted Z scores

A

B

AGA HU children (n=372) AGA HEU children (n=410) SGA HU children (n=39) SGA HEU children (n=51) Time interaction p<0·0001

2·0 1·5 1·0 0·5 0 –0·5 –1·0 –1·5 –2·0 Birth

6 weeks

3 months

6 months

9 months

12 months

Predicted Z scores

C

2·0 1·5 1·0 0·5 0 –0·5 –1·0 –1·5 –2·0

6 weeks

AGA HU children (n=347) AGA HEU children (n=361) SGA HU children (n=38) SGA HEU children (n=48) Time interaction p=0·081

6 weeks

3 months

6 months

9 months

12 months

D

AGA HU children (n=345) AGA HEU children (n=351) SGA HU children (n=37) SGA HEU children (n=46) Time interaction p=0·022 3 months

6 months

AGA HU children (n=345) AGA HEU children (n=368) SGA HU children (n=38) SGA HEU children (n=48) Time interaction p=0·032 9 months

12 months

6 weeks

Approximate age at study visit

3 months

6 months

9 months

12 months

Approximate age at study visit

Figure 4: Interaction plots comparing growth calculated using weight-for-age (A), length-for-age (B), weight-for-length (C), and head circumference-for-age (D) Z scores over time, by categories of HIV exposure and birth size Mean Z score (95% CI) from adjusted mixed-effects linear regression models. AGA=appropriate-for-gestational age. HEU=HIV-exposed uninfected. HU=HIV-unexposed. SGA=small-for-gestational age.

exposure to these interaction effects, we repeated the analyses among the subgroup of women who did not report risky drinking at first antenatal visit (n=723). Although the differences between SGA HEU and SGA HU children were somewhat attenuated compared with in the main analysis (data not shown), results were otherwise similar. For example, in this subgroup of children without maternal risky drinking, when comparing AGA HEU children with AGA HU children the crude mean difference in LAZ score at 12 months was β –0·30 (95% CI –0·50 to –0·10); and for the comparison of SGA HEU children with SGA HU children, the difference was β –0·72 (95% CI –1·36 to –0·09). We had similar findings for WAZ, HCAZ, and BMIZ scores in crude and adjusted analysis (data not shown). We also tested for an interaction between maternal HIV exposure and duration of breastfeeding (comparing those who breastfed for 6 months or longer with those who breastfed for less than 6 months). Although no significant interaction effects were noted for LAZ or HCAZ scores, longer (compared with shorter) breast­ feeding was associated with slightly lower WAZ, WLZ, and BMIZ scores among HEU children but not among HU children (data not shown). For example, mean difference in WLZ score comparing longer breastfeeding with shorter breast­ feeding among HEU children at 12 months was β –0·31 (95% –0·60 to –0·04). By comparison, at 12 months, no observable difference was seen in WLZ scores between HU children with longer breatfeeding compared with HU 242

children with shorter breastfeeding (β 0·11, 95% CI –0·17 to 0·40). However, we saw no effects of breastfeeding duration on prevalence of overweight or underweight (data not shown). Results of the four restricted analyses (ie, analyses limited to children with complete follow-up data; children without maternal risky drinking; children born AGA at term; and children with similar feeding practices, respectively) were similar to those of the main analysis for all growth measures (appendix).

Discussion In the context of universal ART and breastfeeding, we found small but consistent deficits in the early growth trajectories of young, breastfed HEU children compared with HU children drawn from the same peri-urban community in South Africa. Prevalence of undernutrition at age 12 months was low, with linear growth faltering being more common than underweight, wasting, or microcephaly. Consequently, although the absolute difference in LAZ score between HEU and HU children was small, it was associated with a doubled risk of stunting by 12 months. Simultaneously, a high proportion of children in both groups were overweight by age 12 months. In this cohort of young, HIV-uninfected, breastfed South African children, we describe moderately different growth patterns by maternal HIV status and an early onset of the double burden of malnutrition that is increasingly the norm in sub-Saharan Africa.2 www.thelancet.com/child-adolescent Vol 3 April 2019

Articles

Our findings of similar weight and proportional size among young, breastfed HEU and HU children are reassuring given the uniform exposure to maternal ART in utero and throughout breastfeeding. These results approximate those of other prospective, peri-urban South African studies predating universal maternal ART.3,4,9 In comparison, research from other countries in sub-Saharan Africa predating universal maternal ART have reported increased risks of under­weight and wasting among HEU children compared with HU children.3,4,7,12 However, the studies reporting increased risks of underweight and wasting were largely from rural settings with high back­ ground prevalence of undernutrition, advanced maternal disease, or were based on cross-sectional studies that were at risk of selection bias. Nonetheless, the increased risk of stunting after age 9 months among HEU children in our cohort aligns with most other studies of growth among HEU children in resource-limited settings, although the overall prevalence and severity of stunting was substantially lower in our study than in previous reports.3,4,7,11,12 Although the prevalence of stunting among HEU children at 12 months (10%) was much higher than among HU children in our cohort (4%), it is still lower than the general prevalence elsewhere in South Africa, where the most recent national estimates are 31% at this age.23 Stunting is typically a process indicative of chronic malnutrition and closely associated with poverty.24 In our study, most motherchild pairs from both cohorts lived in adverse socio­ economic conditions with evidence of food insecurity, and both groups of children had linear growth faltering after 9 months, around the time of introduction to comp­ lementary foods. Although adjustments for infant feeding did not substantially change our findings, the more pronounced linear growth faltering among HEU children might have been related to earlier cessation of breast­ feeding, as has been reported in other studies.11 Indeed, breastfeeding durations were concerningly short, particularly among HEU children. Our data therefore highlight an ongoing need for improved breastfeeding promotion and support in this context. Although we report largely reassuring data for com­ parative growth of HEU and HU children, our early childhood obesity findings suggest a looming public health crisis. Rapid weight gain in infancy, seen in almost half of our cohort, has been linked with an increased risk for overweight and obesity in childhood and adolescence.22,25 In South Africa, the prevalence of overweight among children younger than 5 years is around 13%.2,23 Our finding of almost 20% of the cohort being overweight by age 1 year in this peri-urban setting provides evidence of a worsening childhood overweight and obesity epidemic, as has been documented in older children and adolescents across sub-Saharan Africa.2,26 Although South Africa has a particularly high burden of obesity across the lifespan,23 other countries in the region have shown similar obesity trends, with an estimated 11% of children younger than 5 years with overweight in Botswana, 9% in Sierra Leone, www.thelancet.com/child-adolescent Vol 3 April 2019

and 8% in Mozambique according to the most recent WHO-UNICEF estimates.2 Childhood obesity predicts adult obesity, with multiple adverse metabolic health effects including dyslipidaemia and insulin resistance.26 However, HEU children might be at increased risk of dyslipidaemia and altered energy metabolism, regardless of obesity.27,28 Whether obesity-related meta­ bolic effects might be exacerbated by concurrent perinatal ART exposure warrants investigation as the obesity epidemic grows across settings with a high prevalence of HIV.2 Our data also highlight the crucial influence of antenatal factors on child health. Specifically, we found a potential interaction effect between in utero HIV exposure and intrauterine growth restriction on linear growth. Although linear growth faltering occurred across the cohort, HEU children who were SGA were at highest risk. We previously reported a similar inter­action between HIV exposure and preterm delivery on neuro­development,16 and hypothesise a potential role for cytomegalovirus in this context.29 Indeed, understanding underlying causal mechanisms will be crucial for the design of HIV-specific interventions for this patient population in the future. Additionally, substantive evidence already exists for interventions to improve perinatal maternal and child nutrition across settings;30 urgent implementation of these strategies is likely to benefit both HEU and HU children. Our study has several limitations and strengths. Singlesite data from peri-urban South Africa might not be generalisable to rural areas or settings with different background prevalences of malnutrition. Our findings might also not extend to children conceived on ART. Although loss to follow-up could predispose our study to selection bias, we saw no substantial differences between children with complete and children with incomplete follow-up, and results were unchanged in complete-case analysis. We could not assess the symmetry of intrauterine growth restriction because of a lack of robust measure­ ments for head circumference and length at birth. Un­ measured confounding is always a concern with observ­ ational research; however, efforts were made to minimise confounding in our study design and analysis. behavioural Notably, we could account for many socio­ factors, which to date have often been neglected in HEU child health research. Our cohort of breastfed HEU children who were universally exposed to maternal ART, with homogeneous ART exposure, is one of the first to be reported since the widespread implementation of new WHO guidelines on universal maternal ART. Other study strengths include robust gestational age estimates, and repeated anthropometric measures over time, which are crucial when assessing a dynamic process such as childhood growth. In the context of universal maternal ART and breast­ feeding, growth trajectories of HEU children were similar to those of analogous HU children. Linear growth faltering was common in both groups, particularly among HEU children with evidence of intrauterine growth 243

Articles

restriction. Concerningly, almost 20% of all chil­dren were overweight by age 1 year. The long-term effects of early onset overweight and stunting in chil­dren exposed to HIV and ART require further in­vestigation. Contributors SMlR was responsible for conceptualisation and implementation of the HIV-unexposed uninfected (HU2) study, assisted with collection of data, did the analyses, and wrote the first draft of the manuscript. LM and EJA conceived the Maternal Child Health Antiretroviral Therapy (MCH-ART) and HU2 studies, and were responsible for study design, acquiring funding, protocol implementation, and overall leadership. KAD and MK provided supervision for child health aspects of the study. KB and TKP were responsible for data management and oversight of data collection. TKP and KKN were the study coordinators. AZ was the senior study manager and provided oversight of all study administration processes. All authors contributed to and approved the final manuscript. Declaration of interests SMlR reports grants from the US President’s Emergency Plan for AIDS Relief (PEPFAR) through the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD), Elizabeth Glaser Pediatric AIDS Foundation, South African Medical Research Council, Fogarty Foundation (US National Insitutes of Health Fogarty International Center Grant), and the Office of AIDS Research during the conduct of the study. EJA reports grants from NICHD during the conduct of the study. All other authors declare no competing interests. Acknowledgments We thank the research and clinical staff who supported this work, and the families who participated in the study. SMlR is also grateful to Tanya Doherty and Vundli Ramokolo for helpful insights, comments, and support of this work. This study was supported by the US President’s Emergency Plan for AIDS Relief through the Eunice Kennedy Shriver National Institute of Child and Human Development under Cooperative Agreement 1R01HD074558. Additional funding came from the Elizabeth Glaser Pediatric AIDS Foundation, South African Medical Research Council (Clinician-Researcher PhD Scholarship), the Fogarty Foundation (US National Institutes of Health Fogarty International Center Grant #5R25TW009340), and the Office of AIDS Research. References 1 UNAIDS. AIDSinfo 2018. http://aidsinfo.unaids.org/ (accessed Sept 9, 2018). 2 WHO. Global Health Observatory data repository. Joint child malnutrition estimates (UNICEF-WHO-The World Bank Group): Levels and trends (2018 edition). Geneva: World Health Organization, April 24, 2018. http://apps.who.int/gho/data/node. main.CHILDOVERWEIGHT?lang=en (accessed Sept 11, 2018). 3 le Roux SM, Abrams EJ, Nguyen K, Myer L. Clinical outcomes of HIV-exposed, HIV-uninfected children in sub-Saharan Africa. Trop Med Int Health 2016; 21: 829–45. 4 Evans C, Jones CE, Prendergast AJ. HIV-exposed, uninfected infants: new global challenges in the era of paediatric HIV elimination. Lancet Infect Dis 2016; 16: e92–107. 5 WHO, United Nations Children’s Fund. Guideline: updates on HIV and infant feeding: the duration of breastfeeding, and support from health services to improve feeding practices among mothers living with HIV. Geneva: World Health Organization, 2016. http://apps. who.int/iris/bitstream/10665/246260/1/9789241549707-eng. pdf?ua=1 (accessed Sept 5, 2016). 6 WHO. Consolidated guidelines on the use of antiretroviral drugs for treating and preventing HIV infection. Geneva: World Health Organization, 2013. 7 Omoni AO, Ntozini R, Evans C, et al. Child growth according to maternal and child HIV status in Zimbabwe. Pediatr Infect Dis J 2017; 36: 869–76. 8 Patel D, Bland R, Coovadia H, Rollins N, Coutsoudis A, Newell ML. Breastfeeding, HIV status and weights in South African children: a comparison of HIV-exposed and unexposed children. AIDS 2010; 24: 437–45.

244

9

10 11 12

13 14

15

16 17 18

19 20

21 22 23

24 25 26 27 28 29

30

Ramokolo V, Lombard C, Fadnes LT, et al. HIV infection, viral load, low birth weight, and nevirapine are independent influences on growth velocity in HIV-exposed South African infants. J Nutr 2014; 144: 42–48. Kuhn L, Kasonde P, Sinkala M, et al. Does severity of HIV disease in HIV-infected mothers affect mortality and morbidity among their uninfected infants? Clin Infect Dis 2005; 41: 1654–61. Sudfeld CR, Lei Q, Chinyanga Y, et al. Linear growth faltering among HIV-exposed uninfected children. J Acquir Immune Defic Syndr 2016; 73: 182–89. Rosala-Hallas A, Bartlett JW, Filteau S. Growth of HIV-exposed uninfected, compared with HIV-unexposed, Zambian children: a longitudinal analysis from infancy to school age. BMC Pediatr 2017; 17: 80. Powis KM, Smeaton L, Hughes MD, et al. In-utero triple antiretroviral exposure associated with decreased growth among HIV-exposed uninfected infants in Botswana. AIDS 2016; 30: 211–20. Bork KA, Cames C, Newell ML, et al. Formula-feeding of HIV-exposed uninfected African children is associated with faster growth in length during the first 6 months of life in the Kesho Bora study. J Nutr 2017; 147: 453–61. Myer L, Phillips TK, Zerbe A, et al. Optimizing antiretroviral therapy (ART) for Maternal and Child Health (MCH): rationale and design of the MCH-ART Study. J Acquir Immune Defic Syndr 2016; 72 (suppl 2): S189–96. le Roux SM, Donald KA, Brittain K, et al. Neurodevelopment of breastfed HIV-exposed uninfected and HIV-unexposed children in South Africa. AIDS 2018; 32: 1781–91. WHO. Training course on child growth assessment: WHO child growth standards. Geneva: World Health Organization; 2008. WHO, UNICEF, USAID, AED, UCDAVIS, IFPRI. Indicators for assessing infant and young child feeding practices. Part 1: definitions. Geneva: World Health Organization, 2008. http://www. who.int/nutrition/publications/infantfeeding/9789241596664/en/ (accessed Feb 11, 2019). Labadarios D, McHiza ZJ, Steyn NP, et al. Food security in South Africa: a review of national surveys. Bull World Health Organ 2011; 89: 891–99. Villar J, Cheikh Ismail L, Victora CG, et al. International standards for newborn weight, length, and head circumference by gestational age and sex: the Newborn Cross-Sectional Study of the INTERGROWTH-21st Project. Lancet 2014; 384: 857–68. WHO. WHO Anthro (version 3.2.2, January 2011) and macros. Geneva: World Health Organization, 2010. http://www.who.int/ childgrowth/software/en/ (accessed Sept 11, 2018). Salgin B, Norris SA, Prentice P, et al. Even transient rapid infancy weight gain is associated with higher BMI in young adults and earlier menarche. Int J Obes (Lond) 2015; 39: 939–44. Statistics South Africa. South Africa demographic and health survey 2016: key indicator report. Statistics South Africa, 2017. https://www.statssa.gov.za/publications/Report%2003-00-09/ Report%2003-00-092016.pdf (accessed Sept 11, 2018). Black RE, Victora CG, Walker SP, et al. Maternal and child undernutrition and overweight in low-income and middle-income countries. Lancet 2013; 382: 427–51. Weng SF, Redsell SA, Swift JA, Yang M, Glazebrook CP. Systematic review and meta-analyses of risk factors for childhood overweight identifiable during infancy. Arch Dis Child 2012; 97: 1019–26. Juonala M, Magnussen CG, Berenson GS, et al. Childhood adiposity, adult adiposity, and cardiovascular risk factors. N Engl J Med 2011; 365: 1876–85. Jao J, Kirmse B, Yu C, et al. Lower preprandial insulin and altered fuel use in HIV/antiretroviral-exposed infants in Cameroon. J Clin Endocrinol Metab 2015; 100: 3260–69. Claudio CC, Patin RV, Palchetti CZ, Machado DM, Succi RC, Oliveira FL. Nutritional status and metabolic disorders in HIV-exposed uninfected prepubertal children. Nutrition 2013; 29: 1020–23. Filteau S, Rowland-Jones S. Cytomegalovirus infection may contribute to the reduced immune function, growth, development, and health of HIV-exposed, uninfected African children. Front Immunol 2016; 7: 257. Bhutta ZA, Das JK, Rizvi A, et al. Evidence-based interventions for improvement of maternal and child nutrition: what can be done and at what cost? Lancet 2013; 382: 452–77.

www.thelancet.com/child-adolescent Vol 3 April 2019