Maternal HIV infection associated with small-for-gestational age infants but not preterm births: Evidence from rural South Africa

Background: Human immunodeficiency virus (HIV) is prevalent in many countries where small-for-gestational age (SGA) and premature delivery are also common. However, the associations between maternal HIV, preterm delivery and SGA infants remain unclear. We estimate the prevalence of SGA and preterm (<37 weeks) births, their associations with antenatal maternal HIV infection and their contribution to infant mortality, in a high HIV prevalent, rural area in South Africa. Methods Data were collected, in a non-randomized intervention cohort study, on all women attending antenatal clinics (20012004), before the availability of antiretroviral treatment. Newborns were weighed and gestational age was determined (based on last menstrual period plus midwife assessment antenatally). Poisson regression with robust variance assessed risk factors for preterm and SGA birth, while Cox regression assessed infant mortality and associated factors. Results Of 2368 live born singletons, 16.6 were SGA and 21.4 were preterm. HIV-infected women (n 1189) more commonly had SGA infants than uninfected women (18.1 versus 15.1; P 0.051), but percentages preterm were similar (21.8 versus 20.9; P 0.621). After adjustment for water source, delivery place, parity and maternal height, the SGA risk in HIV-infected women was higher [adjusted relative risk (aRR) 1.28, 95 confidence interval (CI): 1.061.53], but the association between maternal HIV infection and preterm delivery remained weak and not significant (aRR: 1.07, 95 CI: 0.911.26). In multivariable analyses, mortality under 1 year of age was significantly higher in SGA and severely SGA than in appropriate-for-gestational-age infants [adjusted hazard ratio (aHR): 2.12, 95 CI: 1.183.81 and 2.77, 95 CI: 1.564.91], but no difference in infant mortality was observed between the preterm and term infants (aHR: 1.18 95 CI: 0.791.79 for 3436 weeks and 1.31, 95 CI: 0.582.94 for <34 weeks). Conclusions Maternal HIV infection increases the risk of SGA, but not preterm births, in this cohort. © 2012 The Author. This vertical transmission study (VTS) enrolled HIV-infected and -uninfected pregnant women in nine antenatal clinics in two sites, the largest in rural Umkhanyakude district, northern KwaZulu-Natal and the other outside Durban, South Africa (Coovadia et al., 2007; Bland et al., 2010). Recruitment was from August 2001 to September 2004 with mothers and children followed for 2 years post-natally. Initially, all HIV-infected women and a systematic sample of uninfected women were enrolled in the study. However from July 2003, all women attending antenatal visits were offered enrolment before HIV testing. All pregnant women were tested for HIV and counselled at the antenatal clinic, by lay HIV counsellors, on infant feeding options according to the World Health Organization (WHO)/United Nations programme on AIDS recommendations current at that time (Bland et al., 2007). All HIV-infected women and their infants were offered single-dose nevirapine, but at the time of the study, antiretroviral treatment was not available through the public health services in South Africa. The study was approved by the Biomedical Research Ethics Committee of the University of KwaZulu-Natal. Gestational age at delivery was determined on the basis of last menstrual period (collected at antenatal visits by trained study staff), unless adjusted by midwife assessment on the basis of a clinical assessment. Birthweight was measured by study staff or taken from clinic registers within 72 h of birth. For infants born at home, mothers brought their infant to the clinic the following day (or within 72 h) for immunization (BCG and polio), weighing and registration to obtain a road-to-health card (RTHC); this is the general practice in South Africa since without the RTHC, children cannot access public services such as clinics or school. These infants were weighed by the study staff. However, if women delivered in facilities outside the study area or study staff were not available, we used the weight documented by the Department of Health staff on the RTHC (Rollins et al., 2007). Infants attended monthly clinic visits with a study nurse, scheduled from 6 weeks to 9 months of age, and thereafter quarterly to 24 months. All study clinic assistants who carried out anthropometric measurements at each visit received standardized anthropometry instructions with regular in-service training also given to ensure strict adherence to the techniques. The clinic manager conducted quality assurance checks of measurement techniques throughout the study. The same measuring equipment was used at all of the clinics with precision within 10 g for newborns while for older children, portable electronic weighing scales with taring capability and a precision to 100 g (UNICEF electronic scale 890) was used as recommended by WHO (de Onis et al., 2004). At approximately weekly intervals, the clinic manager checked calibration using standard metal weights. Further, each child had two weights recorded and if their difference was >100 g, a third measurement was taken and the two weights within 100 g were recorded; the mean value was calculated for this analysis (de Onis et al., 2004). Dried blood spot (DBS) samples were collected from all HIV-exposed newborn infants within 72 h of birth, and then at each scheduled clinic visit to determine their HIV status. HIV status was established by quantitative HIV RNA assay with a sensitivity of 80 copies/ml (equivalent to 1600 copies per 50 µl DBS) (Bland et al., 2010). Infants were considered to be perinatally infected if the DBS sample taken at 4–8 weeks was positive, post-natally infected if the 4–8-week sample was negative and any subsequent sample was positive or infected with ‘timing unknown’ if the first sample was taken after 8 weeks and was positive. In the antenatal period, socio-demographic and past and current pregnancy data were collected using structured questionnaires; post-natally, daily infant morbidity and feeding practices were documented at weekly intervals, and maternal weights, mid-upper arm circumference (MUAC) and height were measured (Bland et al., 2010). For these analyses, the weight and MUAC taken at 6 weeks post-delivery were used. Antenatal maternal HIV status was classified into HIV positive and negative; maternal CD4 count was measured in all HIV-positive women. LBW was defined as <2500 g (Kramer, 1987a,b). Infants were classified as preterm if the gestational age at birth was <37 completed weeks, and as SGA if below the 10th percentile of the birthweight-for-gestational age, as recommended by the WHO (WHO, 1995), with the US population-based reference used as the standard for comparability of the prevalence rates of SGA, and the potential effect of maternal HIV infection on SGA, with previous studies elsewhere in Africa (Alexander et al., 1996). To assess severely small-for-gestational-age infants, risk associations were further analyzed at a 3% cutoff (Oken et al., 2003). Size for gestational age was therefore categorized into three groups based on percentiles: AGA, ≥10%; SGA, 3 ≤ 10% and <3%, severely SGA. We only included singleton live births, given the increased risk of LBW and premature delivery for multiple birth; we also excluded stillbirths as four-fifths had missing birthweights. Place of delivery was classified into health facility and home. Parity was classified so as to compare primiparae with multiparae (Mansour et al., 2002) and was categorized as 0, 1–3 or ≥4. Based on the maternal enrollment clinic, residential area was classified as rural, peri-urban or urban, as most people in this setting are likely to attend their nearest clinic, with the median travel time of 81 min (Tanser, 2006). Neonatal mortality was defined as a death in the first 4 weeks of life (Lawn et al., 2005), while infant mortality was defined as the death by 12 months of age. Data were analysed using Stata Version 11.2 (STATA Corps, College Station, TX, USA). Differences between HIV-infected and -uninfected women were assessed using t-test for continuous variables with normal distribution, χ2 test for categorical variables and Fisher's exact test if the numbers were small. Poisson regression, with a value of one attributed to each participant's follow-up time to obtain prevalence ratio estimates (done on all subjects and then separately among the HIV-positive and HIV-negative mothers), was used to assess factors associated with SGA and preterm delivery. To minimize overestimation of the relative risk when using Poisson regression for a categorical covariate of interest, a robust variance procedure was used (Lin and Wei, 1989). This method was considered to be a better alternative than the logistic regression for analysis of cross-sectional data, with the latter producing higher estimates than the former (Coutinho et al., 2008). For ease of comparison with other studies, we repeated our analyses using the conventional binary logistic regression and found essentially similar results to those presented here using Poisson regression. Variables were considered for inclusion in the multivariable model. Goodness-of-fit tests, assessed using the likelihood-based Akaike information criteria (AIC), were used to determine the variables that significantly improved the final Poisson models' fit (Bruin, 2006). Mortality by maternal HIV status and SGA was assessed using Kaplan–Meier survival analysis. Follow-up time was computed as time from birth to withdrawal, loss to follow-up, migration or death if before the first year of life, whichever came first, allowing for child's HIV status as a time-varying variable for infant mortality analyses. Risk factors for infant mortality were assessed in univariable and multivariable Cox regression models with child HIV infection as a time-dependent variable; for the adjusted model, variables with a statistically significant association univariately were included in addition to three included (parity, maternal age and education) a priori (Ndirangu et al., 2010).