Postpartum obesity and weight gain among human immunodeficiency virus-infected and human immunodeficiency virus-uninfected women in South Africa

In South Africa, up to 40% of pregnant women are living with human immunodeficiency virus (HIV), and 30–45% are obese. However, little is known about the dual burden of HIV and obesity in the postpartum period. In a cohort of HIV-uninfected and HIV-infected pregnant women initiating antiretroviral therapy in Cape Town, South Africa, we examined maternal anthropometry (weight and body mass index [BMI]) from 6 weeks through 12 months postpartum. Using multinomial logistic regression, we estimated associations between baseline sociodemographic, clinical, behavioural, and HIV factors and being overweight–obese I (BMI 25 to <35), or obese II-III (BMI >35), compared with being underweight or normal weight (BMI <25), at 12 months postpartum. Among 877 women, we estimated that 43% of HIV-infected women and 51% of HIV-uninfected women were obese I-III at enrollment into antenatal care, and 51% of women were obese I-III by 12 months postpartum. On average, both HIV-infected and HIV-uninfected women gained, rather than lost, weight between 6 weeks and 12 months postpartum, but HIV-uninfected women gained more weight (3.3 kg vs. 1.7 kg). Women who were obese I-III pre-pregnancy were more likely to gain weight postpartum. In multivariable analyses, HIV-infection status, being married/cohabitating, higher gravidity, and high blood pressure were independently associated with being obese II-III at 12 months postpartum. Obesity during pregnancy is a growing public health concern in low- and middle-income countries, including South Africa. Additional research to understand how obesity and HIV infection affect maternal and child health outcomes is urgently needed. To address this aim, we conducted a secondary analysis using data from HIV‐infected women enrolled in the Strategies to Optimize ART Services for maternal and child health (MCH‐ART) trial and a parallel cohort of HIV‐uninfected pregnant women (HIV‐unexposed–uninfected study) conducted in Cape Town, South Africa. Details of both studies have been published previously (le Roux et al., 2019; Myer et al., 2016; Myer et al., 2018). Briefly, the two cohorts had similar inclusion and exclusion criteria, enrolling HIV‐infected pregnant women initiating ART and HIV‐uninfected pregnant women who were > 18 years of age between March 2013 and August 2015 at their first antenatal care (ANC) visit at a primary care center in Gugulethu in Cape Town. Women who were breastfeeding at their first postpartum visit, scheduled 7 days after delivery, were enrolled and followed through 12 months postpartum. Out of 1,087 mother–infant pairs screened at their first postpartum visit, 92 women (79 HIV‐infected women and 14 HIV‐uninfected women; 8% overall) were excluded due to not breastfeeding (le Roux et al., 2019). Gugulethu is an urban community of approximately 300,000 people outside of Cape Town and is characterized by high levels of poverty and HIV (Myer et al., 2018; Strategic Development Information & GIS Department, 2013). Over 95% of women in this setting receive ANC prior to delivery (Myer et al., 2015). Provision of ART and prevention of mother‐to‐child transmission (PMTCT) services are provided at no cost as a part of routine ANC at all public sector clinics, in accordance with local guidelines. Starting in 2013, all HIV‐infected women attending ANC were eligible for lifelong ART, regardless of CD4 count of WHO clinical stage (WHO, 2013). All HIV‐infected women in the study initiated the local first‐line ART regimen of tenofovir (300 mg) + emtricitabine (200 mg)/lamivudine (300 mg) + efavirenz (600 mg), provided as a fixed‐dose combination pill taken once daily. HIV‐uninfected and HIV‐infected women initiating ART were included in the present analysis if they met eligibility criteria, were followed through 12 months postpartum (n = 884), and had a singleton pregnancy (n = 7 excluded). Participants completed study visits at enrollment into ANC (baseline), delivery, 6 weeks, and 3, 6, 9, and 12 months postpartum. Maternal anthropometry was assessed by maternal weight (in kilograms), BMI (calculated as kilograms/meters2), and changes in maternal weight from 6 weeks (baseline) through 12 months postpartum. Postpartum weight change was defined as either no weight change (within +/−2 kg), weight loss more than 2 kg, or weight gain more than 2 kg, between 6 weeks postpartum and each time point. Women were weighed, and their height was measured at enrollment into ANC (median gestational age 20 weeks, range 4–39) and at all postpartum visits by trained data collectors following standard operating procedures. For example, women were weighed on a calibrated digital scale (such as the Charder MS7301 250 Kg Digital Scale) with their shoes and extra layers of clothing removed and were measured using a portable stadiometer (Seca 213 Stature Meter Free‐Standing Stand). Competency checks, repeat training, and random quality control checks were conducted throughout the study period. BMI was categorized as underweight (<18.5), normal (18.5 to <25), overweight (25 to <30), obese I (30 to <35), obese II (35 to <40), and obese III (>40). Maternal weight at delivery was not measured; therefore, we were not able to estimate gestational weight gain. In addition, information on the development of gestational diabetes, pregnancy‐induced hypertension, preeclampsia, or postpartum hypertension or diabetes was not available. At enrollment into ANC, information on baseline sociodemographic, clinical, behavioural, and HIV disease (if applicable) characteristics was collected. Gestational age at enrollment into ANC was determined using ultrasound. A composite poverty score developed by our research team, calculated from current employment, housing type and access to household assets, was used to categorize women as “most,” “moderate,” or “least” disadvantaged (Brittain et al., 2017). Perinatal depression was measured using the Edinburgh Postnatal Depression Scale (range 0–30; Chorwe‐Sungani & Chipps, 2017; Cox, Holden, & Sagovsky, 1987). A score of >13 was used to indicate probable depression (Redinger, Norris, Pearson, Richter, & Rochat, 2018). Alcohol use was measured using the 3‐item Alcohol Use Disorders Identification Test‐Consumption (AUDIT‐C; range 0–12). For women, an AUDIT‐C score >3 indicates hazardous drinking (Bush, Kivlahan, McDonell, Fihn, & Bradley, 1998). Blood pressure was measured at baseline and categorized as normal (<120/80 mm Hg), elevated (systolic 120–129 and diastolic <80 mm Hg), Stage 1 hypertension (systolic 130–139 or diastolic 80–89 mm Hg), or Stage 2 hypertension (systolic >140 or diastolic >90 mm Hg; American College of Obstetricians and Gynecologists, 2018). Due to the few women who had Stage 2 hypertension, women with Stages 1 and 2 hypertension were combined in statistical analyses. Among HIV‐infected women, CD4 cell count (<200, 201 to <350, 350 to <500, >500 cells/mm3) and viral load (<1,000, 1,000 to <10,000, and >10,000 copies/ml), timing of HIV diagnosis (during the current pregnancy or previously) and use of antiretroviral prophylaxis for PMTCT in a previous pregnancy, and previous combination ART use were assessed at enrollment into ANC. The goals of the statistical analysis were to describe maternal weight, BMI, and weight change overtime and by HIV status during the postpartum period, as well as to estimate associations between demographic, clinical, behavioural factors and HIV status, and being overweight or obese at 12 months postpartum. Maternal anthropometry overtime and by HIV status was examined descriptively and graphically. To explore associations with being overweight or obese, we categorized women into one of three groups: underweight or normal weight (BMI <25), overweight or obese I (BMI 25 to <35), or obese II or III (BMI >35). Because only 4% of the population was underweight, effect estimates could not be estimated separately for this group. Therefore, we grouped underweight women with normal weight women in order to retain as much of the sample as possible to maximize statistical precision. We used multinomial logistic regression to estimate odds ratios (OR) for associations between baseline factors and being overweight–obese I (outcome 1), or obese II‐III (outcome 2), compared with being underweight or normal weight (referent), at 12 months postpartum. In bivariable analyses, all factors with a p value <.05 were considered for inclusion into the multivariable model. All variables considered for inclusion into the multivariable model were evaluated for collinearity using pairwise correlation coefficients. For variables with a pairwise correlation coefficient >.50, we selected the variable with the stronger bivariable association for inclusion into the multivariable model. When evaluating associations with BMI at 12 months postpartum, BMI at first ANC visit was highly collinear with BMI at 12 months postpartum and, therefore, was not included due to model convergence issues. Gestational age at enrollment into ANC and breastfeeding duration are likely intermediary variables, between several variables in the model and BMI category at 12 months postpartum, and therefore were not included in models (Ananth & Schisterman, 2017; Hernandez‐Diaz, Schisterman, & Hernan, 2006). As an exploratory, secondary analysis, we explored predictors of change in maternal weight (categorized as weight loss, no weight change, and weight gain using the category definition above) between 6 weeks postpartum and 12 months among the 596 women with a weight measurement at 6 weeks and 12 months postpartum. Bivariable and multivariable analyses were analogous to those described above. Among HIV‐infected women (n = 464), we examined whether HIV‐related factors, including timing of HIV diagnosis, previous PMTCT during pregnancy, viral load, and CD4 count, were associated with BMI at 12 months postpartum. Previous ART use was not included due to the majority of participants initiating ART for the first time. Missing BMI and blood pressure data at enrollment into ANC were common (10–12%) and more frequent among HIV‐infected women. To address potential bias due to missing data, we conducted a sensitivity analysis where we used multiple imputation to impute all missing data from a multivariate normal distribution (N = 50 imputations; Rubin, 1987). We then explored associations between baseline factors and BMI at 12 months postpartum in the imputed data. All statistical analyses were conducted in Stata version 15 (StataCorp, College Station, TX). Ethical approval for both the MCH‐ART and HU2 studies was provided by the University of Cape Town’s Human Research Ethics Committee. The MCH‐ART study also received ethical approval from the Columbia University Institutional Review Board.