Setting and participants
We conducted a cross-sectional HIV/cardiometabolic risk factor survey in 2010 – 2011 using an age-sex stratified random sample of those aged 15+ in the Agincourt sub-district, situated in rural north-eastern South Africa adjacent to the country’s border with Mozambique. The sub-district has been under health and demographic surveillance since 1992, with collection of longitudinal data on all vital events and possible social, economic and environmental modifiers [12,13]. It is one of South Africa’s poorest areas, characterised by limited subsistence farming, shortage of services and high levels of migrant labor coupled with limited local employment opportunities. The area characteristics and the surveillance system are explained in more detail elsewhere [14].
For this cross-sectional study we randomly selected 7662 individuals ages 15+, stratified by age and sex, from an eligible population of 34,413 using the 2009 census update as the sampling frame [15]. In addition, we included 284 adults over age 50 years who participated in the INDEPTH-WHO/SAGE (http://www.who.int/healthinfo/sage/en/) ageing and adult health study in 2006 [16]. Of these individuals, 4362 consented to be interviewed and tested; we restricted our estimation sample to ages 18+ with complete covariate data (n = 3641). We used an indicator of household socioeconomic status [17] from the 2009 census update.
Ethical approval
The study received ethical approvals from the University of the Witwatersrand Human Research Ethics Committee and the Mpumalanga Provincial Research and Ethics Committee. Written consent to participate (or parental consent and assent for minors ages 15–17) was obtained for all participants.
Availability of supporting data
Data extraction of the Agincourt HDSS can be requested from Dr. F. Xavier Gómez-Olivé (F.Gomez-OliveCasas@wits.ac.za). Complete data sharing and collaboration details are available elsewhere [14].
Procedures
A questionnaire on cardiometabolic risk factors, medication use and sexual behaviour was administered by experienced local fieldworkers who visited households up to three times to recruit the sample population. Five dried blood spots provided measures of HIV-status (HIV-, untreated HIV+, and treated HIV+). Treatment status was assessed by self-report. Point-of-care instruments were used to measure glycaemia, total cholesterol and lipid sub-fractions. Respondents were asked about smoking and alcohol history, physical activity, dietary intake, time since the last meal and whether they were using antiretroviral therapy. Physical measurements included height, weight and waist circumference using a flexible stadiometer (Seca, Hamburg, Denmark); and Analysis Scale Body Check (Seca, Hamburg, Denmark). Blood pressure was measured three times with a Boso blood pressure instrument (BOSCH + SOHN, Jungingen, Germany) three minutes apart, taken on the left wrist with participant in sitting positions and at rest; taking the average of the last two measurements. Random blood glucose was measured with a Caresens POP blood glucose meter (i-Sens, Nowon-gu, Seoul Korea); and total cholesterol and lipid sub-fractions were measured using a Cardiochek instrument (Polymer Technology Systems Inc., Indianapolis, IN USA). Socio-demographic characteristics such as years of formal education and employment were extracted from the existing surveillance system database.
Variables used
To evaluate the association between cardiometabolic risk factors and HIV status, we determined different cut-offs for nine indicators by referring to published literature and aimed, whenever possible, to select an internationally recognized definition: high waist circumference as greater than 102 cm for men and 88 cm for women [18]; hypertension as systolic blood pressure greater than or equal to 140 mmHg or diastolic blood pressure greater than or equal to 90 mmHg or current use of antihypertensive medication [18]; high LDL cholesterol as greater than 3 mmol/l [19]; low HDL cholesterol as less than 1.03 mmol/l for men and 1.29 mmol/l for women [18]; high triglycerides as greater than or equal to 1.7 mmol/l [18]; high total cholesterol/ HDL ratio as greater than 8 [19]; diabetes as blood glucose (fasting or not; most of the blood samples were non-fasting) greater than or equal to 11.1 mmol/L [19]; obesity as body mass index of (BMI; kg/m2) 30 or higher [18]; and HIV-serostatus (HIV-, untreated HIV+, and treated HIV+). Personal and household socio-demographic characteristics included sex, age, formal education received (none, 1–5, 6+ years), smoking history (ever smoked or not), physical activity (low, moderate, high) [20], household asset score (in tertiles), employment status (currently employed or not), and alcohol frequency in the past month (no use, 1–3 days/month, 1–4 days/week, or 5+ days/week).
Statistical methods
We calculated the unadjusted prevalence of HIV and cardio-metabolic risk factors by sex; and age-adjusted prevalence using the Agincourt 2009 census population. We used logistic regression to assess associations between cardiometabolic risk factors and HIV-status and a number of socio-demographic variables including sex, age, formal education, smoking history, physical activity, household asset score, employment status, and alcohol intake in the past month. We first fitted baseline models including HIV-status and covariates assumed to be unaffected by HIV-status (education and smoking history). We then estimated the full model with all covariates to determine if these factors moderated the association of HIV-status with each outcome. The HIV-status coefficients were not substantively changed by including all covariates so we present the full estimation results. All analyses were completed using STATA 12.1 and included sampling weights [21].