Setting and Population
The Agincourt sub-district is in South Africa's rural north-east, adjacent to Mozambique, where the Wits/MRC (Agincourt) Rural Public Health and Health Transitions Research Unit has been monitoring causes of death, births and migration in a population of around 70 000 people since 1992[
14]. Each individual has a unique identity number, as does each household. Data is updated annually by census, with trained field workers visiting every household.. This dataset therefore provides a well categorised sampling frame for studies such as this.
The sub-district has 21 villages of varying sizes and is heavily populated at about 170 persons per sq km. Electricity is available in most villages, but not all households. At the time the study was undertaken, some villages had no standpipe to provide clean water and supply from standpipes was limited and sporadic in all other villages. There is high unemployment and, as is common in rural sub-Saharan Africa, considerable labour migration, especially in men. Household plots are too small for subsistence agriculture, although crops supplement the diet [
15].
Most people access plural health care from allopathic health professionals, traditional healers, and prophets linked to churches that offer faith based healing[
16]. There are five publicly funded primary care clinics and a large health centre staffed by nurses, with occasional visits from doctors. Treatment and drugs are free at these clinics. Three district hospitals serve the wider district, and there are a number of private practitioners.
Sample
We stratified the Agincourt villages by size (less than 500 households, and 500 households or more) and by whether the village was a formal settlement (officially recognised) or an informal settlement (settled following an influx of refugees during the Mozambican civil war). Ten villages were randomly selected: four small informal villages, three small formal villages, and three large formal villages (there were no large informal villages). The total population of the selected villages was 28,715, representing 43% of the population of the sub-district. A random sample of individuals from the resident (i.e. excluding migrant workers) population in selected villages aged 35 years and older, provided 526 individuals. This was approximately 10% of the population aged 35 years and over in the selected villages, or 3% of the 16, 705 sub-district population aged 35 years and over). We considered migrant workers or temporary migrants to be people who were linked to a household and provided financial support but lived in the area for less than 6 months of the year because they were employed elsewhere.
Information on age, gender and the household asset score was available from the Agincourt database. The household asset score was based on information collected during the 2001 census about the type and size of dwelling; access to water and electricity; appliances and livestock owned and transport available. The questionnaire contained 34 variables, which were developed in discussion with local field staff and community members and the score was developed using a principal component factor analysis. The score was divided into quintiles labeled: low, medium-low, medium, medium-high and high economic status[
17]. A similar household economic status index was reported to discriminate effectively among levels of economic status in other rural African settings[
18].
We made at least three attempts to visit each individual. A second sample provided substitutes for individuals who were ineligible (for example had died, were migrant workers, were aged < 35 years, or were not living there) or who were untraceable. Subjects who declined to participate, were too ill, or did not keep appointments were not replaced. In three households in a remote village the interview was abandoned for logistic reasons.
Consent and ethics
We obtained informed consent from the village headman and through village community meetings, before starting work in each village. In addition, each participant was asked to consent separately in writing to each component of the study, (the interview, the anthropometric and blood pressure measurements, and venesection). Many participants were illiterate and marked the form with a thumbprint, but we ensured they understood the content of the form. Ethics committee approval was granted by both the London School of Hygiene and Tropical Medicine (755) and the University of the Witwatersrand (M02-04-63).
Interview and clinical examination
In 12 months from July 2002 two specially trained senior nurses from the local community carried out interviews and clinical examinations. The interview covered a past medical history, including hypertension, and smoking habit. Participants were asked about any treatment for high blood pressure in the last week, and asked to show the medication to confirm the name.
The nurses were trained in anthropometric and blood pressure measurement by a member of staff from the Medical Research Council, which carried out the South African Demographic and Health Survey [
19]. They measured height using a portable, collapsible stadiometer (Leicester Height Measure – Seca, Ltd, Birmingham, UK), which was placed on a firm level surface. They then asked participants to remove any headdress and footwear, to stand on the stadiometer platform with their heels together touching the backstop to ensure that their spine at pelvis and shoulder level, as well as the back of their head touched the upright rod of the stadiometer. The nurse then positioned the participant's head so that an imaginary horizontal line joined the corner of the participant's eye and the upper attachment of their ear to their head. They then firmly lowered the measuring arm of the stadiometer onto the participant's head without forcing their head down. Participants were then asked to move away from the stadiometer and the nurse read the measurement to the nearest 0.1 cm.
For waist measurement, the nurse asked the participants to stand upright in a relaxed manner with their arms slightly away from their sides and their feet about 15 cm apart. The nurse then placed the measuring tape around the narrowest part of the bare waist, 2.5 cm above the umbilicus. In difficult cases, for instance, obese persons, the nurse stood behind the participant to find the narrowest part of the waist. If this was not the narrowest portion of the waist, they then measured the smallest circumference between the xiphi sternum and umbilicus. They took the measurement with the measuring tape in a horizontal position to the nearest 0.1 cm after normal expiration, taking care not to insert their fingers under the tape and not to let the tape cut into the participant.
Participants were weighed barefoot and wearing at most light clothing using a UC-321 Precision Health Scale UC-321 scales (A&D Medical). The scale was placed on a flat surface and cleaned with disinfectant. A clean paper towel was placed on top of the scale prior to each measurement. The nurse then switched the scale on and ensured that it read zero. They then asked the participant to step onto the scale and stand with their feet firmly on the middle of the scale without moving while the nurse recorded the reading to the nearest 0.1 kg.
Sitting blood pressure was measured in the left arm using an OMRON 705CP blood pressure monitor and an appropriately sized cuff after the participant had been sitting for five minutes with their arm supported. Three measures were taken two to three minutes apart. Supine arm blood pressure was measured with the same monitor after the participant had been lying for 2 to 3 minutes. Ankle systolic pressures were measured in the posterior tibial artery of both legs using a Doppler ultrasound probe (Huntleigh Mini-Dopplex) and a Mandaus manual digital blood pressure monitor (PMS instruments) with the cuff placed just above the malleoli.
When the nurse identified any problem requiring treatment she gave the participant a letter of referral for the local clinic.
We assessed the quality of the anthropometric measures on several occasions. About midway through the study the MRC nurse trainer visited the site and accompanied the nurses on their visits, assessing their techniques and re-training as required. Several of us (MT, ST, GH and MC) visited the site and accompanied the nurses resulting in approximately monthly quality assessments, and one of us (MC) was available at least fortnightly on site to manage any queries from the nurses.
Blood Analysis
Blood samples were transported in cold boxes, centrifuged locally at the end of each day, and transferred by courier to the University of the Witwatersrand Contract Laboratory Services. Analysis was done using the Roche Cobas Integra 400 System. Total and HDL cholesterol were measured using an enzymatic, colorimetric method,
Clinical Definitions
We used the mean of the second and third blood pressure measurements and the Joint National Committee definition of hypertension (JNC7) [
20] criteria to define hypertension. A person was considered to have high blood pressure if the systolic pressure was at or greater than 140 mmHg, or the diastolic at or greater than 90 mmHg, or they were using anti-hypertensives. As with JNC7, we further divided high blood pressure into Stage 1 (systolic 140–159 mmHg, and/or diastolic 90–99 mmHg) and Stage 2 (systolic 160 mmHg or higher, and/or diastolic 100 mmHg or higher).
In developed populations, the ankle brachial index is a good marker of subclinical peripheral arterial disease and an indicator of overall atherosclerosis[
21]. A cut-off of the ankle brachial index of 0.9 is commonly used to distinguish individuals at high and low risk of clinical atherosclerosis. We used the supine systolic pressure together with the pressure as measured at the ankle (see above) to calculate the ankle brachial index in each leg. We have reported the lowest of the two ankle brachial indices, together with the proportion of participants with an ankle brachial index at or less than 0.9.
Physical activity
At the time, we were unable to find a validated physical activity questionnaire that was suitable for use in this environment where we found no concept of leisure time physical activity, and where walking carrying a load formed a major part of daily activity. We therefore adapted the validated International Physical Activity Questionnaire (IPAQ)[
22], and categorised respondents as moderately active, low active or sedentary on the basis of their responses. Table
1 provides details of how respondents' physical activity was categorised.
Table 1
Definitions of physical activity categories
Any one of: | Has a job which involves at least moderate activity |
| Walks with heavy loads and/or cycles for at least 2.5 hours a week |
| Does 1 hour or more of vigorous non-work activity, e.g. football |
| Has a job that involves standing and walking AND Either walks with loads and/or cycles for at least 1 hour, or walks without loads for at least 2.5 hours a week. |
Low Activity
| Not sedentary but not active enough to fit the vigorous or moderate activity categories |
Sedentary
| Had no job, or a job that only involves sitting and does no walking, cycling, or vigorous non-work activity. |