A detailed description of methods used in the parent PaCT study from which this analysis is derived has been reported elsewhere [10]. Here we only describe methods relevant to results presented in this article.
Study design
The parent study was conducted using a cohort study design between January 2011 and July 2012. Participants were followed for up to six months, with evaluations at baseline, and at six months only. Results presented in this article are a cross-sectional analysis of the baseline data.
Study population
The study was conducted in four sub-Saharan countries, including Tanzania, South Africa, Uganda and Nigeria. Participants were enrolled from 5 different population groups, three defined by occupation, and two by degree of urbanization. Tanzania and South Africa enrolled school teachers, Nigeria enrolled nurses, whereas Uganda enrolled rural and peri-urban residents. Eligible subjects were adults aged 18 years or older, with no intension of migrating outside of their community of residence within the next 6 months (in Uganda), or retiring from service in the next 6 months (for Tanzania, South Africa and Nigeria).
School teachers in Tanzania were enrolled from 18 randomly selected public schools in Dar es Salaam. In South Africa, school teachers were enrolled from government schools in Cape Town Metropolitan area, where 111 schools with 20 or more teachers were invited to participate in the study. In Nigeria, a random sample of nurses were enrolled from two urban hospitals; one located in Abuja city and another in a semi-urban setting 1.5 h outside of Abuja city. In Uganda, participants were enrolled from two geographic locations; a peri-urban community in the Wakiso District 10 miles north of Kampala city, and from a rural community in Bushenyi District 200 miles west of Kampala city. In Wakiso District, a random sample of households was selected from two parishes comprising 13 villages. In Bushenyi District, households were randomly selected from an enumerated list of all the households in each village.
Ethics, consent and permissions
Informed consent was obtained from each subject either by voluntarily posting back a signed form with a completed questionnaire (South Africa and Tanzania) or through documentation with trained interviewers (Nigeria and Uganda). The pilot studies were approved by the: a) Harvard School of Public Health Institutional Review Board, b) Institute of Human Virology Heath Research Ethics Committee in Nigeria; c) Health Research Ethics Committee of the Faculty of Health Sciences, Stellenbosch University in South Africa; d) National Institute for Medical Research in Tanzania; and in Uganda; e) Makerere University School of Public Health Higher Degrees, Research and Ethics Committee; f) Mbarara University of Science and Technology Ethics Committee, and g) the Uganda National Council of Science and Technology.
Measurements
We used a standardized questionnaire to collect data on socio-economic characteristics, history of infectious and chronic disease diagnoses, mental health, injuries, and common risk factors for non-communicable diseases (NCDs) including tobacco use, alcohol use, and physical activity. Most of the questions were adapted from the World Health Organization STEPS instrument developed for use in resource-limited countries [11]. Physical and biochemical measurements were also made following a standardized protocol across the five study sites. Physical measurements included weight, height, and blood pressure. Height was measured without shoes, to the nearest centimetre. Weight was measured with the participants in light clothing and without footwear using a weighing scale to the nearest tenth of a kilogram. Blood pressure measurements were taken on the left arm with the participant in the sitting position using a cali-brated electronic blood pressure device (Welch-Allyn®). Three systolic and diastolic blood pressure measurements were taken at least five minutes apart. The average of the last two blood pressure readings was used in this analysis.
After administering the questionnaire, interviewers requested participants for an appointment to return the following morning so that a blood sample could be obtained to measure plasma glucose levels. Participants were instructed to fast overnight and no exercise or smoking in the morning, in preparation for obtaining a blood sample to conduct a fasting plasma glucose (FPG) measurements. The following morning the interviewer returned and collected the participant’s finger prick blood sample for an FPG test using a digital glucometer (On-Call® Plus, ACON Laboratories). Only participants reporting compliance with an overnight 8-h fast, no exercise or smoking that morning were eligible for the finger prick blood sample collection. Non-compliers were rescheduled to a future date where possible; otherwise this procedure was omitted in such participants. FPG levels were recorded in milli moles per liter (mmol/L).
Statistical analysis
Participants were classified in different blood pressure categories based on the World Health Organization (WHO) cut-off criteria [12, 13]. Thus a participant was classified as being hypertensive if their average systolic blood pressure (SBP) was at least 140 mm Hg, and/or their average diastolic blood pressure (DBP) was at least 90 mm Hg, or if they reported being on regular anti-hypertensive therapy. A participant was classified as being pre-hypertensive if their average SBP was between 120 and 139 mm Hg (inclusive), and/or their DBP was between 80 and 89 mm Hg (inclusive), and not on any anti-hypertensive therapy.
The prevalence of pre-hypertension and hypertension were calculated as the percentage of participants classified as being pre-hypertensive or hypertensive, respectively. To enable comparison of prevalence across the five population groups with differing age-structures, prevalence was directly age-standardized to the World Health Organization’s 2000–2025 world standard population age structure [14, 15].
Body Mass Index (BMI) was computed by dividing the weight (kg) by the height in meters squared (m2) and used to develop the categories of: underweight (less or equal to 18.5), normal weight (18.5–24.9), overweight (25.0–29.9), and obese (30 or higher). Participants were also categorized on the basis of their plasma glucose levels status using the WHO definition and diagnosis of diabetes mellitus criteria as follows: Normal (less than 6.1 mmol/L), Pre-diabetes (6.1 to 6.9 mmol/L), and diabetes or probable diabetes (greater than 6.9 mmol/L, or currently on anti-diabetes mellitus medication) [16].
To identify factors associated with hypertension, the modified Poisson regression model with robust variance was used to estimate both the crude and adjusted prevalence ratios (PR) [17, 18], with their corresponding 95 % confidence intervals (95 % CI). The modified Poisson regression model, a model that uses a robust error variance, was preferred to avoid under estimation of the standard errors for the estimated risk ratios that is usually the case with logistic regression modelling when the prevalence of the outcome is greater than 10 % [17, 18]. Potential confounding and interaction variables that were assessed for inclusion in the model were: age, sex, level of education, body mass index, population group, FPG category, family history of hypertension (first degree relatives only), whether the participant met the WHO physical activity recommendations [19], and tobacco use category. A step-by-step backward elimination of these variables from the model was used to identify those to be retained in the model. The criterion used to retain variables in the model were if a variable was significantly associated with hypertension using a 5 % level of statistical significance (α = 0.05), or if addition of the variable in the model led to a change of at least 10 % in any of the significant PR estimates of any variable already in the model. Cases with missing observations were excluding from the model on a case by case basis. All statistical analyses were performed using STATA version 13 (StataCorp, College Station, Texas, USA).
