Study design, setting and participants
We conducted a cross-sectional analysis comparing the associations between anthropometric and BIA derived adiposity measures with blood pressure and hypertension in India. The Indian Study on Health of Adults (ISHA) is an ongoing population-based cohort study of men and women age 30–69 years from the general population of the town of Barshi (Maharashtra, India). We used baseline data collected between 2015 and 2016 for our analyses (N = 5996). Of these participants, 6 were excluded because of pregnancy.
Study data sources
Upon arrival to each village, and prior to the recruitment of participants, survey teams met with the local Sarpanch or authorized personal of the village to seek permission for the study. In addition, group meetings were also conducted before the start of the survey to create awareness of the study for those residing in the village community.
Trained surveyors collected data for the baseline survey in three stages over a 7-day activity cycle: enumeration (days 1–2), household health survey (days 3–4) and a health checkup camp (days 5–7). First, field teams enumerated the households in the village, collecting information on household size and the usual residing members. Second, trained surveyors (a male and a female) visited the households and interviewed eligible members (30–69 years of age) to obtain detailed information on demographic, socioeconomic, and lifestyle characteristics, along with blood pressure measurements and antihypertensive medication use. After completion of the interview participants were invited to attend a health checkup camp, and were given an invitation card detailing the time and place of their health checkup. Each health checkup camp was set up in the village with direct input from the Sarpanch or any other administrative head to ensure ease of access for participants. Lastly, participants attended the health checkup camp where their blood pressure was measured, along with physical and body composition measurements.
Blood pressure measurement
After 5 min of rest, three blood pressure measurements were taken at heart level in a seated position using the Omron BP-742CAN (Kyoto, Japan) digital automatic monitor. Systolic blood pressure (SBP) and diastolic blood pressure (DBP) were calculated as the average of the three readings. We calculated pulse pressure, mean arterial pressure, and mid-blood pressure using previously reported formulas [23]. Hypertension was defined as SBP ≥140 or DBP ≥ 90, or reported use of antihypertensive medication.
Physical measurements
A measuring tape was used to measure height to the nearest 0.1 cm [23]. Weight was measured to the nearest 0.1 kg using the Tanita MC780MA body composition analyzer (Tokyo, Japan), without footwear and with subtraction of 0.5 kg for clothing weight. WC and HC were both measured to the nearest 0.1 cm using a measuring tape. For WC, measurements were taken at the level of the umbilicus with arms folded across the chest. For HC, measurements were taken at the point yielding the maximum circumference over the buttocks.
Adiposity measures
The anthropometric measures of adiposity were BMI, WC, HC, WHR, and WHtR. BMI was calculated as weight in kilograms divided by the square of height in meters. WHR was calculated as WC in centimeters divided by HC in centimeters, and WHtR as WC in centimeters divided by height in centimeters. The BIA derived measures of adiposity were whole body fat percentage and trunk fat percentage, as estimated with proprietary algorithms by the Tanita body composition analyzer.
Statistical analysis
We performed all analyses separately for men and women. We calculated Pearson’s partial correlation coefficients for the intercorrelations of the different adiposity measures (BMI, WC, HC, WHR, WHtR, whole body fat percentage, trunk fat percentage) and blood pressure components (SBP, DBP, pulse pressure, mean arterial pressure, mid-blood pressure), adjusted for age. In order to compare the associations between anthropometric and BIA derived adiposity measures with blood pressure and hypertension, we evaluated the relationships using several methods. First, we used multiple linear regression models to quantify the associations between adiposity measures with blood pressure components. Second, we used Poisson regression models to examine the relationships between adiposity measures and hypertension. We estimated the means of each blood pressure component, and calculated the prevalence ratios (PR) for hypertension per sex-specific standard deviation (SD) change in an adiposity measure. We used two models for these analyses. In the first model, adjustments were made for age and level of education (illiterate, primary school, middle school, secondary school, college). In the second model, further adjustments were made variously for either an anthropometric or BIA derived measure of general (BMI, whole body fat percentage) or central (WC, trunk fat percentage) adiposity. 198 individuals (3% of total) who reported current use of antihypertensive medication were excluded from the continuous blood pressure analyses.
Lastly, we used receiver operating characteristic (ROC) curves and the area under the curve (AUC), estimated by logistic regression models with adjustments for age and education, to compare the discriminative ability of each adiposity measure for hypertension. The AUC is a measure of the overall discriminative ability of each adiposity measure for hypertension, with values falling between 0.5 and 1.0, representing no discriminative and perfect discriminative ability, respectively. The AUC for all measures of adiposity were compared using a nonparametric approach for the comparison of multiple AUC from ROC curves [24].
In addition to age and education adjustments, further adjustment for tobacco use (non-user, less than daily user, daily user) and alcohol consumption (non-drinker, current drinker) did not substantially affect any of the estimates (results not presented). We performed all statistical analyses using SAS version 9.3 (SAS Institute, Cary, NC, USA), and provide estimates with their respective 95% confidence intervals (CI).