ABC | Volume 113, Nº3, September 2019

Original Article Santos et al. Arterial hypertension in quilombola communities Arq Bras Cardiol. 2019; 113(3):383-390 was performed through random selection in the quilombola communities, using the existing proportion of the population in the communities. These data were provided by the National Institute of Colonization and Agrarian Reform (INCRA). 1 A random sample of clusters was selected in two stage cluster sampling. There are 35 quilomobola communities registered in the State of Sergipe, distributed in eight territories, of which four were randomly selected. Out of these four territories, 15 communities were randomly selected from a total of 19. Between 15% and 20% of the adult population voluntarily participated. For each stage, once the territories and the quilombola communities were registered, the random sampling without replacement was performed using the Stata ® version 15.1 software. The communities studied are far from the city headquarters, in areas of difficult acess. Certain communities (Resina and Pontal da Barra) surround the main river in the region and the sea, respectively. The other communities (Mocambo, Canta Galo, Pirangy, Terra Dura, Forte, Caraíbas, Bongue, Patioba, Ladeiras, Alagamar, Aningas and Quebra Chifre) are situated in large land properties. The quilombola community “Maloca” is the only one that is located in urban area among the other remaining communities in the state. 1 The target population of the research, according to oficial registries, 1 was estimated in 1,979 adult individuals, inhabitants of the 15 quilombola communities. Sample size calculations were done using the G*Power 3 software, 11 respecting the following parameters: 80% power; two-sided alpha = 0.05; covariable distribution pattern; log-normal distribution; potential correlation between predictors, 0.80; expected prevalence of arterial hypertension in the general population (20.4%). 8 According to these parameters, about 350 individuals would be necessary to detect an odds ratio ≥ 1.5 for differences between categorical predictors, in multiple regression logistic analysis. With the aim of preserving these characteristics in a potential situation of missing data, the sample size was increased to about 10%, totalling 390 individuals. The inclusion criteria adopted for individual selection were: age ≥ 18 years; and being registered as quilombolas in the communities where they belong and in the INCRA. The exclusion criteria were: practice of physical exercise in the last 60 minutes; ingestion of alcoholic drinks, coffee or food; use of cigarette or consumption of other substances within the 30 minutes prior to blood pressure measurement; pregnancy; and amputated upper limbs. Clinical and Sociodemographic Data Collection The data were collected using individual interview. The interviewers were trained for this procedure. The interview instrument used was a semi-estructured questionnaire adapted from the following studies: the Brazilian Ministry of Health’s Food Guide 12 and the Evaluation of Physical Activity Program Effectivity in Brazil, 13 both published by the Brazilian Ministry of Health; the National Household Sample Survey ; 14 the criteria of economic classification of the ABEP (Brazilian Association of Market Research Firms), which divides society into economic classes A, B1, B2, C1, C2, D-E, considering household assets, education level and the public services available. 15 The questions related with licit and illicit drugs were based on the Brazilian version of ASSIST (Alcohol, Smoking and Substance Involvement Screening Test) 16 The previous history of diseases was based on the questions asked for admission to hospital due to primary care-sensitive conditions. 17 Then, three blood pressure measurements were perfomed (with a 1-minute interval between each measurement). A Welch Allyn DuraShock™ DS44 (Welch Allyn, Curitiba, Brazil), internationally validated, Aneroid Sphygmomanometer, with nylon cuff and metal clasp, was used. The measurements were performed at the end of the interview. During BP measurements, the individuals remained seated, with their legs uncrossed, feet flat on the floor, back supported by the back of a chair and relaxed. The individual’s left arm was positioned for measurement, followed by the right upper limb. The third measurement was performed on the limb that presented the highest value, always with the arm rested on a table, at heart level. For analysis, the mean of the three measurements was calculated, which corresponded to the research criteria, being considered hypertensive those individuals who had systolic arterial pressure ≥ 140 mmHg and/or diastolic arterial pressure ≥ 90 mmHg. 9 These more conservative measurements have been adopted because the three measurements of the blood pressure were performed in only one day. For this reason, the classification of the American Heart Association was not adopted. 18 The Body mass (BMI) index [kg/m 2 ] was estimated to evaluate the anthropometric measurements (weight and height). The BMI found was categorized according with the following measures: low weight, < 18.5 kg/m 2 ; normal weight, 18.5 to 24.5 kg/m 2 ; overweight, 25 to 29.9 kg/m 2 ; level I obesity, 30 to 34.9 kg/m 2 ; level II obesity, 35 to 39.9 kg/m 2 ; and level III obesity, > 40 kg/m 2 . 19 Statistical analysis Categorical variables were expressed as absolute numbers and percentage. The continuous variables were expressed as mean and standard deviation. To produce robust estimates independent from the distribution pattern of the variables, some tests were specifically adopted. sThe comparisons between continuous variables and two groups were performed using the unpaired student t-test with adjustment for heterogeneity of variance and degrees of freedom using the Satterthwaite method. Comparisons between continuous variables and more than three groups were estimated using the Kruskal-Wallis test. Several logistic regression models for AH were used, starting from the choice of predictors with p < 0.20 in unadjusted analyses. The model's potential increment was assessed after inclusion of squared terms and interaction of predictors. The comparison of the increased prevalence between the quilombola communities and the population in general was performed using the chi-squared adjustment test. To adjust the analysis for the differences between groups and the potential of heteroskedasticity in the quilombola communities, the Huber‑White method was used to estimate clustering, robust standard errors, according with the 15 communities. 384