Study Setting

The ELSA-Brasil study design and cohort profile have been published elsewhere. Briefly, it is a multicenter cohort study of 15,105 civil servants from six cities in Brazil (São Paulo, Belo Horizonte, Porto Alegre, Salvador, Rio de Janeiro, and Vitória). All active or retired employees of the six institutions aged 35 to 74 years were eligible for the study. Exclusion criteria were current or recent (<4 months before the first interview) pregnancy, intention to quit working at the institution in the near future, severe cognitive or communicative impairment, and, if retired, residence outside of a study center’s corresponding metropolitan area. The baseline assessment was a 7-hour onsite examination that included validated questionnaires and clinical and laboratory examinations. This assessment took place from August 2008 to December 2010. Approvals were granted from institutional review boards at all centers, and all individuals signed informed consent agreements.

Study Sample

In the ELSA-Brasil baseline, 13,205 of the study participants (87.4%) underwent CIMT evaluation for both common carotid arteries. After image quality control, we had valid CIMT measurements in both sides for 10,943 of 15,105 participants (72.4%). The study sample definition is displayed in Figure 1 , and a comparison between included and excluded participants is shown in Supplemental Table 1 .

Study sample definition.

Study Variables

In the ELSA-Brasil baseline assessment, race was self-defined according to the classification of the Brazilian national census. The protocol for CIMT measurement has been detailed in previous publications. In brief, a standardized protocol was applied at all centers using a clinical ultrasound scanner (Aplio XG; Toshiba, Tokyo, Japan) with a 7.5-MHz linear transducer. Common carotid artery intima-media thickness measurements were electrocardiographically gated (during three cardiac cycles) and performed in the far wall of a predefined carotid segment 1 cm in length measured from 1 cm below the carotid bifurcation. All participating centers obtained the carotid images and sent these acquisitions to the centralized core reading center in São Paulo. After image quality control, readings were electrocardiographically gated, and the measurements were automated using MIA software, standardizing the interpretation of the carotid scans.

For the main analyses in this report, we defined CIMT as the average of the mean left CIMT and the mean right CIMT values within an individual. Sensitivity analyses were also performed using maximal CIMT (from both common carotid arteries). We analyzed CIMT as a continuous variable and according to the distribution of values within the sample. In a previous study, we used quantile regression model techniques to describe age-, sex-, and race-specific CIMT quartiles for the sample. In this analysis, we also categorized participants according to these modeled quartiles. This procedure was also used in another publication and established the current standard for the Brazilian population. Of particular interest, the cutoff at the 75th percentile of CIMT for a given age, sex, and race (i.e., the fourth quartile in quantile regression models) is considered to be a marker for increased cardiovascular risk. To better contextualize the results of our analysis, we describe in Supplemental Table 2 the 25th, 50th, and 75th percentiles for CIMT values in ELSA-Brasil participants at ages 40, 50, and 60 years.

The definition and scoring for CVH in this analysis was based on the definitions used by the AHA. We evaluated the seven components of the CVH score (diet, physical activity, smoking, body mass index, blood pressure, fasting plasma glucose, and total cholesterol) for each participant and attributed scores of 0, 1, and 2 points, corresponding to poor, intermediate, and ideal profiles ( Table 1 ). The CVH score was calculated as the sum of the individual components (range, 0–14 points). For some analyses, the CVH score was classified as “inadequate” (0–4 points), “average” (5–9 points), or “optimal” (10–14 points). This classification has been used by others.

Information from a food frequency questionnaire was used to calculate all five CVH score components related to diet. We considered the following intake of food components to be adequate: four or more servings of fruit and vegetables per day, ≥7 oz of fish per week, two or more servings of fiber-rich whole grains per day, and ≤450 kcal of sugar-sweetened beverages per week. Sodium consumption was corrected for total energy intake, and sodium consumption of <1,500 mg/d was considered adequate. We evaluated physical activity using the leisure and transportation domains of the International Physical Activity Questionnaire. We considered the AHA standard for physical activity as ideal, consisting of a minimum of 75 min/wk of vigorous activity, 150 min/wk of moderate activity, or 150 min/wk of moderate plus vigorous activity. Time of moderate plus vigorous physical activity between 1 and 149 min/week was considered intermediate, and absence of moderate and/or vigorous physical activity was considered poor. Smoking was assessed by self-report. Past smokers also stated the age at which they stopped smoking. Individuals with ideal smoking profiles were those who had never smoked or whose age at quitting smoking was ≥2 years less than their current age. Individuals with intermediate smoking profiles were past smokers whose age at quitting smoking was <2 years less than their current age.

Criteria for body mass index, blood pressure, fasting plasma glucose, and total cholesterol were the same as those proposed by AHA ( Table 1 ). The anthropometric measurements in ELSA-Brasil were assessed using standard techniques, and body mass index was calculated by dividing weight by the square of height. As in previous publications, the use of medications was classified according to the Anatomical Therapeutic Chemical Classification System codes. Use of antihypertensive medications was defined as the use of medications under codes C02, C03, C07, C08, and C09. Use of hypoglycemic medications was defined as the use of medications under code A10. Use of lipid-lowering medications was defined as the use of medications under code C10.

Statistical Analysis

Categorical variables are presented as proportions and compared using χ ² or Fisher exact tests. Continuous variables are presented as mean ± SD or as median (interquartile range) and were compared using one-way analysis of variance or Kruskal-Wallis tests, as applicable. We analyzed the distribution of CIMT values (categorized in age-, sex-, and race-specific quartiles) according to CVH score as a discrete variable or categorized into strata (inadequate, average, or optimal), using χ ² trend tests for proportions. In addition, we describe the Pearson correlation for the association between CIMT value and CVH score.

We built linear regression models to evaluate the association between CIMT (in millimeters) and CVH score. We present the crude models as well as models adjusted for age, sex, and race. We tested the main models (using both mean and maximal CIMT) for heteroscedasticity of the residuals using the Breusch-Pagan test. Because the tests rejected the null hypothesis, we calculated standard errors and P values using White-corrected covariance matrices. In addition, we performed a systematic search for potential influential points in the main models using the following four measurements: DFFITS, Cook’s distance, leverage values, and covariance ratio. Excluding the observations identified as potential influential points in any of these procedures did not alter significantly the results. Finally, we calculated the variance inflation factors for the independent variables in the main models (age, sex, race, and CVH score). Variance inflation factors were all <1.1, showing the absence of significant multicollinearity. Because CIMT values differ according to sex and race, and β coefficients in linear models may be sensitive to this distribution, we also present models stratified by sex and stratified by race. In addition, we describe the odds ratios from multinomial regression models for the association of CIMT age-, sex-, and race-specific modeled quartiles with CVH score.

As a sensitivity analysis, to reduce the possibility of reverse causation mediated by the adoption of positive lifestyle habits due to nonpharmacologic treatment of hypertension, diabetes or dyslipidemia, we also repeated the analyses while restricting the sample to the 4,505 individuals without self-reported diagnoses of hypertension, diabetes, and/or dyslipidemia in the ELSA-Brasil baseline assessment. The significance level was set at .05. We used R software version 3.2.0 to conduct the analyses.