Cardiovascular diseases (CVD) are currently the leading cause of death in Brazil and worldwide, justifying the implementation of prevention measures against this public health problem. In this sense, cardiovascular risk (CVR) monitoring allows the establishment of priorities among activities that focus on the CVD problem, assisting in the definition of practices and prognostic analyses. Several algorithms and scores are used to assess CVR, but none have been validated for the Brazilian population, even less so for the country’s black population.^1,2 Therefore, there is no consensus on which procedure would be the most appropriate for this purpose.

The Brazilian Ministry of Health has recommended the Framingham Risk Score (FRS), which is intended to estimate CVR for the most severe forms of coronary artery disease (CAD). The Brazilian Society of Cardiology has recommended the Global Risk Score (GRS) to verify the 10-year risk of developing any atherosclerotic disease events.²

Both the FRS and GRS calculations require total cholesterol (TC) and high-density lipoprotein (HDL) data from patients, which are not always available in low-infrastructure settings. However, a simplified version of the GRS replaces these biochemical data with the body mass index (BMI). The simplified GRS (sGRS), by using measures that are easier to obtain, could thus be a better alternative for use in basic health units.³

All these instruments (FRS, GRS, and sGRS) derive from the Framingham Heart Study and are based on the synergism between risk factors for CVD. Therefore, their use is intended to estimate the risk of developing cardiovascular events.^3,4

CVD affects individuals from lower-income populations more severely since they are more exposed to the different risk factors. In Brazil, the black population and particularly the quilombola population, due to institutional racism, remains historically at lower socioeconomic levels; this situation increases risks associated with CVD.⁵

Given the absence of studies on the accuracy of CVR indicators for the Brazilian population, especially for the black population, this study aimed to investigate the accuracy of the FRS and sGRS, using the GRS as reference. Additionally, we described the prevalence of risk factors for CVD in quilombola women from Alagoas.

This work is part of the project entitled II Diagnóstico de saúde e Segurança Alimentar e Nutricional das famílias das comunidades remanescentes dos quilombos do estado de Alagoas, approved by the Research Ethics Committee of Universidade Federal de Alagoas (CAAE No. 33527214.0.0000.5013).

This is a cross-sectional study; data collection was performed from April 2017 to January 2018. The prevalence of high CVR in quilombola communities was considered as the variable of interest to estimate sample size. Since no previous studies had been conducted in Alagoas, the 6% prevalence found in the state of Maranhão was used.⁶ In 2015, around 6465 families were living in quilombola communities in Alagoas. Assuming that each family had one female member, 6465 women were the universe of interest of this research.⁷ Using the StatCalc tool of Epi Info® 7.2.1.0 and admitting a sampling error of 1.5% for a 95% confidence interval (CI), 861 women would be required for this study. In order to compensate for possible sample losses, this number was increased by 10%, totaling 948 women.

In order to achieve this sample number, we randomly chose 34 out of the 68 existing quilombola communities. Women aged between 19 and 59 years from all households in the selected communities were eligible for the study. When there were 2 or more women in the same household, the participating woman was randomly defined. Exclusion criteria were being pregnant or in puerperium, having ingested alcoholic beverages in the last 24 hours, and presenting evident anatomical alterations.

Data collection was performed by interviewers who were properly trained for executing their activities, which occurred under constant supervision. The pilot study used to standardize procedures and test instruments took place before the fieldwork began and was conducted in a community that was not selected for the main study.

Data were collected through interviews using structured forms during home visits. Blood pressure and anthropometric measurements were also obtained. Subsequently, the women were referred to a previously defined place in the community for biochemical tests.

The following variables were obtained for the socioeconomic and demographic characterization of our sample: self-reported skin color (black/brown; other); age group (19 to 39.9; 40 to 49.9; ≥ 50 years); education (< 4; ≥ 4 years of study); marital status (single; stable relationship; widowed/separated); occupation (not working; farming; other); social class⁸ (B+C; D+E; there were no families in class A); participation in the Bolsa Família Program (yes; no); and classification according to the Brazilian Food Insecurity Scale (EBIA), which classifies families as being in situations of mild, moderate, or severe food insecurity.⁹

Regarding health and lifestyle variables, the following conditions were defined: optimal blood pressure (BP), normal BP, prehypertension, and hypertension stages I, II, and III; normoglycemia, pre-diabetes mellitus, and diabetes mellitus (DM); smoking; low body weight, eutrophic, overweight, and obesity grades I, II, and III.

Optimal BP was defined by systolic blood pressure (SBP) < 120 mmHg and diastolic blood pressure (DBP) < 80 mmHg; normal BP was considered as: SBP 120–129 mmHg and DBP 80–84 mmHg; prehypertension: SBP 130–139 mmHg and/or DBP 85–89 mmHg; systemic arterial hypertension (SAH) : SBP ≥ 140 mmHg and/or DBP ≥ 90 mmHg or when the use of antihypertensive medications was reported; stage I SAH: SBP 140–159 mmHg and/or DBP 90–99 mmHg; stage II SAH: SBP 160–179 mmHg and DBP 100–109 mmHg; stage III SAH: SBP ≥ 180 mmHg and DBP ≥ 110 mmHg.¹ SBP and DBP measurements were performed in duplicates, according to the Brazilian Guidelines of Hypertension.¹ An Omron® Hem-7200 portable equipment was used, being periodically calibrated according to the manufacturer’s recommendations. If differences greater than 20 mmHg were observed between the 2 measurements, a third measurement was performed and the most discrepant result was disregarded. The mean of the 2 valid measurements was used in the analysis.

Prediabetes and DM were defined by glycated hemoglobin (HbA1c) levels between 5.7% and 6.4% and ≥ 6.5%, in that order.¹⁰ Hypercholesterolemia, hypertriglyceridemia, and low high-density lipoprotein cholesterol (HDL-c) were determined by the following conditions (mg/dL): low-density lipoprotein (LDL) ≥ 160, triglycerides ≥ 175, and HDL < 50, respectively.¹¹ Dyslipidemia was designated when at least one of these lipid profile alterations was found.

In addition to the biochemical reference values, the use of hypoglycemic drugs or medications to control lipid alterations was also a criterion for defining these conditions.

TC, HDL, and HbA1c measurements were performed using a drop of blood obtained by puncturing the digital pulp, regardless of fasting.¹² These determinations were performed using an Alere Cholestech LDX® System and, in the case of HbA1C, an Alere NycoCard Reader II®, with their respective analysis cassettes.

Smoking was identified as the consumption of tobacco products in the last 3 months and assessed through the Alcohol, Smoking, and Substance Involvement Screening Test (ASSIST), validated for the Brazilian population.¹³

Nutritional status classifications were based on the BMI according to the following categories¹⁴: underweight (< 18.5 kg/m²); eutrophic (18.5–24.9 kg/m²); overweight (25–29.9 kg/m²); obesity grade I (30–34.9 kg/m²); obesity grade II (35–39.9 kg/m²); and obesity grade III (≥ 40 kg/m²). The weight and height of the participant were measured for calculating the BMI. Body mass was obtained using a digital scale (model 813, Seca®) with a 200 kg capacity. A portable stadiometer (model 213, Seca®) was used to measure the participants’ height.

The GRS was used as reference for evaluating the accuracy of FRS and sGRS in predicting CVR. The GRS is determined by attributing points to age, waist circumference, HDL, systolic blood pressure, smoking, and DM variables, as shown in the table presented in the original publication.³ The CVR was classified as follows: ≤ 8 points (low risk); 9 to 12 points (intermediate risk); and ≥ 13 points (high risk).²

The FRS was proposed in 2001 aiming to identify individuals at high risk for atherogenesis and those who were more likely to develop severe forms of CAD when the atherosclerosis process was already established.^15,16 Variables constituting the FRS are age, TC, HDL, smoking, SBP, and use of antihypertensive medication; our scores were established as presented in the original publication.⁴ The FRS was classified as follows: ≤ 19 points (low risk); 20 to 22 points (intermediate risk); and ≥ 23 points (high risk).¹⁷ This classification was proposed by the Brazilian Ministry of Health and favors specificity over sensitivity. In order to avoid this problem, in this study the sum of "intermediate risk" and "high risk" conditions was regarded as high CVR.

As mentioned previously, the sGRS is a simplification in which the procedures of the original GRS were maintained, but TC and HDL data were replaced by the BMI.³

Our sample consisted of 1015 women. Most of them had black or brown skin, 4 or more years of study, a stable marital relationship, belonged to the lowest economic class, were in a situation of food insecurity, and were contemplated by the Bolsa Família Program. Other sociodemographic characteristics are presented in Table 1.

Table 1 also shows the prevalence of cardiovascular risk factors. A total of 22.6% women had hypertension, 25.1% had DM, and 66.8% were overweight (30.3% were obese). The most prevalent lipid alteration was low HDL, followed by hypertriglyceridemia and hypercholesterolemia.

The prevalence of high CVR according to the GRS was 20.1%, which was similar to what was found with the sGRS (20.7%) and much higher than the number observed when using the FRS (4.5%), even when this latter classification grouped “intermediate risk” cases together with the "high risk" category.

We observed a strong association between high CVR and risk factor categories (Table 2). The prevalence of this outcome showed increments ranging from 7.5 to 13.1 times in women presenting SAH stages I to III, respectively. Women with DM had a CVR almost 6 times higher than that of women without this metabolic alteration. Women with stage III obesity had an approximately 3 times higher CVR when compared with eutrophic women. Similar associations also occurred for hypercholesterolemia, hypertriglyceridemia, low HDL, and smoking. Women aged 40 to 49.9 years and those aged 50 or older had frequencies of the outcome of interest that were around 10 and 30 times higher, respectively, than that found in those under 40 years old.

As shown in Figure 1, scales resulting from the FRS and sGRS showed excellent performance in predicting cardiovascular risk (AUC > 0.9; p < 0.001). However, the sGRS showed a higher discriminatory power than the FRS.

Table 3 shows the best CPs found in this study, with a higher accuracy than that obtained with the originally proposed CPs, particularly for the FRS. The CP of 20, which defines intermediate- and high-risk conditions, presented a sensitivity of only 18.6%, with a very high specificity (99.0%). Using the CP of 9, established by the highest J index, sensitivity and specificity reached 98.0% and 69.0%, respectively. The prevalence of high CVR, with this change in CP, increased from 4.5% to 44.4%. Regarding the sGRS, when the CP for high CVR (CP ≥ 13.0) was changed to the one with the best performance (CP ≥ 11.0), sensitivity increased from 87.2% to 97.6%, although with a decrease in specificity (from 96.0% to 88.5%). When considering the best CP, the prevalence of high CVR increased from 20.7% to 28.8%.

CVDs are currently the leading cause of mortality. Therefore, identifying individuals at higher risk is a critical task in public health to enable the adoption of preventive measures. It is important to have accurate and easy-to-operate tools to perform this identification. The results presented herein showed that the sGRS can be used instead of the GRS without any disadvantage. The latter is considered a more complete index because it includes biochemical data that indicate dyslipidemia, which are replaced by the BMI in the sGRS. The performance of the FRS compared to the GRS, although also satisfactory, was worse than that of the sGRS.

When the proposed CPs are used to discriminate patients at high CVR with the FRS, sensitivity is very low (despite a high specificity), which results in a high number of false negatives; this condition is not appropriate for screening CVR for elaborating preventive actions. In this perspective, the GRS should be used and, in the absence of lipid profile data, the sGRS.

Data routinely obtained in basic health units are used in the sGRS, thus this index can be widely applied even in contexts of poor infrastructure.³ Among the information required for the sGRS, only the definition of DM is ideally performed by laboratory tests, which could be an obstacle. However, most Brazilian municipalities have implemented the Blood Glucose Self-Monitoring Program, which provides users with glucose testing supplies.¹⁸ Moreover, the identification of diabetes can be done by self-reporting or when the patient reports the use of hypoglycemic agents.¹⁹

The women followed in this study belonged to quilombola communities, which comprise a social group that is specially subjected to social inequities and survive in a scenario marked by a low socioeconomic status, poor environmental conditions, and a high prevalence of food insecurity and morbidities related to this context. These characteristics are not exclusive to the quilombola women of Alagoas (who our sample plan attempts to represent), being extended to the entire Afrodescendant Brazilian population.^20,21

Some studies have shown a differential distribution of aggravations when comparing individuals from the quilombola population with those not from this population. Silva et al.²¹ investigated the prevalence of food insecurity in 21 rural communities located in the northeast region of Brazil, 9 of which were quilombolas. A prevalence of 64.9% was observed among quilombolas and a prevalence of 42% was found in the other communities, demonstrating the greater social vulnerability of this population.

Using representative samples of quilombola (n = 1631) and non-quilombola (n = 1098) women from the state of Alagoas, Ferreira et al.²² found that among quilombola women, overweight (50.1% vs 44.2%), body fat percentages > 33% (37.1% vs 23.3%), abdominal obesity (53.5% vs 34.3%), and hypertension (34.9% vs 11.4%) predominated in higher proportions. They concluded that quilombola women had a lower socioeconomic level and were more exposed to risk factors, which makes them more susceptible to morbidity and mortality from CVDs and justifies a prioritized attention by public policies.

Due to precarious living conditions, this population would have difficulties in adopting a healthy lifestyle. Regarding their diet, the consumption of ultra-processed foods, which are rich in sodium and poor in nutrients, may be one of the causes of a high prevalence of hypertension.²³ In our study, most women with high CVR had SAH.

Corroborating the results of previous studies, in this study we observed a high prevalence of high CVR in smokers and obese individuals, which are conditions that are potentially reversible by a timely approach in primary care.^17,24

In addition to the risk factors traditionally involved in CVD pathogenesis, it is well known that the racial-ethnic segment investigated by this study faces institutional and personal racism, which constitutes a traumatic event that causes low self-esteem and stress that in turn affect the cardiovascular and hemodynamic conditions and may increase exposure to unhealthy behaviors.^25–28 Incidentally, a previous study with this population showed a high prevalence (65.3%) of common mental disorders.²⁹ In the non-quilombola population of Alagoas, this prevalence was 47.9%.³⁰

In Brazil, the "Reorganization Care Plan for Hypertension and Diabetes Mellitus," implemented in 2001, targeted prevention and treatment at the primary health care level. Nevertheless, the identification of CVR was limited to the presence of risk factors and/or target organ damage.³¹ Subsequently, it was recommended that CVR assessment be carried out by means of multivariable models, adopting risk identification tools based on the Framingham study.^17,32

Among the CVR identification methods evaluated in this study, the FRS using the CP proposed by the Ministry of Health (CP ≥ 20) showed low sensitivity (18.6%) but 99.9% specificity. As mentioned previously, this is not a good performance for a screening instrument, since it results in many false negatives. Based on the ROC curve, we observed that a CP of 9.0 would result in the highest accuracy. Using this new cutoff resulted in sensitivity and specificity values closer to those observed with the sGRS. This index, in turn, would have better accuracy in identifying high CVR in the studied population if the original CP (CP ≥ 13.0) were to be changed for the one defined in this study (CP ≥ 11.0) because its sensitivity would increase from 87.2% to 97.6%, although with a reduction in specificity from 96.0% to 88.5%.

In addition, regarding the low sensitivity observed when using the original cutoff point for the FRS (CP ≥ 20.0), it can be speculated that this is because this protocol was based on a subgroup of atherosclerotic diseases, requiring adjustment to assess CVR in a global scale. This reinforces the relevance of using the sGRS in primary health care. In line with our findings, a study that compared the performance of the GRS and sGRS in predicting CVR in a sample consisting of Black and White Americans³³ concluded that both indexes found similar results. Therefore, the sGRS may be an alternative for initial screenings aiming at appropriate cardiovascular health care considering cost optimization.

Therefore, the efficient management of CVR depends on the access to health services, qualified professional care, and self-care.³⁴ The quilombola communities face political and operational difficulties, since the nearest health services usually have inadequate infrastructure, limited and fragmented care, and few human and material resources.^35,36 Moreover, the adoption of healthy lifestyle habits also depends on the professional approach, which must consider the peculiarities of these individuals, including their beliefs.³⁴