Points are sample averages by sex, race/ethnicity, and age. Lines are cubic spline fits of the points by sex and race/ethnicity. P values correspond to the null hypothesis of no racial/ethnic difference in CAC prevalence controlling for age by sex. Age-matched subgroup, age ≥40 years; examination year ≥1998; n = 21 939. AU indicates Agatston units.
eFigure 1. Consort diagram of study population for baseline, longitudinal, coronary artery calcium, and mortality analyses
eFigure 2. Kaplan-Meier all-cause mortality curves for Hispanic and NHW women and men
Customize your JAMA Network experience by selecting one or more topics from the list below.
Rodriguez F, Leonard D, DeFina L, et al. Association of Educational Attainment and Cardiovascular Risk in Hispanic Individuals: Findings From the Cooper Center Longitudinal Study. JAMA Cardiol. 2019;4(1):43–50. doi:10.1001/jamacardio.2018.4294
Is high educational attainment associated with cardiovascular risk factors, subclinical atherosclerosis, and mortality in Hispanic individuals compared with non-Hispanic white (NHW) individuals?
In this longitudinal cohort study that included 1351 Hispanic and 43 736 NHW participants with high educational attainment, Hispanic individuals had greater prevalence of cardiometabolic risk factors but no difference in predicted cardiovascular risk. No significant ethnic differences in subclinical atherosclerosis or mortality were found.
In a highly educated Hispanic population, there was no evidence for the Hispanic paradox in cardiovascular risk, subclinical coronary atherosclerosis, or mortality.
Hispanic individuals are the fastest growing ethnic group in the United States and face lower socioeconomic status compared with non-Hispanic white (NHW) individuals. However, Hispanic individuals tend to experience better health outcomes than expected, a phenomenon known as the Hispanic paradox. Little is known about how higher socioeconomic status is associated with Hispanic cardiovascular risk factor burden and outcomes.
To determine cardiovascular risk and outcomes among highly educated Hispanic vs NHW individuals in a preventive medicine clinic.
Design, Settings, and Participants
Retrospective cohort analysis of participants from the Cooper Center Longitudinal Study who underwent preventive medical examinations at the Cooper Clinic in Dallas, Texas, from October 1972 to November 2017. Analysis began April 2018.
Ethnicity, self-defined as Hispanic or NHW.
Main Outcomes and Measures
Prevalence of major metabolic risk factors and cardiorespiratory fitness were compared, as were changes among participants with at least 2 visits. Ethnic differences adjusted for age, examination year, and educational attainment were estimated using regression models. Age-matched comparisons of coronary artery calcium scores were performed. All-cause mortality was summarized using the Kaplan-Meier method.
This study included 1351 Hispanic and 43 736 NHW participants aged 20 to 80 years, body mass index between 18.5 and 50.0, and were not missing key cardiometabolic or fitness variables. Both Hispanic and NHW participants had high educational attainment, with a mean of more than 15 years of total education. Hispanic women and men had a higher prevalence of metabolic syndrome (71 of 518 [13.1%] vs 1477 of 13 732 [10.8%] for women and 255 of 833 [30.6%] vs 7902 of 30 004 [26.3%] for men, respectively). Although Hispanic individuals were twice as likely to have diabetes, there was no difference in calculated 10-year atherosclerotic cardiovascular disease risk scores by ethnicity. Both Hispanic and NWH individuals experienced a statistically significant worsening in cardiometabolic parameters during follow-up, although this was not statistically significantly different between groups. In age-matched analyses, there were no significant differences in the prevalence of coronary artery calcium scores between Hispanic and NWH individuals. During a mean (SD) follow-up of 12.9 (7.5) years, there was no difference in mortality between Hispanic and NHW individuals.
Conclusions and Relevance
Hispanic and NHW men and women with high educational attainment had similar atherosclerotic cardiovascular disease risk, subclinical coronary atherosclerosis, and mortality during follow-up. These findings do not support the Hispanic paradox in a highly educated Hispanic population.
Hispanic individuals are the largest minority group in the United States and experience a disproportionate burden of atherosclerotic cardiovascular disease (ASCVD) risk factors.1,2 Although cardiovascular disease is the leading cause of death for Hispanic individuals, most current studies addressing racial/ethnic disparities in ASCVD have focused on differences between black individuals and non-Hispanic white (NHW) individuals, and the ASCVD burden in Hispanic individuals is not yet fully understood.3-6 Some studies have shown that Hispanic individuals experience lower rates of ASCVD and mortality compared with NHW individuals. This health advantage, despite greater adverse risk profiles and lower levels of income and education than NHW individuals, has been termed the Hispanic paradox.7-9
Several explanations have been posited for the observed Hispanic health advantage. One hypothesis is that Hispanic cultural orientation results in healthier behaviors that lead to better outcomes, which may counter the deleterious effects of lower socioeconomic status (SES). Studies have documented that Hispanic individuals with lower levels of acculturation, or adoption of cultural norms in the United States, experience better ASCVD outcomes, although not consistently.10-13 Other studies highlight that the higher burden of metabolic risk factors among Hispanic individuals can be explained by their lower SES because they have lower health literacy and regular access to preventive care.14 Still, when studied in aggregate, US Hispanic individuals paradoxically experience lower rates of cardiovascular disease mortality than NHW individuals.2,15,16
Importantly, Hispanic individuals are a heterogeneous group and most research has been limited to lower SES populations studied cross-sectionally.1 Many contend that race/ethnicity may simply be a proxy for SES. Our study aims to address this gap in knowledge by exploring the longitudinal burden of cardiovascular risk among a well-educated Hispanic population that sought preventive care at the Cooper Clinic in Dallas, Texas, for up to 4 decades. Specifically, we sought to compare the prevalence of ASCVD risk factors, cardiorespiratory fitness, subclinical coronary atherosclerosis, and mortality between Hispanic and NHW individuals.
The Cooper Center Longitudinal Study (CCLS) is a large prospective study of men and women who underwent a preventive medical examination at the Cooper Clinic. The details of this cohort assembly are described elsewhere.17 For the current study, we included participants who self-identified as Hispanic or NHW at their first visit to the Cooper Clinic and excluded other races/ethnicities. The base sample for cross-sectional analysis included 1351 Hispanic and 43 736 NHW individuals between ages 20 and 80 years, body mass index (BMI; calculated as weight in kilograms divided by height in meters squared) between 18.5 and 50.0, and were not missing key variables including blood pressure, triglycerides, high-density lipoprotein cholesterol (HDL-C), or glucose. These same participants contributed to the longitudinal analysis if they had an additional visit within 20 years with nonmissing values for these same variables (eFigure 1 in the Supplement). Individuals were excluded from the coronary artery calcium (CAC) analysis if their examination occurred prior to 1998, they were younger than 40 years, or they were not age-matched by ethnicity.
Patients consented and approved the use of their deidentified data for research. The testing protocols for collecting data entered into the CCLS database are approved annually by the Cooper Institute institutional review board. The study was approved by the Cooper Institute Scientific Review Committee.
All participants completed a medical questionnaire that included self-reported personal and family history of disease. Socioeconomic status was defined by educational attainment in years. A physician performed a physical examination, and anthropometric measures were obtained by trained clinical personnel using standardized procedures. Body mass index was calculated as weight in kilograms per meter squared. Waist circumference was measured in centimeters on a horizontal plane at the level of the umbilicus. Blood pressure levels were measured using the auscultatory technique with a mercury sphygmomanometer. Blood chemistries were analyzed using automated methods.
Glucose, total cholesterol, low-density lipoprotein cholesterol, HDL-C, and triglycerides were obtained after a 12-hour fast. Diabetes was self-reported. Metabolic syndrome was defined using the National Cholesterol Education Program Adult Treatment Panel III guidelines (ie, 3 or more of the following: waist circumference, ≥102 cm [≥88 cm for women]; triglycerides, ≥150 mg/dL [to convert to mmol/L, multiply by 0.0113]; HDL-C, <40 mg/dL [<50 mg/dL for women; to convert to mmol/L, multiply by 0.0259]; systolic blood pressure, ≥130 mm Hg or diastolic blood pressure, ≥85 mm Hg; or a fasting glucose ≥100 mg/dL [to convert to mmol/L, multiply by 0.0555]).18 If waist circumference was missing (n = 4161 participants), we used a BMI cutoff of more than 30 as an alternative per prior guidelines.19
Fitness (aerobic capacity) was estimated in metabolic equivalent tasks (1 metabolic equivalent task = 3.5 mL/kg per minute of oxygen consumption) based on peak treadmill speed and grade. Cardiorespiratory fitness (fitness) was assessed by a maximal treadmill exercise test using a modified Balke protocol.17 Metabolic equivalent task hours per week was estimated using the physical activity questionnaire. Participants provided frequency and duration of leisure time physical activities (ie, treadmill, running, walking, elliptical trainer, outdoor cycling, stationary cycling, swimming laps, vigorous sports, aerobic dance, or other), which was used to calculate hours of activity per week. Hours per week were then weighted by multiplying the value by an estimated metabolic equivalent for each activity derived from the Compendium of Physical Activities and summed across all activities.20
The 10-year risk of ASCVD was calculated for each participant using the pooled cohort equations algorithms with patient baseline demographics (age and sex), lipid panel values (total cholesterol and HDL-C), systolic blood pressure levels, smoking status, diabetes, and treatment for hypertension. Based on the American Heart Association/American College of Cardiology Work Group’s definition, 10-year ASCVD risk was defined as the probability of experiencing a first ASCVD event (nonfatal myocardial infarction or death from coronary heart disease or fatal/nonfatal stroke) during a 10-year period among people without ASCVD at the beginning of the period.21 We also calculated the Framingham general cardiovascular disease risk score using participant sex, age, total cholesterol, HDL-C, treatment for hypertension, smoking status, and diabetes.22
Between January 1998 and November 2017, a subset of this study sample underwent CAC testing. From January 1998 to December 2007, CAC was calculated by the Imatron C-150XP or C-300 electron beam tomography scanner (GE Imatron). From January 2008 to November 2017, a 64-slice scanner (Lightspeed VCT, GE Healthcare) was used. Three-mm thick slices were obtained with 2-mm table increments during a breath-holding protocol. Coronary artery calcification scores were reported in Agatston units, as previously described in the study cohort.23
Participants were excluded if they did not have at least 1 year of follow-up. All participants were followed from the date of their examination until their date of death or December 31, 2014. The National Death Index Plus service was the primary source of vital status and cause of death for this study sample.
Characteristics of participants at their first visit to the Cooper Clinic were summarized by sex and race/ethnicity; racial/ethnic differences by sex were tested using Fisher exact test for nominal characteristics and rank sum statistics for continuous characteristics. Racial/ethnic comparisons of characteristics adjusted for age, educational attainment, and year of examination by sex were based on multiple linear or logistic regression models. Missing educational attainment in 954 participants (2.1%) was multiply imputed by sex and outcome using 25 random draws from the corresponding complete case regression model. Annual rates of change by sex were estimated using mixed-effects regression models of within-participant changes in risk factors on time between visits, with a time × race/ethnicity term to test for race/ethnic differences in rates, random time effects within participants to account for correlation of repeated measurements, and compound symmetric correlation of residuals within participants to account for differences from baseline. Racial/ethnic comparisons of CAC score by sex are sensitive to age. To maintain tight control of the age effect without relying on a specific functional specification (eg, exponential growth), CAC score and prevalence were analyzed in the subgroup formed by matching all Hispanic women and men with NHW women and men of the same (integer) age. Unadjusted comparisons of CAC score relied on age-stratified Wilcoxon (van Elteren test) tests by sex, and unadjusted comparisons of prevalence of CAC more than 0, 10, or 100 Agatston units relied on age-stratified Cochran-Mantel-Haenszel χ2 tests by sex. Comparisons of CAC prevalence adjusted for educational attainment, current smoking, physical activity, fitness, BMI, systolic blood pressure, glucose, triglycerides, total and HDL-C relied on age-matched sets (conditional) logistic regression by sex. Racial/ethnic comparisons of mortality by sex among participants with at least 1 year of follow-up for mortality were based on the Kaplan-Meier method with log-rank test. All analyses were programmed in SAS/STAT, version 9.4 (SAS Institute Inc), and statistical tests were based on 2-sided tests with a significance level of .05.
Our study consisted of 45 087 participants who received preventive medicine examinations at the Cooper Clinic from October 1972 to November 2017. There were a total of 1351 Hispanic individuals (3%) in the study sample. Table 1 shows the demographic and clinical characteristics of the participants. The sample population was highly educated, with no statistical difference in educational attainment level between Hispanic and NHW individuals (mean [SD] educational attainment, 15.4 [2.4] vs 15.6 [2.6] years; P = .08 for women and 16.3 [2.5] vs 16.4 [2.3] years for men, P = .507). Hispanic women and men were more likely to be employed than their NHW counterparts. Hispanic men were less likely to be smokers than NHW men (93 of 833 [11.2%] vs 4380 of 30 004 [14.6%]; P = .005), whereas Hispanic women tended to smoke more than NHW women (38 of 518 [7.3%] vs 782 of 13 732 [5.7%]; P = .12). Hispanic individuals had slightly lower baseline fitness levels compared with NHW individuals.
After adjusting for age, examination year, and educational attainment, Hispanic individuals were more likely to have higher BMIs, waist circumference, glucose levels, and triglyceride levels (Table 1). Hispanic women were more likely to have metabolic syndrome (odds ratio, 1.51; 95% CI, 1.16-1.96] and had twice the odds of having diabetes [odds ratio, 2.07; 95% CI, 1.25-3.43] compared with NHW individuals. Similar patterns were observed for Hispanic men.
Hispanic women and men had lower estimated 10-year risk of ASCVD compared with NHW individuals, although this difference disappeared after adjusting for examination year, age, and educational attainment. Both the ASCVD 10-year risk score and the Framingham general cardiovascular disease risk score increased for Hispanic and NHW individuals during subsequent visits.
During a mean (SD) follow-up of 3.6 (3.3) years, 176 Hispanic women (34%) and 4913 NHW women (36%) had at least 1 follow-up visit, and during a mean (SD) follow-up of 3.2 (3.3) years, 357 Hispanic men (43%) and 14 393 NHW men (48%) had at least 1 follow-up visit. In participants with at least 2 visits, cardiovascular risk factors worsened each year for women (Table 2) and men (Table 3). Hispanic women experienced a nonsignificant increase in fitness (annalized rate of change of 0.01 metabolic equivalent tasks; P = .77), but NHW women experienced a significant decline (annalized rate of change of 0.03 metabolic equivalent tasks; P < .001).
Of 1351 Hispanic individuals, 211 women (15.6%) and 444 men (32.9%) were 40 years or older and underwent routine CAC testing. These were matched by sex and age to 6432 women (14.7%) and 14 852 men (34.0%) of 43 736 NHW individuals (Table 4 and Figure). The mean (SD) age of the resulting age-matched sample was 50.2 (7.2) years for women and 49.1 (7.2) years for men. Hispanic women almost had a significantly greater prevalence of CAC than NHW women (20.9% vs 16.2%; χ2 = 3.74; P = .053), but CAC prevalence was more similar in Hispanic and NHW men (46.4% vs 44.4%; χ2 = 0.76; P = .38).
Of 655 Hispanic individuals who were 40 years or older and underwent routine CAC testing, 11 women (5.2%) and 61 men (13.7%) had at least 2 visits with repeat CAC testing. Similarly, 383 NHW women (6.0%) and 3323 NHW men (22.4%) had at least 2 visits with repeat CAC testing. Hispanic men with baseline CAC score higher than 0 experienced greater annual percentage changes in CAC than NHW men (27.4% vs 21.3%; difference, 6.1%; 95% CI, 0.2-12.0; P = .04). There was no significant difference in the annualized rate of change of CAC among women (17.3% vs 21.0%; difference, −3.8%; 95% CI, −17.3 to 9.8; P = .59).
A total of 40 532 participants were included in the mortality analyses. There were no statistically significant ethnic differences in all-cause mortality during a mean follow-up of 12.9 years (eFigure 2 in the Supplement), including 272 deaths (2.1%) among women (4 Hispanic) and 1045 deaths (3.8%) among men (15 Hispanic). The mortality incidence per 1000 person-years was 0.9 for Hispanic and 1.9 for NHW women (log-rank χ2 = 1.42; P = .23). Hispanic men had a mortality incidence of 1.9 per 1000 person-years and 2.8 for NHW men (log-rank χ2 = 0.51; P = .48).
Over 4 decades, we document persistent health disparities in cardiovascular risk factors for well-educated Hispanic individuals compared with NHW individuals participating in the CCLS. Hispanic individuals had a higher prevalence of metabolic syndrome, diabetes, and lower fitness than NHW individuals. There was no evidence of a Hispanic paradox, with no significant ethnic differences in estimated ASCVD risk, subclinical coronary atherosclerosis (as measured by CAC score), and all-cause mortality.
Our study challenges the well-described epidemiologic association between high SES and improved health outcomes.24-27 Among the frequently used metrics for SES, education ranks among the strongest and most consistent in its association with health risk factors.24 We found a high prevalence of adverse cardiovascular risk profiles (namely obesity, metabolic syndrome, and diabetes) in a highly educated and largely employed cohort of Hispanic individuals. The prevalence of these risk factors was higher among Hispanic individuals compared with NHW individuals of similar SES. After adjusting for educational attainment, there was no difference in predicted ASCVD risk between Hispanic and NHW individuals. In support of these findings, prior work has shown an inverse association between educational attainment and diabetes risk among Hispanic individuals.28 A possible explanation for our findings is that study participants were highly acculturated to the US culture. Increased acculturation has been linked to higher rates of obesity, hypertension, diabetes, and risk of cardiovascular disease.10,11
We measured subclinical coronary atherosclerosis with repeated CAC measurements. Contrary to the Hispanic paradox, we found that Hispanic women tended to have a higher prevalence of CAC at any age and that the rate of CAC increase was higher for Hispanic men compared with NHW men. Prior work has shown that Hispanic individuals have lower CAC scores than NHW individuals, and that CAC correlates well with incident coronary heart disease events across racial and ethnic groups.29 This finding fails to support in populations with higher SES existing notions about the Hispanic paradox. We were unable to detect statistically significant differences in long-term mortality, although this was limited by the low event rate. However, we found no differences between ASCVD 10-year risk scores between Hispanic and NHW individuals, suggesting that even the higher burden of metabolic risk factors among Hispanic individuals does not necessarily translate to worse ASCVD risk or outcomes.
Our study has several strengths. We studied a cohort of highly educated Hispanic individuals who have not traditionally been included in prior studies exploring the Hispanic paradox in ASCVD. We were able to compare Hispanic participants with NHW participants of similar SES. Second, our cohort is well characterized with thoroughly measured data on burden of cardiovascular risk factors, fitness, subclinical atherosclerosis (CAC), and mortality. Finally, we were able to follow the participants longitudinally, a major limitation of prior studies that have focused on cross-sectional associations between Hispanic individuals and ASCVD outcomes.
Our study should also be interpreted in light of several limitations. The CCLS population was relatively homogenous with individuals who were largely self-referred or part of executive wellness programs. This may limit the generalizability of the results to other populations but may mitigate residual confounding in our current analyses. The number of Hispanic individuals in the CCLS registry is small compared with NHW individuals. We were unable to characterize important differences among Hispanic participants including English language proficiency and country of origin. This is a relatively healthy population; hence, the mortality rate was low. We were not able to detect differences in ASCVD prevalence and CAC score may be an imperfect proxy for future ASCVD risk in minority populations.30 However, CAC remains a strong predictor of future ASCVD events among diverse racial/ethnic groups.29 Few Hispanic individuals were included in early years of the CCLS, limiting our ability to detect long-term differences in ASCVD event rates between ethnic groups should they exist. Finally, we relied solely on educational attainment as our indicator of SES, although prior work has shown that education is among the most important predictor of health outcomes. This study complements the limited literature on Hispanic cardiovascular health by studying Hispanic individuals of higher SES followed longitudinally.
In a longitudinal cohort of individuals with higher SES undergoing preventive health examinations, Hispanic and NHW men and women with high educational attainment had similar ASCVD risk, subclinical coronary atherosclerosis, and mortality during follow-up. These findings do not support the Hispanic paradox in a highly educated Hispanic population. More work is needed to understand the relationship between SES, cardiovascular risk, and outcomes in Hispanic individuals.
Corresponding Author: Fatima Rodriguez, MD, MPH, Division of Cardiovascular Medicine, Stanford University, 870 Quarry Rd, Falk CVRC, CV287, Stanford, CA 94305-5406 (firstname.lastname@example.org).
Accepted for Publication: November 2, 2018.
Published Online: December 19, 2018. doi:10.1001/jamacardio.2018.4294
Author Contributions: Drs Rodriguez and Maron had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Rodriguez, DeFina, Barlow, Haskell.
Acquisition, analysis, or interpretation of data: Rodriguez, Leonard, DeFina, Barlow, Willis, Maron.
Drafting of the manuscript: Rodriguez, DeFina, Barlow.
Critical revision of the manuscript for important intellectual content: Rodriguez, DeFina, Willis.
Statistical analysis: Leonard
Administrative, technical, or material support: Rodriguez.
Conflict of Interest Disclosures: None reported.
Additional Contributions: We thank Kenneth H. Cooper, MD, MPH, for establishing the Cooper Center Longitudinal Study, the Cooper Clinic staff for collecting the clinical data, and the Cooper Institute for maintaining the database.