Association Between Living in Food Deserts and Cardiovascular Risk

We performed a cross-sectional analysis of data from subjects aged 20 to 79 years and residing in the Atlanta metropolitan area who were recruited into the META-Health study (Morehouse and Emory Team up to Eliminate Health Disparities; n=712) or the Predictive Health (PH) study (n=709).^24,25 The META-Health (MH) study recruited 753 community participants of whom 721 were eligible and enrolled. Only 712 subjects had valid addresses for determining area characteristics. The Predictive Health study (http://predictivehealth.emory.edu) recruited a cohort of university employees from Emory University and the Georgia Institute of Technology.^24,25 Seven hundred eleven subjects were enrolled in the study, and only 709 had valid addresses for neighborhood analysis. There was no significant difference between those who were included and those who were excluded from our analysis. Those with poorly controlled medical conditions, acute illness, recent hospitalization, or pregnant women were excluded. Detailed information on demographics and anthropometrics was collected. Hypertension, hypercholesterolemia, and diabetes mellitus were defined according to the Joint National Committee, Adult Treatment Panel III, and American Diabetes Association criteria, respectively.^26–28 The Atherosclerotic Cardiovascular Disease in Adults (ASCVD) scores were calculated to estimate the 10-year risk for coronary death or myocardial infarction.²⁹ Subjects signed an informed consent that was approved by the Emory and Georgia Tech institutional review board. Please see the Data Supplement for details.

Each participants’ zip code was entered into the USDA Food Desert Research Atlas that determined their FD status.⁸ A FD refers to a geographic location with both low income and low access to healthy foods. Low-income areas are defined according to criteria developed by the Department of Treasury’s New Markets Tax Credit program as any area where poverty rate is ≥20% or where the median family income is ≤80% of the state-wide median family income. Areas with low access to healthy foods are defined as areas where a significant share of people live ≥1 mile away in urban areas or ≥10 miles in rural areas from supermarket, supercenter, or large grocery store.⁸

Participants were instructed to fast for 12 hours before the study visit. Venous blood was collected, and levels of total cholesterol, low-density lipoprotein cholesterol, high-density lipoprotein cholesterol, and glucose were measured by spectrophotometry. Serum hs-CRP level was measured by immunonephelometry (Siemens/Dade Behring). Markers of OS included plasma levels of the amionthiols, glutathione, and cystine, which were measured using high-performance liquid chromatography as described previously.^19,30,31 Low levels of glutathione and high levels of cystine indicate higher OS. Of 1421 subjects enrolled in MH and PH studies, the following data were missing: hs-CRP (69 MH and 7 PH subjects), GSH (41 MH and 98 PH), and cystine (41 MH and 97 PH subjects) because of technical difficulties in sample drawing or processing. Pulse wave velocity (PWV; 295 MH and 24 PH subjects) and augmentation index (AI, 52 MH and 17 PH subjects) data were missing. Only available data were analyzed.

PWV and AI were estimated in the supine position after an overnight fast using the SphygmoCor device (AtCor Medical, Australia), which records sequential high-quality pressure waveforms at peripheral pulse sites using a high-fidelity tonometer, as described previously.^31–33 AI is a ratio calculated from blood pressure waveform as the percentage of central pulse pressure to the secondary systolic pressure rise caused by the overlap of the reflected pressure waves.³⁴ Details are available in the Data Supplement.

Continuous variables are presented as means±SD when normally distributed, skewed variables (cardiovascular risk scores and hs-CRP) as median (lower and upper quartiles), and categorical variables as proportions. Univariate group differences were compared using the χ² test for categorical variables and independent t tests for continuous variables. Nonparametric testing (Mann–Whitney) was used for comparing skewed variables, and these values were log transformed before analysis. Multiple linear regression was performed to determine the association of FD, area or personal income, and food access with OS measures and vascular function after adjusting for traditional CVD risk factors, including age, sex, race, smoking, hypertension, hyperlipidemia, diabetes mellitus, and body mass index. Interaction between food access, area, and personal income was performed. We intentionally adjusted for CVD risk factors because of baseline differences of risk factor burden between subjects living in different areas of FD. Individual income and education level were highly correlated in our sample. Therefore, we only adjusted for individual income to reduce multicollinearity. P values <0.05 were considered significant. Analyses were performed using SPSS, Inc (version 23).

Baseline demographic and clinical characteristics of the 1421 subjects are presented in Table 1. The sample was 38.5% male, 36.6% black, mean age 49.4±10.2 years. Education-level assessment revealed that 11% had a high school diploma or less, whereas 63.7% were college graduates. Annual household income in 11.2% was <$25 000, and 47.9% had income >$75 000 (Table 1).

The distribution of FD in the Atlanta metro area according to the USDA Food Access Research Atlas is shown in Figure (A). Individuals living in FD areas (n=187, 13.2%) were predominantly black (52% versus 34%) with less college education and lower income compared with those who were not living in FD (Table 1). They also had an unfavorable cardiovascular risk profile with a higher prevalence of hypertension and smoking rates, higher body mass index and fasting blood glucose levels, and a higher ASCVD score (P=0.007). Moreover, they also had higher levels of hs-CRP, lower levels of glutathione, and higher AI compared with those living in non-FD areas (Table 1). These differences remained significant for glutathione and AI after adjustment for traditional risk factors (Tables 1 and 2).

The population was further classified on the basis of residing in neighborhoods and its access to healthy food and area income in addition to individual household income (Table 3; Figure [B and C]). Subjects living in neighborhoods with low access to healthy food were more often black, with lower education level, higher body mass index, and prevalence of hypertension compared with subjects living in favorable food access areas. Similar findings in addition to higher prevalence of diabetes mellitus, smoking, hyperlipidemia, and higher ASCVD score were observed in subjects living in low-income compared with high-income areas. The sample was further classified based on individual income defined as annual household income <$50 000. Similar differences in cardiovascular risk factors were observed in subjects with low compared with high individual income.

The differences in subclinical vascular disease was also examined in subjects based on their neighborhood characteristics and individual SES. There were no differences in the markers of inflammation, OS, or vascular stiffness in those living in low compared with good access to healthy foods (Tables 2 and 3). However, subjects living in low-income areas had a higher hs-CRP levels, lower glutathione levels, higher PWV, and AI compared with subjects residing in higher income areas (Table 3). The differences in hs-CRP and glutathione levels remained significant after adjustment for cardiovascular risk factors (Table 2). Low individual income was also associated with higher levels of hs-CRP, PWV, and AI and lower glutathione level, and the differences remained significant after adjustment for cardiovascular risk factors (Tables 2 and 3).

Subjects were separated into 4 categories based on whether they were living in high-income areas with either good (n=356) or low access to healthy foods (n=597), or in low-income areas with either good (n=281) or low access to healthy food (FD, n=187; Figure I through V in the Data Supplement). Subjects living in low-income areas with low food access (FD) had no significant difference in ASCVD score, hs-CRP, PWV, cystine, or glutathione levels compared with subjects living in areas with low income and good food access areas. Only subjects living in low-income areas, irrespective of whether they had good or poor access to healthy food, had higher hs-CRP level and higher OS (low glutathione) when compared with subjects living in high-income areas, regardless of their food access (Figure I through V in the Data Supplement).

Subjects were separated into 4 categories based on whether they were living in areas with good or poor access to healthy food and whether they had high versus low individual income (Table I and Figure I through V in the Data Supplement). There were no significant differences in cardiovascular risk factors, CVD risk estimation, inflammation, arterial stiffness, and OS between subjects with low individual income living in areas with poor or good access to healthy food. Similarly, no differences in these measures were observed in subjects with high individual income living in varying food access areas. Thus, individuals with high income living in areas with low access to healthy food had lower risk factor burden and ASCVD scores (P=0.001), lower hs-CRP levels (P=0.004), and higher glutathione levels (P=0.005) compared with subjects with lower income living in similar low access area (Figure I through V in the Data Supplement). These data indicate that individual income and not food access was a more important contributor to CVD risk.

The distribution of risk factors, biomarkers, and vascular function was also studied to distinguish the effects of individual versus the area income levels (Table II and Figure I throgh V in the Data Supplement). Subjects with low individual income living in low-income areas were more likely to be black with lower educational level, higher prevalence of diabetes mellitus and smoking, higher hs-CRP levels (P=0.05), higher cardiovascular risk (ASCVD P<0.001), and lower glutathione level (P=0.041) compared with individuals with higher individual income living in low-income areas. Similarly, subjects with low individual income living in high-income areas were more often black with poor education levels, higher ASCVD score, higher Hs-CRP levels, and lower glutathione levels compared with those with high income living in similar high-income areas (Figure I through V in the Data Supplement).

Multivariate analyses were performed after adjustment for traditional cardiovascular risk factors to investigate which of these SES and neighborhood factors were associated with CVD risk markers. When individual income, area income, and food access were entered into the same model, independent predictors of higher ASCVD risk score were both low individual and low neighborhood income. Only low individual income was an independent predictor of both hs-CRP and AI after adjusting for CVD risk factors. Finally, area income (P=0.031) was also an independent predictor of OS (Table 4). Thus, the major predictors of increased CVD risk were individual and neighborhood income, whereas food access and living in FD were not independent predictors of CVD risk. There was no significant interaction between area income, individual income, and food access with ASCVD risk score, OS, and inflammation. The data were unchanged when adjusted for study cohort.

The present study is the first to report the relationship between living in FD (defined as both low income and access to healthy food areas by the USDA), cardiovascular risk factors, and subclinical vascular disease. Previous studies have focused on select neighborhood characteristics and SES in addition to composite scores of neighborhood features.^2,37 For example, areas with poor food quality, low access to healthy foods, or more fast-food restaurants are associated with obesity and diabetes mellitus risks.^38,39 A recent study showed that although living in FDs was associated with decreased fruit and vegetable intake and higher systolic blood pressure, low individual income rather than FDs was associated with higher odds of chronic kidney disease.⁴⁰ Novel findings of our study are the relationships between FD and subclinical CVD measured as vascular dysfunction, inflammation, and OS, and the fact that these are driven by income rather than access to healthy foods.^13,19

There is controversy on the cardiovascular health effects of food access.^41,42 In the Atherosclerosis Risk in Communities study, increased prevalence of convenience stores was associated with a higher incidence of obesity but not diabetes mellitus, hypertension, or dyslipidemia.⁴³ However, other studies did not confirm these findings.^38,44 We found that living in areas with low access to healthy food was not associated with a higher CVD risk, inflammation, OS, or arterial stiffness compared with subjects living in areas with good access to healthy food. Thus, food access by itself, measured by proximity to supermarkets, might not contribute to increased cardiovascular risk, and the relative cost of higher quality food rather than access may be a major barrier to healthy lifestyle and choices.^6,45

The relationship between SES and CVD risk is well established.^2,37,46–48 There is striking geographical variation in the distribution of cardiovascular risk factors and CVD mortality in the United States.⁴⁹ Neighborhood SES, deprivation, lack of cohesion, decreased access to recreational resources, lack of public space, and safety are some of the factors that can affect these outcomes.¹ The income status of neighborhoods can also influence the diversity of food resources, prices, and, thus, food access.^50,51 Our analyses showed that the income status of areas is an independent determinant of cardiovascular risk and subclinical vascular disease.

Our study also showed that it was the individual income, rather than the neighborhood income or food access, that was associated with higher risk of developing CVD, inflammation, and OS. Thus, people with high individual income who lived in low-income areas had lower CVD risk and inflammation compared with subjects with lower individual income who lived in similar area. Moreover, people with high individual income who lived in an area with poor food access had better cardiovascular profile than those with lower individual income who lived in similar area. It seems that the adverse association of FD with health is partly driven by the area income and most importantly by individual income status, rather than access to food. There are several initiatives that promote healthy food access by providing incentives for individuals with low income. Georgia provides incentives for participants in the Supplemental Nutrition Assistance Program to increase the value of Supplemental Nutrition Assistance Program benefits if healthy food is purchased at farmers’ markets. However, few farmers’ markets are located in FDs and few accept Supplemental Nutrition Assistance Program in Georgia.⁵² Providing support services for eligible low-income families may overcome some of the adverse health outcomes driven by low SES. A recent report by the Center for Disease Control and Prevention showed that enrollment in Supplemental Nutrition Assistance Program for Women, Infants, and Children decreased the prevalence of childhood obesity among low-income families.⁵³

The study also demonstrated that residents living in disadvantaged areas including FD, low income, and low-access areas had a higher proportion of black residents. Black neighborhoods have been reported to have more fast-food restaurants,⁵⁴ fewer supermarkets,⁵⁵ fewer healthy options, lower levels of social cohesion, and worse walking environments.⁵⁶ Our data shows that blacks also formed a higher proportion of those with low income, a key driver of increased CVD risk.

Our study has several strengths. It investigates the relationship between components of FD and cardiovascular risk factors in a region of United States where there is tremendous racial and regional disparity in the incidence of CVD. It investigates the health impact of socioeconomic and neighborhood features on subclinical CVD, including inflammation, OS, and arterial stiffness. Limitations include its cross-sectional nature where causality between income and living in FD and subclinical vascular disease cannot be established. Second, our study was conducted in a single urban area in southeastern United States and may not be generalizable to other regions. Third, geographic locations and characteristics can change with time. However, a recent update by the USDA showed relatively small change in low-income and low-access neighborhoods between 2010 and 2015 data.⁵⁷ Previous studies also showed that individuals usually move into neighborhoods with similar SES across their life course.^58,59

People living in FD had a higher prevalence of cardiovascular risk factors, inflammation, OS, and arterial stiffness. These associations are largely driven by area and individual income rather than access to healthy food. The implications of our findings are that at least in urban areas, risk of CVD seems to be associated less with access to healthy food and more with socioeconomic factors. This understanding may help to better tailor resources to affected communities and improve utilization of public health resources.