Residential proximity to greenness mitigates the hemodynamic effects of ambient air pollution

Abstract

Residential proximity to greenness is associated with a lower risk of cardiovascular disease (CVD) and all-cause mortality. However, it is unclear whether the beneficial effects of greenness are linked to a reduction in the effects of ambient air pollutants. We measured arterial stiffness in 73 participants with moderate to high CVD risk. Average levels of ambient PM_2.5 and ozone were calculated from local monitoring stations. Residential greenness was estimated using satellite-derived normalized difference vegetation index (NDVI) for a 200-m and 1-km radius around each participant’s home. Participants were 51% female, average age of 52 yr, and 79% had diagnosed hypertension. In multiple linear regression models, residential NDVI was negatively associated with augmentation index (−3.8% per 0.1 NDVI). Ambient levels of PM_2.5 [per interquartile range (IQR) of 6.9 μg/m³] were positively associated with augmentation pressure (3.1 mmHg), pulse pressure (5.9 mmHg), and aortic systolic pressure (8.1 mmHg). Ozone (per IQR of 0.03 ppm) was positively associated with augmentation index (5.5%), augmentation pressure (3.1 mmHg), and aortic systolic pressure (10 mmHg). In areas of low greenness, both PM_2.5 and ozone were positively associated with pulse pressure. Additionally, ozone was positively associated with augmentation pressure and systolic blood pressure. However, in areas of high greenness, there was no significant association between indices of arterial stiffness with either PM_2.5 or ozone. Residential proximity to greenness is associated with lower values of arterial stiffness. Residential greenness may mitigate the adverse effects of PM_2.5 and ozone on arterial stiffness.

NEW & NOTEWORTHY Previous studies have linked proximity to green spaces with lower cardiovascular disease risk. However, the mechanisms underlying the salutary effects of green areas are not known. In our study of participants at risk of cardiovascular disease, we found that arterial stiffness was positively associated with short-term exposure to PM_2.5, PM₁₀, and ozone and inversely associated with greenness. The association between pollution and arterial stiffness was attenuated in areas of high greenness, suggesting that living green neighborhoods can lessen the adverse cardiovascular effects of air pollution.

INTRODUCTION

Cardiovascular disease (CVD) is the leading cause of death worldwide (1). Although many factors contribute to the risk and progression of the disease, recent estimates suggest that up to 70%–80% of CVD burden could be attributed to non-genetic environmental factors, such as lifestyle choices, socioeconomic status, air pollution, lack of surrounding greenness (2), and residential characteristics (3). Indeed, emerging evidence has shown that living in greener areas results in improved health and is associated with lower mortality (4, 5), and reduced CVD risk (2, 6). A recent analysis of 24,845 adults in the 33 Communities Chinese Health Study found that higher levels of residential greenness were associated with lower systolic blood pressure and lower odds of hypertension (7). We have previously shown that residential proximity to greenness is associated with decreased urinary epinephrine, lower oxidative stress, and a pro-angiogenic state, suggesting that living in green neighborhoods is associated with good cardiovascular health (8).

Despite frequent reports of inverse associations between residential greenness and CVD risk, the mechanism underlying the salutary effects of greenness remain unclear. Several studies show that living in green areas promotes physical activity and social cohesion, reduces mental stress, and mitigates against exposure to air pollution (2). Of these, the effects of greenness on air pollution may be particularly significant. Worldwide, exposure to outdoor air pollution is linked with 3.3 million premature deaths per year (9), and there is substantial evidence showing that exposure to ozone and fine particulate matter (PM_2.5) air pollution has adverse cardiovascular effects (10–13). Short-term exposures to ozone or PM_2.5 have been associated with an increase in susceptibility to acute coronary events (12, 14). Exposure to PM_2.5 is also associated with higher systolic blood pressure in healthy adults (15). Moreover, short-term PM_2.5 and ozone exposures are associated with decreased brachial artery flow-mediated dilation (endothelial dysfunction) (16–18), indicating that many components of the ambient air pollution have adverse cardiovascular effects and impair cardiovascular health.

Nevertheless, it is unclear whether the beneficial effects of neighborhood greenness are related to reduction in area air pollution. Trees have the ability to filter and block particulate matter, and we have recently shown that people living in greener areas were exposed to lower levels of volatile organic compounds (19). Moreover, a study on 2.4 million Canadian adults found that greenness attenuated the association between PM_2.5 and mortality, with no increased risk of mortality associated with PM_2.5 among those living in high greenness areas (20). Similarly, results from the LISAplus and GINIplus cohorts show that the positive association between insulin resistance and PM₁₀ was lost upon adjusting for area normalized difference vegetation index (NDVI) (21). The link between air pollution and area greenness is further supported by a study of Chinese communities, which found that the effects of greenness on metabolic syndrome are, in part, mediated by PM₁₀, nitrogen dioxide, and ozone (22). Nonetheless, it remains unclear whether area greenness is associated with direct measures of cardiovascular function and health.

To understand how residential greenness affects the adverse cardiovascular effects of air pollution, particularly in a vulnerable population, we examined the relationship between residential vegetation (greenness) and central hemodynamics in participants with moderate-to-severe CVD risk and how this relationship is affected by concurrent levels of air pollution. Our working hypothesis was that by decreasing the negative cardiovascular effects of air pollution, residential proximity to areas of high greenness is associated with lower levels of arterial stiffness.

METHODS

Study Population and Design

Study participants (n = 73) were recruited from the University of Louisville Hospital and affiliated clinic system between March 2010 and July 2014. The participants were from the Louisville, KY area and aged 23–84 yr at the time of enrollment. They had moderate-to-high CVD risk (e.g., hypertension, hypercholesterolemia, obesity, diabetes) that required management, or they had overt heart disease. Participants were administered a questionnaire that included demographic information, residential address, smoking history, exercise status, and CVD history. Pulse wave analysis measurements were acquired on the day of recruitment. Medical records were reviewed to provide medication history. Participants provided a urine sample at the time of enrollment. Urinary cotinine, a metabolite of nicotine, was quantified for all participants by ultra-performance liquid chromatography-mass spectrometry, as previously described (23). Participants were not instructed to refrain from drinking coffee or smoking before the exam. Although both coffee consumption and smoking can transiently increase both the augmentation index and the pulse wave velocity (24, 25), these changes are acute and return to baseline within 30 min. Because pulse wave analysis measurements were acquired after a clinic visit (lasting 1–2 h), coffee consumption or smoking before the visit is unlikely to affect measurements of arterial stiffness.

Median household income of the participants at the block group level was collected from the US Census Bureau for 2010, as a proxy for socioeconomic status (26). Participants were recruited through University of Louisville staff emails, posted flyers, and directly through cardiology clinics. Recruitment at clinics consisted of prescreening patients during active enrollment and in-person recruitment of potential participants during waiting periods. Study exclusion criteria included inability to provide written informed consent; vulnerable populations (e.g., pregnant or lactating females and prisoners); lung, liver, kidney, or hematological disease; coagulopathies; substance abuse; chronic cachexia; and severe comorbidities. The study was approved by the Institutional Review Board (IRB number 09.0174) at the University of Louisville, and all participants gave written informed consent.

Residential Greenness

We used satellite-derived normalized difference vegetation index (NDVI) data from the US Geological Survey to estimate exposure to vegetation using methods previously reported (27). The NDVI calculates the ratio of the difference between red and near-infrared reflectance and the sum of these measures, with a range of −1 to 1, with higher values indicating greater vegetation density. We used geographic information systems software from ArcMap9.3+ (ESRI, Redlands, CA) to estimate the peak NDVI value within a 200-m and 1-km radius of each participant’s home. Due to missing addresses, we were unable to calculate NDVI for 8 participants. We also calculated the standard deviation within a 200-m and 1-km radius to estimate the amount of spatial variation in vegetation around individuals’ homes. We chose a 200-m radius to reflect the measure of greenness accessible at each participant’s residence, and 1 km to reflect the general walking distance near the residence.

Air Pollution and Meteorological Data

Data on ambient levels of pollutant concentrations were retrieved from regional EPA-validated monitoring stations in the Louisville, KY region, that report daily pollutant levels (28). Measurements included PM_2.5 (particulate matter with an aerodynamic diameter ≤2.5 μm; μg/m³), PM₁₀ (particulate matter with an aerodynamic diameter ≤10 μm; μg/m³), and ozone (ppm). These values were obtained for the 7-day window preceding and including the day the study participants had their pulse wave analysis completed. We calculated moving averages based on short-term exposure windows of 1, 3, and 7 days. Daily averages were obtained by calculating the daily average from midnight to midnight.

Ozone levels were reported by maximum of the daily 8-h running average concentration—the 8-h time span with the highest average concentration from midnight to midnight on each given day. The present analyses examined the maximum 8 h concentrations of all monitors within the region on the day of the study visit.

Contiguous temperature was obtained from agroclimatology data provided by the NASA applied sciences program (29). Data were collected, processed, and calibrated to ground-based measurements by the NASA prediction of worldwide energy resource project.

Pulse Wave Analysis Measurements

Measures of arterial stiffness (augmentation index, augmentation pressure, and pulse pressure), aortic systolic pressure, and SEVR were derived from pulse wave analysis using a SphygmoCor Cardiovascular Management System (Atcor Medical Pty. Ltd., Sydney, Australia). Data were collected by a trained study nurse practitioner or a medical doctor using a laptop computer installed with SphygmoCor Cardiovascular Management Suite (Version 9.0). Participants were seated and after 5 min of rest, their dominant arm was extended onto a flat surface to ensure that their elbow was at or below the heart. Radial applanation tonometry was conducted, recording 10 s of consistent, quality waveforms. For each patient, the procedure was completed twice, and the values with the highest operator index were used for analysis.

Peripheral waveforms were transformed into central aortic waveforms by the system software using a previously validated transfer function (30). Augmentation index (AIX), augmentation pressure (AP), pulse pressure (PP), systolic pressure, diastolic pressure, and SEVR were all derived from the calculated aortic waveform. The AP is defined as the height above the first systolic shoulder of the aortic waveform, and PP is the difference between systolic and diastolic pressures (31). The AIX is a percentage of AP to PP, that is, (AP/PP) × 100. SEVR, a measure of oxygen supply versus demand, was calculated as the area under the curve during diastole divided by the area under the curve during systole of the central aortic pulse wave. All pulse wave analyses indices are automatically calculated by the SphygmoCor software. Values obtained from the SphygmoCor system are highly reproducible (32–34).

Statistical Analyses

Participant characteristics are expressed as means (±SD) for continuous variables and n (%) for categorical variables. Pearson correlation coefficient was used to measure the linear correlations between the exposure variables. Multiple linear regression was used to examine the relationship between the 1, 3, and 7-day air pollution variables and NDVI with the pulse wave analysis indices. The dependent variables examined were augmentation index, augmentation pressure, pulse pressure, aortic systolic pressure, aortic diastolic pressure, and SEVR. All models were adjusted a priori for sex, race, body mass index (BMI), age, heart rate, and smoking status. Pollutant models were additionally adjusted for temperature. We included an interaction term between low/high NDVI (median cutoff) and 7-day pollutants to assess effect modification by vegetation. We also performed two-pollutant models to determine whether associations found in one-pollutant models were due to covarying pollution levels. Ambient pollutant models were additionally adjusted for residential greenness within 200 m to identify whether inclusion of greenness impacted the associations; however, no differences were observed (data not shown).

Regression results represent change per interquartile range for air pollution metrics, change per 0.1 NDVI for peak greenness, and change per 0.1 SD for the spatial variation of NDVI. As a sensitivity analysis, we analyzed the association between greenness and arterial stiffness additionally adjusting for: 1) median household income at the census block level; 2) urinary cotinine, a metabolite of nicotine; and 3) and exercise (yes/no). Sensitivity power analysis was conducted using n = 73, power = 80%, and α = 0.05, which resulted in a minimum detectable effect size of f² = R²/(1 − R²) = 0.11. All statistical analyses were performed using SAS, version 9.4, software (SAS Institute, Inc., Cary, NC) and GraphPad Prism, version 8 (GraphPad Software, La Jolla, CA).

RESULTS

Of the 73 participants in our study, there were 36 males (49%) and 37 females (51%), and a slightly higher proportion of Blacks (53%) as compared with Whites (Table 1). The mean age of participants was 52.2 yr with a BMI of 32.3. Most participants were nonsmokers (70%) and diagnosed with hypertension (79%).

Table 1.

Demographics characteristics and cardiovascular disease history of study participants

Categorical Variable, n (%)
Sex
Male	36 (49)
Female	37 (51)
Race
White	30 (41)
Black	39 (53)
Other	4 (5)
CVD risk factors
Hypertension	57 (79)
Hyperlipidemia	34 (49)
Diabetes	14 (19)
Current smoker	22 (30)
Exercise	35 (49)
Cardiovascular history
Myocardial infarction	6 (8)
Stroke	6 (8)
Heart failure	8 (11)
Angina	16 (22)
Arrhythmia	24 (34)
Continuous variable, means (SD)
Age, yr	52.2 (12.8)
BMI	32.3 (8.7)
Heart rate beats/min	68.6 (12.5)
Systolic BP, mmHg	134.4 (20.7)
Diastolic BP, mmHg	82.1 (11.8)
Cotinine, mg/g creatinine	555.4 (1,402.0)
Median household income ×10−3	44.3 (29.1)
Outcome measures
Augmentation index, %	25.6 (14.7)
Augmentation pressure, mmHg	11.5 (9.1)
Aortic pulse pressure, mmHg	39.8 (15.5)
Aortic systolic pressure, mmHg	122.7 (20.9)
Aortic diastolic pressure, mmHg	82.9 (12.1)
SEVR, %	169.5 (32.4)

Values are n (%) or means (SD); N = 73 total participants. BMI, body mass index [weight (kg)/height (m)²]; median household income, median household income at census block group level; SD, standard deviation; SEVR, subendocardial viability ratio.

To examine the relationship between residential greenness and arterial stiffness, we calculated the peak NDVI value within zones of different radii around the participant’s home. The average NDVI (200 m) for participants was 0.40 (Table 2). We observed a strong correlation between NDVI at 200-m and 1-km radii (r = 0.83). There was a negative correlation between average NDVI and standard deviation of NDVI. We found that for a 0.1 increase in NDVI (200-m radius), there was a −3.8% (95% CI: −5.9, −1.7) difference in augmentation index (Fig. 1). This association remained significant within a 1-km radius of the participants’ home.

Table 2.

Summary statistics and Pearson correlation coefficients of exposure variables

	n	Means (SD)	Median (IQR)	PM2.5	Ozone	PM10	Temp	200 m	1 km	200 m SD	1 km SD
Pollutant	73
PM2.5, μg/m3		14.1 (5.1)	13.1 (6.9)	1	0.21	0.64**	0.003
Ozone, ppm		0.05 (0.02)	0.05 (0.02)		1	0.49**	0.69**
PM10, μg/m3		23.0 (11.0)	20.7 (11.5)			1	0.37*
Temperature, °C		14.3 (13.8)	15.4 (28.4)				1
NDVI	65
200 m		0.40 (0.12)	0.41 (0.12)					1	0.83**	−0.36*	−0.29*
1 km		0.38 (0.13)	0.40 (0.18)						1	−0.36*	−0.41*
200 m SD		0.13 (0.04)	0.12 (0.05)							1	0.49**
1 km SD		0.16 (0.03)	0.16 (0.04)								1

IQR, interquartile range; NDVI, normalized difference vegetation index; PM_2.5, particulate matter with aerodynamic diameter <2.5 μm; SD, standard deviation. *P < 0.05, **P < 0.001.

Figure 1.

Association between peak NDVI metrics with pulse wave analysis variables. Results are presented as change in mmHg (% for AIX and SEVR) and 95% confidence intervals, per 0.1 NDVI. Linear regression models were used to test for associations between outcomes and NDVI metrics (n = 65). Models are adjusted for sex, race, BMI, age, heart rate, and smoking status. AIX, augmentation index; AP, augmentation pressure; BMI, body mass index; DBP, aortic diastolic pressure; NDVI, normalized difference vegetation index; PP, pulse pressure; SBP, aortic systolic pressure; SEVR, subendocardial viability ratio.

Given that greenness variability has been shown to be inversely associated with CVD outcomes (6), we examined the association between arterial stiffness and the standard deviation of NDVI within a 200-m and 1-km radii of homes. For a 0.1 increase in SD within a 200-m radius, we found negative associations with augmentation index (−7.4%; 95% CI: −13.5, −1.3), augmentation pressure (−4.8 mmHg; 95% CI: −9.0, −0.7), and aortic systolic pressure (−14.7 mmHg; 95% CI: −28.1, −1.3). We did not observe any significant associations at a 1-km radius. Collectively, these results indicate that those who live in areas of higher greenness or higher variability in greenness have better hemodynamic function than those who live in less green areas.

Our previous work has shown that houses in the area of higher greenness have a greater household income (35); therefore, given the widely established association between socioeconomic status and CVD risk (36), we examined whether household income affects the relationship between greenness, air pollution, and arterial stiffness. In the cohort of our study participants, we found a significant correlation between NDVI and median household income. The correlation was 0.60 and 0.74 for 200-m and 1-km NDVI values. Median household income was also correlated with SD of NDVI: −0.40 and −0.45 for 200-m and 1-km SD NDVI (see Supplemental Table S1; all Supplemental material is available at https://doi.org/10.6084/m9.figshare.13517411). Nonetheless, adjusting for median household income did not significantly alter the relationship between NDVI and any of the indices of arterial stiffness (Supplemental Table S2).

In addition to median household income, we conducted sensitivity analysis to adjust for differences in physical activity and smoking; both of which profoundly affect CVD risk (37). However, we found that adjusting for self-reported physical activity levels did not significantly affect the relationship between NDVI and the indices of arterial stiffness, while adjusting for urinary levels of cotinine (a biomarker for tobacco product use) led to only a minor attenuation of the relationship between NDVI and arterial stiffness (Supplemental Table S2). Overall, the results of these sensitivity analyses indicate that the effects of greenspaces on hemodynamic function are largely independent of median household income, physical activity, and tobacco use.

To examine the relationship between hemodynamic function with other environmental factors, we assessed the effects of PM_2.5, PM₁₀, and ozone levels. The median daily concentrations of PM_2.5, PM₁₀, and daily ozone levels in the area were below the national ambient air quality standards (NAAQS) of 35 μg/m³ (24 h standard), 150 μg/m³ (24 h standard), and 0.07 ppm (8 h standard), respectively. A weak correlation was observed between daily PM_2.5 concentration and daily maximum ozone levels (r = 0.21) (Table 2). Daily PM_2.5 and temperature showed little correlation, with r = 0.003. The strongest correlations were observed between PM_2.5 and PM₁₀ (r = 0.64), and between daily ozone and temperature (r = 0.69).

Associations between central hemodynamics with 1-, 3-, and 7-day pollutant levels are presented in Fig. 2. An interquartile range increase in 1-day PM_2.5 (IQR = 6.9 μg/m³) was positively associated with augmentation pressure (3.1 mmHg; 95% CI: 1.11, 5.1), pulse pressure (5.9 mmHg; 95% CI: 1.7, 10.0), aortic systolic pressure (8.1 mmHg; 95% CI: 2.0, 14.1), and inversely associated with SEVR (−6.8%; 95% CI: −13.1, −0.4). Additionally, 1-day ozone levels were positively associated with augmentation index (5.5% 95%CI: 1.3, 9.8), augmentation pressure (3.1 mmHg; 95% CI: 0.03, 6.2), and aortic systolic pressure (10.1 mmHg; 95% CI: 1.2, 19.1) per IQR (0.025 ppm). We also observed positive associations between PM₁₀ and augmentation pressure, pulse pressure, and aortic systolic pressure, ranging from 2.6 to 7.6 mmHg per IQR of 11.5 μg/m³. The associations for augmentation pressure and aortic systolic pressure remained significant at 3- and 7-day lags, while aortic diastolic pressure was found to be positively associate with PM₁₀ at 3-day lag. The associations with PM_2.5 and ozone were lost at 3- and 7-day. In two-pollutant models using a 1-day lag, we observed similar associations as compared with one-pollutant models, with the exception of the model including both PM_2.5 and PM₁₀, in which associations were lost (Supplemental Fig. S1).

Figure 2.

Association between 1-day, 3-day, 7-day average exposure levels with pulse wave analysis variables. Results are presented as change in mmHg (% for AIX and SEVR) and 95% confidence intervals, per IQR of pollutant level. Linear regression models were used to test for associations between outcomes and pollutants (n = 73). Models are adjusted for sex, race, BMI, age, heart rate, smoking status, and temperature. AIX, augmentation index; AP, augmentation pressure; BMI, body mass index; DBP, aortic diastolic pressure; IQR, interquartile range; PP, pulse pressure; SBP, aortic systolic pressure; SEVR, subendocardial viability ratio.

To identify whether greenness may modify the effect of pollutants on arterial stiffness, we tested for interactions between low/high NDVI (200-m radius) and 7-day pollution levels (Fig. 3). For each of the pollutants, we found stronger associations between the levels of pollution and health outcomes in participants living in low greenness areas. The effect of ozone on augmentation pressure, pulse pressure, and aortic systolic pressure was significantly lower in areas of high greenness. Similarly, we also found significant interactions between PM₁₀ and greenness in augmentation pressure and pulse pressure models. No significant associations between health outcomes and pollution were observed for those living in high green areas, indicating that greenness may be protective against pollution-related effects on arterial stiffness.

Figure 3.

Association between 7-day pollutants with pulse wave analysis variables by low/high NDVI (200 m). Results are presented as change in mmHg (% for AIX and SEVR) and 95% confidence intervals, per IQR of pollutant level. Linear regression models with a pollutant and greenness interaction term were used to test whether the association between pollutants and outcomes was modified by greenness (high/low) (n = 65). Models are adjusted for sex, race, BMI, age, heart rate, smoking status, and temperature. AIX, augmentation index; AP, augmentation pressure; BMI, body mass index; DBP, aortic diastolic pressure; IQR, interquartile range; NDVI, normalized difference vegetation index; PP, pulse pressure; SBP, aortic systolic pressure; SEVR, subendocardial viability ratio. *Represents significant interaction between pollutant and low/high NDVI (P < 0.05).

DISCUSSION

The results of this study support the hypothesis that residential greenness is associated with better vascular function in part by attenuating the effects of ambient air pollution. We found that the inverse relationship between augmentation index and peak greenness was independent of radius around the residence, but at smaller radii (200 m) buffer around the home, we observed inverse associations between standard deviation of NDVI and augmentation pressure, aortic pulse pressure, and aortic systolic pressure. Positive associations were also observed with ambient PM_2.5, PM₁₀, and ozone concentrations with arterial stiffness. Greenness significantly modified the effect of PM₁₀ and ozone on arterial stiffness metrics. We observed significant associations between several arterial stiffness metrics and pollutants in low greenness areas, whereas the association between pollutants and arterial stiffness measures was not significant in areas of high greenness.

This study is one of the first to show that levels of residential vegetation are inversely associated with arterial stiffness. One previous study, using the Whitehall II cohort, found an inverse association between long-term greenness exposure and arterial stiffness at baseline; however, they did not find an association with progression of arterial stiffness (38). This study differed from ours in that it was predominantly male (74.5%), white (92%), and measured carotid-femoral pulse wave velocity (cv-PWV). Moreover, our study cohort had a higher baseline risk of CVD than the Whitehall II cohort.

Several investigators have reported that living in areas of high greenness is associated with reduced risk of cardiovascular mortality and morbidity (2). However, the mechanisms underlying the association between greenness and cardiovascular health are not fully understood. Our finding that greenness was associated with augmentation index, but not other metrics of arterial stiffness, including aortic systolic pressure, implies that greenness may affect arterial stiffness, independent of blood pressure changes. These observations are consistent with our previous work showing that individuals who live in areas of high greenness have higher levels of circulating angiogenic cells (8). Given that these cells aid in endothelial repair (8), these data, along with our present observations, lend support to the notion that living in areas of high greenness may be associated with better vascular health, in general, and better endothelial function in particular. Indeed, several previous studies have demonstrated a link between blood pressure and vegetation levels (2, 7). There is extensive data showing that arterial stiffness is a robust and sensitive predictor of both cardiovascular events and mortality, as well as vascular aging (39). However, we did not find a strong association of blood pressure with NDVI, which may be in part because many of our study participants were hypertensive and on blood pressure medications. Nevertheless, the strong association between arterial stiffness and NDVI indicates that there may be a sensitive link between living in green areas and vascular health.

In addition to the association between arterial stiffness and NDVI, we observed significant inverse associations between NDVI standard deviation within a 200-m radius and aortic systolic pressure, augmentation pressure, and aortic pulse pressure. The standard deviation of NDVI captures the heterogeneity of greenness within a given radius. Vegetation is not randomly distributed, and it commonly clusters in specific areas. A high standard deviation reflects areas that contain both low and high vegetation. Some common neighborhood characteristics that contribute to high variability include trees along roadways and the presence of parks with parking lots. A cross-sectional study of 11,404 adults in Australia found that the odds of hospitalization for heart disease or stroke was 37% lower among adults in areas of highly variable greenness, compared with those in low variability areas (6). There are two hypothesized explanations for why high variability in greenness could be protective of cardiovascular health. First, areas of high greenness variability may be associated with increased physical activity (6). Additionally, areas of high variability in greenness consist of roads aligned with trees, which provide a buffer to traffic pollutants and potentially other environmental factors such as noise (40).

Our finding that arterial stiffness was only associated with standard deviation at the 200-m radius, but not the 1-km radius provides further evidence in support of the notion that areas of high NDVI SD may be areas intersected by major roadways lined by trees. As roadways are sources of pollutants, street trees within a 200-m radius around the individuals’ residence would be more relevant in blocking exposure to pollution. Indeed, previous research has shown that residential greenness levels in smaller buffers (50 m and 100 m) are associated with reduced levels of indoor PM_2.5 (41).

We found that in addition to NDVI, arterial stiffness was strongly associated with ambient air pollutants. The link between arterial stiffness and air pollutants has been reported in several previous studies, however, with mixed results. For instance, a large, cross-sectional study consisting of hypertensive, and normotensive controls did not find any association between arterial stiffness correlates and PM₁₀ (42). However, they did find a positive association between PM₁₀ and augmentation pressure, but only in males. In a study of the Framingham Offspring and Third Generation cohorts, short-term exposures to PM_2.5 are not associated with vasodilator responses; however, PM_2.5 is associated with baseline amplitude, suggesting adverse alterations in vascular tone (43). A longitudinal study consisting of 370 males with a mean age of 78 yr found a significant association between PM_2.5 and augmentation index at 3 days, but only marginal association (P < 0.10) was observed between augmentation index and ozone (44). Acute exposure to PM_2.5 has been shown to increase augmentation index values in welders after a day of occupational exposure (45), and similarly, augmentation index values were increased in healthy males exposed to diesel exhaust (46).

We found that ozone, but not PM_2.5, was significantly associated with higher augmentation index. This suggests that ozone-induced effects on arterial stiffness are independent of PM_2.5 exposure and potentially stronger. These findings are in agreement with a study by Wu et al., which found that a 4.8% increase in right cardio-ankle vascular index (r-CAVI), an arterial stiffness index, is associated with a 17.6-ppb personal ozone exposure (47). No association was found between PM_2.5 and r-CAVI, however, when PM exposure is segregated to between 1 μm and <2.5 μm, a significant increase of 2.5% r-CAVI is reported. Other studies have also reported the more potent effects of ozone compared with PM with respect to cardiovascular morbidity and mortality. Henrotin et al. (48) reported that ischemic stroke is associated with ozone levels, but not PM₁₀. Similarly, Hong et al. found that ozone exposure has a greater effect on stroke mortality than particulate matter (49).

Many epidemiological studies have demonstrated an independent association between blood pressure and exposure to PM_2.5 (50, 51). Some studies have also shown that central pressures provide a better prediction of CVD risk than does peripheral pressure. We found a strong association between aortic systolic pressure and PM_2.5 levels; an IQR of 6.9 μg/m³ PM_2.5 was associated with an 8.1-mmHg higher aortic systolic pressure. While these results are two to three orders of magnitude larger than similar studies (50), our study comprises patients with high CVD risk who may be more susceptible to the harmful effects of air pollution. Augmentation pressure and pulse pressure were also associated with PM_2.5 concentrations, even after adjusting for temperature, while augmentation index was not. Both augmentation pressure and pulse pressure have a stronger correlation with aortic systolic pressure, as compared with augmentation index, r = 0.66, 0.81, and 0.32, respectively. Together, these findings indicate that: 1) in comparison with augmentation index, augmentation pressure, and pulse pressure may be less specific measures of arterial stiffness and 2) that PM_2.5 may have a stronger effect on blood pressure than arterial stiffness. However, we found an inverse relationship with NDVI, where the effects on arterial stiffness seems to be more consistent and stronger than the effects of blood pressure, suggesting that either the effects of greenness are weaker than those of air pollutants or that they operate via different mechanisms that overlap only partially. Indeed, additional adjustment of greenness in our pollutant models showed minimal changes in associations.

Aortic blood pressure has been demonstrated to have an important role in regulating left ventricular function. The left ventricular afterload can be described as the pressure multiplied by the time area under the curve during systole (SPTI). Thus, left ventricular afterload is inversely related to left ventricular ejection fraction, as an increase in SPTI will be accompanied by a decrease in ejection fraction. Accordingly, the time area under the curve during diastole (DPTI) is a potential indicator of subendocardial blood supply. The ratio of DPTI to SPTI, SEVR, provides a hemodynamic estimation of myocardial perfusion relative to cardiac workload. A decrease in SEVR has been associated with impaired coronary flow reserve (52), and severity of Type I and Type II diabetes (53, 54).

Previous work has shown that short term exposure to high levels of PM_2.5 was associated with significant decreases in SEVR (55). A key finding in the present study is the association between both ambient PM_2.5 (1 day) and ozone concentrations (7 days in low greenness areas) with SEVR. To the best of our knowledge, this is the first study to report such findings, both for low levels of PM_2.5 and ozone. Similarly, we found a borderline inverse association between 7-day PM₁₀ and SEVR in areas of low greenness (P value = 0.058), whereas the relationship was attenuated for both PM₁₀ and ozone in areas of high greenness. This suggests that exposure to ambient air pollutants is associated with SEVR, and thus, could lead to increased risk of myocardial infarction; however, living in areas of high greenness may provide protection against changes in SEVR.

In our models stratified by low and high greenness, we found significant associations between pollutants and augmentation pressure, pulse pressure, aortic systolic pressure, and SEVR in the low greenness areas. These associations were attenuated in areas of high greenness. This is consistent with previous studies that have found that greenness attenuated the effects of PM_2.5 on mortality (20, 56, 57). Interestingly, stratifying by low and high greenness also revealed significant associations, which we did not observe in our full sample. This highlights the importance of accounting for greenness levels in future air pollution studies.

In addition to modifying the effects of air pollution, proximity to greenness may improve cardiovascular health by decreasing mental stress. Exposure to natural outdoor environments has been found to be associated with better mental health and could facilitate stress reduction (58), and neighborhood greenness is associated with lower levels of self-perceived stress and depression (59), particularly in older adults. In our work, we have found that higher levels of residential greenness are inversely associated with urinary levels of the stress hormone—epinephrine (8). Hence, it seems plausible that some of the effects of greenness on arterial stiffness may be mediated by a reduction in mental stress. Because mental stress is associated with increases in arterial stiffness (60), lower levels of mental stress in residents of greener neighborhoods could contribute to the lower levels of arterial stiffness in these participants. Therefore, further work is clearly warranted to delineate the relative contribution of mental stress and other psychosocial factors to the relationship between green spaces and cardiovascular health.

Our study had several limitations that should be considered. First, this is a cross-sectional study with a small sample size; thus, causal relationships cannot be inferred. Moreover, pulse wave analysis measurements were only conducted using a single temporal measurement, which may not accurately represent the arterial stiffness. However, our study was designed to determine the association between arterial stiffness and contemporaneous levels of air pollutants. Because the levels of air pollution change daily, arterial stiffness measured on different days could not be used to confirm the results of previous measurements (made at a different level of air pollution). Thus, pooling multiple measurements made on days with different levels of air pollution would “washout” the effects of air pollution. In addition to day-to-day variability, arterial stiffness also show diurnal variation (61), and therefore our measurements, made at different time of the day, could be confounded by circadian variability in arterial stiffness. However, stratification of the indices of arterial stiffness did not show any significant difference between measurements made early, in the middle or late in the day (Supplemental Table S3), suggesting that our results are unlikely to be confounded by circadian variability in arterial stiffness. Moreover, pulse wave analysis is a noninvasive method that provides indirect correlates of arterial stiffness using a transfer function to derive aortic pressure waveforms from radial pressure waveforms. This transfer function may result in measurement error of the true aortic stiffness measurements. Finally, exposure measurements relied on a diffuse monitoring network, which may not be representative of the participants’ individual exposure levels and could lead to exposure misclassification. Moreover, we measured NDVI around the participants’ residence, which only accounts for a proportion of each participants’ daily activity.

In summary, changes in arterial stiffness were positively associated with short-term exposure to PM_2.5, PM₁₀, and ozone and inversely associated with greenness, in our community-based study of at-risk participants. The association between pollution and arterial stiffness was attenuated in areas of high greenness. These findings indicate that living in green areas may be conducive for vascular health and that the salubrious effects on greenness may be attributable, in part, to attenuated exposure to air pollutants such as PM and ozone. Further studies are required to assess the relationship between greenness, air pollution, and CVD risk.

GRANTS

This research was supported in part by grants from the WellPoint Foundation and National Institute of Environmental Health Sciences Grants ES029846, ES019217, and ES023716.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

D.W.R., R.Y., D.J.C., N.D., R.J.K., A.P.D., S.N.R., and A.B. conceived and designed research; D.W.R. analyzed data; D.W.R., D.J.C., S.N.R., and A.B. interpreted results of experiments; D.W.R. prepared figures; D.W.R. drafted manuscript; D.W.R., R.Y., D.J.C., N.D., R.J.K., S.N.R., and A.B. edited and revised manuscript; D.W.R., R.Y., D.J.C., N.D., R.J.K., A.P.D., S.N.R., and A.B. approved final version of manuscript.