Residential Proximity to Major Roads, Exposure to Fine Particulate Matter and Coronary Artery Calcium: The Framingham Heart Study

Abstract

Objective

Long-term exposure to traffic and particulate matter air pollution is associated with a higher risk of cardiovascular disease, potentially via atherosclerosis promotion. Prior research on associations of traffic and particulate matter with coronary artery calcium Agatston score (CAC), an atherosclerosis correlate, has yielded inconsistent findings. Given this background, we assessed whether residential proximity to major roadway or fine particulate matter (PM_2.5) were associated with CAC in a Northeastern U.S. study.

Approach and Results

We measured CAC up to two times from 2002–2005 and 2008–2011 among Framingham Offspring or Third Generation Cohort participants. We assessed associations of residential distance to major roadway and residential PM_2.5 (2003 average; spatiotemporal model) with detectable CAC, using generalized estimating equation regression. We used linear mixed effects models to assess associations with log_e(CAC). We also assessed associations with CAC progression. Models were adjusted for demographic variables, socioeconomic position markers and time. Among 3,399 participants, 51% had CAC measured twice. CAC was detectable in 47% of observations. At first scan, mean age was 52.2 years (standard deviation 11.7); 51% male. There were no consistent associations with detectable CAC, continuous CAC or CAC progression. We observed heterogeneous associations of distance to major roadway with odds of detectable CAC by hypertensive status; interpretation of these findings is questionable.

Conclusions

Our findings add to prior work and support evidence against strong associations of traffic or PM_2.5 with the presence, extent or progression of CAC in a region with relatively low levels of and little variation in PM_2.5.

GRAPHIC ABSTRACT

Distance to road scaled to living at 75^th (416.9 m) vs 25^th (57.5 m) percentile from a major roadway; PM_2.5 scaled to 1.4 µg/m³ (IQR). Adjusted for demographic markers, socioeconomic position, time.

INTRODUCTION

Long-term exposure to particulate matter air pollution is associated with a higher risk of cardiovascular disease (CVD) morbidity and mortality^1–4. One pathway through which this could occur is by atherosclerosis promotion^1,5,6, with potential mechanisms including a systemic inflammatory and oxidative stress response, autonomic nervous system imbalance and, possibly, the transport of particulate matter or its constituents directly into arterial blood circulation.^1,5,6 Studies in susceptible animal models have found particulate matter exposure leads to atherosclerosis progression.^7,8 Epidemiological studies have also provided evidence of positive associations between ambient particles and traffic with atherosclerosis markers.^9–16

Positive associations between PM_2.5 and cardiovascular or all-cause mortality have been observed¹⁷ in regions with relatively low PM_2.5 levels, including New England.¹⁸ Studies in Massachusetts have found positive associations between traffic exposure and adverse cardiovascular outcomes^19,20 and between PM_2.5 and acute myocardial infarction.²¹ In a multi-city U.S. study, traffic exposure was associated with left ventricular mass and microvascular abnormalities.^22,23 PM_2.5 or traffic have been associated with impaired conduit artery and microvascular function among Framingham Heart Study participants living in the Northeastern U.S.^24,25 In a Boston-area study, black carbon, a correlate of traffic, was associated with carotid intima-media thickness (CIMT), a marker of subclinical atherosclerosis.¹¹

However, prior U.S. and German studies of residential proximity to a major roadway and particulate matter exposure with coronary artery calcium Agatston score (CAC), a marker of a later stage in the atherosclerotic disease process than CIMT, have yielded inconsistent results.^9,10,16 Given this background, we aimed to assess whether there were associations of these exposures with CAC among participants from the Framingham Heart Study living in the Northeastern U.S., a region with relatively low PM_2.5 levels. CAC provides a quantitative estimate of total coronary atheroma (both calcified and non-calcified plaque).²⁶ An independent predictor of coronary heart disease (CHD) and CVD^27,28, CAC has been used as a CVD prognostic tool.²⁶ In Framingham Heart Study participants, CAC improved discrimination and risk reclassification for major CHD beyond traditional risk factors.²⁸ We assessed associations of CAC measured up to two times during the periods 2002–2005 and 2008–2011 with residential distance to a major roadway and with exposure to spatially-resolved PM_2.5 at home address in the Framingham Offspring and Third Generation Cohorts. Residential distance to a major roadway, here defined as A1, A2 or A3 road (U.S. Census Features Class), is a surrogate of exposure to local traffic emissions. PM_2.5 is emitted by both local and regional pollution sources.

MATERIALS AND METHODS

Materials and Methods are available in the online-only Data Supplement.

RESULTS

Study Participants

Table 1 describes participant characteristics of the 5,118 observations from 3,399 participants. Average age of participants was 52.2 years and 59.0 years during the first and second rounds of multi-detector computed tomography (MDCT) scans, respectively. Overall, women contributed to 51% of observations and college graduates contributed to 45% of observations. Participants from the second round of MDCT scans were less likely than those from the first round to be current smokers (7 versus 13%) and more likely to report being on hypertension medication (35 versus 19%) or lipid medication (38 versus 14%).

Table 1.

Characteristics of Study Participants

Characteristics	mean ± SD or n (%)
	CT 2002–2005 (N=2,854)			CT 2008–2011 (N=2,266)
Age at MDCT Scan (years)	52.2	±	11.7	59.0	±	11.8
Male Sex, %	1,443		(51)	1,078		(48)
Offspring, %	1,123		(39)	1,054		(47)
Education
Some College	915		(32)	735		(32)
College Graduate	1,277		(45)	1,021		(45)
Median Census Value Owner-Occupied Housing	222,497	±	101,214	222,910	±	103,729
Current Smokers, %	370		(13)	167		(7)
Former Smokers, %	1,060		(37)	956		(42)
Pack-years
Current Smokers	29.8	±	13.7	32.7	±	15.8
Former Smokers	16.9	±	17.7	16.9	±	16.6
Alcohol (average drinks/week)	4.9	±	7.3	4.6	±	7.2
Physical Activity Index (dimensionless)	37.6	±	7.4	36.1	±	6.3
Menstrual periods stopped*, %	738		(52)	905		(76)
Diabetes History, %	146		(5)	193		(9)
Clinically apparent CVD at MDCT scan, %	154		(5)	116		(5)
Body Mass Index (kg/m2)	27.8	±	5.3	28.4	±	5.4
Hypertension medications, %	535		(19)	787		(35)
Systolic Blood Pressure (mm Hg)	122	±	16	122	±	16
Diastolic Blood Pressure (mm Hg)	76	±	9	75	±	9
Lipid medications, %	389		(14)	855		(38)
Triglycerides (mg/dL)	127	±	89	118	±	80
Total cholesterol	197.0	±	34.9	191.8	±	34.9
High-density cholesterol	53.7	±	16.6	59.0	±	18.0
10-year predicted risk of CVD†, %	0.05	±	0.07	0.07	±	0.09
MDCT Scan Results
CAC>0, %	1,284		(45)	1,143		(50)
CAC, among those with CAC>0‡	70.5		(278.5)	137.1		(419.4)

^*Among women.

^†American College of Cardiology/American Heart Association 2013 10-year predicted risk of atherosclerotic CVD; Median, interquartile range.

^‡Median, interquartile range.

Data calculated from 5,118 observations, from 3,399 participants with at least one CAC measurement.

Exposure Distributions

Table 2 summarizes the distributions of exposures. The median distance to the nearest major roadway, defined as A1, A2 or A3 road (U.S. Census Features Class), was 201 m and 23% of observations came from participants who lived within 50 m of a major roadway. Among 1,937 residential locations within 150 m of a major roadway, the largest roadway within 150 m was an A1 for 73, an A2 for 209, and an A3 for 1,655 locations. Median PM_2.5 in 2003 was 10.7 µg/m³, which is lower than the current U.S. Environmental Protection Agency annual PM_2.5 National Air Quality Standard of 12 µg/m³.

Table 2.

Distributions of Proximity to a Major Roadway, PM_2.5

Exposure	Median (IQR) or n [%]	Range (min, max)	Range (5th to 95th)	Range (25th to 75th)
Proximity to a Major Roadway (m)*	201 (359)	0.01–999.7	6.9–815.1	57.5–416.9
Total PM2.5 (µg/m3), 2003†	10.7 (1.4)	2.9–26.7	8.2–12.6	9.9–11.4
Total PM2.5 (µg/m3), 2003–2009‡	9.8 (1.1)	2.6–17.2	7.2–11.1	9.2–10.3
Residential Proximity in Categories*
<50 m	1,063 [23%]
50 to <200 m	1,212 [27%]
200 to <400 m	1,071 [23%]
400 to <1,000 m	1,218 [27%]

IQR is Interquartile Range;

^*Proximity to a major roadway analyses restricted to individuals living <1,000 m from a major road; 4,564 observations (554 observations, 11% lived ≥1,000 m).

^†PM_2.5 2003 data calculated from 5,118 observations.

^‡PM_2.5 2003–2009 calculated from 5,116 observations.

Associations with Odds of Detectable CAC and Average CAC

We found no associations of residential proximity to a major roadway or average residential PM_2.5 (2003 or 2003–2009) exposure with the odds of detectable CAC (Table 3). We also observed no associations of these exposures with average natural log-transformed CAC among those with detectable CAC (Table 3). Some point estimates were in the opposite direction than expected. For instance, compared with living 400 to <1,000 m from a major roadway, living <50 m from a major roadway was associated with a 5.8% lower CAC, though CIs were wide.

Table 3.

Adjusted Associations of Proximity to Major Roadway and PM_2.5 with CAC^*

	CAC>0†			Linear Mixed Effects (among those with CAC>0)‡
	Odds Ratio	95% CI		Percent Difference	95% CI
<50 m	0.95	(0.74,	1.22)	−5.8	(−21.4,	13.4)
50 to <200 m	0.94	(0.74,	1.18)	−10.2	(−24.9,	6.1)
200 to <400 m	0.90	(0.71,	1.14)	6.8	(−12.9,	26.2)
400 to <1,000 m	1.00
Log of Distance to a Major Road§	1.03	(0.93,	1.16)	7.2	(−2.6,	16.7)
2003 PM2.5 (µg/m3)∥	0.98	(0.90,	1.06)	−4.2	(−10.0,	3.8)
2003–2009 PM2.5 (µg/m3)∥	0.98	(0.89,	1.07)	0.4	(−6.5,	9.6)

^*Adjusted for covariates described in the online-only Data Supplement.

^†For binary outcome, distance to roadway: 4,564 observations (2,205 CAC>0); PM_2.5 2003: 5,118 observations (2,427 CAC>0); PM_2.5 2003–2009: 5,116 observations (2,426 CAC>0).

^‡Model log_e(CAC) among those with CAC>0. Distance to roadway: 2,205 observations; PM_2.5 2003: 2,427 observations; PM_2.5 2003–2009: 2,426 observations. Percentile bootstrapped CIs (n=1,000 clustered bootstrap samples).

^§Scaled to the difference between 75^th (416.9 m) vs the 25^th (57.5 m) percentile from a major road.

^∥Scaled to the 2003 PM_2.5 IQR (1.4 µg/m³).

Associations with Odds of Detectable CAC Progression and Average Annual Change in CAC

Among 1,719 participants with CAC measured during MDCT round 1 and round 2, 41% had detectable CAC progression. We observed no association between residential distance to a major roadway or PM_2.5 and the odds of detectable CAC progression (Table 4). We observed a weak association of residential proximity to a major roadway with average annual change in CAC: living further from a roadway was associated with a higher average annual change in CAC (Table 4). There was no association of PM_2.5 with average annual change in CAC (Table 4).

Table 4.

Adjusted Association of Proximity to a Major Roadway, PM_2.5 with Detectable CAC Progression, Change in CAC

	Detectable CAC Progression*			Associations with Change in CAC*
	OR	95% CI		Mean CAC change/year	95% CI
<50 m	0.92	(0.66,	1.29)	−3.6	(−9.3,	1.9)
50 to <200 m	0.78	(0.56,	1.08)	−2.0	(−8.3,	3.3)
200 to <400 m	0.90	(0.65,	1.25)	−1.1	(−7.6,	4.4)
400 to <1,000 m	1.00
Log Distance to Road†	1.06	(0.91,	1.22)	2.2	(0.1,	4.3)
2003 PM2.5 (µg/m3)‡	1.02	(0.91,	1.14)	−0.8	(−2.3,	0.6)
2003–2009 PM2.5 (µg/m3)§	1.06	(0.93,	1.20)	−0.8	(−2.4,	0.8)

^*Adjusted for covariates described in the online-only Data Supplement. Percentile bootstrapped CIs (n=1,000 clustered bootstrap samples) for change in CAC analyses.

^†Scaled to the difference between 75^th (416.9 m) vs 25^th (57.5 m) percentile from a major road. Detectable CAC progression: 1,536 observations (647 detectable progression). Change in CAC: 4,564 observations (1,536 two CAC measurements).

^‡Scaled to 2003 PM_2.5 IQR (1.4 µg/m³). 2003 PM_2.5 detectable CAC progression: 1,719 observations (711 detectable progression). 2003–2009 PM_2.5 detectable CAC progression: 1,718 observations (711 detectable progression). 2003 PM_2.5 change in CAC: 5,118 observations (1,719 two CAC measurements). 2003–2009 PM_2.5 change in CAC: 5,116 observations (1,718 two CAC measurements).

Sensitivity Analyses

We found evidence of non-linearity for the associations of PM_2.5 (2003 and 2003–2009) with natural log-transformed CAC—there was a suggested positive association at lower PM_2.5 levels and suggested negative association at higher PM_2.5 levels, though CIs were very wide (Supplemental Figure I).

When we only adjusted for age and sex, results were similar. In the main analyses, we adjusted for age and age² at scan, sex, body mass index, smoking status, pack-years, individual-level education, median census-tract value of owner-occupied housing units, cohort and time. We did not observe materially different results when we further adjusted for: physical activity index, alcohol intake, menopausal status, diabetes, anti-hypertensive medication, systolic blood pressure, diastolic blood pressure, lipid-lowering medication, total cholesterol, high-density lipoprotein cholesterol and triglycerides. Results were also similar when we adjusted for year of CT scan as a categorical variable.

Results were similar when we restricted to those free of CVD. We found no consistent pattern of heterogeneity of associations with the presence or extent of CAC by age, sex, cohort, 10-year risk of atherosclerotic CVD or smoking status (Supplemental Table I). We observed heterogeneity of the association of distance to a major roadway with odds of detectable CAC by hypertensive status or use of anti-hypertensive medications. Living closer to a major roadway was associated with higher odds of detectable CAC among those who had hypertension or used anti-hypertensive medications and lower odds of detectable CAC among those without hypertension or who did not use anti-hypertensive medications. However, we did not observe such patterns of association with the extent of CAC, nor with associations of PM_2.5 with CAC. Given the lack of consistent heterogeneity by hypertension or anti-hypertensive treatment, interpretation of these findings is questionable.

Results were similar when we assessed associations of living close to a major roadway (<150 m) with CAC. We did not observe an association of distance to nearest A1 or A2 road with the presence, extent, detectable progression or annual change of CAC. Results were similar when we ran analyses separately for MDCT round 1 and MDCT round 2.

Carrying out the CAC>0 analyses with a mixed logistic model yielded similar results, though with wider CIs. We observed no strong associations of distance to a major roadway or PM_2.5 with log_e(CAC+1) or with CAC greater than the 75 or 90^th age and sex-specific healthy referent cut-points.

DISCUSSION

In this study in a region with relatively low levels of and variation in PM_2.5, we found no consistent associations between residential distance to a major roadway or PM_2.5 with the presence or extent of CAC or with CAC progression. Sensitivity analyses yielded generally robust findings.

Prior studies of associations of these exposures with CAC have yielded somewhat inconsistent results. In the Heinz Nixdorf Recall Study, based in an industrial region of Germany, residential distance to a major road was associated with elevated CAC and higher continuous CAC.⁹ PM_2.5 was associated with CAC only among individuals who had not recently worked full time. In the US-based MESA, cross-sectional associations of PM_2.5 and thoracic particles (<10 µm) with CAC were weak (not statistically significant) or inconsistent.¹⁰ However, more recently, MESA has found PM_2.5 and nitrogen oxides, a marker of traffic-related air pollution, were positively associated with CAC progression.¹⁶

Our results add to prior research assessing associations of PM_2.5 and distance to a major roadway with coronary atherosclerosis. We carried out thorough analyses and did not find evidence of strong associations of these exposures with CAC. Unlike the Heinz Nixdorf Recall Study, we did not find that living closer to a major road was associated with more extensive CAC. These findings may be due to regional differences in pollution and population characteristics. Average PM_2.5 levels in our study region were generally much lower than in the Heinz Nixdorf Recall Study region.⁹ Diesel cars are more common in Germany than in the U.S. and diesel exhaust may be more harmful than gasoline exhaust. Additionally, compared with our study, participants in the Heinz Nixdorf Recall Study were older (mean age 60.2 years).⁹ On the other hand, our findings were more consistent with the cross-sectional results from MESA than with those from the Heinz Nixdorf Recall Study. Participants were also, on average, older in the cross-sectional analysis of MESA (mean 62.0 years)¹⁰ than in our study. Additionally, average PM_2.5 levels in the MESA cross-sectional study were generally higher than in our study region. At the different MESA sites¹⁰, mean annual 2001 PM_2.5 averages (in µg/m³) ranged from 12.82 (Minnesota) to 24.10 (California).

With repeated measures of CAC, we were able to assess for evidence of associations between distance to a major roadway and PM_2.5 exposures with CAC progression. In contrast to MESA¹⁶, we did not find evidence of strong positive associations of PM_2.5 or distance to a major roadway with CAC progression. Of note, there was a relatively short time between the first and second MDCT scans (average 6.1 years) in our study.

Studies have also assessed the associations of PM_2.5 or residential distance to a major roadway with other atherosclerosis surrogates. For instance, the Heinz Nixdorf Recall Study and MESA examined associations with calcification in the thoracic and abdominal aorta, respectively, which both predict incident CVD.^29,30 In the Heinz Nixdorf Recall Study, PM_2.5 was associated with more extensive thoracic aortic calcium.¹³ In MESA, there was a weak association of PM_2.5 with the presence but not extent of abdominal aortic calcium.³¹ Many studies have found associations between particulate air pollution and CIMT.^11,14,15 Particulate air pollution has been somewhat more consistently associated with CIMT than with CAC. This may be due in part to different ranges and composition of ambient particulate air pollution in various study regions. Additionally, CAC and CIMT are correlates of different aspects of subclinical atherosclerosis³²; CIMT represents an earlier stage in vascular injury. Relatively recent exposures may contribute more to earlier stages of disease than to progression to arterial calcification. It would be of interest to study the association of particulate air pollution with soft plaques in the coronary arteries.

Importantly, the lack of associations of residential proximity to a major road or residential estimates of PM_2.5 exposure with CAC in our study population does not mean that ambient air pollution does not cause atherosclerosis among people living in the Northeastern U.S. For example, our group has previously reported associations between traffic-related air pollution and CIMT in a Boston-area study¹¹. There are several pathways through which particulate matter exposure could lead to atherosclerosis.^1,5,6 First, particulate matter inhalation can lead to a pulmonary inflammatory and oxidative stress response, which can ‘spill-over’, yielding systemic oxidative stress and inflammation.^1,5,6 This may lead to atherosclerosis progression through vascular inflammation and impaired vascular function.^1,5 Second, particulate matter inhalation can activate the sympathetic nervous system^1,5, which may yield vasoconstriction, plaque instability and endothelial dysfunction. Third, there is some evidence that particulate matter or its constituents can directly transport into systemic circulation, yielding downstream effects such as coagulation, platelet function, vascular inflammation and atherosclerosis.^1,5,6

This study has several limitations. As the analyses are observational, there is potential for residual or unmeasured confounding. However, we have adjusted for many potential confounders, including both individual- and area-level socioeconomic position markers. Estimates of distance to a major roadway and PM_2.5 are subject to unavoidable measurement error. However, we do not expect this error to be related to the presence or extent of CAC. Additionally, we aimed to calculate distance to roadway from the actual house, which should be a better correlate of exposure at a person’s residence than would distance to the edge of the property. Importantly, the goal of our exposure assessment is to use residential location to correctly rank an individual’s exposure, not to assign correct absolute levels to each individual. In the main analyses, we assessed distance to nearest A1, A2 or A3 roadway and did not differentiate between types of major roadway (few participants lived close to an A1 or A2 roadway). We expect that traffic volume is, on average, generally highest on A1 roads (primary highways with limited access) and lowest on A3 roads (secondary and connecting roads) and could potentially range, on average, from less than 10,000 to more than 150,000 vehicles per day. In analyses looking at distance to A1 or A2 roadway only, results were similar. The association of categories of distance to a major roadway with log-transformed CAC may be difficult to interpret, so we also assessed associations of living close to a major roadway (<150 m versus ≥150 m). We also did not consider year-to-year variability in exposures, though adjusting for year of CT scan (as a categorical variable) did not change results. Additionally, using PM_2.5 from 2003–2009 instead of from 2003 yielded similar findings, suggesting results were not sensitive to PM_2.5 index period. As we do not have detailed long-term residential history, we were unable to study a long-term exposure window (e.g., 20 years). For a chronic disease process like atherosclerosis, exposure over a period of many years might have a greater impact on the disease process than relatively recent exposure.

Measuring CAC is only one of many approaches aimed at quantifying atherosclerosis. Additionally, the ability of CAC to predict future CHD may be most informative when combined with other risk factors.³³ However, prior work has shown a consistent association between CAC and risk of incident CHD and CVD. A meta-analysis found that compared with individuals with a CAC of 0, those with CAC>400 had 10 times the odds of incident CHD (95% CI: 3.1–34).²⁷ In the Framingham Heart Study, many cardiovascular risk factors, including age, sex and Framingham risk score, have been associated with CAC³⁴ and CAC is associated with incident CHD and CVD in this cohort.²⁸ As the participants in this study are predominantly Caucasian and of middle to upper middle class, these results might not be generalizable to populations with other characteristics.

The study also has several strengths. We used spatially resolved PM_2.5 exposures that benefit from spatial resolution of land use regression and spatiotemporal resolution of satellite data. We use two different measures of exposure to air pollution: distance to a major roadway and PM_2.5. While distance to a major roadway is correlated with exposure to local traffic-related exposures, PM_2.5 captures both local and regional sources of air pollution. Additionally, measuring CAC twice among some participants enabled us to conduct analyses assessing the associations of distance to a major roadway and PM_2.5 with CAC progression.

In conclusion, we observed no evidence that residing closer to a major roadway or having higher ambient residential PM_2.5 exposure was strongly associated with the presence, extent, or progression of CAC among individuals residing in a region with relatively low levels of and little variation in PM_2.5 levels. These findings add to the totality of evidence on the association of traffic and PM_2.5 exposure and coronary atherosclerosis in humans.

Highlights.

• Long-term exposure to particulate air pollution and traffic are associated with a higher risk of cardiovascular disease, possibly via promotion of atherosclerosis.

• Studies assessing associations of particulate air pollution and traffic with CAC, a correlate of subclinical atherosclerosis, have had inconsistent findings.

• We found no strong associations between residential proximity to a major roadway or fine particulate matter (PM_2.5) air pollution exposure and CAC among Framingham Heart Study participants living in the Northeastern U.S., a region with low levels of and low variability in PM_2.5.

Nonstandard abbreviations

CAC

Coronary artery calcium Agatston score

CIMT

Carotid intima-media thickness

MDCT

Multi-detector computed tomography