Ambient Air Pollution Is Associated With the Severity of Coronary Atherosclerosis and Incident Myocardial Infarction in Patients Undergoing Elective Cardiac Evaluation

The Cleveland Clinic GeneBank study is a single‐site sample repository generated from consecutive patients undergoing elective diagnostic coronary angiography or elective cardiac computed tomographic angiography with extensive clinical and laboratory characterization and longitudinal observation. Participant recruitment occurred between 2001 and 2007, and all patients provided written informed consent prior to being enrolled. Ethnicity was self‐reported, and information regarding demographics, medical history, and medication use was obtained by patient interviews and confirmed by chart reviews at baseline enrollment. Assessment of functional capacity, as a measure of physical activity, was estimated at enrollment based on the self‐administered Duke Activity Status Index questionnaire.¹² All clinical outcome data were verified by source documentation. At baseline, CAD was defined as adjudicated diagnoses of stable or unstable angina, MI (adjudicated definition based on defined electrocardiographic changes or elevated cardiac enzymes), angiographic evidence of ≥50% stenosis in ≥1 major epicardial vessel, and/or a history of known CAD (documented MI, CAD, or history of revascularization). Coronary atherosclerosis severity at baseline was defined as the number of major epicardial vessels with ≥50% stenosis. All quantitative determinations of coronary stenosis were adjudicated by a cardiologist blinded to participant identity. Prospective cardiovascular risk was assessed by the incidence of all‐cause mortality or nonfatal MI or stroke during 3 years of follow‐up from the time of enrollment. Participants were contacted annually either directly in person, by telephone follow‐up, or by other means. In the case of a participant being deceased, a preidentified and pre–agreed upon proxy was contacted. Nonfatal events were defined as MI or stroke in patients who survived at least 48 hours following the onset of symptoms, and all adjudicated outcomes were confirmed using source documentation. The GeneBank Study has been used previously for discovery and replication of novel genes and risk factors for CAD.^{13, 14, 15, 16, 17, 18, 19, 20} The present study was approved by the institutional review boards of the Cleveland Clinic and the University of Southern California Keck School of Medicine.

Samples were collected from overnight fasted participants on the day of elective cardiac catheterization. Plasma aliquots were isolated from whole blood collected into EDTA tubes, maintained at 0 to 4°C immediately following phlebotomy, processed within 4 hours of blood draw, and stored at −80°C until analysis. Plasma levels of total cholesterol, low‐ and high‐density lipoprotein cholesterol, triglycerides, and high‐sensitivity C‐reactive protein were measured on the Architect platform (Abbott Diagnostics).

Daily concentrations of PM_2.5 and NO₂ in the United States from 1998 through 2010 were downloaded from the US Environmental Protection Agency's (EPA) Air Quality System (AQS) database (https://www.epa.gov/aqs). This national database contains hourly and daily outdoor air pollution concentration data back to the late 1970s, and the network has remained fairly stable since 2000 for PM_2.5 and NO₂. Data for PM_2.5 and NO₂ were primarily limited to those collected with Federal Reference Method samplers and Federal Equivalent Method monitors. Non–Federal Reference Method PM_2.5 and NO₂ data were used only when Federal Reference or Equivalent Method measurement data were not available for a location. Because ozone was not routinely monitored across the year in many locations, it was not included in this study. Automated quality control checks on the concentration ranges and persistence were applied to the AQS data. Because the national air‐monitoring networks began measuring PM_2.5 and NO₂ in 1999, few data exist prior to that date. The hourly PM_2.5 and NO₂ data were averaged into standard daily exposure metrics, and monthly averages were calculated from the daily average pollutant data. A 75% data completeness criterion was used in determining monthly averages. Because the historical daily PM_2.5 measurements were often made once every third or sixth day rather than daily, the completeness criterion was applied based on the expected completeness for a 1‐in‐6‐day sampling schedule. Monthly air‐quality exposure values were spatially interpolated from the air‐quality monitoring locations of the residential ZIP code coordinates (based on geographic centroid) of each participant at the time of enrollment into the GeneBank study. The station‐specific monthly air‐quality data were spatially interpolated using inverse distance‐squared weighting. The data from up to 4 air‐quality measurement stations were included in each interpolation. Because of the regional nature of PM_2.5 and NO₂ concentrations, a maximum interpolation radius of 50 km was used; however, when a residence was located within 5 km of ≥1 station with valid observations, the interpolation was based solely on the concentrations from the stations within 5 km. The same 75% completeness criteria were applied to the estimates of average exposures for each exposure period. Estimated levels of PM_2.5 and NO₂ were based on 46 and 4 monitoring sites, respectively.

The National Land Cover Database (NLCD) for 2011 was used to assess the extent of industrial development near each participant's reported residence.²¹ The NLCD provides many categories of land cover at the native 30‐m resolution of the Landsat Thematic Mapper. Specifically, we used the land cover class (class 24: developed, high intensity) that combines commercial, industrial, high‐density residential, and transportation land use and is characterized by 80% to 100% impervious surfaces. The percentage of industrial‐like land cover in the postal ZIP code area, defined by the 2010 5‐digit ZIP code boundaries, of each participant's residence was computed in a geographic information system (ArcGIS version 10.3; Esri). This measure was significantly correlated with levels before and after enrollment of PM_2.5 (r=0.25 and r=0.29, respectively; P<0.0001) and NO₂ (r=0.22 and r=0.20, respectively; P<0.0001).

Primary outcomes included the degree of coronary atherosclerosis severity at baseline (0, 1–2, or ≥3 epicardial vessels with ≥50% stenosis) and prospective incident events (nonfatal MI, stroke, all‐cause mortality) over 3 years of follow‐up. Participants who experienced an event within 14 days of enrollment were excluded from the analyses to omit acute events during the initial baseline period. Air pollution variables for each participant included estimated levels of exposure to PM_2.5 (in μg/m³) and NO₂ (in parts per billion) during the 36 months prior to baseline enrollment and the 36 months after enrollment for cross‐sectional and prospective analyses, respectively. Multinomial logistic regression was applied to evaluate the effect of air pollution on coronary atherosclerosis severity among participants with CAD at baseline. Cox proportional hazards models were used to estimate the effect of ambient air pollution on prospective events among all participants and in a subset with CAD at baseline. To test for confounding after adjustment for a priori covariates (age, sex, smoking, and education), we performed sensitivity analyses with the following variables: obesity (body mass index <30 versus ≥30), statin therapy use (yes or no), high‐sensitivity C‐reactive protein, and coronary atherosclerosis severity (0, 1–2 or ≥3 epicardial vessels with ≥50% stenosis). In addition, we tested whether cardiovascular risk factors modified the effects of air pollution on CAD outcomes by further adjusting for Framingham Adult Treatment Panel III (ATP III) risk score (including total and high‐ and low‐density lipoprotein cholesterol levels, presence of atherosclerosis, family history of premature coronary heart disease, smoking, hypertension, diabetes mellitus, and age). None of the above potential covariates changed the estimates considerably (>10%); therefore, the final model included the a priori selected covariates of age, sex, current smoking (yes or no), and education level (college or higher, high school, less than high school) as adjusting variables. To assess potential residual confounding, physical activity and commercial/industrial land use were included in the models to test their influence on the effect estimates. Multipollutant models were further adjusted for the other respective copollutant. Additional analyses for association of PM_2.5 levels with incident MI were performed after stratifying by the presence of coronary atherosclerosis (≥1 epicardial vessel with ≥50% stenosis) and/or statin therapy use and by median age (≥64 years), sex, education level, current smoking, or obesity (body mass index ≥30). Adjusted hazard ratios or odds ratios with 95% CIs are reported with 2‐sided P values. Interaction P values were obtained from likelihood ratio tests. All analyses were performed using SAS 9.3 (SAS Institute Inc).

The clinical characteristics of the GeneBank participants in this study are described in Table 1 and Table S1. To avoid the potential for referral bias, only those participants whose residential ZIP codes were reported as being in Ohio (n=6575) were included. As expected for a patient population undergoing elective coronary angiography for clinical evaluation, the majority of participants at enrollment were male, had prevalent CAD, and were using statin therapy. In addition, a significant fraction of participants were obese or had diabetes, and most had attained at least a high school level of education (Table 1 and Table S1).

Air pollution variables included average estimated daily exposure levels of PM_2.5 and NO₂ during the 36 months prior to enrollment and the 36 months after enrollment for cross‐sectional and prospective analyses, respectively. As shown in Figure 1, GeneBank participants who reported their residential ZIP codes as being located in Ohio were clustered in metropolitan regions, particularly in the area surrounding Cleveland. Estimated exposure levels for PM_2.5 in the 36 months before and after enrollment were available for ≈6100 Ohio residents and ranged from 10 to 21 μg/m³, whereas NO₂ levels were available for ≈4600 participants and ranged from 5 to 23 parts per billion (Figure 2). Exposure levels of each pollutant during the 36 months before and after enrollment were strongly correlated with each other, with comparable means, whereas PM_2.5 and NO₂ levels were only weakly correlated with each other regardless of exposure period (Figure S1). Based on data from the EPA, PM_2.5 levels in Ohio during the pre‐ and postenrollment periods (1998–2010) were comparable to US national and regional trends, whereas NO₂ levels were lower. This may have been caused by the small number of monitoring sites for NO₂, which led to fewer participants receiving assignments for this pollutant compared with PM_2.5 and limiting exposure contrast.

We determined the association of air pollution levels before study entry with the extent of coronary atherosclerosis among patients with CAD, defined based on angiographic evidence at baseline or a positive history of CAD. After adjustment for age, sex, education level, and smoking, a 2‐SD (2.2‐μg/m³) increase in exposure to PM_2.5 over the 36 months preceding enrollment was associated with significantly increased likelihood (odds ratio 1.43, 95% CI 1.11–1.83; P=0.005) of having mild coronary disease, defined as 1 to 2 vessels with ≥50% stenosis, compared with patients with a history of CAD but no vessels with ≥50% stenosis at baseline (Table 2). The association of PM_2.5 with severe coronary atherosclerosis, defined as ≥3 vessels with ≥50% stenosis, was even more pronounced (odds ratio 1.63, 95% CI 1.26–2.11; P<0.001) (Table 2). A test of heterogeneity demonstrated that the effect of PM_2.5 on severe coronary atherosclerosis was significantly different from the effect on mild disease (heterogeneity P=0.03) (Table 2). Additional adjustment for the Framingham ATP III risk score, which captures several other CVD risk factors, did not appreciably change the effect estimates (Table 2). The magnitude and significance of the association of PM_2.5 with the extent of coronary atherosclerosis were also not markedly affected after further adjustment for additional potential covariates (Table S2). In comparison, increased NO₂ levels (2 SD; 4.1 parts per billion) were not associated with the likelihood of having mild or severe atherosclerotic disease (Table 3).

We next investigated whether air pollution levels were associated with prospective risk of incident MI, stroke, and all‐cause mortality among all GeneBank participants. A 2‐SD increase in PM_2.5 levels over 36 months of follow‐up after angiographic evaluation was associated with significantly increased risk of MI (hazard ratio 1.33, 95% CI 1.02–1.73; P=0.03) (Table 4). The effect estimates for the association of PM_2.5 and MI risk were not substantially changed in models that adjusted for Framingham ATP III risk score (Table 4) and other potential covariates (Table S3) or in a multipollutant model adjusting for NO₂ levels (Table S4). Restricting these analyses to only patients with angiographically determined CAD at baseline yielded similar effect estimates (Table S3). To address the potential for referral bias further, we also carried out an analysis that included only participants from the greater Cleveland metropolitan area (n=3437). The effect estimate obtained from this subanalysis with ≈2400 fewer participants (hazard ratio 1.12; 95% CI 0.96–1.30; P=0.15) was comparable and directionally consistent with the analysis that included all Ohio residents (Table 4). By comparison, there were no associations between PM_2.5 or NO₂ levels with risk of all‐cause mortality (Tables 4 and 5), whereas a 2‐SD increase in NO₂ levels was associated with elevated risk of stroke (hazard ratio 1.72, 95% CI 1.03–2.87; P=0.04) (Table 5); however, this association was no longer significant after adjustment for Framingham ATP III risk score (Table 5). It is possible that the lack or attenuation of significance in some of these analyses may have been due to reduced sample size. There were, for example, fewer numbers of participants for whom NO₂ levels or complete data on the full set of covariates were available (Tables S3 and S4).

We next performed stratified analyses and formal tests of interaction to determine whether the effect of PM_2.5 on incident MI was modulated by coronary atherosclerosis at baseline or by the use of statin therapy. Stratifying the analyses by participants who had coronary atherosclerosis at baseline, who used statin medications, or both did not reveal statistical evidence that the effects of PM_2.5 on MI risk were modulated by these factors (Table S5). There was also no significant statistical evidence that risk of MI was modulated by 3‐way interactions among PM_2.5, coronary atherosclerosis, and statin therapy use (3‐way interaction P=0.33). Last, we investigated whether any of the effects of PM_2.5 on risk of MI were modified by a priori covariates and performed analyses stratified by median age, sex, education level, smoking, coronary atherosclerosis severity, and obesity (body mass index ≥30), which also included formal tests of interaction. There were no significant effect modifications among these covariates, PM_2.5, and risk of MI (Table S6).