Synthesis of Harvard Environmental Protection Agency (EPA) Center studies on traffic-related particulate pollution and cardiovascular outcomes in the Greater Boston Area

ABSTRACT

The association between particulate pollution and cardiovascular morbidity and mortality is well established. While the cardiovascular effects of nationally regulated criteria pollutants (e.g., fine particulate matter [PM_2.5] and nitrogen dioxide) have been well documented, there are fewer studies on particulate pollutants that are more specific for traffic, such as black carbon (BC) and particle number (PN). In this paper, we synthesized studies conducted in the Greater Boston Area on cardiovascular health effects of traffic exposure, specifically defined by BC or PN exposure or proximity to major roadways. Large cohort studies demonstrate that exposure to traffic-related particles adversely affect cardiac autonomic function, increase systemic cytokine-mediated inflammation and pro-thrombotic activity, and elevate the risk of hypertension and ischemic stroke. Key patterns emerged when directly comparing studies with overlapping exposure metrics and population cohorts. Most notably, cardiovascular risk estimates of PN and BC exposures were larger in magnitude or more often statistically significant compared to those of PM_2.5 exposures. Across multiple exposure metrics (e.g., short-term vs. long-term; observed vs. modeled) and different population cohorts (e.g., elderly, individuals with co-morbidities, young healthy individuals), there is compelling evidence that BC and PN represent traffic-related particles that are especially harmful to cardiovascular health. Further research is needed to validate these findings in other geographic locations, characterize exposure errors associated with using monitored and modeled traffic pollutant levels, and elucidate pathophysiological mechanisms underlying the cardiovascular effects of traffic-related particulate pollutants.

Implications: Traffic emissions are an important source of particles harmful to cardiovascular health. Traffic-related particles, specifically BC and PN, adversely affect cardiac autonomic function, increase systemic inflammation and thrombotic activity, elevate BP, and increase the risk of ischemic stroke. There is evidence that BC and PN are associated with greater cardiovascular risk compared to PM_2.5. Further research is needed to elucidate other health effects of traffic-related particles and assess the feasibility of regulating BC and PN or their regional and local sources.

Introduction

The health effects of air pollution exposure range widely and include adverse pulmonary, cardiovascular, neonatal, and neurological outcomes (Dockery et al., 1993; Fleisch et al., 2015; Franklin, Zeka, and Schwartz, 2007; Kioumourtzoglou et al., 2016; Pope et al., 2002; Rice et al., 2015; Wilker et al., 2015). Since 1999, investigators at the Harvard Environmental Protection Agency (EPA) Center have conducted research to characterize air pollution exposures and sources, assess health effects at the molecular and physiological level, and identify factors that modify these health effects. In the past 15 years, traffic emissions have been identified as an important source of particle exposure that is associated with many adverse cardiovascular outcomes (Brook et al. 2010).

Traffic is a major source of both gaseous and particulate pollutants. Traffic-related particles with an aerodynamic diameter size d_a ≤ 2.5 μm (fine particles, PM_2.5) are of significant public health concern, as they penetrate to the deep alveolar regions of the lungs and have been linked to numerous adverse health effects (Dockery et al. 1993; Hoek et al. 2002; Künzli et al. 2000; Laden et al. 2006; Lepeule et al. 2012; Seaton et al. 1995). As such, PM_2.5 is among the criteria pollutants regulated by the U.S. EPA. Fine particles are formed as a result of fossil fuel combustion by motor vehicles and stationary sources such as power plants. As black carbon (BC) is a combustion by-product, BC may be a more precise proxy for traffic-related particles than PM_2.5 (Laden et al. 2000a).

To assess the health effects of traffic pollutants, recent Harvard EPA Center studies have primarily utilized BC, particle number (PN), and proximity to major roadways as markers of traffic exposure. PN represents mostly particles with an aerodynamic diameter of less than 100 nm, or ultrafine particles (UFPs), which are predominantly freshly generated combustion particles. In comparison, BC particles are a mix of freshly generated UFPs (d_a<0.1 μm) and aged traffic particles (mostly in the accumulation mode; 0.1< d_a<1.0 μm) in the Boston metropolitan area (Kang, Koutrakis, and Suh 2010). BC is a marker of local and regional traffic pollutants and measured as the light-absorbing component of particulate matter. Distance to major roadways is a proxy for long-term exposure to traffic-related air pollution as well as acute exposures, given that individuals with high long-term residential traffic exposures are also more likely to have high acute exposures. In addition, distance to roadways has been shown to be a good proxy for outdoor concentrations of traffic-related air pollutants (Brauer et al. 2003; Hochadel et al. 2006; Hoek et al. 2001).

Investigators outside of the Harvard EPA Center have also assessed the effect of traffic-related pollution on cardiovascular health. However, many of these studies use less specific measures of traffic pollution, including PM₁₀, ozone, carbon monoxide, nitrogen oxides, and sulfur dioxide (Bilenko et al. 2015; Foraster et al. 2014; Korek et al. 2015; Lanki et al. 2015; Shields et al. 2013; Sørensen et al. 2012). While these pollutants are of significant concern for public health, non-traffic sources also contribute substantially to these pollutants. In comparison, BC, PN, and proximity to major roadways represent more specific proxies for traffic particles. As such, we focus on comparing select studies that utilize these specific markers of traffic exposure in order to highlight salient findings on the effect of traffic-related particulate pollutants on cardiovascular health.

Studies assessing source-specific effects of particulate pollutants have demonstrated that traffic-related particles are associated with cardiovascular morbidity and mortality (Janssen et al. 2002; Laden et al. 2000b; Lanki et al. 2006). In fact, particles from traffic emissions were shown to be more strongly associated with cardiovascular mortality than particles from coal combustion in six U.S. cities (Laden et al. 2000b). In addition, the effect of PM₁₀ exposure on the risk of cardiovascular hospital admissions has been found to vary with the proportion of traffic particles (Janssen et al. 2002). Therefore, it is important to elucidate the effects of primary traffic particles, such as BC and PN, on cardiovascular health.

Cardiovascular morbidity and mortality have been consistently associated with particulate pollutant exposure (Brook 2008; Brook et al. 2010; Miller et al. 2007; Pope et al. 2004). Mechanisms underlying these associations include systemic inflammation, oxidative stress, hypercoagulability, vascular dysfunction, atherosclerosis, and autonomic dysfunction. Studies from the Harvard EPA Center have demonstrated the impact of traffic-related air pollutants on these physiologic measures and associated cardiovascular outcomes. A broad literature search reveals that a significant proportion of large cohort studies utilizing traffic-specific markers such as BC and PN have been conducted by Harvard investigators. In this paper, we synthesize findings from large New England-based epidemiological studies that investigated the impact of traffic-related particle exposures on cardiovascular mortality, hypertension, ischemic stroke, heart rate variability, and molecular markers of endothelial dysfunction, inflammation, and thrombosis.

Cardiovascular mortality

Long-term exposure to particulate air pollution increases risk of cardiovascular mortality in the general population (Di et al. 2017; Dockery et al. 1993; Miller et al. 2007; Pope et al. 2002). Residential proximity to a major roadway has been utilized as a surrogate for long-term traffic exposure to demonstrate associations with all-cause mortality (Finkelstein, Jerrett, and Sears 2004; Hoek et al. 2002), cardiopulmonary mortality (Gehring et al. 2006; Hoek et al. 2002), stroke mortality (Maheswaran and Elliott 2003), sudden cardiac death (Hart et al. 2014), and fatal coronary heart disease (Hart et al. 2014). In addition, living near a major road has been associated with myocardial infarction (Hart et al. 2013), hypertension (Kingsley et al. 2015), impaired conduit artery and microvascular function (Wilker et al. 2014), vascular end-organ damage (Van Hee et al. 2009), atherosclerosis (Hoffmann et al. 2007), and elevated C-reactive protein (CRP) levels (Lanki et al. 2015).

Long-term traffic exposures are more strongly associated with fatal cardiovascular outcomes and post-discharge mortality compared to nonfatal events (Hart et al. 2014; Rosenlund et al. 2006, 2008; von Klot et al. 2009). The link between particulate pollution and cardiac death may be explained by cardiac electrical instability (Zanobetti et al. 2009), as cardiac repolarization abnormalities may increase the risk of ventricular arrhythmia and sudden death. In addition, PM_2.5 and BC exposures have been associated with ventricular ectopy (Zanobetti et al. 2014) and ventricular fibrillation (Dockery et al. 2005), which both increase the risk of cardiovascular mortality.

The Harvard EPA Center investigators were among the first to assess the mortality risks of residential proximity to high-traffic roadways for survivors of heart failure (Medina-Ramón et al. 2008), ischemic stroke (Wilker et al. 2013), and acute myocardial infarction (AMI) (Rosenbloom et al. 2012). These studies reported high post-discharge mortality rates of 76% (1,055 out of 1,389 within 5 years) for heart failure, 56% (950 out of 1,683 within a median of 4.6 years) for ischemic stroke, and 30% (1,071 out of 3,547 within 10 years) for AMI. In all three studies, living closer to a high-traffic road was associated with higher post-discharge mortality risk (Figure 1). Living within 100 m of a major road was associated with a 30% (95% confidence interval [CI]: 13%, 49%), 20% (95% CI: 1%, 43%), and 27% (95% CI: 1%, 60%) increased risk of mortality after acute heart failure, ischemic stroke, and AMI, respectively. In contrast, mortality risks among individuals living within greater distance-to-road categories (e.g., 100–200 m or 200–400 m) were not statistically significant. Of note, these studies differed in follow-up time, reference distance group, and definition of high-traffic road as summarized in Table 1.

Table 1. Description of roadway proximity studies assessing cardiovascular mortality risk.

Study	Cohort	Exposure	Reference group	Follow-up period
Medina-Ramón et al. (2008)	1,389 acute HF patients in Worcester, MA	US Census feature class A1 or public bus route	>100 m of highway or >50 m of bus route	5 years
Wilker et al. (2013)	1,683 ischemic stroke pts at BIDMC	Distance to the nearest high-traffic roadway (>10,000 vehicles/day)	>400 m	4.6 years (median)
Rosenbloom et al. (2012)	3,547 post-AMI patients from 64 US hospitals (82% in New England)	US Census feature class A1 or A2	>1,000 m	10 years

Figure 1. Post-hospitalization mortality risk of roadway proximity among cardiovascular patients. Risk estimates are derived from Medina-Ramón et al. (2008) (post-heart failure), Wilker et al. (2013) (post-ischemic stroke), and Rosenbloom et al. (2012) (post-acute MI).

Roadway proximity is representative of both chronic and acute traffic pollutant exposures, as individuals with high chronic residential exposures to traffic-related pollutants are more likely to also have high acute exposures. However, there are important limitations of roadway proximity as a surrogate for traffic pollutant exposure. First, it is not possible to disentangle the impact of other aspects of near-roadway exposure, such as traffic noise, and estimate their relative contribution to cardiovascular risk. Second, without traffic volume data, it is assumed that exposures are similar within a given roadway classification, leading to non-differential exposure classification errors and underestimation of the associated risk. Third, socioeconomic status is a potential confounder given its association with both roadway proximity as well as survival after a serious cardiac event (Kaplan and Keil 1993). Lastly, additional exposure misclassification occurs without information on the amount of time spent at home and other home characteristics (e.g., soundproofing, topography, wind speed, and orientation relative to prevailing winds and to roadways). Despite these limitations, distance to roadway has been shown to be a good proxy for traffic-related air pollutant concentrations (Hoek et al. 2001; Janssen et al. 2002). A recent analysis of speciated PM_2.5 data from over 650 indoor samples from 340 homes in the Greater Boston Area illustrated that indoor traffic-related particles increased with roadway proximity, demonstrating roadway proximity as a good proxy for personal exposure to traffic pollutants (Huang et al. 2018).

In summary, living closer to a major roadway is associated with an increased risk of mortality in patients with underlying cardiovascular disease. There are several mechanisms mediating the association between traffic pollutant exposure and mortality among cardiovascular patients, including cardiac autonomic dysfunction, atherosclerosis, inflammation, and oxidative stress. The following sections discuss the effects of traffic exposure on these cardiovascular outcomes in greater detail.

Heart rate variability

The relationship between air pollution exposure and adverse cardiovascular effects is thought to be mediated in part by autonomic cardiac dysfunction. Heart rate variability (HRV) is under the control of the autonomic nervous system and a reduction in HRV has been associated with myocardial infarction (Tsuji et al. 1996), ischemic events (Kop et al. 2001), atrial fibrillation (Bettoni and Zimmermann 2002), and sudden death (Odemuyiwa et al. 1991). Many studies have demonstrated the association between air pollution exposure and reduced HRV (Adam et al. 2014; Baja et al. 2013; Gold et al. 2000; Park et al. 2007, 2005; Pieters et al. 2012; Pope et al. 1999). Certain individuals appear to have greater susceptibility, including the elderly (Schwartz et al. 2005), those with pre-existing cardiovascular disease or diabetes (Zanobetti et al. 2010), and those with reduced anti-oxidative defenses (Chahine et al. 2007).

HRV is a physiological measure for cardiac autonomic tone that describes the variation of the time interval between heart beats, and there are a number of HRV indices that have been utilized by air pollution epidemiologists. HRV indices can be broadly divided into those that primarily indicate parasympathetic activity (e.g., high frequency [HF], root mean squared differences between adjacent RR intervals [r-MSSD], proportion of adjacent normal-to-normal heartbeat intervals differing by more than 50 ms [PNN50]), and those that predominantly reflect sympathetic activity (e.g., total power [TPow], standard deviation of normal-to-normal intervals [SDNN], low frequency [LF]). The predominance of sympathetic and/or diminished parasympathetic activity is believed to drive the association between HRV and adverse cardiovascular outcomes.

Harvard EPA Center studies have demonstrated a greater reduction in HRV associated with BC exposure compared to PM_2.5 exposure among cardiovascular patients and elderly individuals. Zanobetti et al. (2010) studied the effect of PM_2.5 and BC exposure on HRV in 46 patients who were recently hospitalized in Boston for coronary artery disease, and lived within 40 km of the Harvard central monitoring site (Figure 2). This study found that decreases in r-MSSD and HF–-measures of parasympathetic tone–-were associated with both PM_2.5 and BC for exposure windows ranging from the current hour to the 5-day moving average. Notably, the effect sizes were larger for BC compared to PM_2.5 exposure: a decrease in HF by 16.4% (95% CI: −20.7%, −11.8%) was associated with an interquartile range (IQR) increase in 5-day average BC, compared to a decrease of 6.9% (95% CI: −10.5%, −3.1%) for an IQR increase in 5-day average PM_2.5. A panel study of 28 elderly Boston residents by Schwartz et al. (2005) also reported that BC exposure had the largest association with HRV reduction among PM_2.5 and other air pollutants (including secondary PM, ozone [O₃], sulfur dioxide [SO₂], nitrogen dioxide [NO₂], and carbon monoxide [CO]). Figure 2 illustrates that associations between HRV reduction and BC were more often statistically significant and larger in magnitude than those with PM_2.5 in both Zanobetti et al. (2010) and Schwartz et al. (2005).

Figure 2. Effect of traffic exposure on heart rate variability (HRV). Risk estimates reported per interquartile range (IQR) increase in PM_2.5 or BC.

Harvard investigators also conducted a personal exposure study of 44 seniors in suburban St. Louis, MO and consistently found larger associations for BC compared to PM_2.5 across all the HRV measures, which included SDNN, r-MSSD, PNN50, HF, and others (Adar et al. 2007). Together, these studies suggest that BC exposures may have greater effects on reducing HRV than PM_2.5 exposure. Of note, in areas where coal-burning non-traffic sources are the predominant source of air pollution (e.g., Steubenville, OH), PM_2.5 and sulfate (a non-traffic coal burning PM component) were equally strongly associated with HRV, reflecting important cardiovascular risks of non-traffic sources of air pollutant exposure (Luttmann-Gibson et al. 2014, 2006).

In contrast to the aforementioned studies, variable results have been reported in Harvard EPA Center studies on the larger, Normative Aging Study (NAS) cohort (Bell, Rose, and Damon 1966, 1972). Short-term exposure to BC was not associated with any of the HRV measures in a cross-sectional study (Park et al. 2005), but was associated with an increase in LF/HF (a measure of sympathetic tone) in a longitudinal analysis (Baja et al. 2013). Another study utilized residential exposures derived from spatiotemporal models and found an association between medium-term (28- and 56-day moving average) BC exposure and a reduced HF, but not with short-term BC exposure (Mordukhovich et al. 2015). Therefore, differences in study design (e.g., cross-sectional vs. longitudinal analysis) and exposure characterization (e.g., central monitor vs. spatiotemporal modeling, short- vs. medium-term) may explain variability in results within the same cohort.

Many studies outside of the Harvard EPA Center have illustrated the adverse effects of BC exposure on HRV not only among individuals with cardiovascular disease (Huang et al. 2012), diabetes (Sun et al. 2015), high body mass index (BMI ≥25 kg/m²) (Dong et al. 2018), and chronic obstructive pulmonary disease (COPD) (Pan et al. 2018b), but also among healthy individuals (Cole-Hunter et al. 2015; Pan et al. 2018a). There is also evidence of the adverse effects of increased PN levels on HRV among diabetic patients (Sun et al. 2015) as well as healthy individuals (Weichenthal et al. 2011). Therefore, traffic-related pollutant exposure is associated with cardiac autonomic dysfunction among not only individuals with existing co-morbidities but also healthy adults.

In summary, exposures to traffic-related particles adversely affect cardiac autonomic function. Short-, medium-, and long-term exposures to PM_2.5 and BC are associated with decreased HRV, particularly those that represent parasympathetic tone (e.g., HF, rMSSD, and SDNN). As reduced HRV is a known predictor of poor cardiac outcomes, it is important to identify traffic exposure risks among susceptible populations.

Inflammatory response

Inhalation of particulate pollutants elicits a systemic inflammatory response that is initiated by cytokines released by lung cells engulfing particles that deposit on the lung surface. These cytokines stimulate the bone marrow to produce white blood cells and the liver to produce acute-phase proteins (e.g., CRP, fibrinogen) and coagulation factors. Inflammatory responses promote thrombosis and contribute to and accelerate the progression of atherosclerosis (Libby, Ridker, and Hansson 2009; Suwa and Hogg 2002), subsequently increasing the risk of complications such as myocardial infarction and stroke. As such, many studies have examined the effect of traffic exposure on inflammatory markers known to be risk factors for cardiovascular disease.

Fibrinogen

As a thrombogenic factor, fibrinogen is an important determinant of blood viscosity and an independent risk factor for cardiovascular disease (Lip 1995). Harvard investigators have found associations between BC and PN exposure and increased fibrinogen levels in the NAS cohort (Table 2, Figure 3) (Bind et al. 2012, 2014; Zeka et al. 2006). In contrast, associations between PM_2.5 exposures and fibrinogen were mostly non-significant. Bind et al. (2014) found greater increases in fibrinogen among individuals with genetic susceptibility to oxidative stress, endothelial dysfunction, and metal processing dysfunction. These increases were associated with exposures to PN and BC but not PM_2.5. Conversely, outside of the Harvard EPA Center, a Multi-Ethnic Study of Atherosclerosis cohort study of over 6,800 individuals from 6 U.S. cities (Baltimore, MD; Chicago, IL; Winston-Salem, NC; Los Angeles, CA; New York, NY; and St. Paul, MN) have reported increases in fibrinogen with PM_2.5 exposures but not BC (Hajat et al. 2015).

Table 2. Summary of fibrinogen risk estimates associated with PM_2.5, BC, and PN exposure.

Study	Cohort	Study Period	Moving averages	Significant Findings*
Zeka et al. (2006)	NAS (n = 710) Mean age: 73.0; 100% men	2000–2004	2-, 7-, 28-d	PM2.5: none (+) BC: 28-d (+) PN: 2-, 7-d
Bind et al. (2012)	NAS (n = 704) Mean age: 73.2; 100% men	2000–2009	4-h, 1-, 3-, 7-, 14-, 21-, 28-d	PM2.5: none (+) BC: 3-, 7-d (+) PN: 3-, 14-, 21-, 28-d
Bind et al. (2014)	NAS (n = 822) Mean age: 72.4; 100% men	1999–2011	7-d	PM2.5: none (+) BC: 7-d (+) PN: 7-d
Li et al. (2017)	Framingham (n = 3,996) Mean age: 53.6; 46% men	1998–2011	1- to 7-d	PM2.5: none (-) BC: 2-d, 3-d

Notes: *Bind et al. (2014): results for high oxidative stress, endothelial dysfunction, and metal processing dysfunction

(+): positive associations reported

(-): negative associations reported

Figure 3. Effect of traffic exposure on inflammatory markers–-C reactive protein (CRP) and fibrinogen. Risk estimates reported per 1 μg/m³ increase in BC, 10 μg/m³ increase in PM_2.5, and 15,000 particles/cm³ increase in PNC.

Of note, majority of studies outside of the Harvard EPA Center have not found consistent associations between particles and fibrinogen, with some reporting negative (Panasevich et al. 2009; Steinvil et al. 2008), positive (Ghio et al. 2003; Liao et al. 2005), or null associations (Baccarelli et al. 2007; Rückerl et al. 2007, 2014). In the Greater Boston Area, the Framingham Offspring and Third Generation cohort study of nearly 4,000 participants found a negative association between BC and fibrinogen levels (Li et al. 2017). Authors concluded this could possibly be due to residual or unmeasured confounding. Therefore, while NAS cohort studies support the relationship between traffic exposure and increased fibrinogen, other studies show mixed results and evaluation of additional inflammatory markers are needed.

C-reactive protein (CPR)

C-reactive protein (CRP) is an acute-phase protein that rises in the presence of inflammation. The aforementioned studies also measured CRP levels and reported weaker associations between traffic exposure and CRP compared to fibrinogen. This is most likely due to larger variability in CRP measurements reported by many studies. Most notably, in a repeated measures analysis by Harvard investigators, elderly men in the NAS cohort with high genetic susceptibility for metal processing dysfunction had increases in CRP level with greater PN exposure, but not with BC or PM_2.5 exposure (Bind et al. 2014). Consistent with these findings, a subsequent quantile analysis of the same cohort showed that among participants with CRP levels > 2 mg/L (i.e., 60th percentile), an IQR increase in 28-day average PN was associated with an increase in CRP, while no associations were found with BC or PM_2.5 (Bind et al. 2016). Therefore, PN exposure may confer greater cardiovascular risk than PM_2.5 exposure.

Unlike PN, associations between BC exposure and CRP have not been found in NAS cohort studies. However, a Harvard EPA Center study on 88 elderly men with COPD in Eastern Massachusetts found a 12% higher CRP level (95% CI: 2, 23%) per IQR increase in 7-day average indoor BC (Garshick et al. 2018). Furthermore, a large study on nearly 4,000 participants in the Framingham Offspring and Third Generation cohorts reported a 5.8% higher CRP level (95% CI: 0.5, 11.4) in association with a 0.5 µg/m³ higher 5-day moving average BC, although associations with other moving averages (1- to 7-days) were null (Li et al. 2017).

Outside of the Harvard EPA Center, a study conducted in a Los Angeles cohort of 29 elderly subjects with a history of coronary artery disease found significant associations between CRP and same-day BC and PN exposure, but not for PM_2.5 (Delfino et al. 2008). In addition, this study compared associations with components of indoor PM and found more robust associations with PM of outdoor origin (e.g., PN, elemental carbon), supporting that traffic-related particles and freshly generated particles trigger inflammatory responses. A follow-up study with a larger cohort (60 elderly subjects with history of coronary artery disease living in Los Angeles) found that CRP was elevated with same-day PN exposure but not BC or PM_2.5 exposure (Delfino et al. 2009). This finding may reflect differences in the particle size distributions of PN and BC. PN is composed of nanoparticles and ultrafine particles, which differ in aerodynamic properties. For instance, particles 6–12 nm are more sensitive to air stagnation than larger particles 50–100 nm, as evidenced by higher concentrations of nanoparticles near highways in cooler months in Los Angeles (Zhu et al. 2004). This adds to the evidence that PN exposure may be more strongly associated with adverse cardiovascular outcomes than PM_2.5 exposures. This is especially notable given the relatively higher exposure error of PN due to its considerable spatiotemporal variability.

ICAM-1 and VCAM-1

ICAM-1 and VCAM-1 are proteins found on cell surfaces that mediate the recruitment of white blood cells during an inflammatory response. The soluble form of ICAM-1 and VCAM-1 in the plasma are measured in studies. While ICAM-1 and VCAM-1 are both involved in recruiting white blood cells to sites of inflammation, increased plasma ICAM-1 levels may indicate a general inflammatory state and upregulation in non-endothelial cells, whereas VCAM-1 may have a more restricted distribution within the vascular system (Pradhan, Rifai, and Ridker 2002).

Among Harvard EPA Center studies, ICAM-1 and VCAM-1 were found to be associated with BC and PN exposure in the NAS cohort (Figure 4) (Bind et al. 2012; Madrigano et al. 2010). In addition to NAS cohort studies, Garshick et al. (2018) assessed a smaller cohort of 88 COPD patients and found no association between BC and VCAM-1, whereas O’Neil et al. (2007) found a positive association in a cohort of 92 diabetic patients. Like fibrinogen, there was evidence that VCAM-1 was more strongly associated with BC and PN than PM_2.5 (Table 3). In Madrigano et al. (2010), increased VCAM-1 levels were associated with BC but not PM_2.5. In O’Neil et al. (2007), both BC and PM_2.5 exposures were associated with increases in VCAM-1, but BC exposure was linked to greater increases in VCAM-1 levels than PM_2.5 exposure. In addition, PN exposure was more strongly associated with increases in VCAM-1 than PM_2.5 and BC exposures in Bind et al. (2012). Given that exposure errors for PN are generally larger than for PM_2.5 due to greater spatial variability, a greater PN risk estimate despite a less precise exposure measure suggests that the true association between VCAM-1 and PN is likely even greater in magnitude.

Table 3. Summary of VCAM-1 risk estimates associated with PM_2.5, BC, and PN exposure.

Study	Cohort	Study Period	Findings*
Madrigano et al. (2010)	NAS (n = 809; 1,819 measures) Mean age: 72.3; 100% men	1999–2008	2-d 1 μg/m^3 BC, 4.5%, [1.1%, 8.0%] 2-d 10 μg/m^3 PM2.5, 1.7% [−0.9%, 4.3%]
Bind et al. (2012)	NAS (n = 704; up to 4 measures per participant) Mean age: 73.2; 100% men	2000–2009	3-d IQR PN, 10.3% [6.6%, 14.2%] 3-d IQR BC, 3.0% [0.5%, 5.6%] 3-d IQR PM2.5, 2.7% [0.6%, 4.8%]
O’Neil et al. (2007)	92 clinical trial participants at Joslin Diabetes Center and BIDMC Mean age: 56.6; 60% men	1998–2002	5-d IQR BC, 23.8% [8.4%, 41.4%] 5-d IQR PM2.5, 8.6% [0.1%, 17.8%]
Garshick et al. (2018)	COPD patients from Eastern MA (n = 88) Mean age: 72.7; 100% men	Nov 2012 -Dec 2014	2-d IQR BC, −0.03% [−1.48%, 1.44%] 3-d IQR BC, −0.07% [−1.71%, 1.59%] 5-d IQR BC, −0.09% [−1.68%, 1.53%]

Note: * 95% confidence interval reported

Figure 4. Effect of traffic exposure on inflammatory markers ICAM and VCAM. Risk estimates reported per 1 μg/m³ increase in BC, 10 μg/m³ increase in PM_2.5, and 15,000 particles/cm³ increase in PNC.

Both Bind et al. (2012) and Alexeeff et al. (2011) investigated the NAS cohort during similar time periods (1999–2008 vs. 2000–2009), but Alexeeff et al. utilized a spatiotemporal land use regression model that was previously developed and validated for estimating daily BC concentrations for each participant’s address (Gryparis et al. 2007). Briefly, this model integrated daily monitored values from 82 different monitoring sites in the Greater Boston Area and predicted BC levels using land use measures (e.g., cumulative traffic density, population density, distance to major roadways), geographic information system location (e.g., latitude, longitude), meteorological factors (e.g., apparent temperature, wind speed, planetary boundary layer height), and other characteristics (e.g., day of week, day of season). For a 1 µg/m³ increase in a 28-day average BC, modeled BC concentrations yielded a 5.1% increase in ICAM-1 (95% CI: 0.74, 9.6%) whereas measured BC values yielded a higher risk estimate of 8.8% increase in ICAM-1 (95% CI: 5.3, 12.2%) (Figure 5, Table 4). Similarly, increases in VCAM-1 levels were lower with the modeled compared to measured BC concentrations.

Table 4. Risk estimates derived from measured versus modeled BC concentrations (risk estimates reported per 1 µg/m³ increase in BC).

Study	Outcome	Moving Avg	Risk (monitored BC)	Risk (modeled* BC)
Bind et al. (2012)	VCAM ICAM	28-d	6.4% (1.3, 11.8%) 8.8% (5.3, 12.2%)
Alexeeff et al. (2011)	VCAM ICAM	28-d		3.4% (−2.2, 9.2%) 5.1% (0.7, 9.6%)
Wilker et al. (2010)	DBP (mmHg) SBP (mmHg)	7-d	5.5 mmHg (4.5, 6.4) 7.3 mmHg (5.5, 9.1)
Schwartz et al. (2012)	DBP (mmHg) SBP (mmHg)	7-d		2.6 mmHg (1.0, 4.2) 3.1 mmHg (0.2, 6.0)
Wellenius et al. (2012)	Ischemic Stroke	1-d	21.0% (4.0, 41.6%)	13.7% (1.7, 28.1%)
Mostofsky et al. (2012)	Ischemic Stroke	1-d	52.8% (17.2, 99.5%)

Note: *Land use regression models were used to derive BC concentrations in these studies.

Figure 5. Effect of BC exposure on ICAM and VCAM using land use regression models. Risk estimates reported per 1 μg/m³ increase in BC.

Using a validated land use regression model allows investigators to address individual-level differences in exposures rather than classifying exposures based on measurements at the nearest monitor. A validation study illustrated that 1) the predicted BC estimates are highly correlated with actual exposure measurements from 30 additional monitoring locations in Boston separate from the 82 monitoring stations used to fit the land use model, and that 2) actual exposure measurements are much more closely correlated with the predicted BC measurements than exposure estimates based on a central-site monitor (Gryparis et al. 2007). Therefore, risk estimates derived from our land use regression model may reflect a more precise health risk estimate, particularly for pollutants like BC that are more spatially variable.

Outside of the Harvard EPA Center, very few studies have assessed the association of BC and PN exposure with ICAM-1 and VCAM-1. The aforementioned Multi-Ethnic Study of Atherosclerosis cohort study found no association between ICAM-1 with PM_2.5 or BC in 2,865 participants in 6 U.S. cities (Hajat et al. 2015). While this study included a large number of samples, an important limitation was that an annual average BC concentration was used due to limited availability of daily central monitoring data on BC. A recent personal exposure study conducted among 53 traffic police officers in Kathmandu, Nepal showed no association between PM_2.5 or BC with ICAM-1, VCAM-1, or CRP (Shakya et al. 2019). This study was limited by sample size and short study duration (i.e., 5–6 days per participant). Therefore, the Harvard EPA Center studies uniquely provide valuable insights into the association of BC and PN exposure with increases in ICAM-1 and VCAM-1 levels.

Other inflammatory markers

In addition to fibrinogen, CRP, ICAM-1, and VCAM-1, studies have assessed a host of other markers of inflammation and thrombosis (e.g., interleukin (IL)-1β, IL-6, IL-8, tumor necrosis factor (TNF)-α, sedimentation rate, white blood cell (WBC) count, sTNF-RII, von Willebrand Factor, endothelin-1b and, vascular endothelial growth factor (VEGF) (Fang et al. 2012; Garshick et al. 2018; Liu et al. 2009; Wittkopp et al. 2013). Other studies have explored the effect of traffic pollutant exposures on novel biomarkers of oxidative stress, illustrating evidence of decreased activity of in erythrocyte antioxidant enzymes (e.g., glutathione peroxidase-1 and superoxide dismutase) (Delfino et al. 2009) and increases in lipid peroxidation and oxidative DNA damage (Grady et al. 2018). In aggregate, these studies suggest that systemic responses–-in particular systemic inflammation and pro-thrombotic activity–-mediate the relationship between traffic pollutant exposure and cardiovascular disease. Several Harvard EPA Center studies have also found an association between BC and ST-segment depression (Chuang et al. 2008; Gold et al. 2005), which suggests that particle exposure may induce low-grade myocardial inflammation and repolarization changes. Together, these studies provide evidence that traffic pollutants increase cardiovascular disease risk through an inter-related process of inflammation and endothelial dysfunction.

Blood pressure

Air pollution exposure induces autonomic imbalance, oxidative stress, and systemic inflammation. These, in turn, promote endothelial dysfunction (Madrigano et al. 2010), impaired vascular reactivity (Wilker et al. 2014), and hypertension (Brook et al. 2009). Studies have reported associations between elevated BP and short-term PM_2.5 and BC exposure among elderly subjects (Delfino et al. 2010; Liu et al. 2009; Zanobetti et al. 2004a), healthy individuals (Kubesch et al. 2015; Louwies et al. 2015), and newborns (van Rossem et al. 2015). Residential proximity to major roadways have also been shown to be associated with hypertension among postmenopausal women (Kirwa et al. 2014). However, other studies have found null or inconsistent links between traffic pollutants and blood pressure (Ebelt, Wilson, and Brauer 2005; Fuks et al. 2014; Ibald-Mulli et al. 2004). Human chamber inhalation studies have demonstrated PM_2.5-induced increases in BP, particularly in high-traffic areas (Brook et al. 2009). Personal exposure studies on indoor and outdoor cycling have also reported an association between short-term (2 hour) ultrafine particle exposure and increases in systolic BP and adverse effects on micro-vascular function among healthy individuals (Kubesch et al. 2015; Weichenthal, Hatzopoulou, and Goldberg 2014). In addition, a year-long particle filter intervention has been shown to lower systolic BP in a Taiwanese cohort (Chuang et al. 2017). Furthermore, PM_2.5, BC, and PN have been found to be associated with a higher baseline pulse amplitude, suggesting an adverse effect on vascular tone (Ljungman et al. 2014).

A number of Harvard EPA Center studies have investigated the association between traffic pollutants and BP in the NAS cohort (Figure 6, Table 5). These NAS studies consistently reported increases in both systolic and diastolic BP associated with BC exposure (Bind et al. 2016; Mordukhovich et al. 2009; Schwartz et al. 2012; Wilker et al. 2010; Zhong et al. 2016). In Mordukhovich et al. (2009), an elevation in BP was observed with BC exposure but not PM_2.5 in the NAS cohort (n = 461, 1999–2007). Wilker et al. (2010) studied a larger NAS cohort (n = 789, 1995–2008) and reported even greater BP increases associated with BC exposure. Furthermore, Bind et al. (2016) found larger elevations in systolic BP with BC exposure compared to PM_2.5 exposure, particularly among participants with higher baseline systolic BP. For instance, among participants with systolic BP > 155 mmHg (i.e., 90th percentile), an IQR increase in BC exposure was associated with a higher systolic BP by 7.2 mmHg (95% CI: 5.5, 8.8) compared to 3.6 mmHg (95% CI: 1.6, 5.7) for PM_2.5 exposure. This finding adds to the evidence that BC exposure may have greater consequences to cardiovascular health than PM_2.5 exposure.

Table 5. Summary of blood pressure risk estimates associated with PM_2.5, BC, and PN exposure.

Study	Cohort	Study Period	Moving averages	Significant Findings*
Mordukhovich et al. (2009)	NAS (n = 461) Mean age: 71.1; 100% men	1999–2007	7-d	PM2.5: none BC: 7-d (DBP & SBP)
Schwartz et al. (2012)	NAS (n = 853) Mean age: 72.6; 100% men	1996–2008	1-, 7-, 28-d, 3-mo, 12-mo	BC: 1-, 7-, 28-d, 3-mo, 12-mo (DBP) 7-, 28-d, 3-mo, 12-mo (SBP)
Wilker et al. (2010)	NAS (n = 789) Mean age: 72.3; 100% men	1995–2008	7-d	BC: 7-d (DBP & SBP)
Zhong et al. (2016)	NAS (n = 675) Age 55–100; 100% men	1999–2012	2-, 5-, 7-, 14-, 21-, 28-d	BC: 2-, 5-, 7-, 14-, 21-, 28-d (DBP & SBP)
Bind et al. (2016)	NAS (n = 1,112) Median age: 69; 100% men	1995–2013	28-d	PM2.5: 28-d (DBP & SBP) BC: 28-d (DBP & SBP) PNC: 28-d (DBP & SBP)

Notes: DBP: diastolic blood pressure, SBP: systolic blood pressure.

Figure 6. Effect of traffic exposure on systolic and diastolic blood pressure. Risk estimates reported per 1 μg/m³ increase in BC and 10 μg/m³ increase in PM_2.5.

Schwartz et al. (2012) also studied the NAS cohort (n = 853, 1996–2008), but employed a spatiotemporal land use regression model to estimate daily BC concentrations. Modeled BC concentrations in Schwartz et al. (2012) yielded a lower risk estimate than Wilker et al. (2010), which utilized monitored BC concentrations. A 1 μg/m³ increase in 7-day average BC was associated with an increase in diastolic BP by 5.5 mmHg (95% CI: 4.5, 6.4 mmHg) in Wilker et al. (2010) compared to 2.6 mmHg (95% CI: 0.98, 4.2 mmHg) in Schwartz et al. (2012). Similar patterns between modeled and monitored BC concentrations were observed with risk estimates of systolic BP increases. The utilization of land use regression models is an attempt to reduce exposure error as exposure misclassification often leads to an underestimation of the health effects of air pollution (Zeger et al. 2000). A slightly lower risk estimate with modeled compared to measured BC concentrations is consistent with the pattern of findings with the risk estimates for ICAM-1 and VCAM-1 levels (Table 4). In addition, Wellenius et al. (2012) explicitly compared and reported similar but slightly lower risk estimates for ischemic stroke derived from modeled compared to observed BC concentrations. Further studies comparing risk estimates using both BC measurements from central monitoring sites and modeled BC concentrations are needed.

Other Boston-based Harvard Center studies have found increases in BP with both BC and PM_2.5 exposures among cardiac rehabilitation patients (Zanobetti et al. 2004b) and individuals with type 2 diabetes (Hoffmann et al. 2012). Outside of the Harvard EPA Center, another Boston-based traffic exposure study assessed the effect of same-day PM_2.5, BC, and PN on BP among a cohort of 220 individuals who primarily reside near highways (Chung et al. 2015). This study found increases in diastolic BP associated with PN exposure but not with PM_2.5 or BC. In conjunction with the previously discussed robust effect of PN exposure on fibrinogen and CRP levels, this result provides additional evidence for the harmful effect of PN on cardiovascular health. The null associations with BC and PM_2.5 in Chung et al. (2015) may be due to differences in exposure window (e.g., same-day vs. 7-day) and study cohort, as susceptible populations (e.g., patients with history of cardiovascular disease, diabetic patients, elderly) may be at greater risk of BP increases with BC and PM_2.5 exposures. A recent study on the same Boston-based cohort found that the association between PN and BP were stronger among non-Hispanic white participants and among diabetics (Corlin et al. 2018).

In summary, associations between BP elevation and PM_2.5 and BC exposures were observed across different Boston-based cohorts in different time periods, using both monitored air pollution measurements and spatiotemporal model predicted values. In addition, there is evidence that BC and PN exposures pose greater risk of BP increases than PM_2.5 exposure.

Ischemic stroke

There is extensive literature on the adverse effect of both short-term (days to weeks) and long-term (months to years) exposure to air pollution and stroke (Ljungman and Mittleman 2014). A Harvard EPA Center study using a hybrid satellite remote sensing and land use regression approach to estimate particle exposure in New England found that both acute and chronic PM_2.5 exposures were associated with stroke (Kloog et al. 2012). Another study geocoded the addresses of all deaths in Massachusetts between 1995 and 2002, and found that an IQR increase in BC was associated with 2.3% (95% CI, 1.2, 3.4%) increase in daily deaths, with an even larger effect on stroke mortality risk (4.4%; 95% CI, −0.2, 9.3%) (Maynard et al. 2007).

Among Harvard EPA Center studies, Wellenius et al. (2012) and Mostofsky et al. (2012) are directly comparable, large cohort studies (1,705 and 1,060 patients, respectively) that found an elevated risk of ischemic stroke with exposure to PM_2.5 as well as BC (Figure 7). Both studies utilized a time-stratified case crossover design and included patients who were hospitalized at the Beth Israel Deaconess Medical Center (BIDMC) with neurologist-confirmed ischemic stroke. Mostofsky et al. 2012 additionally demonstrated that the effect of BC exposure was independent of the PM_2.5 effect. In that study, a second model adjusted for PM_2.5 in estimating the BC exposure risk, and a third model assessed the effect of residual BC mass that is not correlated with PM_2.5 concentrations. These risk estimates were similar to those estimated by the single-pollutant model, illustrating the BC effect was independent of PM_2.5 exposure. In addition, BC risk estimates were slightly larger than PM_2.5 risk estimates.

Figure 7. Estimated effects of traffic exposure on ischemic stroke risk. Risk estimates reported per IQR increase in BC and PM_2.5.

Outside of the Harvard EPA Center, epidemiological studies on the association between stroke and exposure to BC or PN are sparse. A recent large time-stratified case-crossover study (n = 2,742) based in Barcelona, Spain reported an increased risk of ischemic stroke risk among patients with large-artery atherosclerosis to be associated with BC exposure, but not with PM_2.5 (Vivanco-Hidalgo et al. 2018). In summary, exposure to traffic-related pollutants is associated with an increased risk of ischemic stroke. There is evidence that the effect of BC is independent of the effect of PM_2.5 exposure, and that the risk is greater with BC exposure than with PM_2.5 exposure.

Conclusion

The adverse effect of air pollution on cardiovascular health is well-established and supported by traffic studies from the Harvard EPA Center. Specific markers of traffic exposure (e.g., roadway proximity, BC, PN) are associated with a number of adverse cardiovascular outcomes, including cardiac autonomic dysfunction, systemic inflammation and thrombosis, elevated BP, and ischemic stroke. Studies both within and outside of the Harvard EPA Center suggest that BC and PN exposures are more strongly associated with cardiovascular outcomes than PM_2.5 exposures. There is epidemiological evidence that such associations may be physiologically mediated by inflammatory responses as well as microvascular and autonomic dysfunction; however, further toxicology and targeted animal and controlled human studies elucidating the underlying pathophysiological mechanism are needed. In addition, monitored BC and PN measurements as estimates of personal traffic exposure and the magnitude of exposure errors need further characterization. Strong associations between PN and inflammatory markers (e.g., VCAM-1, CRP) have been observed despite the wide variability in PN measurements–-and subsequently larger exposure errors–-suggesting that the true association may be even greater in magnitude. In addition, the use of land use models may yield a more precise health risk estimate especially for more spatially variable pollutants. More studies directly comparing risk estimates derived from monitored and modeled measurements of particulate pollutants are needed in order to characterize the influence of exposure errors on study results. Further research is necessary to support generalizability of our findings to populations with greater diversity in race, socioeconomic status, geographic location, age, and health status. With the wealth of evidence that traffic emissions as measured by BC and PN are particularly harmful to human health, controlling their local and regional sources is important for protecting public health.

Abbreviations

BC	=	black carbon
BP	=	blood pressure
BIDMC	=	Beth Israel Deaconess Medical Center
CO	=	carbon monoxide
CRP	=	C-reactive protein
EPA	=	Environmental Protection Agency
ICAM-1	=	intercellular adhesion molecule-1
MOBILIZE	=	Maintenance of Balance, Independent Living, Intellect, and Zest in the Elderly
NO2	=	nitrogen dioxide
NAS	=	Normative Aging Study
PM2.5	=	fine particulate matter
PN	=	particle number
UFP	=	ultrafine particle
VCAM-1	=	vascular cell adhesion molecule-1

Acknowledgment

This publication was made possible by the National Institute of Environmental Health Sciences grants P01-ES009825, P30-ES000002, R01-ES015172-01, R01-ES024332, R01-ES019853, R01-ES022657-01A1, R00-ES015774, and K23-ES026204 and US EPA grants RD-83479801, RD-83587201, and RD-83241601. Its contents are solely the responsibility of the grantee and do not necessarily represent the official views of the U.S. EPA. Further, U.S. EPA does not endorse the purchase of any commercial products or services mentioned in the publication. The Veterans Administration’s Normative Aging Study is supported by the Cooperative Studies Program/Epidemiology Research and Information Centers of the U.S. Department of Veterans Affairs and is a component of the Massachusetts Veterans Epidemiology Research and Information Center (MAVERIC), Boston, MA. Research reported in this publication was also supported by the John Harvard Distinguished Science Fellows Program within the FAS Division of Science of Harvard University and by the Office of the Director, National Institutes of Health under Award Number DP5OD021412. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health, the U.S. Department of Veterans Affairs, or the United States Government.

Additional information

Funding

This work was supported by the NIH Office of the Director [DP5OD021412]; National Institute of Environmental Health Sciences [ES000002, ES009825, ES015172-01, ES015774, ES024332, ES019853, ES022657-01A1, ES015774, ES026204, R01-ES019853, R01-ES022657-01A1, K23-ES026204, R01-ES019853, R01-ES022657-01A1, K23-ES026204]; and U.S. Environmental Protection Agency [RD-83241601, RD-83479801, RD-83587201].

Notes on contributors

Iny Jhun

Iny Jhun and Jina J. Kim are graduates of the Harvard T. H. Chan School of Public Health in Boston, MA.

Iny Jhun and Bennet Cho are students at Harvard Medical School in Boston, MA.

Jina Kim

Iny Jhun and Jina J. Kim are graduates of the Harvard T. H. Chan School of Public Health in Boston, MA.

Bennet Cho

Iny Jhun and Bennet Cho are students at Harvard Medical School in Boston, MA.

Diane R. Gold

Diane R. Gold, Joel Schwartz, Brent A. Coull, Murray Mittleman, Francesca Dominici, and Petros Koutrakis are research professors at the Harvard T. H. Chan School of Public Health in Boston, MA.

Diane R. Gold is also a professor of medicine at Harvard Medical School and affiliated with the Brigham and Women’s Hospital in Boston, MA.

Antonella Zanobetti

Antonella Zanobetti is a principal research scientist at the Harvard T. H. Chan School of Public Health in Boston, MA.

Mary B. Rice

Mary B. Rice and Elissa H. Wilker are assistant professors of medicine at the Harvard Medical School and affiliated with the Beth Israel Deaconess Medical Center in Boston, MA.

Murray A. Mittleman

Murray Mittleman is also an associate professor of medicine at Harvard Medical School and affiliated with the Beth Israel Deaconess Medical Center in Boston, MA.

Eric Garshick

Eric Garshick is a professor medicine at the Harvard Medical School and affiliated with the Veterans Affairs Boston Healthcare System in Boston, MA

Pantel Vokonas

Pantel S. Vokonas is a professor of preventive medicine and epidemiology at the Boston University School of Medicine and affiliated with the Veterans Affairs Boston Healthcare System in Boston, MA.

Marie-Abele Bind

Marie-Abele Bind is a research associate at the Department of Statistics and John Harvard Distinguished Science Fellow, Division of Science at Harvard University in Cambridge, MA.

Elissa H. Wilker

Elissa H. Wilker is also an adjunct assistant professor at the Harvard T. H. Chan School of Public Health in Boston, MA and affiliated with Sanofi Genzyme in Cambridge, MA.

Helen Suh

Helen Suh is a professor in the Department of Civil and Environmental Engineering at Tufts University in Medford, MA.