Air pollution is associated with increased risk of venous thromboembolism: the Multi-Ethnic Study of Atherosclerosis

Air pollution is a harmful ubiquitous hazard associated with elevated risk of atherosclerotic cardiovascular disease.^1-3 Venous thromboembolism (VTE), which consists of deep vein thrombosis and pulmonary embolism, annually affects ≈1 million Americans⁴ and may also be linked to air pollution, although research evaluating this association in the United States is sparse. Even associations of modest magnitude are of interest in the context of air pollution, because the exposure is widespread, and thus the population-attributable risk may be large.

Several reviews have summarized mechanistic evidence that may link air pollution to the development of VTE. Pollution is believed to be prothrombotic and antifibrinolytic, and to induce endothelial dysfunction, platelet aggregation, and inflammation.^5-7 Numerous individual studies have evaluated the association of air pollution with markers of coagulation and inflammation.^8-15 For example, a 40 parts per billion (ppb) increase in long-term oxides of nitrogen (NO_x) was associated cross-sectionally with a 7% (95% confidence interval [CI], 2%-13%) higher level of D-dimer after controlling for confounders in a Multi-Ethnic Study of Atherosclerosis (MESA) publication.¹⁰ In addition, higher fine particulate matter (fine particulate matter ≤2.5 micrometers in aerodynamic diameter [PM_2.5]) exposure was associated with greater concentrations of C-reactive protein (CRP), fibrinogen, and E-selectin.¹⁰ In an evaluation of air pollution and markers of coagulation, an Italian study among healthy adults reported that higher concentrations of coarse particulate matter (particulate matter ≤10 micrometers in aerodynamic diameter [PM₁₀]) and gaseous pollutants (carbon monoxide, nitrogen dioxide [NO₂]) were associated with shorter prothrombin time, but not with shorter activated partial thromboplastin time.¹⁵ 

Existing studies evaluating the association between air pollution and VTE yielded mixed results, with some finding exposure to air pollutants to be harmful^16-26 and others reporting no significant association.^27-30 Relatively few were conducted in the United States or used a traditional prospective cohort approach, which allows precise longitudinal assessment of end points and exposures and the ability to account for confounding factors at the individual level. A prospective study with well-characterized air pollution measures would provide valuable information about whether an association between chronic exposure to air pollution and VTE exists.

The characterization of air pollution in the US-based MESA is unparalleled for a large-scale cohort,³¹ with community- and participant-level estimated concentrations aggregated to estimate indexes of multiple air pollutants that were updated each fortnight of follow-up over up to 18.5 years. Using these unique data, we tested the hypotheses that greater exposure to PM_2.5, NO_x, NO₂, and ozone (O₃) is associated with elevated incident VTE risk.

MESA is a community-based prospective cohort study that began in 2000 to 2002 when 6814 men and women, self-identifying with 4 racial-ethnic groups, were recruited from 6 US communities. Individuals were classified as Hispanic, non-Hispanic Black, non-Hispanic White, or non-Hispanic Chinese based on their answers to race and ethnicity questions modeled on the year 2000 Census. White participants were recruited from all study sites, whereas participants identifying as Black were recruited from Forysth County, North Carolina; Chicago, Illinois; New York, New York; Baltimore, Maryland; and Los Angeles, California; those identifying as Hispanic from St. Paul, Minnesota; New York, New York; and Los Angeles, California; and those identifying as Chinese from Chicago, Illinois; and Los Angeles, California.^32,33 To be eligible for inclusion, potential participants had to be free from clinically recognized cardiovascular disease, and report that they had no other serious medical conditions that would prevent long-term participation in this study. The study was approved by the institutional review boards of the affiliated centers, and all participants provided written informed consent before inclusion.

The MESA Air ancillary study^3,31 assessed air pollution exposure at baseline and throughout follow-up using a residence-specific, spatiotemporal approach. For this analysis, we followed up participants from baseline through the end of 2018 for incident VTE events. We excluded participants who had no or missing VTE follow-up information (n = 31), reported using anticoagulants (n = 39), or were missing key covariates (n = 22). We also excluded those missing air pollution exposure data during follow-up (n = 71). Our final analytic data set included 6651 participants.

The methods of MESA Air pollution measurements and modeling have been described previously.^31,34 Briefly, residence-specific spatiotemporal pollution models were created, incorporating community-specific measurements, agency monitoring data, and geographical predictors that estimated concentrations of PM_2.5 and NO_x between 1999 and 2018 at the participants’ residence. Input data include information from 27 long-term sites, 771 community snapshot locations (simultaneous 2-week measurements collected on 3 occasions), and outside 697 participants’ homes between 2005 and 2009 to supplement measurements from the Environmental Protection Agency Air Quality System for monitoring locations.³ Long-term exposure of each participant to PM_2.5, PM₁₀, NO_x, and NO₂ was estimated as the average of fortnightly residence-specific predictions.

MESA surveillance includes telephone interviews with participants (or proxies) every 9 to 12 months that query hospital admissions and deaths. International Classification of Diseases (ICD) discharge codes were recorded for all participants who reported a hospitalization. For participants who died, next-of-kin interviews were conducted, including inquiries about any hospitalizations preceding death. VTE was defined by the presence, in any position, of the following International Classification of Diseases, Ninth Revision (ICD-9) hospital discharge codes: 415, 415.x (except 415.0), 451, 451.1x, 451.2, 451.81, 451.9, 453.1, 453.2, 453.4x, 453.5x, 453.8, 453.82, and 453.9; or the following International Classification of Diseases, Tenth Revision (ICD-10) revision codes: I26, I26.0, I26.0x, I26.9, I26.90, I26.92, I26.99, I80.1, I80.1x, I80.2, I80.20, I80.20x, I80.21, I80.21x, I80.22, I80.22x, I80.23, I80.23x, I80.29, I80.29x, I80.3, I80.8, I80.9, I82.1, I82.22, I82.220, I82.221, I82.4, I82.40, I82.40x, I82.41, I82.41x, I82.42, I82.42x, I82.43, I82.43x, I82.44, 82.44x, I82.49, I82.49x, I82.4Y, I82.4Yx, I82.4Z, I82.4Zx, I82.5, I82.50, I82.50x, I82.51, I82.51x, I82.52, 82.52x, I82.53, I82.53x, I82.54, I82.54x, I82.59, I82.59x, I82.5Y, I82.5Yx, I82.5Z, I82.5Zx, I82.9, I82.90, and I82.91. MESA did not capture outpatient VTE events, nor did it review medical records to validate hospital VTE discharge codes. At baseline, participants were asked if they had cancer previously, and cancer events have been identified throughout follow-up via ICD-9 and ICD-10 codes. Noncancer VTE events were defined as those that did not have evidence of cancer preceding or at the time of the VTE event. In a sensitivity analysis, we also restricted to VTE events occurring in the first position, and (separately) excluded VTE events with pulmonary disease ICD codes (ICD-9: 460-519; ICD-10: J00-J99) in the first position.

Individual covariate information was collected at baseline by centrally trained study personnel. Questionnaires queried self-reported information on age, race/ethnicity, sex, study site, education, income, smoking status, secondhand smoke exposure,³⁵ physical activity,³⁶ and neighborhood socioeconomic status (SES).³⁷ Height and weight were measured by standardized procedures, and body mass index (BMI) was calculated as weight in kilograms divided by the square of height in meters (kg/m²). In sensitivity analyses, we stratified by prevalent respiratory disease at baseline, defined by self-reported asthma, self-reported emphysema, or medication use for asthma (inhaled steroids or oral anti-inflammatory drugs).

Fasting blood samples at baseline were collected and handled according to standardized protocols.³⁸ Plasma biomarkers were analyzed at the MESA central laboratory and biorepositry at the Laboratory for Clinical Biochemistry Research at the University of Vermont. We identified markers of inflammation and coagulation that were measured in the full sample. D-dimer was determined by immunoturbidometry on the Sta-R analyzer (Liatest D-DI; Diagnostica Stago, Parsippany, NJ). Fibrinogen and CRP were measured using the BNII nephelometer (N Antiserum to Human Fibrinogen and N High Sensitivity CRP; Dade Behring Inc, Deerfield, IL), whereas interleukin 6 (IL-6) was measured by ultrasensitive enzyme-linked immunosorbent assay (Quantikine HS Human IL-6 Immunoassay; R&D Systems, Minneapolis, MN) and factor VIII using the Sta-R analyzer (STA-Deficient VIII; Diagnostica Stago).

Descriptive characteristics for quartiles of average air pollution during follow-up were tabulated. Cox proportional hazards regression was used to evaluate the association between time-varying air pollution indices and risk of incident VTE. Person time began at baseline and accrued until whichever occurred first of the following: incident VTE, fortnightly exposure estimates were unable to be updated for the individual, death, loss-to-follow-up, or administrative censoring on December 31, 2018. Time-varying exposures were modeled using the Andersen-Gill counting method.³⁹ Specifically, follow-up time for each MESA participant was split into 30-day time bins for this analysis, and air pollution exposure was assessed at each time bin such that the exposure estimate in a given bin represents the participant-specific estimated concentration averaged over the 2 years before that month of follow-up. In sensitivity analyses, we explored associations restricting follow-up to the first 10 years, and additionally adjusting for neighborhood SES.

For each pollutant, a series of nested models were used. Model 1 adjusted for age, race/ethnicity, sex, study site, and education. Model 2 further adjusted for smoking status (3 options: current/former/never), secondhand smoke exposure (5 options: none, 1-9 hours/week, ≥10 hours/week, current smoker, and missing), physical activity, and BMI. Model 3, our mediation model, included the same variables as model 2, but additionally adjusted for baseline biomarkers that might be on the causal pathway between air pollution and incident VTE. Importantly though, biomarkers were measured at a single time point at baseline, not updated each fortnight as pollution measures were.

Multiplicative interactions for air pollution measures with age, race/ethnicity, study site, smoking status, BMI, and factor VIII were tested. Analyses were repeated with noncancer VTE as the outcome (cancer-associated VTE cases were censored at date of VTE occurrence). Several additional sensitivity analyses were conducted restricting follow-up to the first 10 years, exploring a more specific definition of VTE, and additionally accounting for neighborhood SES, smoking status, exposure to secondhand smoke, respiratory disease at baseline, and pulmonary disease concurrent with incident VTE. All analyses were performed using SAS version 9.4 (SAS Inc, Cary, NC).

The 6651 participants included in the present analysis were on average 62.1 ± 10.2 years old, 53.0% were female, and 38.4% identified as White, 27.5% as Black, 22.0% as Hispanic, and 12.0% as of Chinese ancestry. The recruitment was balanced by field site, with 15% to 20% of participants coming from each site. Mean±SD PM_2.5 was 13.5±3.0 μg/m³, NO₂ 17.9±8.2 ppb, NO_x 36.1±19.6 ppb, and O₃ 22.2±3.7 ppb.

Table 1 reports participant characteristics by quartiles of average PM_2.5. In general, participants with higher PM_2.5 exposure tended to be older and non-White. The racial/ethnic pattern to some degree reflects variation in air pollution by site, and site-specific differences in racial/ethnic group at study enrollment.^32,33 Concentrations of D-dimer, IL-6, and CRP at baseline were similar by PM_2.5 quartile, whereas higher fibrinogen and factor VIII concentrations were observed with greater average PM_2.5 exposure in follow-up. The air pollutants were intercorrelated, with higher PM_2.5 associated with greater higher levels of NO₂ and NO_X, and lower levels of O₃. supplemental Tables 1-3 (available on the Blood website) provide participant characteristics stratified by quartiles of average NO₂, NO_X, and O₃, respectively.

A total of 248 VTE events accrued over a median follow-up of 16.7 years. When modeled continuously, greater concentrations of all pollutants, except for O₃, were significantly associated with greater risk of incident VTE across all 3 studied models (Table 2). Specifically, in model 2, the hazard ratio (HR) (95% CI) for incident VTE associated per 3.6 μg/m³ higher PM_2.5 was 1.39 (1.04-1.86). For 13.3 ppb higher NO₂ concentration, the HR was 2.74 (1.57-4.77), whereas for 30.0 ppb higher NO_x, it was 2.21 (1.42-3.44). The associations were virtually identical with additional adjustment in model 3 for baseline biomarkers of coagulation and inflammation that may mediate the association. When considering only noncancer VTE (n events = 156), the magnitude was slightly stronger for PM_2.5 (model 2 HR, 1.59; 95% CI, 1.15-2.18) and NO₂ (model 2 HR, 3.14; 95% CI, 1.65-5.95) but virtually identical for NO_X. There were no multiplicative interactions by age, race, or smoking status. There was modest evidence (P_interaction = .04) that the association between PM_2.5 and VTE was of slightly higher magnitude at the Chicago site. To facilitate cross-study comparisons, supplemental Table 4 provides HRs (95% CIs) of the pollutants and risk of VTE with the pollutants scaled to 1 unit greater estimated concentration.

Numerous sensitivity analyses were conducted to explore the robustness of our findings. Specifically, we:

1. restricted follow-up to the first 10 years (Table 3),

2. defined incident VTE by requiring ICD codes to be in the first position (supplemental Table 5),

3. additionally adjusted for neighborhood SES (supplemental Table 6),

4. excluded current smokers (Table 4),

5. excluded current smokers and individuals exposed to secondhand smoke (supplemental Table 7),

6. stratified by prevalent respiratory disease status at baseline (supplemental Table 8), and

7. censored those (n = 13) with a pulmonary disease ICD code in the first position at the time of the incident VTE event (supplemental Table 9).

Briefly, the findings were similar across most sensitivity analyses. A few exceptions were that (a) magnitudes of association tended to be stronger with shorter follow-up (Table 3) and (b) there was no evidence of association among the 11% of participants with evidence of respiratory disease at baseline; however, precision was poor for this analysis. All other sensitivity analyses yielded findings similar to those of the main results.

Higher ambient concentrations of long-term average PM_2.5, NO₂, and NO_X were associated with greater risk of developing hospitalized VTE in this prospective cohort of 6651 individuals from 6 US communities. Findings persisted in an array of sensitivity analyses, such as restricting to nonsmokers and individuals without pulmonary disease. This study, which harnessed deeply characterized air pollution metrics in the context of a US community-based cohort, complements existing evidence by suggesting that chronic exposure to air pollution is linked to greater VTE risk in the general population. Areas with chronic exposure to ambient air pollution may experience slightly higher VTE rates than areas with less exposure. Importantly, air pollution is ubiquitous, so even modest associations can result in a large number of events. This study supports the case for global efforts of pollution reduction to curtail pollution-related adverse health outcomes, which we demonstrate includes risk of VTE.

The associations of both acute and chronic exposure to air pollution can be examined in relation to VTE risk. Both research questions have value; the MESA prospective cohort analyses conducted herein are designed to assess VTE risk associated with chronic air pollution exposure. Of studies that have evaluated short-term (triggering) effects via case-crossover designs, one found a significant association with PM_2.5,¹⁸ another found a significant association with PM_2.5 for pulmonary embolism but not for deep vein thrombosis,²² whereas a third concluded PM_2.5 and ozone were not associated with VTE risk.³⁰ 

Studies evaluating chronic exposure to pollutants and VTE have used case-control designs,^24-26 cohort designs,^16,17,27 or ecologic designs linking aggregate data on air pollution in a region with VTE hospitalization rates in that same region.^18,19,28,29 Some have used proximity to roadways as a surrogate for air pollution,^25,26 whereas many others measured specific pollutants (most often PM_2.5 and/or PM₁₀) using a variety of methods to ascertain exposure. Many studies were based in Italy,^{19,22,24-26,28,29} although studies have been conducted in the United States,^18,27 Sweden,¹⁶ Korea,¹⁷ Chile,²⁰ China,²³ and Iran.²¹ Most,^18-21 but not all,^28,29 of the ecologic studies suggested that air pollution is associated with VTE, although effect sizes have been modest. Case-control studies also typically reported a link between pollution and VTE,^23-26 often with larger effect sizes.^24-26 

Findings from MESA and other prospective cohorts suggest modest associations of chronic exposure to particulate matter and gaseous pollutants with VTE. Of prior studies, the one most similar to our analysis used data from the Swedish Malmö Diet and Cancer cohort.¹⁶ Here, greater exposure to PM_2.5 in the decade (modeled as a moving average) before the VTE event was associated with increased risk of VTE (HR, 1.17; 95% CI, 1.01-1.37 per interquartile range of 1.2 μg/m³). The magnitude of this finding is comparable to the current finding when a similar scaling is considered (see supplemental Table 4: HR for 1 μg/m³ higher PM_2.5, 1.10; 95% CI, 1.01-1.19). The Swedish Malmö Diet and Cancer cohort did not observe a significant association of any other pollutants (PM₁₀, NO_x, or black carbon) with incident VTE. Likewise, there was no association with PM_2.5 when modeled as average exposure in the year prior.¹⁶ An analysis of the Korean National Health Insurance Service–National Health Screening Cohort,¹⁷ which linked particulate matter and pollutant data from the Korean Nationwide Meteorological Observatory, reported an increased risk of VTE with PM₁₀ (HR, 1.06; 95% CI, 1.05-1.07 per 1 μg/m³), SO₂ (HR, 1.12; 95% CI, 1.08-1.16 per 1 ppb), and O₃ (HR, 1.04; 95% CI, 1.03-1.05 per 1 ppb), respectively. Here, in MESA, we identified adverse associations of NO₂ and NO_X with incident VTE, but not O₃.

Our study is one of the few evaluating chronic exposure to air pollution and VTE in the United States, a nation with substantial racial and geographic inequities in air pollution exposure.⁴⁰ An analysis conducted in the northeastern United States found that estimated exposure to PM_2.5 within a zip code (yearly moving average) was associated with a higher rate of VTE hospital admissions in the same region.¹⁸ Although the study had a large number of VTE events, unlike MESA Air, it had less comprehensive ascertainment of individual air pollution exposure, lacked information on potential confounding or mediating factors, and was not a closed cohort. Another US-based prospective cohort study assessed exposure to air pollution based on proximity to roadways, finding no association with incident VTE, but was underpowered to detect modest associations.²⁷ 

The primary strength of this study is the careful characterization of participants’ exposures to air pollution, which included assessment of 4 pollutants, updated each fortnight, over an average of nearly 17 years of follow-up.³⁴ In particular, the MESA Air prediction model takes advantage of cohort-specific monitoring campaigns. Yet, as with most long-term observational studies of air pollution exposure, the predicted exposure estimates used in this analysis were generated from a prediction model and not from individual-level continuous monitoring. Other strengths include the diverse sample from 6 geographically disparate communities, targeted multiethnic recruitment, and prospective design. However, given that only 6 US communities were included, the results may not be generalizable to the entire United States, especially to areas where average concentrations of pollutants were outside the ranges observed in the MESA communities. Limitations warrant discussion. First, incident VTE events were defined using hospitalization ICD codes. Some misclassification likely occurred because VTE events were not adjudicated by medical record review, and because VTE cases managed as outpatient would have been missed. Second, coagulation and inflammatory mediators were assessed only at baseline. The lack of consistent repeat measures of hemostatic and inflammatory biomarkers greatly hinders our ability to rigorously evaluate whether these biomarkers mediate the association between pollution and VTE. Last, the observational design inherently limits direct causal inference, although it has some advantages over retrospective studies.

Chronic exposure to air pollution, as defined by higher concentrations of PM_2.5, NO₂, and NO_X for up to 18.5 years, was associated with increased risk of developing VTE. This study is the most deeply characterized analysis of pollution and VTE to date that we are aware of, and adds to mounting evidence that air pollution adversely impacts health.

The authors thank the other investigators, the staff, and the participants of the Multi-Ethnic Study of Atherosclerosis (MESA) for their valuable contributions. A full list of participating MESA investigators and institutions can be found at http://www.mesa-nhlbi.org.

Funding received for MESA: This research was supported by contracts 75N92020D00001, HHSN268201500003I, N01-HC-95159, 75N92020D00005, N01-HC-95160, 75N92020D00002, N01-HC-95161, 75N92020D00003, N01-HC-95162, 75N92020D00006, N01-HC-95163, 75N92020D00004, N01-HC-95164, 75N92020D00007, N01-HC-95165, N01-HC-95166, N01-HC-95167, N01-HC-95168, and N01-HC-95169 from the National Institutes of Health (NIH), National Heart, Lung, and Blood Institute (NHLBI), and by grants UL1-TR-000040, UL1-TR-001079, and UL1-TR-001420 from the NIH/National Center for Advancing Translational Sciences. Funding received for MESA Air: this publication was developed under the Science to Achieve Results research assistance agreements, No. RD831697 (MESA Air) and RD-83830001 (MESA Air Next Stage), awarded by the US Environmental Protection Agency (EPA). It has not been formally reviewed by the EPA. The views expressed in this document are solely those of the authors, and the EPA does not endorse any products or commercial services mentioned in this publication. P.L.L. was partially supported by NIH/NHLBI grant K24 HL159246.

Contribution: P.L.L. and A.R.F. conceived the idea; J.R.M. analyzed the data, with support from M.T.Y. and C.L.L.; J.D.K. led collection of the Multi-Ethnic Study of Atherosclerosis air pollution data; and all authors aided in interpreting the data and critically reviewed the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Pamela L. Lutsey, Division of Epidemiology and Community Health, School of Public Health, University of Minnesota, 1300 S 2nd St, Suite 300, Minneapolis, MN 55454; email: lutsey@umn.edu.