OUP user menu

Long-term exposure to air pollution is associated with survival following acute coronary syndrome

(CC)
Cathryn Tonne, Paul Wilkinson
DOI: http://dx.doi.org/10.1093/eurheartj/ehs480 First published online: 20 February 2013

Abstract

Aims The aim of this study was to determine (i) whether long-term exposure to air pollution was associated with all-cause mortality using the Myocardial Ischaemia National Audit Project (MINAP) data for England and Wales, and (ii) the extent to which exposure to air pollution contributed to socioeconomic inequalities in prognosis.

Methods and results Records of patients admitted to hospital with acute coronary syndrome (ACS) in MINAP collected under the National Institute for Cardiovascular Outcomes Research were linked to modelled annual average air pollution concentrations for 2004–10. Hazard ratios for mortality starting 28 days after admission were estimated using Cox proportional hazards models. Among the 154 204 patients included in the cohort, the average follow-up was 3.7 years and there were 39 863 deaths. Mortality rates were higher for individuals exposed to higher levels of particles with a diameter of ≤2.5 µm (PM2.5; PM, particulate matter): the fully adjusted hazard ratio for a 10 µg/m3 increase in PM2.5 was 1.20 (95% CI 1.04–1.38). No associations were observed for larger particles or oxides of nitrogen. Air pollution explained socioeconomic inequalities in survival to only a small extent.

Conclusion Mortality from all causes was higher among individuals with greater exposure to PM2.5 in survivors of hospital admission for ACS in England and Wales. Despite higher exposure to PM2.5 among those from more deprived areas, such exposure was a minor contribution to the socioeconomic inequalities in prognosis following ACS. Our findings add to the evidence of mortality associated with long-term exposure to fine particles.

  • Air pollution
  • Myocardial infarction
  • Acute coronary syndrome
  • Mortality
  • Socioeconomic inequalities
  • Cohort

Introduction

Persuasive evidence continues to develop that exposure to air pollution, especially particulate matter (PM), is associated with heart disease, even at the relatively low concentrations currently found in the UK.1,2 However, few studies have investigated the influence of long-term exposure to air pollution on survival and subsequent cardiac events among survivors of myocardial infarction (MI) and the findings have been inconsistent.35 Individuals with pre-existing diseases such as ischaemic heart disease may be more susceptible to the adverse health effects of PM exposure.6 Survivors of ischaemic cardiovascular events therefore may serve as a useful model of vulnerability, potentially allowing for the detection of the effects of PM on prognosis with relatively few years of follow-up.

Outcome after MI has repeatedly been found to show a strong socioeconomic gradient,715although the underlying pathways are not well understood.16,17 Higher levels of air pollution frequently occur in more deprived areas across different populations.1820 This observation raises the possibility that exposure to air pollution may explain, in part, socioeconomic gradients in prognosis among MI patients.21

We therefore investigated (i) whether long-term exposure to air pollution was associated with all-cause mortality using the Myocardial Ischaemia National Audit Project (MINAP) data base for England and Wales, and (ii) the extent to which exposure to air pollution contributed to socioeconomic inequalities in prognosis.

Methods

Study population

The cohort consisted of patients admitted to hospital for acute coronary syndrome (ACS) identified through MINAP2224 who resided in England and Wales at the time of admission. We included patients with a final diagnosis (at discharge) of ST elevation MI (STEMI) and non-ST elevation MI (non-STEMI) between 1 January 2004 and 31 March 2007. Diagnosis of STEMI required the presence of electrocardiographic changes of ST elevation consistent with infarction (ST elevation ≥2 mm in contiguous chest leads and/or ST elevation ≥1 mm in two or more standard leads), and the presence of enzyme (twice the upper limit of reference range) or troponin elevation (above locally accepted cut-off value). Diagnosis of non-STEMI required electrocardiographic changes consistent with the diagnosis (new ST or T-wave changes except ST elevation), and the presence of enzyme or troponin elevation. Consistent with other researchers,3,25 we excluded from the follow-up the 28 days immediately following admission. We restricted our analysis to patients with complete data on date of admission, age, sex, and vital status who were older than 25 and who lived inside the geographic range of modelled air pollution for England and Wales (n = 154 204) (Supplementary material online, Table S1). Vital status was obtained from the Office of National Statistics. The follow-up continued until the date of death or the end of the study (1 April 2010). Our study was approved by the London School of Hygiene and Tropical Medicine research ethics committee (No. 5412).

Patients' demographic characteristics, postcode of residence, smoking and medical history, in-hospital treatment, and discharge drugs are recorded in the MINAP database, usually at the time of care.22 Small-area deprivation data were based on the Index of Multiple Deprivation (IMD) 2007 for England and IMD 2008 for Wales calculated at the level of Lower Layer Super Output Area (LLSOA). We used the score from income, education, and employment domains of IMD, since the composite IMD measure includes data on air pollution levels.26 Deciles of each score were generated using cut-points from the distribution across all LLSOAs in England, for residents of England, or the distribution from Welsh LLSOAs for residents of Wales.

Air pollution exposure

Annual average background concentrations of nitrogen dioxide (NO2), oxides of nitrogen (NOx), and two size fractions of particulate matter with diameter ≤10 µm (PM10) and ≤2.5 µm (PM2.5) in μg/m3 for the years 2004–10 were obtained from the Department for Environment Food and Rural Affairs (http://uk-air.defra.gov.uk/data/pcm-data). Concentrations were modelled at a resolution of 1 km × 1 km, using an approach described elsewhere.27,28 Briefly, point and area sources were modelled using ADMS, a second-generation Gaussian dispersion model.29 Contributions from rural background were based on measurements at rural ambient monitors and a network of diffusion tubes. Sources contributing to background concentrations of PM included point and area sources of primary particles, regional primary particles, secondary organic and inorganic aerosols, sea salt, and a residual contribution. Each contribution was modelled separately or derived from measurements as described elsewhere.27,28 Because PM concentrations prior to 2004 were modelled using other units, we restricted our analysis to admissions beginning in 2004 to ensure that exposures for each year were in comparable, gravimetric units.

For 2009, the background annual average NOx model had an r2 of 0.8 compared with 92 verification monitoring sites.28 The PM10 model had an r2 of 0.9 at 19 verification sites (tapered element oscillating microbalance sites corrected for volatile losses). The agreement between modelled PM2.5 at background sites (20 FDMS sites) was fairly good: r2 = 0.7.28

MINAP data were released with the coordinates of the residential postcode centroid rounded to 100 m to protect patient confidentiality. Individual level exposure was defined as the average concentration at model grid points within 1 km of each patient's postcode centroid for each year. Rather than considering oxides of nitrogen and PM as independent exposures (Spearman correlations in Supplementary material online, Table S2), we included NO2 and NOx because they can be considered surrogates for particles generated from specific sources such as traffic and had greater spatial variability compared with PM10 and PM2.5.

Statistical analysis

Hazard ratios were estimated using a Cox proportional hazards model (coxph) in the R software package (2.13.2). Time was modelled as days of follow-up starting 28 days after hospital admission. Each admitting hospital was allowed to have its own baseline hazard rate. Air pollution concentrations were included as a time-varying exposure, where the annual average air pollution was assigned to the person-time falling within the corresponding calendar year. To adjust for longer term time trends in air pollution and risk of death, we adjusted for calendar time using a natural spline with 2 degrees of freedom (df).

We adjusted for several clinical- and individual-level demographic characteristics as well as area-level deprivation, which were predictors of prognosis and correlated with air pollution concentrations. Variables were included in the following form: age—natural spline with 3 df; reperfusion—categories of no reperfusion treatment (reference), lysis, or percutaneous coronary intervention; area-level income—deciles; smoking—categories of ex-smoker, current, non-current with unknown history, or never smoked (reference); binary indicators for STEMI, white ethnicity, history of diabetes, angina, and MI prior to the first admission recorded in MINAP. Whether a patient was discharged from hospital on ACE-inhibitors, aspirin, beta-blockers, or statins was also modelled using binary indicators. To adjust for large-scale spatial variation in mortality or re-admission due to factors other than air pollution, we included indicators for Wales and the nine Government Office Regions in England. Our analysis investigates the association of within region, within hospital catchment area variation in air pollution and mortality. It therefore takes into account clustering of individuals with similar attributes other than those included in our model within a fairly small area. We assessed violations of the proportional hazard assumption by including an interaction term with the follow-up for each covariate in the fully adjusted model (Model 5). Regression models were based on observations with complete data for all covariates included in the model.

Sensitivity analyses

We investigated whether our results for the association between air pollution and survival were sensitive to our adjustment for deprivation, modelling each hospital with its own baseline, and the degrees of freedom for the time trend. The fully adjusted model (Model 5) was fit using deciles of education and employment deprivation rather than income. To account for clustering of patients within hospital, but allow some sharing of information across hospitals, we included hospital as a random effect rather than a stratification variable. We also fit the fully adjusted model with 3 rather than 2 degrees of freedom for time.

Attributable burden

We estimated the number of deaths brought forward due to exposure to PM2.5 in this cohort by first estimating the attributable fraction and then applying it to the number of deaths observed during the follow-up. Similar to a previous study,30 our calculation of attributable fraction compared the risk of exposure to PM2.5 based on the distribution of exposure observed in the cohort relative to 4 µg/m3, an estimate of PM2.5 due to natural sources alone (details in Supplementary material online) and used the hazard ratio estimated from the fully adjusted proportional hazards model.

Results

Among the 154 204 patients who met our inclusion criteria, the average duration of the follow-up was 3.7 years (SD = 1.6). There were 39 863 deaths (26% of patients) during the follow-up. Deaths according to year of the follow-up are presented in Supplementary material online, Table S3. Patient's demographic characteristics, medical history, treatments received while in hospital, area-level deprivation are presented in Table 1. Patients were on average 68 years old and were predominantly white males. The average exposure during the follow-up for residents of England was 18.8 µg/m3 for NO2, 17.0 µg/m3 for PM10, and 11.0 µg/m3 for PM2.5. Patients living in London had the highest exposures compared with other regions (Table 2). The relationship between income deprivation and PM2.5 according to admitting hospital is presented in Supplementary material online, Figure S1.

View this table:
Table 1

Characteristics of patients hospitalized with acute coronary syndrome in England and Wales between 2004 and 2007

n (% missinga)
Mean (SD) age154 204 (0)68 (13) years
Male154 204 (0)66.6%
Ethnicity141 236 (8)
 White90.3%
 Non-white9.7%
Smoking141 493 (8)
 Never24.5%
 Ex-smoker33.9%
 Current32.1%
 Non-current, unknown history9.5%
Medical history prior to admission
 Hypertension146 866 (5)45.3%
 Diabetes146 129 (5)16.6%
 Angina146 236 (5)27.1%
 Cerebrovascular disease141 598 (8)6.8%
 Heart failure141 812 (8)4.4%
 Previous AMI148 698 (4)19.9%
Final diagnosis
 ST elevation47.1%
 Non-ST elevation154 204 (0)52.9%
Reperfusion151 641 (2)
 None59.3%
 Lysis36.7%
 Primary PCIb4.0%
Discharge drugs
 ACE-inhibitor128 946 (16)83.4%
 Beta-blocker130 891 (15)76.9%
 Aspirin132 951 (14)93.2%
 Statin131 290 (15)93.8%
Mean (SD) area level (LSOA) deprivation
England145 132 (0)
 % residents in income deprivation16.4 (12.2)
 % residents in employment deprivation11.1 (7.1)
Wales9072 (0)
 % residents in income deprivation22.6 (20.0)
 % residents in employment deprivation22.7 (20.0)
  • AMI, acute myocardial infarction; PCI, percutaneous coronary intervention; ACE, angiotensin-converting enzyme-inhibitor; LSOA, lower super output area.

  • aPercentage missing based on 154 204 sample size, whereas percentage distributions for a given covariate are based on number with complete information for that covariate.

  • bEuropean Society of Cardiology guidelines published in 2005 state that primary PCI is the treatment of choice for patients with STEMI presenting in a hospital with PCI facility and an experienced team.46 Among patients with STEMI, the percentage receiving primary PCI was 4% in 2004; 7% in 2005; 12% in 2006; 14% in 2007.

View this table:
Table 2

Distribution of exposure to air pollution within person-time of the follow-up

Mean (SD) exposure by region (μg/m3)nNO2NOxPM10PM2.5
England145 13218.8 (6.8)28.3 (12.7)17.0 (2.7)11.0 (1.9)
 North East12 04517.7 (5.2)25.6 (9.0)13.7 (1.7)8.4 (1.2)
 North West22 15219.9 (6.0)29.6 (10.9)15.0 (2.3)9.5 (1.6)
 Yorkshire16 99818.4 (5.0)27.5 (9.3)16.1 (2.0)10.2 (1.3)
 East Midlands14 18517.0 (5.1)25.4 (9.6)17.6 (1.6)11.4 (1.2)
 West Midlands13 60920.8 (6.8)32.8 (13.8)17.6 (2.3)11.4 (1.6)
 East of England18 46216.6 (4.2)24.8 (7.7)18.2 (1.3)12.0 (0.9)
 London12 94930.5 (6.1)50.2 (12.5)21.7 (1.8)14.1 (1.1)
 South East20 81717.5 (4.7)25.1 (8.5)17.9 (1.5)11.8 (1.0)
 South West13 91512.8 (5.1)17.4 (8.1)15.7 (1.6)9.8 (1.2)
Wales907212.9 (5.5)17.7 (8.7)14.6 (1.9)9.1 (1.3)

Exposure to a 10 µg/m3 increase in PM2.5 during the same calendar year as the follow-up was associated with a 20% (95% CI 4–38) increase in death from all causes after adjusting for area-level income deprivation and other confounders (Table 3). There were no associations with the other pollutants and mortality in the fully adjusted models. There was evidence of non-proportional hazards of death by lysis, current smoking, and discharge on ACE-inhibitors in the fully adjusted models. However, the estimated associations were very similar to those presented for Model 5 in Table 3 after accounting for non-proportional hazards with an interaction term: confidence intervals around the HR for PM2.5 ranged from 1.03 to 1.37. There was no evidence of effect modification of the fully adjusted PM2.5 association with mortality by STEMI vs. non-STEMI (P = 0.29) or by reperfusion (P = 0.25 for interaction with primary PCI and P = 0.26 for lysis).

View this table:
Table 3

Hazard ratios and 95% CI for air pollution and income deprivation associated with all-cause mortality

Model (M): covariatesaEventsNO2 (per 10 µg/m3)NOx (per 10 µg/m3)PM10 (per 10 µg/m3)PM2.5 (per 10 µg/m3)Income (most vs. least deprived decile)
M1: age, sex, time39 8631.11 (1.08, 1.13)1.05 (1.04, 1.07)1.17 (1.10, 1.26)1.44 (1.29, 1.60)1.50 (1.43, 1.58)
M2: M1+reperfusion+region38 9171.10 (1.07, 1.13)1.05 (1.04, 1.07)1.18 (1.10, 1.27)1.48 (1.33, 1.66)1.50 (1.42, 1.58)
M3: M2+final diagnosis, smoking, ethnicity, diabetes, angina, previous MI30 7841.07 (1.04, 1.10)1.04 (1.02, 1.05)1.13 (1.05, 1.23)1.42 (1.26, 1.62)1.35 (1.27, 1.43)
M4: M3+discharge drugs25 8221.06 (1.03, 1.10)1.03 (1.01, 1.05)1.12 (1.03, 1.23)1.40 (1.22, 1.60)1.30 (1.22, 1.39)
M5: M4+pollutant and income (mutually adjusted)25 8221.01 (0.98, 1.04)1.00 (0.99, 1.02)1.01 (0.92, 1.10)1.20 (1.04, 1.38)1.28 (1.20, 1.37)b
  • aAll models stratified by admitting hospital. Covariates modelled as time trend (natural spline with 2 df), age (natural spline with 3 df), reperfusion (none, lysis, PCI), STEMI (yes/no), smoking history (ex, current, non-smoker with unknown history, or never smoker), white ethnicity (yes/no), history of diabetes (yes/no), history of angina (yes/no), previous MI (yes/no), prescription for ACE-inhibitors (yes/no), beta-blockers (yes/no), aspirin (yes/no), or statins (yes/no) at discharge.

  • bIncome hazard ratio adjusted for PM2.5.

Adjusting for small-area income deprivation substantially attenuated the PM2.5 association with mortality (Model 5, Table 3). The association between income deprivation and mortality was attenuated by smoking status and existing conditions like diabetes and previous MI (Model 3, Table 3). Further adjustment for PM2.5 only slightly attenuated the income deprivation and mortality association (Model 5, Table 3).

The fully adjusted PM2.5 results were not sensitive to the use of education or employment for adjustment of deprivation (Table 4) or an additional degree of freedom to model time. Less stringent control for differences across hospitals using a random intercept for hospital resulted in a slightly smaller HR compared with the main analysis.

View this table:
Table 4

Sensitivity of the PM2.5 associations with mortality following admission for acute coronary syndrome

Hazard ratio (95% CI) per 10 µg/m3 PM2.5
Fully adjusted model
 Education in place of income1.22 (1.06, 1.40)
 Employment in place of income1.20 (1.04, 1.38)
 Income with random effect for hospital1.17 (1.04, 1.33)
 Income with time trend using 3 df1.20 (1.04, 1.38)

We estimated that the mortality rate would be reduced by 12% if this cohort were exposed to naturally occurring PM2.5 rather than their modelled exposure. This translates to 4783 deaths brought forward due to exposure to PM2.5 from man-made sources.

Discussion

In this study of a large population of patients with advanced ischaemic heart disease, long-term exposure to air pollution was associated with all-cause mortality. The association was strongest for PM2.5 concentrations; there was no evidence of an association for the other pollutants. Whereas small-area income deprivation explained a substantial amount of the association between PM2.5 and mortality, exposure to PM2.5 explained little of the large socioeconomic gradient in mortality rate. The association between PM2.5 and mortality was not sensitive to adjustment for other measures of socioeconomic deprivation, including a random intercept for admitting hospital, or more flexibility in our model of time trends.

The primary strengths of our study are its large size, including a patient population from all of England and Wales, and detailed data on in-hospital treatments, discharge drugs, and patient characteristics which allowed for comprehensive confounder adjustment. One of the main limitations was the lack of cause-specific death, which limited our ability to analyse mortality outcomes in more detail. The majority of deaths among these ACS survivors is likely to be due to cardiovascular causes,7,31 although we cannot rule out that deaths influenced by air pollution were due to non-cardiovascular causes. The spatial resolution of the exposure model was fairly modest and corresponded to background concentrations, limiting our ability to investigate the role of local sources such as traffic. Exposure was assigned to individuals based on their postcode of residence at time of first admission. Complete data on residential history were not available; however, we explored the stability of residential postcodes over time among individuals who were re-admitted to hospital. There was no change in postcode for 75% of patients who were re-admitted, and the postcode centroid changed by ≤300 m for 90%. Although individual-level deprivation data were unavailable, we adjusted for area-level deprivation as well as for individual-level smoking and clinical history, which are likely to be important mediators of the relationship between an individual's socioeconomic position and prognosis. Several studies have shown that area-level measures of deprivation are more correlated with air pollution exposure, and comprehensive adjustment using area-level measures can essentially remove the correlation between individual-level deprivation and air pollution exposure.26,32 Although we adjusted for drugs prescribed at discharge, no data were available on drugs taken during the follow-up or for secondary prevention measures, which may have resulted in some residual confounding.

This study advances the field in several ways. First, our study provides the most comprehensive accounting for important differences in case management across hospitals through a combination of using detailed clinical data on treatments within the MINAP database and allowing each hospital to have its own baseline hazard. Differences across hospitals and potential clustering of patients within hospitals have not been accounted for in similar studies.3,5 Second, the most comparable study in terms of size assigned exposure at the county level using ambient monitors and was consequently restricted to urban areas.5 Third, the detailed data within MINAP allowed for comprehensive adjustment for confounding, particularly for smoking and discharge drugs which were not available for previous studies.3 Finally, our study provides evidence that particulate air pollution exposure does not contribute substantially to socioeconomic inequalities in post-MI prognosis. This hypothesis has been discussed in the literature, but not tested.33

Our results for PM2.5 are consistent with a broad body of evidence indicating that PM in this size range is especially relevant for cardiovascular mortality.3436 The biological plausibility of our findings is supported by accumulating evidence that pulmonary oxidative stress and inflammation in response to inhaled particles lead to systemic oxidative damage and inflammation37 as well as consequent endothelial dysfunction,38 increased thrombosis,39 and plaque vulnerability. Brief exposure to combustion-related particles among MI survivors has been shown to promote myocardial ischaemia and inhibit fibrinolytic capacity.40 Evidence from both human and animal studies suggests that long-term exposure to particulate air pollution enhances the progression and instability of underlying atherosclerosis via inflammatory processes, thereby promoting further ischaemic events.4143 This evidence is supported by epidemiological findings of an association with the occurrence of MI as well as cardiovascular mortality.35,44 Our findings of a positive association between long-term exposure to air pollution and mortality were specific to PM2.5, which was somewhat unexpected given that exposure to PM2.5 was highly correlated with PM10. The confidence intervals for the PM10 association were wider than those of NO2 and NOx, for which the evidence of a null association was clearer.

Studies with large numbers of events among the general population are required to answer the question of whether MI survivors are more susceptible to the effects of air pollution compared with the general population. We were not able to address this question in this study. If MI survivors are found to be more susceptible, this could have important public health implications. Thirty-day mortality for ACS has fallen year on year in England and Wales,45 leading to a growing number of MI survivors for whom air pollution may pose an elevated risk that receives relatively little attention in clinical settings.

In conclusion, we observed an association between long-term exposure to PM2.5 and all-cause mortality among patients with a previous ACS event. Exposure to air pollution explained relatively little of the large socioeconomic gradient in survival. The extent to which this population is at higher risk of death compared with the general population and implications for secondary prevention in this population requires further investigation.

Funding

C.T. was funded by the Economic and Social Research Council (RES-064-27-0026). This independent research was commissioned and funded by the Policy Research Programme in the UK Department of Health as part of the Health Effects of Outdoor and Indoor Air Pollution initiative (PR-AP-1107-10002). The Myocardial Ischaemia National Audit Project is commissioned and funded by the Healthcare Quality Improvement Partnership. The views expressed are not necessarily those of the Department. Funding for the Open Access was provided by the Economic and Social Research Council (RES-064-27-0026).

Conflict of interest: none declared.

Acknowledgements

We thank all those participating in the Myocardial Ischaemia National Audit Project (MINAP) and especially those staff responsible for data collection. We also thank Shakoor Hajat and Ben Armstrong, London School of Hygiene and Tropical Medicine, for helpful input on the statistical analysis, and AEA technology for assistance with the air pollution data.

This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by-nc/3.0/), which permits non-commercial use, distribution, and reproduction in any medium, provided that the original authorship is properly and fully attributed; the Journal, Learned Society and Oxford University Press are attributed as the original place of publication with correct citation details given; if an article is subsequently reproduced or disseminated not in its entirety but only in part or as a derivative work this must be clearly indicated. For commercial re-use, please contact journals.permissions@oup.com

References

View Abstract