Cardiovascular diseases and particularly coronary heart disease (CHD) associated with coronary atherosclerosis are the most common cause of death worldwide.1 The first presentation with CHD is fatal in at least 25% of people, as a result of sudden arrhythmic death or a fatal myocardial infarct2; therefore, identifying the triggers for this first presentation is of great importance for public health.
There has been increased awareness of the potential importance of environmental factors for CHD events. Such factors include environmental tobacco smoke and, more recently, air pollution. Small particles have consistently been shown to be associated with cardiovascular mortality and hospitalization for cardiac events.3–7 However, studies investigating the association between air pollution and acute cardiac events give less consistent results.8–15 This, as the Committee on the Medical Effects of Air Pollutants16 has suggested, may be due to low levels of pollution in the cities where the studies have taken place in combination with low statistical power. The Committee recommended that further studies of acute cardiac events are required. The current study has a much larger study population, with lower air pollution concentrations than in the previous studies.
Particles with aerodynamic diameters smaller than 2.5 μm (PM2.5) or 10 μm (PM10) seem to be primarily responsible for the observed cardiovascular effects of air pollution. A review of several studies found that cardiovascular death rates increase by about 1% for every 10 μg/m3 increase in PM2.5.17 This increase is due not only to the deaths of critically-ill people who would have died in the near future regardless of exposure,16 but also to the likely underlying susceptibility related to atherosclerosis, which is frequently subclinical and asymptomatic in its early stages. Therefore, to best inform public health measures, it is useful to understand the triggers that determine whether on any given day, the disease remains silent or presents with an acute coronary event or sudden arrhythmic death.
Schwartz18 examined the relationship between airborne particles and daily deaths in 10 US cities, and found a higher risk of out-of-hospital death compared with death in-hospital. Because CHD is the most common cause of death in the US, this observation suggests that out-of-hospital cardiac arrests would be an important health outcome to consider when investigating the role of air pollutants, but little is known about this relationship.
In this study, we aimed to further clarify the association between outdoor air pollution and sudden cardiac events, using a large, comprehensive data base with information on out-of-hospital cardiac arrest.
Study Population and Outcome Data
Melbourne is a large city with a population of about 3.8 million. We included cases whose residential postcodes were partially or completely within the 2006 urban growth boundary of Metropolitan Melbourne.
The subjects comprised those who had a paramedic-attended out-of-hospital cardiac arrest in the Melbourne metropolitan area between January 2003 and December 2006 inclusive. The Victorian Ambulance Cardiac Arrest Registry captures all cardiac arrests in Melbourne attended by the Ambulance service. It is one of the largest out-of-hospital registries in the world. Information extracted from the registry included date of birth, sex, date and time of event, postcode of the event, postcode of the home address of the patient, arrest location (eg, house, street or public place) etc. On average about 3350 cardiac arrests are attended each year by the Melbourne Ambulance Service.
We restricted the cardiac arrests to those that occurred within metropolitan Melbourne to persons who were at least 35 years of age living in metropolitan Melbourne. We excluded arrests for which the precipitating event was identified by the ambulance personnel as clearly not being cardiac, leaving those presumed to have cardiac etiology (n = 8434). We thus excluded 2117 arrests that were reported to have been precipitated by noncardiac events: trauma (n = 308), including road traffic accidents, falls, stabbing, shooting, and industrial accidents; overdose or poisoning (n = 306); terminal illness (n = 740); respiratory (n = 265); neurologic (n = 50); hanging or drowning (n = 325); unknown (n = 62), and others (n = 61). We included only people age 35 years or over, because younger people have relatively more cardiac arrests for reasons other than CHD (such as genetic diseases that predispose to fatal arrhythmias and hypertrophic or dilated cardiomyopathy). We restricted the study area to metropolitan Melbourne because of the location of the air pollution monitoring site.
Ambient Air Pollution and Meteorology Data
Ambient air pollution and meteorologic data were collected over the same period as outcome data. We obtained data for daily average PM2.5, PM10, ozone (O3), carbon monoxide (CO), nitrogen dioxide (NO2), and sulfur dioxide (SO2) from the EPA Victoria at one central monitoring station in Melbourne. Although there are 2 EPA sites in Melbourne collecting PM2.5 data, we included data only from the inner East monitoring site. More than 25% of data during the study period were missing from the inner West. Furthermore, the PM2.5 concentrations showed a very high correlation of 0.95 between the 2 sites. PM10 and PM2.5 were measured by tapered element oscillating microbalances. The data for the coarse fraction PM2.5–10 were obtained by subtracting the daily average PM2.5 concentration from the PM10 concentration. PM10 and PM2.5 levels in Melbourne were generally well below the Australian National Environment Protection Measure for Ambient Air Quality standards. Daily average observations of temperature and humidity were obtained from the Bureau of Meteorology from a monitoring site at Melbourne Airport.
A case-crossover design19 was used with a time-stratified referent period to select control days associated with each index case, where case day was the day the arrest occurred. Time was stratified into months and day of week, and the reference day associated with the date of each index case was used as the referent control day(s) within the same month (eg, exposure of a case on a Monday in January was compared with exposures on all other Mondays in January). By using this approach rather than the symmetric bi-directional design, we eliminated confounding by day of week, monthly trends, and seasonal and long-term trends in the exposure variables.20,21 We calculated the Spearman rank correlations between pollutants using the difference of the concentrations on the index case day and the average of the associated reference control days.
The occurrence of an out-of-hospital cardiac arrest was the outcome variable. The primary exposure variables were daily average PM2.5, PM10, PM2.5–10, O3, CO, NO2, and SO2. We included daily average temperature and relative humidity as potential confounders. Conditional logistic regression models were used to evaluate the association between the pollutants and cardiac arrest. As a first step, we developed single-pollutant models separately and then included confounders, retaining those that made a significant contribution to the most parsimonious models. Parameter estimates from these models (calculated from the hazard ratio for the interquartile range of the pollutant) may be interpreted as proportional increases in risk.
We also developed lagged models using this approach. We developed models for single-day lags (0–3 day lags, where lag 0 was the average concentration on the day of the arrest, lag 1 the average concentration on the day prior to the arrest, etc.) and for the average of the day of the arrest and the day previous to the arrest (lag 0–1). Lags were investigated to allow for possible latency following exposure. Data were stratified by sex to test for interaction. In addition the data were analyzed by 3 age groups (35–64 years, 65–74 years, 75 years and older).
As an analysis of the sensitivity of observed effects to different choices of analytic methods, we fit time-series models to the association of particulates and out-of-hospital cardiac arrest incidence. A semiparametric Poisson regression model was used to model out-of-hospital cardiac arrest using a generalized additive model (GAM), where the penalized regression splines were used to estimate smoothing spline.22 In all models we included variables for day-of-the-week effects, a smooth function for the day of the week (times) and weather confounders such as relative humidity and daily average temperature.
We created a variable to represent “hot” days (defined as maximum temperature >35°C) and “cold” days (defined as minimum temperature <3°C) and fitted lags of 0–1 for hot days and lags of 0–6 for cold days, to investigate the effect of heat and cold.
All analyses were performed using Stata Release 9.023 and R software version 188.8.131.52
Study Population and Exposure Characteristics
A total of 8434 paramedic-attended out-of-hospital cardiac arrests in Melbourne over the 4-year study period were included in this analysis—an average of 5.8 arrests per day. The mean (±SD) age of the study population was 72 (±14) years, with a maximum of 103 years. There were almost twice as many arrests among men (n = 5447, 65%) compared with women (n = 2987, 35%). The average age at which the arrest occurred in women was 75 years, almost 5 years higher than for men. Table 1 summarizes the distribution of daily average values for air pollution and meteorologic variables during the study period.
The interquartile range of daily average values for all pollutants was similar to the interquartile range of the differences in values between the day of the arrest and their associated control day (Table 1). The concentration on the case days for PM2.5 were on average 0.16 μg/m3 higher than on the control days; for PM10 the concentration on the case days were on average 0.23 μg/m3 higher than on the control days
Air Pollution and Out-of-Hospital Cardiac Arrests
Prior to fitting case-crossover models, we fitted GAMs to evaluate the functional form between each pollutant, meteorological variable, and out-of-hospital cardiac arrest. The estimated effect for temperature was 1.001 degrees of freedom, suggesting a linear functional form. The results of the case-crossover analysis are presented in Table 2. Of all the air pollutants, the strongest associations were found with PM10 and PM2.5. No association was found with the coarse fraction of PM10 (PM2.5–10). The increase in risk associated with PM2.5 on the day of the arrest (lag 0) was estimated at 2.44% (95% CI = 0.54% to 4.37%) per 4.26 μg/m3 (IQR) increase in PM2.5. For the previous day (lag 1) this increase in risk was 2.46% (0.33% to 4.65%) and for the average of lag 0 and lag 1 this increase in risk was 3.61% (1.29% to 5.99%). Longer lag periods for PM did not show such strong relationships. There was also an association between CO and cardiac arrests, but not as strong as for PM2.5. There was no clear relationship for O3, SO2, or NO2 in any lag period.
We developed 2-pollutant models for PM2.5, PM10, and PM2.5–10 with NO2, O3 and CO. Of all the particulate fractions, PM2.5 had the strongest association with out-of-hospital cardiac arrest. The gaseous pollutants did not modify this effect (Fig. 1). Because the concentrations of SO2 were extremely low in Melbourne, we did not include this pollutant in two-pollutant modeling analyses. The correlations between PM2.5 and the gaseous pollutants were 0.49, 0.13 and 0.55 for NO2, O3 and CO, respectively.
Subgroup analyses (Fig. 2) showed higher risks in men than in women. The increase in risk of an out-of-hospital cardiac arrest for an interquartile range increase in PM2.5 for men was 4.00% (95% CI = 1.18% to 6.90%) and for women 2.77% (−1.28% to 6.99%). The analysis by age showed highest rise among 65–74 year olds and lowest among those over 75 years. Further investigation of the age effect by sex found that the effect in the 65–74 year olds was stronger in men (5.60% [0.33 to 11.15]) compared with 3.21% [−4.87 to 11.96] in women).
The results from the sensitivity analysis using GAM models confirmed the results from the case cross-over analysis—that is, PM2.5 was associated with incidence of out-of-hospital cardiac arrest.
We fit lags of 0–1 for hot days and lags of 0–6 for cold days; the only association was for cold days at lag 4. None of these additional results changed the estimated effect of PM2.5.
Previous research has shown an association of hourly PM2.5 exposure with out-of-hospital cardiac arrest.25 In our data, this effect persisted for up to 2 days following an increase in PM2.5. The particle fraction responsible for the observed effects was likely to be PM2.5 because no association was found with the coarse fraction (PM2.5–10) and a slightly stronger association was found with PM2.5 than with PM10. CO was also associated with out-of-hospital cardiac arrest, but not as strongly. The estimates for O3 were large but the confidence intervals were wider than for the other pollutants, likely due to a higher degree of measurement error. NO2 and SO2 did not show an association with out-of-hospital cardiac arrest, although the levels of SO2 in this study were very low. Men showed a slightly stronger association than women, and those age 64–74 years were slightly more susceptible to the effects of PM2.5.
A study from Boston11 investigated the association between PM2.5 and the occurrence of a myocardial infarction (MI) by identifying 772 participants on average 4 days after the MI. The study did not find an association for the 24 hours immediately before the MI, but it did for 24–48 hours before. The OR was about 1.25 with an increase of 20 μg/m3 PM2.5. We found a 2.16% increase in out-of-hospital cardiac arrest the day before the arrest for an IQR increase, which would correspond to an increase of 10.55% for an increase of 20μg/m3. However these 2 studies are not directly comparable. For example, the time scales and cardiac outcome were somewhat different—we looked at the day prior to out-of-hospital cardiac arrest and the Boston study looked at 24–48 hours prior to MI. Another difference is that our study (where only 16% survived to hospital) included all events whereas the patients in the Boston study had to survive at least 4 days. Methods similar to those used in the Boston study were carried out in King County, Washington, with 5793 cases who had MI; no consistent association with particulate air pollution was found.14
To date, only 4 studies have specifically investigated air pollution and out-of-hospital cardiac arrests or cardiac deaths. Two small case-crossover studies in Washington State, did not find any association between increased levels of PM2.5, PM10, SO2, or CO and out-of-hospital cardiac arrests.26,27 A case-crossover study from Rome found that particle number concentration, PM10 and CO were associated with out-of-hospital coronary deaths, but PM2.5 was not measured.9 A study from Indianapolis did not find any association with PM2.5 on the day of the arrest or 1–3 days before the arrest, although it did find an association with the PM2.5 concentration during the hour of (witnessed) out-of-hospital cardiac arrest.25 The present study is the first to suggest an effect between the concentrations of fine airborne particles on the day prior to and on the day of arrest and occurrence of an out-of-hospital cardiac arrest (regardless of whether the person survived).
Air-pollutant concentrations in the Rome study were on average 3 times higher than in Melbourne, although the investigators did not measure PM2.5. The authors found a 6.1% increase in out-of-hospital coronary deaths for an increase of 29.7 μg/m3 PM10 for Lag 0–1.9 This increase of PM10 would correspond to a 10.1% increase in out-of-hospital cardiac arrest in the present study. However, the study from Rome investigated out-of-hospital coronary deaths whereas the present study looked at out-of-hospital cardiac arrests. During our study period, 84% of out-of-hospital cardiac arrest patients did not make it to the emergency department alive. Both studies showed clear associations between particulate air pollution and cardiovascular outcomes, providing further evidence for the importance of fine particulate air pollution as a possibly preventable trigger of acute out-of-hospital cardiac events.
Furthermore, in subgroup analyses by age, both the Rome study9 and ours suggest that people age 65–74 years are most susceptible to particulate air pollution. This may reflect the more frequent and extensive CHD among older people and the competing causes of death such as cancer in the very elderly. Alternatively the apparently lower risk of air pollution in the oldest age group might reflect survivor bias. However our findings should be interpreted with caution as confidence intervals are large.
Melbourne has relatively low levels of air pollution compared with other major cities. In Seattle and Indianapolis, the average 24-hour air pollutant concentrations were at least twice as high as Melbourne's, and no associations with out-of-hospital cardiac arrests were found in these areas.10,25,27 However the statistical power of these studies was lower. The differences in associations found between air pollution and out-of-hospital cardiac arrests in different countries could also be due to differences in chemical composition of particulate matter. Bell et al28 investigated the associations of chemical composition of PM2.5 and hospital admissions in 106 counties in the US; they found that the difference in the risk was in part explained by the chemical composition of the particulate matter, with vanadium, elemental carbon, or nickel explaining most of the difference. However, the main source of nickel and vanadium in PM2.5 is oil combustion, and this is unlikely to play a larger role in Melbourne than in other parts of the world.
Effects observed at low PM2.5 concentrations such as occur in Melbourne should be taken into account in setting particulate air pollution standards, especially since in Australia the enforceable 1-day standard for PM10 is currently 50 μg/m3, but there is only an advisory standard for PM2.5 of 25 μg/m3. The present study suggests an increase in the risk of cardiac effects at concentrations below the current air quality standards in Australia.
PM2.5 remained strongly associated with out-of-hospital cardiac arrest regardless of whether NO2, CO or O3 was included in the model. High correlation between pollutants make it difficult to interpret two-pollutant models, although in this study the concentrations on the case day and their related control days (as defined in the case-crossover analysis) was low for PM2.5 and O3 (P = 0.13). The correlation was higher between PM2.5 and CO (P = 0.55) and PM2.5 and NO2 (P = 0.49). Melbourne experiences extremely low levels of SO2, which means that confounding by this pollutant is unlikely.
A major strength of this study was the large sample size of 8434 cardiac arrests with presumed cardiac etiology. The Victorian Ambulance Cardiac Arrest Registry is one of the largest such registries in the world. The Registry includes information on every incident of out-of-hospital cardiac arrest attended by ambulance in Melbourne. The large size and comprehensive nature of the Registry minimized the risks of selection bias.
A limitation of our study was the lack of information on personal risk factors such as smoking, alcohol intake, and obesity. In addition we had no information on comorbidities such as hypertension, chronic obstructive pulmonary disease, etc. However, in a case-crossover study this is less of a problem; individual risk factors are controlled by the design, with cases serving as their own controls. We also do not know whether the person had previously been diagnosed with a coronary heart disease and was already on medication, or whether the cardiac arrest was the first presentation of CHD. It has been suggested that those who are already treated for CHD are less susceptible to the effects of pollution.29 While the same person could have been included in the database more than once, this was unlikely because the survival rate is low. In Melbourne, only 6% of adults who have an out-of-hospital cardiac arrest with a presumed cardiac etiology survive to hospital discharge (approximately 1 month after the event), according to data from Ambulance Victoria.
Another limitation of this study (and with most other air pollution studies) is the use of a central monitoring station to estimate exposure for the entire Melbourne population. People with an arrest no doubt varied in their exposure depending on distance from the central monitoring station and time spent indoors. However, an advantage of our data is that we did not need to estimate or impute data, as we had a very complete dataset with measured daily average concentrations over a 4-year period.
In summary, our results suggest that exposure to particulate air pollution, especially PM2.5, is associated with an increase in the occurrence of out-of-hospital cardiac arrests in Melbourne.
We thank the Ambulance Victoria, EPA Victoria, CSIRO Marine and Atmospheric Sciences and the Bureau of Meteorology for providing the data for this study.
1.WHO. Preventing chronic diseases: a vital investment. Geneva, Switzerland: World Health Organisation; 2005.
2.Myerburg RJ, Kessler KM, Castellanos A. Sudden cardiac death: epidemiology, transient risk, and intervention assessment. Ann Intern Med
3.Dominici F, Peng RD, Bell ML, et al. Fine particulate air pollution and hospital admission for cardiovascular and respiratory diseases. JAMA
4.Franklin M, Zeka A, Schwartz J. Association between PM2.5 and all-cause and specific-cause mortality in 27 US communities. J Expo Sci Environ Epidemiol
5.Pope CA, Burnett RT, Thurston GD, et al. Cardiovascular mortality and long-term exposure to particulate air pollution: epidemiological evidence of general pathophysiological pathways of disease. Circulation
6.Samet JM, Zeger SL, Dominici F, Curriere F, Coursac I, Dockery DW. The national morbidity, mortality, and air pollution study. Part I: Methods and methodologic issues. Res Rep Health Eff Inst
. 2000;94(pt 1):5–14; discussion 75–84.
7.Zanobetti A, Schwartz J. Air pollution and emergency admissions in Boston, MA. J Epidemiol Community Health
8.Checkoway H, Levy D, Sheppard L, Kaufman J, Koenig J, Siscovick D. A case-crossover analysis of fine particulate matter air pollution and out-of-hospital sudden cardiac arrest. Res Rep Health Eff Inst
. 2000;99:5–28;discussion 29–32.
9.Forastiere F, Stafoggia M, Picciotto S, et al. A case-crossover analysis of out-of-hospital coronary deaths and air pollution in Rome, Italy. Am J Respir Crit Care Med
10.Levy D, Sheppard L, Checkoway H, et al. A case-crossover analysis of particulate matter air pollution and out-of-hospital primary cardiac arrest. Epidemiology
11.Peters A, Dockery DW, Muller JE, Mittleman MA. Increased particulate air pollution and the triggering of myocardial infarction. Circulation
12.Peters A, Liu E, Verrier RL, et al. Air pollution and incidence of cardiac arrhythmia. Epidemiology
13.Rich KE, Petkau J, Vedal S, Brauer M. A case-crossover analysis of particulate air pollution and cardiac arrhythmia in patients with implantable cardioverter defibrillators. Inhal Toxicol
14.Sullivan J, Sheppard L, Schreuder A, Ishikawa N, Siscovick D, Kaufman J. Relation between short-term fine-particulate matter exposure and onset of myocardial infarction. Epidemiology
15.Vedal S, Rich K, Brauer M, White R, Petkau J. Air pollution and cardiac arrhythmias in patients with implantable cardioverter defibrillators. Inhal Toxicol
16.COMEAP. Cardiovascular disease and air pollution: A report by the Committee on the Medical Effects of Air Pollution. London: Department of Health; 2006.
17.Pope CA, Dockery DW. Health effects of fine particulate air pollution: lines that connect. J Air Waste Manag Assoc
18.Schwartz J. Assessing confounding, effect modification, and thresholds in the association between ambient particles and daily deaths. Environ Health Perspect
19.Maclure M. The case-crossover design: a method for studying transient effects on the risk of acute events. Am J Epidemiol
20.Bateson TF, Schwartz J. Control for seasonal variation and time trend in case-crossover studies of acute effects of environmental exposures. Epidemiology
21.Bateson TF, Schwartz J. Selection bias and confounding in case-crossover analyses of environmental time-series data. Epidemiology
22.Wood SN, Augustine NH. GAMs with integrated model selection using penalized regression alpines and applications to environmental modelling. Ecol Modell
23.Stata Statistical Software
[computer program]. Release 9. College Station, TX: StataCorp LP; 2003.
24.R [computer program]. Version 2.8.0. Vienna, Austria: R Foundation for Statistical Computing; 2004.
25.Rosenthal FS, Carney JP, Olinger ML. Out-of-hospital cardiac arrest and airborne fine particulate matter: a case-crossover analysis of emergency medical services data in Indianapolis, Indiana. Environ Health Perspect
26.Levy D, Lumley T, Sheppard L, Kaufman J, Checkoway H. Referent selection in case-crossover analyses of acute health effects of air pollution. Epidemiology
27.Sullivan J, Ishikawa N, Sheppard L, Siscovick D, Checkoway H, Kaufman J. Exposure to ambient fine particulate matter and primary cardiac arrest among persons with and without clinically recognized heart disease. Am J Epidemiol
28.Bell ML, Ebisu K, Peng RD, Samet JM, Dominici F. Hospital admissions and chemical composition of fine particle air pollution. Am J Respir Crit Care Med
29.Barclay JL, Miller BG, Dick S, et al. A panel study of air pollution in subjects with heart failure: negative results in treated patients. Occup Environ Med