rss
J Epidemiol Community Health 68:37-43 doi:10.1136/jech-2013-203116
  • Air pollution

A systematic review of air pollution and incidence of out-of-hospital cardiac arrest

  1. Judith Finn1,3,4,6
  1. 1Discipline of Emergency Medicine, The University of Western Australia, Crawley, Western Australia, Australia
  2. 2Combined Universities Centre for Rural Health, The University of Western Australia, Crawley, Western Australia, Australia
  3. 3Faculty of Health Sciences, Curtin University, Bentley, Western Australia, Australia
  4. 4St John Ambulance (WA), Belmont, Western Australia, Australia
  5. 5School of Population Health, The University of Western Australia, Perth, Western Australia, Australia
  6. 6School of Public Health and Preventive Medicine, Monash University, Melbourne, Victoria, Australia
  1. Correspondence to Dr Tiew-Hwa Katherine Teng, Discipline of Emergency Medicine, The University of Western Australia, Crawley, WA 6009, Australia; katherine.teng{at}uwa.edu.au
  • Received 18 July 2013
  • Revised 12 September 2013
  • Accepted 18 September 2013
  • Published Online First 7 October 2013

Abstract

Introduction Studies have linked air pollution with the incidence of acute coronary artery events and cardiovascular mortality but the association with out-of-hospital cardiac arrest (OHCA) is less clear.

Aim To examine the association of air pollution with the occurrence of OHCA.

Methods Electronic bibliographic databases (until February 2013) were searched. Search terms included common air pollutants and OHCA. Studies of patients with implantable cardioverter defibrillators and OHCA not attended by paramedics were excluded. Two independent reviewers (THKT and TAW) identified potential studies. Methodological quality was assessed by the Newcastle-Ottawa Scale.

Results Of 849 studies, 8 met the selection criteria. Significant associations between particulate matter (PM) exposure (especially PM2.5) and OHCA were found in 5 studies. An increase of OHCA risk ranged from 2.4% to 7% per interquartile increase in average PM exposure on the same day and up to 4 days prior to the event. A large study found ozone increased the risk of OHCA within 3 h prior to the event. The strongest risk OR of 3.8–4.6% per 20 parts per billion ozone increase of the average level was within 2 h prior to the event. Similarly, another study found an increased risk of 18% within 2 days prior to the event.

Conclusions Larger studies have suggested an increased risk of OHCA with air pollution exposure from PM2.5 and ozone.

Background

Out-of-hospital cardiac arrest (OHCA) is a significant public health issue affecting an estimated 310 000 Americans/year.1 While OHCA incidence and outcome vary around the globe,2 case fatality is consistently high, with overall survival to hospital discharge usually less than 10%.2 Thus, there is an imperative to better understand factors that ‘trigger’ the onset of cardiac arrest.

There is increasing evidence to support the association of ambient air pollution with overall cardiovascular mortality and morbidity.3–5 Exposure to higher than usual levels of airborne air pollutants over a few hours to several days has been reported to increase the risk of myocardial infarction,4 arrhythmia,6–8 stroke9–12 and heart failure,13–16 particularly in susceptible patients.12–18 Inconsistent results have been found in the relationship between OHCA and air pollution.3 However, there has not been a systematic review of these studies and the aim of our paper is to identify, evaluate and summarise the studies of air pollution and OHCA to examine the hypothesis that air pollution is associated with the incidence of OHCA.

Methods

Search strategy

A literature search was performed to identify studies that had analysed the association between OHCA attended by emergency medical services (EMS) and exposure to any air pollutant of interest (see Exposure section) in any lag period. The following bibliographic databases were searched (by authors THKT and TAW): MEDLINE (1946-February 2013); EMBASE (1980–February 2013), CINAHL (1982–February 2013), AUSTHealth (1997–February 2013) and the Cochrane Library (2004–February 2013). Scopus and Biosis Previews were searched for additional environmental science literature. Regional electronic bibliographic databases: Chinese Biomedical Literature Database (CBM), China Knowledge Resource Integrated Database (Cnki), CiNii (Japan), KoreaMed (Korea), IndMED (India) and LILACS (for Latin America and the Carribean) were also examined. Reference lists of relevant review articles and journals were hand-searched, and “Google” and “Google Scholar” search engines were used to search the internet.

Terms were mapped to the appropriate MeSH/EMTREE subject headings and “exploded”: [“cardiac arrest” OR “heart arrest” OR “sudden cardiac death”] AND [“air pollution” OR “air pollutants” OR “particulate matter” OR “airborne particles” OR “fine particles” OR environmental exposure” OR “soot” OR “elemental carbon” OR “ carbon monoxide” OR “nitrogen dioxide” OR “nitrogen oxides” OR “ozone” OR “sulfur dioxide” OR “ sulphur dioxide”] (see online supplementary appendix 1).

Study selection

Inclusion criteria were comparative studies and articles published in any language in peer-reviewed journals that examined the relationship between air pollution and OHCA in adults and children, including neonates. Exposures from ambient airborne pollutant levels: particulate matter (PM) <2.5 m in aerodynamic diameter (PM2.5), PM10, ultrafine particles (UFP), nitrogen oxides (NOx), ozone, sulfur dioxide (SO2) and carbon monoxide (CO) were included.

Animal studies, toxicological studies, summaries, commentaries, reviews, case reports, editorials, duplicates and articles only published in abstract form were excluded. Studies of patients with implantable cardioverter defibrillators and OHCA not attended by EMS personnel were also excluded. No time limit on journal publication date was set. Where there were multiple reports related to the same study, the most comprehensive publication was selected for inclusion.

All abstracts and titles were screened by two independent reviewers (THKT and TAW). Full-text articles of studies that met selection criteria were reviewed for eligibility for inclusion in the systematic review. If there was disagreement, arbitration was sought from a third reviewer (JF).

Data extraction

A standardised checklist was used to extract data from studies that met the inclusion criteria. Data collected were: study design, study setting, study population, demographic and baseline characteristics (including age, sex, aetiology, presenting cardiac rhythm if available and comorbidities), sample sizes, EMS, methodology, exposure (pollutants), exposure levels, number of monitoring sites, effect measurement, control conditions and outcome measurement. Where necessary, further information was sought from the lead authors of the studies by personal communication. The data extraction and risk of bias assessment were conducted independently by authors, THKT and TAW.

Risk of bias assessment

The Newcastle-Ottawa Scale19 was used to assess risk of bias in individual studies. Using the ‘star system’, each study was evaluated on three broad perspectives: (1) selection of the study groups; (2) comparability of the groups and (3) ascertainment of the exposure or outcome of interest. Study quality was graded as poor (1–3 stars), intermediate (4–6 stars) or high (7–9 stars). Emphasis was placed on the following components: ascertainment of OHCA occurrence (as a measure of outcome measurement); air-quality measurement (as a measure of exposure measurement noting potential exposure misclassification); representativeness of the study sample; and the extent of adjustment for confounders and sensitivity analysis undertaken.

Strategy for data synthesis

Narrative and tabular summaries of study characteristics were presented as guided by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement.20 The occurrence of OHCA attended by EMS personnel was the primary outcome. Summaries of exposure effects were reported using risk ratio as a measure of effect size. Statistical significance was defined as p<0.05. A protocol was prospectively registered online with PROSPERO—an international prospective register of systematic reviews.21

Results

Of the 849 studies identified by the systematic search, the following were excluded: 118 (duplicates); 335 (no relevance to the research question); 161 (examined air pollution and all-cause mortality, other cause-specific mortality or hospitalisations or were editorials or case studies); 16 (were related to infant mortality, sudden infant death syndrome, children's and fetal health and not specific to OHCA). Twenty-one citations were considered potentially eligible and full-text reports were retrieved. Of these, 9 studies that examined the association of ambient air pollution and ventricular arrhythmias detected by implantable cardioverter defibrillators, one study with abstract only, and another study which examined out-of-hospital coronary deaths (ICD 9 codes 410-414)22 were excluded. Three papers23–25 described different aspects of the same study, so were considered as one study. Finally eight studies were included in the review (see online supplementary table S1).

Study characteristics

Eight studies examined the association of specific air pollutants with OHCA,23 ,26–32 with a total of 38 060 OHCA cases (range 36223–11 67729). All were case-crossover studies.23 ,26–32 (see online supplementary table S1). Five studies were conducted in the USA,23 ,26–29 two in Europe (Denmark,32 Finland31) and one in Australia.30 There were no studies reported from the developing countries. Exposure levels and measurement techniques, study characteristics, EMS systems, sample size and reporting methods varied widely.

Studies examined multiple pollutants, (including UFP,31 ,32 PM2.5,23 ,26–29 ,31 ,32 PM10,23 ,28 ,30–32 PM2.5–10,28 ,30 ,31 CO,23 ,26 ,29–32 N0x,26 ,30 ,31 S0223 ,26 ,28 ,30 ,31 and ozone23 ,26 ,29–32), with the exception of Rosenthal et al27 that only examined ambient PM2.5 (table 1).

Table 1

Air pollutants, lags and effect measurement

Four studies23 ,28–30 restricted their study to adults, while four had no age restriction (see online supplementary appendix 2). Of the OHCA cases, mean age was 65 years, 64.5% were men and none were trauma-related. However, the study populations were heterogeneous and ranged from OHCA cases of presumed cardiac aetiology where resuscitation was attempted by paramedics,26 ,29 non-dead on arrival (DOA) cases,27 ,29 and OHCA cases identified from EMS data irrespective of the aetiologies,27 ,32 to a selective subpopulation of married residents whose spouses participated in a personal interview.23 Recent studies 26 ,29–32 had larger sample sizes than earlier studies.

Study quality

Six studies26 ,27 ,29–32 were graded high quality and two23 ,28 were graded intermediate quality (see online supplementary appendix 3).

Outcome data

Outcome data were derived mainly from the EMS records of the respective regions studied. OHCA cases were ascertained by reviewing medical records or death certificates,23 ,28 ,31 by study physicians23 ,26 ,31 ,32 or both,31 or restricted to non-DOA by EMS personnel.27 ,29

Exposure data

For each study, air pollution from ambient air monitoring sites was used as a proxy for individual exposure. Various methods were used for exposure ascertainment. Three different methods of particle mass measurement were used in the studies reviewed: nephelometry,23 ,28 β-attenuation instrument26 ,29 ,32 and tapered element oscillating microbalance (TEOM).27 ,30–32 Two studies,23 ,28 used the nephelometry method for estimation of the PM2.5 fraction. Wichmann et al32 used TEOM as well as β-attenuation instrument to measure exposure levels for PM2.5 and PM10, as well as .a differential mobility particle sizer to measure particle number concentration as a proxy for UFP. In two studies26 ,29 that examined ozone, exposure levels were the 8-h daily maximum, while the 24 h daily average level was used by other investigators.30–32

The number of air monitors used ranged from 127 ,30 to 3326 (for PM2.5); 130 to 328 for PM10; 130 to 22 for NO229; up to 14 for SO226, up to 1926 for CO and up to 4729 for ozone. Silverman et al26 and Ensor et al29 used the most extensive air-monitoring networks (see online supplementary table S1) to capture the exposure data of the pollutants of interest.

Exposure metrics for the pollutants were measured as average hourly or 24 h exposure for all pollutants (including PM2.5, PM10, UFP, NOx, SO2 and CO), with the exception of ozone. Ozone level was measured using an 8-h mean of daily maximum exposure metric26 ,29 or 24 h mean exposure.30–32

Mean exposure levels for PM2.5 ranged from 6.3530 to 18.423 µg/m3 and from 10.3132 to 31.923 µg/m3 for PM10. For ozone, it ranged from 13.3430 to 25.9432 ppb. Mean ozone level in Helsinki31 was reported as 46.8 g/m3 (equivalent to 23.87 ppb).

Lag periods varied from 0 to 6 days. Lag 0 h (or day) was the same hour (or same day) exposure as the day of OHCA event. Lag 6 days was exposure 6 days prior to the event.

Wichmann et al32 validated the results of the association of OHCA with PM using data from two detection systems for PM2.5 and PM10. Validation of the case-crossover results obtained using Poisson time series methodology was reported by three other studies.26 ,29 ,30

Although most studies used multivariable regression analyses, Checkoway et al23 did not adjust for temperature and relative humidity. Rosenthal et al27 did not adjust for effects of copollutants.

Overall results

Particulate matter

Five studies26 ,29–32 found significant positive associations between PM exposure (especially PM2.5) and OHCA, with increased OHCA risk ranging from 2.4% to 7% per IQ increase in average PM exposure on the same day up to 4 days prior to the event. However, three studies23 ,27 ,28 did not (figures 1 and 2).

Figure 1

Association of particulate matter 2.5 and out-of-hospital cardiac arrest. *, HR; **, relative risk.

Figure 2

Association of particulate matter 10 with out-of-hospital cardiac arrest.

Ozone

Five studies26 ,29–32 measured ozone, with positive associations observed in two.29 ,31 In a study with 11 677 OHCA cases using 47 monitors, Ensor et al29 found increased odds of OHCA even with short-term exposure (within 3 h prior to the event), with the highest increases of 3.8–4.6% per 20 ppb ozone increase from the average level in lags of 0–2 h. Rosenthal et al31 found the odds of OHCA increased by 18% (3.0–35%) in lag 2 days (2 days prior to the event; figure 3). Earlier studies26 ,30 did not find any association of ozone with OHCA. Wichmann et al32 only used ozone levels for statistical adjustment in the analysis of the association between OHCA and PM and found adjustment for ozone did not affect the relationship between OHCA and PM.

Figure 3

Association of ozone with out-of-hospital cardiac arrest. **, Relative risk.

Other pollutants

Pollutants other than PM and ozone were measured in seven studies but only Wichmann et al32 found positive associations between NOx and CO in lag 3 days with OHCA. Other studies did not find any association in any lag period for any gaseous pollutant.

Five studies26 ,27 ,29 ,30 ,32 examined the effects of air pollutants (specifically on PM2.5) for different age groups. Results were inconclusive, with conflicting results reported for the age groups examined. Two studies29 ,31 reported increased OHCA risk of 4.9–6% (95% CI 0.0% to 11.0%) in the age group 65–74 years for IQR increase of PM2.5. Rosenthal et al27 found a marked increase in OHCA risk (25% (95% CI 5.0% to 49.0%)) with PM2.5 for patients aged 60–75 years. By contrast Wichmann et al32 and Silverman et al26 found no age effect.

Reports of effect sizes were also heterogeneous; they included ORs, HRs, and relative risks based on interquartile increases in the 24 h mean pollutant level or increases of 10 g/m3 in the mean pollutant level. Therefore neither quantitative assessment of heterogeneity nor meta-analysis could be undertaken.

Discussion

Our systematic review evaluated evidence of the effect of ambient air pollution on the occurrence of OHCA attended by EMS personnel. We found an increased risk of OHCA with ambient PM in the majority of the studies,26 ,29–32 consistent with results of cardiorespiratory, cardiovascular and all-cause mortality from other studies over the past 20 years.3 ,4 ,33–35 Ozone was associated with OHCA in two of the five studies it was measured, while for other pollutants there were very few positive associations.

All of the studies were conducted in developed countries and measured pollution levels were often within the existing air quality standards for those countries. Despite air pollution being a major policy issue in many parts of Asia, and concentrations regularly exceeding those in developed countries, our literature search did not find any studies that examined the association of air pollution with the occurrence of OHCA from the developing countries: in the year 2000 two-thirds of 800 000 deaths and 4.6 million lost years of healthy life were attributed to air pollution.36

In a review of PM and CVD, the American Heart Association concluded there was sufficient evidence of a causal relationship between PM2.5 exposure and cardiovascular morbidity and mortality.3 Of the three23 ,27 ,28 studies that found no association of PM2.5 with OHCA, two23 ,28 used nephelometry to estimate the PM2.5 fraction. This method was not endorsed by the Environment Protection Agency37 as the data do not consistently reflect true PM levels and cannot be used to determine compliance to National Ambient Air Quality Standards. Furthermore, these studies23 ,27 ,28 had smaller sample sizes, lower concentrations of PM2.5 and/or different PM2.5 composition compared with later studies.

PM concentrations in Melbourne,30 Copenhagen32 and Helsinki31 were similar. Not surprisingly, the effect estimates of PM2.5 and PM10 with OHCA conducted in these cities were also similar. Although ozone was recognised to be an air pollutant that could increase the risk of mortality, until recently there was no reported association with OHCA.38 A positive association between ozone and OHCA was reported in two studies.29 ,31 Notably, Ensor et al29 collected the most comprehensive data to date on ozone in terms of study duration, number of pollution monitors and number of OHCA cases. The study results by Ensor et al29 and Rosenthal et al31 could possibly be attributed to a higher exposure concentration (mean 23.87–25.52 ppb) compared to Dennekamp30 (mean 13.34 ppb). Heterogeneity in the studies and interactions between air pollutants could account for the lack of association in Silverman et al26 (mean not reported).

Only one positive association of NOx or CO with OHCA was found in our review.32 Pollutants NOx, CO and SO2 have been reported to have moderately high collinearity.32 ,39 One large 17-city study, the China Air Pollution and Health Effects study,40 reported 10 g/m3 increase in 2-day moving average of NOx was associated with increase in cardiovascular and respiratory mortality, but did not specifically address the association of NOx with OHCA.

Exposure assessment for air pollution studies is challenging. Potential exposure misclassification or error can arise when ambient air pollution is used as a surrogate for individual exposure. Ambient pollution levels at a given location may not reflect an individual's true exposure. The cross-sectional relationship between personal exposure and fixed monitoring sites concentrations varies from poor41 to reasonably good,42 although correlations improve when assessed on a longitudinal basis.43 Furthermore, exposures to air pollutants take place over time and at multiple locations over a subject’s daily activities. It is thus not feasible to have accurate exposure estimates for individual participants. Zeger et al44 highlighted the difference between population-average exposure and air concentration as an important potential source of bias that would influence the magnitude of the observed effect.

The number of monitoring sites varied considerably between the studies and most used five or less. Increased aggregation of monitoring station data has been reported to improve the representativeness of the exposure estimates by decreasing exposure misclassification, which is more profound when using individual stations versus regional averages.45 Consequently, studies that utilised single sites for air pollution data assessment27 ,30 ,32 are more likely to suffer from potential exposure misclassification than those with multiple sites.

Ascertainment of particle mass measurement with three different methods: nephelometry,23 ,28 β-attenuation instrument26 ,29 ,32 and TEOM27 ,30–32; and the imprecision of exposure estimates also account for the degree of heterogeneity between studies. Regardless, Ensor et al29 and Silverman et al26 utilised the most extensive networks for air pollution data and their large studies demonstrated positive associations between air pollutants and OHCA.

In all the case-crossover studies, the time-stratified referent selection method was used to select control days, with the day of OHCA event as case day and the same day of the week in the same month and year as control days. This method minimises bias due to non-stationarity of air pollution time series data,23 and controls for confounding of exposure due to seasonal patterns.

The main reasons for downgrading of study quality were due to sample populations and exposure assessments. Two of these studies23 ,28 used nephelometric measures as a surrogate for PM2.5, and had limited generalisability. Checkoway et al23 was limited to 362 patients, a subset (but not a random sample) of all paramedic-attended OHCA at the study location. Sullivan et al28 comprised participants who belonged to the health maintenance organisation and were insured.

Our systematic review has several limitations. The high degree of heterogeneity between studies and reporting methods meant that a meta-analysis could not be undertaken. Eligible studies reviewed were few and all were observational with inherent bias of observational studies. Potential confounders, such as diurnal patterns,46–48 imprecision of exposure estimates and economic deprivation that increases the rates of morbidity and mortality attributed to air pollution49 were not considered.

The strength of our study lies in its comprehensive search of multiple databases, including regional bibliographic databases and its inclusion of multiple air pollutants. Our review identified a gap in the literature from developing countries. Future studies should be directed to understanding the relationship in high exposure areas where mean PM levels are 10 times higher36 than those reported in this review.

Conclusion

The associations between OHCA occurrence and short-term exposure of ambient PM and ozone are inconsistent, but recent studies have found positive associations. Large studies using similar sampling techniques are needed. Research is also needed to identify patients who are most at risk and to quantify personal exposure and risk.

What is already known on this subject

  • There has been increasing evidence to support the association of ambient air pollution with overall cardiovascular mortality and morbidity. However, inconsistent results have been found in the relationship between out-of-hospital cardiac arrest and air pollution.

What this study adds

  • Recent big studies have found positive associations of ambient particulate matter and ozone with out-of-hospital cardiac arrest. Large studies using similar sampling techniques are needed. Research is also needed to identify patients who are most at risk and to quantify personal exposure and risk.

Footnotes

  • Contributors All authors made substantial contribution with the revisions to the manuscript to improve the intellectual content and gave the final approval of the revised manuscript. More clearly defined roles are as follows: authors THKT, TAW and JF contributed to the conception, design, data extraction and reviews, analysis and interpretation of the data. JF and IJ and AT provided the clinical expertise. PF provided the air pollution expertise. AB assisted with the statistical advice and graphs. HT assisted with graphs and tables.

  • Funding This work was partially supported by the NH&MRC Centre for Research Excellence grant (CRE #1029983) for the Australian Resuscitation Outcomes Consortium (Aus-ROC).

  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.

References

Free sample

This recent issue is free to all users to allow everyone the opportunity to see the full scope and typical content of JECH.
View free sample issue >>

Don't forget to sign up for content alerts so you keep up to date with all the articles as they are published.