PM2·5 is an important but modifiable environmental risk factor, not only for pulmonary diseases and cancers, but for cardiovascular health. However, the evidence regarding the association between air pollution and acute cardiac events, such as out-of-hospital cardiac arrest (OHCA), is inconsistent, especially at concentrations lower than the WHO daily guideline (25 μg/m3). This study aimed to determine the associations between exposure to ambient air pollution and the incidence of OHCA.
Methods
In this nationwide case-crossover study, we linked prospectively collected population-based registry data for OHCA in Japan from Jan 1, 2014, to Dec 31, 2015, with daily PM2·5, carbon monoxide (CO), nitrogen dioxide (NO2), photochemical oxidants (Ox), and sulphur dioxide (SO2) exposure on the day of the arrest (lag 0) or 1–3 days before the arrest (lags 1–3), as well as the moving average across days 0–1 and days 0–3. Daily exposure was calculated by averaging the measurements from all PM2·5 monitoring stations in the same prefecture. The effect of PM2·5 on risk of all-cause or cardiac OHCA was estimated using a time-stratified case-crossover design coupled with conditional logistic regression analysis, adjusted for daily temperature and relative humidity. Single-pollutant models were also investigated for the individual gaseous pollutants (CO, NO2, Ox, and SO2), as well as two-pollutant models for PM2·5 with these gaseous pollutants. Subgroup analyses were done by sex and age.
Findings
Over the 2 years, 249 372 OHCAs were identified, with 149 838 (60·1%) presumed of cardiac origin. The median daily PM2·5 was 11·98 μg/m3 (IQR 8·13–17·44). Each 10 μg/m3 increase in PM2·5 was associated with increased risk of all-cause OHCA on the same day (odds ratio [OR] 1·016, 95% CI 1·009–1·023) and at lags of up to 3 days, ranging from OR 1·015 (1·008–1·022) at lag 1 to 1·033 (1·023–1·043) at lag 0–3. Results for cardiac OHCA were similar (ORs ranging from 1·016 [1·007–1·025] at lags 1 and 2 to 1·034 [1·021–1·047] at lag 0–3). Patients older than 65 years were more susceptible to PM2·5 exposure than younger age groups but no sex differences were identified. CO, Ox, and SO2 were also positively associated with OHCA while NO2 was not. However, in two-pollutant models of PM2·5 and gaseous pollutants, only PM2·5 (positive association) and NO2 (negative association) were independently associated with increased risk of OHCA.
Interpretation
Short-term exposure to PM2·5 was associated with an increased risk of OHCA even at relatively low concentrations. Regulatory standards and targets need to incorporate the potential health gains from continual air quality improvement even in locations already meeting WHO standards.
Introduction
The adverse health effects of air pollution have been established in epidemiological studies over recent decades. At a global scale, ambient air pollution was responsible for 4·1 million deaths in 2016.
As the most extensively studied type of air pollution, PM2·5 is accepted to have causal associations with cardiovascular disease.
Several independent groups have observed cardiovascular outcomes associated with PM2·5 exposures lower than existing standards and guidelines such as the WHO daily 24-h average guideline value of 25 μg/m3.
Out-of-hospital cardiac arrest (OHCA) is a major medical emergency and public health problem, with global survival rates of less than 10%.
Growing evidence supports the associations between acute air pollution exposure and OHCA risk in case-crossover studies.
However, inconsistent results for PM2·5 and gaseous pollutants have also been reported, especially with PM2·5 levels lower than the WHO standard.
A study
of 5973 cases in Stockholm, for example, found no association between PM2·5 and OHCA where the average daily PM2·5 was 8·1 μg/m3, although others have found positive associations with similar PM2·5 exposure.
These results indicate that a larger-scale study will be helpful to elucidate associations in areas with low PM2·5 concentrations. In addition, the acute risks from air pollution have been well characterised for susceptible people such as those older than 75 years of age,
but less is understood about sex differences. A larger study could assist in characterising associations and detecting specific population subgroups for better protection.
Research in context
Evidence before this study
Out-of-hospital cardiac arrest (OHCA) is a major medical emergency and public health problem, with global survival rates at less than 10%. Given the dismal prognosis, preventive strategies are the best solution. In addition to the modification of individual risk factors, growing evidence supports concern about ambient air pollution and its association with cardiovascular diseases including OHCA. However, results regarding the short-term exposure to ambient PM2·5 and the incidence of OHCA have been inconsistent in previous studies, especially at PM2·5 levels lower than the WHO standard (25 μg/m3). Moreover, vulnerable groups such as older people who are exposed to air pollution consistently exhibit enhanced risk in epidemiological studies; however, less is known about sex differences. All main previous reports on the topic known to us are listed in the appendix (p 10).
Added value of this study
This nationwide case-crossover study in Japan shows an independent association between increase in daily PM2·5 exposure and risk of OHCA, even at PM2·5 levels lower than regulation standard and guidelines. The study includes 249 372 all-cause OHCAs, of which 149 838 were due to cardiac causes. No sex dimorphisms were observed. To our knowledge, this is by far the largest case-crossover study on this topic to estimate the risk and detect susceptible populations. Our data also show the association between short-term exposure to carbon monoxide, photochemical oxidants, and sulphur dioxide and increased risk of all-cause OHCAs, but not with nitrogen dioxide.
Implications of all the available evidence
Despite the air quality of the study generally meeting the current standards, significant associations between PM2·5 and OHCA were observed. These facts, combined with previous publications, suggest the requirement for world leaders, governments, and policy makers to reassess the current air quality standards and to improve air quality further. As no boundary exists in air quality between countries, a global approach to tackle this crucial health issue is necessary for our planet.
Thus, we did this nationwide case-crossover study in Japan with OHCA cases from the All-Japan Utstein registry to determine the associations between short-term exposure to ambient air pollution and the incidence of OHCA.
Methods
Study area and outcome data
Japan has a territory of 377 973·89 km2 with 47 prefectures, and had a population of 127·1 million in 2015.
The OHCA records we used were collected by the All-Japan Utstein registry of the Fire and Disaster Management Agency (FDMA) of Japan, from Jan 1, 2014, to Dec 31, 2015. The FDMA is responsible for emergency medical responses, and there were 752 fire stations with an ambulance dispatch centre in 2014.
Details of the registry have been described previously.
Briefly, it consists of a nationwide, prospectively collected repository of OHCA data, which is recorded according to the Utstein-style internationally standardised reporting guidelines.
Each record includes patient information about the location, date, sex, age, OHCA origin, and other core data elements, recorded by the local emergency medical service personnel managed by the FDMA.
Cardiac arrest was defined as the cessation of cardiac mechanical activity as confirmed by the absence of signs of circulation.
Cardiac arrests were classified as being of cardiac or non-cardiac origin. Non-cardiac origins were further classified into cerebrovascular disease, respiratory disease, malignant tumour, trauma, submersion, drug overdose, or any other non-cardiac causes. These origins were determined by the physicians in charge at the emergency medical service. This study was approved by the Tasmanian Human Research Ethics Committee (H0017657).
Ambient air pollution and meteorological data
Data on PM2·5 and gaseous pollutants including carbon monoxide (CO), nitrogen dioxide (NO2), photochemical oxidants (Ox), and sulphur dioxide (SO2) were obtained from the Environmental Database of the National Institute for Environmental Studies, Japan.
Each prefecture in Japan has several city-based PM2·5 monitoring stations. The daily PM2·5 for each prefecture was calculated by averaging the measurements from all stations in the prefecture. The differences between daily maximum and minimum PM2·5 concentrations from these stations within the same day and the same prefecture were also calculated. The means of these maximum differences within 1 year were then calculated from each prefecture for subsequent sensitivity analysis. Daily averages for each prefecture for gaseous pollutants were calculated from hourly concentrations of CO, NO2, Ox, and SO2.
Meteorological data were from Japan’s weather stations, the Local Meteorological Observatories, and included daily temperature and relative humidity. These were obtained for the study period from the Japan Meteorologic Agency website.
We calculated daily meteorological data for each prefecture by averaging daily mean temperature and relative humidity from all Local Meteorological Observatory stations each day, excluding those in non-habitable areas with extreme conditions (eg, on top of Mount Fuji). A correlation matrix was constructed between ambient air pollution and meteorological variables to avoid any potential collinearity issue.
Statistical analysis
The linked data were analysed using a time-stratified case-crossover design coupled with conditional logistic regression analysis. The time-stratified case-crossover design was first proposed in 1991 by Maclure
and is a well established study design to assess the association between transient exposure to air pollution and emergency health events; it is a common design in air pollution epidemiological studies.
The day the OHCA occurred is defined as the case day, whereas the days of the same day of the week during the same month and year (eg, all other Tuesdays of the same month and year) were controls. Each individual patient thus acts as their own control as the study design compares the exposure on the case day with exposure on the control days. This design has been verified as an efficient approach to control time-independent and time-dependent confounders in evaluating short-term health impacts associated with air pollution.
The main analysis was the multivariable model of the effects of PM2·5. Odds ratios (ORs) and their 95% CIs were estimated using a conditional logistic regression model by merging the air quality and meteorological data at the prefecture level with each OHCA case. Since the response period of the incidence of OHCA after exposure to air pollution is unknown, the analyses were done with lagged exposures for 0–3 days, on the basis of previous case-crossover studies.
Lag 0 was the average exposure concentration on the day of the OHCA, lag 1 was the average exposure concentration on the day before the OHCA, and so forth. The moving average of 0–1 day (lag 0–1) was calculated by averaging the exposure of the day of the OHCA episode and the day preceding the OHCA. The average exposure of 3 days before the OHCA and the day of the OHCA was defined as lag 0–3. Meteorological data including temperature and relative humidity during the same lag period were used as potential confounders in the multivariable model. We modelled temperature and relative humidity as a natural cubic spline with three degrees of freedom to test the possible non-linear effects of meteorological variables. We chose other degrees of freedom as sensitivity analyses. Single-pollutant models for the effects of CO, NO2, Ox, and SO2 on OHCAs were run separately to the main analysis, and two-pollutant models were also investigated for PM2·5 with other gaseous pollutants (CO, NO2, Ox, and SO2).
To better understand the association between ambient air exposure and different origins of OHCAs, separate analyses were done for all-cause OHCAs and the arrests presumed to have cardiac causes in both the single-pollutant and two-pollutant models. Additional stratification for sex and age (≤35 years, 36–64 years, 65–74 years, and ≥75 years) were also investigated for PM2·5. Furthermore, we did sensitivity analyses for the prefectures that had an annual mean maximum difference in PM2·5 concentrations of less than 10 μg/m3, as well as four specific prefectures reflecting various sources of pollutants (Tokyo and Osaka as major cities with pollutants from vehicles, and Hyogo and Nagasaki as coastal cities with pollutants from ships). ORs were expressed as per unit change in CO and per 10-unit change in PM2·5, NO2, Ox, and SO2. A correlation analysis was done to describe the relationship between ambient air pollutants and meteorological variables, expressed with the Pearson correlation coefficient r. All analyses were done using R version 3.5.0.
Role of the funding source
There was no funding source for this study. BZ and KN had access to all the data. All authors were responsible for the decision to submit the manuscript.
Results
249 372 OHCAs were documented during the 2 years of study (table 1). The mean age of patients affected was 74·4 years (SD 17·4) and 57% were male. 149 838 (60·1%) of the included OHCAs were presumed to have cardiac origin. Of 99 534 patients with OHCAs of non-cardiac origin, 20 857 arrests were caused by respiratory diseases and 8030 were caused by cerebrovascular diseases (table 1). 245 614 (98·5%) of the OHCAs occurred while PM2·5 concentrations were lower than the Japan or USA standards of 35 μg/m3, and 229 233 (91·9%) occurred at concentrations lower than the WHO standard of 25 μg/m3 (figure 1).
Table 1Characteristics of the study participants with OHCAs, by year
2014 (n=125 951)
2015 (n=123 421)
Total (n=249 372)
Age, years
74·3 (17·4)
74·6 (17·3)
74·4 (17·4)
≤35
4991 (4·0%)
4744 (3·8%)
9735 (3·9%)
36–64
21 865 (17·4%)
20 704 (16·8%)
42 569 (17·1%)
65–74
22 622 (18·0%)
22 302 (18·1%)
44 924 (18·0%)
≥75
76 473 (60·7%)
75 671 (61·3%)
152 144 (61·0%)
Sex
Female
54 206 (43·0%)
53 000 (42·9%)
107 206 (43·0%)
Male
71 745 (57·0%)
70 421 (57·1%)
142 166 (57·0%)
OHCA origin
Cardiac
76 141 (60·5%)
73 697 (59·7%)
149 838 (60·1%)
Non-cardiac
49 810 (39·5%)
49 724 (40·3%)
99 534 (39·9%)
Cerebrovascular diseases
4085 (3·2%)
3945 (3·2%)
8030 (3·2%)
Respiratory diseases
10 296 (8·2%)
10 561 (8·6%)
20 857 (8·4%)
Malignant tumours
3779 (3·0%)
3904 (3·2%)
7683 (3·1%)
External causes
8592 (6·8%)
7803 (6·3%)
16 395 (6·6%)
Other non-cardiac causes
23 058 (18·3%)
23 511 (19·0%)
46 569 (18·7%)
Data are mean (SD) or n (%). Each year contributed approximately 50% to the total for each category. OCHA=out-of-hospital cardiac arrest.
Summary statistics of daily ambient air quality and meteorological data show median PM2·5 levels of 11·98 μg/m3 (IQR 8·13–17·44) and NO2 levels of 8·55 parts per billion (ppb; IQR 5·91–12·37; figure 1; table 2). The number of monitoring stations and mean of maximum differences of PM2·5 concentrations for each prefecture are shown in the appendix (p 2), alongside a scatterplot matrix of the air pollutants and meteorological data (appendix p 25). PM2·5 concentrations were moderately or weakly positively correlated with other air pollutants (r <0·43). Ox was moderately negatively correlated with CO, whereas NO2 showed a weak positive correlation with SO2 (appendix p 25). Temperature and humidity were weakly correlated with air pollutants; humidity was negatively correlated whereas the direction of correlation oscillated for temperature. Therefore, these variables were put into the same model.
Table 2Description of daily ambient air pollutants and meteorological data for the days of OHCAs
Days missing data
Mean (range)
Percentile
IQR
5%
25%
50%
75%
95%
PM2·5 (μg/m3)
11
13·62 (0·27 to 71·01)
4·61
8·13
11·98
17·44
28·37
9·31
CO (ppm)
35
0·35 (0·0 to 1·41)
0·18
0·27
0·33
0·41
0·57
0·14
NO2(ppb)
2
9·75 (0·0 to 45·02)
3·45
5·91
8·55
12·37
20·19
6·46
Ox (ppb)
2
30·51 (2·22 to 79·20)
13·50
22·96
29·93
37·31
49·53
14·35
SO2(ppb)
2
1·83 (0·0 to 106·75)
0·38
0·93
1·52
2·37
4·28
1·44
Temperature (°C)
0
15·2 (−10·2 to 32·1)
1·7
8·0
16·2
22·0
27·5
14·0
Humidity
0
72% (0 to 100)
50%
64%
73%
81%
91%
17%
34 310 datapoints were included for each parameter. CO=carbon monoxide. NO2=nitrogen dioxide. Ox=photochemical oxidants. SO2=sulphur dioxide. ppm=parts per million. ppb=parts per billion. OHCA=out-of-hospital cardiac arrest.
When examining the association between daily lag exposure to PM2·5 and OHCA, positive associations were observed for all lags, with an increased risk of all-cause OHCA ranging from OR 1·015 (95% CI 1·008–1·022) at lag 1 to 1·033 (1·023–1·043) at lag 0–3 per 10 μg/m3 increase in PM2·5 exposure, and an OR of 1·016 (1·009–1·023) on the same day (ie, lag 0; figure 2; appendix p 5). Similarly, significant associations were shown between cardiac OHCA and PM2·5 exposures at all lags, with an increased risk ranging from 1·016 (1·007–1·025) at lags 1 and 2 to 1·034 (1·021–1·047) at lag 0–3 per 10 μg/m3 increase in PM2·5. Effect size estimates were generally greater for average exposure over lag 0–3 than for individual lags or lag 0–1.
Figure 2Association of OHCA with daily lag exposure to PM2·5
These findings were corroborated in sensitivity analyses for prefectures with an annual mean maximum difference in PM2·5 concentrations of less than 10 μg/m3. The excess risk of OHCA associated with PM2·5 exposure at all lags remained similar to those for the whole population for all-cause OHCA (ranging from OR 1·017 [95% CI 1·005–1·030] at lag 1 to 1·038 [1·021–1·056] at lag 0–3) and cardiac OHCA (ranging from 1·019 [1·003–1·035] at lag 0 to 1·039 [1·017–1·062] at lag 0–3; appendix p 3). Similar effects sizes but fewer significant associations (due to smaller sample sizes and statistical power) between PM2·5 and OHCA were observed in Tokyo, Osaka, and Hyogo (appendix p 3). Significant and robust estimates (OR 1·040–1·065) were seen for lag 0–3 in each representative prefecture, except for Nagasaki, which had fewer OHCAs (n=2560) than the other representative prefectures (ranging from 9625 cases in Hyogo to 25 387 in Tokyo; appendix p 3).
In the stratified analysis, both sexes had similar effect estimates for the association between short-term PM2·5 exposure and all-cause OHCA, with increased risk ranging from OR 1·015 (1·006–1·024) at lag 1 to 1·032 (1·019–1·045) at lag 0–3 for men and from 1·014 (1·004–1·025) at lag 1 to 1·035 (1·020–1·050) at lag 0–3 for women, although the association was not significant for women at lag 0 (figure 3; appendix p 5). Among patients older than 65 years, PM2·5 exposures over all 4 days considered were significantly associated with incidence of all-cause OHCA (figure 3). The estimates of increased risk were similar between those aged 65–74 years and those aged 75 years or older (figure 3). No significant associations were detected for patients younger than 64 years, except for the patients aged 36–64 years at lag 3 (appendix p 5).
Figure 3Association of all-cause and cardiac-cause OHCA with daily lag exposure to PM2·5 by sex and age
Similar findings were observed for cardiac OHCAs (figure 3). No sex-attributable dimorphism was seen in the excess risks for OHCA with cardiac origin (ranging from OR 1·016 [95% CI 1·005–1·028] at lag 2 to 1·035 [1·018–1·052] at lag 0–3 in men and from 1·015 [1·002–1·029] at lags 1 and 2 to 1·033 [1·014–1·053] at lag 0–3 in women, again with no significant association for women at lag 0; appendix p 6). Similar to all-cause OHCA, no associations between short-term PM2·5 exposure and cardiac OHCA were identified for patients younger than 65 years. Significant associations were found in patients aged 75 years or older across all 4 days. In those aged 65–74 years, however, excess risks of cardiac OHCA were only significant at lags 0, 0–1, and 0–3, although the effect sizes were similar to those of all-cause OHCA (figure 3; appendix p 6).
In the stratification by age and sex, we also did sensitivity analyses for all-cause OHCA and cardiac OHCA in prefectures with annual mean maximum differences in PM2·5 concentrations of less than 10 μg/m3 (appendix p 4). The results followed a similar pattern to the main analyses, with slightly higher risk estimates but fewer significant associations. For all-cause OHCA, increased risks were found in men (ranging from OR 1·019 [95% CI 1·003–1·036] at lag 1 to 1·040 [1·017–1·063] at lag 0–3) and patients aged 75 years or older (from 1·022 [1·007–1·038] at lag 1 to 1·051 [1·029–1·073] at lag 0–3) over all 4 days; increased risks were also found for women at lags 2, 3, and 0–3, ranging from 1·022 (1·004–1·041) at lag 2 to 1·036 (1·011–1·062) at lag 0–3 (appendix p 4). For cardiac OHCA, similar associations were observed with PM2·5 between sexes, although fewer were significant (appendix p 4). Significant associations ranged from 1·023 (1·002–1·045) at lag 2 to 1·040 (1·011–1·071) at lag 0–3 for men and occurred only at lag 3 (1·026 [1·002–1·051]) and lag 0–3 (1·037 [1·004–1·072]) for women (appendix p 4). Increased risks were still found across all 4 days for patients older than 75 years, ranging from 1·025 (1·006–1·045) at lag 1 to 1·055 (1·028–1·083) at lag 0–3 (appendix p 4).
In the single-pollutant models of other pollutants, significantly positive effects were observed for CO, Ox, and SO2 on all-cause OHCA. Each 1 parts-per-million increase in CO was associated with the increased risks of OHCA, ranging from OR 1·080 (95% CI 1·023–1·140) at lag 1 to 1·185 (1·095–1·284) at lag 0–3 (figure 4). Each 10 ppb increase in Ox was associated with increased OHCA risk at lags 1, 0–1, and 0–3, from 1·009 (1·003–1·016) at lag 1 to 1·011 (1·001–1·021) at lag 0–3. For SO2, these increased risks were present at lags 1, 3, 0–1, and 0–3, ranging from 1·071 (1·023–1·121) at lag 1 to 1·167 (1·075–1·267) at lag 0–3 per 10 ppb increase (figure 4). Interestingly, NO2 exposure was negatively associated with all-cause OHCA at lag 1 (figure 4; appendix p 7).
Figure 4Association of OHCA with daily lag exposure to gaseous pollutants
Among patients with OHCA of cardiac origin, the risk estimates for CO mostly became stronger, except for at lag 2, which was similar to all-cause OHCAs (figure 4). For SO2, the only increased risk was 1·122 (1·008–1·247) at lag 0–3. No significant associations were found for Ox with cardiac OHCA (figure 4; appendix p 7). NO2 had negative association at lag 2 (figure 4; appendix p 7).
In two-pollutant models for all-cause and cardiac OHCAs, the association of PM2·5 remained significant over all 4 days, except for at lags 0 and 1 for cardiac OHCAs in the PM2·5 plus CO model, and the associations of CO, Ox, and SO2 disappeared (appendix pp 8–9, 26). Greater risk estimates were observed for PM2·5 when models included NO2 compared with the single pollutant models, whereas the association of NO2 remained negative (appendix pp 8–9, 26).
Sensitivity analyses achieved by varying the degrees of freedom for the temperature and relative humidity in the above models confirmed our results (appendix pp 13–24). No differences in statistical significance were observed among different degrees of freedom.
30262-1&id=gr1.jpg){kind=link}
30262-1&id=gr2.jpg){kind=link}
30262-1&id=gr3.jpg){kind=link}
30262-1&id=gr4.jpg){kind=link}