Objective To assess the relationship between fine particulate matter (PM2.5) concentration and ischaemic heart disease (IHD) morbidity and mortality.
Methods A time-series study conducted in Beijing from 1 January 2010 to 31 December 2012. Data on 369 469 IHD cases and 53 247 IHD deaths were collected by the Beijing Monitoring System for Cardiovascular Diseases, which covers all hospital admissions and deaths from IHD from Beijing's population of 19.61 million.
Results The mean daily PM2.5 concentration was 96.2 μg/m3 with a range from 3.9 to 493.9 μg/m3. Only 15.3% of the daily PM2.5 concentrations achieved WHO Air Quality Guidelines target (25 μg/m3) in the study period. The dose–response relationships between PM2.5 and IHD morbidity and mortality were non-linear, with a steeper dose–response function at lower concentrations and a shallower response at higher concentrations. A 10 μg/m3 increase in PM2.5 was associated with a 0.27% (95% CI 0.21 to 0.33%, p<2.00×10−16) increase in IHD morbidity and a 0.25% (95% CI 0.10 to 0.40%, p=1.15×10−3) increase in mortality on the same day. During the 3 years, there were 7703 cases and 1475 deaths advanced by PM2.5 pollution over expected rates if daily levels had not exceeded the WHO target.
Conclusions PM2.5 concentration was significantly associated with IHD morbidity and mortality in Beijing. Our findings provide a rationale for the urgent need for stringent control of air pollution to reduce PM2.5 concentration.
Statistics from Altmetric.com
Industrialisation has historically been accompanied by air pollution.1 With spectacular economic growth during the past three decades, China has become the largest energy consumer and the second-largest economy in the world. Meanwhile, air pollution has become a serious issue in China.2 From 21 February 2014 to 26 February 2014, heavy smog lingered over Beijing, and the government raised the city's smog alert to the second highest level for the very first time and maintained it for 132 h because the concentration of fine particulate matter (PM2.5)—that which is ≤2.5 µm in aerodynamic diameter—measured over 400 μg/m3.3
A meta-analysis that combined the results from 13 studies solely using either mortality or admission data reported that PM2.5 pollution was significantly associated with the risk of myocardial infarction.4 However, to fully reflect the association between PM2.5 and ischaemic heart disease (IHD), both non-fatal and fatal cases and both in-hospital and out-of-hospital deaths, should be considered. Moreover, this meta-analysis did not include studies from developing countries where PM2.5 pollution is more severe, so the association between PM2.5 and IHD at very high levels of pollution is still unclear.
A cohort study explored the exposure–response relationship for cardiovascular mortality in relation to PM2.5 from active smoking, second-hand smoke and ambient air pollution, and found that the exposure–response function is extremely steep at very low estimated PM2.5 exposure levels and flattens out at high levels.5 However, the shape of the dose–response relationship between PM2.5 and cardiovascular risk in a real and severe air pollution environment is still unclear.
Therefore, we conducted this study in Beijing to explore the short-term dose–response relationship between PM2.5 and IHD morbidity and mortality at city level. Such information is crucial for evidence-based policy making regarding clinical practice, public health, air pollution and even economic development.
PM2.5 and weather data
The US embassy Beijing Air Quality Monitor is atop the embassy building located in Chaoyang district, reporting hourly PM2.5 concentrations using an automatic Met One BAM-1020 β attenuation monitor (Met One Instruments, Grants Pass, Oregon, USA).6 ,7 A previous study has demonstrated that the data from the monitor exhibited approximately the same trend as citywide PM2.5 concentrations.8 In online supplementary figure S1, the station is at the centre of the red circle, with a radius of 40 km defined as previous studies.9 Coverage includes 79.2% of Beijing's total population and all areas of high population density (>5000 people/km2). The area also covers 97.8% (44/45) of the tertiary hospitals and 79.3% (69/87) of the secondary hospitals in Beijing that admit IHD cases. If the cumulative time of missing hourly PM2.5 data was <12 h during a day, that day was considered as qualified and the daily averaged PM2.5 concentration was calculated directly (1060 days). Otherwise, daily averages were calculated using linear interpolation (36 days).
Meteorological data on daily mean temperature, relative humidity and dew-point temperature were obtained from the China Meteorological Data Sharing Service System.10
The resident population of Beijing was 19.61 million, according to data from the 2010 National Population Census. IHD cases were identified by the Beijing Monitoring System for Cardiovascular Diseases which links the routinely collected records in the Beijing Hospital Discharge Information System and the Beijing Vital Registration Monitoring System.11 The Hospital Discharge Information System is operated by the Beijing Public Health Information Center and covers discharges from all government and private hospitals at secondary or tertiary level in Beijing.12 Hospital discharges of IHD were identified according to the primary discharge diagnosis with the International Classification of Diseases Tenth Revision (ICD-10) codes of I20-I25. The Vital Registration Monitoring System is managed by Beijing Center for Disease Control and Prevention and covers all deaths in Beijing.13 Deaths due to IHD, including both those in and out of hospital, were identified according to the underlying causes of death, with ICD-10 codes of I20–I25 and I46.
Cases of IHD were identified by linking records of hospital admissions and deaths. A total of 340 819 IHD admissions and 53 247 IHD deaths were collected by our monitoring systems during the 1096 days. Any hospital admissions or death records for the same patient that occurred within 28 days (n=24 597) were assumed to relate to one case and were excluded.14 ,15 Finally, 369 469cases of IHD morbidity were analysed. These included 199 209 cases of acute IHD (I20.0 and I21–I22), 111 851cases of chronic IHD (I25) and 58 409 cases of other types of IHD (I20.1, I20.8, I20.9 and I23–I24). Mortality outcomes included 53 247 IHD deaths, among which 13 867 and 39 380 deaths occurred in and out of hospital, respectively.
Daily data of IHD cases or deaths, PM2.5 concentration and weather variables were linked by date and, therefore, can be analysed with a time-series design. Because morbidity and mortality from IHD were rare, we fit the following generalised additive Poisson model to explore the association between PM2.5 and IHD:where E(Yt) represents the number of IHD cases or deaths at day t; the day of week was controlled for as a categorical variable; ps represents penalised spline function and β represents the log-relative risk of IHD morbidity or mortality associated with a unit increase of PM2.5. Relative risks of IHD morbidity and mortality with a 10 μg/m3 increase in PM2.5 concentration were calculated. Percentage change equals relative risk minus 1 and then multiplies by 100. The Z test was used to compare the two relative risks derived from subgroup analysis.16
Degrees of freedom (df) for calendar time, temperature and dew-point temperature were selected based on the parameters used in the previous studies,17 ,18 and were further tested by sensitivity analyses.
We used the smoothing function to graphically analyse the dose–response relationship between the log-relative risk of morbidity or mortality and PM2.5 concentration. To estimate the associations between PM2.5 concentration and IHD morbidity and mortality, we fitted the models with different lag structures from lag 0 day to lag 4 days. Considering that single-day lag models might underestimate the associations,19 we used the 3-day and 5-day moving averages of PM2.5 concentrations (lag 0–2 days and lag 0–4 days, respectively). We adjusted for temperature and dew-point temperature on the day when the cases or deaths occurred in the single-day lag models, and adjusted for the 3-day means of the two meteorological variables in the lag 0–2 days model, or the 5-day means in the lag 0–4 days model.
Potential autocorrelations in the case and death data were accessed by the plots of autocorrelation functions for the residuals from the current-day models, and no clear evidence of autocorrelation was found (see online supplementary figure S2).
The IHD cases and deaths advanced by PM2.5 pollution was estimated based on the population-attributable risk fraction (PARF), which was calculated as (relative risk–1)/relative risk, assuming that the prevalence of exposure to air pollution was 100%.
The following equations were used to calculate the IHD cases and deaths advanced by PM2.5 pollution over expected rates if daily levels had not exceeded the target:
IHD cases advanced by PM2.5 pollution=
IHD deaths advanced by PM2.5 pollution= where PMt is the PM2.5 concentration at day t. PARFlag0 is the PARF for lag 0 day; 337 and 49 is the average number of daily cases and deaths, respectively. When (PMt–Target) <0, the number of cases or deaths advanced by PM2.5 pollution at day t was set to 0.
All statistical analyses were performed using R Programming Language (V.3.0.2, R Development Core Team) using the NLME, MGCV and TSMODEL packages.
IHD morbidity and mortality
Of the 369 469 IHD cases that occurred during the study period, 59.9% of the data were male cases and 44.6% were under 65 years of age. Among 53 247 IHD deaths, 74.0% occurred out of hospital (table 1). On average, 337 IHD cases (range: 111 to 743) and 49 IHD deaths (range: 21 to 92) occurred per day.
PM2.5 concentration and weather conditions
During the 1096 days, the mean daily PM2.5 concentration was 96.2 μg/m3, with a range from 3.9 to 493.9 μg/m3 (table 2). Only 47.4% (519 days) of the daily PM2.5 concentrations achieved the target of Chinese Ministry of Environmental Protection 2010 (75 μg/m3), 22.2% (243 days) achieved the target of the US National Ambient Air Quality Standards (35 μg/m3) and 15.3% (168 days) achieved the target of WHO Air Quality Guidelines (25 μg/m3). Weather conditions are also shown in table 2.
Associations between PM2.5 concentration and IHD risk
There were clear dose–response relationships of PM2.5 concentration with IHD morbidity and mortality (figure 1). These relationships are non-linear with a steeper dose–response function at lower concentrations (0–75 μg/m3) and a shallower response at higher concentrations.
As shown in table 3, after adjusting for day of the week, seasonality and other time-varying influences and weather conditions, a 10 μg/m3 increase in PM2.5 was associated with a 0.27% (95% CI 0.21 to 0.33%, p<2.00×10−16) increase in IHD morbidity and a 0.25% (95% CI 0.10 to 0.40%, p=1.15×10−3) increase in IHD mortality on the same day. Significant lag associations of PM2.5 with IHD morbidity were observed at lag 1, 2 and 3 days, whereas, no significant lag association with IHD mortality was noted. Both 3-day and 5-day average concentrations significantly associated with IHD morbidity and mortality.
The current day and lag 0–2 days associations of PM2.5 with different types of IHD were all significant (figure 2). A significantly stronger lag association was found for chronic IHD cases versus acute cases (lag 3 days) and for non-fatal cases versus fatal cases (lag 1, 2, 3 and 4 days), while no significant difference was observed for the associations of PM2.5 with in-hospital deaths versus out-of-hospital deaths.
The associations of PM2.5 with IHD by age and gender
Among people aged ≥65 years, current-day, 3-day and 5-day average of PM2.5 were significantly associated with both IHD morbidity and mortality (figure 3). However, for those aged <65 years, significant associations were only found with IHD morbidity. A significantly stronger association of PM2.5 with IHD morbidity on the same day was found for people aged ≥65 years than for those aged <65 years (p=0.028). The associations of 3-day or 5-day average concentration with both IHD morbidity and mortality were significantly stronger among the older people (p<0.05).
The associations of current-day, 3-day and 5-day average of PM2.5 with IHD morbidity and mortality were all significant among men and women (figure 3). The current-day association with IHD morbidity was stronger in women than in men (p=0.019). However, no gender difference was observed for IHD mortality.
IHD cases and deaths advanced by PM2.5 pollution
As shown in table 4, there were 4099 cases and 855 deaths advanced by PM2.5 pollution over expected rates if daily levels had not exceeded the Chinese target (75 μg/m3) during the study period. The corresponding numbers were 6860 cases and 1312 deaths for the US target (35 μg/m3) and 7703 cases and 1475 deaths for WHO target (25 μg/m3).
Results of sensitivity analyses
Online supplementary figure S3 presents the percentage change in IHD morbidity and mortality associated with a 10 μg/m3 increase in PM2.5, under different df for calendar time, temperature and dew temperature. The estimate became stable when df for calendar time is closer to 10, while increasing df has little effect on the estimates for temperature and dew temperature, suggesting that the findings on the associations of PM2.5 concentration with IHD morbidity and mortality are robust.
To evaluate the potential effect of excluding the 24 597 cases reported within 28 days for the same patient, sensitivity analyses were conducted by including these cases. As shown in online supplementary table S1, similar results have been found.
This study provides strong evidence of the associations of PM2.5 with IHD morbidity and mortality in Beijing. These dose–response relationships are non-linear with a steeper dose–response function at lower concentrations and a shallower response at higher concentrations.
The exploration of the dose–response relationship between PM2.5 concentration and IHD risk is crucial to determine the pattern and scope of the adverse response. A cohort study conducted in 1.2 million American adults evaluated the exposure–response relationship for cardiovascular mortality in relation to PM2.5 from active cigarette smoking, second-hand smoke and ambient air pollution.5 The results suggested a relatively steep dose–response function at very low levels of PM2.5 and a flattening out at high levels, which is consistent with our results. To the best of our knowledge, our study provides the first city-level investigation of the dose–response relationship between PM2.5 pollution and IHD risk in a real and severe air pollution environment. The finding that an increase in PM2.5 at very low levels was sufficient to induce a significant adverse response on IHD morbidity and mortality implies that there is no threshold for safety of PM2.5 pollution. This result is supported by other time-series studies conducted in the USA17 ,20 and also supported by a recent cohort study conducted in Europe, which indicated that even in concentration ranges well below the present European annual mean limit value (<25 µg/m3), PM2.5 concentration is associated with natural-cause mortality and incidence of acute coronary events.21 ,22 These findings strongly support that policymakers would be justified in reducing targets for PM2.5 concentration. Although at higher PM2.5 concentrations the slope of the curve became shallower, the risk continued to rise as PM2.5 concentration was higher, implying that there is no saturation effect for the risk of IHD at a level of PM2.5 concentration as high as 500 µg/m3.
Several time-series studies conducted in Western countries have evaluated the association of PM2.5 and cardiovascular admissions, and the results are inconsistent. The Medicare Cohort Air Pollution Study (MCAPS) using Medicare files with 11.5 million individuals older than 65 years of age, reported that a 10 μg/m3 increase in PM2.5 concentration was related to 0.44% (95% CI 0.02% to 0.86%) increase in IHD admission,17 an association close to the finding for patients at the same age group in our study. However, the study conducted in Finland, including 20 007 admissions for coronary heart disease, found no association between PM2.5 and coronary heart disease admissions, probably due to the extremely low level of air pollution in that study.23
Possibly because of the higher level of PM2.5 pollution in Beijing, a longer lag association was noted in our study than the previous results, which reported a lag of 1–2 days.17 ,24 Interestingly, we found a significantly stronger lag association for chronic IHD cases versus acute cases and for non-fatal cases versus fatal cases, a difference which has not been reported previously as far as we know. A possible reason for this difference is that the time lag between symptoms onset and the hospitalisation in chronic IHD or non-fatal IHD cases was longer. Besides, the findings from a recent study may partly explain the difference from the perspective of underlying biological mechanisms. The study revealed that exposure to higher levels of air pollution lead to acute inflammatory and prothrombotic responses in the early lag period, followed by a gradual decrease in effect estimates to often negative effects due to compensatory mechanisms.25
The updated American Heart Association scientific statement on Particulate Matter Air Pollution and Cardiovascular Disease suggested that the elderly are a susceptible population; however, the evidence is limited, especially for morbidity.26 The MCAPS study found the association of PM2.5 with IHD admission rate was stronger among the population aged ≥75 years than among those aged 65–74 years, but that the difference was not significant.17 In this study, we provide clear evidence to support that people older than 65 years are more sensitive to PM2.5 pollution. These findings indicate that older persons are more sensitive to the elevated PM2.5 and should stay at home when PM2.5 concentration is extremely high. Using a face mask may also be considered, given the evidence that it could reduce symptoms and improve a range of cardiovascular health measures in patients with IHD.27 Furthermore, we found a stronger association of PM2.5 with IHD morbidity on the same day in women than in men. This is supported by the previous systematic review which reported that more studies of adults report stronger associations among women, possibly because of increased deposition of fine particles, higher airway hyper-responsiveness to oxidants or relatively lower socioeconomic status.28
Compared with previous studies, our study has several strengths. First, morbidity and mortality data were collected from an established monitoring system which covers all deaths and all hospital admissions of IHD in Beijing. The large number of patients allowed us to examine the associations at high levels of validity and reliability. Second, all hospital admissions and death records for the same patient that occurred within 28 days were excluded, so the potential dilution of the effect from interhospital transfers and rehospitalisations related to the same event was reduced as much as possible. Third, given the high level of pollution in Beijing, we were able to evaluate the dose–response relationship in a very wide range of PM2.5 concentrations.
This study also has some limitations. First, multiple comparisons have been conducted to evaluate the associations of PM2.5 with IHD among several subgroups. Therefore, false positive associations would likely have occurred simply by chance. Second, because public data from other stations were not available until 2013,29 our assessment of PM2.5 was derived entirely from the single monitoring station in the US embassy, Beijing. This station covers all areas with high density of population, but not some remote areas in Beijing. The inclusion of patients residing farther than 40 km from the monitoring station may result in exposure misclassification.
In conclusion, our results have demonstrated that PM2.5 concentration significantly associated IHD morbidity and mortality in Beijing. The dose–response relationships between PM2.5 and IHD morbidity and mortality were non-linear, with a steeper dose–response function at lower concentrations and a shallower response at higher concentrations. Stringent control of air pollution to reduce PM2.5 concentration is urgently needed.
What is already known on this subject?
Air pollution with particulate matter (PM2.5) ≤2.5 µm in aerodynamic diameter increases cardiovascular risk and is a serious problem in Beijing, China. However, most particle–outcome relationship studies in China focused solely on mortality.
What might this study add?
This study adds to the literature that a non-linear dose–response association exists between PM2.5 and ischaemic heart disease (IHD) morbidity and mortality, with a steeper dose–response function at lower concentrations and a shallower response at higher concentrations in an environment with high concentration of PM2.5, and provides a better understanding of the upper ranges of the dose–response curve.
How might this impact on clinical practice?
These findings indicate that measures should be taken by patients with IHD to prevent the harm of PM2.5 pollution. Patients with IHD, especially those older patients, should be advised to stay indoors or use a mask on days with unusually high air pollution levels.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online supplement
Contributors JL had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: W.X, GL, DZ, JL. Acquisition of data: WX, GL, XX, ZW, WL, GL, JL. Analysis and interpretation of data: WX, MW, JS, ZJ, QZ, JL. Drafting of the manuscript: WX, DZ, JL. Critical revision of the manuscript for important intellectual content: WX, GL, DZ, WW, JL. Statistical analysis: WX, JL. Administrative, technical, or material support: WX, GL, DZ, JL. Study supervision: JL.
Funding This work was supported by Capital Health Research and Development of Special (contract No. 2011-1005-01), the National Natural Science Foundation of China (contract No. 81302503), and the National Science & Technology Pillar Program During the Twelfth Five-Year Plan Period of China (contract No. 2011BAI08B01).
Competing interests None.
Ethics approval Our project ‘Morbidity and Mortality from Cardiovascular Disease in Beijing: a Surveillance Study’ (contract No. 2011-1005-01), has been approved by the Anzhen Hospital Ethics Committee.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.