Size-fractioned particulate air pollution and cardiovascular emergency room visits in Beijing, China

Environmental Research 121 (2013) 52–63

Contents lists available at SciVerse ScienceDirect

Environmental Research journal homepage: www.elsevier.com/locate/envres

Size-fractioned particulate air pollution and cardiovascular emergency room visits in Beijing, China e ¨ Liqun Liu a,b,c,n, Susanne Breitner a,b, Alexandra Schneider a, Josef Cyrys a,d, Irene Bruske , Ulrich Franck f, f f f,g h Uwe Schlink , Arne Marian Leitte , Olf Herbarth , Alfred Wiedensohler , Birgit Wehner h, Xiaochuan Pan c, H-Erich Wichmann b,e, Annette Peters a,i a

Helmholtz Zentrum M¨ unchen—German Research Center for Environmental Health, Institute of Epidemiology II, Neuherberg, Germany Ludwig-Maximilians-Universit¨ at M¨ unchen, IBE-Chair of Epidemiology, Munich, Germany c Peking University, Health Science Center, School of Public Health, Beijing, China d University of Augsburg, Environmental Science Center (WZU), Augsburg, Germany e Helmholtz Zentrum M¨ unchen—German Research Center for Environmental Health, Institute of Epidemiology I, Neuherberg, Germany f Helmholtz Center for Environmental Research—UFZ, Core Facility Studies, Leipzig, Germany g University of Leipzig, Leipzig, Germany h Leibniz Institute for Tropospheric Research (IfT), Physics Department, Leipzig, Germany i Focus Network Nanoparticles and Health (NanoHealth), Helmholtz Zentrum M¨ unchen—German Research Center for Environmental Health, Neuherberg, Germany b

a r t i c l e i n f o

a b s t r a c t

Article history: Received 12 January 2012 Received in revised form 10 October 2012 Accepted 23 October 2012 Available online 30 January 2013

Background: Although short-term exposure to ambient particulate matter has increasingly been linked with cardiovascular diseases, it is not quite clear how physical characteristics of particles, such as particle size may be responsible for the association. This study aimed at investigating whether daily changes in number or mass concentrations of accurately size-segregated particles in the range of 3 nm– 10 mm are associated with daily cardiovascular emergency room visits in Beijing, China. Methods: Cardiovascular emergency room visit counts, particle size distribution data, and meteorological data were collected from Mar. 2004 to Dec. 2006. Particle size distribution data was used to calculate particle number concentration in different size fractions, which were then converted to particle mass concentration assuming spherical particles. We applied a time-series analysis approach. We evaluated lagged associations between cardiovascular emergency room visits and particulate number and mass concentration using distributed lag non-linear models up to lag 10. We calculated percentage changes of cardiovascular emergency room visits, together with 95% confidence intervals (CI), in association with an interquartile range (IQR, difference between the third and first quartile) increase of 11-day or 2-day moving average number or mass concentration of particulate matter within each size fraction, assuming linear effects. We put interaction terms between season and 11-day or 2-day average particulate concentration in the models to estimate the modification of the particle effects by season. Results: We observed delayed associations between number concentration of ultrafine particles and cardiovascular emergency room visits, mainly from lag 4 to lag 10, mostly contributed by 10–30 nm and 30–50 nm particles. An IQR (9040 cm  3) increase in 11-day average number concentration of ultrafine particles was associated with a 7.2% (1.1–13.7%) increase in total, and a 7.9% (0.5–15.9%) increase in severe cardiovascular emergency room visits. The delayed effects of particulate mass concentration were small. Regarding immediate effects, 2-day average number concentration of Aitken mode (30–100 nm) particles had strongest effects. An IQR (2269 cm  3) increase in 2-day average number concentration of 30–50 nm particles led to a 2.4% ( 1.5–6.5%) increase in total, and a 1.7% (  2.9–6.5%) increase in severe cardiovascular emergency room visits. The immediate effects of mass concentration came mainly from 1000–2500 nm particles. An IQR (11.7 mg m  3) increase in 2-day average mass concentration of 1000–2500 nm particles led to an around 2.4% (0.4–4.4%) increase in total, and a 1.7% (  0.8–4.2%) increase in severe cardiovascular emergency room visits. The lagged effect

Keywords: Accurately size-fractioned ambient particulate matter Cardiovascular emergency room visits Beijing Distributed lag non-linear model Particle number concentration Ultrafine particles

n

¨ Corresponding author at: Helmholtz Zentrum Munchen—German Research Center for Environmental Health, Institute of Epidemiology II, Neuherberg, Germany. E-mail addresses: [email protected] (L. Liu), [email protected] (S. Breitner), ¨ [email protected] (A. Schneider), [email protected] (J. Cyrys), [email protected] (I. Bruske), [email protected] (U. Franck), [email protected] (U. Schlink), [email protected] (A. Marian Leitte), [email protected] (O. Herbarth), [email protected] (A. Wiedensohler), [email protected] (B. Wehner), [email protected] (X. Pan), [email protected] (H.-E. Wichmann), [email protected] (A. Peters). 0013-9351/$ - see front matter & 2012 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.envres.2012.10.009

L. Liu et al. / Environmental Research 121 (2013) 52–63

53

curves of number and mass concentrations of 100–300 nm particles or 300–1000 nm particles were quite similar, indicating that using particulate number or mass concentrations seemed not to affect the cardiovascular effect (of particles within one size fraction). The effects of number concentration of ultrafine particles, sub-micrometer particles (3–1000 nm) and 10–30 nm particles were substantially higher in winter comparing with in summer. Conclusions: Elevated concentration levels of sub-micrometer particles were associated with increased cardiovascular morbidity. Ultrafine particles showed delayed effects, while accumulation mode (100–1000 nm) particles showed immediate effects. Using number or mass concentrations did not affect the particle effects. & 2012 Elsevier Inc. All rights reserved.

1. Background Ambient particulate matter (PM) has increasingly been linked with cardiovascular diseases during the last decades (Brook et al., 2004). Observed effects of short-term fluctuations of ambient PM on the cardiovascular system include ischemia and arrhythmia in patients with coronary artery disease, altered heart rate and autonomic control, altered blood pressure, systemic inflammatory response, a pro-thrombotic state and endothelial dysfunction (Peters, 2005). The exposure-response functions between shortterm exposure to PM and cardiovascular mortality and morbidity are generally considered to be near-linear and without a threshold (Samoli et al., 2005), indicating that even quite low concentrations of PM could result in adverse influence. Different size fractions of PM are under investigation: PM with an aerodynamic diameter smaller than 10 mm (PM10) or 2.5 mm (PM2.5 or fine particles), PM with diameters between 2.5 mm and 10 mm (PM10 2.5 or coarse particles), as well as particles smaller than 100 nm (PM0.1 or ultrafine particles) (Brook et al., 2004). Both subcategories of PM10 (coarse and fine particles) showed short-term effects on cardiovascular mortality or morbidity (Brunekreef and Forsberg, 2005). Some findings suggested that the associations were stronger for finer than for coarser particles, or the effects of coarse particles on short-term total mortality no longer existed after adjusting for fine particles in a two-pollutant analysis (Brunekreef and Forsberg, 2005; Kan et al., 2007; Peng et al., 2008; Wichmann et al., 2000). A possible explanation for the stronger associations between fine particles and cardiovascular diseases is that they are deposited on the bronchial tree and alveoli; ultrafine particles are even able to cross over into the bloodstream (Ljungman, 2009). The direct stimulation of blood vessels, as well as the particle-induced pulmonary reflexes and pulmonary inflammation, eventually leading to arrhythmia or myocardial ischemia, are all considered as possible biological mechanisms linking PM with cardiovascular diseases (Brook et al., 2004). Ultrafine particles are most often measured by number per cubic centimeter (Brook et al., 2010); their high particle number concentration and large active surface area (plus small size), thus high deposition efficiency in the pulmonary region, make them a great contributor to the observed cardiovascular effects (Delfino et al., 2005; Pekkanen and Kulmala, 2004). PM10, fine and coarse particles are typically measured by their mass per volume of air (mg m  3) (Brook et al., 2010). Meteorological conditions (such as air humidity) can modify the spectrum of the size fractions with consequences to health effects (Leitte et al., 2009). Due to the limited availability of appropriate measurement data, there are only few epidemiological studies on the shortterm effects of more accurately size-segregated particles on daily cardiovascular mortality or morbidity (Atkinson et al., 2010; Branis et al., 2010; Halonen et al., 2009; Peters et al., 2009; ¨ Stolzel et al., 2003). Moreover, in a global context, most studies about health effects of smaller particles have been conducted in North America or Europe; there is a relative lack of studies on Asian population. Different air pollution mixtures and levels in

Asian areas might influence the associations between human health and air pollution. This study aimed at investigating whether daily changes in ambient concentrations of particle size fractions in the range of 3 nm–10 mm are associated with cardiovascular emergency room visits in Beijing, China. Moreover, we aimed to better delineate whether using particle number or mass concentration may affect the associations.

2. Material and methods 2.1. Study area and period We conducted this study in Beijing, China, from 4 Mar 2004 to 31 Dec 2006 (1033 days). Beijing has an area of about 16,808 km2 consisting of eight urban and ten suburban districts (Fig. 1), with a population size of approximately 15,380,000 in 2005 (http://baike.baidu.com/view/2621.htm). It is located in the North China Plain surrounded by mountains of 1000–1500 m in altitude to the west, north, and northeast, and the Bohai Sea on the southeast side. Typical warm temperate semihumid continental monsoon climate leads to hot, humid summers and cold, dry winters. Springs and autumns are both of relatively short duration. 2.2. Emergency room visits data We obtained standard medical record forms from the Emergency Department of Peking University’s Third Hospital, which is located in the Haidian District (Fig. 1). Only the forms of patients who visited and left the emergency room within one day were available; the forms of patients with more severe problems who were transferred to the In-patient Department were not available for our

Fig. 1. Beijing (shaded area is the urban area) and the locations of data sources.

54

L. Liu et al. / Environmental Research 121 (2013) 52–63

database as their forms were transferred together with the patients. A database including diagnosis and date of visit based on information in the forms was built. The diagnoses were coded according to the International Classification of Diseases 10th Revision (ICD-10) by trained fourth grade or senior medical students in Peking University Health Science Center, who had medical knowledge and mostly had hospital internship experience. In order to assure the accuracy of coding, afterwards a nosological expert from the Third Hospital was asked to code all the diagnoses appeared to make a standard diagnosis-code list. According to the list the original percentage of misclassification was about 4%. Furthermore, about 20% (randomly chosen 4–5 days per month) of the emergency room visit information were double entered; the identity of the numeric content between the two entries was higher than 90%. We divided cases into sub-categories of ischemic heart diseases (I20–I25), arrhythmia (I47–I49), heart failure (I50), cerebrovascular diseases (I60–I69), and other diseases within the I00–I99 category as well as cause-unknown sudden death (R96). We combined all cases as ‘‘total circulatory emergency room visits’’. Moreover, we combined the first four sub-categories (cases of ischemic heart diseases, arrhythmia, heart failure, and cerebrovascular diseases), which are more severe or even fatal, as ‘‘severe cardiovascular emergency room visits’’. 2.3. Meteorological data We obtained meteorological data including daily mean temperature, relative humidity, and barometric pressure from the China Meteorological Data Sharing Service System (station 54511, located at N391480 E1161280 in the south eastern part of Beijing within Daxing District, see Fig. 1). 2.4. Particulate concentration data Particle size distribution data were sampled on top of a six-floor building inside the campus of Peking University, which is also located in the Haidian District (Fig. 1). The setup of the measurement station is described in detail elsewhere (Wehner et al., 2004 and 2008). The measurement station is a few hundred meters away from major roads (no heavy traffic) and about 20 m above ground. The campus is a primarily residential and commercial area without industrial sources or agricultural activities. Local emission sources within a radius of 1 km could be vehicular traffic, fuel combustion for domestic cooking and heating, and construction. An earlier examination of the spatial variability of PM2.5 mass and chemical composition in 1999–2002 showed only minor differences between the campus site and a downtown site (Wehner et al., 2004 and 2008; Leitte et al., 2011). Furthermore, average particle number size distributions at the Peking University measurement site and another regional measurement site, located around 50 km to the south of the Peking University, were shown to be similar in summer (Yue et al., 2009), confirming that the measurement site may be considered as an urban background station. Aerosol number size distributions were continuously measured between 3 nm and 10 mm. Measurements were done by a Twin Differential Mobility Particle Sizer which covered the size range from 3 nm to 800 nm (mobility diameter) (Birmili et al., 1999), and an Aerodynamic Particle Sizer (TSI model 3221) which covered the size range from 800 nm to 10 mm (aerodynamic diameter). The data were corrected for losses due to diffusion and sedimentation within the inlet line as described by Wehner et al. (2004). Data on number size distributions were used to calculate particle number concentrations (cm  3). These data were converted to particle mass concentrations (mg m  3) assuming spherical particles with a mean particle density of 1.5 g cm  3. The assumption of this density was based on previous measurements of chemical compositions of particles in Beijing (Wehner et al., 2008; Leitte et al., 2011). We calculated daily mean number and mass concentrations for the size fractions (nm) 3–10, 10–30, 30–50, 50–100, 100–300, 300–1000, 1000–2500 and 2500–10,000; also for 3–100 (ultrafine particles), 3–1000 (PM1 or sub-micrometer particles), 3–2500 (PM2.5) and 3–10,000 (PM10). In our study, the number concentration of particles 41000 nm and the mass concentration of particles o 100 nm were quite low; in other words, the contribution to particle number by 1000–2500 nm and 2500–10,000 nm particles, as well as the contribution to particle mass by 3–10 nm, 10–30 nm, 30–50 nm and 50–100 nm particles, were quite low in comparison to other fractions. Therefore, we excluded those fractions from the analyses. We also did not investigate the effects of number concentration of PM2.5 and PM10 (as they would be nearly the same as the effect of number concentration of PM1), as well as the effect of mass concentration of ultrafine particles. 2.5. Statistical analyses We used generalized semi-parametric Poisson regression to model the natural logarithm of the expected daily emergency room visit counts as a function of the predictor variables. Data were analyzed using the package ‘‘mgcv’’ version 1.7–12 and the package ‘‘dlnm’’ version 1.6.2 in the statistical software R version 2.14.1 (R Development Core Team, 2011).

In the beginning of the statistical process, a confounder model was built for total or severe cardiovascular emergency room visits, respectively, without particle exposure. First, to control for long-term trend and seasonality in the emergency room visits data, we used a penalized spline of date order. The smoothing basis was thin plate regression spline (same for all the other penalized splines); the dimension of the basis (k value) was decided by minimizing the absolute value of the sum of the partial autocorrelation function of the model’s residuals up to 30 lags (Touloumi et al., 2006). This minimization of the partial autocorrelation function of the residuals intends to minimize the correlation between residuals from proximate observations in the data series, to match the standard assumption of uncorrelated residuals (Gasparrini and Armstrong, 2010). Second, to control short-term periodicity in the emergency room visits data, we used dummy variables for day of the week (Monday to Sunday). Third, to ensure sufficient adjustment of the effects of meteorological confounders, we used penalized splines of daily mean air temperature and relative humidity. The dimension of the basis (k value) was set to 4; the lag of temperature and humidity was chosen based on the minimization of generalized cross validation criteria (Simon, 2006) of the model. The minimization of generalized cross validation criteria aims to maximize the ability of the model to predict new observations arising from the same phenomenon which produced the data (Gasparrini and Armstrong, 2010). Air temperature of the previous day and relative humidity of the current day (at which the cardiovascular event happened) were included in the cofounder model for total cardiovascular emergency room visits; while both air temperature and relative humidity of the current day were included in the confounder model for severe cardiovascular emergency room visits. Besides the four confounders above, dummy variables for public holidays (holiday and nonholiday) and penalized spline of daily mean barometric pressure were considered as well, but only public holidays was included as it improved model fit. At last, after choosing confounders, we re-adjusted the dimension of the basis (k value) for the spline of date order, also by minimizing the partial autocorrelation function of the residuals. The final k values were 11 for both confounder models. After the confounder models were fixed, we separately added PM number or mass concentrations within each size fraction lag by lag from lag 0 to lag 14 to them (by penalized spline, k¼ 4, same as the splines of air temperature and relative humidity), and plotted the exposure-response functions (not shown) in order to check the linearity of particle effects. We found that all the exposure-response functions between number concentration of PM and cardiovascular emergency room visits could be considered as linear; while some exposure-response functions for mass concentration of PM within several size fractions showed a J-shape. Moreover, these exposure-response functions showed that particle effects on cardiovascular emergency room visits could exist until around lag 10. Based on the information obtained by investigating the exposure-response functions, we decided to investigate the cumulative lagged effects of PM up to 11 days by applying distributed lag non-linear models (Gasparrini et al., 2010). First, we built a cross-basis matrix for number or mass concentrations of PM within each size fraction, defining the relationship between particle concentration and cardiovascular emergency room visits, as well as the lagged effect. We assumed all the effects of number concentration of PM are linear (just as simple distributed lag models); while we modeled the relationship between cardiovascular emergency room visits and mass concentration of PM through a simple B-spline. We specified the lagged effect of PM up to 11 days, with a third degree polynomial function. We then added each cross-basis matrix, respectively, to the confounder models and fit the distributed lag non-linear models (simple distributed lag models for number concentration of PM). After model fitting the curves of lag-specific effects of number concentrations of PM within each size fraction were plotted. As well the 3-dimensions (3D) graphs representing associations which vary non-linearly along the space of mass concentrations of PM within each size fraction and lags were plotted. Furthermore, we added 11-day moving average number or mass concentrations of PM within each size fraction to the confounder models (by penalized spline, k¼4, same as the splines of air temperature and relative humidity), and plotted the exposure-response functions. And we calculated percentage changes of total or severe cardiovascular emergency room visits, together with 95% confidence intervals (CI), in association with an interquartile range (IQR, difference between the third and first quartile) increase of the 11-day moving average number or mass concentrations of PM within each size fraction, assuming linear effects. Moreover, we investigated the effects of 2-day (lag 0 and lag 1) moving average number or mass concentrations of PM, as this average has been used in previous literature and for number concentration of 100–300 nm and 300–1000 nm particles and mass concentration of particles in our study seemed to be a good choice. We plotted exposure-response functions and calculated percentage changes of cardiovascular emergency room visits as well, using the same ways described above for 11-day moving average concentration. A previous study has shown that in Beijing, the effect of PM2.5 on cardiovascular emergency room visits were significantly higher in spring compared to the other seasons (Su et al. unpublished work). Therefore, we were also interested in knowing if season modified the particle effects. We defined a dummy variable for season (spring—April to May, summer—June to August, autumn—September to October, winter—November to March) (http://www.bast.net.cn/kjhd/kxpj/kprx/ 2005/11/7/47746.shtml). Interaction terms between season and 11-day or 2-day

L. Liu et al. / Environmental Research 121 (2013) 52–63 average particulate concentration were added to the models in order to estimate the particle effects of the corresponding subgroups. We used likelihood ratio test to determine whether there were indeed differences between the subgroups.

55

death counts data of Beijing urban area was reported to follow a seasonal pattern (Breitner et al., 2011). 3.2. Particle concentration data

2.6. Sensitivity analyses To explore the robustness of the models, we additionally forced lag 0 mean barometric pressure (chosen based on the minimization of generalized cross validation criteria of the model as well) into the base models and then re-analyzed the particle effects. Moreover, for distributed lag non-linear models we changed the degree of the polynomial order from 3 to 4. We also increased or reduced the k value of the penalized spline of date order by 1, 2 and 3. The final k values were 11 for both confounder models, so it means that we tried k equal to 8, 9, 10, 12, 13 and 14 as sensitivity analysis.

3. Results 3.1. Cardiovascular emergency room visits data and meteorological data Table 1 presents the overall cardiovascular emergency room visits counts during the study period, as well as the descriptive statistics for daily total and severe cardiovascular emergency room visits, air temperature, relative humidity, and barometric pressure. Severe cardiovascular emergency room visits counts represented 67% of the total counts. As shown in Supplementary Fig. 1, daily air temperature, relative humidity and barometric pressure all followed seasonal patterns, but with different directions and magnitudes. We could not confidently detect a seasonal pattern within cardiovascular emergency room visits data; in contrast, daily cardiovascular

Table 2 presents the descriptive statistics for daily number and mass concentration of PM in different size fractions. According to time-series plots (data not shown), all particle concentrations were higher in colder periods and lower in warmer periods. Clear seasonal patterns (peak in winter and trough in summer) could be seen for number concentration of 3–10 nm, 50–100 nm and 100–300 nm particles, as well as for mass concentration of 100–30 nm particles. There was a decline from 2004 to 2006 in number concentration of 3–10 nm, 10–30 nm and 30–50 nm particles, as well as of ultrafine particles and PM1. Spearman rank correlations among particle concentration are shown in Supplementary Table 1. As expected, number concentration of particles within the Nucleation mode range (o30 nm), within the Aitken mode range (30–100 nm) and within the accumulation mode range (100–1000 nm), respectively, were highly correlated with each other. Number concentration of particles within different mode ranges was not substantially correlated. Modest negative correlation has also been seen between the smallest 3–10 nm particles and 300–1000 nm particles. Again as expected, mass concentration of particles within the accumulation mode range was highly correlated with each other. And modest correlations were seen between mass concentration of 1000–2500 nm particles and 100–300 nm as well as 300–1000 nm particles. The number and mass concentration of the two particle size fractions within accumulation mode range were also highly cross-correlated with each other.

Table 1 Descriptive statistics of daily cardiovascular emergency room visit counts, daily mean air temperature, relative humidity and barometric pressure. Count during the study period Total cardiovascular emergency room visitsa Severe cardiovascular emergency room visitsb

13,026 8698

Temperature (1C) Relative humidity (%) Barometric pressure (hPa)

Daily mean

Standard deviation

13 8

5 4

14.2 53 1012

10.7 20 10

Minimum 1 1  10.1 8 988

Median

Maximum

12 8

30 24

16.4 54 1012

32.1 93 1043

a

Total cardiovascular emergency room visits ¼ Visits due to cardiovascular diseases (I00–I99) and cause-unknown sudden death (R96). Severe cardiovascular emergency room visits ¼Visits due to ischemic heart diseases (I20–I25), arrhythmia (I47–I49), heart failure (I50) and cerebrovascular diseases (I60–I69). b

Table 2 Descriptive statistics of daily mean particulate number concentration and mass concentration. Size fraction (nm)

Missing (%)

Number concentration (cm  3) Daily mean

3–10 10–30 30–50 50–100 100–300 300–1000 1000–2500 2500–10,000 3–100 3–1000 3–2500 3–10,000

7.0 7.0 7.0 7.0 7.0 18.8 18.8 18.8 7.0 18.8 18.8 18.8

3367 6732 4890 6792 6430 882

Standard deviation 4250 3736 1839 2881 3583 725

Mass concentration (lg m  3) Mininmu Median Maximum Daily mean 85 1182 893 634 337 28

1804 5862 4691 6454 5904 712

40703 29914 13608 19393 21169 4775

19622 27624

76283 86864

a a

21781 29297

9616 10226

5612 7441 a a

Standard deviation

Mininum Median Maximum

a a a a

29.3 62.0 16.4 23.7

18.3 52.6 15.0 21.4

1.4 2.2 0.3 0.4

93.4 109.8 136.4

68.6 77.9 93.3

3.8 7.8 11.4

26.2 49.4 12.7 18.4

105.1 323.9 139.4 222.0

80.0 92.9 117.0

412.5 451.5 539.0

a

a The contributions to particle number by 1000–2500 nm and 2500–10,000 nm particles, as well as the contributions to particle mass by 3–10 nm, 10–30 nm, 30–50 nm and 50–100 nm particles were quite low in comparison to other fractions. Therefore, we excluded those fractions from analyses. We also did not investigate the effects of number concentration of 3–2500 nm and 3–10,000 nm particles (as they would be nearly the same as the effect of number concentration of 3–1000 nm particles), as well as the effect of mass concentration of 3–100 nm particles.

56

L. Liu et al. / Environmental Research 121 (2013) 52–63

Fig. 2. Relative risks (together with 95% confidence intervals) for TOTAL cardiovascular emergency room visits in association with an interquartile range increase in NUMBER concentration of particulate matter of each size fraction, obtained by distributed lag non-linear models. Models were estimated with lags up to 11 days using a third degree polynomial. Overall 11-day relative risks are indicated in each plot.

Modest correlation could be seen between number concentration of 3–10 nm particles and mass concentration of 300–1000 nm particles (negative), and between number concentration of 50–100 nm particles and mass concentration of 100–300 nm particles. Modest correlation could also be seen between number concentration of 300–1000 nm particles and mass concentration of 1000–2500 nm particles. 3.3. Regression results Fig. 2 and Supplementary Fig. 2 present the lagged effects of sub-micrometer particles on total and severe cardiovascular emergency room visits, respectively, in association with an IQR increase in number concentration. When looking at the same particle size fraction, the images of the effect curves of total and severe cardiovascular emergency room visits were very similar, while the ones of severe cardiovascular emergency room visits had broader 95% CI, probably due to fewer cases compared with total counts. The increase of number concentration of the four particle size fractions smaller than 100 nm showed delayed effects from lag 4 to lag 10, among which the effects of 10–30 nm and 30–50 nm particles were stronger then the effects of 3–10 nm and 50–100 nm particles. The increase of number concentration of the four particle size fractions within 30–1000 nm showed more immediate effects mainly from lag 0 to lag 1. Therefore, delayed and immediate effects could both be seen with the increase of number concentration of PM1. Note, that the delayed effects of the two larger size fractions, ultrafine particles and PM1, actually referred especially the two size fractions 10–30 nm and 30–50 nm. When using mass instead of

number concentration of 100–300 nm or 300–1000 nm particles, the lagged effect curves (not shown) were quite similar as the curves of number concentration of 100–300 nm or 300–1000 nm particles, indicating that using particulate number or mass concentration seemed not to affect the cardiovascular effect (of particles within one size fraction). Fig. 3 and Supplementary Fig. 3 present the 3D graphs for total and severe cardiovascular emergency room visits, represent the associations which vary non-linearly along the space of mass concentrations of particles within each size fraction and lags. The increase of mass concentration of all particle size fractions showed immediate effects mainly from lag 0 to lag 2. Figs. 4–7 and Supplementary Figs. 4–7 present the exposureresponse functions of total and severe cardiovascular emergency room visits, respectively, associated with 11-day or 2-day moving average particle concentration. When looking at the fragment of each function associated with 11-day average number or mass concentration, where most of the concentration data rooted, only the function between 11-day average number concentration of 3–10 nm particles and cardiovascular emergency room visits showed relatively obvious J-shape; the other functions did not substantially deviate from linearity. With regard to the functions associated with 2-day average number or mass concentration, the one between 2-day average mass concentration of 2500– 10,000 nm particles and cardiovascular emergency room visits showed kind of J-shape. The wag at the end of some functions could ascribe to very little particle concentration data. Table 3 and Supplementary Table 2 present the percentage changes (95% CI) of total and severe cardiovascular emergency room visits associated with 11-day and 2-day moving average

L. Liu et al. / Environmental Research 121 (2013) 52–63

57

Fig. 3. 3-dimensions graphs for TOTAL cardiovascular emergency room visits representing the associations which vary non-linearly along the space of MASS concentrations of particulate matters within each size fraction and lags 0–10, obtained by distributed lag non-linear models.

particulate number and mass concentration (assuming linear effect). The delayed effect of ultrafine particles number concentration was strongest, an IQR (9040 cm  3) increase in 11-day average number concentration of ultrafine particles led to a 7.2% (1.1–13.7%) increase in total cardiovascular emergency room visits, and a 7.9% (0.5–15.9%) increase in severe cardiovascular emergency room visits. 10–30 nm and 30–50 nm particles contributed most to these effects. The delayed effects of particulate mass concentration were small. Among all particle size fractions, 11-day average mass concentration of 1000–2500 nm and 2500– 10,000 nm particles had strongest effects. An IQR (9.2 mg m  3 and 14.5 mg m  3) increase in 11-day average mass concentration of these two size fractions led at most to a 2.4% (  0.2–5.2%) increase in cardiovascular emergency room visits. When looking at PM1, PM2.5 and PM10, the strongest effect was the one of mass concentration of PM10 on total cardiovascular emergency room visits, an IQR (53.7 mg m  3) increase in 11-day average mass concentration of PM10 led to a 0.6% (  2.1–3.5%) increase in total cardiovascular emergency room visits. Among all particle size fractions, 2-day average number concentration of Aitken mode particles had strongest effects. An IQR (2269 cm  3) increase in 2-day average number concentration of 30–50 nm particles led to a 2.4% (  1.5–6.5%) increase in total, and a 1.7% (  2.9–6.5%) increase in severe cardiovascular emergency room visits. The immediate effect of mass concentration came mainly from 1000–2500 nm particles. An IQR (11.7 mg m  3) increase in 2-day average mass concentration of 1000–2500 nm particles lead to an around 2.4% (0.4–4.4%) increase in total, and a

1.7% (  0.8–4.2%) increase in severe cardiovascular emergency room visits. When looking at PM1, PM2.5 and PM10, an IQR (68.5 mg m  3, 79.8 mg m  3 and 94.5 mg m  3) increase in 2-day average mass concentration led to 1.4–2.0% ( 1.4–5.0%) increases in total cardiovascular emergency room visits. In contrast, however, increases in PM1, PM2.5 and PM10 showed no harmful effects on severe cardiovascular emergency room visits. The effects of (both 11-day and 2-day average) number concentration of ultrafine particles, PM1 and 10–30 nm particles were substantially higher in winter (data not shown). The effects of number concentration of particles in other size fractions, as well as the effects of mass concentration of particles showed no seasonal difference. 3.4. Sensitivity analyses Forcing lag 0 mean barometric pressure additionally into the base models did not substantially change the shapes of the curves and graphs or the percentage changes (95% CI) in Table 3 (data not shown). After varying the degree of polynomial order of the distributed lag non-linear models from 3 to 4, the lagged effect curves kept similar trends but became rougher (as expected), and the 3D graphs also became slightly rougher (data not shown). Reducing the k value of the penalized spline of date order from 11 to 10, 9 or 8 induced changes on some particle effect magnitudes but not on directions; increasing the k value of the penalized spline of date order from 11 to 12, 13 or 14 made almost no change in particle effects (data not shown).

58

L. Liu et al. / Environmental Research 121 (2013) 52–63

Fig. 4. Exposure-response functions (together with 95% confidence intervals) for TOTAL cardiovascular emergency room visits associated with 11-day moving average NUMBER concentration of particulate matter of each size fraction.

4. Discussion 4.1. Effects of ultrafine particles, sub-micrometer particles and their sub-size fractions Using particle data from the same measurement station, Breitner et al. (2011) reported 2-days delayed associations between daily cardiovascular mortality in the Beijing urban area and number concentration of Aitken mode particles and particles smaller than 800 nm. In a study conducted in London by Atkinson et al. (2010), the association between particulate number concentration and cardiovascular deaths was observed at lag 1. In studies conducted in Erfurt, Germany, an increase in number concentration of 10–30 nm particles, 30–50 nm particles and ultrafine particles at lag 4 were associated with the most increased risk ¨ for daily mortality (Peters et al., 2009; Stolzel et al., 2003). In our study on cardiovascular emergency room visits, the delayed effects of ultrafine particles and the four size fractions under ultrafine range, as well as sub-micrometer particles, started to appear at lag 4 and lasted until lag 10, suggesting that the effects of ultrafine and sub-micrometer particles (measured in number concentration) on cardiovascular morbidity may appear later and last longer than the effects on cardiovascular mortality. Branis et al. (2010) reported that in Prague, Czech Republic, the increases in 7-day average number concentration of Aitken mode,

accumulation mode and total submicron particles all elevated cardiovascular hospital admissions. The percentage changes in cardiovascular hospital admissions associated with an increase of 1000 particles cm  3 were 1.8% (95% CI: 0.7–2.9%), 16.4% (95% CI: 5.2–28.7%) and 1.1% (95% CI: 0.4–1.8%), respectively. The effect of accumulation mode particles was strongest. In our results, the increase in 11-day average number concentration of accumulation mode particles had almost no effects; a 1000 particles cm  3 increase in 11-day average number concentration of 10–30 nm, 30–50 nm and 3–1000 nm particles was associated with 2.1% (0.3–4.1%), 2.9% (0–6.0%) and 0.6% (0–1.2%) changes in total cardiovascular emergency room visits. Halonen et al. (2009) investigated the associations between number concentration of Nucleation mode, Aitken mode and accumulation mode particles and acute hospital admissions of a population aged 65 years or older due to coronary heart disease (CHD), stroke and arrhythmia in Helsinki, Finland. They found that only the 5-day average number concentration of Aitken mode particles was significantly associated with arrhythmia admission. The percentage change was 4.1% (95% CI: 0.3–8.0%) per 2467 particles cm  3 increase. Arrhythmia admission accounted for 20% severe cardiovascular emergency room visits in our study; we found a 6.0% (  0.8– 13.3%) increase in severe cardiovascular emergency room visits per 2076 cm  3 increase in 11-day average number concentration of 30–50 nm particles.

L. Liu et al. / Environmental Research 121 (2013) 52–63

59

Fig. 5. Exposure-response functions (together with 95% confidence intervals) for TOTAL cardiovascular emergency room visits associated with 11-day moving average MASS concentration of particulate matter of each size fraction.

Branis et al. (2010) also reported significant associations between cardiovascular hospital admissions and number concentration of Aitken mode, accumulation mode and total submicron particles at lag 0 to lag 2, comparable to what we found. The percentage changes in cardiovascular hospital admissions associated with an increase of previous day (lag 1) 1000 particles cm  3 were 1.3% (95% CI: 0.5–2.1%), 7.8% (95% CI: 0–16.2%) and 0.8% (95% CI: 0.2–1.3%), respectively. In our results, a 1000 particles cm  3 increase in 2-day average number concentration of Aitken mode, accumulation mode and sub-micrometer particles was associated with 2.4% (1.2–6.1%), 0.9% ( 1.7–3.6%) and 2.2% (2.2–6.8%) changes in total cardiovascular emergency room visits. On the other hand, Andersen et al. (2008) failed in finding any significant associations between number concentration of 6–700 nm particles or ultrafine particles and cardiovascular admissions of elderly people (Z65 years) in Copenhagen, Denmark. Only very few studies have used mass concentration of accumulation mode or sub-micrometer particles, mainly focusing on cardiovascular mortality rather than cardiovascular hospital admission. Breitner et al. (2011) reported an increased ischemic heart disease mortality associated with increases in lag 2 mass concentration of 100–300 nm, 300–800 nm and o800 nm particles in Beijing, China. Perez et al. (2009) found increases in cardiovascular and cerebrovascular mortality associated with an increase in lag 1 mass concentration of PM1 in a study conducted in Barcelona. In a study conducted in Spokane, Washington, cardiac emergency room visits and hospital admissions were

considered, but no consistent associations between them and mass concentration of PM1 were found (Slaughter et al., 2005).

4.2. Effects of 1000–2500 nm particles and fine particles A number of epidemiologic studies have examined the adverse effect of fine particles on cardiovascular morbidity (Guo et al., 2010; Jalaludin et al., 2006; Hwang et al., 2004; Linares and Dı´az, 2010). Guo et al. (2010) explored the risk of hypertension emergency hospital visits in Beijing, China associated with fine particles. They reported a 8.4% (95% CI: 2.8–13.9%) change associated with a 10 mg m  3 increase in 5-day average PM2.5 concentration. Jalaludin et al. (2006) found in Sydney, Australia, the percentage changes of all cardiovascular attendances associated with an increase of 4.8 mg m  3 in lag 0–1 average PM2.5 were 0.85% (95% CI: 0.18–1.52%). We could see in Table 3 that the percentage change of total cardiovascular emergency room visits associated with an IQR (79.8 mg m  3) increase in 2-day average mass concentration of PM2.5 in our study was 2.0% (0.8–5.0%). The IQR in our study was around 40 times to 4.8 mg m  3 in the Sydney study, while the percentage change was only a bit more than 2 times; this might give a hint on possible population adaptation to air pollution in Beijing (although the linearity of effects does not point towards an adaptation). There is a lack of time-series study exactly examining the adverse effect of 1000–2500 nm particles (PM2.5 1) on cardiovascular hospital admission. But two panel studies (Chang et al., 2007; Wu et al., 2010) have

60

L. Liu et al. / Environmental Research 121 (2013) 52–63

Fig. 6. Exposure-response functions (together with 95% confidence intervals) for TOTAL cardiovascular emergency room visits associated with 2-day moving average NUMBER concentration of particulate matter of each size fraction.

shown that increased PM2.5 1 mass concentration could raise the risk of arterial stiffness and heart rate variability reduction. 4.3. Effects of coarse particles and PM10 A large number of epidemiologic studies have examined the ¨ adverse effect of PM10 on cardiovascular health outcomes (Ruckerl et al., 2011). The short-term (time lag: 0–5 days) effects of PM10 were seen on cardiovascular diseases hospital admission of people of every age group; however, the effect size varied a lot (even within each age group) (Morris, 2001). The percentage changes in the admission associated with per 10 mg m  3 increase in PM10 ranged from 0.5% to 4.8% (Morris, 2001). With regard to coarse particles and cardiovascular morbidity, Peng et al. (2008) reported an association between a 10 mg m  3 increase in PM10 2.5 in the United States and a 0.4% (95% CI: 0.1–0.7%) increase in cardiovascular disease admission on the same day. Host et al. (2008) reported an association between a 10 mg m  3 increases in PM10 2.5 in six French cities and 6.4% (95% CI: 1.6–11.4%) increases in ischemic heart disease hospitalization of elderly people. Besides, two studies (Lipsett et al., 2006; Yeatts et al., 2007) conducted on the base of repeatedly measured data of certain group of subjects also reported that increase in coarse particles concentration was associated with increased risk of systemic inflammation and decreased heart rate variability. Zanobetti and Schwartz (2009) reported a 0.32% (95% CI: 0–0.64%) increase in cardiovascular mortality associated with a 10 mg m  3 increase in 2-day average coarse particles. In contrast, Kan et al. (2007) did not find a

significant effect of coarse particles on cardiovascular mortality in Shanghai, China. We could see in Table 3 that the percentage changes of total cardiovascular emergency room visits associated with IQR (94.5 mg m  3 and 20.2 mg m  3) increases in 2-day average mass concentration of PM10 and PM10 2.5 in our study were 2.0% (95% CI:  0.84.9%) and 1.2% (95% CI:1.3 3.6%), respectively. 4.4. Strengths and limitations This study was conducted in a highly polluted city and based on data of accurately size-segregated particles. The emergency room visit data was collected from Peking University’s Third Hospital, located in the Haidian district, where patients within 10 km of the measurement site were likely to be treated (personal communication with hospital doctors) (Leitte et al., 2011). As the urban area of Beijing is 1368 km2 with about 7,072,000 registered permanent residents in 2005 (Liu et al., 2011), this area about 314 km2 then contained approximately 1,623,251 permanent residents. The daily ERVT count reached 13, ensuring the statistical power of the analysis. We were able to gather fourth grade or senior medical students (mostly) with hospital internship experience to turn the diagnoses to ICD-10 codes and input them to the database. And we applied two steps, getting a standard diagnosiscode list for comparison and double entry, to assure good quality of the emergency room visit data. Nevertheless, this study also has some limitations. First, we collected particle data from only one measurement site;

L. Liu et al. / Environmental Research 121 (2013) 52–63

61

Fig. 7. Exposure-response functions (together with 95% confidence intervals) for TOTAL cardiovascular emergency room visits associated with 2-day moving average MASS concentration of particulate matter of each size fraction. Table 3 Percentage changes (together with 95% confidence intervals) for TOTAL cardiovascular emergency room visits associated with an interquartile range increase in 11-day or 2-day moving average particulate number or mass concentration. Size fraction (nm)

3–10 10–30 30–50 50–100 100–300 300–1000 1000–2500 2500–10,000 3–100 3–1000 3–2500 3–10,000

11-Day moving average )number concentration (cm  3)

11-Day moving average mass concentration (lg m  3)

2-Day moving average number concentration (cm  3)

2-Day moving average mass concentration (lg m  3).

Interquartile range

Interquartile range

interquartile range

Interquartile range

Percentage range change % (95% CI)

18.5 51.1 11.7 20.2

0.4 1.3 2.4 1.2

68.5 79.8 94.5

1.4 (  1.4, 4.3) 2.0 ( 0.8, 5.0) 2.0 ( 0.8, 4.9)

2715 2907 1945 2824 2523 402

9040 10310

Percentage change % (95% CI)

3.8 6.2 5.7 2.2  1.0  0.7

(  0.4, 8.3) (0.8, 12.0) (0.0, 11.7) (  3.1, 7.7) (  4.8, 3.0) (  3.4, 2.0)

11.8 29.9 9.2 14.5

7.2 (1.1, 13.7) 5.8 (  0.5, 12.4) 40.7 43.0 53.7

Percentage change % (95% CI)

 1.2  0.3 2.4 2.0

( 4.5, ( 2.9, ( 0.2, ( 1.4,

2.3) 2.4) 5.2) 5.6)

 0.3 ( 3.2, 2.6) 0.3 ( 2.4, 3.0) 0.6 ( 2.1, 3.5)

therefore, measurement error due to greater spatial variability of Beijing could be present in this study, especially for numbers of traffic-associated particles. Nevertheless, the average particle number size distributions at the Peking University measurement site and another regional measurement site located around 50 km to the south of the University were shown to be comparable

3334 3689 2269 3428 3692 704

10340 11990

Percentage change % (95% CI)

 0.8 0.6 2.4 2.4 0.6 0.9

( 3.3, ( 3.3, ( 1.5, ( 1.2, ( 2.5, ( 1.7,

1.8) 4.7) 6.5) 6.1) 3.7) 3.6)

1.1 ( 3.0, 5.3) 2.2 ( 2.2, 6.8)

( 2.5, 3.3) (  1.3, 4.0) (0.4, 4.4) (  1.3, 3.6)

(Leitte et al., 2011). And the Peking University measurement site and the Third Hospital were quite near to each other (see Fig. 1); so the measurement site is assumed to have measured the average exposure of the district. Moreover, as discussed by Gasparrini and Armstrong (2010), the error in assigning individuals to central site levels is primarily of a

62

L. Liu et al. / Environmental Research 121 (2013) 52–63

Berkson-type (Zeger et al. 2000), which in linear models does not lead to bias in estimation, although reduces the precision. In our study, the association between number concentration of particles and cardiovascular emergency room visit is linear. Therefore, the use of data from one monitoring station should not lead to bias in the effect estimation of particle number. Second, we collected emergency room visit data also from only one hospital, and only patients within a certain area of Beijing (patients within 10 km of the measurement site, as mentioned above) were likely to be treated there. This area is an urban area, has no substantial difference from other areas of the urban part of Beijing, except that many universities are located there. Therefore, it can be speculated that the education level and socioeconomic status of the population of this area might be higher than the average level of the whole Beijing urban area. Third, the diagnoses of these emergency department visits are made by physicians. A physician would make all possible diagnoses about one certain patient, and write them in the standard medical record forms we obtained, in the order of high to low probability. We took the first (the most possible) diagnosis. In some cases, the first ‘‘diagnosis’’ is not a disease but a symptom, for example palpitation, chest pain, short of breath, dizziness, fever and so on, or a combination of several symptoms, which means that the physician could not make a fairly certain diagnosis during the time when the patient was in the emergency department. In these cases, the physician would suggest the patient to go for a further check, but would still write down all possible diagnoses after the symptom, in the order of high to low probability, based on the past anamneses the patient could offer and the available medical examination. In these cases, we took the second (still considered as the most possible) diagnosis, and we assume that the diagnostic uncertainty is higher. Fourth, only the forms of patients who visited and left the emergency room within one day were available, the exclusion of more serious cases (those who were admitted to the hospital) may underestimate the risk, reinforcing the results. Fifth, due to too many missing records in age and gender, we were not able to investigate particle effects by age or gender subgroups.

5. Conclusions The results from our study add to the evidence that elevated concentration level of sub-micrometer particles are associated with increased cardiovascular morbidity. Ultrafine particles showed delayed effects, while the effects of accumulation mode particles were rather immediate. This might indicate that particles within different size ranges play their effects through different pathways. The different lags by which the effects of certain particle size fractions appear should be considered when taking preventive measures to improve public health.

Author contributions LL performed the statistical analyses and drafted the manuscript. SB and AS guided the statistical analyses and the interpretation of the results, and revised the manuscript critically. JC, AW and BW performed air pollution data collection and data processing, and revised the manuscript critically. IB was involved in the study design and revised the manuscript critically. UF, US, AML and OH were involved in the study design and in air pollution data processing, and revised the manuscript critically. XP obtained emergency room visit data and meteorological data, and revised the manuscript critically. HEW was substantially involved in the study design and revised the manuscript critically.

AP was substantially involved in the study design, guided the interpretation of the results, and revised the manuscript critically.

Acknowledgments This research was funded by the German Research Foundation (DFG) (grants PE 1156/1–2 and WI 621/16-1). Parts of this work were funded by a scholarship being awarded to Liqun Liu (File no. 2008601213) under the State Scholarship Fund by the China Scholarship Council (CSC). We would like to thank the Emergency Department of Peking University Third Hospital for providing the medical record forms, the Institute for Tropospheric Research (IfT) for providing the monitoring devices, and the State Key Joint Laboratory of Environmental Simulation and Pollution Control in Peking University, Beijing, China for operating the particle monitoring. We also would like to thank Dr. Rebecca Thwing Emeny in Helmholtz Zentrum Muenchen for reading and editing the manuscript to give advice on language correlations.

Appendix A. Supporting information Supplementary data associated with this article can be found in the online version at http://dx.doi.org/10.1016/j.envres.2012. 10.009.

References Andersen, Z.J., Wahlin, P., Raaschou-Nielsen, O., Ketzel, M., Scheike, T., Loft, S., 2008. Size distribution and total number concentration of ultrafine and accumulation mode particles and hospital admissions in children and the elderly in Copenhagen, Denmark. Occup. Environ. Med. 65, 458–466, http://dx. doi.org/10.1136/oem.2007.033290. Atkinson, R.W., Fuller, G.W., Anderson, H.R., Harrison, R.M., Armstrong, B., 2010. Urban ambient particle metrics and health: a time-series analysis. Epidemiology 21, 501–511. Birmili, W., Stratmann, F., Wiedensohler, A., 1999. Design of a DMA-based size spectrometer for a large particle size range and stable operation. J. Aerosol Sci. 30, 549–553. Branis, M., Vyskovska, J., Maly, M., Hovorka, J., 2010. Association of size-resolved number concentrations of particulate matter with cardiovascular and respiratory hospital admissions and mortality in Prague, Czech Republic. Inhal. Toxicol. 22 (Suppl 2), 21–28. ¨ Breitner, S., Liu, L., Cyrys, J., Bruske, I., Franck, U., Schlink, U., Leitte, A.M., Herbarth, O., Wiedensohler, A., Wehner, B., Hu, M., Pan, X., Wichmann, H.E., Peters, A., 2011. Sub-micrometer particulate air pollution and cardiovascular mortality in Beijing, China. Sci. Total Envir. 409 (24), 5196–5204. Brook, R.D., Franklin, B., Cascio, W., Hong, Y., Howard, G., Lipsett, M., Luepker, R., Mittleman, M., Samet, J., Smith Jr, S.C., Tager, I., 2004. Air pollution and cardiovascular disease: a statement for healthcare professionals from the expert panel on population and Prevention Science of the American Heart Association. Circulation 109, 2655–2671. Brook, R.D., Rajagopalan, S., Pope III, C.A., Brook, J.R., Bhatnagar, A., Diez-Roux, A.V., Holguin, F., Hong, Y., Luepker, R.V., Mittleman, M.A., Peters, A., Siscovick, D., Smith Jr, S.C., Whitsel, L., Kaufman, J.D., 2010. Particulate matter air pollution and cardiovascular disease: an update to the scientific statement from the American Heart Association. Circulation 121, 2331–2378. Brunekreef, B., Forsberg, B., 2005. Epidemiological evidence of effects of coarse airborne particles on health. Eur. Respir. J. 26, 309–318, http://dx.doi.org/ 10.1183/09031936.05.00001805. Chang, L., Tang, C., Pan, Y., Chan, C., 2007. Association of heart rate variability of the elderly with personal exposure to PM1, PM1–2.5, and PM2.5–10. Bull. Environ. Contam. Toxicol. 79, 552–556. Delfino, R.J., Sioutas, C., Malik, S., 2005. Potential role of ultrafine particles in associations between airborne particle mass and cardiovascular health. Environ. Health Perspect. 113, 934–946. Gasparrini, A., Armstrong, B., 2010. Time series analysis on the health effects of temperature: advancements and limitations. Environ. Res. 110 (6), 633–638. Gasparrini, A., Armstrong, B., Kenward, M.G., 2010. Distributed lag non-linear models. Stat. Med. 29 (21), 2224–2234. Guo, Y., Tong, S., Zhang, Y., Barnett, A.G., Jia, Y., Pan, X., 2010. The relationship between particulate air pollution and emergency hospital visits for hypertension in Beijing, China. Sci. Total Environ. 408 (20), 4446–4450.

L. Liu et al. / Environmental Research 121 (2013) 52–63

Halonen, J.I., Lanki, T., Yli-Tuomi, T., Tiittanen, P., Kulmala, M., Pekkanen, J., 2009. Particulate air pollution and acute cardiorespiratory hospital admissions and mortality among the elderly. Epidemiology 20, 143–153. Host, S., Larrieu, S., Pascal, L., Blanchard, M., Declercq, C., Fabre, P., Jusot, J.F., Chardon, B., Le Tertre, A., Wagner, V., Prouvost, H., Lefranc, A., 2008. Shortterm associations between fine and coarse particles and hospital admissions for cardiorespiratory diseases in six French cities. Occup. Environ. Med. 65, 544–551, http://dx.doi.org/10.1136/oem.2007.036194. Hwang, J., Hu, T., Chan, C., 2004. Air pollution mix and emergency room visits for respiratory and cardiac diseases in Taipei. J. Data Sci. 2, 311–327. Jalaludin, B., Morgan, G., Lincoln, D., Sheppeard, V., Simpson, R., Corbett, S., 2006. Associations between ambient air pollution and daily emergency department attendances for cardiovascular disease in the elderly (65þ years), Sydney, Australia. J. Expo. Sci. Environ. Epidemiol. 16, 225–237, http://dx.doi.org/ 10.1038/sj.jea.7500451. Kan, H., London, S.J., Chen, G., Zhang, Y., Song, G., Zhao, N., Jiang, L., Chen, B., 2007. Differentiating the effects of fine and coarse particles on daily mortality in Shanghai, China. Environ. Int. 33, 376–384. Leitte, A.M., Petrescu, C., Franck, U., Richter, M., Suciu, O., Ionovici, R., Herbarth, O., Schlink, U., 2009. Respiratory health, effects of ambient air pollution and its modification by air humidity in Drobeta-Turnu Severin, Romania. Sci. Total Environ. 407, 4004–4011. Leitte, A.M., Schlink, U., Herbarth, O., Wiedensohler, A., Pan, X., Hu, M., Richter, M., Wehner, B., Tuch, T., Wu, Z., Yang, M., Liu, L., Breitner, S., Cyrys, J., Peters, A., Wichmann, H.E., Franck, U., 2011. Size-segregated particle number concentrations and respiratory emergency room visits in Beijing, China. Environ. Health Perspect. 119, 508–513. Linares, C., Dı´az, J., 2010. Short-term effect of concentrations of fine particulate matter on hospital admissions due to cardiovascular and respiratory causes among the over-75 age group in Madrid, Spain. Public Health 124 (1), 28–36. Lipsett, M.J., Tsai, F.C., Roger, L., Woo, M., Ostro, B.D., 2006. Coarse particles and heart rate variability among older adults with coronary artery disease in the Coachella Valley, California. Environ. Health Perspect 114, 1215–1220. Liu, L., Breitner, S., Pan, X., Franck, U., Leitte, A.M., Wiedensohler, A., von Klot, S., Wichmann, H.E., Peters, A., Schneider, A., 2011. Associations between air temperature and cardio-respiratory mortality in the urban area of Beijing, China: a time-series analysis. Environ. Health 10, 51, http://dx.doi.org/ 10.1186/1476-069X-10-51. Ljungman P. Cardiovascular effects of short-term exposure to air pollution: Exploring potential pathways and susceptible subgroups. /http://hdl.handle. net/10616/38213S. Morris, R.D., 2001. Airborne particulates and hospital admissions for cardiovascular disease: A quantitative review of the evidence. Environ. Health Perspect. 109 (suppl 4), 495–500. Pekkanen, J., Kulmala, M., 2004. Exposure assessment of ultrafine particles in epidemiologic time-series studies. Scand. J. Work Environ. Health 30 (Suppl 2), 9–18. Peng, R.D., Chang, H.H., Bell, M.L., McDermott, A., Zeger, S.L., Samet, J.M., Dominici, F., 2008. Coarse particulate matter air pollution and hospital admissions for cardiovascular and respiratory diseases among Medicare patients. JAMA 299, 2172–2179. ¨ Perez, L., Medina-Ramo´n, M., Kunzli, N., Alastuey, A., Pey, J., Pe´rez, N., Garcia, R., Tobias, A., Querol, X., Sunyer, J., 2009. Size fractionate particulate matter, vehicle traffic, and case-specific daily mortality in Barcelona, Spain. Environ. Sci. Technol 43 (13), 4707–4714.

63

Peters, A., 2005. Particulate matter and heart disease: Evidence from epidemiological studies. Toxicol. Appl. Pharmacol. 207 (2), S477–S482. Peters, A., Breitner, S., Cyrys, J., Stolzel, M., Pitz, M., Wolke, G., Heinrich, J., Kreyling, W., Kuchenhoff, H., Wichmann, H.E., 2009. The influence of improved air quality on mortality risks in Erfurt, Germany. Res. Rep. Health Eff. Inst. 137, 5–77. ¨ Ruckerl, R., Schneider, A., Breitner, S., Cyrys, J., Peters, A., 2011. Health effects of particulate air pollution: A review of epidemiological evidence. Inhal. Toxicol. 23 (9-11), 555–592. Samoli, E., Analitis, A., Touloumi, G., Schwartz, J., Anderson, H.R., Sunyer, J., Bisanti, L., Zmirou, D., Vonk, J.M., Pekkanen, J., Goodman, P., Paldy, A., Schindler, C., Katsouyanni, K., 2005. Estimating the exposure-response relationships between particulate matter and mortality within the APHEA multicity project. Environ. Health Perspect 113 (1), 88–95. Simon, N.Wood, 2006. Generalized Additive Models: An Introduction with R. Taylor & Francis Group, LLC. Slaughter, J.C., Kim, E., Sheppard, L., Sullivan, J.H., Larson, T.V., Claiborn, C., 2005. Association between particulate matter and emergency room visits, hospital admissions and mortality in Spokane, Washington. J. Expo. Anal. Environ. Epidemiol. 15 (2), 153–159. ¨ Stolzel M., Peters A., Wichmann H.-E. Daily mortality and fine and ultrafine particles in Erfurt, Germany. Revised Analyses of Time-Series Studies of Air Pollution and Health, Health Effects Institute, 2003, pp. 231–240. Touloumi, G., Samoli, E., Pipikou, M., Le Tertre, A., Atkinson, R., Katsouyanni, K., 2006. Seasonal confounding in air pollution and health time-series studies: Effect on air pollution effect estimates. Stat. Med 25 (24), 4164–4178. Wehner, B., Birmili, W., Ditas, F., Wu, Z., Hu, M., Liu, X., Mao, J., Sugimoto, N., Wiedensohler, A., 2008. Relationships between submicrometer particulate air pollution and air mass history in Beijing, China, 2004–2006. Atmos. Chem. Phys 8, 6155–6168. Wehner, B., Wiedensohler, A., Tuch, T.M., Wu, Z.J., Hu, M., Slanina, J., Kiang, C.S., 2004. Variability of the aerosol number size distribution in Beijing, China: New particle formation, dust storms, and high continental background. Geophys. Res. Lett. 31, L22108, http://dx.doi.org/10.1029/2004GL021596. ¨ Wichmann, H.E., Spix, C., Tuch, T., Wolke, G., Peters, A., Heinrich, J., Krevling, W.G., Heyder, J., 2000. Daily mortality and fine and ultrafine particles in Erfurt, Germany part I: Role of particle number and particle mass. Res. Rep. Health Eff. Inst 98, 5–94. Wu, C., Kuo, I., Su, T., Li, Y., Lin, L., Chan, C., Hsu, S., 2010. Effects of personal exposure to particulate matter and ozone on arterial stiffness and heart rate variability in healthy adults. Am. J. Epidemiol. 171, 1299–1309. Yeatts, K., Svendsen, E., Creason, J., Alexis, N., Herbst, M., Scott, J., Kupper, L., Williams, R., Neas, L., Cascio, W., Devlin, R.B., Peden, D.B., 2007. Coarse particulate matter (PM2.5–10) affects heart rate variability, blood lipids, and circulating eosinophils in adults with asthma. Environ. Health Perspect. 115, 709–714. Yue, D., Hu, M., Wu, Z., Wang, Z., Guo, S., Wehner, B., Nowak, A., Achtert, P., Wiedensohler, A., Jung, J., Kim, Y.J., Liu, S., 2009. Characteristics of aerosol size distributions and new particle formation in the summer in Beijing. J. Geophys. Res., 114, http://dx.doi.org/10.1029/2008JD010894. Zanobetti, A., Schwartz, J., 2009. The effect of fine and coarse particulate air pollution on mortality: a national analysis. Environ. Health Perspect 117, 898–903. Zeger, S.L., Thomas, D., Dominici, F., Samet, J.M., Schwartz, J., Dockery, D., Cohen, A., 2000. Exposure measurement error in time-series studies of air pollution: Concepts and consequences. Environ. Health Perspect 108 (5), 419–426.