American Journal of Epidemiology © The Author 2014. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health. All rights reserved. For permissions, please e-mail: [email protected].

Vol. 179, No. 9 DOI: 10.1093/aje/kwu026 Advance Access publication: March 18, 2014

Original Contribution Short-Term Associations of Cause-Specific Emergency Hospitalizations and Particulate Matter Chemical Components in Hong Kong

Vivian Chit Pun, Ignatius Tak-Sun Yu, Hong Qiu, Kin-Fai Ho, Zhiwei Sun, Peter K.K. Louie, Tze Wai Wong, and Linwei Tian*

Initially submitted August 16, 2013; accepted for publication January 24, 2014.

Despite an increasing number of recent studies, the overall epidemiologic evidence associating specific particulate matter chemical components with health outcomes has been mixed. The links between components and hospitalizations have rarely been examined in Asia. We estimated associations between exposures to 18 chemical components of particulate matter with aerodynamic diameter less than 10 μm (PM10) and daily emergency cardiorespiratory hospitalizations in Hong Kong, China, between 2001 and 2007. Carbonaceous particulate matter, sulfate, nitrate, and ammonium accounted for two-thirds of the PM10 mass. After adjustment for time-varying confounders, a 3.4-μg/m3 increment in 2-day moving average of same-day and previous-day nitrate concentrations was associated with the largest increase of 1.32% (95% confidence interval: 0.73, 1.92) in cardiovascular hospitalizations; elevation in manganese level (0.02 μg/m3) was linked to a 0.91% (95% confidence interval: 0.19, 1.64) increase in respiratory hospitalizations. Upon further adjustment for gaseous copollutants, nitrate, sodium ion, chloride ion, magnesium, and nickel remained significantly associated with cardiovascular hospitalizations, whereas sodium ion, aluminum, and magnesium, components abundantly found in coarser PM10, were associated with respiratory hospitalizations. Most positive links were seen during the cold season. These findings lend support to the growing body of literature concerning the health associations of particulate matter composition and provide important insight into the differential health risks of components found in fine and coarse modes of PM10. air pollution; chemical constituents; hospitalization; particulate matter; time series

Abbreviations: CI, confidence interval; lag0n, n-day moving average of current day to previous n days; PM10, particulate matter with aerodynamic diameter less than 10 μm; PM2.5, particulate matter with aerodynamic diameter less than 2.5 μm.

elemental carbon, each making up about 10%–30% of the overall mass load (2). Because not all particulate matter is the same, the ultimate goal is to identify the components and sources most directly responsible for adverse health effects to allow for more targeted regulations (3, 4). As large-scale data collection on particulate matter chemical composition began in the past decade, researchers have been able to investigate associations between acute exposure to particulate matter components and death and nonfatal health outcomes. Although early air pollution studies conducted in Western countries identified evidence of adverse health effects of carbonaceous species (i.e., elemental and organic carbon) and nickel (5–10), the overall epidemiologic

Particulate matter pollution has been the most widespread health threat among the criteria air pollutants. Assessments of most particulate matter–health associations have largely focused on the ambient concentration of total mass (weight) of particulate matter with aerodynamic diameter less than 10 μm (PM10) and particulate matter with aerodynamic diameter less than 2.5 μm (PM2.5). However, significant heterogeneity in risk estimates suggests that the aspect of particulate matter most harmful to health may not be best quantified by mass measurement alone. Particulate matter is a complex mixture of numerous components (i.e., chemical species) (1). The predominant components are sulfate, nitrate, ammonium, sea salt, mineral dust, organic compounds, and black or 1086

Am J Epidemiol. 2014;179(9):1086–1095

Downloaded from http://aje.oxfordjournals.org/ at Temple University on November 14, 2014

* Correspondence to Dr. Linwei Tian, 4/F, Jockey Club School of Public Health and Primary Care, Chinese University of Hong Kong, Prince of Wales Hospital, Shatin-NT, Hong Kong, China (e-mail: [email protected]).

Effects of PM10 Components on Hospitalization Risks 1087

METHODS Data

Between January 1, 2001, and December 31, 2007, the Hong Kong Environmental Protection Department collected 24-hour filter samples of PM10 at 6 general air quality monitoring stations (Figure 1). These stations cover urban areas with different land uses (e.g., residential, commercial, industrial, and a mix of these), new towns, and rural areas in Hong Kong, and they capture the air quality to which the general population is exposed on a regular basis (23). Twenty-six PM10 chemical components were speciated from the samples via various analytical methods as reported elsewhere (22). Upon exclusion of contaminated components and those that had more than 25% of samples below the method’s detection limit or that had more than 25% of missing samples, we included station-specific measurements of 18 PM10 components. They were elemental carbon, organic carbon, Am J Epidemiol. 2014;179(9):1086–1095

nitrate, sulfate, ammonium ion, chloride ion, sodium ion, potassium ion, aluminum, arsenic, calcium, cadmium, iron, magnesium, manganese, nickel, lead, and vanadium. Measurements of organic carbon were converted to those of organic matter by multiplying by a factor of 1.6 (24). To adjust for copollutants and meteorological conditions, we calculated daily mean concentrations of nitrogen dioxide and sulfur dioxide, as well as PM10 mass and 8-hour mean (10:00 AM–6:00 PM) concentration of ozone from hourly tapered element oscillating microbalance air pollutant data collected at the same monitoring stations; we obtained daily mean temperature and relative humidity data from the Hong Kong Observatory for the same study period. We acquired from the Hong Kong Hospital Authority daily counts of emergency hospital admissions into publicly funded hospitals between January 1, 2001, and December 31, 2007. These hospitals provide 24-hour accident and emergency services and represent 90% of the hospital beds that serve Hong Kong residents (Figure 1) (19). We considered hospitalizations for all cardiovascular causes (International Classification of Diseases, Ninth Revision, codes 390–459) and respiratory causes (International Classification of Diseases, Ninth Revision, codes 460–519). Hospitalizations due to influenza (International Classification of Diseases, Ninth Revision, code 487) were excluded from the respiratory category and treated as a potential confounder in the data analysis. Statistical analysis

Because each monitoring station collected PM10 samples, on average, every sixth day on a distinct sampling schedule, there might have been samples taken simultaneously from multiple stations or only a single station, or there might have been no samples taken from any stations on a particular day. Collectively, 71% of the study days were covered by measurements from at least 1 station. To compute the territory-wide mean concentrations of PM10 components, we applied a centering method to remove the station-specific influence on the measurements of each component by 1) calculating the annual mean (Xi) for each monitoring station i; 2) subtracting the annual mean from the daily mean concentration for station i on each sample day j (Xij ); 3) adding the annual mean of all stations (X) to the resulting centered values (Xij – Xi) for each station and sampling day to produce Xij00 ¼ Xij  Xi þ X; and 4) taking the average of Xij00 over all stations (25). The final PM10 exposure time series contained nonmissing territorywide mean concentrations of 18 chemical components for 1,805 days over the 7-year study period (71%), which is equivalent to data for 5 days a week. All pollutant concentrations are expressed in μg/m3 except for elemental carbon and organic matter, which are reported in μg carbon/m3. We applied generalized additive models with autoregressive terms to estimate associations between PM10 chemical components and cause-specific emergency hospital admissions (26). Smoothing splines with 8 degrees of freedom per year for time trend, 6 degrees of freedom for current-day temperature and previous 3-day moving average, and 3 degrees of freedom for current-day relative humidity and previous 3-day moving average were adopted a priori to minimize

Downloaded from http://aje.oxfordjournals.org/ at Temple University on November 14, 2014

findings linking particulate matter components to health risks remain mixed (11). Similarly, no consensus has been found in Asian epidemiologic studies regarding the health effects of specific particulate matter components. Two recent studies conducted in Xi’an, China, reported a positive association between PM2.5 components originating from fossil fuel combustion (e.g., elemental carbon) and daily cardiorespiratory mortality risk, especially during cold seasons (12, 13). A Shanghai, China, study linked elevations in the black carbon level to increases in total and cardiovascular mortality risk after adjustment for PM2.5 mass, but found no association with respiratory mortality risk (14); Korean researchers found significant associations of total and daily cardiovascular mortality risk with only magnesium and ammonium at the P < 0.05 level and not with the other components they examined (15). Particulate matter composition, and thus its toxicity, varies temporally and geographically depending upon the cityspecific sources of origin (e.g., local or regional emissions) and interactions between sources, meteorological conditions, and other time-varying factors. Hence, more composition studies under different atmospheres are needed to enhance our understanding of particulate matter–related health effects. Hong Kong is a coastal urban city on the boundary region of the Asian continent and the Pacific Ocean. Its air pollution levels often exceed the recommended quality guidelines of the World Health Organization (Geneva, Switzerland), posing a serious public health threat. Although 2 decades of active air pollution research have associated particulate matter mass with elevated daily risk of death (16, 17) and emergency hospital admissions in Hong Kong (18–21), individual particulate matter chemical components and sources responsible for adverse health effects have rarely been examined. In Hong Kong, particulate matter components originate from several major emission sources, such as motor vehicles, soil and road dust, and polluted sea salt, as well as industrial and agricultural combustion from the adjacent Pearl River Delta region (22). In this study, we used the PM10 composition data that have been measured consecutively for more than a decade to estimate the associations between chemical components and daily cardiorespiratory emergency hospital admissions.

1088 Pun et al.

General air monitoring station Public hospital

Yuen Long

Sham Shui Po Kwun Tong Tung Chung Central Western

N 0 km

4 km

8 km

Figure 1. Locations of 6 air quality general monitoring stations and 26 publicly funded hospitals that contributed air pollution and hospital admissions data included in the study, Hong Kong, China, 2001−2007.

problems associated with multiple testing and core model selection strategies (5, 27). We also adjusted for day of week, public holidays, and influenza epidemics (16). We first used single-pollutant models to examine associations of emergency hospitalization with exposure to individual PM10 components on the same day (lag0) and for single-day exposure on the previous 1–3 days (lag1 to lag3) while adjusting for time-varying confounders. Then, we focused mainly on exposure to a 2-day moving average of same-day and previousday pollutant concentrations (lag01) as a priori lag structure based on previous Hong Kong studies (16, 17). To further control for potential confounding effects of gaseous copollutants, we constructed multipollutant models to investigate the independent associations between PM10 components and emergency hospitalizations in the presence of nitrogen dioxide, sulfur dioxide, and ozone. We evaluated additional multipollutant models in which we also adjusted for PM10 mass. Pearson’s correlations were used to summarize the relationships between air pollutants. Moreover, we conducted separate analyses for warm seasons (May–October) and cold seasons

(November–April) with 4 degrees of freedom per year for time trend per season (11). For sensitivity analysis, we reanalyzed the time-series models after either imputing the missing component concentrations for the days without any samples (751 days) by linear interpolation using the na.approx function in the R zoo package (28) or replacing the missing data with nonmissing measurement values from the previous day. Risk estimates were reported as the percent increase [(relative risk − 1) × 100%] in daily emergency hospital admissions for an interquartile-range increment in air pollution concentrations. Where appropriate, 95% confidence intervals and 2-sided P values are reported. We performed time-series analyses using the gam function in the R mgcv package (29). RESULTS

Over the 7-year study period, we identified 400,011 cardiovascular and 587,422 respiratory emergency hospital admissions in Hong Kong, which are equivalent to 157 (standard deviation, 24) and 230 (standard deviation, 49) Am J Epidemiol. 2014;179(9):1086–1095

Downloaded from http://aje.oxfordjournals.org/ at Temple University on November 14, 2014

Tsuen Wan

Effects of PM10 Components on Hospitalization Risks 1089 Table 1. Summary Statistics of Emergency Hospital Admissions, Meteorological Measurements, and Concentrations of PM10 and Its Components in Hong Kong, China, 2001–2007 Variablea

Daily Mean (SD)

Minimum

Maximum

IQR

No. of emergency hospital admissions Cardiovascular

157 (24)

66

243

30

Respiratory

220 (49)

89

518

61

Meteorological conditions Temperature, °C

23.6 (4.9)

8.2

31.8

8.1

Relative humidity, %

78.3 (9.9)

31.3

98.1

11.4

Nitrogen dioxide

59.4 (22.3)

14.3

171.3

29.2

Sulfur dioxide

21.7 (15.1)

2.8

138.5

13.3

Ozone

42.9 (28.6)

2.9

206.5

36.2

PM10

55.9 (32.5)

Pollutant concentration, μg/m3

296.3

44.3

0.3 (0.3)

−0.1

2.9

0.3

Ammonium

3.3 (2.8)

−0.5

22.8

Arsenic

0.006 (0.007)

Cadmium

0.002 (0.003)

0.0 0.0

0.07

3.6 0.01

0.08

0.002

Calcium

0.8 (0.6)

−0.1

6.2

0.6

Chloride ion

0.9 (1.1)

−0.4

11.9

0.8

Elemental carbon

3.7 (1.6)

0.6

21.6

1.8

Iron

0.6 (0.4)

0.0

3.9

0.5

Lead

0.07 (0.07)

0.0

1.0

0.09

0.3 (0.2)

0.0

1.9

0.2

Magnesium Manganese Nickel

0.02 (0.02)

0.0

0.1

0.02

0.007 (0.006)

0.0

0.1

0.005

3.7 (3.3)

−0.4

36.8

3.4

Organic matter

13.3 (8.8)

0.5

69.2

11.1

Potassium ion

0.6 (0.6)

0.0

5.0

0.7 1.2

Nitrate

1.5 (1.0)

−0.1

8.8

Sulfate

11.5 (7.5)

0.8

63.6

9.8

Vanadium

0.01 (0.01)

0.0

0.1

0.01

Sodium ion

Abbreviations: IQR, interquartile range; PM10, particles with aerodynamic diameter less than 10 μm; SD, standard deviation. a Over the 7-year study period, there were 2,556 (100%) days of complete emergency hospitalization, meteorological, and gaseous pollutant data and 1,805 (71%) days of nonmissing data for PM10 mass and its 18 chemical components.

admissions per day, respectively (Table 1). The mean daily average temperature and relative humidity were 23.6°C and 78.3%, respectively. The daily mean concentration for PM10 was 55.9 μg/m3, with organic matter, sulfate, nitrate, elemental carbon, and ammonium ion accounting for 23.8%, 20.6%, 6.6%, 6.5%, and 5.9% of the overall particulate matter mass, respectively. These major components were moderately to highly correlated with nitrogen dioxide (r = 0.61–0.80) and nil to mildly correlated with sulfur dioxide and ozone (Web Table 1 available at http://aje.oxfordjournals. org/). Moderate to strong correlations were also seen between many components (e.g., r = 0.76 for nickel and vanadium; r = 0.93 for sulfate and ammonium ion). The majority of PM10 components had consistently higher concentrations during the cold season (data not shown). Am J Epidemiol. 2014;179(9):1086–1095

The single-pollutant models of single-day exposure lags exhibited similar patterns of associations as those from previous studies in which the risk estimates of PM10 components varied by lag structures and health outcomes (Figure 2). We observed the largest significant associations of PM10 mass ( per 44.3 μg/m3) with emergency hospitalizations for cardiovascular causes at lag0 (1.53%, 95% confidence interval (CI): 0.68, 2.38; Figure 2A) and with hospitalizations for respiratory causes at lag1 (0.95%, 95% CI: 0.22, 1.68; Figure 2B). Elemental carbon, organic matter, nitrate, potassium ion, iron, manganese, nickel, lead, and vanadium demonstrated significant links with cardiovascular hospitalizations in more than half of the lags examined, whereas only magnesium and manganese were significantly associated with respiratory hospitalizations at 2 lags (i.e., lag0 and lag1).

Downloaded from http://aje.oxfordjournals.org/ at Temple University on November 14, 2014

3.5

Aluminum

1090 Pun et al.

Change per IQR Increment, %

A)

2

1

0

−1 0123

0123

0123

0123

0123

0123

0123

0123

0123

0123

0123

0123

0123

0123

0123

0123

0123

0123

PM10

EC

OM

NO−3

SO2− 4

NH+4

Na+

K+

Cl−

Al

As

Ca

Cd

Fe

Mg

Mn

Ni

Pb

V

Pollutants and Lag Days

Change per IQR Increment, %

B)

2

1

0

−1 0123 PM10

0123 EC

0123

0123

0123

0123

OM

NO−3

SO2− 4

NH+4

0123 +

Na

0123 +

K

0123 Cl

−

0123

0123

0123

0123

0123

0123

0123

0123

0123

0123

Al

As

Ca

Cd

Fe

Mg

Mn

Ni

Pb

V

Pollutants and Lag Days

Figure 2. Percent change in emergency hospital admissions for A) cardiovascular and B) respiratory diseases per interquartile range (IQR) increment in particles with aerodynamic diameter less than 10 μm (PM10) and component concentrations at single-day exposure lags (the current day (lag0) or up to 3 days prior to current day (lag3)), based on single-pollutant models adjusted for meteorological factors, seasonal and temporal trend, day of week, and influenza epidemic, Hong Kong, China, 2001–2007. Statistically significant associations are represented by lines with hollow circles. Bars, 95% confidence intervals. Al, aluminum; As, arsenic; Ca, calcium; Cd, cadmium; Cl–, chloride ion; EC, elemental carbon; Fe, iron; K+, þ  2 potassium ion; Mg, magnesium; Mn, manganese; Na+, sodium ion; NH4 ; ammonium; Ni; nickel; NO3 ; nitrate; OM, organic matter; Pb, lead; SO4 ; sulfate; V, vanadium.

Figures 3A and 3B show the cause-specific risk estimates at lag01 exposure based on single-pollutant models. A 3.4-μg/m3 increment in nitrate concentration was associated with the largest significant increase of 1.32% (95% CI: 0.73, 1.92) in cardiovascular hospitalizations at lag01, followed by interquartile range–increment in organic matter level (1.09%, 95% CI: 0.39, 1.81; Figure 3A). Positive associations were also observed for elemental carbon, nickel, lead, iron, manganese, and most ions (sodium ion, potassium ion, and chloride ion), which corresponded to 0.41%–0.90% increases in cardiovascular hospitalizations. In contrast, fewer significant associations with respiratory hospitalizations were found (Figure 3B). Elevation in manganese level ( per 0.02 μg/m3) was associated with a 0.91% (95% CI: 0.19, 1.64) increase in respiratory hospitalizations, followed by

interquartile range–increment in sodium ion, magnesium, aluminum, and sulfate levels (0.63%–0.77% increases in hospitalizations). No risk estimates were sensitive to alternative regression models in which we imputed missing data (data not shown). Figures 3C and 3D show that most of the positive associations, which were statistically significant based on singlepollutant models, diminished in multipollutant models in which we simultaneously controlled for nitrogen dioxide, sulfur dioxide, and ozone, in addition to adjusting for time-varying (e.g., seasonal) confounders. The cardiovascular admission risks of nitrate (1.03%, 95% CI: 0.38, 1.67) and nickel (0.49%, 95% CI: 0.001, 0.97) were slightly attenuated in the presence of gaseous copollutants, whereas the associations of sodium ion (1.09%, 95% CI: 0.46, 1.73), chloride Am J Epidemiol. 2014;179(9):1086–1095

Downloaded from http://aje.oxfordjournals.org/ at Temple University on November 14, 2014

0123

Effects of PM10 Components on Hospitalization Risks 1091

DISCUSSION

The majority of recent epidemiologic studies of particulate matter pollution and health have been based on PM2.5 mass and its composition because PM2.5 is often presumed to be more relevant as an exposure indicator than larger particles (i.e., PM10). However, because of the complex physicochemical properties of particulate matter at different size fractions, the physiopathological mechanism underlying adverse health effects has yet to be fully understood (30). Thus, the contribution of larger particles to poor health should be further investigated. Our study is one of the few that has focused on exposure to PM10 components. In contract to a previous London, United Kingdom, study that also examined the link between health and various PM10 components but found evidence of an association of only nitrate and sulfate with respiratory admissions (31), we observed positive associations between multiple PM10 components and emergency hospitalizations for cardiorespiratory causes in Hong Kong at the various lags examined. We found that combustion-related particles (i.e., elemental carbon, organic matter, nitrate, potassium ion, iron, manganese, nickel, and lead) were associated with elevated cardiovascular hospitalizations, and sulfate was associated with respiratory hospitalizations based on single-pollutant models. These components reflect the major combustion sources in Hong Kong, which are local gasoline and diesel vehicle exhausts, residual oils from marine vessels, and regional industrial and agricultural combustion (22). Our findings are in accord with those of studies previously reporting evidence of adverse effects of PM2.5 components from combustion of coal and heavy oil on mortality risk and hospital admissions (5, 6, 8, 32–34). We took a step further to adjust for gaseous copollutants and found that most of the significant associations diminished, suggesting possible confounding effects. This is consistent with previous Hong Kong studies showing that the addition of nitrogen dioxide and sulfur dioxide reduced the magnitude and statistical significance of PM10 associations (16, 25). Burnett et al. (35) also reported the diminishment of particle mass and sulfate associations with cardiorespiratory hospitalizations in Toronto, Canada, after simultaneously adjusting for gaseous pollutants. ConAm J Epidemiol. 2014;179(9):1086–1095

versely, Cao et al. (12) showed that the inclusion of nitrogen dioxide and sulfur dioxide did not mask the positive association between combustion-related particles and mortality risk in Xi’an, China. Because most existing studies have relied on single-pollutant models, the potential confounding effect of gaseous pollutants on associations between particulate matter components and health should be investigated further. Our analysis showed that the associations of nitrate and nickel with cardiovascular hospitalizations remained significant after controlling for gaseous copollutants. Current epidemiologic evidence on the health risks of nitrate has been conflicting. Most studies that have examined the associations of nitrate have found no link with cardiovascular hospitalizations (27, 31, 32) and death (13, 15, 36), whereas 2 studies conducted in London, United Kingdom, and California suggested nitrate to be more important for respiratory than for cardiovascular hospitalizations (31, 33). Nitrate is acidic in nature. Although there has been no convincing animal toxicological evidence of nitrate effects at ambient levels, it is hypothesized that acidic aerosols may lower the pH within the airways by depositing hydrogen ions, thereby triggering adverse reactions (37, 38). Nitrate can be found in both fine and coarse particles (Web Figure 1). In Hong Kong, nitrogen oxides emitted from power plants, industry, and motor vehicles both locally and from the adjacent Pearl River Delta region generate most of the finer-mode nitrate through oxidation (39). When maritime and polluted urban air masses mix together, coarse-mode nitrate can also be formed through the reaction of nitric acid with sodium chloride (40, 41). In contrast, nickel is found almost exclusively in the fine mode of PM10 (i.e., PM2.5), and it serves as an index of heavy oil combustion from marine vessels in coastal cities like Hong Kong. Our observed nickel association is in agreement with another recent Hong Kong study that showed better nickel correspondence than vanadium correspondence to cardiovascular hospitalization (42), as well as with other PM2.5 composition studies reported in Xi’an, China, and the United States (5–7, 12). Sodium ion and chloride ion are considered signature elements for sea salt particles in coastal areas, and magnesium is related to sea salt, as well (22). In Hong Kong, 16% of PM10 mass originated from sea salts, which is much higher than the 9% reported in studies conducted in Barcelona, Spain, and Copenhagen, Denmark (22, 43, 44). Mar et al. (45) reported that sea salt was consistently linked to increased cardiovascular and total mortality risk in Atlanta, Georgia, whereas Andersen et al. (44) did not associate sea salt with hospital admissions among children or the elderly in Copenhagen. These components are most abundantly found in the coarser mode of PM10. In our study, sodium ion, chloride ion, and magnesium remained significantly associated with cardiovascular hospitalizations after adjustment for gaseous copollutants; risk estimates of sodium ion and magnesium were also robust for respiratory hospitalizations. Most studies did not link sodium to adverse health (8, 11, 12, 15, 27) except for 2 studies conducted in New York, New York, and 26 US communities (6, 7). Epidemiologic evidence of the impact of chloride has been highly mixed (8, 12, 13, 15, 34, 36, 46), and evidence on magnesium has been more scarce. Only 1 Korean study has associated magnesium in PM2.5 with respiratory death (15). The heterogeneity in findings may arise

Downloaded from http://aje.oxfordjournals.org/ at Temple University on November 14, 2014

ion (0.58%, 95% CI: 0.24, 0.92), and magnesium (0.6%, 95% CI: 0.1, 1.09) were stronger based on the multipollutant models compared with the single-pollutant models (Figure 3C). In addition, only sodium ion (0.65%, 95% CI: 0.11, 1.19), aluminum (0.56%, 95% CI: 0.02, 1.1), and magnesium (0.63%, 95% CI: 0.21, 1.04) remained significantly associated with respiratory hospitalizations after adjustment for gaseous copollutants (Figure 3D). Further adjustment for PM10 mass in the multipollutant models did not substantially change the regression results (data not shown). Table 2 shows that sodium ion and chloride ion were associated with cardiorespiratory hospitalizations in the cold season. In the cold season, nickel was significantly associated with cardiovascular hospitalizations, and aluminum and magnesium were significantly associated with respiratory hospitalizations. For nitrate, a significant association with cardiovascular hospitalizations was seen in the warm season.

1092 Pun et al.

A)

B) PM10

PM10

EC

EC

OM

OM

NO−3

NO−3

SO2− 4

SO2− 4

NH+4

NH+4

Na+

Na+

+

+

K

K

● ● ● ● ● ●

Cl−

Pollutants

Cl−

Pollutants

● ●

Al As

Cd

Cd

Fe

Fe

Mg

Mg

Mn

Mn

Ni

Ni

Pb

Pb

V

V

0.0

0.5

1.0

1.5

2.0

2.5

Change per IQR Increment, %

● ● ● ● ● ● ● ● ●

−1.0

−0.5

0.0

0.5

1.0

1.5

2.0

2.5

Change per IQR Increment, %

Figure 3 continues

from the use of PM2.5 composition instead of PM10 composition, which also comprises larger crustal materials. Besides originating from sea salt, sodium and chloride can be generated from combustion emissions from coal, vehicles, and smelters, and magnesium can be generated from soil and/or road dust (1, 15, 47). We observed a significant association of aluminum with respiratory hospitalizations after controlling for gaseous copollutants. Aluminum is found most abundantly in the coarser mode of PM10. It is classified as crustal material derived from exposed soil, unpaved roads, and construction activities and is moderately correlated with magnesium (22). Several studies showed significant links between soil and/ or road dust and adverse health outcomes (8, 43, 44, 48– 50), whereas some did not (45, 51–54). Overall, epidemiologic studies on aluminum are limited. Our findings add to those previously reported in several ways. First, this is one of the few studies that has focused on exposure to PM10 components, whereas the majority of epidemiologic studies have been based on PM2.5 components. Unlike cities where PM2.5 composition data are measured in consecutive years, PM2.5 chemical components in Hong Kong were measured only every sixth day for 3 nonconsecutive years. Hence, the data are sparse, and time-series analyses have limited statistical power to assess the associations between PM2.5 components and health outcomes. The application of PM10 composition data allowed us to detect

adverse health associations of not only components that are in the fine mode, but also those that have substantial fractions in the coarser mode of PM10. Second, we conducted the first Asian study to investigate the links between particulate matter components and emergency hospital admissions with adjustment for gaseous copollutants, whereas all other Asian studies of this sort focused on mortality risk. Emergency hospital admissions were unplanned and unscheduled admissions, which could be more sensitive to environmental effects than overall hospitalizations and mortality risk (20). Third, with nearly 1 million emergency hospitalizations over 7 years, our study was well powered to detect statistically significant associations. We acknowledge the limitations of the study, including the every-sixth-day sampling scheme for the PM10 composition analysis. Exposure misclassification error might exist, though the risk estimates were insensitive to alternative interpolation methods. Our findings of the associations of PM10 components with hospitalizations should be interpreted with caution for several reasons. First, measurement error cannot be eliminated from our study. Components with very low ambient concentrations (e.g., cadmium) might be subject to more instrument or laboratory errors. Likewise, components from local sources (e.g., elemental carbon from traffic) might be subject to more error than those regionally transported, given their higher spatial heterogeneity (55). Components with large measurement errors may have smaller regression estimates and lower Am J Epidemiol. 2014;179(9):1086–1095

Downloaded from http://aje.oxfordjournals.org/ at Temple University on November 14, 2014

Ca

−0.5

●

As

Ca

−1.0

●

Al

Effects of PM10 Components on Hospitalization Risks 1093

C)

D) PM10

PM10

EC

EC

OM

OM

NO−3

NO−3

SO2− 4 NH+4

SO2− 4

Na

NH+4 Na+

+

K+

K+

Cl−

−

Pollutants

Pollutants

Cl

Al As

Al As Ca

Cd

Cd

Fe

Fe

Mg

Mg

Mn

Mn

Ni

Ni

Pb

Pb

V

V

−1.0

−0.5

0.0

0.5

1.0

1.5

2.0

2.5

−1.0

−0.5

Change per IQR Increment, %

0.0

0.5

1.0

1.5

2.0

2.5

Change per IQR Increment, %

Figure 3. Percent change in cause-specific emergency hospital admissions per interquartile range (IQR) increment in 2-day moving average (current day and previous day, lag01) of daily particles with aerodynamic diameter less than 10 μm (PM10) and component concentrations. A) Singlepollutant models for cardiovascular disease hospitalization; B) single-pollutant models for respiratory disease hospitalization; C) multipollutant models for cardiovascular disease hospitalization with additional adjustment for nitrogen dioxide (NO2), sulfur dioxide (SO2), and ozone (O3); D) multipollutant models for respiratory disease hospitalization with additional adjustment for NO2, SO2, and O3. All models were adjusted for meteorological factors, seasonal and temporal trend, day of week, and influenza epidemic. Statistically significant associations are represented by lines with hollow circles. Bars, 95% confidence intervals. Al, aluminum; As, arsenic; Ca, calcium; Cd, cadmium; Cl–, chloride ion; EC, elemental carbon; Fe, iron; K+, þ  2 potassium ion; Mg, magnesium; Mn, manganese; Na+, sodium ion; NH4 ; ammonium; Ni; nickel;NO3 ; nitrate; OM, organic matter; Pb, lead; SO4 ; sulfate; V, vanadium.

Table 2. Percent Change in Emergency Hospital Admissions per Interquartile Range Increment in PM10 and Chemical Components at lag01 by Seasona, Adjusted for Meteorological Factors, Seasonal and Temporal Trend, Day of Week, Influenza Epidemic, and Gaseous Copollutantsb, Hong Kong, China, 2001–2007 Cardiovascular Admissions Pollutant

Warm Season

Respiratory Admissions

Cold Season % Change

95% CI

Warm Season % Change

95% CI

Cold Season

% Change

95% CI

% Change

95% CI

PM10

0.91

−0.69, 2.54

0.46

−0.61, 1.55

0.16

−1.33, 1.68

0.55

−0.36, 1.47

Nitrate

0.78

0.13, 1.43

0.67

−0.22, 1.56

−0.15

−0.76, 0.47

0.41

−0.34, 1.17

Sodium ion

0.25

−0.60, 1.11

1.30

0.34, 2.28

−0.04

−0.81, 0.74

0.96

0.15, 1.77

Chloride ion

0.19

−0.18, 0.55

0.49

−0.01, 1.00

−0.16

−0.49, 0.18

0.39

−0.02, 0.81

Aluminum

0.35

−0.78, 1.50

0.15

−0.65, 0.60

−0.06

−1.11, 1.00

0.81

0.13, 1.49

Magnesium

0.43

−0.4, 1.26

0.44

−0.25, 1.14

−0.10

−0.85, 0.66

0.82

0.26, 1.39

Nickel

0.19

−0.37, 0.75

0.79

−0.02, 1.62

0.00

−0.52, 0.52

0.33

−0.37, 1.02

Abbreviations: CI, confidence interval; PM10, particles with aerodynamic diameter less than 10 μm. a Warm season, May–October; cold season, November–April. b Gaseous pollutants include nitrogen dioxide, sulfur dioxide, and ozone.

Am J Epidemiol. 2014;179(9):1086–1095

Downloaded from http://aje.oxfordjournals.org/ at Temple University on November 14, 2014

Ca

1094 Pun et al.

2. 3. 4. 5. 6.

7. 8. 9. 10. 11. 12.

ACKNOWLEDGMENTS 13.

Author affiliations: Jockey Club School of Public Health and Primary Care, Chinese University of Hong Kong, Hong Kong Special Administrative Region of the People’s Republic of China (Vivian Chit Pun, Ignatius Tak-Sun Yu, Hong Qiu, Kin-Fai Ho, Tze Wai Wong, Linwei Tian); School of Public Health and Family Medicine, Capital Medical University, Beijing, the People’s Republic of China (Zhiwei Sun); and Hong Kong Environmental Protection Department, Hong Kong Special Administrative Region of the People’s Republic of China (Peter K.K. Louie). This work was supported bu the Environmental Conservation Fund ( project ECF 35/2010) and the Health and Medical Research Fund (grant 11120311). We thank the Environmental Protection Department of the Hong Kong Special Administrative Region for supplying the air monitoring data; the Hospital Authority for the hospital admission data; and the Hong Kong Observatory for the weather data for this study. Conflict of interest: none declared.

14. 15.

16.

17.

18. 19.

REFERENCES 1. Bell ML, Dominici F, Ebisu K, et al. Spatial and temporal variation in PM2.5 chemical composition in the United States

20.

for health effects studies. Environ Health Perspect. 2007; 115(7):989–995. Pöschl U. Atmospheric aerosols: composition, transformation, climate and health effects. Angew Chem Int Ed Engl. 2005; 44(46):7520–7540. The US National Research Council. Research Priorities for Airborne Particulate Matter: IV. Continuing Research Progress. Washington, DC: National Academy Press; 2004. Health Effects Institute. Understanding the Health Effects of Components of the Particulate Matter Mix: Progress and Next Steps. Cambridge, MA: Health Effects Institute; 2002. Bell ML, Ebisu K, Peng RD, et al. Hospital admissions and chemical composition of fine particle air pollution. Am J Respir Crit Care Med. 2009;179(12):1115–1120. Ito K, Mathes R, Ross Z, et al. Fine particulate matter constituents associated with cardiovascular hospitalizations and mortality in New York City. Environ Health Perspect. 2011; 119(4):467–473. Zanobetti A, Franklin M, Koutrakis P, et al. Fine particulate air pollution and its components in association with cause-specific emergency admissions. Environ Health. 2009;8:58. Cakmak S, Dales RE, Gultekin T. Components of particulate air pollution and emergency department visits in Chile. Arch Environ Occup Health. 2009;64(3):148–155. Metzger KB, Tolbert PE, Klein M, et al. Ambient air pollution and cardiovascular emergency department visits. Epidemiology. 2004;15(1):46–56. Laden F, Neas LM, Dockery DW, et al. Association of fine particulate matter from different sources with daily mortality in six U.S. cities. Environ Health Perspect. 2000;108(10):941–947. Zhou J, Ito K, Lall R, et al. Time-series analysis of mortality effects of fine particulate matter components in Detroit and Seattle. Environ Health Perspect. 2011;119(4):461–466. Cao J, Xu H, Xu Q, et al. Fine particulate matter constituents and cardiopulmonary mortality in a heavily polluted Chinese city. Environ Health Perspect. 2012;120(3):373–378. Huang W, Cao J, Tao Y, et al. Seasonal variation of chemical species associated with short-term mortality effects of PM2.5 in Xi’an, a central city in China. Am J Epidemiol. 2012;175(6): 556–566. Geng F, Hua J, Mu Z, et al. Differentiating the associations of black carbon and fine particle with daily mortality in a Chinese city. Environ Res. 2013;120:27–32. Son J-Y, Lee J-T, Kim K-H, et al. Characterization of fine particulate matter and associations between particulate chemical constituents and mortality in Seoul, Korea. Environ Health Perspect. 2012;120(6):872–878. Wong CM, Atkinson RW, Anderson HR, et al. A tale of two cities: effects of air pollution on hospital admissions in Hong Kong and London compared. Environ Health Perspect. 2002; 110(1):67–77. Wong CM, Vichit-Vadakan N, Kan H, et al. Public Health and Air Pollution in Asia (PAPA): a multicity study of short-term effects of air pollution on mortality. Environ Health Perspect. 2008;116(9):1195–1202. Lee SL, Wong WH, Lau YL. Association between air pollution and asthma admission among children in Hong Kong. Clin Exp Allergy. 2006;36(9):1138–1146. Wong TW, Lau TS, Yu TS, et al. Air pollution and hospital admissions for respiratory and cardiovascular diseases in Hong Kong. Occup Environ Med. 1999;56(10):679–683. Qiu H, Yu IT, Tian L, et al. Effects of coarse particulate matter on emergency hospital admissions for respiratory diseases: a time-series analysis in Hong Kong. Environ Health Perspect. 2012;120(4):572–576. Am J Epidemiol. 2014;179(9):1086–1095

Downloaded from http://aje.oxfordjournals.org/ at Temple University on November 14, 2014

statistical significance (56). The issue of representativeness associated with particulate matter components may hinder the interpretations of the relative strength of the observed associations in monitor-based studies of ambient particulate matter composition. Second, we did not adjust for exposure to multiple components in the regression models because of the moderate to strong correlations among PM10 components. It is possible that the chosen PM10 component is responsible for the observed risk estimates, or the association could be due to the component’s correlation with another component that is the true toxic agent, or multiple components could be jointly responsible. Finally, we did not directly measure coarse particulate matter (i.e., PM2.5–10) composition, nor did we have a long enough period of PM2.5 composition data to subtract coarse particulate matter from PM10 composition measurements. Thus, we cannot assess the health effects of chemical components in coarse particulate matter. In summary, we found evidence that combustion-related particles (i.e., nitrate and nickel), sea salt-related particles (i.e., sodium ion, chloride ion, and magnesium), and particles related to soil/road dust (i.e., aluminum) were significantly associated with cause-specific emergency hospital admissions in Hong Kong in the presence of gaseous copollutants, particularly during the cold season. This study lends support to the growing body of literature concerning the adverse effects of particulate matter chemical composition and also provides insight into the differential health effects of components that are most abundantly found in either fine or coarse fractions of PM10.

Effects of PM10 Components on Hospitalization Risks 1095

Am J Epidemiol. 2014;179(9):1086–1095

40. Zhuang H, Chan CK, Fang M, et al. Formation of nitrate and non-sea-salt sulfate on coarse particles. Atmos Environ. 1999; 33(26):4223–4233. 41. Lee T, Yu X, Ayres B, et al. Observations of fine and coarse particle nitrate at several rural locations in the United States. Atmos Environ. 2008;42(11):2720–2732. 42. Tian L, Ho K, Louie PKK, et al. Shipping emissions associated with increased cardiovascular hospitalizations. Atmos Environ. 2013;74:320–325. 43. Ostro B, Tobias A, Querol X, et al. The effects of particulate matter sources on daily mortality: a case-crossover study of Barcelona, Spain. Environ Health Perspect. 2011;119(12): 1781–1787. 44. Andersen ZJ, Wahlin P, Raaschou-Nielsen O, et al. Ambient particle source apportionment and daily hospital admissions among children and elderly in Copenhagen. J Expo Sci Environ Epidemiol. 2007;17(7):625–636. 45. Mar TF, Ito K, Koenig JQ, et al. PM source apportionment and health effects. 3. Investigation of inter-method variations in associations between estimated source contributions of PM2.5 and daily mortality in Phoenix, AZ. J Expo Sci Environ Epidemiol. 2006;16(4):311–320. 46. Valdés A, Zanobetti A, Halonen JI, et al. Elemental concentrations of ambient particles and cause specific mortality in Santiago, Chile: a time series study. Environ Health. 2012;11:82. 47. Johnson TM, Guttikunda S, Wells GJ, et al. Tools for Improving Air Quality Management: A Review of Top-Down Source Apportionment Techniques and Their Application in Developing Countries. Washington, DC: The World Bank Group; 2011. 48. Lall R, Ito K, Thurston GD. Distributed lag analyses of daily hospital admissions and source-apportioned fine particle air pollution. Environ Health Perspect. 2011;119(4): 455–460. 49. Penttinen P, Vallius M, Tiittanen P, et al. Source-specific fine particles in urban air and respiratory function among adult asthmatics. Inhal Toxicol. 2006;18(3):191–198. 50. Gent JF, Koutrakis P, Belanger K, et al. Symptoms and medication use in children with asthma and traffic-related sources of fine particle pollution. Environ Health Perspect. 2009;117(7):1168–1174. 51. Maciejczyk P, Chen LC. Effects of subchronic exposures to concentrated ambient particles (CAPs) in mice. VIII. Source-related daily variations in in vitro responses to CAPs. Inhal Toxicol. 2005;17(4-5):243–253. 52. Sarnat JA, Marmur A, Klein M, et al. Fine particle sources and cardiorespiratory morbidity: an application of chemical mass balance and factor analytical source-apportionment methods. Environ Health Perspect. 2008;116(4):459–466. 53. Halonen JI, Lanki T, Yli-Tuomi T, et al. Particulate air pollution and acute cardiorespiratory hospital admissions and mortality among the elderly. Epidemiology. 2009;20(1): 143–153. 54. Schreuder AB, Larson TV, Sheppard L, et al. Ambient woodsmoke and associated respiratory emergency department visits in Spokane, Washington. Int J Occup Environ Health. 2006;12(2):147–153. 55. Ito K, Xue N, Thurston G. Spatial variation of PM2.5 chemical species and source-apportioned mass concentrations in New York City. Atmos Environ. 2004;38(31):5269–5282. 56. Lipfert FW, Wyzga RE. Air pollution and mortality: the implications of uncertainties in regression modeling and exposure measurement. J Air Waste Manage Assoc. 1997; 47(4):517–523.

Downloaded from http://aje.oxfordjournals.org/ at Temple University on November 14, 2014

21. Qiu H, Yu IT, Wang X, et al. Differential effects of fine and coarse particles on daily emergency cardiovascular hospitalizations in Hong Kong. Atmos Environ. 2013;64:296–302. 22. Yuan Z, Yadav V, Turner JR, et al. Long-term trends of ambient particulate matter emission source contributions and the accountability of control strategies in Hong Kong over 1998– 2008. Atmos Environ. 2013;76:21–31. 23. Hong Kong Environmental Protection Department. Air quality reports. http://www.aqhi.gov.hk/en/download/air-qualityreports.html. Accessed August 12, 2013. 24. Turpin BJ, Lim H. Species contributions to PM2.5 mass concentrations: revisiting common assumptions for estimating organic mass. Aerosol Sci Technol. 2001;35(1):602–610. 25. Wong CM, Ma S, Hedley AJ, et al. Effect of air pollution on daily mortality in Hong Kong. Environ Health Perspect. 2001; 109(4):335–340. 26. Hastie TJ, Tibshirani RJ. Generalized Additive Models. Boca Raton, FL: Chapman & Hall/CRC; 1990. 27. Peng RD, Bell ML, Geyh AS, et al. Emergency admissions for cardiovascular and respiratory diseases and the chemical composition of fine particle air pollution. Environ Health Perspect. 2009;117(6):957–963. 28. Zeilis A, Grothendieck G. zoo: S3 infrastructure for regular and irrecular time series. J Stat Soft. 2005;14(6):1–27. 29. Wood SN. Fast stable restricted maximum likelihood and marginal likelihood estimation of semiparametric generalized linear models. J R Stat Soc Ser B Stat Methodol. 2011;73(1): 3–36. 30. Host S, Larrieu S, Pascal L, et al. Short-term associations between fine and coarse particles and hospital admissions for cardiorespiratory diseases in six French cities. Occup Environ Med. 2008;65(8):544–551. 31. Atkinson RW, Fuller GW, Anderson HR, et al. Urban ambient particle metrics and health: a time-series analysis. Epidemiology. 2010;21(4):501–511. 32. Kim S-Y, Peel JL, Hannigan MP, et al. The temporal lag structure of short-term associations of fine particulate matter chemical constituents and cardiovascular and respiratory hospitalizations. Environ Health Perspect. 2012;120(8): 1094–1099. 33. Ostro B, Roth L, Malig B, et al. The effects of fine particle components on respiratory hospital admissions in children. Environ Health Perspect. 2009;117(3):475–480. 34. Ostro BD, Feng WY, Broadwin R, et al. The impact of components of fine particulate matter on cardiovascular mortality in susceptible subpopulations. Occup Environ Med. 2008;65(11):750–756. 35. Burnett RT, Cakmak S, Brook JR, et al. The role of particulate size and chemistry in the association between summertime ambient air pollution and hospitalization for cardiorespiratory diseases. Environ Health Perspect. 1997;105(6):614–620. 36. Ostro B, Feng W-Y, Broadwin R, et al. The effects of components of fine particulate air pollution on mortality in California: results from CALFINE. Environ Health Perspect. 2007;115(1):13–19. 37. Reiss R, Anderson EL, Cross CE, et al. Evidence of health impacts of sulfate-and nitrate-containing particles in ambient air. Inhal Toxicol. 2007;19(5):419–449. 38. Kelly FJ, Fussell JC. Size, source and chemical composition as determinants of toxicity attributable to ambient particulate matter. Atmos Environ. 2012;60:504–526. 39. Yuan ZB, Lau AKH, Zhang HY, et al. Identification and spatiotemporal variations of dominant PM10 sources over Hong Kong. Atmos Environ. 2006;40(10):1803–1815.