Science of the Total Environment 408 (2010) 6086–6091

Contents lists available at ScienceDirect

Science of the Total Environment

j ourna l homepage: www.e lsev ie r.com/ locate /sc i totenv

Ambient air pollution and daily mortality in Anshan, China: A time-stratifiedcase-crossover analysis

Renjie Chen a,1, Guowei Pan b,1, Haidong Kan a,⁎, Jianguo Tan c, Weimin Song a, Zhenyu Wu a, Xiaohui Xu d,Qun Xu e, Cheng Jiang f, Bingheng Chen a

a School of Public Health and Key Lab of Public Health Safety of the Ministry of Education, Fudan University, Shanghai, Chinab Liaoning Provincial Center for Disease Control and Prevention, Shenyang, Chinac Shanghai Climate Center, Shanghai, Chinad Department of Epidemiology and Biostatistics, College of Public Health and Health Professions, University of Florida, Gainesville, FL, USAe Department of Epidemiology and Health Statistics, Institute of Basic Medical Sciences, Chinese Academy of Medical Sciences, Peking Union Medical College, Beijing, Chinaf Shanghai Municipal Center of Disease Control and Prevention, Shanghai, China

⁎ Corresponding author. P.O. Box 249, 130 Dong-AnTel./fax: +86 21 6404 6351.

E-mail address: haidongkan@gmail.com (H. Kan).1 These authors contributed equally to this work.

0048-9697/$ – see front matter © 2010 Elsevier B.V. Adoi:10.1016/j.scitotenv.2010.09.018

a b s t r a c t

a r t i c l e i n f o

Article history:Received 30 May 2010Received in revised form 6 September 2010Accepted 13 September 2010

Keywords:Air pollutionCase-crossoverMortality

Few case-crossover studies were conducted in China to investigate the acute health effects of air pollution.Weconducted a time-stratified case-crossover analysis to examine the association between air pollution and dailymortality in Anshan, a heavily-polluted industrial city in northeastern China. Dailymortality, air pollution, andweather data in 2004–2006 in Anshan were collected. Time-stratified case-crossover approach was used toestimate the effect of air pollutants (PM10, SO2, NO2 and CO) on total and cardiopulmonary mortality. Controlswere selected as matched days of the week in the same month. Potential effect modifiers, such as gender andage, were also examined. We found significant associations between air pollution and daily mortality fromcardiovascular diseases in Anshan. A 10 μg/m3 elevation of 2-day moving average (lag 01) concentration inPM10, SO2, NO2 and CO corresponded to 0.67% (95% CI: 0.29%, 1.04%), 0.38% (95% CI: −0.06%, 0.83%), 2.11%(95% CI: 0.22%, 4.00%) and 0.04% (95% CI: 0.01%, 0.07%) increase of cardiovascular mortality. The associationsfor total and respiratory mortality were generally positive but statistically insignificant. The air pollutionhealth effects were significantly modified by age, but not by gender. Conclusively, our study showed thatshort-term exposure to air pollution was associated with increased cardiovascular mortality in Anshan. Thesefindings may have implications for local environmental and social policies.

Road, Shanghai 200032, China.

ll rights reserved.

© 2010 Elsevier B.V. All rights reserved.

1. Introduction

Outdoor air pollution has become a major concern for publichealth world widely. Short-term exposure to outdoor air pollution hasbeen linked to increased mortality, increased rates of hospitaladmissions and emergency department visits, exacerbation ofchronic respiratory conditions (e.g., asthma), and decreased lungfunction (Kampa and Castanas, 2008). Most of these studies wereconducted in developed countries, and only a small number of studieshave been conducted in Asia developing countries (Health EffectsInstitute, 2004). There remains a need for studies in cities ofdeveloping countries, where characteristics of outdoor air pollution(e.g. air pollution level and mixture, transport of pollutants), andsocio-demographic status of local residents (e.g. disease pattern, agestructure, and socioeconomic characteristics), may be different from

North America and Europe, where most epidemiologic studies wereconducted.

Due to long-standing heavy reliance on coal in energy structureand drastic industrialization, urbanization and motorization in recentdecades, air pollution levels in Chinese cities are among the highest inthe world (Kan, 2009). Time-series analysis has been used to examinethe relationship between outdoor air pollution and daily mortality/morbidity in several large Chinese cities, including Beijing (Guo et al.,2009; Xu et al., 1994), Chongqing (Venners et al., 2003), Shanghai(Kan et al., 2008), and Shenyang (Xu et al., 2000). Recent multi-citytime-series analysis conducted in several Chinese cities providesfurther evidence supporting coherence and plausibility of the acuteeffect of air pollutants on cardio-respiratory health (Wong et al.,2008). Case-crossover and time-series designs are often viewed ascompeting methods in air pollution epidemiogy (Carracedo-Martinezet al., 2010); however, few case-crossover studies were conducted inChina to investigate the acute health effects of air pollution.

In the current study, we conducted a time-stratified case-crossoveranalysis to examine the association between ambient air pollutionand daily mortality in Anshan, a heavily-polluted industrial city innortheastern China.

Table 1Summary statistics of daily death numbers, air pollutant concentrations and weatherconditions in Anshan, China (2004–2006).

Mean±SD Min P(25) P(50) P(75) Max

Daily death countsAll-cause 27.6±6.1 12 23 27 31 65Cardiovascular 14.1±4.3 4 11 14 17 34Respiratory 1.9±1.5 0 1 2 3 8

Air pollutants concentrations (μg/m3)PM10 110.9±60.2 15 69 98 139 469SO2 59.0±74.3 3 10 26 83 57NO2 25.5±16.3 5 14 21 33 116CO 1094.7±787.1 10 515 895 1522 5489

WeatherTemperature (°C) 11.4±12.7 −19 1 14 23 36Humidity (%) 55.2±16.0 16 44 55 67 97

6087R. Chen et al. / Science of the Total Environment 408 (2010) 6086–6091

2. Materials and methods

2.1. Data

Our study area includes the urban Districts of Anshan (624 km2)and had a population of 1.46 million by the end of 2006. We excludedthe sub-urban Districts due to inadequate air pollution monitoringstations in that area. We obtained detailed death records of urbanresidents in Anshan from January 1, 2004 to December 31, 2006 fromLiaoning Provincial Center for Disease Control and Prevention.Information recorded on the death records include gender, age, dateof death, and cause of death determined by the InternationalClassification of Diseases, Tenth Revision (ICD-10). We classifiedmortality data into deaths due to total non-accidental causes (ICD-10:A00-R99), cardiovascular diseases (ICD-10: I00-I99), and respiratorydiseases (ICD-10: J00-J98).

Air quality indicators in our study include particulate matter withaerodynamic diameter of ≤10 μm (PM10), sulfur dioxide (SO2),nitrogen dioxide (NO2) and carbon monoxide (CO). We obtaineddaily air quality data from the China National EnvironmentalMonitoring Center (CNEMC). CNEMC is part of the State Environ-mental Protection Ministry, China, and has authority over the entireenvironmental monitoring network at all levels – provincial, city –

throughout China. The daily concentrations for each pollutant wereaveraged from the available monitoring results of six fixed-sitestations in Anshan. According to relevant rules of Chinese govern-ment, the location of these stations should not be in the direct vicinityof traffic or of industrial sources. Moreover, the locations should notbe influenced by local pollution sources and should also avoidbuildings, or housing large emitters such as coal-, waste-, or oil-burning boilers, furnaces, and incinerators. Thus our monitoringresults reflect the general background urban air pollution level inAnshan rather than local sources such as traffic or industrialcombustion.

To allow adjustment for the effect of weather on mortality,meteorological data (daily mean temperature and humidity) wereobtained from Anshan Meteorological Bureau.

2.2. Statistical analysis

We analyzed the air pollution, mortality and weather data with acase-crossover design. Case-crossover design, which was firstlyproposed by Maclure to study the transient effect of an intermittentexposure (Maclure, 1991), has been used to study the acute healtheffects of air pollution (Carracedo-Martinez et al., 2010; Jaakkola,2003). The case-crossover approach is a design inwhich only cases aresampled, and their exposure at the time of their failure is comparedwith some estimate of their typical level of exposure. This approachonly requires exposure data for cases and can be regarded as a specialtype of case–control study in which case serves his/her referent.Therefore, case-crossover design has the advantage of controlling forpotential confounding caused by fixed individual characteristics, suchas gender, race, diet, and age. Recently, time-stratified referentselection, i.e. time is divided fixed strata and using the remainingdays in a stratum as referents for a case that falls in that stratum, wasshown to be able to avoid some subtle selection bias issues and resultin a unbiased effect estimate using conditional logistic regressionmodels (Levy et al., 2001). In our analysis, we selected control daysmatched on the same day of the week in the same month of the sameyear when a death occurred (a case day). This control selectionstrategy was expected to adjust for the effects of long-term trend,seasonality, and day of week by design (Zanobetti and Schwartz,2005).

Previous studies reported the association between weatherconditions (temperature and relative humidity) and mortality wasgenerally nonlinear (Basu and Samet, 2002). We therefore adjusted

for temperature and relative humidity on the concurrent day of deathusing a natural smooth spline with 3 degrees of freedom (Zanobettiand Schwartz, 2009).

Both total non-accidental and cause-specific mortality wereassessed. We fitted both single-pollutant models and models with adifferent combination of pollutants to assess the stability ofpollutants' effect. Single-day lag models were reported to underesti-mate the cumulative effect of air pollution on mortality (Kan et al.,2008;Wong et al., 2008); therefore, we used 2-day moving average ofcurrent and previous day concentrations of air pollutants (lag0–1) forour main analyses as a priori lag. As a sensitivity analysis, we alsoexamined the effect of air pollutants with different lag structuresincluding both single-day lag (from lag0 to lag4) and multi-day lag(lag01 and lag04). In single-day lag models, a lag of 0 day (lag0)corresponds to the current-day pollutant concentration, and a lag of1 day (lag1) refers to the previous day concentration; in multi-day lagmodels, lag04 corresponds to 5-day moving average of pollutantconcentration of the current and previous 4 days.

To explore the potential effect modifier, we stratified the analysesby gender (males and females) and age group (5–64 years, 65–74 years, and ≥75 years). Deaths under age 5 were too few andtherefore were excluded from our analysis. We tested the statisticalimportant significance of differences between effect estimates of thestrata divided by potential effect modifiers by calculating the 95%

confidence interval (CI) as⌢Q1−

⌢Q2

� �F1:96

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiðS⌢E1Þ2 + ðS⌢E2Þ2

r, where

Q1 and Q2 are the effect estimates for the two categories (i.e. malesand females), and SE1 and SE2 are their respective standard errors(Zeka et al., 2006).

We carried out time-stratified referent selection using thestatistical software package SAS, version 9.1 (SAS Institute Inc, Cary,NC), and conducted conditional logistic regression analysis using theCOXPH procedure in R 2.11.0 (R Development Core Team, 2010). Theresults are presented as the percent change in daily mortality per10 μg/m3 increase of 2-day moving average (lag0–1) pollutantconcentrations unless specified otherwise.

3. Results

From 2004 to 2006 (1096 days), a total of 31,847 deaths (18,369males and 13,478 females) were identified in the study population. Thepercentages of total deaths by age group were 30.7% for 5–64 years,31.1% for 65–74 years and 37.4% for ≥75 years. On average, there wereapproximately 27.6 non-accidental deaths per day, including 14.1 fromcardiovascular disease and 1.9 from respiratory disease (Table 1).

During our study period, the mean daily average concentrations ofPM10, SO2, NO2 and CO were 110.9, 59.0, 25.5 and 1094.7 μg/m3,

Table 2Pearson correlation coefficients between daily air pollutant concentrations andweatherconditions in Anshan (2004–2006).

SO2 NO2 CO Temperature Humidity

PM10 0.37⁎ 0.55⁎ 0.59⁎ −0.11⁎ −0.12⁎

SO2 0.60⁎ 0.50⁎ −0.73⁎ −0.10⁎

NO2 0.73⁎ −0.31⁎ −0.05CO −0.22⁎ 0.09⁎

Temperature 0.28⁎

⁎ pb0.01.

6088 R. Chen et al. / Science of the Total Environment 408 (2010) 6086–6091

respectively. The mean daily average temperature and humidity were11.4 °C and 55.2% in Anshan (Table 1).

Table 2 shows the Pearson correlation coefficients between airpollutants' concentrations, temperature and humidity. PM10, SO2, NO2

and CO were closely correlated with each other with the Pearsoncorrelation coefficients ranging from 0.37 to 0.73 (pb0.01 for allcoefficients). Generally, air pollutant concentrations were negativelycorrelated with temperature and humidity.

In the single-pollutant model, positive and statistically significantassociations were found between levels of air pollutants (PM10, NO2,and CO) and cardiovascular mortality (Table 3). An increase of 10 μg/m3 in 2-daymoving average (lag 01) concentrations of PM10, SO2, NO2

and CO corresponds to 0.67% [95% confidential interval (CI): 0.29%,1.04%], 0.38% (95% CI: −0.06%, 0.83%), 2.11% (95% CI: 0.22%, 4.00%)and 0.04% (95% CI: 0.01%, 0.07%) increase of cardiovascular mortality,respectively. For total and respiratory mortality, the associations weregenerally positive but statistically insignificant. SO2 was not signifi-cantly associated with any mortality outcomes.

In the multiple-pollutant model, the effects of PM10 on cardiovas-cular mortality remained significant after adding SO2, NO2 or CO(Table 3). The effects of NO2 and CO on cardiovascular mortalitydecreased and became insignificant after adjustment for co-pollu-tants. We generally did not find significant association for total orrespiratory mortality in the multiple-pollutant model.

Fig. 1 presents the lag pattern of air pollutants' effect oncardiovascular mortality. Statistically significant associations wereobserved for some, but not all, lag structures of pollutants'concentrations. Generally, multi-day exposures (lag01 to lag04)have larger effects than single-day exposure (lag0 to lag4). Theeffects of air pollutants (PM10, NO2 and CO) on cardiovascular

Table 3Percent increase of daily mortality associated with 10 μg/m3 increase of pollutant concentr

Total mortality

PM10 Without adjustment 0.24 (−0.03, 0.51Adjusted for SO2 0.16 (−0.17, 0.50Adjusted for NO2 0.13 (−0.23, 0.48Adjusted for CO 0.11 (−0.24, 0.47Adjusted for SO2+NO2+CO 0.07(−0.33, 0.46

SO2 Without adjustment 0.27 (−0.05, 0.60Adjusted for PM10 0.16 (−0.23, 0.56Adjusted for NO2 0.13 (−0.29, 0.55Adjusted for CO 0.14 (−0.25, 0.53Adjusted for PM10+NO2+CO 0.08 (−0.36, 0.52

NO2 Without adjustment 1.30 (−0.06, 2.67Adjusted for PM10 0.89 (−0.91, 2.69Adjusted for SO2 0.94 (−0.85, 2.74Adjusted for CO 0.66 (−1.32, 2.65Adjusted for PM10+SO2+CO 0.36 (−1.90, 2.61

CO Without adjustment 0.02 (0.00, 0.05)⁎

Adjusted for PM10 0.02 (−0.01, 0.05Adjusted for SO2 0.02 (−0.01, 0.04Adjusted for NO2 0.01 (−0.02, 0.05Adjusted for PM10+SO2+NO2 0.01 (−0.02, 0.05

⁎ pb0.05.

mortality were statistically significant for most multi-day laggeddays we examined.

Daily deaths due to respiratory disease were too few and thereforewere excluded from our stratified analysis by gender and age (Tables 4and 5). The air pollution effects on total and cardiovascular mortalityvaried by gender (Table 4). For total mortality, sex-specific effectestimates were insignificant for all pollutants, except that CO hadstatistically significant effect in males. For cardiovascular mortality,effect estimates of PM10 and CO in males were statistically significantand slightly higher than in females; in contrast, the effect of SO2 andNO2 in females was slightly higher than in males. It should beemphasized that for both total and cardiovascularmortality, between-gender differences were insignificant for all pollutants we examined.

For residents aged 5–64 years or 65–74 years, we did not findstatistically significant association (Table 5). For residents aged 75 orabove, the effects of all four air pollutants on both total andcardiovascular mortality were statistically significant. The differencesbetween the 5–64 and ≥75 groups were statistically significant formost pollutants.

4. Discussion

Evidence gained in this case-crossover analysis showed thatoutdoor air pollution (PM10, NO2 and CO) was associated withmortality from cardiovascular disease in Anshan in 2004–2006. Theassociations for total and respiratory mortality were generallypositive but statistically insignificant. The elderly were morevulnerable to air pollution than younger people. Our analysis showedthat short-term exposure to air pollution was associated withincreased cardiovascular mortality in Anshan, China. These findingsmay have implications for environmental and social policies inAnshan, and for the local government to take steps to protecthuman health.

Most previous studies on the short-term association between airpollution and health outcomes have been based on Poisson regressiontime-series approach. Case-crossover design offers the ability tocontrol confounders by design rather than by statistical modeling,thus offering the opportunity to improve causal inference of airpollution health effects. Because each subject serves as his or her owncontrol, invariant subject-specific covariates such as gender and agedo not act as confounders (Maclure, 1991). Also, by choosing thecontrol periodwithin a fewweeks of death, this approach reduced any

ations with single and multiple-pollutant models in Anshan (mean and 95% CI).

Cardiovascular mortality Respiratory mortality

) 0.67 (0.29, 1.04)⁎ 0.21(−0.82, 1.24)) 0.73 (0.26, 1.19)⁎ 0.31(−1.01, 1.62)) 0.69 (0.19, 1.18)⁎ 0.44(−0.97, 1.85)) 0.66 (0.16, 1.16)⁎ 0.36(−1.01, 1.73)) 0.70 (0.16, 1.25)⁎ 0.46(−1.06, 1.98)) 0.38 (−0.06, 0.83) 0.04(−1.16, 1.24)) −0.12 (−0.67, 0.43) −0.18(−1.72, 1.35)) 0.11 (−0.48, 0.69) 0.14(−1.57, 1.85)) 0.13 (−0.41, 0.67) 0.05(−1.44, 1.54)) −0.14 (−0.75, 0.48) 0.01 (−1.77, 1.79)) 2.11 (0.22, 4.00)⁎ −0.18(−5.39, 5.02)) −0.15 (−2.66, 2.35) −1.72(−8.85, 5.42)) 1.81 (−0.69, 4.32) −0.63(−8.04, 6.78)) 1.06 (−1.70, 3.81) −0.52(−8.46, 7.41)) −0.01 (−3.14, 3.13) −1.55 (−10.99, 7.89)

0.04 (0.01, 0.07)⁎ 0.00(−0.08, 0.09)) 0.00 (−0.04, 0.04) −0.02(−0.13, 0.09)) 0.03 (−0.01, 0.07) 0.00(−0.11, 0.10)) 0.02 (−0.02, 0.07) 0.01(−0.12, 0.14)) 0.00 (−0.05, 0.05) −0.01 (−0.14, 0.13)

PM10

-1.0

-0.5

0.0

0.5

1.0

1.5

0 1 2 3 4 0-1 0-2 0-3 0-4Lag

0 1 2 3 4 0-1 0-2 0-3 0-4Lag

0 1 2 3 4 0-1 0-2 0-3 0-4Lag

0 1 2 3 4 0-1 0-2 0-3 0-4

Lag

Exc

ess

risk

(%)

-1.0

-0.5

0.0

0.5

1.0

1.5

Exc

ess

risk

(%)

SO2

NO2

-2.0

0.0

2.0

4.0

Exc

ess

risk

(%)

CO

-0.1

0.0

0.1

0.1

Exc

ess

risk

(%)

Fig. 1. Percent increase of daily cardiovascular mortality associated with 10 μg/m3 increase of pollutant concentrations, using different lag structure of pollutants.

6089R. Chen et al. / Science of the Total Environment 408 (2010) 6086–6091

potential confounding role of the time trends, seasonality, andweather variables. Using time-series generalized additive model(GAM) approach, we found similar association between air pollutionand cardiovascular mortality in Anshan (data not shown). It is notclear that the case-crossover design is necessarily superior or inferiorto the time-series approach, but the fact that both case-crossover andtime-series methods provided relatively similar results suggested thatthe association between air pollution and cardiovascular mortalitywas reasonably robust and probably not due tomethodological bias orconfounders.

With the rapid economic development and consequent improve-ment in living conditions, nutrition, and health care in China,cardiovascular disease has placed much more disease burden onpopulation than ever. It was estimated that cardiovascular disease isnow the leading cause of death in the Chinese population of adults40 years of age and older, accounting for around 40% of total mortality(He et al., 2005). Our finding adds to increasing evidence thatexposure to higher levels of air pollution is associated with adversecardiovascular consequences. A recent scientific statement from theAmerican Heart Association concluded that short-term or long-termexposure to air pollution is associated with increased risk ofcardiovascular disease and death (Brook et al., 2004). Recently therehave been many studies concerned with the potential mechanisms

Table 4Gender-specific percent increase of daily mortality of Anshan residents associated with 10

Cause of death Mean daily death no. PM10

Total mortalityMale 15.7 0.28 (−0.08, 0.64)Female 11.9 0.19 (−0.22, 0.60)

Cardiovascular mortalityMale 8.0 0.74 (0.24, 1.24)⁎

Female 6.2 0.57 (0.00, 1.14)⁎

⁎ pb0.05.

linking air pollution and cardiovascular diseases. Air pollutants mayadversely affect the cardiovascular system directly and indirectly.Direct effects may occur via agents that readily cross the pulmonaryepithelium into the circulation, such as gaseous pollutants (e.g.,nitrogen oxides) and possibly ultra-fine particles (UFPs), along withsoluble constituents of particles (e.g., transition metals); indirecteffects may occur via induction of pulmonary oxidative stress andinflammation, leading to endothelial dysfunction and systemicinflammation (Bhatnagar, 2006; Nawrot et al., 2006). All these directand indirect effects may in turn activate hemostatic pathways andimpair cardiovascular function.

In reality, people cannot selectively inhale some air pollutants andnot others. Identification of the specific pollutants contributing mostto the health hazard of the air pollution mixture may have importantimplications for environmental and social policies. In China, however,many researchers have reported controversial findings. Our studyfound significant effect of PM10 on cardiovascular mortality in bothsingle-pollutant and multi-pollutant models while the effect ofgaseous pollutants (NO2 and CO) decreased and became insignificantafter adjustment for PM10 (Table 3). Similarly, in Shenyang, totalsuspended particle (TSP), but not gaseous pollutants, was a significantpredictor of cardiovascular death (Xu et al., 2000). In contrast, it wasSO2, but not TSP nor find particle, that was significantly associated

μg/m3 increase in air pollutant concentrations (mean and 95% CI).

SO2 NO2 CO

0.25 (−0.18, 0.68) 1.55 (−0.26, 3.36) 0.04 (0.01, 0.07)⁎

0.30 (−0.18, 0.79) 0.94 (−1.14, 3.01) 0.01 (−0.03, 0.04)

0.31 (−0.28, 0.91) 1.90 (−0.62, 4.43) 0.04 (0.00, 0.08)⁎

0.47 (−0.20, 1.13) 2.28 (−0.58, 5.14) 0.03 (−0.02, 0.08)

Table 5Age-specific percent increase of daily mortality of Anshan residents associated with 10 μg/m3 increase in air pollutant concentrations (mean and 95% CI).

Cause of death Mean daily death no. PM10 SO2 NO2 CO

Total mortality5–64 7.74 −0.17 (−0.70, 0.35) −0.22 (−0.85, 0.41) −0.02 (−2.66, 2.61) 0.01 (−0.03, 0.06)65–74 8.90 0.22 (−0.26, 0.70) 0.24 (−0.32, 0.80) 0.90 (−1.52, 3.33) 0.02 (−0.02, 0.06)≥75 10.74 0.54 (0.11, 0.97)⁎ 0.66 (0.15, 1.17)⁎ 2.80 (0.65, 4.95)⁎ 0.03 (0.00, 0.07)⁎

Cardiovascular mortality5–64 3.27 0.55 (−0.24, 1.34) 0.03 (−0.90, 0.97) 0.12 (−3.90, 4.14) 0.03 (−0.04, 0.09)65–74 4.64 0.12 (−0.55, 0.78) 0.12 (−0.64, 0.89) 0.76 (−2.57, 4.09) 0.02 (−0.04, 0.07)≥75 6.23 1.15 (0.59, 1.71)⁎ 0.77 (0.10, 1.44)⁎ 4.40 (1.60, 7.19)⁎ 0.06 (0.01, 0.11)⁎

⁎ pb0.05.

6090 R. Chen et al. / Science of the Total Environment 408 (2010) 6086–6091

with daily cardiovascular mortality in Beijing (Xu et al., 1994) andChongqing (Venners et al., 2003). This controversy may reflect thedifferent characteristics of the study sites as well as differences inanalytic techniques used in various studies. It is also worth noting thatthe observed health effects attributed to the ambient gaseouspollutants, e.g. SO2 and NO2, might result from exposures to particles(Sarnat et al., 2005).

Various factors may modify the health effects of air pollution. Wedid not find significant evidence for effect modification by gender.Some studies have reported stratified analyses of air pollution healtheffects by gender, and the patterns are not conclusive (Clougherty,2010). Like a few other studies (Stieb et al., 2002, 2003), we foundsignificant effects of air pollution in the elderly only. Two age groups,the elderly and the very young, are presumed to be at greater risk forair pollution-related effects (Schwartz, 2004). Small numbers ofdeaths in the age group 0–4 in our study limited the power to detectthe effects of air pollution even if they exist. Older people tend to befrailer and to have a higher prevalence of cardio-respiratory diseasesand symptoms, therefore it is plausible that they may have increasedsusceptibility to the detrimental effects of air pollutants (Kelsall et al.,1997). Considering the population aging process and the change indisease pattern in China (from the traditional infectious disease tochronic non-communicable disease, e.g. cardiopulmonary diseases),the impact of air pollution on public health could be very extensive inChina. Additional investigations of potential modifying factors inChina will help in public policy making, risk assessment and standardsetting.

Some limitations must also be taken into account. First, as in otherstudies in this field, we derived daily average levels of air pollutantsandweather variables from outdoormonitoring stations as proxies forindividual exposure levels, so that the personal exposure may havediffered substantially from estimated exposure. Measurement errormay have substantial implications for interpreting epidemiologicstudies on air pollution, particularly for the case-crossover or time-series analyses (Zeger et al., 2000). It is possible that this type of errormay introduce bias to our analysis results. However, due to lack ofavailable information on personal exposure, we could not quantifysuch a bias. Second, as Bateson and Schwartz pointed (Bateson andSchwartz, 2004), the statistical power is slightly lower in the case-crossover approach compared with the time-series analysis, so wemight have ignored some significant effects or differences betweenstrata. Third, validity of cause of death was limited, since we had onlyaccess to the death certificates. Thus, misclassification between deathcauses may be possible.

5. Conclusions

Current air pollution level in Anshan was related to an increasedmortality from cardiovascular disease. Our finding strengthens theevidence of the short-term effect of air pollution onmortality in China(Kan et al., 2009) and worldwide (Pope and Dockery, 2006), and mayhave implications for local environmental and social policies.

Acknowledgement

The study was supported by the National Basic Research Program(973 program) of China (2011CB503802), Gong-Yi Program of ChinaMinistry of Environmental Protection (200809109, 200909016),National Natural Science Foundation of China (30800892), ShanghaiPu Jiang Program (09PJ1401700), and Program for New CenturyExcellent Talents in University (NCET-09-0314). Weimin Song wassupported by the National High Technology Research and Develop-ment Program of China (863 Program) (2007AA06Z409). Jianguo Tanwas supported by the Gong-Yi Program of China MeteorologicalAdministration (GY200706019).

The authors declare they have no competing financial interests.

References

Basu R, Samet JM. Relation between elevated ambient temperature and mortality: areview of the epidemiologic evidence. Epidemiol Rev 2002;24:190–202.

Bateson TF, Schwartz J. Who is sensitive to the effects of particulate air pollution onmortality? A case-crossover analysis of effect modifiers. Epidemiology 2004;15:143–9.

Bhatnagar A. Environmental cardiology: studying mechanistic links between pollutionand heart disease. Circ Res 2006;99:692–705.

Brook RD, Franklin B, Cascio W, Hong Y, Howard G, Lipsett M, et al. Air pollution andcardiovascular disease: a statement for healthcare professionals from the ExpertPanel on Population and Prevention Science of the American Heart Association.Circulation 2004;109(21):2655–71.

Carracedo-Martinez E, Taracido M, Tobias A, Saez M, Figueiras A. Case-crossoveranalysis of air pollution health effects: a systematic review of methodology andapplication. Environ Health Perspect 2010;118(8):1173–82.

Clougherty JE. A growing role for gender analysis in air pollution epidemiology. EnvironHealth Perspect 2010;118:167–76.

Guo Y, Jia Y, Pan X, Liu L, Wichmann HE. The association between fine particulate airpollution and hospital emergency room visits for cardiovascular diseases in Beijing,China. Sci Total Environ 2009;407(17):4826–30.

He J, Gu D, Wu X, Reynolds K, Duan X, Yao C, et al. Major causes of death among menand women in China. N Engl J Med 2005;353(11):1124–34.

Health Effects Institute. Health effects of outdoor air pollution in developing countriesof Asia: a literature review; 2004. Boston, MA, USA.

Jaakkola JJ. Case-crossover design in air pollution epidemiology. Eur Respir J Suppl2003;40:81s–5s.

Kampa M, Castanas E. Human health effects of air pollution. Environ Pollut 2008;151:362–7.

Kan H. Environment and health in china: challenges and opportunities. Environ HealthPerspect 2009;117:A530–1.

Kan H, London SJ, Chen G, Zhang Y, Song G, Zhao N, et al. Season, sex, age, and educationas modifiers of the effects of outdoor air pollution on daily mortality in Shanghai,China: the Public Health and Air Pollution in Asia (PAPA) study. Environ HealthPerspect 2008;116(9):1183–8.

Kan H, Chen B, Hong C. Health impact of outdoor air pollution in China: currentknowledge and future research needs. Environ Health Perspect 2009;117(5):A187.

Kelsall JE, Samet JM, Zeger SL, Xu J, et al. Air pollution and mortality in Philadelphia,1974–1988. Am J Epidemiol 1997;146(9):750–62.

Levy D, Lumley T, Sheppard L, Kaufman J, Checkoway H. Referent selection in case-crossover analyses of acute health effects of air pollution. Epidemiology 2001;12(2):186–92.

Maclure M. The case-crossover design: a method for studying transient effects on therisk of acute events. Am J Epidemiol 1991;133:144–53.

Nawrot TS, Nemmar A, Nemery B. Air pollution: to the heart of the matter. Eur Heart J2006;27(19):2269–71.

Pope III CA, Dockery DW. Health effects of fine particulate air pollution: lines thatconnect. J Air Waste Manag Assoc 2006;56:709–42.

6091R. Chen et al. / Science of the Total Environment 408 (2010) 6086–6091

R Development Core Team. R: a language and environment for statistical computing,version 2.11.0. Vienna: R Foundation for Statistical Computing; 2010.

Sarnat JA, Brown KW, Schwartz J, Coull BA, Koutrakis P. Ambient gas concentrations andpersonal particulate matter exposures: implications for studying the health effectsof particles. Epidemiology 2005;16(3):385–95.

Schwartz J. Air pollution and children's health. Pediatrics 2004;113:1037–43.Stieb DM, Judek S, Burnett RT. Meta-analysis of time-series studies of air pollution and

mortality: effects of gases and particles and the influence of cause of death, age, andseason. J Air Waste Manag Assoc 2002;52(4):470–84.

Stieb DM, Judek S, Burnett RT. Meta-analysis of time-series studies of air pollution andmortality: update in relation to the use of generalized additive models. J Air WasteManag Assoc 2003;53(3):258–61.

Venners SA, Wang B, Xu Z, Schlatter Y, Wang L, Xu X. Particulate matter, sulfur dioxide,and daily mortality in Chongqing, China. Environ Health Perspect 2003;111(4):562–7.

Wong CM, Vichit-Vadakan N, Kan H, Qian Z. 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):1195–202.

Xu X, Gao J, Dockery DW, Chen Y. Air pollution and dailymortality in residential areas ofBeijing, China. Arch Environ Health 1994;49(4):216–22.

Xu Z, Yu D, Jing L, Xu X. Air pollution and daily mortality in Shenyang, China. ArchEnviron Health 2000;55(2):115–20.

Zanobetti A, Schwartz J. The effect of particulate air pollution on emergency admissionsfor myocardial infarction: a multicity case-crossover analysis. Environ HealthPerspect 2005;113:978–82.

Zanobetti A, Schwartz J. The effect of fine and coarse particulate air pollution onmortality: a national analysis. Environ Health Perspect 2009;117:898–903.

Zeger SL, Thomas D, Dominici F, Samet JM, Schwartz J, Dockery D, et al. Exposuremeasurement error in time-series studies of air pollution: concepts andconsequences. Environ Health Perspect 2000;108(5):419–26.

Zeka A, Zanobetti A, Schwartz J, et al. Individual-level modifiers of the effects ofparticulate matter on daily mortality. Am J Epidemiol 2006;163(9):849–59.