ta)

Qu

an Untre

ineseMedicine, Beijing, China

a r t i c l e i n f o

Science of the Total Environment 409 (2011) 49234928

Contents lists available at SciVerse ScienceDirect

Science of the Tot

l se 2011 Elsevier B.V. All rights reserved.

1. Introduction

Air masses always contain many pollutants in differing amounts,including both particulate matter (PM) and gaseous pollutants.Although the strongest evidence linking air pollutants with adversehealth effects thus far is for PM (Pope and Dockery, 2006), manystudies have reported associations for gaseous pollutants such asnitrogen dioxide (NO2) (Samoli et al., 2006), ozone (O3) (Bellet al., 2004a), sulfur dioxide (SO2) (Kan et al., 2010), and carbonmonoxide (CO). CO is a colorless, odorless, and tasteless air toxinwhich is produced by incomplete combustion of hydrocarbons. Inurban areas, CO is primarily generated by motor vehicle emission.

et al., 2006; Yang et al., 1998). Recent multi-city analyses conductedin the U.S. and Europe provide further evidence supportingcoherence and plausibility of the associations (Bell et al., 2009;Samoli et al., 2007). However, most of these studies were conductedin developed countries.

Coal is still the major source of energy, constituting about 75% of allenergy sources. Consequently, air pollution in China predominantlyconsists of coal smoke, with suspended particulate matter (PM) andsulfur dioxide (SO2) as the principal air pollutants. In terms of PMand SO2, China may have the worst air pollution level in the world(Kan et al., 2011). In large cities, however, with the rapid increase inthe number of motor vehicles, air pollution has gradually changedPreviously, epidemiologic studies have repotions of ambient CO with daily mortality andvascular diseases (Allred et al., 1989; Dales,

Corresponding author at: P.O. Box 249, 130 DonChina. Tel./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 2011 Elsevier B.V. Alldoi:10.1016/j.scitotenv.2011.08.029ndings were generally insensitive to alternative model specications. In conclusion, ambient CO wasassociated with increased risk of daily mortality in these three cities. Our ndings suggest that the role ofexposure to CO and other trafc-related air pollutants should be further investigated in China.Article history:Received 7 June 2011Received in revised form 9 August 2011Accepted 10 August 2011Available online 10 September 2011

Keywords:Air pollutionCAPESCarbon monoxideMortalityTime-seriesa b s t r a c t

Ambient carbon monoxide (CO) is an air pollutant primarily generated by trafc. CO has been associated withincreased mortality and morbidity in developed countries, but few studies have been conducted in Asiandeveloping countries. In the China Air Pollution and Health Effects Study (CAPES), the short-term associationsbetween ambient CO and daily mortality were examined in three Chinese cities: Shanghai, Anshan andTaiyuan. Poisson regression models incorporating natural spline smoothing functions were used to adjustfor long-term and seasonal trend of mortality, as well as other time-varying covariates. Effect estimateswere obtained for each city and then for the cities combined. In both individual-city and combined analysis,signicant associations of CO with both total non-accidental and cardiovascular mortality were observed. Inthe combined analysis, a 1 mg/m3 increase of 2-day moving average concentrations of CO corresponded to2.89% (95%CI: 1.68, 4.11) and 4.17% (95%CI: 2.66, 5.68) increase of total and cardiovascular mortality, respec-tively. CO was not signicantly associated with respiratory mortality. Sensitivity analyses showed that ourrted short-term associa-morbidity from cardio-2004; Riojas-Rodriguez

from the conventtion/motor vehicair pollution (e.gants), meteoroloChina are differeknowledge, no pacute health effecountries.

g-An Road, Shanghai 200032,

rights reserved.Department of Epidemiology and Biostatistics, C Public Health and Health Professions, University of Florida, Gainesville, FL, United Statesf Department of Preventive Medicine, School of Management, Beijing University of Chineseg ollege ofAmbient carbon monoxide and daily morPollution and Health Effects Study (CAPES

Renjie Chen a,b,1, Guowei Pan c,1, Yanping Zhang d,1,Bingheng Chen a,b, Haidong Kan a,b,a School of Public Health, Key Lab of Public Health Safety of the Ministry of Education, Fudb G_RIoCE (Research Institute for the Changing Global Environment) and Fudan Tyndall Cec Liaoning Provincial Center for Disease Control and Prevention, Shenyang, Chinad Taiyuan Municipal Center for Disease Control and Prevention, Taiyuan, Chinae Department of Epidemiology and Health Statistics, Institute of Basic Medical Sciences, Ch

j ourna l homepage: www.elity in three Chinese cities: The China Air

n Xu e, Guang Zeng f, Xiaohui Xu g,

niversity, Shanghai, China, Fudan University, Shanghai, China

Academy of Medical Sciences, Peking Union Medical College, Beijing, China

al Environment

v ie r .com/ locate /sc i totenvional coal combustion type to themixed coal combus-le emission type. Also, the characteristics of outdoor. air pollution level, and fate and transport of pollut-gical conditions and socio-demographic patterns innt from North America and Western Europe. To ourrior studies have been carried out to examine thects of ambient CO in China, or even Asian developing

The objective of this paper is to examine the short-term associa-tions between ambient CO and daily mortality in three Chinese cities Shanghai, Anshan and Taiyuan. This study is a component of theChina Air Pollution and Health Effects Study (CAPES) initiated bythe China Ministry of Environmental Protection.

2. Materials and methods

2.1. Data

The locations of Shanghai, Anshan and Taiyuan are described inFig. 1. Shanghai is the economic center and one of the largest citiesof China. Anshan is a heavily-polluted industrial city in northeasternChina. Taiyuan is the capital city of Shanxi province. Our study areaswere restricted to the urban areas of the three cities, due to inade-quate air pollution monitoring stations in the suburban areas.

The study periods were 2006 to 2008 for Shanghai, 2004 to 2006for Anshan, and 2004 to 2008 for Taiyuan. The sources of mortalitydata were Shanghai Municipal Center of Disease Control and Preven-tion; Liaoning Provincial Center of Disease Control and Prevention;and Taiyuan Municipal Center of Disease Control and Prevention.The causes of death were coded according to International Classica-tion of Diseases, 10 (ICD-10). The mortality data were classied intodeaths due to total non-accidental causes (ICD-10: A00-R99), cardio-vascular disease (ICD-10: I00-I99), and respiratory disease (ICD-10:J00-J98).

The sources of air pollutant concentrations were Shanghai Envi-

element oscillating microbalance (Thermo Environmental Instru-ments Inc., TEOM Series 1400a), ultraviolet uorescence (ThermoEnvironmental Instruments Inc., Model 43A), and chemilumines-cence (Thermo Environmental Instruments Inc., Model 42C) wereused for the measurement of CO, PM10, SO2, and NO2, respectively.For the calculation of 24-hour mean concentrations, at least 75% ofthe one-hour values must be available on that particular day. If astation had more than 25% of the values missing for the whole periodof analysis, the entire station was excluded from the analysis. In eachcity, the location of monitoring stations was mandated not to be inthe direct vicinity of trafc or of industrial sources, and not to beinuenced by local pollution sources and should also avoid buildings,or housing large emitters such as coal-, waste-, or oil-burning boilers,furnaces, and incinerators.

To allow adjustment for the effect of weather conditions onmortality, meteorological data (daily mean temperature and relativehumidity) were obtained from one monitoring station at each city.

2.2. Statistical analysis

The CAPES project follows the same protocol as the Public Healthand Air Pollution in Asia (PAPA) program of the Health EffectsInstitute (Wong et al., 2008, 2010). Specically, the protocolcomprises specication for selection of monitoring stations, qualityassurance or quality control for the data collection, health outcomesand air pollutants to be included in the analysis. The protocol alsoincluded the methods to standardize data management including

4924 R. Chen et al. / Science of the Total Environment 409 (2011) 49234928ronmental Monitoring Center (6 stations), Anshan EnvironmentalMonitoring Center (2 stations), and Taiyuan Environmental Monitor-ing Center (9 stations). At each city, the daily concentrations for eachpollutant were averaged from the available monitoring results ofmultiple stations. Automatic continuous monitoring system was setup at each city to measure daily air pollution levels. Air qualityindicators included CO, particulate matter with aerodynamicdiameter of 10 m or less (PM10), sulfur dioxide (SO2), and nitrogendioxide (NO2). 24-hour average concentrations for CO, SO2,PM10, and NO2 were calculated. The methods based on light absor-bance (Thermo Environmental Instruments Inc., Model 48C), taperedFig. 1. Locations of air monitoring station and weather mcompilation of daily data. The methods were individualized to suitthe local situation, including the specications for selection ofmonitoring stations and quality assurance and quality control proce-dures for data collection on health outcomes and air pollutants to beincluded in the analysis.

The daily death, air pollution and weather are linked by date andtherefore can be analyzed with a time-series design (Zeger et al.,2006). Because counts of daily mortality data in our study approxi-mately follow a Poisson distribution and the relations betweenmortality and explanatory variables are mostly nonlinear (Dominiciet al., 2004), we used overdispersed generalized linear Poissononitoring station in Shanghai, Taiyuan and Anshan.

models (quasi-likelihood) with natural spline (ns) smoothers toanalyze the mortality, air pollution, and covariates data. This methodhas accounted for the over-dispersion in the condence interval and pvalues.

To control long-term and seasonal trends of daily mortality andweather conditions, generalized linear modeling, with ns smoothers,was used to model daily mortality (Bell et al., 2004b; Burnett et al.,2004). The partial autocorrelation function (PACF) was used toguide the selection of degree of freedom (df) for time trend in thecore models. Specically, 46 df per year was used for time trend.When the absolute magnitude of the PACF plot was less than 0.1 forthe rst two lag days, the core models were regarded as adequate(Peng et al., 2006). If this criterion was not met, auto-regressionterms were used to reduce autocorrelation (Kan et al., 2008). Day of

20042006 for Anshan, and 20042008 for Taiyuan), the mean dailydeath numbers for all non-accidental causes, cardiovascular causesand respiratory causes were 125.3, 49.6 and 12.9, respectively, inShanghai; 27.6, 14.1, and 1.9 in Anshan; and 24.2, 8.9, and 1.9 in Tai-yuan (Table 1). Among all deaths, cardiorespiratory causes accountedfor 49.9% in Shanghai, 58.0% in Anshan, and 44.6% in Taiyuan.

Generally, the CO levels in three Chinese cities were similar withthose reported in developed countries (Table 1) (Bell et al., 2009;Samoli et al., 2007). The averaged daily concentrations of CO were1.3 mg/m3 for Shanghai (20062008), 1.1 mg/m3 for Anshan (2004

4925R. Chen et al. / Science of the Total Environment 409 (2011) 49234928the week (DOW) was included as dummy variable in the models.Residuals of the core models were examined to check whetherthere were discernable patterns and autocorrelation by means ofresidual plots and PACF plots. Test for normality of the residualswas also conducted.

CO concentrations were added into the core model to assess itsassociation with daily mortality in each city. Combined estimates ofCO were calculated using a xed- or random-effects model. Estimateswere weighted by the inverse of the sum of within- and between-study variance. Homogeneity tests were performed by means ofChi-square tests for the differences in sum of squares betweenindividual and weighted average of the estimates.

Because the assumption of the linearity between CO and the log ofmortality may not be justied, the smoothing function with 3 df wasused to graphically describe their relationships. We compared thelinear and spline models by computing the difference between thedeviances of the tted two models (Samoli et al., 2005). This differ-ence followed a chi-square distribution with degrees of freedombeing the difference in the degrees of freedom of the tted models.

Both single- and multi-pollutant models were tted to assess thestability of effect estimate of CO; up to two pollutants were includedper model. Single-day lag models were reported to underestimate thecumulative effect of air pollution on mortality (Bell et al., 2004b);therefore, we used 2-day moving average of current and previousday concentrations (lag01) of CO for our main analyses. Given thatit is not easy to determine the optimal values of df for time trend, sen-sitivity analyses were conducted to test the impact of alternative dfvalues on the estimated effect of CO.

All analyses were conducted in R 2.10.1 using the MGCV package(R Development Core Team, 2010). The results are presented as thepercent change in daily mortality per 1 mg/m3 increase of COconcentrations.

3. Results

Table 1 summarizes the air pollution and mortality data in thethree cities. In our research periods (20062008 for Shanghai,

Table 1Descriptive data on the study period, population, daily deaths, air pollutant concentra-tions and weather conditions (mean and S.D.) in three Chinese cities.

Shanghai Anshan Taiyuan

Study period (year) 20062008 20042006 20042008Study population (millions) 6.5 1.5 1.1No. of total deaths 125.3 (22.4) 27.6 (6.1) 24.2 (7.9)No. of cardiovascular deaths 49.6 (12.4) 14.1 (4.3) 8.9 (4.2)No. of respiratory deaths 12.9 (5.4) 1.9 (1.5) 1.9 (1.7)CO (mg/m3) 1.3 (0.5) 1.1 (0.8) 1.8 (1.0)PM10 (g/m3) 86.4 (53.1) 110.9 (60.2) 132.1 (65.4)SO2 (g/m3) 52.6 (29.9) 59.0 (74.3) 77.1 (8.0)NO2 (g/m3) 55.9 (21.2) 25.5 (16.3) 22.7 (8.7)Temperature (C) 17.8 (8.9) 11.3 (12.7) 11.2 (10.7)Humidity (%) 69.6 (11.7) 55.2 (16.0) 55.1 (18.1)2006), and 1.8 mg/m3 for Taiyuan (20042008).The correlation coefcients of CO with NO2 were higher than with

PM10 or SO2, probably because both CO and NO2 are primarily gener-ated by trafc sources (Table 2). In all three cities, CO was weakly cor-related with temperature and relative humidity.

In the single-pollutant models, we found signicant associationsbetween CO levels and daily mortality from total non-accidentalcauses and from cardiovascular diseases in each of the three cities(Table 3). An increase of 1 mg/m3 of 2-day moving average concen-trations of CO corresponds to 2.41% [95% condence interval (CI):0.64, 4.19], 1.97% (95%CI: 0.13, 3.81), 3.90% (95%CI: 2.54, 5.26) in-crease of total non-accidental mortality; and 3.85% (95%CI: 1.29,6.40), 2.83% (95%CI: 0.27, 5.38), and 5.38% (95%CI: 3.28, 7.49)increase of cardiovascular mortality in Shanghai, Anshan and Taiyuan,respectively. We did not observe signicant associations of CO withrespiratory mortality in all three cities. In the three-city combinedanalysis, there was no signicant heterogeneity for the associationsof CO with either total or cardiovascular mortality. An increase of1 mg/m3 of 2-daymoving average concentrations of CO correspondedto 2.89% (95%CI: 1.68, 4.11) and 4.17% (95%CI: 2.66, 5.68) increase oftotal and cardiovascular mortality, respectively (Table 3).

In the three Chinese cities, the associations of CO with total andcardiovascular mortality were only minimally altered by addingPM10 and SO2 into the models (Table 3). However, adjustment forNO2 decreased the associations and rendered some of them statisti-cally insignicant in the city-specic analysis. In the three-city pooledestimates, the effects of CO remained signicant after adjustment forany co-pollutants.

There were almost linear relationships between CO and total/cardiovascular mortality in all three cities (Fig. 2). We did not ob-serve any obvious threshold concentration below which CO has noeffect on mortality outcomes. The differences in the deviance be-tween the linear and spline models were statistically insignicantfor all models we examined.

CO showed similar lag patterns for their effects on total and car-diovascular mortality (Fig. 3). For single-day lags, the risks decreasedfrom lag-days 0 to 4; multi-day exposures (lag01 and lag04) usuallyhave larger effects than single-day exposure (lag0 to lag4). The effectsof CO on total and cardiovascular mortality were statistically signi-cant for most lagged days we examined.

Within the range of 312, the change of df/yr for time trend didnot substantially affect the estimated effects of CO (data notshown), suggesting that our ndings are relatively robust in thisaspect. We also compared the effects of CO with alternative df valuesfor weather conditions. Within the range of 312, the change of df for

Table 2Spearman coefcients of CO with co-pollutants and weather variables in three Chinesecities*.

Shanghai Anshan Taiyuan

PM10 0.72 0.59 0.52SO2 0.71 0.50 0.58NO2 0.80 0.73 0.48Temperature 0.40 0.31 0.38Relative humidity 0.09 0.09 0.12

pb0.01 for all coefcients.

4926 R. Chen et al. / Science of the Total Environment 409 (2011) 49234928temperature and humidity resulted in almost identical estimated ef-fects of CO on mortality outcomes.

4. Discussion

This combined analysis summarizes the results from three Chinesecities concerning the short-term effects of CO on daily mortality. Sig-nicant associations of CO with total and cardiovascular mortalitywere found, and these ndings were generally insensitive to alterna-tive model specications. In the combined analysis, we found signi-cant effects of CO on total and cardiovascular mortality afteradjustment for PM10, SO2 or NO2. To our knowledge, this is the rstmulti-city analysis in China or even in Asia to report the acute healtheffect of CO. Our ndings are consistent with previous ndings of am-bient CO in the US (Bell et al., 2009) and Europe (Samoli et al., 2007).

In our multi-city combined analysis, 1 mg/m3 increase of CO wasassociated with 2.89% (95%CI: 1.68, 4.11) increase of total non-accidental mortality. The magnitudes of our estimates for CO are gen-erally comparable with previous single-city, multi-city, and meta an-alyses worldwide. For instance, in Phoenix (US), Mar et al. (2000)found that 1 mg/m3 increase of CO was associated with 2.79% (95%CI:

Table 3Percent increase (mean and 95% CIs) of daily mortality with a 1 mg/m3 increase of COconcentration under single and multi-pollutant models effect estimates of individualcities and combined effectsa.

City Model Total mortality Cardiovascularmortality

Shanghai Single-pollutant 2.41 (0.64, 4.19) 3.85 (1.29, 6.40)Adjusted for PM10 2.31 (0.22, 4.40) 3.00 (0.30, 6.30)Adjusted for SO2 2.27 (0.13, 4.41) 1.91 (1.45, 5.27)Adjusted for NO2 0.24 (3.03, 2.54) 2.24 (1.80, 6.28)

Anshan Single-pollutant 1.97 (0.13, 3.81) 2.83 (0.27, 5.38)Adjusted for PM10 2.49 (0.11, 4.86) 1.92 (1.38, 5.22)Adjusted for SO2 1.60 (0.45, 3.65) 2.48 (0.38, 5.33)Adjusted for NO2 1.85 (0.72, 4.42) 2.77 (0.81, 6.34)

Taiyuan Single-pollutant 3.90 (2.54, 5.26) 5.38 (3.28, 7.49)Adjusted for PM10 3.47 (1.85, 5.08) 6.36 (3.88, 8.85)Adjusted for SO2 2.82 (1.31, 4.32) 4.94 (2.63, 7.26)Adjusted for NO2 1.95 (0.43, 3.47) 3.77 (1.41, 6.13)

Pooled estimates Single-pollutant 2.89 (1.68, 4.11) 4.17 (2.66, 5.68)Adjusted for PM10 2.91 (1.79, 4.04) 3.94 (1.13, 6.76)Adjusted for SO2 2.36 (1.31, 3.42) 3.38 (1.46, 5.31)Adjusted for NO2 1.53 (0.35, 2.72) 3.23 (1.46, 5.00)

a Two-day moving averaging (lag 01) concentrations were used in Table 3.0.84, 2.64) increase of total non-accidental mortality. In 11 Canadi-an cities, Burnett et al. (2004) estimated that 1 mg/m3 increase ofCO was associated with 2.00% (95%CI: 1.39, 2.61) increase of totalmortality. Also, in a meta analysis of 18 time-series studies of COand daily mortality, Stieb et al. (2003) estimated that the excess all-causes mortality change (single-pollutant models) associated with1 mg/m3 of CO was 1.27% (95%CI 0.87%1.67%). A large-scale multi-city time-series analyses in 19 European cities estimated that 1 mg/m3 increase of CO corresponded to 1.20% (95%CI: 0.631.77%) in-crease in total deaths (Samoli et al., 2007). The heterogeneity of var-ious ndings may reect differences in the characteristics of the studysites. Numerous factors, including indoor air pollution, weather pat-terns, sensitivity of local residents to pollution (e.g. socioeconomicstatus, age, smoking rate), and air pollution levels, may affect themagnitudes of exposureresponse relationships. Although onlythree Chinese cities were studied, our results may begin to allay con-cerns regarding the generalizability of the results of the substantial,but largely Western, literature on the effects of short-term exposureto CO. The results, which are broadly consistent with previous re-search, suggest that neither genetic factors nor longer-term exposureto highly polluted PM/SO2 substantially modies the acute effect ofCO on daily mortality in China. We assume two reasons for this con-sistency: 1. the CO levels are comparable in China and Westerncountries; 2. unlike PM with various components and toxicity, COin different countries shares the same biological mechanisms to affecthuman health.

The shape of exposureresponse relationships is crucial for publichealth assessment and there has been growing demand for providingthe relevant curves. Consistent with previous studies in the US andEurope (Bell et al., 2009; Samoli et al., 2007), the exposureresponserelationship between CO and mortality in three Chinese cities gener-ally supported a linear association without threshold. It should benoted that we found signicant effects of CO even below the levelsof air quality standard in China (4 mg/m3 for 24-h averageconcentration of CO) (Fig. 2). Therefore, current air quality standardof CO might not be sufcient to protect the public health in China.Further control of CO pollution is likely to result in health benets.A reduction in morbidity and mortality after the implementation ofan intervention program will add evidence to the hypothesis of acausal link between CO pollution and ill health.

CO is produced by incomplete combustion of hydrocarbons. Itsmain source in urban area is vehicle exhaust emissions. Accompaniedwith the rapid socioeconomic development, the number of motor ve-hicles increases drastically in urban China, and exhaust emissions arebecoming one of the major contributors to urban air pollution. Evi-dence from epidemiological studies on the health effects of trafc-related air pollutants [nitrogen oxides (NOx), CO, PM, O3 and its pre-cursors] has accumulated in China (Kan et al., 2011). However, mostof these studies focus on occupationally exposed populations (e.g.trafc policemen, drivers or conductors) and vulnerable population(e.g. children) (Chen et al., 2009; Wang et al., 2009; Zhou et al.,2001). No Chinese studies examined mortality in relation totransport-related air pollution, which has been well investigated inWestern countries. Our ndings of CO suggest that the role of expo-sure to trafc-related air pollution should be investigated further inChina.

The association between ambient CO and daily mortality is biolog-ically plausible. CO is one of the few air pollutants that we know itsbiologically toxic form, carboxyhemoglobin (COHb) (McGrath,2000). COHb reduces the oxygen-carrying capacity of the blood andimpairs the release of oxygen from hemoglobin to extravascular tis-sues. Therefore, the toxic effects of CO are most evident in organsand tissues with high oxygen consumption such as the brain, theheart, exercising skeletal muscle, and the developing fetus. The effectsof exposure to low CO concentrations, such as the levels found in am-bient air, are far more subtle and considerably less threatening thanthose occurring in CO poisoning (McGrath, 2000). The underlyingmechanisms for effects of low-level exposure are unclear, but likelyinclude reduced exercise capacity and exacerbation of cardiovascularsymptoms in individuals with impaired cardiovascular systems. COhas also been associated with alteration of the cardiac autonomic reg-ulation in population-based studies (Liao et al., 2004) and in panelstudies (Timonen et al., 2006).

Our analysis has strengths and limitations. These three Chinesecities offer advantages for the study of the CO-mortality relationshipin that they are generally very densely populated. As in most previoustime-series studies, we simply averaged the monitoring results acrossvarious stations as the proxy for population exposure level to CO. Thesimple averaging methodmay raise a number of issues given that pol-lutant measurements can differ frommonitoring location to monitor-ing location and that ambient monitoring results differ from personalexposure level to air pollutants (Sarnat et al., 2005). Numerous fac-tors, such as air conditioning and ventilation rate between indoorand outdoor air, may affect the monitoring results from xed stationsas surrogates of personal exposure to air pollutants (Janssen et al.,2002). Because we were unable to measure the true population expo-sures in these three cities, we could not determine the direction of thebias and its impact on our conclusions. Wewere not able to collect the

1-h maximum CO, though previous study has found similar effect

4927R. Chen et al. / Science of the Total Environment 409 (2011) 49234928Shanghai: Total mortality

logR

R0.

00.

050.

10estimates of CO using 1-h maximum and 24-h averaged concentrations(Bell et al., 2009). Also, wewere not able to obtain the age-specicmor-tality data of the three Chinese cities, which limited our ability to ex-plore the modifying effect of age on the health impact of CO.Moreover, it is still uncertain whether temperature is a confounder oreffect moderator (i.e., a synergistic effect) of the CO-mortality associa-tion. Future research should study the interaction between CO andtemperature.

In summary, we found signicant adverse effects of ambient COlevels on mortality from all natural and cardiovascular causes, evenwell below the health-based air quality standards in China. The effectestimates were generally consistent with other CO studies in the

CO (microgram per cubic meter)

CO (microgram per cubic meter)

CO (microgram per cubic meter)

logR

Rlo

gRR

1000-0.

05

Anshan: Total mortality

-0.

10

Taiyuan: Total mortality

1000-0.

1

1500 2000 2500 3000

10005000 1500 2000 2500 3000

-0.

050.

00.

050.

100.

00.

10.

20.

3

2000 3000 4000 5000

Fig. 2. Smoothing plots of CO against total and cardiovascular mortality risk (df=3) in eacindicate the estimated mean percentage of change in daily mortality, and the dotted lines rlogR

R

Cardiovascular mortality

0.0

0.05

0.10

0.15remaining world, and relatively robust to adjustment for co-pollutants, various lagged exposure, and varying smoothness oftime. To our knowledge, this is the rst multicity CO study in China.The ndings might have implications for standard revision of ambientCO and trafc policy formations.

Acknowledgment

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 and201209008), National Natural Science Foundation of China (30800892),

CO (microgram per cubic meter)

CO (microgram per cubic meter)

CO (microgram per cubic meter)

logR

Rlo

gRR

-0.

05

Cardiovascular mortality

-0.

10

Cardiovascular mortality

-0.

1

10005000 1500 2000 2500 3000

1000 1500 2000 2500 3000

1000 2000 3000 4000 5000

0.0

0.1

0.2

0.3

0.4

0.5

-0.

050.

00.

050.

100.

15

h city. X-axis is the 2-day average (lag01) CO concentrations (g/m3). The solid linesepresent twice the standard error.

Kan H, Wong CM, Vichit-Vadakan N, Qian Z. Short-term association between sulfur di-oxide and daily mortality: the Public Health and Air Pollution in Asia (PAPA) study.Environ Res 2010;110:25864.

Kan H, Chen R, Tong S. Ambient air pollution, climate change, and population health inChina. Environ Int 2011. doi:10.1016/j.envint.2011.03.003.

Liao D, Duan Y, Whitsel EA, Zheng ZJ, Heiss G, Chinchilli VM, et al. Association of higherlevels of ambient criteria pollutants with impaired cardiac autonomic control: apopulation-based study. Am J Epidemiol 2004;159:76877.

Mar TF, Norris GA, Koenig JQ, Larson TV. Associations between air pollution and mor-tality in Phoenix, 19951997. Environ Health Perspect 2000;108:34753.

McGrath JJ. Biological plausibility for carbon monoxide as a copollutant in PM epidemi-ologic studies. Inhal Toxicol 2000;12(Suppl 4):91-107.

Peng RD, Dominici F, Louis TA. Model choice in time series studies of air pollution andmortality. J R Stat Soc Ser a-Stat Soc 2006;169:17998.

Pope III CA, Dockery DW. Health effects of ne particulate air pollution: lines that con-nect. J Air Waste Manag Assoc 2006;56:70942.

R Development Core Team. R: a language and environment for statistical computing,version 2.10.1. Vienna: R Foundation for Statistical Computing; 2010.

Riojas-Rodriguez H, Escamilla-Cejudo JA, Gonzalez-Hermosillo JA, Tellez-Rojo MM,Vallejo M, Santos-Burgoa C, et al. Personal PM2.5 and CO exposures and heartrate variability in subjects with known ischemic heart disease in Mexico City.J Expo Sci Environ Epidemiol 2006;16:1317.

Samoli E, Analitis A, Touloumi G, Schwartz J, Anderson HR, Sunyer J, et al. Estimatingthe exposureresponse relationships between particulate matter and mortalitywithin the APHEA multicity project. Environ Health Perspect 2005;113:8895.

Samoli E, Aga E, Touloumi G, Nisiotis K, Forsberg B, Lefranc A, et al. Short-term effects ofnitrogen dioxide on mortality: an analysis within the APHEA project. Eur Respir J2006;27:112938.

Fig. 3. Percent increase of total and cardiovascular mortality associated with 1 mg/m3

increase of CO, using different lag structures (lag0lag4, lag01, and lag 04). X-axis islag structures. Y-axis is percent increase of mortality.

4928 R. Chen et al. / Science of the Total Environment 409 (2011) 49234928and Program for New Century Excellent Talents in University (NCET-09-0314).

The authors declare they have no competing nancial interests.

References

Allred EN, Bleecker ER, Chaitman BR, Dahms TE, Gottlieb SO, Hackney JD, et al. Short-term effects of carbon monoxide exposure on the exercise performance of subjectswith coronary artery disease. N Engl J Med 1989;321:142632.

Bell ML, McDermott A, Zeger SL, Samet JM, Dominici F. Ozone and short-term mortalityin 95 US urban communities, 19872000. Jama 2004a;292:23728.

Bell ML, Samet JM, Dominici F. Time-series studies of particulate matter. Annu Rev Pub-lic Health 2004b;25:24780.

Bell ML, Peng RD, Dominici F, Samet JM. Emergency hospital admissions for cardiovas-cular diseases and ambient levels of carbon monoxide: results for 126 UnitedStates urban counties, 19992005. Circulation 2009;120:94955.

Burnett RT, Stieb D, Brook JR, Cakmak S, Dales R, Raizenne M, et al. Associations be-tween short-term changes in nitrogen dioxide and mortality in Canadian cities.Arch Environ Health 2004;59:22836.

Chen CH, Kan HD, Huang C, Li L, Zhang YH, Chen RJ, et al. Impact of ambient air pollu-tion on public health under various trafc policies in Shanghai, China. Biomed En-viron Sci 2009;22:2105.

Dales R. Ambient carbonmonoxide may inuence heart rate variability in subjects withcoronary artery disease. J Occup Environ Med 2004;46:121721.

Dominici F, McDermott A, Hastie TJ. Improved semi-parametric time series models ofair pollution and mortality. J Am Stat Assoc 2004;468:93848.

Janssen NA, Schwartz J, Zanobetti A, Suh HH. Air conditioning and source-specic par-ticles as modiers of the effect of PM(10) on hospital admissions for heart and lungdisease. Environ Health Perspect 2002;110:439.

Kan H, London SJ, Chen G, Zhang Y, Song G, Zhao N, et al. Season, sex, age, and educa-tion as modiers of the effects of outdoor air pollution on daily mortality in Shang-hai, China: the Public Health and Air Pollution in Asia (PAPA) Study. EnvironHealth Perspect 2008;116:11838.Samoli E, Touloumi G, Schwartz J, Anderson HR, Schindler C, Forsberg B, et al. Short-term effects of carbon monoxide on mortality: an analysis within the APHEA pro-ject. Environ Health Perspect 2007;115:157883.

Sarnat JA, Brown KW, Schwartz J, Coull BA, Koutrakis P. Ambient gas concentrationsand personal particulate matter exposures: implications for studying the health ef-fects of particles. Epidemiology 2005;16:38595.

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;53:25861.

Timonen KL, Vanninen E, de Hartog J, Ibald-Mulli A, Brunekreef B, Gold DR, et al. Effectsof ultrane and ne particulate and gaseous air pollution on cardiac autonomiccontrol in subjects with coronary artery disease: the ULTRA study. J Expo Sci Envi-ron Epidemiol 2006;16:33241.

Wang S, Zhang J, Zeng X, Zeng Y, Chen S. Association of trafc-related air pollution withchildren's neurobehavioral functions in Quanzhou, China. Environ Health Perspect2009;117:16128.

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. Envi-ron Health Perspect 2008;116:1195202.

Wong CM, Vichit-Vadakan N, Vajanapoom N, Ostro B, Thach TQ, Chau PY, et al. Part 5.Public health and air pollution in Asia (PAPA): a combined analysis of four studiesof air pollution and mortality. Res Rep Health Eff Inst 2010:377418.

Yang W, Jennison BL, Omaye ST. Cardiovascular disease hospitalization and ambientlevels of carbon monoxide. J Toxicol Environ Health A 1998;55:18596.

Zeger SL, Irizarry R, Peng RD. On time series analysis of public health and biomedicaldata. Annu Rev Public Health 2006;27:5779.

ZhouW, Yuan D, Ye S, Qi P, Fu C, Christiani DC. Health effects of occupational exposuresto vehicle emissions in Shanghai. Int J Occup Environ Health 2001;7:2330.

Ambient carbon monoxide and daily mortality in three Chinese cities: The China Air Pollution and Health Effects Study (CAPES)1. Introduction2. Materials and methods2.1. Data2.2. Statistical analysis

3. Results4. DiscussionAcknowledgmentReferences