Traffic Air Pollution and Risk of Death from Breast Cancer in Taiwan : Fine Particulate Matter ( PM 2 . 5 ) as a Proxy Marker

The relationship between breast cancer mortality and air pollution was examined using an ecological design. The study areas consisted of 61 municipalities in Taiwan. Air quality data for recorded concentrations of fine particulate matter (PM2.5) from study municipalities for 2006–2009 were obtained as a marker of traffic emissions and were used as a proxy for polycyclic aromatic hydrocarbons (PAHs) exposure. Age-standardized mortality rates for breast cancer mortality were calculated for the study municipalities for the years 1999–2008. A weighted multiple regression model was used to calculate the adjusted risk ratio in relation to PM2.5 levels. Results showed that individuals who resided in municipalities with the highest PM2.5 were at an increased risk of death from breast cancer. This study is the first to suggest that exposure to high levels of PM2.5, a proxy measure of PAHs, may be associated with an increased risk of death from breast cancer. The findings of this study warrant further investigation into the role of air pollutants in the risk of breast cancer.


INTRODUCTION
The incidence of breast cancer varied widely internationally, with the highest rates observed in North America and Europe and the lowest rates in Asia (Schottenfeld and Frauimeni, 1996).In Taiwan, the incidence rates for breast cancer was 51.95 per 100,000 in 2008 (DOH, 2008).Breast cancer is the fourth leading cause of female cancer mortality (DOH/Taiwan, 2011).The ageadjusted mortality rate for breast cancer was 11 per 100,000 in 2010.There is substantial geographic variation in female breast cancer mortality within the country (NHRI/ Taiwan, 2004), which suggests an environmental risk factor may be involved.
The known risk factors for breast cancer are primarily related to the reproductive life of women and inheritance (e.g., BRCA1 and BRCA2 gene mutations) (Coyle, 2004).
Only about one-third of new cases of breast cancer are attributable to known risk factors, and much of the etiology remains unexplained (Coyle, 2004;Crouse et al., 2010).Therefore it has been hypothesized that environmental exposures may also contribute to breast cancer risk (Laden and Hunter, 1998;Coyle, 2004).
Ambient outdoor air pollution has been implicated as a cause of various health problems including cancer (Dominici et al., 2005;Boffetta, 2006;Curtis et al., 2006;Samet and Krewski, 2007).Air pollution is a complex mixture of different gaseous and particulate components; thus it is difficult to define an exposure measure of relevance when the biological mechanisms are largely unknown (Boffetta and Nyberg, 2003).Air pollution from motor vehicle exhaust has been one of the most studied environmental factors (Craig et al., 2008;Chen and Li, 2010;Chuang et al., 2010a;Shen et al., 2010;Tsai et al., 2011).Exhaust from traffic is a complex mixture of many chemical compounds, including benzene, polycyclic aromatic hydrocarbons (PAHs) and benzo [a]pyrene (B[a]P) (Chuang et al., 2010b;Wu et al., 2010;Huang et al., 2011a, b;Kao et al., 2011).The International Agency for Research on Cancer (IARC) classified the emission of diesel exhaust engine compounds as probably carcinogenic (Group 2A) and gas engine exhaust compounds as possibly (Group 2B) carcinogenic to humans (IARC, 1989).
According to occupational epidemiologic studies, excess risks of breast cancer were reported to be associated with occupational exposure to PAHs among pre-and postmenopausal women (Petralia et al., 1999;Gammon et al., 2002;Labreche et al., 2010).Given that these same pollutants are present in the motor vehicle exhaust, it is plausible that traffic-related exposures may contribute to the risk of breast cancer (Crouse et al., 2010).Several studies reported an increased risk for breast cancer in association with surrogate measures of exposure to vehicle exhaust, including emissions associated with heavy traffic density (Lewis-Michl et al., 1996), total suspended particulates (TSP) (Bonner et al., 2005), estimated exposures to B[a]P derived from a traffic emissions model (Nie et al., 2007), and estimated nitrogen dioxide (NO 2 ) derived from a landuse regression model (Crouse et al., 2010).
In Taiwan, the major source of PAHs is traffic exhaust (Chang et al., 2003).PAHs can be present in particulate or gaseous phase.Gaseous PAHs usually contain more fractions of less carcinogenicity and lower molecular weights, while particulate PAHs contain more fractions of higher carcinogenicity and higher molecular weights (Chang et al., 2006).PAHs with high molecular weights are often more carcinogenic than those with low molecular weights (Ravindra et al., 2001).Therefore particulate phase PAHs are considered to be the most hazardous substances to human health.Most of the PAHs were in the particulate phase.Ninety to 95% of particulate phase PAHs are physically associated with particular matter (PM) < 3.3 μm (Ravinndra et al., 2001;Bonner et al., 2005).These small particles are thought to have biological relevance because they can be inhaled and deposited in the lower respiratory tract (Bonner et al., 2005).
The study reported here was designed to explore further whether the risk of death from breast cancer is associated with exposure to PAHs, using fine particulate matter (PM 2.5 ), a marker of traffic-related air pollution, as a proxy for PAHs exposure in Taiwan.

Study Areas
Taiwan is divided into 359 administrative districts, which are referred to in this study as municipalities.In 1994, a network of 66 air pollution monitoring stations was established in 66 municipalities by the Environmental Protection Administration (EPA) in Taiwan.Among these, two stations were excluded because they were situated in the National Park areas and served as background monitoring sites.In addition, three stations were removed due to the reconstruction of the buildings where there were built during 2000-2007, and were excluded from analysis.This elimination left 61 monitoring stations situated in 61 municipalities for study (Fig. 1).These 61 municipalities provide an opportunity to investigate the impact of PM 2.5 air pollution on the risk of death from breast cancer.

PM 2.5 Exposure Estimates
There is an extensive air pollution monitoring network in Taiwan, managed by Taiwan EPA as mentioned above, which routinely monitors five "criteria" pollutants (PM 10 , O 3 , CO, SO 2 , NO 2 ).However, PM 2.5 was not regularly monitored.The monitoring stations were fully automated and provided daily readings of polutant levels.
PM 2.5 concentrations in Taiwan have been measured continuously since 2006.The monitoring stations provided data for PM 2.5 in mean daily mass concentration (μg/m 3 ).The availability of the monitoring network for PM 2.5 provides an opportunity to investigate the impact of fine particle air pollution on female breast cancer mortality.We obtained PM 2.5 data for study municipalities from the EPA's air quality monitoring stations for 2006-2009.The municipality of residence for all cases and controls was identified from death certificates.The municipality of residence formed the only basis for estimating their fine particulate air pollution exposure.

Mortality
Information concerning both the number of death and Taiwan's midyear population by sex, age, calendar year and municipality from 1999-2008 was obtained from the Department of Health which is in charge of the death registration system in Taiwan.In this study, female breast cancer deaths were those that were defined by the International Classification of Disease, Injury and Causes of Death (ICD-9 code 174).As the age distribution was not comparable among the 61 municipalities, the age standardized rates (ASR) were computed by the direct method, using the world population in 2000 as the standard population (Ahmad et al., 2000).

Fertility
Reproduction, particularly early full-term pregnancy, seems to exert a protective effect against breast cancer (Kelsey, 1979;Wigle, 1977;Rosenwaike, 1980;Remennick, 1989).One indicator of reproduction was available at the municipality level in Taiwan where fertility rates were recorded for different time periods from 1971 till now.In this study the fertility rates from 1971 for females aged 20-24 years were used to study the possible beneficial effect that early pregnancy might have in the prevention of breast cancer mortality (MOI/Taiwan, 1973).Fertility rates for earlier than 1971 would have been preferable; however, data were not available.An assumption was thus made that a municipality with high average fertility rates in 1971 would also have higher than average rates prior to 1971.

Socioeconomic Factors
It was found that mortality attributed to cancer was associated with urbanization gradients (Yang and Hsieh, 1998;Reynolds et al., 2004a;Hall et al., 2005).In this study, an urbanization index (Tzeng and Wu, 1986) was used to adjust for possible confounding resulting from different urbanizational levels among the municipalities.The urbanization index used in this study serves as a proxy for a large number of explanatory variables such as population density, age composition, mobility, economic activity and family income, educational level, environmental factors, and health service-related facilities which are related to the etiology of mortality.Each municipality in Taiwan was given a degree of urbanization category, 1-8.A municipality with the highest urbanization score, such as the Taipei metropolitan area, was classified in category 1, while mountainous areas with the lowest score were assigned to category 8.This index was used previously (Chiu et al., 2006;Liu et al., 2008).For the analyses, the urban-rural classification was further subdivided into two categories: I, urban areas (categories 1-4); and II, rural areas (categories 5-8).

Statistics
In the analysis, the 61 municipalities were divided into tertiles according to the levels of PM 2.5 and the fertility rates.The risk ratios were calculated in relation to the group with the lowest exposure level.A weighted multiple regression model was used to calculate the adjusted risk ratio.In addition to the levels of PM 2.5 , both fertility rates and urbanization index were included in the multiple regression analysis to adjust for the possible influence on the increased risk of death from breast cancer and to obtain a better estimate of the impact of PM 2.5 on the risk of death from breast cancer.Since Taiwan municipalities vary greatly in size, the regression model used took into account the size of the municipalities.For each municipality the weight used was the square root of person-years (Draper and Smith, 1998).A Wald chi-square test for linear trend was also performed by entering PM 2.5 exposures as a 3level ordinal variable (with the values 0-2) (Rothman and Greenland, 1998).Values of p < 0.05 were considered statistically significant.

RESULTS AND DISCUSSION
Average annual age-standardized breast cancer mortality rates per 100,000 (ASR) for 1999-2008 by tertiles of PM 2.5 levels and fertility rates are listed in Table 1.
The crude risk ratios for death from breast cancer were significantly higher for the group with the highest levels of PM 2.5 .Further, municipalities with high fertility rates have a statistically significant lower risk of death from breast cancer (Table 2).
Results of the weighted multiple regression analyses are shown in Table 2. Adjustments for possible confounders (the urbanization level and fertility rate) did not markedly alter the risk ratio.The adjusted RR (95% CI) were 1.12 (0.96-1.32) for the group with PM 2.5 levels between 30.48 μg/m 3 and 39.41 μg/m 3 and 1.19 (1.03-1.38)for the group with PM 2.5 levels between 39.48 μg/m 3 and 51.10 μg/m 3 , compared to the group with the lowest PM 2.5 levels.Trend analyses showed statistically significant trend in risk of death attributed to breast cancer with increasing PM 2.5 levels.When we included PM 2.5 levels as a continuous variable, similar trend in results were observed as those found when PM 2.5 levels were examined as a categorical variable (a mortality rate of 0.065/10 5 increase per 1 μg/m 3 of PM 2.5 increase).In contrast, there was a statistically significant decreasing trend in the risk of death from breast cancer with increasing fertility rates.
The results of the present study showed that individuals who resided in municipalities with the highest PM 2.5 levels, a proxy measure of PAHs, were at a significantly increased risk of death from breast cancer.This finding is essentially consistent with previous epidemiologic studies which examined the hypothesis that breast cancer may be associated with exposure to traffic air pollution (Bonner et al., 2005;Nie et al., 2007;Crouse et al., 2010) and industrial pollution (Pan et al., 2011).
There have been a number of epidemiological studies that assessed the increased risk of death from breast cancer development from living in an urban rather than a rural area.In general, breast cancer mortality had been considerably higher in urban populations than non-urban ones (Blot et al., 1977;Goldsmith, 1980;Greenberg, 1983;Muir et al., 1987;Kelesy and Berkowitz, 1988;Yang and Hsieh, 1998).The only "urban factor" consistently mentioned in the literature is air pollution, suggesting that residing in an urban area is a reliable surrogate for increased air pollution exposure (Greenberg, 1983).
The specific exposure chemicals responsible for the elevation of risk of death from breast cancer have not been identified with certainty, but a possible candidate may be motor exhaust emissions because traffic emissions are the major source of air pollution in urban areas.An effect of vehicle exhaust emissions on the risk of death from breast cancer is plausible.Airborne particles emitted from petroleum or diesel engines contain numerous PAHs and BaP.The mutagenic and carcinogenic effects of PAHs and BaP are well documented in experimental studies (IARC, 1989;Musafia-Jeknic et al., 2005;Krewski and Rainham, 2007) and have been shown to cause mammary cancers in female rats (Maltoni et al., 1997).In addition, PAHs may also have estrogenic and antiestrogenic properties that could affect breast cancer risk (Santodonato, 1997;Meek, 1998;Brody and Rudel, 2003).The postulation that a mechanism by which vehicle exhaust emissions might increase the risk of breast cancer occurrence could be that airborne particles containing PAHs, which are lipophilic and may therefore reach elevated concentrations in breast tissue and promote carcinogenesis in the cells of the breast (Morris and Seifter, 1992).
Traffic counts and proximity to roads have commonly served as surrogates for exposure to traffic-related potential carcinogens (Reynolds et al., 2004b).Concentrations of these compounds are higher within 500 to 1000 feet of busy roads and freeways based on measured traffic-related air pollutant levels (Dubowsky et al., 1999;Raaschou-Nielsen et al., 2000, 2001).Lewis-Michl et al. (1996) conducted a study in Nassau and Suffolk Counties, New York State, suggested a possible increased breast cancer risk among women living close to areas characterized by relatively heavy traffic compared with other areas (OR = 1.29, 95% CI = 0.77-2.15).To our knowledge, it is not known which exposure assessment approach best reflects chronic personal exposure to traffic-related air pollution (Raaschou-Nielsen and Reynolds, 2006;Bedeschi et al., 2007).
Benzene, total suspended particles (TSP) and NO 2 are commonly used markers for traffic-related air pollution (Muzykl et al., 1998;Van Wijnen and Van der Zee, 1998;Raaschou-Nielsen et al., 2001;Bedeschi et al., 2007).Measurements of the spatial distribution of benzene in the atmosphere showed that the highest outdoor concentrations within urban areas tended to occur adjacent to main roads (Leung and Harrison, 1999).However, benzene levels were not recorded in fixed, outdoor monitoring stations in Taiwan.Gunier et al. (2006) has used geographic information systembased methods to estimate PAH exposure.Crouse et al. (2010) developed a land-use regression model to predict concentrations of NO 2 across Montreal and used NO 2 as a marker for traffic-related air pollution.They found evidence of an association between the incidence of postmenopausal breast cancer and exposure to ambient concentrations of NO 2 .For each increase of 5 ppb NO 2 , the adjusted odds ratio was 1.31 (95% CI = 1.00-1.71).However, it should be noted that NO 2 is only partly related to local sources (mainly traffic) as NO is quickly converted to NO 2 in reaction with ambient ozone.Only in areas with significant use of diesel vehicles reduce the problem of using NO 2 as a proxy for traffic pollution (Palmgren et al., 1996).In the Western New York exposures and breast cancer (WEB) study, Bonner et al. (2005) found that TSP exposures, a surrogate for PAHs exposure, were associated with increased postmenopausal breast cancer risk.In a second analysis from that study, Nie et al. (2007) used a traffic emission model to estimate BaP exposure, a surrogate for PAHs, found evidence of increased breast cancer risk among women exposed to high traffic emissions exposure.To our knowledge, this study is the first to suggest that exposure to high levels of PM 2.5 , a proxy measure of PAHs, may be associated with an increased risk of death from breast cancer.
It is mandatory to register any birth, death, marriage/ divorce, and migration in the household registration offices in Taiwan.Demographic and vital statistics data derived from the household registration system are considered to be reliable and complete.Nonetheless, data on the accuracy of breast cancer diagnosis are not available in Taiwan and misclassification is possible.However, this misclassification is likely to be nondifferential (i.e., unlikely to be related PM 2.5 ) and therefore would tend to underestimate rather than overestimate the true association.
Since the measure of effect in this study is mortality rather than incidence, migration during the interval between cancer diagnosis and death also needs to be considered.During this period, cancer diagnosis may influence a decision to migrate and possibly introduce bias.Data are not available for the differences in survival rates of breast cancer patients between high and low PM 2.5 areas.If there is a trend toward migration to more urban or high PM 2.5 areas because of proximity to medical care for example, a spurious association between PM 2.5 and cancer death would result.Our study population is stable in terms of mobility compared with populations in most industrialized countries, probably of culture factors (Yu et al., 2006).Furthermore, urbanization level was included as a control variable in the analysis.Since it is conceivable that municipalities with similar urbanization levels may have similar migration rates, this probably minimized the migration problem in our study.
Previous studies reported associations between increased risk of breast cancer and early-life exposure (at first birth or menarche) to markers of air pollution (Bonner et al., 2005;Nie et al., 2007), but no associations with exposures 10 to 20 years before diagnosis.The lifetime residential history of each study subject is not available.The PM 2.5 levels are also not available before 2006.We could not examine associations with early-life exposures.Our study may thus have been limited by potential exposure misclassification.While these sources of misclassification are important, such misclassification of exposure is most likely to be non-differential (i.e., unlikely to be associated with breast cancer), which would reduce the magnitude of association rather than introduce a positive bias in the association.
Our study employed an ecologic design using group level exposure data.It was presumed that individuals residing in the municipalities of higher levels of PM 2.5 experienced a greater exposure to PAH from motor vehicles.Nonetheless, PAHs concentrations may differ substantially within a municipality and therefore group exposure levels may not necessarily correspond to individual exposure levels (Reynolds et al., 2003).Further, potential exposure misclassification may also have resulted from differing individual time-activity and personal exposure from indoor sources (such as environmental tobacco smoke) (Elliott et al., 2000).However, this misclassification is likely to be non-differential.Our results thus probably underestimate the true estimates of the relative risk of death from breast cancer associated with exposure to PM 2.5 .
Body mass index, age at menarche, age at full-term pregnancy, parity, menopausal status and age at menopausal status, duration of lifetime lactation, use of lactation suppression hormones, and use of oral contraceptives are documented risk factors for breast cancer occurrence in Taiwan (Chie et al., 1996a(Chie et al., , b, 1997a(Chie et al., , b, 1998)).These risk factors represent potential confounders which needs to be taken into account when investigating the role of traffic air pollution in breast cancer mortality.There is, unfortunately, no information available on these variables for individual municipalities and thus could not be adjusted for directly in the analysis.In this study, an integrated indicator of urbanization level and a reproductive indicator (fertility rates) were used to adjust for possible confounding factors resulting from different socioeconomic characteristics and childbearing practices in municipalities.Further, even though more complete information would have been desirable, one measure of the study's internal validity is that the observed associations for PM 2.5 pointed in the direction expected based on previous investigations.
Recent studies suggest that smoking (Cui et al., 2006;Johnson et al., 2011;Luo et al., 2011;Xue et al., 2011) and environmental tobacco smoking (ETS) (Johnson et al., 2011) increase the risk of breast cancer.There is unfortunately no information available on the prevalence rate of smoking and ETS for individual municipality and thus it could not be adjusted for in the analysis.However, the prevalence of smoking among females was very low (only about 4.2%) in Taiwan (Cheng et al., 2001).We think that the degree to which not controlling for this variable may have affected our results should be small if it existed at all.Furthermore, there is no reason to believe that there would be correlation between ETS and the levels of PM 2.5 .
Screening has been shown to be effective in reducing breast cancer mortality (Nelson et al., 2009).Unfortunately, there is no information available on the prevalence of screening utilization for individual municipality studied and therefore it could not be adjusted for directly in the analysis.If breast cancer screening is correlated with the levels of PM 2.5 , a spurious association between PM 2.5 and breast cancer death could result.However, there is no reason to believe that there would be any correlation between the habits of undergoing breast cancer screening and the levels of PM 2.5 .Nonetheless, this is a limitation that should be considered.
PM 2.5 is considered to be the most relevant measure for biological effects of air pollution because these fine particles are respired into the lower respiratory tract (Pope and Dockery, 2006).Our findings should not be interpreted as meaning that PM 2.5 is a causal factor; it is more likely a marker of the complex mixture that is derived from motor vehicle exhaust (Crouse et al., 2010).We speculated that PAHs physically associated with PM 2.5 may be the agent responsible for the association between PM 2.5 and risk of death from breast cancer risk.However, it is important to note that there are numerous other potential carcinogens in traffic emissions as well as PAHs (Nie et al., 2007).We cannot rule out the possibility that other compounds present in PM 2.5 are affecting breast cancer risk or are acting synergistically with PAHs (Bonner et al., 2005).

CONCLUSIONS
In summary, this study is the first to suggest that exposure to high levels of PM 2.5 , a proxy measure of PAHs, may be associated with an increased risk of death from breast cancer.The findings of this study warrant further investigation into the role of air pollutants in the etiology of breast cancer development.

Fig. 1 .
Fig. 1.Locations of the monitoring stations in this study.

Table 2 .
Risk ratio (RR) of mortality from breast cancer in relation to PM 2.5 levels based on a weighted multiple regression model.