Secondary Logo

Journal Logo

Air Pollution

Elemental Composition of Particulate Matter and the Association with Lung Function

Eeftens, Marloesa,b,c; Hoek, Gerarda; Gruzieva, Olenad; Mölter, Annae; Agius, Raymonde; Beelen, Roba; Brunekreef, Berta,f; Custovic, Adnang; Cyrys, Josefh,i; Fuertes, Elainej,k; Heinrich, Joachimj; Hoffmann, Barbaral,m; de Hoogh, Keesn; Jedynska, Aleksandrao; Keuken, Mennoo; Klümper, Claudial; Kooter, Ingeborgo; Krämer, Ursulal; Korek, Michald; Koppelman, Gerard H.p; Kuhlbusch, Thomas A. J.q; Simpson, Angelag; Smit, Henriëtte A.f; Tsai, Ming-Yib,c,r; Wang, Menga; Wolf, Kathrinj; Pershagen, Görand; Gehring, Ulrikea

Author Information
doi: 10.1097/EDE.0000000000000136

Particulate matter (PM) has been widely regarded as serious problem for public health.1–3 Environmental guidelines and standards refer to the mass concentration of particles smaller than 10 μm (PM10) or smaller than 2.5 μm (PM2.5).4–6 However, PM is emitted from a wide variety of sources (eg, combustion; wear of brakes, tires, and clutches; industrial emissions; and crustal materials),7,8 and there is increasing epidemiologic evidence that its various constituents affect health in different ways.8 Knowing more specifically which elements or sources are responsible for the toxicity of PM could be instrumental in preventing adverse health effects more effectively.8,9

Lung function is an objective marker for respiratory health and a predictor for cardiorespiratory morbidity and mortality.10 Negative long-term effects of PM on lung function have been repeatedly shown.2,11 A recently published study also found effects of both nitrogen oxides and PM2.5 on forced expiratory volume in the first second and on peak expiratory flow in the same cohorts used in the present paper and found higher levels of heterogeneity for PM (especially PM10) than for nitrogen oxides.12 Associations between PM and forced expiratory volume in the first second differ between studies,11,13 which may be partly due to spatial differences in PM composition and, hence, in toxicity of particulate air pollutants. Recent reviews identified little conclusive evidence regarding the ability of specific PM constituents to affect long-term respiratory health.1,2,14,15 Some studies have looked at short-term effects of PM composition on lung function,16,17 but to our knowledge, the effects of long-term exposure to specific metal constituents on lung function have not been studied. There is a need for studies that disentangle the differential effects on lung function that might result from various PM constituents.18 This paper aims to investigate the associations between 8 elemental constituents of fine PM (copper [Cu], iron [Fe], potassium [K], nickel [Ni], sulfur [S], silicon [Si], vanadium [V], and zinc [Zn]) and lung function in 5 European birth cohorts, using standardized assessment methods. Furthermore, we examine whether previously observed negative effects of PM2.5 and PM10 on lung function can be attributed to any of these 8 elemental constituents.

The current study is part of the ESCAPE (European Study of Cohorts for Air Pollution Effects; www.escapeproject.eu) and TRANSPHORM (Transport related Air Pollution and Health impacts—Integrated Methodologies for Assessing Particulate Matter; www.transphorm.eu) projects.

METHODS

Study Population

Our analysis included 5 European birth cohorts: the Barn Allergi Milio Stockholm Epidemiologi (BAMSE) study from Sweden; the German Infant Nutritional Intervention study—plus influence of pollution and genetics (GINIplus) and the Lifestyle-related factors on the Immune System and development of Allergies in childhood—plus the influence of traffic emissions and genetics (LISAplus), both from Germany; the Manchester Asthma and Allergy Study (MAAS) from the United Kingdom; and the Prevention and Incidence of Asthma and Mite Allergy (PIAMA) study from the Netherlands. All cohorts were initiated in the mid- to late 1990s to study the onset and development of asthma and allergies during childhood in relation to various environmental factors. More information about study designs and study populations is provided in eAppendix 1 (http://links.lww.com/EDE/A803). Data from subcohorts GINIplus North (Wesel) and GINIplus South (Munich) were analyzed separately to prevent between-area differences in estimated exposure affecting the results. Data from GINIplus and LISAplus North (both from Wesel, Ruhr Area) were analyzed jointly because the same procedures were followed. From these cohorts, we selected all participants with successful lung function measurements at ages 6–8 years, exposure estimates at the 6/8-year address, and complete information on all main confounders. We included 1808 children from Stockholm County, 600 from Munich, 851 from Wesel, 500 from Manchester, and 900 from The Netherlands (eAppendix 1, http://links.lww.com/EDE/A803).

Exposure Assessment

We selected 8 elements (Cu, Fe, K, Ni, S, Si, V, and Zn), reflecting a variety of anthropogenic sources such as brake linings (Cu, Fe, Zn) and tire wear (Zn), industrial (smelter) emissions (Fe, Zn), crustal materials (Si, K), fossil fuel combustion (Ni, V, S), and biomass burning (K), considering existing evidence for toxicity14,15,19,20 and ensuring a high percentage (>75%) of detected samples. Elemental concentrations were estimated for each study subject’s address using land-use regression models developed for each of the 5 study areas.21 Briefly, between October 2008 and February 2010, three 14-day PM measurements in the cold, intermediate, and warm seasons were taken at 20 locations in the Stockholm County, Manchester, Ruhr Area, and Munich/Augsburg study areas and at 40 sites in the Netherlands and Belgium. We used Harvard impactors to collect samples of PM2.5 and PM10 on Teflon filters and analyzed all filters for elemental composition using X-ray fluorescence, as described elsewhere.21 For each site, results from the 3 measurements were averaged to estimate the annual average by adjusting for temporal variation using a centrally located background reference site, which was operated for an entire year in each study area.22 Predictor variables on various metrics of nearby traffic intensity, population/household density, and land use (eg, industry, ports, urban green) were derived from geographic information systems in circular buffers ranging from 25 to 5000 m and were evaluated to explain spatial variation of elemental concentrations.23 Area-specific land-use regression models were made separately for each element within each PM size fraction, as presented previously.21 A brief overview of model performance can be found in eTable 1 (http://links.lww.com/EDE/A803). With the exception of Zn, elements were predominantly found within either PM2.5 (K, Ni, S, V) or PM10 (Cu, Fe, Si).21 This made it less useful to subtract PM2.5 from PM10 for elemental composition than for mass. We assumed that the composition of PM10 was representative primarily for the coarse component. If no significant predictors could be included in a land-use regression model, we did not estimate cohort exposures, as all would be the same. Therefore, there were no exposures estimated for the PM2.5 Ni in Stockholm, for PM2.5 V in Munich, or for PM2.5 K in Wesel and Manchester. The relevant geographic predictors were derived for the cohort addresses. Whenever a predictor had a lower or higher value for one of the cohort addresses than for any of the measurement sites, its value was truncated to (respectively) the minimum or maximum at the measurement sites. Exposures to each element were estimated for the birth addresses and at the addresses for age 6 or 8 years. The latter was the main exposure in these analyses because we observed stronger associations with lung function for the older ages than we did for the birth address in our previous analyses.12

Health Outcomes

At age 6 years (in the Munich and Wesel cohorts) or age 8 years (in the Stockholm County, Manchester, and Dutch cohorts), all children performed spirometry tests and a peak expiratory flow test (except for the Manchester cohort), adhering to the guidelines recommended by the European Respiratory Society and the American Thoracic Society.24 More detailed descriptions of lung function measurements can be found in eAppendix 1 (http://links.lww.com/EDE/A803). As our primary outcome, we assessed forced exhaled volume in the first second, which was available for all cohorts, has high reproducibility and reflects airway obstruction.25 Six-year-old children can generally perform reliable spirometry but have shorter expiratory times. Therefore, forced expiratory volume in the first second cannot always be reliably determined. For the German cohorts, which were measured at age 6, we used forced expiratory volume in the first 0.5 seconds instead. In addition, we studied forced vital capacity and peak expiratory flow, which were available for all except the German and Manchester cohorts, respectively.

Statistical Analyses

We used linear regression to analyze associations between exposure estimates and lung function (log-transformed forced expiratory volume in the first second, forced vital capacity, and peak expiratory flow). Elemental concentrations were entered as continuous variables without transformation. Effect estimates were calculated for fixed increments, chosen by rounding the mean P10–P90 range for all ESCAPE study areas. We further checked for collinearity of predictor variables using a correlation matrix and variance inflation factor.

Data analyses were performed at the Institute for Risk Assessment Sciences (Utrecht University, the Netherlands) for the Dutch and German cohorts using SAS (SAS 9.2, Cary, NC). Analyses for the Stockholm County and Manchester cohorts were done locally, using STATA (Version 11, StataCorp LP, College Station, TX), for the Manchester cohort using SPSS v20 (SPSS Inc., Armonk, NY). Study-specific estimates were combined using random effects meta-analyses to account for heterogeneity between the cohorts.26 We used the I2 statistic to quantify heterogeneity.27 Meta-analyses were performed in STATA (Version 12, StataCorp LP, College Station, TX) using the “metan” command.

Confounding Variables

A common set of potential confounders were defined a priori for all cohorts, following our previous analysis.12 In a crude model, we adjusted only for sex and for log-transformed age, height, and weight. In addition, we adjusted for individual categorical confounders that were not on the causal pathway but were associated with the outcome (P < 0.05) in at least 1 cohort: recent respiratory infections, nonnative ethnicity (binary, as few children were of nonnative ethnicity), parental socioeconomic status, atopic mother, atopic father, breastfeeding, maternal smoking during pregnancy, environmental tobacco smoke, dampness or mould in the home, furry pets in the home, and study area (only for the Stockholm County cohort). We did not adjust for short-term exposures, as daily concentration data on PM constituents were not available for the study areas, and no indication of confounding of long-term associations by short-term exposures was found previously.12 All confounder information was available from questionnaires for all cohorts. Covariates were defined as similarly as possible, given the available information.

Two-pollutant Models

Because of the correlations between many elemental constituents (particularly Cu, Fe, S, Si) and PM mass, associations found for these elements may be biased and instead reflect a relationship with PM mass. We used 2-pollutant models with elemental concentrations adjusting for PM2.5 or PM10 mass to assess the independent effects of specific constituents. We also present associations with PM2.5 and PM10 mass, adjusted for each of their constituents. For the main outcome, forced expiratory volume in the first second, we also present 2-pollutant models for all other combinations of the pollutants and constituents.

Sensitivity Analyses

Some elemental constituents were highly correlated with PM mass, which resulted in collinearity in some PM mass–adjusted models. In order to check whether this changed the effect estimates, we removed from the meta-analysis all models for which the variance inflation factor of the elemental constituent was higher than 3.

We investigated whether effect estimates changed, when we excluded from the meta-analysis those cohorts and elements for which the land-use regression models had relatively low cross-validation R2,21 by excluding 26 (out of 80) exposure models with leave-one out cross-validation R2 lower than 0.50.

In our previous analysis,12 lung function was associated with exposure at the 6- or 8-year address and not with exposure at the birth address. To verify the contribution of exposure at different time points, we conducted an analysis, including both estimated exposures from the birth address and the 6- or 8-year address as co-pollutants, for those children who had moved since birth. In addition, we stratified the main analyses for children who had moved and who had not moved since birth and for asthmatic and nonasthmatic children.

RESULTS

Characteristics of the Study Population

Distributions of lung function indicators and population characteristics for each cohort are shown in Table 1. The Stockholm County cohort had a relatively low percentage of children with atopic parents. The Manchester cohort had a higher percentage of asthmatic children compared with other cohorts.

TABLE 1
TABLE 1:
Description of the Health and Confounder Characteristicsa of the Study Population Included in the Main Analyses

Distribution and Correlation of Air Pollution Exposures

The within-cohort variability of estimated pollutant concentrations differed widely among the cohorts (Figure 1). There was large within-cohort variation for elements from non-tailpipe emissions (Cu, Fe, Zn), especially for PM10. Exposure levels for these elements were comparable for all cohorts. Similar levels were also observed for Ni, although the within-cohort variation was limited for the Munich and Manchester cohorts, because exposure levels were low. There was little consistency across cohorts for K, S, Si, and V exposures. However, exposure contrasts for PM10 Si (and to some extent PM10 K) were large in the Stockholm County cohort and for PM2.5 and PM10 V in the Dutch cohort.

FIGURE 1
FIGURE 1:
Distribution (mean, median, interquartile range, 5th and 95th percentile) of estimated annual average concentrations of PM2.5 and PM10 for copper (Cu), iron (Fe), potassium (K), nickel (Ni), sulfur (S), silicon (Si), vanadium (V), and zinc (Zn). No usable land-use regression models were available for PM2.5 K for the Wesel and Manchester cohorts, for PM2.5 Ni for the Stockholm County cohort, and for PM2.5 V for the Munich cohort; see also text section on exposure to air pollutants.

Correlations between elements and PM mass were much higher in Stockholm County, Wesel, and the Netherlands than in Munich and Manchester (Table 2). Correlations between the measured concentrations were different for some elements, especially for the Munich cohort (eFigure 1, http://links.lww.com/EDE/A803). A potential explanation is that the study area covered by this cohort is highly heterogeneous and thus may not be covered as adequately by the monitoring area as the study areas covered by the other cohorts. Further correlations between elements are presented in eTables 2A–E (http://links.lww.com/EDE/A803). Correlations between the estimated elemental exposures were slightly higher for traffic-related elements (eg, Cu, Fe) and for elements whose mass makes up a relatively large fraction of PM mass (eg, K, S, Si).

TABLE 2
TABLE 2:
Pearson Correlation (R) Between Elemental Composition and PM Mass for Each Cohort

Associations Between Air Pollution and Lung Function

Associations between elemental concentrations and forced expiratory volume in the first second varied among cohorts (Figure 2, Table 3). Heterogeneity was often larger for the elemental constituents than for PM mass (Table 3). A 2 ng/m3 increase in PM10 Ni was associated with a 1.6% (95% confidence interval [CI] = 0.4% to 2.7%) lower forced expiratory volume in the first second, and a 200 ng/m3 increase in PM10 S was associated with a 2.3% (−0.1% to 4.6%) reduction for the same outcome, consistent across 4 of the 5 cohorts. Weights of the cohorts in the meta-analysis are shown in eTable 3 (http://links.lww.com/EDE/A803). There were no large differences between the crude and confounder-adjusted models (eTable 4, http://links.lww.com/EDE/A803). Generally, CIs for the elemental constituents were increased and most negative associations disappeared when we additionally adjusted for PM mass (Table 3). However, associations for PM10 Ni (−1.3% [−3.1% to 0.5%]) and PM10 S (−2.1% [−4.8% to 0.7%]) weakened only slightly and also did not change much after adjustment for the other elemental co-pollutants or for NO2 or PM2.5 absorbance (eFigure 2, http://links.lww.com/EDE/A803). The association between PM10 mass and forced expiratory volume in the first second was reduced to null when adjusted for PM10 Ni or PM10 S, but not when adjusted for the other elements (Figure 3). Negative associations of PM2.5 mass with forced expiratory volume in the first second generally remained after adjustment for various elemental constituents (Figure 3).

TABLE 3
TABLE 3:
Confounder-adjusted and Confounder and PM Mass–adjusted Model Associations Between Estimated Air Pollution Levels and Forced Expiratory Volume in the First Second: Results from Random Effects Meta-analyses Expressed as Percent Change with 95% CIs, I2, and P Value of Heterogeneity of Effect Estimates Between Cohorts
FIGURE 2
FIGURE 2:
Confounder-adjusted cohort-specific and combined associations between estimated air pollution levels and forced expiratory volume in the first second. Adjusted for age, sex, height and weight, recent respiratory infections, nonnative ethnicity/nationality, parental education, allergic mother, allergic father, breastfeeding, maternal smoking during pregnancy, smoking at home, mold/dampness at home, and furry pets at home. BAMSE indicates Stockholm County, Sweden; GINI South, Munich, Germany; GINI/LISA North, Wesel, Germany; MAAS, Manchester, United Kingdom; PIAMA, The Netherlands. Associations are presented for an increase in 5 ng/m3 PM2.5 Cu, 20 ng/m3 PM10 Cu, 100 ng/m3 PM2.5 Fe, 500 ng/m3 PM10 Fe, 50 ng/m3 PM2.5 K, 100 ng/m3 PM10 K, 1 ng/m3 PM2.5 Ni, 2 ng/m3 PM10 Ni, 200 ng/m3 PM2.5 S, 200 ng/m3 PM10 S, 100 ng/m3 PM2.5 Si, 500 ng/m3 PM10 Si, 2 ng/m3 PM2.5 V, 3 ng/m3 PM10 V, 10 ng/m3 PM2.5 Zn, 20 ng/m3 PM10 Zn.
FIGURE 3
FIGURE 3:
Associations between PM mass and forced expiratory volume in the first second, forced vital capacity, and peak expiratory flow, after adjustment for elemental composition. Solid and dotted lines show the effect estimate and 95% CIs of the single-pollutant associations between PM2.5 (above) and PM10 (bottom) and forced expiratory volume in the first second (left), forced vital capacity (middle), and peak expiratory flow (right), as reported by Gehring et al.12 No exposure estimates were available for PM2.5 Ni in the Stockholm County cohort, PM2.5 V in the Munich cohort, and PM2.5 K in the Wesel and Manchester cohorts (see “Exposure Assessment” section), which means the 2-pollutant models adjusted for these elements are based on fewer studies. Therefore, for these elements, single-pollutant PM2.5 associations were recalculated to include only those cohorts that also had valid element exposures. Dots with error bars show the remaining associations with PM2.5 and PM10 mass, after adjusting for each of the elements mentioned below.

Forced vital capacity and peak expiratory flow were also negatively associated with Ni and S (eTables 5 and 6; eFigures 3 and 4, http://links.lww.com/EDE/A803). In addition, negative associations were found with PM2.5 K, PM2.5 V, and PM10 Zn for forced vital capacity and with PM2.5 Cu, PM10 K, PM10 Si, and PM2.5 and PM10 V for peak expiratory flow in the confounder-adjusted models. Heterogeneity in associations was lower for peak expiratory flow than for the other 2 lung function parameters. Associations with Ni and S persisted after additional adjustment for PM mass, whereas most associations attenuated (eTables 5 and 6, http://links.lww.com/EDE/A803). The negative associations between forced vital capacity and PM2.5 and PM10 mass were attenuated after adjustment for PM2.5 Cu and Fe, PM10 Ni and V, and both size fractions of S (Figure 3). Associations of peak expiratory flow with PM2.5 and PM10 mass disappeared after adjusting for PM2.5 Ni and PM10 K, respectively, but not after adjustment for the other elemental pollutants (Figure 3).

Sensitivity Analyses

Estimated concentrations of PM mass were highly correlated (R > 0.80) with PM2.5 Cu (Dutch cohort), PM10 Fe (Wesel and Dutch cohorts), PM10 K (Stockholm County cohort), PM10 Ni (Wesel cohorts), PM2.5 S (Dutch cohort), and PM10 Si (Stockholm County and Dutch cohorts) (Table 2). This led to variance inflation factors above 3 for all health outcomes in PM mass–adjusted models. Combined effect estimates with and without these specific cohorts and exposures are shown in eTable 7 (http://links.lww.com/EDE/A803). The negative association between PM10 Ni and forced expiratory volume in the first second and peak expiratory flow became slightly stronger after exclusion of the Wesel cohort. For forced expiratory volume in the first second, the combined estimate became −2.1% (−3.7% to −0.7%) compared with −1.3% (−3.1% to 0.5%). Otherwise, results remained unchanged when we left out models with variance inflation factors above 3.

When we restricted the meta-analyses to those cohorts for which the exposure models had leave-one out cross-validation R2 of at least 0.50 (eTable 1, http://links.lww.com/EDE/A803), the association of PM10 Ni remained strong, but associations with PM10 S disappeared (eTable 8, http://links.lww.com/EDE/A803). PM2.5 V land-use regression models had low leave-one out cross-validation R2 for all cohorts except the Dutch cohort. Hence, our sensitivity analysis for PM2.5 V was based entirely on the estimate for the Dutch cohort (eTable 8, http://links.lww.com/EDE/A803).

Few associations were seen between exposure at the birth address and lung function (eTable 9, http://links.lww.com/EDE/A803). A consistent negative association with PM2.5 Zn at the birth address was found for all outcomes. However, among the children who had moved since birth (47% of total), exposure at the 6- or 8-year address was more negatively associated with lung function than exposures at the birth address for many exposure–outcome combinations. However, CIs largely overlap (eFigure 5, http://links.lww.com/EDE/A803). Stratified analyses did not reveal systematically different associations for children who had and had not moved since birth (eTable 10, http://links.lww.com/EDE/A803).

In the sensitivity analysis stratified by asthma status, we found that for the majority of exposures, associations with forced expiratory volume in the first second were more negative in asthmatic children (eTable 11, http://links.lww.com/EDE/A803). However, CIs for asthmatic and nonasthmatic children largely overlap.

DISCUSSION

We observed small reductions of forced expiratory volume in the first second, forced vital capacity, and peak expiratory flow in children, associated with increased exposure to various PM elements, especially Ni and S. The negative associations with Ni and S remained unchanged after adjustment for PM mass, suggesting that these elements may be independently associated with lung function. However, the earlier reported associations12 between PM2.5 mass and lung function were not consistently explained by any of the 8 elements. Associations with elements were more heterogeneous across cohorts than were PM mass associations.

Comparison with Previous Studies

Negative effects of long-term exposure to PM mass on lung function have been shown repeatedly2,11—in our recent analysis12 and previously in the cohorts used in the present study.28–30 To date, no studies have looked at lung function in relation to long-term exposure to PM constituents. Studies of PM from specific sources (such as desert dust,31 woodsmoke,32 and traffic11) have found associations with forced expiratory volume in the first second and forced vital capacity. We found the most consistent associations with the elements Ni and S. Previous toxicological studies found adverse effects of residual oil fly ash components (particularly Ni and V) on human bronchial epithelium cells.33 Epidemiologic studies also report short-term effects of Ni, sulfate (which includes the element S), and V for various health outcomes.14,15,19,20 PM containing Ni, sulfate, and V is emitted from the burning of residual oil (eg, from shipping) and some industrial processes.34 Port areas appear in several land-use regression models for elements,21 including Ni, S, and V in the Dutch cohort and PM10 S, PM10 Si, and PM2.5 Zn in the Wesel cohorts.

Heterogeneity

Heterogeneity as assessed by the I2 statistic was low to moderate for the various elements, though generally higher than for particle mass. It is likely that the large differences in elemental exposure contrast between cohorts contributed to the heterogeneity found in the effect estimates. These between-cohort contrasts were smaller for NO2, PM2.5, PM2.5 absorbance, and PM10.12 A presentation of relative, rather than absolute, differences in lung function helped to limit the heterogeneity between cohorts. However, in this and our previous study,12 we cannot rule out the possibility that the moderate heterogeneity may be due to age at lung function measurements, characteristics of the cohorts, and lung function parameters (forced expiratory volume in the first second vs. first half-second).

PM Mass and Elemental Constituents Associations

The previously observed associations between PM2.5 mass and lung function were not explained by any of the 8 elements. A first explanation could be that measurement and exposure estimation errors were smaller for PM2.5 mass than for the elements, resulting in less attenuation. Spatial contrasts for several elemental constituents are strikingly larger than those for PM mass, both within and between study areas. The ability of land-use regression models to explain these contrasts differed among areas and among elements (eTable 1, http://links.lww.com/EDE/A803).21 The explained variance of models for K, Ni, S, Si, and V was generally lower than for PM mass. For the non-tailpipe components Cu, Fe, and Zn, models were as predictive as for PM mass. Although land-use regression models for Cu, Fe, Si, and Zn had higher explained variance, and thus exposure estimation was presumably more accurate, these elements did not have the most profound associations with lung function. Moreover, when we omitted models with low cross-validation R2, in a sensitivity analysis, this did not change the results (eTable 8, http://links.lww.com/EDE/A803).

Secondly, it is possible that other (elemental or organic) components of PM (eg, polyaromatic hydrocarbons or quinones) are related to decreased lung function.14,15 Thirdly, some constituents were highly correlated with PM mass, and hence it was challenging to isolate the effects of individual constituents from this complex mixture. A sensitivity analysis showed that combined effect estimates for individual elements and PM mass remained largely unchanged after adjustment for other co-pollutants.

Strengths and Limitations

The 5 cohorts included in this meta-analysis were comparable in study design and recruitment period (1994–1999). Children of highly educated parents and allergic parents were somewhat overrepresented in the analysis population (eTable 12, http://links.lww.com/EDE/A803).12 As the effect estimates in the present study remained largely unchanged after adjustment for potential confounders including parental education and parental allergy, this likely did not result in serious bias. Similar methods were employed for health effect assessment, whereas exposure assessment methods were completely harmonized across all study areas, and each center used the same statistical methods. The time of day and period of the year when the lung function measurements were taken was not standardized within or between cohorts. However, there were no systematic differences regarding time and season between areas with high and low levels of air pollution within the different cohorts. Lung function was measured once per participant; thus, the analysis was cross-sectional. Although it is possible that we found associations due to multiple testing, we tried to limit the number of models by a priori defining the same exposures and confounders for all cohorts.

Effects of Long-term Exposure at the Birth Versus 6- and 8-year Address

As before,12 we found stronger associations with the exposures at the age 6- and 8-year address than at the birth address (eTable 9, http://links.lww.com/EDE/A803). This could be due to better exposure estimation at the more recent address, as recent measurements formed the basis for land-use regression modeling. It is also possible that lung function is affected primarily by recent exposure and that any growth deficits resulting from exposure earlier in life may be compensated for by moving to a less polluted area.35 A sensitivity analysis restricted to children who had moved since birth showed that PM2.5 and PM10 mass for the 6- and 8-year address had a stronger negative association with lung function than those calculated for the birth address. For Ni and S, no clear difference was found (eFigure 5, http://links.lww.com/EDE/A803). Although it has been suggested that growth deficits need not be permanent if the child moves to a less polluted environment,35,36 studies have shown that children with reduced lung function at an early age still have substantial deficits later in life.37

This study marks a starting point in unraveling the adverse effects on lung function resulting from differences in the composition of PM. We detected small effects on lung function associated with annual average levels of some evaluated elements, particularly nickel and sulfur. The associations remained similar after adjustment for PM mass, suggesting a possible independent effect of these constituents. Reduced lung function was more consistently associated with increased PM mass.

ACKNOWLEDGMENTS

We thank everyone involved in exposure monitoring and modeling, particularly Matthias Birk, Thomas Kusch, Kees Meliefste, and Marjan Tewis.

REFERENCES

1. Brunekreef B, Holgate ST. Air pollution and health. Lancet. 2002;360:1233–1242
2. Pope CA 3rd, Dockery DW. Health effects of fine particulate air pollution: lines that connect. J Air Waste Manag Assoc. 2006;56:709–742
3. Rückerl R, Schneider A, Breitner S, Cyrys J, Peters A. Health effects of particulate air pollution: a review of epidemiological evidence. Inhal Toxicol. 2011;23:555–592
4. 2008 Available at: http://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=CELEX:32008L0050:EN:NOT. Accessed 1 June 2013
5. World Health Organization. Air Quality Guidelines: Global Update 2005: Particulate Matter, Ozone, Nitrogen Dioxide and Sulfur Dioxide. 2006 World Health Organization Regional Office for Europe, Scherfigsvej 8, DK-2100, Copenhagen Ø, Denmark Available at: http://www.euro.who.int/__data/assets/pdf_file/0005/78638/E90038.pdf. Accessed 1 June 2013
6. Forsberg B, Hansson HC, Johansson C, Areskoug H, Persson K, Järvholm B. Comparative health impact assessment of local and regional particulate air pollutants in Scandinavia. Ambio. 2005;34:11–19
7. Kim E, Hopke PK, Pinto JP, Wilson WE. Spatial variability of fine particle mass, components, and source contributions during the regional air pollution study in St. Louis. Environ Sci Technol. 2005;39:4172–4179
8. Schwarze PE, Ovrevik J, Låg M, et al. Particulate matter properties and health effects: consistency of epidemiological and toxicological studies. Hum Exp Toxicol. 2006;25:559–579
9. Peng RD, Dominici F, Pastor-Barriuso R, Zeger SL, Samet JM. Seasonal analyses of air pollution and mortality in 100 US cities. Am J Epidemiol. 2005;161:585–594
10. Künzli N, Tager IB. Air pollution: from lung to heart. Swiss Med Wkly. 2005;135:697–702
11. Götschi T, Heinrich J, Sunyer J, Künzli N. Long-term effects of ambient air pollution on lung function: a review. Epidemiology. 2008;19:690–701
12. Gehring U, Gruzieva O, Agius RM, et al. Air pollution exposure and lung function in children: the ESCAPE project. Environ Health Perspect. 2013;121:1357–1364
13. Hoek G, Pattenden S, Willers S, et al. PM10, and children’s respiratory symptoms and lung function in the PATY study. Eur Respir J. 2012;40:538–547
14. Kelly FJ, Fussell JC. Size, source and chemical composition as determinants of toxicity attributable to ambient particulate matter. Atmos Environ. 2012;60:504–526
15. Stanek LW, Sacks JD, Dutton SJ, et al. Attributing health effects to apportioned components and sources of particulate matter: an evaluation of collective results. Atmos Environ. 2011;45:5655–5663
16. Roemer W, Hoek G, Brunekreef B, et al. PM10 elemental composition and acute respiratory health effects in European children (PEACE project). Pollution Effects on Asthmatic Children in Europe. Eur Respir J. 2000;15:553–559
17. Strak M, Janssen NA, Godri KJ, et al. Respiratory health effects of airborne particulate matter: the role of particle size, composition, and oxidative potential-the RAPTES project. Environ Health Perspect. 2012;120:1183–1189
18. Brunekreef B. The color of smoke. Epidemiology. 2010;21:903–904
19. Mostofsky E, Schwartz J, Coull BA, et al. Modeling the association between particle constituents of air pollution and health outcomes. Am J Epidemiol. 2012;176:317–326
20. Bell ML, Ebisu K, Peng RD, Samet JM, Dominici F. Hospital admissions and chemical composition of fine particle air pollution. Am J Respir Crit Care Med. 2009;179:1115–1120
21. de Hoogh K, Wang M, Adam M, et al. Development of land use regression models for particle composition in twenty study areas in Europe. Environ Sci Technol. 2013;47:5778–5786
22. Eeftens M, Tsai MY, Ampe C, et al. Spatial variation of PM2.5, PM10, PM2.5 absorbance and PMcoarse concentrations between and within 20 European study areas and the relationship with NO2 - Results of the ESCAPE project. Atmos Environ. 2012;62:303–317
23. Eeftens M, Beelen R, de Hoogh K, et al. Development of land use regression models for PM(2.5), PM(2.5) absorbance, PM(10) and PM(coarse) in 20 European study areas; results of the ESCAPE project. Environ Sci Technol. 2012;46:11195–11205
24. Miller MR, Hankinson J, Brusasco V, et al.ATS/ERS Task Force. Standardisation of spirometry. Eur Respir J. 2005;26:319–338
25. Enright PL, Lebowitz MD, Cockroft DW. Physiologic measures: pulmonary function tests. Asthma outcome. Am J Respir Crit Care Med. 1994;149(2 pt 2):S9–S18 discussion S19
26. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials. 1986;7:177–188
27. Higgins JP, Thompson SG. Quantifying heterogeneity in a meta-analysis. Stat Med. 2002;21:1539–1558
28. Eenhuizen E, Gehring U, Wijga AH, et al. Traffic-related air pollution is related to interrupter resistance in 4-year-old children. Eur Respir J. 2013;41:1257–1263
29. Nordling E, Berglind N, Melén E, et al. Traffic-related air pollution and childhood respiratory symptoms, function and allergies. Epidemiology. 2008;19:401–408
30. Schultz ES, Gruzieva O, Bellander T, et al. Traffic-related air pollution and lung function in children at 8 years of age: a birth cohort study. Am J Respir Crit Care Med. 2012;186:1286–1291
31. Middleton N, Yiallouros P, Kleanthous S, et al. A 10-year time-series analysis of respiratory and cardiovascular morbidity in Nicosia, Cyprus: the effect of short-term changes in air pollution and dust storms. Environ Health. 2008;7:39
32. Naeher LP, Brauer M, Lipsett M, et al. Woodsmoke health effects: a review. Inhal Toxicol. 2007;19:67–106
33. Maciejczyk P, Chen LC. Effects of subchronic exposures to concentrated ambient particles (CAPs) in mice. VIII. Source-related daily variations in in vitro responses to CAPs. Inhal Toxicol. 2005;17:243–253
34. Viana M, Kuhlbusch T, Querol X, et al. Source apportionment of particulate matter in Europe: a review of methods and results. J Aerosol Sci. 2008;39:827–849
35. Avol EL, Gauderman WJ, Tan SM, London SJ, Peters JM. Respiratory effects of relocating to areas of differing air pollution levels. Am J Respir Crit Care Med. 2001;164:2067–2072
36. Rojas-Martinez R, Perez-Padilla R, Olaiz-Fernandez G, et al. Lung function growth in children with long-term exposure to air pollutants in Mexico City. Am J Respir Crit Care Med. 2007;176:377–384
37. Gauderman WJ, Avol E, Gilliland F, et al. The effect of air pollution on lung development from 10 to 18 years of age. N Engl J Med. 2004;351:1057–1067

Supplemental Digital Content

© 2014 by Lippincott Williams & Wilkins, Inc