ehp-118-1021

7/28/2019 ehp-118-1021

http://slidepdf.com/reader/full/ehp-118-1021 1/6

Environmental Health Perspectives • volume 118 | number 7 | July 2010 1021

Research | Children’s Health

Te role o air pollution in the development o new-onset asthma remains controversial, and

the contribution o this environmental risk ac-tor to the pandemic remains unclear (Eder et al.2006; Sarnat and Holguin 2007). Althoughincreasing evidence indicates that living nearheavy trac is associated with increased rates o asthma, some well-designed studies have oundonly weak or no associations (Heinrich and Wichmann 2004; Otedal et al. 2009).

Tese inconsistencies may reect incom-plete assessment o exposures to trac in themicroenvironments in which children spendmost o their time. Exposure at locationsother than home, especially at school wherechildren spend a large proportion o waking

hours and may engage in physical activity that would increase the ventilation rate and dose o inhaled pollutants, may have strong inuenceson occurrence o asthma. However, ew stud-ies have examined the eect o trac-relatedpollution at schools on asthma rates among children. wo cross-sectional Dutch studiesthat examined this question reported higherrates o respiratory symptoms among childrenin schools near roadways with heavy traic,especially truck traic (Janssen et al. 2003;van Vliet et al. 1997). A northern Caliornia study ound that schools downwind rom busy

reeways had higher concentrations o oxides o nitrogen (NOx ) and higher asthma prevalence

rates (Kim et al. 2004). However, in a subse-quent analysis o these same data, the eect o school exposure was attenuated and no longersignifcant ater adjusting or modeled residen-tial trac-related exposure (Kim et al. 2008).

Nitrogen dioxide (NO2) is routinely measured at regulatory monitoring stations.However, these measurements relect bothbackground regional concentrations and localsources near the stations. Some studies haveassessed local exposure to trac-related pollu-tion using measured NO2 (or NOx ) as surro-gate or the complex mixture o trac-relatedpollutants that occurs in close proximity to

roadways (Brauer et al. 2007; Gaudermanet al. 2005; Jerrett et al. 2008; Kim et al. 2004;Kramer et al. 2000; Nordling et al. 2008).NO2 can easibly be measured at a large num-ber o locations, and it has been widely used asa proxy or the mixture o trac-related pol-lutants that vary markedly depending on dis-tance rom roadways, season, wind speed, and wind direction. However, the mixture o pol-lutants in close proximity to roadways includestransition metals and organic aerosols, whichare more plausible causes o asthma than NO2 (Li et al. 2003).

Te Southern Caliornia Children’s HealthStudy (CHS) was designed to investigate thechronic eects o air pollution on respiratory health. Previous analyses have ound associa-tions o asthma with residential distance tomajor roads and modeled and measured pol-lutant markers or intracommunity variationin exposure to trac (Gauderman et al. 2005; Jerrett et al. 2008; McConnell et al. 2006;Salam et al. 2007b) and with regional pollut-ants in susceptible children (Islam et al. 2008;McConnell et al. 2002). In the current analy-sis, a prospective, longitudinal evaluation o new cases o asthma has allowed assessment o exposure beore the development o asthma inan ongoing cohort o school children recruitedin kindergarten and frst grade, an age at whichphysician-diagnosed asthma becomes reliableand valid (Martinez et al. 1995). Exposure totraic-related pollutants has been estimatedat participating schools and at residences. Inaddition, NO2, ozone (O3), and particulatematter < 10 µm (PM10) and < 2.5 µm in aero-dynamic diameter (PM2.5) have been measuredcontinuously during the lietime o most study participants at a location in each community representative o exposure in the neighbor-

hoods where children live (Kunzli et al. 2003;McConnell et al. 2006). Tis design allowedexamination o the joint eects o trac-relatedpollution exposure at school and home and o regional measured pollution at community

Address correspondence to R. McConnell, Departmento Preventive Medicine, USC Keck School o Medicine, 1540 Alcazar St., CHP 236, Los Angeles,CA 90033 USA. elephone: (323) 442-1096. Fax:(323) 442-3272. E-mail: [email protected]

Supplemental Material is available online(doi:10.1289/ehp.0901232 via http://dx.doi.org/).

E. Rappaport and J. Manila managed data, andB. Penold assigned exposures estimated rom resi-

dential trac. R. Ma, R.. Burnett, and E. Hughesassisted with the development and implementationo the random efects program.

Tis study was supported by National Instituteso Health grants 5P30ES007048, 5P01ES009581,5 P01ES011627, 1R01ES016535, 5R01ES014447,5R03ES014046, and 5R01HL61768; U.S.Environmental Protection Agency grants R826708,RD831861, R831845, and R82735201; the SouthCoast Air Quality Management District, and theHastings Foundation.

F.L. is employed by Sonoma echnology, Inc,(Petaluma, CA). Te other authors declare they haveno actual or potential competing interests.

Received 22 July 2009; accepted 22 March 2010.

Childhood Incident Asthma and Trafc-Related Air Pollution at Home

and School

Rob McConnell,1 Talat Islam,1 Ketan Shankardass,2 Michael Jerrett,3 Fred Lurmann,4 Frank Gilliland,1 Jim Gauderman,1 Ed Avol,1 Nino Künzli,5 Ling Yao,6 John Peters,1 and Kiros Berhane 1

1University of Southern California, Los Angeles, California, USA; 2St. Michael’s Hospital, Toronto, Ontario, Canada; 3University of California, Berkeley, California, USA; 4Sonoma Technology, Inc, Petaluma, California, USA; 5Swiss Tropical and Public Health Institute,Basel and University of Basel, Switzerland; 6United Health Group, City of Hope Hospital Medical Center, Los Angeles, California, USA

B ackground: Traic-related air pollution has been associated with adverse cardiorespiratory eects, including increased asthma prevalence. However, there has been little study o eects o tra-c exposure at school on new-onset asthma.

oBjectives: We evaluated the relationship o new-onset asthma with trafc-related pollution near homes and schools.

Methods: Parent-reported physician diagnosis o new-onset asthma (n = 120) was identiied during 3 years o ollow-up o a cohort o 2,497 kindergarten and rst-grade children who wereasthma- and wheezing-ree at study entry into the Southern Caliornia Children’s Health Study. Weassessed trafc-related pollution exposure based on a line source dispersion model o trafc volume,distance rom home and school, and local meteorology. Regional ambient ozone, nitrogen dioxide(NO2), and particulate matter were measured continuously at one central site monitor in each o 13study communities. Hazard ratios (HRs) or new-onset asthma were scaled to the range o ambient

central site pollutants and to the residential interquartile range or each trafc exposure metric.r esults: Asthma risk increased with modeled trafc-related pollution exposure rom roadways near homes [HR 1.51; 95% condence interval (CI), 1.25–1.82] and near schools (HR 1.45; 95% CI,1.06–1.98). Ambient NO2 measured at a central site in each community was also associated withincreased risk (HR 2.18; 95% CI, 1.18–4.01). In models with both NO2 and modeled trafc expo-sures, there were independent associations o asthma with trafc-related pollution at school and home, whereas the estimate or NO2 was attenuated (HR 1.37; 95% CI, 0.69–2.71).

conclusions: Trafc-related pollution exposure at school and homes may both contribute to thedevelopment o asthma.

k ey words: air pollution, asthma, child, epidemiology, vehicular trafc. Environ Health Perspect 118:1021–1026 (2010). doi:10.1289/ehp.0901232 [Online 6 April 2010]

7/28/2019 ehp-118-1021


McConnell et al.

1022 volume 118 | number 7 | July 2010 • Environmental Health Perspectives

monitors. We hypothesized that exposure atschool and home both contribute to the risk o new-onset asthma.

Methods

Study population. Characteristics o this cohorthave been described previously (McConnell

et al. 2006). Briely, 5,349 children attend-ing kindergarten and frst grade were enrolledinto a new CHS cohort during the 2002–2003school year rom 45 schools in 13 communities.Communities were selected to represent therange and mixture o regional particulate pol-lutants, NO2, and O3 in southern Caliornia.

Parents provided inormed consent and com-pleted a baseline and yearly questionnaire withinormation about demographic character-istics, respiratory illness, and risk actors orasthma at study entry. Children with a history o physician-diagnosed asthma at study entry (n = 715, 13.4%) were excluded rom thisollow-up, because the ocus was on risks or

incident asthma. o ensure that the study pop-ulation was ree o any previously undiagnosedasthma, we then also excluded children with a history o wheeze (n = 868, 16.2%) and addi-tional children with missing inormation about wheeze, or missing or a “don’t know” answerabout asthma (n = 394, 7.4%). Tese exclu-sions resulted in a total o 3,372 children clas-sifed as disease ree at baseline. We excludedanother 340 children with no residential tracinormation because home address could notbe geocoded [disproportionately rom a singlecommunity, as described in the SupplementalMaterial (doi:10.1289/ehp.0901232)] andanother 535 children who did not participatein any o the annual ollow-up questionnaires.Te fnal sample included 2,497 children.

New-onset asthma and covariates. Children with physician-diagnosed asthma reportedon a yearly questionnaire during 3 years o ollow-up were deined to have new-onsetasthma. Children missing questionnaires inany year continued to contribute person-timeto the models until they answered yes (at whichtime they were censored) or were lost to ol-low-up. Te date o new onset o asthma couldnot be precisely deined. hereore, the dateo onset was assigned to the midpoint o theinterval between the date o the questionnaire

when asthma diagnosis was frst reported andthe date o the previous questionnaire reporting asthma status, and this date was used or esti-mating ollow-up time in all statistical analyses.Sociodemographic characteristics, exposure tocigarette and wildfre smoke, health insurance,housing characteristics, history o allergy, andparental asthma were assessed by questionnaire[see Supplemental Material (doi:10.1289/ehp.0901232) or details].

Community ambient air pollution mea-surements. Ambient levels o O3, NO2, PM10,and PM2.5 were measured continuously dur-ing ollow-up at a central site monitor in each

community, as described in the SupplementalMaterial (doi:10.1289/ehp.0901232) and inprevious reports (Peters et al. 1999). Becausethere is marked diurnal variation in O3 andexposure occurs largely between 1000 and1800 hours, this average was used in all analy-ses. Average temperature and relative humid-ity were also obtained during ollow-up rommeasurements at the monitoring stations.

Local traffic-related pollutant exposure. Exposure to trac-related pollution was esti-mated using methods described previously (McConnell et al. 2006), and several metrics

Table 1. Characteristics o participants at study entry and incidence o asthma on ollow-up.

Characteristic n a (%) Cases IRb (95% CI)

Age at entry (years)< 6 496 (20) 22 17.0 (11.2–25.9)6 1,178 (47) 58 19.0 (14.7–24.6)> 6 823 (33) 40 18.9 (13.9–25.8)

Race/ethnicityHispanic white 1,380 (55) 59 16.8 (13.0–21.7)Non-Hispanic white 905 (36) 45 18.9 (14.0–25.2)Arican American 77 (3.1) 6 33.9 (15.2–75.4)Asian 97 (3.9) 8 30.8 (15.4–61.5)Other/unknown 38 (1.5) 2 22.7 (5.7–90.8)

SexFemale 1,307 (52) 62 18.2 (14.2–23.3)Male 1,190 (48) 58 19.2 (14.8–24.8)

History o allergy

Yes 764 (34) 56 28.7 (22.1–37.4)No 1,508 (66) 50 12.8 (6.7–24.4)Play team sport

Yes 918 (39) 47 19.7 (14.8–26.2)No 1,463 (61) 67 17.9 (9.2–34.5)

Parental history o asthmaYes 389 (17) 31 31.3 (22.0–44.6)No 1,899 (83) 76 15.4 (7.2–33.3)

Maternal smoking during pregnancyYes 156 (6.3) 11 27.9 (15.5–50.5)No 2,308 (93.7) 107 18.0 (5.4–60.3)

Secondhand smokeYes 164 (6.7) 7 17.0 (8.1–35.6)No 2,281 (93.3) 110 18.7 (4.1–84.0)

MildewYes 532 (22) 29 21.1 (14.7–30.4)No 1,911 (78) 85 17.3 (7.9–37.9)

Pests in homeYes 1,603 (67) 75 18.1 (14.4–22.7)No 777 (33) 36 18.2 (9.7–33.9)

Dogs in homeYes 730 (30) 30 15.7 (11.0–22.4)No 1,706 (70) 86 19.7 (9.1–42.8)

Cats in homeYes 462 (19) 17 14.0 (8.7–22.5)No 1,972 (81) 98 19.4 (7.2–52.2)

Indoor NO2 sourceYes 1,799 (73) 88 18.8 (15.3–23.2)No 675 (27) 31 18.3 (9.9–33.9)

Wildfre exposurec

Yes 251 (12) 16 24.6 (15.1–40.1)No 1,764 (88) 79 17.3 (6.2–48.2)

Health insurance

Yes 2,135 (88) 106 19.2 (15.9–23.3)No 300 (12) 10 13.1 (5.7–30.3)

Household income (US$)≤ 14,999 329 (14) 16 19.5 (11.9–31.8)15,000–49,999 714 (31) 38 21.1 (15.4–29.0)> 50,000 1,292 (55) 57 16.6 (12.8–21.6)

Parental educationLess than high school 508 (21) 23 17.9 (11.9–26.9)At least high school 452 (19) 20 17.5 (11.3–27.1)Some college 850 (36) 47 20.0 (14.9–26.9)College and above 560 (24) 24 13.9 (9.1–21.3)

Total 2,497 120 18.7 (15.6–22.3)

a Numbers may not add up to 2,497 in some subgroups because o missing data. b Crude incidence (IR) rate per 1,000person-years. c Occurred in 2003 during frst year o ollow-up.

7/28/2019 ehp-118-1021


Childhood incident asthma and traffic-related air pollution


o local trac related exposure were compiled.Participant residence and school addresses weregeocoded, and we estimated distances to thenearest reeway or other highways or arterialroads and trac density within 150 m o eachchild’s residence and school. Concentrationso pollutants rom local vehicle emissions athomes and schools were estimated separately

rom reeway and nonreeway sources, using a line source dispersion model o the incrementalcontribution by these nearby sources to oxideso nitrogen above the regional backgroundlevels (Benson 1989). he modeled annualconcentration estimates (parts per billion) werebased on distance to roadways, vehicle counts,vehicle NOx emission rates, wind speed anddirection, and height o the mixing layer in eachcommunity. However, these modeled expo-sures reect the mixture o multiple pollutantsrom nearby trac, and the high correlation o pollutants in the mixture precludes identiying the eect o any specifc pollutant in the mix-ture, as described in the Supplemental Material(doi:10.1289/ehp.0901232). For brevity, thesemodeled exposures will be reerred to as trac-related pollutants (RP). An estimate o com-bined school and home exposure to trac wasmade by weighting the estimates by approxi-mate time at school (35 hr weekly, or 21%)and assigning the remaining 79% o time to thehome. Health eect estimates based on modelsin which the contribution o school to the com-bined exposure was reduced to 16% to accountor a lack o school exposure during summerrecess were very similar (data not shown).

Statistical methods. We ftted a multilevelCox proportional hazards model that allows

or assessment o residual variation in time toasthma onset and also or clustering o chil-dren around schools and communities (Ma et al. 2003). Letting ui and uij denote com-munity- and school (within community)-levelrandom eects, with uij assumed to be positiveand independent conditional on ui , we ft theollowing model:

hijl (t )= h0s (t ) uij exp(β X ijl + δT Z ijl ), [1]

where, or the l th subject in the i th com-munity, and j th school, hijl (t ) is the hazardunction at age t ; h0s (t ) is the baseline hazard

unction in stratum s (defned by age at study entry and sex), X ijl is a trac pollution expo-sure metric, and Z ijl are other individual cova-riates, such as secondhand smoke exposure,pets in the home, and other possible con-ounders. All models included race/ethnicity.

he model allowed or joint evaluationo the eects o exposure to traic-relatedpollutants at homes and at schools and toambient pollutants measured at commu-nity central sites, with eects scaled to theinterquartile range (IQR) or each metric o residential exposure (e.g., or RP rom the

line source dispersion model) and to the totalrange across the 13 communities, respectively.raic exposure at homes and school werecorrelated. hereore, in models including both exposures, home trac estimates weredeviated rom the corresponding school met-ric to minimize the chance o collinearity andallow or valid independent eect estimates

(or example, r -square or school with deviatedhome RP were < 0.38 or reeway, nonreeway, and total RP). Although the school-specifc estimates o exposure were o primary interest, models including the average resi-dential exposure corresponding to each school(and residential exposures deviated rom thisaverage) were also itted [see SupplementalMaterial (doi:10.1289/ehp.0901232) ordetails]. Potential conounders were evalu-ated one at a time, based on whether the esti-mates or the pollutant associations changedby > 10%. We assessed heterogeneity o tra-ic pollution eects by level o community central site regional pollutant measurementsby comparing nested models using a partiallikelihood ratio test with and without inter-action terms. We examined any potentialnonlinearity in the exposure–response rela-tionship using cubic spline terms, piecewisepolynomials joined smoothly at a number o break points (Hastie and ibshirani 1990),or the exposure terms and comparing thenested models using a partial likelihood ratiotest. All analyses were conducted using sotware designed to run within R sotware (R Development Core eam, Vienna, Austria)or implementing random eects Cox propor-tional hazards models (Krewski et al. 2009;

Ma et al. 2003). All hypotheses were tested

assuming a 0.05 signifcance level and a two-sided alternative hypothesis.

Results

Most o the 2,497 disease-ree children includedin this analysis were ≤ 6 years o age (range,4.8–9.0 years) at study entry (able 1). Tere were 120 new cases o asthma, resulting in an

incidence rate o approximately 18.7 cases per1,000 person-years (based on 6,434 person-years o ollow-up). Rates did not dier by ageat study entry or sex. About hal the children were Hispanic whites, and they had the low-est rates o asthma (16.8/1,000 person years), whereas Arican Americans had the highest rate(33.9/1,000 person-years), but this was basedon only six cases. Substantially higher rates were also observed or children with history o allergy, parental history o asthma, and mater-nal smoking during pregnancy.

Both residential and school RP met-rics had a wide range with skewed distribu-tions (able 2). For example, the mean andmedian or nonreeway residential RP were7.3 and 6.1 ppb, respectively, and the IQR was 8.0 ppb, with a total range rom 0.08 toa maximum o 55.1 ppb. A ew homes andschools had very low reeway RP (largely in a single community that has no reeways). Tere was ≥ 3-old variation in the multiyear averageNO2, PM10, and PM2.5 rom the lowest tohighest pollution community or measure-ments at the central site monitors. O3 levelsvaried by 2-old.

Te associations o new-onset asthma werestrongest with nonreeway RP. he haz-ard ratio (HR) at homes was 1.51; 95% con-

idence interval (CI), 1.25–1.81; p < 0.001

Table 2. Distribution o annual average residential trafc-modeled pollution and o community central sitemeasurements.

Mean Median IQR Minimum Maximum Range

Residential trafc (ppb)

Nonreeway TRP 7.3 6.1 8 0.08 55.1 55

Freeway TRP 11.1 7.3 13 < 0.0001 134.5 134.5

Total TRP 18.4 14.6 20.9 0.23 144.1 143.9

School trafc (ppb)

Nonreeway TRP 6.1 5.9 5.9 0.3 19.7 19.4

Freeway TRP 10.9 8.2 15.2 < 0.0001 39.9 39.9

Total TRP 17 11.9 18.5 0.7 51.4 50.7

Central-site measurements

NO2

(ppb) 20.4 21.8 12.8 8.7 32.3 23.6

PM10 (µg/m3) 35.5 34 11.7 17.6 61.5 43.9

PM2.5 (µg/m3) 13.9 15.1 9.7 6.3 23.7 17.4

1000–1800 hours O3 (ppb) 44.6 43.6 11.1 29.5 59.8 30.3

Table 3. Association between new-onset asthma and modeled exposure at home and school.

Trafc-related exposurea Home HRb (95% CI) School HRb (95% CI) Combinedc HRb (95% CI)

Nonreeway TRP 1.51 (1.25–1.81)* 1.45 (1.06–1.98) 1.61 (1.29–2.00)*

Freeway TRP 1.12 (0.95–1.31) 1.08 (0.86–1.34) 1.12 (0.94–1.35)

Total TRP 1.32 (1.08–1.61)** 1.20 (0.91–1.58) 1.34 (1.07–1.68)

a Scaled to the IQR at homes or each metric (rom Table 2). b HR (95% CI, adjusted or race/ethnicity and or baselinehazards strata o age at study entry and sex) with random eects or community and school. c Combined weighted or

time at home and school. *p < 0.001. **p < 0.01.

7/28/2019 ehp-118-1021


McConnell et al.


(able 3). he HR or exposure at schools was almost as large as or residential exposure(HR = 1.45; 95% CI, 1.06–1.98), and thecombined average exposure weighted or timeat school and home had a slightly strongerassociation (HR = 1.61; 95% CI, 1.29–2.00).hese associations were not conounded by any o the covariates in able 1 or by com-

munity relative humidity, which we havepreviously ound to conound RP exposure(Jerrett et al. 2008). Tere also were no sig-niicant interactions o either residential orschool nonreeway RP with sex, allergy, orparental history o asthma, which we havepreviously shown modiied eects o traicexposure (McConnell et al. 2006). Similareect sizes were observed in sensitivity analy-ses restricted to lietime residents at the sameaddress, analyses excluding children diagnosedin the frst year o ollow-up and excluding children who changed schools during theollow-up period. here was little evidenceo nonlinearity in the exposure–response rela-tionship based on sensitivity analyses compar-ing the ft o a smoothed cubic spline modelo asthma with a linear model ( p-value > 0.80)or the partial likelihood ratio test or models with 3 and 5 knots compared with the linearmodel. Weaker associations were observed with total RP at homes (HR = 1.32; 95%CI, 1.08–1.61) and school (HR = 1.20; 95%CI, 0.91–1.58). Tere were weaker and non-signiicant associations with RP modeledrom reeways. Nonreeway RP was mod-erately to strongly correlated with reeway RP at homes (R = 0.64) and at schools[R = 0.70; Supplemental Material, able E-1,

(doi:10.1289/ehp.0901232)]. However, inmodels co-adjusted or reeway and nonreeway RP, the nonreeway eect est imatesor school and home were similar to theunadjusted estimates.

O the regional community central-sitepollutants, NO2 was associated with morethan double the risk o new-onset asthma (HR = 2.17; 95% CI, 1.18–4.00; p = 0.01) overthe range o exposure (23.6 ppb) across the 13study communities (able 4). In a model withNO2, school and residential nonreeway RPexposure, the estimate or NO2 was attenuated(HR = 1.37; 95% CI, 0.69–2.71; able 5).

However, the adjusted associations o asthma with nonreeway RP at homes (HR = 1.46)and schools (HR = 1.45) in able 5 werevery similar to the unadjusted eect estimates(HR = 1.51 and 1.45, respectively, shown inable 3). In sensitivity analyses, a similar pat-tern o attenuation o regional NO2 eect esti-mates was observed in models weighted or

time at school and home (results not shown),and in models adjusted or total (reeway plusnonreeway) RP exposure at school andhome [Supplemental Material, able E-2(doi:10.1289/ehp.0901232)]. Tere were nei-ther statistically signifcant interactions o cen-tral site pollutants with residential RP norany consistent patterns suggestive o dierenteects in high and low regional pollutant com-munities (results not shown). We have previ-ously shown that O3 modifed the associationo team sports participation with new-onsetasthma in an older cohort (McConnell et al.2002), but there was neither a main eect o sports nor conounding or eect modifcationo eects o residential RP or ambient centralmonitor pollutant exposure.

Distance to reeways and other major roadsand trac volume on those roads have beenused in other studies as independent predictorso traic exposure, so we examined the dis-tribution [Supplemental Material, able E-3(doi:10.1289/ehp.0901232)] and correla-tions o these metrics (Supplemental Material,able E-1) and the association at school andhome with asthma (Supplemental Material,ables E-4 and E-5). In general, these exposuremetrics were weakly to moderately correlated with the RP metrics but were not consistently

associated with asthma in our data set.

Discussion

Te study is unique in its prospective assess-ment o the relationship o new-onset child-hood asthma to community regional airpollution and near-source traic-relatedexposure at home and in a large number o schools. Te results indicate that associationso asthma with trac-related pollution romnearby sources at schools were independent o estimated eects o exposures at homes, andthese eects were o similar size over a rangeo exposure common in Southern Caliornia.

Evidence o a school eect comparable withthat associated with the much larger amounto time spent at home could potentially beexplained by physical education and otherexercise at school that may increase ventila-tion rate and dose o pollutants to the lungsand thereby increase the risk associated withexposure. rac-related pollutant levels may

also be considerably higher during the morn-ing hours, when children are arriving at school,especially during temperature inversions thatoccur largely in the winter months when chil-dren are attending school (Kim et al. 2002;Ning et al. 2007). It is possible that schoolexposure may have reected time spent in otherlocations in the child’s neighborhood ratherthan school-specifc exposure, and this possibil-ity merits urther investigation. However, timeat school during the school year accounts ormore than one-third o all waking hours, andin this age group it is likely that most o thechild’s remaining nonschool time was spentat home (Xue et al. 2004). Te increased risk associated with RP could not be explainedby conounders commonly associated withasthma or by ambient NO2 or other currently regulated regional pollutants measured at thecentral-site monitoring stations. Tese resultsstrengthen an emerging body o evidence romboth toxicologic and epidemiologic studies thatair pollutants rom nearby trac contribute tothe development o asthma (Salam et al. 2008).

NO2 measured at community monitor-ing stations has been associated with wheezeprevalence in an older CHS cohort (Peterset al. 1999) and with increased risk o asthma incidence in a Japanese cohort (Shima et al.

2002). We have recently reported associationso both incident and prevalent asthma withambient residential NO2 measured outsidehomes o a relatively small sample o childrenrom another CHS cohort in many o thesame communities, but NO2 in that study was a marker or intracommunity variationin the mixture o RPs (Gauderman et al.2005; Jerrett et al. 2008). Other studies alsoindicate that associations o respiratory dis-ease with intracommunity variation in NO2 relect the eects o other RP rather thanNO2 (Kramer et al. 2000). A study that meas-ured NO2 and NOx in central neighborhood

Table 4. Association o new-onset asthma withcommunity central site pollutant measurements.

Pollutant HRa (95% CI)

NO2 2.17 (1.18–4.00)

PM10 1.35 (0.64–2.85)

PM2.5 1.66 (0.91–3.05)

O3 0.76 (0.38–1.54)

a HR (95% CI) across the range o exposure in the 13 com-munities (23.6 ppb or NO2, 43.9 µg/m3 PM10, 17.4 µg/m3 PM2.5, and 30.3 ppb or 1000–1800 hours O3), adjusted orrace/ethnicity and or baseline hazards strata o age atstudy entry and sex with random eects or communityand school.

Table 5. Mutually adjusted associations o new-onset asthma with community central site pollutantmeasurements and nonreeway TRP at home and school.a

Central sitepollutant

HRb (95% CI) or ambientpollutant, adjusted or

TRP at home and school

HRb (95% CI) or home TRP,adjusted or school TRPand ambient pollutant

HRb (95% CI) or schoolTRP, adjusted or home TRP

and ambient pollutant

NO2 1.37 (0.69–2.71) 1.46 (1.16–1.84)* 1.45 (1.03–2.06)PM10 1.40 (0.62–3.17) 1.46 (1.16–1.85)* 1.53 (1.10–2.12)*PM2.5 1.30 (0.66–2.56) 1.48 (1.19–1.85)* 1.49 (1.07–2.08)O3 1.01 (0.49–2.11) 1.50 (1.20–1.86)* 1.54 (1.10–2.14)*

a Mutually adjusted across each row (i.e., eect o each community pollutant was examined separately in a modelincluding both home and school TRP). b HR (95% CI) or central-site pollutants scaled across the range o exposure in the13 communities (23.6 ppb or NO2, 43.9 μg/m3 PM10, 17.4 μg/m3 PM2.5, and 30.3 ppb or 1000–1800 hours O3); householdnonreeway TRP was deviated rom school, scaled to the IQR or home exposure (8 ppb rom Table 2). * p < 0.01.

7/28/2019 ehp-118-1021


Childhood incident asthma and traffic-related air pollution


locations ound that associations with asthma were attenuated by adjustment or modeledresidential trac exposure (Kim et al. 2004,2008). In the current study, the attenuatedassociation between asthma and NO2 con-tinuously measured at the community moni-tor in models with adjustment or RP alsosuggests that NO2 was not causally related to

asthma. However, interpretation o this resultis not clear because exposure measured at thecommunity monitor may have misclassifedexposures o children in parts o the com-munity with signifcant local sources o NO2. We did not fnd evidence or main eects o regional O3 and PM with asthma, consistent with results rom previous analyses in olderCHS cohorts, although we have previously shown that O3 modifed the eect o outdoorexercise in a genetically susceptible subgroup(Islam et al. 2009; McConnell et al. 2002).

Associations with asthma were signifcantor RP exposure estimates modeled romlocal nonreeway roadway proximity, traicvolume, and meteorology. Tere was little evi-dence or an eect o major roadway proxim-ity alone, or traic density, or or pollutionrom reeways. Te absence o a reeway RPeect suggests that causal pollutants may behighly reactive, resulting in steep spatial gradi-ents rom reeway sources and little exposureat the longer distances o homes and schoolsto reeways compared with distance to othermajor roads [Supplemental Material, able E-3(doi:10.1289/ehp.0901232)]. Several previ-ous studies have ound that the largest gradi-ents in trac-related pollutants occur within150 m rom roadways (Gilbert et al. 2003;

Zhu et al. 2002), and variability has oten beenbest explained by trac volume within 300 m(Briggs et al. 2000; Gilbert et al. 2003; Rosset al. 2006). High-volume trac on secondary roads may also produce pollutants resulting rom requent stops and accelerations, whichhave been ound to be associated with asthma symptoms (Ryan et al. 2005). However, in anolder cohort rom some o these same commu-nities, we observed associations o reeway-mod-eled exposures and residential reeway proximity with respiratory outcomes (Gauderman et al.2005, 2007). Dierent trac metric associa-tions with respiratory health may have been due

to variation in the distribution o reeway andnonreeway traic around homes in the di-erent cohorts and the relatively small numbero children living near reeways (< 7% within150 m in the current cohort). It is also possiblethat the accuracy o exposure models or thereeway and nonreeway sources varied in thetwo cohorts or that the greater mobility o olderchildren resulted in health eects associated with spending time in areas such as parks inclose proximity to reeways near homes.

An imp or tant distinc tion bet wee nthe RP and simpler traic metrics is the

inclusion o meteorology (average annual wind speed and direction and height o themixing layer) in addition to proximity andvolume. rac density alone [SupplementalMate r i a l , ab le E -4 (do i :10 .1289/ehp.0901232)] and proximity to a majorroadway (Supplemental Material, ables E-4and E-5) had weaker positive associations

with incident asthma that were not statisti-cally signifcant. Tereore, inconsistent fnd-ings in some o the (largely cross-sectional)studies that have evaluated associations withasthma (Heinrich and Wichmann 2004) may reect the ailure to account or the impact o meteorology on exposure. We have recently measured NOx at > 900 locations in thesecommunities. Approximately two thirds o the within-community variability in NOx was explained by RP, a substantially higheramount o variation than was explained by any other trac metric (unpublished data). We have also previously observed that bothRP and residential traic proximity wereassociated with increased asthma prevalencein this cohort (McConnell et al. 2006).

Te prospective design is a strength o thisstudy. In other prospective studies o birthcohorts in Sweden and the Netherlands, mod-eled trac-related pollutants, including NOx and/or PM2.5, were associated with incidenceo wheeze and atopy (Nordling et al. 2008)and doctor-diagnosed asthma (Brauer et al.2007) in children < 5 years o age, when thediagnosis o asthma may be dicult to distin-guish rom transient wheeze not predictive o asthma. As these cohorts mature, the relation-ship to asthma is likely to become more clear.

For example, associations between PM2.5 lightabsorption, roadway proximity, and asthmaticbronchitis and atopy were ound in a study o two German birth cohorts that ollowed somechildren to 6 years o age (Morgenstern et al.2008). Associations with allergic symptomsin these cohorts were less robust at an earlierage (Gehring et al. 2002). Other recent resultsrom prospective studies o birth cohorts(Gehring et al. 2009) and o children (Shima et al. 2002) ollowed to older ages have gener-ally shown positive associations between mark-ers o residential RP and asthma, although a recent Norwegian study ound no association

with modeled exposure at the birth address o a large cohort o children (Otedal et al. 2009).Genetic studies examining pathways likely tomediate eects o air pollution also strengthenthe causal inerence that trac-related pollut-ants may cause asthma. We have shown else- where that the risk o asthma associated withtrac exposure and with trac-related ambi-ent PM was modifed in a predictable way by unctional gene variants in pathways associ-ated with asthma, including inammation andairway remodeling (Salam et al. 2007a) andoxidative stress (Islam et al. 2008; Salam et al.

2007b). In another analysis o the data romthis cohort, we have shown that the estimatedeect o RP was modifed by socioeconomicstatus and psychosocial stress (Shankardasset al. 2009).

Tere are some limitations to these data.Our conservative estimates o RP or thecenter o school buildings are likely to under-

estimate the exposure o students in build-ings closer to the roadways, especially during dropo and pickup o students at the begin-ning and end o the school day. Te impact o this underestimate o exposure on the assess-ment o health eects is not clear, although ingeneral it might be expected that this exposuremisclassifcation would be nondierential withrespect to asthma and would result in underes-timation o the eect o trac exposure. Otherlimitations include the relatively short (3-year)period o ollow-up and the retrospective ques-tionnaire assessment o early-lie risk actors, asthis was not a pregnancy cohort. Our cohortexcluded children who let the community beore school entry and who had asthma or wheeze at study entry. However, our cohortallowed us to examine risk actors in schoolchildren, an age during which there has beenlittle study o new-onset asthma.

Asthma is a clinical syndrome with nosensitive and speciic diagnostic test avail-able to confrm clinical assessments. Becauseo the clinical nature o the assessment, thereported physician diagnosis o asthma we usedhas been recommended and widely used asa method to classiy asthma status in epide-miologic studies, although this approach haslimitations (Asher et al. 1995; Burr 1992).

I misclassiication o true asthma based onthis approach were random with respect toexposure to air pollution, then the observedassociations would be likely to have under-estimated the eect o exposure. Bias couldexplain our results i asthma misclassifcation were related also to exposure. Tis could occuri exposure were associated with access to careand dierences in practice among physiciansthat have the potential to inluence asthma diagnosis (Samet 1987). However, adjustmentor actors that mediate access to care including amily income, parental education, and having medical insurance did not alter our results,

indicating that conounding due to dieren-tial access to care was unlikely to explain thesefndings. Te validity o questionnaire-reportedphysician diagnosis has also been establishedin a subset o CHS participants (Salam et al.2007a). In addition, any child with a history o wheezing at study entry was excluded romthe analysis, so it is unlikely that the exacerba-tion o undiagnosed prevalent asthma by tracexplained these associations. Loss to ollow-upcould also potentially result in bias becauseo selection. Sociodemographic characteristicsvaried between the analysis cohort and children

7/28/2019 ehp-118-1021


McConnell et al.


not contributing to the analysis because o lossto ollow-up or missing or poorly matchedgeocodes or addresses precluding estimationo residential exposure [Supplemental Material,able E-6 (doi:10.1289/ehp.0901232)].Hispanic children were less likely than non-Hispanic white children to be included in theollow-up analysis, as were children without

insurance and with lower parental educationand income. However, adjusting or these ac-tors had little impact on the pattern o observedeects. In addition, there was little dierencebetween exposure to nonreeway RP in thestudy sample (7.3 ppb) and in the group notincluded because o loss to ollow-up (7.6 ppb).Tereore, it is unlikely that selection bias dueto loss to ollow-up explained our results.

Conclusions

An estimated 6.2 million children have asthma in the United States, making it the most com-mon chronic disease in childhood (Moormanet al. 2007), and rates have increased mark-edly in developed countries over the past severaldecades (Braman 2006). Morbidity results inimpaired quality o lie or the aected childand other amily members, in increased useo health services, and in school absences thathave large social and economic costs (Wang et al. 2005). Our results indicate that childrenexposed to higher levels o trac-related air pol-lution at school and home are at increased risk o developing asthma. Almost 10% o publicschools in Caliornia are located within 150 mo roadways with > 25,000 vehicles daily (Greenet al. 2004). Students in urban areas in easternU.S. cities are even more likely than children in

Los Angeles to attend schools near major highways (Appatova et al. 2008). Tereore, expo-sure to RP is potentially an important publichealth problem aecting large populations o children. Planning transportation and otherurban development to limit population expo-sure to trac exhaust, as well as more eectivecontrol o vehicular emissions, may result insubstantial long-term public health benefts.

RefeRences

Appatova AS, Ryan PH, Lemasters GK, Gruinshpun SA. 2008.

Proximal exposure o public schools and students to

major roadways: a nationwide US survey. J Environ Plann

Manage 51:631–646.Asher MI, Keil U, Anderson HR, Beasley R, Crane J, Martinez F,

et al. 1995. International Study o Asthma and Allergies in

Childhood (ISAAC): rationale and methods. Eur Respir J

8(3):483–491.

Benson P. 1989. Caline4—A Dispersion Model or

Predicting Air Pollution Concentration Near Roadways.

Sacramento:Caliornia Department o Transportation,

Ofce o Transportation Laboratory.

Braman SS. 2006. The global burden o asthma. Chest 130(suppl

1):4S–12S.

Brauer M, Hoek G, Smit HA, de Jongste JC, Gerritsen J,

Postma DS, et al. 2007. Air pollution and development

o asthma, allergy and inections in a birth cohort. Eur

Respir J 29(5):879–888.

Briggs DJ, de Hoogh C, Gulliver J, Wills J, Elliott P, Kingham S,

et al. 2000. A regression-based method or mapping

trafc-related air pollution: application and testing in our

contrasting urban environments. Sci Total Environ 253(1–

3):151–167.

Burr ML. 1992. Diagnosing asthma by questionnaire in epidemi-

ological surveys [Editorial]. Clin Exp Allergy 22(5): 509–510.

Eder W, Ege MJ, von Mutius E. 2006. The asthma epidemic.

N Engl J Med 355(21):2226–2235.

Gauderman WJ, Avol E, Lurmann F, Kuenzli N, Gilliland F,

Peters J, et al. 2005. Childhood asthma and exposure to

trafc and nitrogen dioxide. Epidemiology 16(6):737–743.

Gauderman WJ, Vora H, McConnell R, Berhane K, Gilliland F,

Thomas D, et al. 2007. Eect o exposure to trafc on lung

development rom 10 to 18 years o age: a cohort study.

Lancet 369(9561):571–577.

Gehring U, Cyrys J, Sedlmeir G, Brunekree B, Bellander T,

Fischer P, et al. 2002. Traic-related air pollution and

respiratory health during the frst 2 yrs o lie. Eur Respir J

19(4):690–698.

Gehring U, Wijga AH, Brauer M, Fischer P, de Jongste JC,

Kerkho M, et al. 2009. Traic-related air pollution and

the development o asthma and allergies during the frst 8

years o lie. Am J Respir Crit Care Med 181(6):596–603.

Gilbert NL, Woodhouse S, Stieb DM, Brook JR. 2003. Ambient

nitrogen dioxide and distance rom a major highway. Sci

Total Environ 312(1–3):43–46.

Green RS, Smorodinsky S, Kim JJ, McLaughlin R, Ostro B.

2004. Proximity o Caliornia public schools to busy roads.

Environ Health Perspect 112(1):61–66.

Hastie TJ, Tibshirani RJ. 1990. Generalized Additive Models.

London:Chapman & Hall.

Heinrich J, Wichmann HE. 2004. Traic related pollutants in

Europe and their eect on allergic disease. Curr Opin

Allergy Clin Immunol 4(5):341–348.

Islam T, Berhane K, McConnell R, Gauderman WJ, Avol E,

Peters JM, et al. 2009. Glutathione-S -transerase (GST) P1,

GSTM1, exercise, ozone and asthma incidence in school

children. Thorax 64(3):197–202.

Islam T, McConnell R, Gauderman WJ, Avol E, Peters JM,

Gilliland FD. 2008. Ozone, oxidant deense genes, and risk

o asthma during adolescence. Am J Respir Crit Care Med

177(4):388–395.

Janssen NA, Brunekree B, van Vliet P, Aarts F, Melieste K,

Harssema H, et al. 2003. The relationship between air pollu-

tion rom heavy trafc and allergic sensitization, bronchial

hyperresponsiveness, and respiratory symptoms in Dutchschoolchildren. Environ Health Perspect 111:1512–1518.

Jerrett M, Shankardass K, Berhane K, Gauderman WJ,

Künzli N, Avol E, et al. 2008. Trafc-related air pollution

and asthma onset in children: a prospective cohort study

with individual exposure measurement. Environ Health

Perspect 116:1433–1438.

Kim JJ, Huen K, Adams S, Smorodinsky S, Hoats A, Malig B,

et al. 2008. Residential trafc and children’s respiratory

health. Environ Health Perspect 116:1274–1279.

Kim S, Shen S, Sioutas C. 2002. Size distribution and diurnal and

seasonal trends o ultrafne particles in source and recep-

tor sites o t he L os Angeles basin. J A ir Waste Manag

Assoc 52(3):297–307.

Kim JJ, Smorodinsky S, Lipsett M, Singer BC, Hodgson AT,

Ostro B. 2004. Trafc-related air pollution near busy roads:

the East Bay C hildren’s Respiratory Health Study. Am J

Respir Crit Care Med 170(5):520–526.

Kramer U, Koch T, Rant U, Ring J, Behrendt H. 2000. Trafc-

related air pollution is associated with atopy in childrenliving in urban areas. Epidemiology 11(1):64–70.

Krewski D, Jerrett M, Burnett RT, Ma R, Hughes E, Shi Y,

et al. 2009. Extended ollow-up and spatial analysis o the

American Cancer Society study linking particulate air pol-

lution and mortality. Res Rep Health E Inst 140:5–114.

Kunzli N, McConnell R, Bates D, Bastain T, Hricko A, Lurmann F,

et al. 2003. Breathless in Los Angeles: the exhausting

search or clean air. Am J Public Health 93(9):1494–1499.

Li N, Hao M, Phalen RF, Hinds WC, Nel AE. 2003. Particulate air

pollutants and asthma. A paradigm or the role o oxidative

stress in PM-induced adverse health eects. Clin Immunol

109(3):250–265.

Ma R, Krewski D, Burnett RT. 2003. Random eects Cox models:

a Poisson modeling approach. Biometrika 90:157–169.

Martinez FD, Wright AL, Taussig LM, Holberg CJ, Halonen M,

Morgan WJ. 1995. Asthma and wheezing in the irst six

years o lie. The Group Health Medical Associates. N Engl

J Med 332(3):133–138.

McConnell R, Berhane K, Gilliland F, London SJ, Islam T,

Gauderman WJ, et al. 2002. Asthma in exercising children

exposed to ozone: a cohort study. Lancet 359(9304):386–391.

McConnell R, Berhane K, Yao L, Jerrett M, Lurmann F,

Gilliland F, et al. 2006. Trafc, susceptibility, and childhood

asthma. Environ Health Perspect 114:766–772.

Moorman JE, Rudd RA, Johnson CA, King M, Minor P, Bailey C,

et al. 2007. National surveillance or asthma—United

States, 1980–2004. MMWR Surveill Summ 56(8):1–54.

Morgenstern V, Zutavern A, Cyrys J, Brockow I, Koletzko S,

Kramer U, et al. 2008. Atopic diseases, allergic sensitiza-

tion, and exposure to trafc-related air poll ution in chil-

dren. Am J Respir Crit Care Med 177(12):1331–1337.

Ning Z, Geller MD, Moore KF, Sheesley R, Schauer JJ,

Sioutas C. 2007. Daily variation in chemical characteristics

o urban ultrafne aerosols and inerence o their sources.

Environ Sci Technol 41(17):6000–6006.

Nordling E, Berglind N, Melen E, Emenius G, Hallberg J,

Nyberg F, et al. 2008. Traic-related air pollution and

childhood respiratory symptoms, unction and allergies.

Epidemiology 19(3):401–408.

Otedal B, Nystad W, Brunekree B, Nastad P. 2009. Long-term

trafc-related exposures and asthma onset in schoolchil-

dren in Oslo, Norway. Environ Health Perspect 117:839–844.

Peters JM, Avol E, Navidi W, London SJ, Gauderman WJ,

Lurmann F, et al. 1999. A study o twelve Southern

Caliornia communities with diering levels and types o

air pollution. I. Prevalence o respiratory morbidity. Am J

Respir Crit Care Med 159(3):760–767.

Ross Z, English PB, Scal R, Gunier R, Smorodinsky S, Wall S,

et al. 2006. Nitrogen dioxide prediction in southern

Caliornia using land use regression modeling: potential

or environmental health analyses. J Expo Sci Environ

Epidemiol 16(2):106–114.

Ryan PH, LeMasters G, Biagini J, Bernstein D, Grinshpun SA,

Shukla R, et al. 2005. Is it trafc type, volume, or distance?

Wheezing in inants living near truck and bus traic.

J Allergy Clin Immunol 116(2):279–284.

Salam MT, Gauderman WJ, McConnell R, Lin PC, Gilliland FD.

2007a. Transorming growth actor-β1 C-509T polymor-

phism, oxidant stress, and early-onset childhood asthma.

Am J Respir Crit Care Med 176(12):1192–1199.

Salam MT, Islam T, Gilliland FD. 2008. Recent evidence oradverse eects o residential proximity to trafc sources

on asthma. Curr Opin Pulm Med 14(1):3–8.

Salam MT, Lin PC, Avol EL, Gauderman WJ, Gilliland FD. 2007b.

Microsomal epoxide hydrolase, glutathione S -transerase

P1, trafc and childhood asthma. Thorax 62(12):1050–1057.

Samet JM. 1987. Epidemiologic approaches or the identifca-

tion o asthma. Chest 91(6 suppl):74S–78S.

Sarnat JA, Holguin F. 2007. Asthma and air quality. Curr Opin

Pulm Med 13(1):63–66.

Shankardass K, McConnell R, Jerrett M, Milam J, Richardson J,

Berhane K. 2009. Parental stress increases the eect o air

pollution on childhood asthma incidence. Proc Natl Acad

Sci USA 106:12406–12411.

Shima M, Nitta Y, Ando M, Adachi M. 2002. Eects o air pol-

lution on the prevalence and incidence o asthma in chil-

dren. Arch Environ Health 57(6):529–535.

van Vliet P, Knape M, de Hartog J, Janssen N, Harssema H,

Brunekree B. 1997. Motor vehicle exhaust and chronic

respiratory symptoms in children living near reeways.Environ Res 74(2):122–132.

Wang LY, Zhong Y, Wheeler L. 2005. Direct and indirect costs

o asthma in school-age children [Abstract]. Prev Chronic

Dis 2(1):A11.

Xue J, McCurdy T, Spengler J, Ozkaynak H. 2004. Understanding

variability in time spent in selected locations or 7–12-year

old children. J Expo Anal Environ Epidemiol 14(3):222–233.

Zhu Y, Hinds WC, Kim S, Sioutas C. 2002. Concentration and

size distribution o ultrafne particles near a major high-

way. J Air Waste Manag Assoc 52(9):1032–1042.

Documents

ehp-118-1021