UBC Faculty Research and Publications

A Cohort Study of Traffic-Related Air Pollution Impacts on Birth Outcomes Brauer, Michael; Lencar, Cornel; Tamburic, Lillian; Koehoorn, Mieke; Demers, Paul; Karr, Catherine May 31, 2008

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


52383-Brauer_M_et_al_Cohort_study_traffic_air_pollution.pdf [ 140.78kB ]
JSON: 52383-1.0074698.json
JSON-LD: 52383-1.0074698-ld.json
RDF/XML (Pretty): 52383-1.0074698-rdf.xml
RDF/JSON: 52383-1.0074698-rdf.json
Turtle: 52383-1.0074698-turtle.txt
N-Triples: 52383-1.0074698-rdf-ntriples.txt
Original Record: 52383-1.0074698-source.json
Full Text

Full Text

680 VOLUME 116 | NUMBER 5 | May 2008 • Environmental Health PerspectivesResearch | Children’s HealthNumerous studies have indicated associationsbetween exposure to ambient air pollutionand adverse pregnancy outcomes. Such asso-ciations, if determined to be causal, are likelyto result in significant public health impactsgiven the widespread exposure to air pollu-tion and the fact that low birth weight(LBW) or preterm births are subsequentlyassociated with long-term sequelae such asdevelopmental disability and chronic lungdisease (Cano et al. 2001; Dik et al. 2004).Determination of a causal relationshipbetween air pollution and adverse pregnancyoutcomes would have implications for burdenof disease measures and add to the impor-tance of strategies to mitigate the healtheffects of air pollution exposure.Previous studies have been reviewed indetail. Sˇrám et al. (2005) concluded that evi-dence is sufficient to support a causal associa-tion between ambient concentrations ofparticulate matter and LBW, but evidence ofeffects for other pollutants and for other out-comes such as preterm birth is less robust.Maisonet et al. (2004) concluded that studiesto date support small effects of air pollutionon preterm birth and small for gestational agebirth (SGA), but not full-term LBW. In asystematic review, Glinianaia et al. (2004)suggested that evidence of associations with airpollution and fetal growth or pregnancy dura-tion is limited and inconclusive and argued forpopulation-based cohort designs using high-quality individual exposure estimates. Thesereviews highlight the difficulties in interpret-ing an evidence base with differences amongmethods and with important limitations.First, most studies are either time-series stud-ies (Dugandzic et al. 2006; Liu et al. 2003,2007; Mannes et al. 2005; Sagiv et al. 2005)that relate relatively short-term changes in airpollution concentrations to temporal changesin rates of adverse pregnancy outcomes or, lessfrequently, cohort analyses that compare out-comes between locations with differing levelsof ambient air pollution (Salam et al. 2005)based on interpolated ambient monitoringnetwork data. Between-city comparisons aresubject to potential confounding becausecovariates may be highly correlated with airpollution, whereas time-series studies areproblematic to interpret because they relateshort-term changes in air pollution that aredriven primarily by meteorology to outcomes.They inherently assume that the impact of airpollution on birth outcomes is acute, requireknowledge of the relevant periods of pregnancyduring which air pollution may have impacts,and are subject to potential confounding byseasonally varying factors. As reviewed byGlinianaia et al. (2004), a number of studieshave suggested stronger relationships betweenbirth outcomes and exposure during specificperiods of pregnancy based on comparison ofstatistical effect sizes. However, results acrossstudies have not consistently identified specificperiods of exposure that are most closely linkedto adverse pregnancy outcomes. Increasingly, air pollution researchershave identified important spatial variability inair pollution concentrations within airsheds(Hoek et al. 2002b; Lewne et al. 2004; Zhanget al. 2004; Zhu et al. 2004). In many situa-tions these contrasts are of greater magnitudethan between-city or temporal contrasts(Jerrett et al. 2005). Such spatial contrasts,primarily related to measures of proximity totraffic corridors, have been associated with anumber of health impacts including mortality(Hoek et al. 2002a; Maynard et al. 2007;Miller et al. 2007; Nafstad et al. 2004;Roemer and van Wijnen 2001), asthma andrespiratory symptoms (Bayer-Oglesby et al.2006; Brauer et al. 2002, 2007; Gaudermanet al. 2005, 2007; McConnell et al. 2006;Ryan et al. 2005; Smargiassi et al. 2006), andotitis media (Brauer et al. 2006).Application of within-airshed spatial con-trasts in birth outcome studies are few (Leemet al. 2006; Parker et al. 2005; Ritz and Yu1999; Ritz et al. 2000; Slama et al. 2007;Wilhelm and Ritz 2003, 2005). These stud-ies, though provocative, have been limitedlargely to Southern California—a metropoli-tan area with relatively high levels of ambientair pollution. They relied on interpolatedambient monitoring data or simple roadAddress correspondence to M. Brauer, School ofEnvironmental Health, The University of BritishColumbia, 2206 East Mall, Vancouver BC V6T1Z3Canada. Telephone: (604) 822-9585. Fax: (604)822-9588. E-mail: brauer@interchange.ubc.caThe research was supported in part by HealthCanada via an agreement with the British ColumbiaCentre for Disease Control to the Border AirQuality Study. Additional support was provided bythe Center for Health and Environment Research atThe University of British Columbia, funded by theMichael Smith Foundation for Health Research.M.K. was supported in part by a Michael SmithFoundation for Health Research Scholar Award.The authors declare they have no competingfinancial interests.Received 4 October 2007; accepted 22 January 2008.A Cohort Study of Traffic-Related Air Pollution Impacts on Birth OutcomesMichael Brauer,1 Cornel Lencar,1 Lillian Tamburic,2 Mieke Koehoorn,1,3 Paul Demers,1,3 and Catherine Karr41School of Environmental Health, 2Centre for Health Services and Policy Research, and 3Department of Health Care and Epidemiology,The University of British Columbia, Vancouver, British Columbia, Canada; 4Department of Pediatrics, University of Washington, Seattle,Washington, USABACKGROUND: Evidence suggests that air pollution exposure adversely affects pregnancy outcomes.Few studies have examined individual-level intraurban exposure contrasts.OBJECTIVES: We evaluated the impacts of air pollution on small for gestational age (SGA) birthweight, low full-term birth weight (LBW), and preterm birth using spatiotemporal exposure metrics.METHODS: With linked administrative data, we identified 70,249 singleton births (1999–2002)with complete covariate data (sex, ethnicity, parity, birth month and year, income, education) andmaternal residential history in Vancouver, British Columbia, Canada. We estimated residentialexposures by month of pregnancy using nearest and inverse-distance weighting (IDW) of studyarea monitors [carbon monoxide, nitrogen dioxide, nitric oxide, ozone, sulfur dioxide, and partic-ulate matter < 2.5 (PM2.5) or < 10 (PM10) µm in aerodynamic diameter], temporally adjusted landuse regression (LUR) models (NO, NO2, PM2.5, black carbon), and proximity to major roads.Using logistic regression, we estimated the risk of mean (entire pregnancy, first and last month ofpregnancy, first and last 3 months) air pollution concentrations on SGA (< 10th percentile), termLBW (< 2,500 g), and preterm birth.RESULTS: Residence within 50 m of highways was associated with a 26% increase in SGA [95%confidence interval (CI), 1.07–1.49] and an 11% (95% CI, 1.01–1.23) increase in LBW.Exposure to all air pollutants except O3 was associated with SGA, with similar odds ratios (ORs)for LUR and monitoring estimates (e.g., LUR: OR = 1.02; 95% CI, 1.00–1.04; IDW: OR = 1.05;95% CI, 1.03–1.08 per 10-µg/m3 increase in NO). For preterm births, associations were observedwith PM2.5 for births < 37 weeks gestation (and for other pollutants at < 30 weeks). No consistentpatterns suggested exposure windows of greater relevance.CONCLUSION: Associations between traffic-related air pollution and birth outcomes were observedin a population-based cohort with relatively low ambient air pollution exposure.KEY WORDS: air pollution, birth weight, carbon black, carbon monoxide, nitrogen dioxide, nitricoxide, particulate matter, pregnancy, pregnancy outcome, preterm birth, soot, sulfur dioxide, vehi-cle emissions. Environ Health Perspect 116:680–686 (2008). doi:10.1289/ehp.10952 available viahttp://dx.doi.org/ [Online 23 January 2008]proximity measures rather than high-resolu-tion spatial contrasts in concentrations. Wesought to assess the relationship between repro-ductive outcomes and spatial and temporallyvarying levels of air pollution in the metropoli-tan area of Vancouver, British Columbia,Canada, a city with relatively low levels ofambient air pollution. We estimated exposuresat the individual level, for a population-basedcohort using both monitor-based methods andland use regression models based on proximityto traffic sources, land use, population density,and topographic features. Even in Vancouver,an area with a dense ambient monitoring net-work, exposure assessment based on regulatorymonitoring network data is more suited tocharacterizing temporal variability. Land useregression models, even those with temporalcomponents, as in this analysis, focus on high-resolution spatial variability in air pollutantconcentrations.The literature describing associationsbetween air pollution and birth outcomes hasfocused on clinically defined outcomes ofLBW and preterm birth, defined in a variety ofways, which complicates comparisons. Theunderlying biological processes—fetal growthrestriction and inadequate gestational length—are incompletely understood and imperfectlyrepresented in routinely available perinatalmeasurements available in vital statisticrecords. We elected to focus on SGA births asa primary outcome measure, because birthweight as a function of gestational age has adirect effect on perinatal morbidity and mor-tality (Pollack and Divon 1992).LBW may result from complex and multi-ple pathways of fetal growth restriction attrib-uted to maternal, fetal, or placental factors.Three broad categories of biological factors havebeen suggested to play a role in inadequate fetalgestation: abnormality of the biological clock,abnormal implantation, and infection andinflammation (Mattison et al. 2003).The current theories provide multiplesites at which environmental factors mayinfluence biological factors to modulate fetalgrowth and induce preterm birth. However,specific toxicologic mechanisms including rel-evant timing during gestational developmentare not known. We explored each of theseprocesses, fetal growth restriction and inade-quate gestational length, separately, andexplored the influences of exposure timing inearly and late pregnancy.MethodsCohort. The study area is the greater Vancouvermetropolitan region. We constructed thecohort by extracting data from a series of linkedadministrative data sets obtained from theBritish Columbia Ministry of Health, theBritish Columbia Vital Statistics Agency, andthe British Columbia Perinatal DatabaseRegistry. Health data are available through anapproved process (Chamberlayne et al. 1998)via the British Columbia Linked HealthDatabase for research purposes and governed bya data access agreement between the researchersand the data stewards. Medical services andhospitalization data were provided and gov-erned by the Ministry of Health, Governmentof British Columbia; and vital statistics data bythe British Columbia Vital Statistics Agency.These data were further merged through anadditional data access agreement with recordsin the provincial perinatal database governed bythe British Columbia Reproductive CareProgram. The study protocol was approved bythe Institutional Review Board (BehaviouralResearch Ethics Board) of The University ofBritish Columbia. Briefly, we identified vitalstatistics records for 92,158 children born inthe study area during a 4-year period(1999–2002); 77,342 had mothers with veri-fied complete residential history within thestudy area during the 9 months of pregnancy.Exclusions were made for 976 multiple births,8 children with no recorded birth weight orparity status, 2,998 with missing maternal age,1,691 with a missing native status (ethnicity)indicator, and 1,420 who were missing specificcensus covariates (neighborhood income,maternal education) of interest. From the77,342 with residential history in the studyarea, these exclusions left 70,249 subjects(90.8%) for analyses. Additional subjects wereexcluded from specific (monitoring network-based) analyses if measurements were not avail-able for suitably proximal monitors. Residential history. We compiled mater-nal residential histories from the beginning ofpregnancy until birth from postal codes andassociated dates recorded in provincial healthplan registry files, and from all hospital dis-charge and physician billing records for eachmother. Approximately 10% of the compiledrecords included invalid or nonresidential(including urban post office boxes) postalcodes and were excluded from the residentialhistory assignment. For each subject a longitu-dinal residential history was constructed fromthe remaining data. Where changes in postalcodes were observed during the follow-upperiod, transitions were set as the midpointbetween dates if nonoverlapping, or at the firstdate of the next postal code if overlapping.Administrative data sources. Birth weightand duration of gestation. We used vital statis-tics records to identify SGA births, defined asthose with birth weights below the 10th per-centile of the cohort, stratified by sex, for eachweek of gestation. LBW at term were thosewith at least 37 weeks gestation and birthweight < 2,500 g, and preterm births werethose with < 37 weeks duration of gestation, asindicated on the vital statistics birth records.Subgroup analyses were conducted for births of< 30, 30–34, and 35–37 weeks gestation andfor birth weights below the 5th percentile ofthe cohort for each week of gestation.Covariates. For each mother we also col-lected available data for several covariates ofinterest that may be associated with the healthoutcomes. Sex, parity, and the month and yearof birth were available from Vital Statisticsrecords. First Nations (“status Indians”) statusof the baby was obtained for each individualfrom hospital discharge records. Maternal ageand maternal smoking during pregnancy wasobtained from Perinatal Database Registry filesthat were linked to each individual birthrecord. Because no individual-level data wereavailable for income and maternal level of edu-cation, we assigned subjects to neighborhood-level income quintiles and maternal educationquartiles using Census data based on their resi-dence at the resolution of the census dissemi-nation area (DA). Dissemination areas are thesmallest geographic areas for which allCanadian Census data are disseminated andcorresponds to one or more neighboring blockswith target populations of 400–700 persons(Puderer 2001). Air pollution. Monitoring network data.Exposure to air pollution for each cohortmember was assigned by three differentapproaches, two based on the regulatory mon-itoring network and one based on dedicatedsampling campaigns. The regulatory monitor-ing network was operated by the BritishColumbia Ministry of Environment and MetroVancouver and includes daily measurements at24 monitors for ozone, 22 for nitric oxide/nitrogen dioxide, 14 for sulfur dioxide, 19 forcarbon monoxide, 19 for particulate matter< 10 µm in aerodynamic diameter (PM10), and7 for PM < 2.5 µm in aerodynamic diameter(PM2.5). The monitoring data were assigned toindividuals at their 6-digit postal code of resi-dence. The 6-digit postal code typically cor-responds to one block-face in urban areas; areasmay be considerably larger in rural areas withlow population density.Concentrations were assigned to postalcodes by nearest monitor and inverse-distanceweighting (IDW) approaches. This approachprovided high temporal resolution (dailymeasures for most days) with less precise spa-tial resolution than land use regression esti-mates. For the nearest monitor assignment,we assigned for each day a concentration fromthe operational monitor closest to the postalcode of interest and within 10 km. We thencomputed monthly averages for each individ-ual for the full duration of their pregnancy.For the IDW approach we used an inverse-distance (1/distance) weighted average of thethree closest monitors within 50 km to com-pute a monthly mean concentration. For bothapproaches, a month was considered missingif there was a gap of > 5 consecutive days inTraffic pollution and birth outcomesEnvironmental Health Perspectives • VOLUME 116 | NUMBER 5 | May 2008 681air monitoring data or if there were > 10missing days in a given month. Using themonthly averages, we then computed meanexposures for each mother for the full dura-tion of pregnancy, the first and last 30 days ofpregnancy, and the first and last 3 months ofpregnancy. Exposures were updated withchange in postal code of residence andweighted by time spent in multiple resi-dences. Postal code information for motherswas obtained from the provincial health regis-tration and health care contact records.Land use regression model. Exposureassessment based on a land use regression(LUR) model developed for the study region(Henderson et al. 2007) provided improvedlocal spatial resolution. Briefly, 116 passivesamplers to collect NO and NO2 weredeployed for two 14-day periods at 116 sitesin the study area. Mean concentrations dur-ing these two periods were highly correlatedwith and closely approximated annual aver-ages from regulatory monitoring networkdata. In addition, PM2.5 mass was measuredonce at a subset of 25 locations during a2-month sampling period. Integrated 1-weekaverage PM2.5 samples were collected onTeflon (Teflo; Pall Corp., East Hills, NY,USA) filters using Harvard Impactors (AirDiagnostics and Engineering, Harrison, ME,USA) at a flow rate of 4 L/min. Five samplingunits were rotated between the 25 sites andone was collocated with a tapered elementoscillating microbalance (TEOM; ThermoElectron Corp., East Greenbush, NY, USA)sampler at a regulatory monitoring networkstation. For a subset of 39 sites, we measuredshort-term levels of particle absorbance (blackcarbon) on one occasion using a particle sootabsorption photometer (Radiance Research,Seattle, WA, USA) in a mobile monitoringplatform. These measurements were adjustedfor temporal variation based on repeatedmeasurements at a centrally located site toresult in estimated annual average concentra-tions. We have previously demonstratedstrong correlations (R2 0.7–0.8) (Noullettet al. 2006; Rich 2003) between particleabsorbance and traditional measurements ofelemental carbon (Cyrys et al. 2003).For each of the 116 (and the subsets of 25and 39) measurement sites, 55 variables weregenerated in a geographic information system(GIS) (ArcGIS; ESRI, Redlands, CA, USA),and linear regression models of NO, NO2 andblack carbon were built with the most predic-tive covariates. For NO, the model had an R2 of0.62 and included the number of major roadswithin 100-m and 1,000-m radius circularbuffers of the measurement sites, the number ofsecondary roads within a 100-m buffer, thepopulation density within a 2,500-m radius,and elevation. For NO2, the model (R2 = 0.56)included the same variables as well as theamount of commercial land use within 750 m.For PM2.5 the model (R2 = 0.52) included theamount of commercial and industrial land usewithin 300 m, the amount of residential landuse within 750 m, and elevation. For black car-bon the model (R2 = 0.56) included the num-ber of secondary roads within a 100-m buffer,distance to the nearest highway, and theamount of industrial land use within 750 m.Evaluations, based on comparison to additionalmeasurements and cross-validation analysis,indicated that the PM2.5 and black carbonmodels performed much more poorly than didthe NO or NO2 models. Using the LUR mod-els, smooth spatial surfaces of predicted (annualaverage) concentrations were generated for theentire study area at a resolution of 10 m. Thesurfaces were then smoothed (Focal Statistics,ArcGis Spatial Analyst; ArcGIS) to removeabrupt changes and edge effects so as to moreaccurately reflect the measured effect of proxim-ity to roadways (Gilbert et al. 2003). For eachLUR model, the corresponding monitoringnetwork data for each pollutant were fit with amonthly dummy variables and a covariate forlinear trend (Times Series Forecasting System,version 9; SAS Institute Inc., Cary, NC, USA).For black carbon, the PM2.5 trend was usedbecause there were no corresponding regulatorymonitoring network data. From these models,we applied month–year adjustment factors toeach surface to estimate monthly average con-centrations. Using these, we then computedindividual subject exposures for the sameexposure windows as described above for themonitor-based approaches.Road proximity. Finally, we calculatedroad proximity for home postal codes of allcohort members. Road classifications (DMTIArcView street file data set for BritishColumbia, Canmap Streetfiles, version 2006.3;DMTI Spatial, Markham, Ontario, Canada)were used to determine whether a home postalcode was within 50 or 150 m of an expresswayor primary highway, within 50 or 150 m of asecondary highway or major road/arterial road,or within 150 m of a secondary highway ormajor road or within 50 m of an expressway orprimary highway.ResultsOf the 70,249 live births considered in analy-sis, 36,138 (51.4%) were male and 542(0.77%) had First Nations status indicated.Mean parity was 1.8 and the mean maternalage was 31.1 years. A total of 7.4% (5,198) ofthe cohort reported maternal smoking duringpregnancy. The mobility of the cohort asascertained from residential histories indicatedthat 60.7% (42,649) had a single address dur-ing pregnancy, whereas 34.9% (24,537) hadtwo addresses, and 4.4% (3,063) had three tofive addresses. 89.2% of the postal codes werereferenced to a block face, and an additional8.5% referenced to a block.As expected for the greater Vancouver area,air pollution concentrations were low relativeto air quality standards and internationalguidelines (Table 1). Mean concentrations andranges at the middle of the concentration dis-tributions were similar for the monitor-basedand the LUR estimates. As expected, the LURestimates had smaller minimum and greatermaximum values than the IDW estimates.Brauer et al.682 VOLUME 116 | NUMBER 5 | May 2008 • Environmental Health PerspectivesTable 2. Prevalence of SGA, LBW, and pretermbirth in the cohort (n = 70,249). Outcome No. (%)SGA (< 10th percentile) 6,939 (10.4)LBW (< 2,500 g at ≥ 37 weeksa) 894 (1.3)Gestation < 37 weeks 3,748 (5.3)Gestation < 35 weeks 820 (1.2)Gestation < 30 weeks 170 (0.2)an = 66,501 births at full (≥ 37 weeks) term. Table 1. Distributions of air pollution concentrations estimated by different approaches.Pollutant Model Mean Min Max IQRNO Nearest 23.7 0.5 93.3 14.3NO IDW 22.1 2.9 63.6 14.9NO LUR 30.7 1.4 145.5 14.4NO2 Nearest 34.4 9.7 63.8 9.3NO2 IDW 32.5 15.3 53.6 11.3NO2 LUR 31.6 0.0 63.8 9.3CO Nearest 633.1 124.3 1,409.0 194.7CO IDW 613.8 305.4 1,062.3 172.1PM2.5 Nearest 5.3 0.3 37.0 1.0PM2.5 IDW 5.1 1.0 7.6 1.1PM2.5 LUR 4.0 0.0 11.3 1.5BC LUR 1.6 0.0 6.2 1.2PM10 Nearest 12.7 5.6 35.4 1.6PM10 IDW 12.5 8.4 16.6 1.4SO2 Nearest 5.7 0.0 24.9 3.5SO2 IDW 5.3 0.3 17.8 3.0O3 Nearest 28.0 2.3 69.2 8.2O3 IDW 28.3 10.4 48.2 8.4Abbreviations: BC, black carbon; IQR, interquartile range; Max, maximum; Min, minimum. Weighted concentrations of 3monitors within 50 km of residential postal code. Nearest: concentration from nearest monitor within 10 km of residentialpostal code. All concentrations were computed for full duration of pregnancy. All concentrations in µg/m3 except for BC,which is in 10–5 particles/m.Correlations between IDW estimates ofCO, NO, NO2, and SO2 were all > 0.8 andnegatively correlated with O3 (r = –0.7 to–0.8). Correlations between LUR estimatesCO, NO, NO2, SO2 and PM10 or PM2.5were lower (r = 0.1 to 0.5). Correlations forthe same pollutant between IDW and LURestimates were moderate (r = 0.55 for NO,0.37 for NO2) except for the LUR PM2.5 esti-mates, which were not correlated with IDWestimates of either PM2.5 or PM10. IDW andnearest monitor estimates for the same pollu-tant were all highly correlated (r = 0.70 to0.89). Given these high correlations, andbecause the nearest monitor exposure estimatesdisplayed sharp spatial patterns in exposurerelated to monitor location (Marshall et al.2008), we focused on the IDW estimates.Exposures estimated for various periods ofpregnancy were, in general, highly correlatedwith the full term of pregnancy average expo-sure, limiting our ability to assess the impact ofexposure during specific windows of pregnancy.Correlations between exposure during the firstor last 3 months of pregnancy and the full termof pregnancy were 0.63–0.89 for IDW esti-mates of gaseous pollutants, 0.48–0.64 forIDW estimates of PM, and 0.59–0.94 for LURestimates.Table 2 presents the prevalence of SGA,LBW at term, and preterm births in thecohort. As reported by others (Ghosh et al.2007), we observed higher rates of SGA forfemales (104.9 vs. 103.6 per 1,000 live births)and higher rates of preterm birth for males(56.9 vs. 49.6 per 1,000 live births).For NO, NO2, CO, PM10, PM2.5, andblack carbon we observed small but consistentincreased risks of SGA (Table 3). Odds ratios(ORs) for exposure estimates based on thenearest monitor approach were similar to thosefor the inverse distance weighted estimates[e.g., for NO, OR = 1.03; 95% confidenceinterval (CI), 1.01–1.05 for nearest, vs. OR =1.05; 95% CI, 1.03–1.08 for IDW] and aretherefore not presented further in the analysis.Because O3 was highly negatively correlatedwith all of the primary traffic-related air pollu-tants (r = –0.83 for IDW CO; r = –0.86 forIDW NO), associations were largely protective(e.g., OR = 0.89; 95% CI, 0.84–0.94 for IDWSGA) and are not presented further.Although our prior expectations were thatLUR exposure assessments would provideincreased precision and variability of individ-ual exposure, associations were not consis-tently larger for the LUR estimates, althoughtheir CIs were smaller in most cases. Similarresults (not shown) were found for the subsetof SGA births with birth weights below the5th percentile of the cohort for each week ofgestation. We did not observe (results notshown) any consistent patterns indicatinglarger effect estimates for exposures duringspecific periods of pregnancy. ORs for early(first month, first 3 months) or late pregnancy(last month, last 3 months) exposure windowswere remarkably similar to those for thefull duration of pregnancy. Similar patternswere observed for LBW (Table 4), althoughassociations, except for IDW NO2, did notreach statistical significance because of thesmaller number of LBW cases.Tables 5 and 6 present the analysis of thesimple road proximity measures. In contrastto the LUR models that indicated a smallTraffic pollution and birth outcomesEnvironmental Health Perspectives • VOLUME 116 | NUMBER 5 | May 2008 683Table 3. Crude and adjusted ORsa for SGA.Exposure (entire pregnancy) SGA cases (n) Crude OR (95% CI) Adjustedb OR (95% CI)NO–IDW (10 µg/m3) 6,351 1.03 (1.00–1.05) 1.05 (1.03–1.08)NO–LUR (10 µg/m3) 6,750 1.02 (1.01–1.04) 1.02 (1.00–1.04)NO2–IDW (10 µg/m3) 6,351 1.12 (1.08–1.16) 1.14 (1.09–1.18)NO2–LUR (10 µg/m3) 6,750 1.02 (0.99–1.05) 0.99 (0.96–1.02)CO–IDW (100 µg/m3) 6,351 1.03 (1.01–1.05) 1.06 (1.03–1.08)PM2.5–IDW (1 µg/m3) 6,307 0.99 (0.97–1.01) 1.02 (0.98–1.05)PM2.5–LUR (1µg/m3) 6,750 1.03 (1.02–1.05) 1.02 (1.00–1.03)BC–LUR (10–5/m) 6,750 1.03 (1.01–1.05) 1.01 (0.99–1.03)PM10–IDW (1 µg/m3) 6,351 1.02 (1.00–1.05) 1.02 (0.99–1.05)SO2–IDW (1 µg/m3) 6,351 1.01 (1.00–1.02) 1.01 (1.00–1.02)BC, black carbon. aORs per standardized increases (indicated in parentheses), roughly corresponding to interquartile range. bAdjusted forinfant sex, First Nations status, parity, maternal age, maternal smoking during pregnancy, month–year of birth, income(quintile-census), maternal education (quartile-census).Table 4. Crude and adjusted ORsa for LBW at full (≥ 37 weeks) term.Exposure (entire pregnancy) LBW cases (n) Crude OR (95% CI) Adjustedb OR (95% CI)NO–IDW (10 µg/m3) 894 1.02 (0.96–1.09) 1.03 (0.96–1.10)NO–LUR (10 µg/m3) 874 1.02 (0.98–1.08) 1.01 (0.96–1.07)NO2–IDW (10 µg/m3) 894 1.12 (1.02–1.23) 1.11 (1.01–1.23)NO2–LUR (10 µg/m3) 874 1.01 (0.94–1.09) 0.97 (0.89–1.05)CO–IDW (100 µg/m3) 894 1.02 (0.96–1.07) 1.02 (0.96–1.09)PM2.5–IDW (1 µg/m3) 889 0.98 (0.92–1.06) 0.98 (0.92–1.05)PM2.5–LUR (1µg/m3) 874 1.05 (1.02–1.09) 1.03 (0.99–1.07)BC–LUR (10–5/m) 874 1.03 (0.97–1.09) 1.00 (0.95–1.07)PM10–IDW (1 µg/m3) 894 1.01 (0.95–1.08) 1.01 (0.95–1.08)SO2–IDW (1 µg/m3) 894 1.00 (0.97–1.02) 0.99 (0.97–1.02)BC, black carbon. aORs per standardized increases (indicated in parentheses), roughly corresponding to interquartile range. bAdjusted forinfant sex, First Nations status, parity, maternal age, maternal smoking during pregnancy, month–year of birth, income(quintile-census), maternal education (quartile-census).Table 5. Crude and adjusted ORs for SGA and road proximity.No. of SGA cases Measure (% of total subjects) Crude OR (95% CI) Adjusteda OR (95% CI)< 150 m highwayb or 1,218 (1.73) 1.01 (0.95–1.08) 0.99 (0.92–1.06)< 50 m major roadc< 50 m highway 176 (0.25) 1.31 (1.12–1.54) 1.26 (1.07–1.49)< 150 m highway 413 (0.59) 0.96 (0.86–1.07) 0.93 (0.83–1.03)< 50 m major road 842 (1.20) 1.04 (0.97–1.13) 1.03 (0.95–1.12)< 150 m major road 1,611 (2.29) 1.05 (0.99–1.12) 1.04 (0.98–1.11)aAdjusted for infant sex, First Nations status, parity, maternal age, maternal smoking during pregnancy, month–year ofbirth, income (quintile-census), maternal education (quartile-census). bDMTI type 1 and 2 road (expressway: 114,000 vehi-cles/day; principal highway: 21,000 vehicles/day). cDMTI type 3 and 4 road (secondary highway: 18,000 vehicles/day;major road: 15,000 vehicles/day).Table 6. Crude and adjusted odds ratios for LBW at full term and road proximity.No. of LBW casesMeasure (% of total subjects) Crude OR (95% CI) Adjusteda OR (95% CI)< 150 m highwayb or 358 (0.51) 0.98 (0.82–1.17) 0.95 (0.79–1.13)< 50 m major roadc< 50 m highway 49 (0.07) 1.32 (0.87–2.00) 1.22 (0.81–1.87)< 150 m highway 124 (0.18) 1.06 (0.80–1.40) 1.01 (0.76–1.33)< 50 m major road 250 (0.36) 0.97 (0.79–1.20) 0.96 (0.77–1.18)< 150 m major road 476 (0.68) 0.95 (0.81–1.12) 0.94 (0.79–1.10)aAdjusted for infant sex, First Nations status, parity, maternal age, maternal smoking during pregnancy, month–year ofbirth, income (quintile-census), maternal education (quartile-census). bDMTI type 1 and 2 road (expressway: 114,000 vehi-cles/day; principal highway: 21,000 vehicles/day). cDMTI type 3 and 4 road (secondary highway: 18,000 vehicles/day;major road: 15,000 vehicles/day).magnitude increased risk of SGA and LBW inrelation to traffic-related air pollutants, a strongassociation was observed for mothers whoresided within 50 m of an expressway or high-way. Although the number of subjects meetingthis criterion was small (0.25% of all births forSGA), there was a 26% increased risk of anSGA birth compared with those mothers resid-ing > 50 m from an expressway or highway(mean of > 21,000 vehicles per day). ForLBW, we observed an 11% increase in risk forthis group. No increased risk was observed forthose living within 150 m of a highway orwithin 50 m of a major road (mean of15,000–18,000 vehicles per day). The LURmodels are derived in part from the density ofroads of within specific radii, but also includeland use and population density variables.Simple road proximity measures were notstrongly predictive of measured concentrationsin the LUR models (Henderson et al. 2007).For the preterm birth outcome of < 37weeks, we did not observe any consistent asso-ciations with any of the pregnancy averageexposure metrics except for PM2.5 (IDW: OR= 1.06; 95% CI, 1.01–1.11). In addition,pregnancy average PM2.5 exposure was relatedto preterm births < 35 (IDW: OR = 1.12;95% CI, 1.02–1.24) and < 30 weeks (IDW:OR = 1.13; 95% CI, 0.92–1.39). For theoutcome of birth < 30 weeks, although therewere very few cases, we found associationswith the different pregnancy average exposuremetrics that were similar to those observed forSGA (Table 7). No associations wereobserved between the simple road proximitymeasures and preterm birth < 37 weeks, andthere were no cases of births < 30 weeks thatwere within 50 m of a highway. In addition,there were no consistent trends for early orlate gestational period exposure to be morestrongly associated with preterm births.DiscussionWe explored several outstanding questions inair pollution reproductive epidemiology in apopulation-based study using individual esti-mates of exposure to air pollutants. Weobserved small-magnitude associations betweena number of traffic-related pollutants and SGAbirth weight. Consistent associations were alsoobserved between PM2.5 and preterm births(< 37 weeks). For other pollutants, associationswith gestational duration were observed onlyfor the small subset of births at < 30 weeks ges-tation. Given high correlations between NO,NO2, CO, and SO2, it was not possible to dif-ferentiate impacts of specific pollutants.Associations with SGA were, however, gener-ally of lower magnitude for PM10 and PM2.5,and we did not detect any associations withblack carbon, although our black carbonmodel performed relatively poorly in evalua-tion comparisons and was limited by a smallnumber of measurements and seasonal adjust-ment by PM2.5 rather than black carbontrends. Our results suggest a general associationbetween pollutants dominated by trafficsources (NOx and CO) and LBW in thisstudy. This conclusion is supported by ourfinding of a strong association between resi-dence within 50 m of a highway or expressway,but not other measures of traffic proximity,and both SGA and LBW at full term.Overall, results from the traffic-based landuse regression models agree with analyses basedon exposures estimated from the ambientmonitoring network. Although effect estimatesbased on LUR models did have somewhatsmaller CIs than those based on monitoringnetwork data, they did not indicate increasedestimates of effect, as one might expect givenimproved spatial resolution of the LUR modelsand the potential for reduced exposure misclas-sification. It is possible that LUR exposure esti-mates may be more appropriate for primarypollutants, such as NO and black carbon, thatvary the most spatially, whereas monitor-basedestimates are more appropriate for secondarypollutants such as NO2 and PM2.5 that displayless spatial heterogeneity. In contrast to theimproved spatial resolution of land use regres-sion-based exposure estimates, exposures deter-mined from ambient monitors are directlyrelated to a larger number of measurements,include more precise temporal information,and capture a different spatial scale of variabil-ity in ambient air pollution (Marshall et al.2008). Exposure estimates from land useregression and monitoring network data forthe same pollutant were only moderately corre-lated and appeared to be somewhat indepen-dent, with each capturing different aspects ofspatiotemporal variability in exposure(Marshall et al. 2008).Because exposures were estimated only forhome addresses, it is also possible that subjectmobility was related to varying degrees ofexposure misclassification for the differentmodeling approaches. An evaluation of theLUR and IDW models for (short-term) meas-ured personal exposures of pregnant womenindicated that for NO and NO2, especially forthose who were the least mobile, LUR modelswere a stronger predictor of personal exposureand better explained between-subject variabil-ity in exposure. Monitor-based estimates betterexplained within-subject (temporal) variabilityin exposure. For PM2.5 and black carbon,monitor-based estimates (of PM2.5) were morehighly correlated with personal exposures thanwere the LUR models (Nethery et al. 2007),Overall, these differences between themonitor-based and LUR model exposure esti-mates suggest that our finding of risks associ-ated with both types of exposure estimatesindicate some independence of these risks. Inanalysis of air pollution and childhood lungfunction, independent effects from bothregional and local traffic-related air pollutionhave been reported (Gauderman et al, 2007).Although simple measures of road prox-imity have been shown to predict substan-tially smaller percentages of variability inmeasured concentrations of NOx, PM2.5, andblack carbon compared with LUR models(Brauer et al. 2003; Henderson et al. 2007),road distance measures are straightforward,precise, directly relevant to land use policy,and easy to assess and apply in areas withouthigh monitor density. In particular, it is possi-ble that the LUR model estimates, thoughmore specific measures of individual air pollu-tants, may incorporate more exposure mis-classification to the harmful components oftraffic than do the measures of road proxim-ity. We found increased risks of SGA andLBW only for those subjects closest to roadsof highest traffic intensity in our study area.Our findings of associations between airpollution and SGA are consistent with reportsfrom a number of studies (Dejmek et al.1999; Hansen et al. 2007; Liu et al. 2003,2007; Parker et al. 2005) and conclusions ofreviews (Maisonet et al. 2004; Sˇrám et al.2005), including studies that focus on spatialcontrasts in exposure. For example, studies inSouthern California (Ritz and Yu 1999;Wilhelm and Ritz 2005), which has a relativelydense ambient monitoring network, foundBrauer et al.684 VOLUME 116 | NUMBER 5 | May 2008 • Environmental Health PerspectivesTable 7. Crude and adjusted ORsa for preterm births < 30 weeks.Exposure (entire pregnancy) Cases (n) Crude OR (95% CI) Adjusted b OR (95% CI)NO–IDW (10 µg/m3) 170 1.19 (1.04–1.35) 1.26 (1.08–1.47)NO–LUR (10 µg/m3) 166 1.08 (0.97–1.19) 1.05 (0.94–1.18)NO2–IDW (10 µg/m3) 170 1.13 (0.92–1.39) 1.12 (0.89–1.40)NO2–LUR (10 µg/m3) 166 1.11 (0.95–1.31) 1.08 (0.91–1.29)CO–IDW (100 µg/m3) 170 1.13 (1.01–1.28) 1.16 (1.01–1.33)PM2.5–IDW (1 µg/m3) 168 0.99 (0.86–1.14) 1.13 (0.92–1.39)PM2.5–LUR (1µg/m3) 166 1.08 (1.00–1.18) 1.07 (0.98–1.16)BC–LUR (10–5/m) 166 1.01 (0.89–1.15) 0.99 (0.87–1.13)PM10–IDW (1 µg/m3) 170 1.11 (0.97–1.29) 1.13 (0.95–1.35)SO2–IDW (1 µg/m3) 170 1.02 (0.96–1.08) 1.02 (0.96–1.09)BC, black carbon. aORs per standardized increases (in parentheses), roughly corresponding to interquartile range. bAdjusted for infant sex,First Nations status, parity, maternal age, maternal smoking during pregnancy, month–year of birth, income (quintile-cen-sus), maternal education (quartile-census).that exposure to higher levels of ambient CO,based on the nearest monitoring stationswithin 2 miles of the mother’s home address,during the last trimester was associated withincreased risk of LBW (OR = 1.22; 95% CI,1.03–1.44 for 3-month average exposures> 5.5 ppm). Similar effects were also observedfor PM10, with stronger associations found forsubjects living within 1 mile of a monitoringstation. Most recently, Slama and colleagues(2007) applied LUR models for PM2.5, NO2,and black carbon to a smaller cohort of 1,016full-term births (≥ 37 weeks) > 2,500 g inMunich. An increased prevalence ratio of 1.13(95% CI, 1.00–1.29) for birth weight < 3,000g at term was associated with a 1-µg/m3increase in PM2.5 concentration, with associa-tions also observed for increased concentrationsof black carbon or NO2.For preterm births, our study, even withits large population, had limited numbers ofcases with which to detect a relationship withair pollution. We found consistent associa-tions with PM2.5 but not other pollutants forbirths < 37, 35, or 30 weeks. For risk of verypreterm birth (< 30 weeks), we observed ele-vated ORs for a larger number of pollutants(NO, NO2, CO, PM10, and PM2.5 but notSO2). As with SGA, we did not observe anyassociation with black carbon. In addition tothe poor performance of our black carbonLUR model in evaluation tests, the applicationof this model was further limited by our needto use PM2.5 monitoring data, as opposed toblack carbon measurements, for seasonaladjustment.A general limitation for our modelingapproach for analysis of preterm births arisesfrom the fact that entire pregnancy exposureaveraging periods will, by definition, be differ-ent for cases and noncases. Further, the last3 months and last month of pregnancy occur atdifferent gestational periods for cases and non-cases. The importance of these differences inexposure is less significant in our data, becauseair pollution exposures for different pregnancyperiods were rather highly correlated.Measures of road proximity, which do notsuffer from this potential problem, were notassociated with preterm births. This analysis,however, was limited by the low numbers ofmothers with preterm births who resided< 50 m of a highway (for example, no births< 30 weeks).In this analysis we did not identify specificexposure windows (early or late pregnancy) ofgreater or lesser relevance for SGA or pretermbirth, although exposures from different periodsof pregnancy were highly correlated. Reportsfrom other studies have been inconsistent withregard to the importance of early or later preg-nancy exposures. The recent study of Slamaand colleagues (2007) showed a tendency forthird-trimester exposures to exhibit strongerassociations with LBW, but these associationswere also highly correlated with exposures aver-aged over the full pregnancy. Multiple potentialmechanisms by which air pollution may affectfetal growth and birth weight have been pro-posed, but there is little information regardingspecific etiology on which greater importance ofspecific exposure windows can be based.This study had several advantages overprevious analyses of the relationship betweenair pollution and birth outcomes. First, thestudy was population based and estimatedexposures at the individual level using bothinterpolated ambient monitoring networkdata and temporally adjusted LUR models tocharacterize spatio-temporal variability inexposure to a greater degree than in previousstudies. A recent exception is the study fromMunich, which also applied temporallyadjusted LUR models, although to a smallerand more restrictive cohort (Slama et al.2007). In addition, we accounted for residen-tial mobility during pregnancy when assign-ing exposures. Failure to account for mobility,especially in studies such as ours using verysmall areas to assign exposures, is likely toresult in misclassified exposures (Fell et al.2004). In our study we identified 35% of thepopulation as changing a residence duringpregnancy, highlighting the importance ofaccounting for mobility in residence-basedassessment of exposure. Mobility rates amongpregnant women reported in the literaturerange from 12% (Fell et al. 2004) to 33%(Canfield et al. 2006), although most studieswere small population case–control studies.Further, by a unique linkage of multiple data-bases, we acquired individual information onseveral important covariates.Largely because of the use of administra-tive databases, this study does have severalimportant limitations. We defined fetalgrowth restriction as a weight below the 10thpercentile for gestational age. This definitionis imperfect and controversial because it doesnot make a distinction among fetuses who areconstitutionally small, growth restricted andsmall, and growth restricted but not small.The data elements in our administrative dataset did not allow us to make these distinctions(e.g., we did not have data on parental size orbirth length). The definition used, however,represents a common definition of fetalgrowth restriction and increases the opportu-nity for comparisons among studies.We were also limited by unavailability ofindividual data for indicators of socioeconomicstatus. In this case we had to rely on neighbor-hood-level census data, where there is a likeli-hood of misclassification at the individual level.In addition, no information was available onmaternal ethnicity, except First Nations status,as ethnicity is known to affect birth weight dis-tributions (Janssen et al. 2007). We also hadno information regarding nutrition or prenatalcare. Further, although unlike most previousstudies we did account for residential mobilityin assigning exposures, our ascertainment ofmobility was imperfect because we had noinformation on the actual moving date; mov-ing was assumed from consistent changes inaddress information collected during healthcare system contacts. This limitation may haveaffected our sensitivity to detect more acuteeffects of air pollution—for example, impactson preterm birth. In addition, exposures werebased only on residential address and did notconsider time weighting by other importantmicroenvironments, which more closelyapproximates measured exposures (Netheryet al. 2007).ConclusionsOur findings in a population-based study addto an expanding literature that links severaltraffic-derived air pollutants (e.g., NO, NO2,CO) to adverse birth outcomes, particularlyincreased risk of SGA birth weight. In addi-tion, we observed consistent associationsbetween PM2.5 exposure and risk of pretermbirth. The overall importance of proximity tohighways, even in an area with relatively lowlevels of ambient air pollutants, suggestsopportunities for urban planning approachesto prevention. The toxicologically relevantcomponents of traffic-related air pollutionremain unclear and deserve increased attentionin future studies.REFERENCESBayer-Oglesby L, Schindler C, Hazenkamp-von Arx ME, Braun-Fahrlander C, Keidel D, Rapp R, et al. 2006. Living nearmain streets and respiratory symptoms in adults: theSwiss cohort study on air pollution and lung diseases inadults. Am J Epidemiol 164(12):1190–1198.Brauer M, Gehring U, Brunekreef B, de Jongste J, Gerritsen J,Rovers M, et al. 2006. Traffic-related air pollution and otitismedia. Environ Health Perspect 114:1414–1418.Brauer M, Hoek G, Smit HA, de Jongste JC, Gerritsen J,Postma DS, et al. 2007. Air pollution and development ofasthma, allergy and infections in a birth cohort. EurRespir J 29(5):879–888.Brauer M, Hoek G, van Vliet P, Meliefste K, Fischer P, GehringU, et al. 2003. Estimating long-term average particulate airpollution concentrations: Application of traffic indicatorsand geographic information systems. Epidemiology14(2):228–239.Brauer M, Hoek G, Van Vliet P, Meliefste K, Fischer PH, WijgaA, et al. 2002. Air pollution from traffic and the develop-ment of respiratory infections and asthmatic and allergicsymptoms in children. Am J Respir Crit Care Med166(8):1092–1098.Canfield MA, Ramadhani TA, Langlois PH, Waller DK. 2006.Residential mobility patterns and exposure misclassifica-tion in epidemiologic studies of birth defects. J Expo SciEnviron Epidemiol 16(6):538–543.Cano A, Fons F, Brines J. 2001. The effects on offspring of pre-mature parturition. Hum Reprod Update 7(5):487–494.Chamberlayne R, Green B, Barer ML, Hertzman C, LawrenceJW, Sheps SB. 1998. Creating a population-based linkedhealth database: a new resource for health servicesresearch. Can J Public Health 89:270–273.Cyrys J, Heinrich J, Hoek G, Meliefste K, Lewne M, Gehring U,et al. 2003. Comparison between different traffic-relatedparticle indicators: Elemental carbon (EC), PM2.5 mass,Traffic pollution and birth outcomesEnvironmental Health Perspectives • VOLUME 116 | NUMBER 5 | May 2008 685Brauer et al.686 VOLUME 116 | NUMBER 5 | May 2008 • Environmental Health Perspectivesand absorbance. J Expo Anal Environ Epidemiol 13(2):134–143.Dejmek J, Selevan SG, Benesˇ I, Solansky I, Sˇrám RJ. 1999. Fetalgrowth and maternal exposure to particulate matter dur-ing pregnancy. Environ Health Perspect 107:475–480.Dik N, Tate RB, Manfreda J, Anthonisen NR. 2004. Risk ofphysician-diagnosed asthma in the first 6 years of life.Chest 126(4):1147–1153.Dugandzic R, Dodds L, Stieb D, Smith-Doiron M. 2006. Theassociation between low level exposures to ambient airpollution and term low birth weight: a retrospective cohortstudy. Environ Health 5:3.Fell DB, Dodds L, King WD. 2004. Residential mobility duringpregnancy. Paediatr Perinat Epidemiol 18(6):408–414.Gauderman WJ, Avol E, Lurmann F, Kuenzli N, Gilliland F,Peters J, et al. 2005. Childhood asthma and exposure totraffic and nitrogen dioxide. Epidemiology 16(6):737–743.Gauderman WJ, Vora H, McConnell R, Berhane K, Gilliland F,Thomas D, et al. 2007. Effect of exposure to traffic on lungdevelopment from 10 to 18 years of age: a cohort study.Lancet 369(9561):571–577.Ghosh R, Rankin J, Pless-Mulloli T, Glinianaia S. 2007. Does theeffect of air pollution on pregnancy outcomes differ bygender? A systematic review. Environ Res 105(3):400–408.Gilbert NL, Woodhouse S, Stieb DM, Brook JR. 2003. Ambientnitrogen dioxide and distance from a major highway. SciTotal Environ 312(1-3):43–46.Glinianaia SV, Rankin J, Bell R, Pless-Mulloli T, Howel D. 2004.Particulate air pollution and fetal health: a systematicreview of the epidemiologic evidence. Epidemiology15(1):36–45.Hansen C, Neller A, Williams G, Simpson R. 2007. Low levels ofambient air pollution during pregnancy and fetal growthamong term neonates in Brisbane, Australia. Environ Res103(3):383–389.Henderson SB, Beckerman B, Jerrett M, Brauer M. 2007.Application of land use regression to estimate long-termconcentrations of traffic-related nitrogen oxides and fineparticulate matter. Environ Sci Technol 41(7):2422–2428.Hoek G, Brunekreef B, Goldbohm S, Fischer P, van den BrandtPA. 2002a. Association between mortality and indicatorsof traffic-related air pollution in the Netherlands: a cohortstudy. Lancet 360(9341):1203–1209.Hoek G, Meliefste K, Cyrys J, Lewne M, Bellander T, Brauer M,et al. 2002b. Spatial variability of fine particle concentra-tions in three European areas. Atmos Environ36(25):4077–4088.Janssen PA, Thiessen P, Klein MC, Whitfield M, MacNab YC,Cullis-Kuhl S. 2007. Standards for the measurement ofbirth weight, length and head circumference at term inneonates of European, Chinese and South Asian ancestry.Open Medicine 1(2):e74–88.Jerrett M, Arain A, Kanaroglou P, Beckerman B, Potoglou D,Sahsuvaroglu T, et al. 2005. A review and evaluation ofintraurban air pollution exposure models. J Expo AnalEnviron Epidemiol 15(2):185–204.Leem JH, Kaplan BM, Shim YK, Pohl HR, Gotway CA, Bullard SM,et al. 2006. Exposures to air pollutants during pregnancyand preterm delivery. Environ Health Perspect 114:905–910.Lewne M, Cyrys J, Meliefste K, Hoek G, Brauer M, Fischer P,et al. 2004. Spatial variation in nitrogen dioxide in threeEuropean areas. Sci Total Environ 332(1-3):217–230.Liu S, Krewski D, Shi Y, Chen Y, Burnett RT. 2003. Associationbetween gaseous ambient air pollutants and adversepregnancy outcomes in Vancouver, Canada. EnvironHealth Perspect 111:1773–1778.Liu S, Krewski D, Shi Y, Chen Y, Burnett RT. 2007. Associationbetween maternal exposure to ambient air pollutants dur-ing pregnancy and fetal growth restriction. J Expo SciEnviron Epidemiol 17(5):426–432.Maisonet M, Correa A, Misra D, Jaakkola JJ. 2004. A review ofthe literature on the effects of ambient air pollution onfetal growth. Environ Res 95(1):106–115.Mannes T, Jalaludin B, Morgan G, Lincoln D, Sheppeard V,Corbett S. 2005. Impact of ambient air pollution on birthweight in Sydney, Australia. Occup Environ Med62(8):524–530.Marshall JD, Nethery EN, Brauer M. 2008. Within-urban vari-ability in ambient air pollution: comparison of estimationmethods. Atmos Environ 42(6):1359–1369. Mattison DR, Wilson S, Coussens C, Gilbert D (eds). 2003. TheRole of Environmental Hazards in Premature Birth:Workshop Summary. Institute of Medicine. Roundtable onEnvironmental Health Sciences, Research, and Medicine.Washington, DC:National Academies Press.Maynard D, Coull BA, Gryparis A, Schwartz J. 2007. Mortalityrisk associated with short-term exposure to traffic parti-cles and sulfates. Environ Health Perspect 115:751–755.McConnell R, Berhane K, Yao L, Jerrett M, Lurmann F, GillilandF, et al. 2006. Traffic, susceptibility, and childhood asthma.Environ Health Perspect 114:766–772.Miller KA, Siscovick DS, Sheppard L, Shepherd K, Sullivan JH,Anderson GL, et al. 2007. Long-term exposure to air pollu-tion and incidence of cardiovascular events in women.N Engl J Med 356(5):447–458.Nafstad P, Haheim LL, Wisloff T, Gram F, Oftedal B, Holme I,et al. 2004. Urban air pollution and mortality in a cohort ofNorwegian men. Environ Health Perspect 112:610–615.Nethery E, Leckie S, Teschke K, Brauer M. 2007. From meas-ures to models: an evaluation of air pollution exposureassessment for epidemiologic studies of pregnant women.Occup Environ Med; doi:10.1136/oem.2007.035337 [Online10 December 2007].Noullett M, Jackson PL, Brauer M. 2006. Winter measurementsof children’s personal exposure and ambient fine particlemass, sulphate and light absorbing components in a north-ern community. Atmos Environ 40(11):1971–1990.Parker JD, Woodruff TJ, Basu R, Schoendorf KC. 2005. Air pol-lution and birth weight among term infants in California.Pediatrics 115(1):121–128.Pollack RN, Divon MW. 1992. Intrauterine growth retardation:definition, classification, etiology. Clin Obstet Gynecol35(1):99–107.Puderer H. 2001. Introducing the Dissemination Area for the2001 Census: An Update. Georgraphy Working PaperSeries No. 2000-4. Ottawa:Statistics Canada.Rich K. 2003. Air Pollution and Patients with Implanted CardiacDefibrillators: An Epidemiological Analysis and Assessmentof Exposure [MSc Thesis]. Vancouver, BC, Canada:TheUniversity of British Columbia.Ritz B, Yu F. 1999. The effect of ambient carbon monoxide onlow birth weight among children born in southernCalifornia between 1989 and 1993. Environ Health Perspect107:17–25.Ritz B, Yu F, Chapa G, Fruin S. 2000. Effect of air pollution onpreterm birth among children born in southern Californiabetween 1989 and 1993. Epidemiology 11(5):502–511.Roemer WH, van Wijnen JH. 2001. Daily mortality and air pollu-tion along busy streets in Amsterdam, 1987-1998.Epidemiology 12(6):649–653.Ryan PH, LeMasters G, Biagini J, Bernstein D, Grinshpun SA,Shukla R, et al. 2005. Is it traffic type, volume, or distance?Wheezing in infants living near truck and bus traffic.J Allergy Clin Immunol 116(2):279–284.Sagiv SK, Mendola P, Loomis D, Herring AH, Neas LM, SavitzDA, et al. 2005. A time-series analysis of air pollution andpreterm birth in Pennsylvania, 1997–2001. Environ HealthPerspect 113:602–606.Salam MT, Millstein J, Li YF, Lurmann FW, Margolis HG,Gilliland FD. 2005. Birth outcomes and prenatal exposureto ozone, carbon monoxide, and particulate matter: resultsfrom The Children’s Health Study. Environ Health Perspect113:1638–1644.Slama R, Morgenstern V, Cyrys J, Zutavern A, Herbarth O,Wichmann HE, et al. 2007. Traffic-related atmospheric pol-lutants levels during pregnancy and offspring’s term birthweight: a study relying on a land-use regression exposuremodel. Environ Health Perspect 115:1283–1292.Smargiassi A, Berrada K, Fortier I, Kosatsky T. 2006. Trafficintensity, dwelling value, and hospital admissions for res-piratory disease among the elderly in Montreal (Canada):a case-control analysis. J Epidemiol Community Health60(6):507–512.Sˇrám RJ, Binková B, Dejmek J, Bobak M. 2005. Ambient air pol-lution and pregnancy outcomes: a review of the literature.Environ Health Perspect 113:375–382.Wilhelm M, Ritz B. 2003. Residential proximity to traffic andadverse birth outcomes in Los Angeles County, California,1994–1996. Environ Health Perspect 111:207–216.Wilhelm M, Ritz B. 2005. Local variations in CO and particulateair pollution and adverse birth outcomes in Los AngelesCounty, California, USA. Environ Health Perspect113:1212–1221.Zhang KM, Wexler AS, Zhu YF, Hinds WC, Sioutas C. 2004.Evolution of particle number distribution near roadways.Part II: The ‘road-to-ambient’ process. Atmos Environ38(38):6655–6665.Zhu YF, Hinds WC, Shen S, Sioutas C. 2004. Seasonal trends ofconcentration and size distribution of ultrafine particlesnear major highways in Los Angeles. Aerosol Sci Technol38:5–13.


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items