ehp.1306837 outdoor air pollution, preterm birth, and low birth weight

8/12/2019 Ehp.1306837 Outdoor Air Pollution, Preterm Birth, And Low Birth Weight

1/7

Environmental Health Perspectives VOLUME122 |NUMBER4 |April 2014 425

Research|Childrens HealthAll EHPcontent is accessible to individuals with disabilities. A fully accessible (Section 508compliant)HTML version of this article is available at http://dx.doi.org/10.1289/ehp.1306837 .

Outdoor Air Pollution, Preterm Birth, and Low Birth Weight: Analysis of theWorld Health Organization Global Survey on Maternal and Perinatal HealthNancy L. Fleischer,1Mario Merialdi,2Aaron van Donkelaar,3Felipe Vadillo-Ortega,4,5Randall V. Martin,3,6Ana Pilar Betran,2Joo Paulo Souza,2and Marie S. ONeill7

1University of South Carolina Arnold School of Public Health, Columbia, South Carolina, USA; 2Department of Reproductive Health andResearch, World Health Organization, Geneva, Switzerland; 3Department of Physics and Atmospheric Science, Dalhousie University,Halifax, Nova Scotia, Canada; 4School of Medicine, Universidad Nacional Autnoma de Mxico, Mexico City, Mexico; 5Instituto

Nacional de Medicina Genomica, Mexico City, Mexico; 6Harvard-Smithsonian Center for Astrophysics, Cambridge, Massachusetts, USA;7University of Michigan School of Public Health, Ann Arbor, Michigan, USA

BACKGROUND:Inhaling fine particles (particulate matter with diameter 2.5 m; PM2.5) caninduce oxidative stress and inflammation, and may contribute to onset of preterm labor and otheradverse perinatal outcomes.

OBJECTIVES:We examined whether outdoor PM2.5was associated with adverse birth outcomesamong 22 countries in the World Health Organization Global Survey on Maternal and PerinatalHealth from 2004 through 2008.

METHODS:Long-term average (20012006) estimates of outdoor PM2.5were assigned to 50-kmradius circular buffers around each health clinic where births occurred. We used generalized esti-mating equations to determine associations between clinic-level PM2.5levels and preterm birth andlow birth weight at the individual level, adjusting for seasonality and potential confounders at indi-

vidual, clinic, and country levels. Country-specific associations were also investigated.

RESULTS:Across all countries, adjusting for seasonality, PM 2.5was not associated with pretermbirth, but was associated with low birth weight [odds ratio (OR) = 1.22; 95% CI: 1.07, 1.39 forfourth quartile of PM2.5(> 20.2 g/m

3) compared with the first quartile (< 6.3 g/m3)]. In China,the country with the largest PM2.5range, preterm birth and low birth weight both were associated

with the highest quartile of PM2.5only, which suggests a possible threshold effect (OR = 2.54; CI:1.42, 4.55 and OR = 1.99; CI: 1.06, 3.72 for preterm birth and low birth weight, respectively, forPM2.5 36.5 g/m

3compared with PM2.5< 12.5 g/m3).

CONCLUSIONS:Outdoor PM2.5concentrations were associated with low birth weight but not pre-term birth. In rapidly developing countries, such as China, the highest levels of air pollution may beof concern for both outcomes.

CITATION:Fleischer NL, Merialdi M, van Donkelaar A, Vadillo-Ortega F, Martin RV, Betran AP,Souza JP, ONeill MS. 2014. Outdoor air pollution, preterm birth, and low birth weight: analy-sis of the World Health Organization Global Survey on Maternal and Perinatal Health. EnvironHealth Perspect 122:425430; http://dx.doi.org/10.1289/ehp.1306837

IntroductionAir pol lut ion is associated with increasedmorbidity and mortality for multiple healthindicators, including cardiovascular disease,lung cancer, acute respiratory infections,asthma, and pregnancy outcomes (Brunekreefand Holgate 2002; Glinianaia et al. 2004;Kampa and Castanas 2008; Lacasana et al.2005; Maisonet et al. 2004; rm et al.2005). Inequity in health outcomes associatedwith air pollution occurs among people liv-ing in low-income countries compared withhigh-income countries, and for poor people

living in countries at all levels of develop-ment (ONeill et al. 2008). Preterm birth(< 37 weeks gestation) and low birth weight(LBW) (< 2,500 g) have been associated withair pollution exposure, but the weight ofthe evidence is not yet sufficient to establishcausality at this time (Maisonet et al. 2004;rmet al. 2005). LBW is a consequence ofreduced length of gestation and/or restrictedfetal growth in utero (Kramer 2003). Bothprematurity and growth restriction makeimportant contributions to morbidity andmortality during infancy, and in the long

term these conditions may put adults at riskfor a wide range of adverse health outcomes(Longo et al. 2013; Rogers and Velten 2011).

Air pollutants may be part of a complexset of factors that increase the risk of pretermbirth or LBW through processes related toinflammation, oxidative stress, endocrine dis-ruption, and impaired oxygen transport acrossthe placenta (Slama et al. 2008). Exposureto airborne particles with diameter 2.5 m(PM2.5) is of particular relevance in relationto pregnancy outcomes. hese particles canbe inhaled into the deep regions of the lung,

and oxidative stress and inflammation maybe among the mechanistic pathways throughwhich exposure to this pollutant may con-tribute to onset of preterm labor (Slama et al.2008). In addition, previous research showsthat fine particles are more spatially homo-geneous than other pollutants, and outdoormeasurements of these particles may serve asa useful proxy index of personal exposure to arange of pollutants (Sarnat et al. 2005).

Most studies of air pollution and adversebirth outcomes have been conducted in com-munities in high-income countries, with very

few data in low- and middle-income countries.Few studies have examined cross-country com-parisons of the relationship between air pol-lution and birth outcomes, where differencesin pollution levels may be most extreme. TeWorld Health Organization (WHO) GlobalSurvey on Maternal and Perinatal Health(WHOGS) database (Shah et al. 2008) offersa unique opportunity to link global estimatesof fine particulate matter with pregnancy out-comes in many areas of the world where thisline of investigation has yet to be undertaken.

he aim of this paper is to examinethe relationship between PM2.5 and pre-term birth and LBW among 22 countriesin the WHOGS.

Methods

Population. Te WHOGS is a multicountry,cross-sectional survey that collected data onall deliveries in participating facilities for 23months, depending on the annual volume ofdeliveries of the facility. Data were collectedfor > 290,000 women in 373 institutions in 24countries in Africa, Latin America, and Asia.Te WHOGS was implemented in Africa and

the Americas between September 2004 andMarch 2005, and in Asia between October2007 and April 2008. Te survey had a strati-fied multistage cluster sampling design, withfour countries sampled from each of the 14WHO-defined subregions that are under thebroader regions of Africa, the Americas, andAsia (except in two subregions with only threecountries each). Te capital city and two ran-domly selected provinces were included, fol-lowed by a random sampling of up to seven

Ad dr es s co rre sp on de nc e to N. L. Fl ei sc he r,Department of Epidemiology and Biostatistics,Arnold School of Public Health, University of South

Carolina, 915 Greene St., 4th Floor, Columbia, SC29208 USA. elephone: (803) 777-6220. E-mail:[email protected] thank . Sunbury for her assistance with geo-

coding the clinics.Support for this work provided by the World

Health Organization (contract 200174365) and R01grants ES016932 and ES017022 from the NationalInstitute of Environmental Health Sciences,National Institutes of Health.

Te authors declare they have no actual or poten-tial competing financial interests.

Received: 22 March 2013; Accepted: 4 February2014; Advance Publication: 7 February 2014; FinalPublication: 1 April 2014.
http://dx.doi.org/10.1289/ehp.1306837http://dx.doi.org/10.1289/ehp.1306837http://dx.doi.org/10.1289/ehp.1306837mailto:[email protected]://dx.doi.org/10.1289/ehp.1306837mailto:[email protected]://dx.doi.org/10.1289/ehp.1306837


2/7

Fleischer et al.

426 VOLUME122 |NUMBER4 |April 2014 Environmental Health Perspectives

health institutions in each location with at least1,000 deliveries in the year before the survey.Facility data on available services were col-lected at each site, as were data on all womenwho delivered in the facilities during the studyperiod. Individual-level data were abstractedfrom medical records by trained data collec-tors. We obtained written permission from

the ministry of health of each country andthe director of each health facility. Individualinformed consent was not obtained becausethis study was a cluster-level study in whichdata were extracted from medical records withno individual identification. Te ethics reviewcommittee of WHO and that of each countryapproved the study protocol. Detailed meth-odology of the WHOGS has been describedelsewhere (Shah et al. 2008).

he WHOGS defined preterm birth asgestational age of < 37 weeks at delivery, asdetermined by the best available obstetric esti-mate of gestational age. LBW was defined as< 2,500 g at birth. Because of heterogeneityin the quality of the estimated gestational ageacross the survey countries, we did not usethe traditional cut-off of LBW being amongonly full-term births. Only live, spontaneous,singleton births were included in the analyses.All analyses were restricted to facilities with< 20% prevalence of preterm birth (to main-tain as much comparability as possible for theestimation of gestational age) and > 100 birthsrecorded during the 2- to 3-month samplingperiod (having < 100 suggests problems withthe completeness of sampling of births orthat the facility may have had fewer than theinclusion criterion of 1,000 deliveries/year).

In addition, countries with fewer than halfof the randomly selected facilities from thatcountry meeting our inclusion criteria werealso excluded. Data from five African coun-tries, eight countries in the Americas, and nineAsian countries are included in this analysis(22 of 24 countries in the WHOGS).

Demographic and pregnancy-related fac-tors, including age, maternal education (years),parity, prenatal care (number of antenatal vis-its), and infant sex were treated as potentialconfounders in the analysis. All variables werecontinuous in the models, except infant sex.Women with missing data on the birth out-

comes or any of the potential confounders wereexcluded from the analysis. A median of 0.7%of women per facility were missing data forpreterm birth, with 0.4% per facility missingfor LBW. Data on the birth outcomes wererarely missing; education was the most com-mon source of missing data, although infor-mation on prenatal care was also frequentlymissing in some countries. Of the 305 facilitiesin the analysis, 43 had > 10% of women with-out complete data on the birth outcomes andpotential confounders. Sensitivity analyses wererun excluding the 43 facilities, in addition to

two other facilities from one country since all ofits other facilities had high levels of missingness.

Air pol lut ion exposure ass essment andother indicators. Remote sensing data providea useful estimate of pollution levels in theabsence of extensive local ground-based moni-tor networks, particularly when the nearestmonitor is located > 100 km away (Lee et al.

2012). Such monitoring networks are rarein less wealthy regions of the world (Cohenet al. 2005). Air pollution exposure for thisstudy is therefore represented with global esti-mates of PM2.5as developed by van Donkelaaret al. (2010). hese values provide a long-term average (20012006) global estimateof PM2.5at approximately 10 km 10 kmresolution. hey are derived from a combi-nation of observations from the ModerateResolution Imaging Spectroradiometer(MODIS) (Levy et al. 2007) and MultiangleImaging Spectroradiometer (MISR) (Dineret al. 1998) instruments from the erra satel-lite, and simulations with the GEOS-Chemchemical transport model (www.geos-chem.org). Te resultant PM2.5data were validatedusing ground-based data and have an expected1-sigma uncertainty of 1 g/m3 + 25%.Further details have been published elsewhere(van Donkelaar et al. 2010).

Our objective was to use these PM2.5concentrations to estimate exposure duringpregnancy among the women whose data werecaptured by the WHOGS. Health facilitiesparticipating in the WHOGS were geocodedusing the exact address or city information,as available, to determine the closest pos-sible geographical coordinates using Google

Earth (http://www.google.com/earth/). Next,50-kmradius circular buffers were createdaround the coordinates of interest, and aver-age PM2.5concentrations within these bufferswere then matched to the health facilities. Tisbuffer size was chosen to represent a realis-tic distance within which women giving birthat the facilities might live, because residentialaddresses of these women were not availablein the survey. Te seasonal impact of samplingand uncertainty on satellite-derived PM2.5esti-mates unfortunately limited their direct use ona monthly basis. Rather, adjustment was madefor the impact of seasonality on the relation-

ship between PM2.5levels and adverse birthoutcomes using simulated seasonality fromthe GEOS-Chem model. A scalar variable ofPM2.5deviation from the overall 20012006average was simulated for each calendar monthand multiplied by the original PM2.5 levelto estimate exposure for the calendar monthpreceding each womans delivery date. heseasonally adjusted PM2.5 values were usedin the regression analyses. Te month beforebirth was chosen for seasonal adjustmentbecause of the strong seasonal patterns of airpollution exposure in some locations, and

the potential importance of exposure duringthe third trimester to adverse birth outcomes(Ritz and Wilhelm 2008; Woodruff et al.2009). Exposure in the first trimester has alsobeen associated with adverse birth outcomes.Terefore, we also performed sensitivity analy-ses using an average of the scalar variable fromthe first 3 months of pregnancy to adjust for

seasonal variation from the overall average.For some locations, data were availablefrom air pollution monitors located within50 km of the clinics. In these cases, a com-parison between PM2.5levels measured bythe ground monitors and the levels estimatedfrom the satellite imagery was possible usingsupplemental data published with the originalsatellite estimates publication (van Donkelaaret al. 2010). We calculated ratios of themeasured to estimated PM2.5 and averagedthe ratios for the corresponding metropoli-tan area. In the instance where the ratio waseither > 2.0 or < 0.50, we added half thedifference between the average of the mea-sured concentrations in the metropolitan areanear the clinic and those estimated from theremote sensing imagery, and incorporated thisadjusted estimate in sensitivity analyses. Aftercalculating the ratio of available ground-basedmonitored PM2.5and satellite-estimated data,only one study city, So Paulo, Brazil, had atleast a 2-fold difference between the methodsfor the second sensitivity analysis. Its ratio ofmeasured concentration to satellite estimateswas 2.6. hus, for the Brazilian clinics nearSo Paulo, 8.24 g/m3was added to the satel-lite-derived estimate of PM2.5within 50 km ofthe clinic. Sensitivity analyses were run using

the adjusted Brazil estimates.We also examined several country-level

indicators in relation to adverse birth out-comes and air pollution levels. Gross domesticproduct (GDP) per capita [in internationaldollars at purchasing power parity (PPP) rates]was obtained from the Central Intel ligenceAgency s Wor ld Factbook (Cent ra lIntelligence Agency 2007), and populationliving in urban areas (percent), per capitahealth care expenditure (current U.S. dollars),and the country-level Gini coefficient, a mea-sure of income inequality with values from0 (equality) to 1 (inequality), were obtained

from the World Banks World DevelopmentIndicators (World Bank 2009), except for theGini coefficient data for Algeria, Cuba, andJapan, which were obtained from the WorldIncome Inequality Database (United NationsUniversityWorld Institute for DevelopmentEconomics Research 2008). Data from 2006were used for all country-level variables, or theclosest year if data were unavailable for 2006.

Statistical analysis. Birth outcomes forwomen from the same health facil ities maybe correlated, thereby violating the inde-pendence assumption of basic regression


3/7

Global study of air pollution and birth outcomes


models. herefore, we used generalized esti-mating equation (GEE) models (Liang andZeger 1986) to account for the nested struc-ture of the data (individual women withinhealth facilities within countries) when esti-mating the associations between seasonallyadjusted, clinic-level PM2.5exposure levels andbirth outcomes. wo GEE models were run

for each outcome. In the first model, a globalestimate was obtained combining all countrieswhile controlling for mothers age, education,parity, and prenatal care and the infants sex.Te second model was also adjusted for othercountry-level covariates (GDP per capita,urbanicity, antenatal care coverage, per capitahealth care expenditure, and the Gini coef-ficient) when determining the global estimate.PM2.5effect estimates were calculated per10-g/m3increments and as quartiles, in sepa-rate models. Te quartiles were based on thedistribution for the entire study population.

We ran corresponding country-specificGEE models for China and India, the twocountries with the widest ranges of PM2.5lev-els. New quartile cut points for these modelswere based on country-specific distributions.

Results

Data from 192,900 live, spontaneous, single-ton births from 22 countries in Africa, Asia,and Latin America were used in our analyses(able 1). he prevalence of preterm birthranged from 3.0% in Vietnam, to 11.1% inTailand. Algeria had the lowest prevalenceof LBW at 3.5%, and India had the highest

at 20.4%. Paraguay had the lowest facility-level average PM2.5levels during 20012006.Facilities in China and India had the largestranges of average PM2.5, and also the facilitieswith the highest average levels. PM2.5levelsaveraged across 20012006 for each facilitycan be seen in Figure 1.

Te odds of preterm birth among women

exposed to higher seasonally adjusted PM2.5levels were not different from those for womenexposed to lower levels of PM2.5based onmodels with and without adjustment forcountry-level variables (able 2). When assess-ing the results by PM2.5quartiles, odds ratios(ORs) were close to the null for all exposurequartiles, without evidence of a positive trend.

For LBW, women in the highest twoquartiles had higher odds of LBW babiescompared with women in the lowest quar-tile of PM2.5exposure [OR = 1.19; 95%CI: 1.06, 1.33 for quartile 3 (PM2.511.96to < 20.16 g/m3) vs. quartile 1 (PM2.5< 6.298 g/m3); OR = 1.22; 95% CI: 1.07,1.39 for quartile 4 (PM2.5 20.16 g/m

3) vs.quartile 1 (PM2.5 < 6.298 g/m

3)] (able 2).hese results were slightly attenuated,but remained statistically significant, whenadjusted for country-level variables.

Because of the large variability of PM2.5levels in China and India, we examined eachcountry separately using country-specificquartiles of exposure (able 3). In China,we found a higher odds of preterm birth andLBW among mothers in the highest quartile ofPM2.5exposure ( 36.5 g/m

3) compared with

those in the lowest quartile (< 12.5 g/m3)(OR = 2.54; 95% CI: 1.42, 4.55 for pretermbirth; OR = 1.99; 95% CI: 1.06, 3.72 forLBW). Linear trends based on PM2.5mod-eled as a simple continuous variable were alsostatistically significant for each birth outcomein China (OR = 1.11; 95% CI: 1.04, 1.17and OR = 1.07; 95% CI: 1.01, 1.14 for pre-

term and LBW in association with 10-g/m3

increases in PM2.5, respectively). In India, wesaw evidence for an inverse association betweenPM2.5levels and both preterm birth and LBW.Results for preterm birth were not statisticallysignificant for either the linear estimate or thequartile analysis. However, for the quartileanalysis of LBW we saw an inverse associa-tion, whereby women in the highest quartile ofPM2.5 ( 70.3 g/m

3) exposure had a lowerodds of LBW babies compared to women inthe lowest quartile (< 18.8 g/m3) (OR = 0.82;95% CI: 0.75, 0.90). Te linear trend for thisrelationship was also statistically significant.

We ran sensitivity analyses excluding facil-ities with a high level of missingness and forPM2.5adjusted from ground-based monitors.For preterm birth, results were comparableto the main analysis when we excluded facili-ties with the large proportions of births withmissing data, and when we adjusted exposurelevels for women who gave birth in Brazilianclinics near So Paulo using data fromground-based monitors; there was no evidencethat preterm birth was associated with PM2.5(data not shown). For LBW, results from thesensitivity analyses were qualitatively similar to

Table 1.Description of preterm birth and air pollution characteristics, by country, WHO Global Survey on Maternal and Perinatal Health, 20042008.

Region and countryNo. of

facilities

Live,spontaneous,singletonbirths (n)

Pretermbirth (%) LBW (%)

Mothersage [years

(mean SD)]

Motherseducation [years

(mean SD)]Parity

(mean SD)Antenatal visits

(mean SD)

MeanPM2.5[g/m

3(range)a]

Seasonally adjustedPM2.5[g/m

3(range)]

AfricaAlgeria 17 12,718 3.7 3.5 30.3 5.8 8.6 4.6 2.6 1.7 5.0 2.6 10.716.7 3.611.7Congo, DR 19 7,067 7.4 11.8 27.2 6.8 8.3 3.8 3.3 2.4 3.6 1.5 11.716.8 6.624.4Kenya 19 16,694 9.4 7.3 24.7 5.7 9.8 2.9 2.1 1.4 4.1 2.3 4.25.5 2.57.7Niger 7 4,826 3.2 10.4 26.6 6.5 3.8 4.4 3.4 2.4 2.9 1.5 27.734.1 3.044.0Nigeria 17 6,538 8.6 5.3 27.6 6.0 9.4 5.6 3.0 2.2 5.7 3.9 17.535.4 7.153.5

AsiaCambodia 5 5,170 5.6 6.9 26.8 5.5 7.0 3.8 1.8 1.2 4.4 2.1 13.315.8 16.823.9China 21 9,221 4.8 3.6 26.2 4.7 8.7 3.5 1.4 0.6 5.9 3.5 6.498.1 2.6145.2India 13 14,622 10.3 20.4 24.5 3.6 6.1 4.1 1.8 1.0 3.2 2.8 19.663.9 10.6109.3Japan 10 2,191 4.3 7.9 31.0 4.9 14.0 2.0 1.6 0.8 11.8 3.2 8.720.9 11.834.9Nepal 8 7,042 9.2 11.4 23.5 4.2 6.1 4.6 1.6 0.9 3.8 2.1 20.646.3 11.761.6Philippines 17 11,326 7.6 14.1 26.3 6.4 10.4 2.6 2.2 1.6 4.5 2.8 8.211.0 12.723.2Sri Lanka 14 6,381 8.1 14.2 27.6 5.5 10.4 2.6 1.9 1.0 8.6 3.3 5.57.7 4.914.9Thailand 11 7,344 11.1 9.2 26.9 6.1 9.6 4.0 1.7 0.9 7.9 3.6 8.921.7 15.835.6Vietnam 15 11,800 3.0 4.1 27.7 4.7 12.4 2.8 1.6 0.7 5.8 2.8 9.643.4 12.154.3

Latin AmericaArgentina 14 7,745 6.6 5.7 26.6 6.5 9.3 3.1 2.3 1.7 6.2 3.0 4.58.3 3.610.1Brazil 19 10,735 7.1 8.4 24.1 6.0 7.6 3.1 2.2 1.6 5.9 2.6 1.46.3 1.39.3Cuba 17 7,841 4.0 3.9 26.3 6.4 11.4 2.6 1.7 0.8 11.2 2.7 7.19.2 3.78.5Ecuador 13 9,845 7.0 10.3 24.6 6.3 9.0 3.6 2.3 1.5 5.3 3.0 4.013.0 2.814.1Mexico 20 14,497 7.5 7.5 25.0 6.0 8.4 3.3 2.2 1.4 6.5 3.0 10.821.9 4.320.1Nicaragua 6 4,001 6.3 7.0 23.0 5.8 6.8 3.6 2.2 1.5 4.0 2.6 7.28.1 1.010.9Paraguay 6 2,466 8.6 5.6 25.4 6.4 8.8 3.4 2.4 1.8 4.9 2.9 3.34.5 3.16.8Peru 17 12,830 5.8 5.0 26.2 6.4 10.1 3.1 2.0 1.3 6.2 3.1 8.620.6 6.930.9

DR, Democratic Republic.aMean PM2.5refers to the mean values during 20012006 for each facility in each country.


4/7

Fleischer et al.


the main analysis, though model 2 ORs wereno longer statistically significant (= 0.05)for the upper two quartiles compared with thelowest quartile (data not shown).

When we performed sensitivity analysesof associations with PM2.5 levels that wereadjusted for levels during the first trimesterrather than levels during the month before

birth, associations of preterm birth and LBWwith PM2.5modeled as a continuous vari-able were negative for model 2 (OR = 0.59;95% CI: 0.49, 0.70 and OR = 0.75; 95% CI:0.68, 0.83 for a 10-g/m3increase in expo-sure for preterm birth and LBW, respectively).For both outcomes, ORs for exposures inthe second and third quartiles versus the firstquartile (< 4.8 g/m3) were nonsignificant inall models, but the highest quartile of PM2.5exposure ( 27.3 g/m3) was negatively associ-ated with preterm birth and LBW (OR = 0.25;95% CI: 0.09, 0.67 and OR = 0.59; 95% CI:0.38, 0.92, respectively, for model 2.)

DiscussionWe investigated the relationship between airpollution and pregnancy outcomes acrosscountries from vastly different regions ofthe world. By using data from women inthe WHOGS and PM2.5 levels derivedfrom remote sensing data, we were able toestimate associations for a study populationthat included women from areas of the worldwhere it is often difficult to acquire reliabledata on both pregnancy outcomes and airpollution concentrations. Estimated PM2.5exposures were not associated with pretermbirth based on our analysis, but LBW was sig-

nificantly higher among women who deliveredin facilities where PM2.5concentrations wereabove the median (i.e., > 12.0 g/m3) com-pared with women delivering at facilities withaverage PM2.5 levels < 6.3 g/m

3. In China,the country with the largest range of PM2.5exposure levels, both preterm birth and LBW

were significantly higher among women withestimated exposure to at least 36.5 g/m3ofPM2.5compared with women in the lowestquartile of exposure (< 12.5 g/m3).

For preterm birth, we found null resultswhen looking at PM2.5levels across countries.In the United States and other high-incomecountries, PM2.5has been associated with

preterm birth in many studies (Brauer et al.2008; Chang et al. 2012; Darrow et al. 2009;Huynh et al. 2006; Kloog et al. 2012; Ritzet al. 2007; Wu et al. 2009, 2011), althoughtwo studies reported no association (Gehringet al. 2011; Rudra et al. 2011). Few studieshave been published on the relation betweenPM2.5 and preterm birth in low- and mid-dle-income countries, or across countries atdifferent levels of development. PM10wasassociated with preterm birth in a study of> 374,167 births from Seoul, South Korea,in 19982000 using Cox models and expo-sure by trimester (Suh et al. 2009). In China,PM

10was associated with preterm birth in a

time series analysis of daily births in 2004 inShanghai (n = 3,346 preterm births) (Jianget al. 2007), and in a time series analysis of142,312 births in 2007 in Guangzhou (Zhaoet al. 2011). Misclassification of the expo-sure or preterm birth or uncontrolled con-founding by co-exposures in this sample ofmostly low- and middle-income countriescould have biased associations toward the nullin our analysis. In China, only the highestquartile of PM2.5 exposure was associatedwith preterm birth compared with the lowestquartile. It may be that, given co-exposuresto other environmental factorswhich may

act as uncontrolled confounders (e.g., poornutrition due to seasonal availability) or effectmodifiers (e.g., indoor air pollution) of therelationshipthe impact of air pollution maybe most prominent at higher levels in mid-dle-income countries, of which China is animportant example. Te null results in India

could be attributable to a downward bias dueto misclassification of the pollution exposureor the measurement of preterm birth, or dueto other co-exposures or environmental fac-tors, as described above. Another possibility isthat the most severely affected fetuses did notsurvive to be counted as live births, resultingin the appearance of protective effects in the

highest exposure categories. In addition, thequartile cut points in both China and Indiawere much higher than the cut points fromthe overall analysis. Exposure in the lowestquartile in China and India may have beenso high already that a relationship betweenPM2.5exposure and preterm birth and LBWwould not be detectable with the first quartileas the reference level of exposure.

LBW was positively associated with PM2.5exposure when data were pooled across all22 countries in our analysis, consistent withfindings in the United States and other high-income countries showing an increased risk ofLBW with higher levels of PM

2.5(Bell et al.

2007; Kloog et al. 2012; Morello-Frosch et al.2010; Parker et al. 2005; Wilhelm et al. 2012);this increase was also evident in a recent meta-analysis of data from nine mostly high-incomecountries (Dadvand et al. 2013). Other studies,however, have found no relationship (Braueret al. 2008; Gehring et al. 2011). Again, evi-dence from low- and middle-income coun-tries is scarce. A study of 891 newborns born19941999 and randomly selected from amongparticipants in a casecontrol study fromtwo districts in the Czech Republic found anincreased risk of LBW associated with PM2.5;analyses did not adjust for potential confound-

ers (Rossner et al. 2011). Other studies lookingat PM10 found a higher risk of LBW associ-ated with exposure. A cross-sectional study inSo Paulo, Brazil, of 179,460 live births dur-ing 1997 found that PM10exposure duringthe first trimester of pregnancy was associatedwith LBW (Gouveia et al. 2004). However,

Figure 1.Map showing estimated PM2.5levels in 50-kmradius buffers around clinics in 22 countries, 20012006.

0 10 20 30 40 50 60 70 80 90

PM2.5(g/m3)


5/7

Global study of air pollution and birth outcomes


a cross-sectional study of births from 2002(n = 77,987) in Rio de Janeiro, Brazil, reportedno association between PM10(PM with diam-eter 10 m) exposure and LBW, regardlessof trimester of exposure (Junger and de Leon2007). In Seoul, South Korea, a cross-sectionalstudy of births from 20022003 found a higherrisk of LBW associated with annual PM10

exposure (Seo et al. 2007); a similar study of177,660 births from 2004 in seven Koreancities found the same relationship (Seo et al.2010). We again saw evidence for a thresh-old effect for LBW in China, where womenexposed to at least 36.5 g/m3of PM2.5hadhigher odds of experiencing LBW. In India,contrary to expectation, women in the high-est quartile of PM2.5exposure experienced alower risk of LBW compared to women in thelowest quartile. Tis may be attributable to anumber of factors, including co-exposures (suchas indoor air pollution) or residual confound-ing (such as malnutrition). Tese factors mayoutweigh any potential effect of outdoor airpollution that we would expect to see. Also,as noted above, other potential explanationsinclude selective survival of fetuses that werenot as severely affected, and also that the quar-tiles cut points are higher in China and Indiathan in the overall analysis, resulting in com-parisons with lowest quartiles that themselvescontained fairly high exposure levels.

Limitations and strengths. Tis study hasa number of limitations. Te survey data werecross-sectional, so we were unable to assessthe dynamic relationship between variationsin preterm birth and other adverse outcomes,and their relationships with air pollution over

time. Related to this, the exposure assessmentis a 6-year average of particulate matter ratherthan a point-specific exposure assessment. Wedid have some data mismatch in timing, sincethe exposure was assessed for 20012006,whereas some of the birth data were collected

from 2007 through 2008 (Asia). We treatedthe 6-year average exposure as a proxy forlong-term exposure. Because pollution lev-els are typically correlated over time, the useof the 6-year average data as a proxy shouldbe representative of the period during whichbirths in Asia were recorded. If pollution levelsincreased significantly in the Asian countries

in 20072008, the association between pol-lution and adverse birth outcomes would beunderestimated. A further limitation of thedata mismatch was that the WHOGS was col-lected only during particular months in eachregion. Since pregnancy outcomes and air pol-lution show seasonal variation, we may nothave captured significant changes in weatherthat may have occurred annuallyfor exam-ple, that would alter the relationship betweenair pollution and adverse birth outcomes.

Additionally, there may be critical periodsduring the pregnancy when fetuses are par-ticularly vulnerable to the effects of air pollu-tion. We did adjust our pollution estimates forseasonal differences, to help account for someof these issues. However, when we conducteda sensitivity analysis using the first trimes-ter as the critical period for seasonal adjust-ment rather than the month before birth, wefound null and protective effects, contrary toexpectation. Because the first trimester wasdetermined by subtracting the gestational agefrom the birth date, the period may not havebeen accurately obtained given the potentialissues with determination of gestational age(see more details below). In addition, becausethe sampling of births was done during spe-cific months in each region, and generally

excluded the months AprilMay throughAugustSeptember, we do not have a full pic-ture of the exposures throughout the year. Ifannual fluctuations were important during theyears of sampling, we may not be accuratelycapturing the true exposure.

We also assumed that the particulate mat-ter measurement at the facility is representa-tive of the exposure to each of the womenwho delivered at the facility. By including abuffer of 50 km around the clinic, we tried toensure that most women who used the clinichad an appropriate exposure value assigned tothem. Because PM2.5is one of the most spa-

tially homogeneous markers of air pollution,this assumption is often applied in air pollu-tion epidemiology (Miller et al. 2007; Parket al. 2006). Additionally, these particles canbe well correlated with individual exposures(Sarnat et al. 2005). However, it is possiblethat women traveled > 50 km to the clinics.For this reason and others, exposure couldhave been misclassified.

Misclassification is also possible for the out-come variables, preterm birth and LBW. Tecapacity to accurately measure some variablesin resource-poor countries (e.g., gestational ageand birth weight) is a well-understood chal-lenge. Although gestational age was calculatedby the best available obstetric estimate at eachclinic, the precision of this estimate may varybetween clinics within countries, and betweencountries. Because the definition of pretermbirth relies on gestational age, it may have beenmisclassified. We attempted to minimize themisclassification of LBW by not restrictingthe definition to full term infants, althoughmisclassification may still exist.

We also have limited data on individualcharacteristics. Smoking information was notcollected in the survey, and the only measure ofthe mothers weight was latest weight beforedelivery, the date of which was not recorded;

so we were unable to accurately calculate bodymass index for the women and thus did notinclude it as a confounder. We also had noinformation on indoor air pollutants, whichwould be particularly important in the poorercountries where women often cook with

Table 2.Adjusted ORs (95% CI) for preterm birth and LBW associated with a10-g/m3increase in PM2.5and with quartiles of PM2.5(relative to the lowestquartile) after adjusting exposure estimates to account for seasonality, WHOGlobal Survey on Maternal and Perinatal Health, 20042008.

Outcome Model 1 Model 2

Preterm birthPM2.5(10 g/m

3) 0.96 (0.91, 1.02) 0. 96 (0.90, 1.02)< 6.35 1.0 (Reference) 1.0 (Reference)6.35 to < 12.32 1.08 (0.95, 1.22) 1.08 (0.95, 1.24)12.32 to < 22.20 1.05 (0.90, 1.23) 1.06 (0.90, 1.25) 22.20 0.96 (0.79, 1.17) 0.96 (0.79, 1.18)

LBWPM2.5(10 g/m

3) 1.00 (0.97, 1.03) 0.99 (0.96, 1.01)< 6.298 1.0 (Reference) 1.0 (Reference)6.298 to


6/7

Fleischer et al.


biomass fuels indoors. Because most of thecountries in our study are low- and middle-income countries, this is of particular concern.Other limitations to conducting these types ofstudies in poorer countries, and across coun-tries at different levels of development, are thatthere may be other area-level confounders thataffect the relationship of interest. For instance,

season may affect both air pollution levels andnutrient availability [e.g., antioxidant vitamins(Casanueva et al. 2005)], which certainly affectpregnancy outcomes. Poorer countries arealso more like to have worse air quality (withfewer regulations restricting pollution) andmore vulnerable populations in general (Cohenet al. 2005). We tried to account for some ofthese between-country variations by includingcountry-level markers of economic developmentand inequality, but we recognize that these maynot be sufficient controls for these differences.

Despite the limitations, this study hasmany strengths. Tis is the first multicountrystudy to analyze air pollution as a potentialdeterminant for preterm birth and LBW thatincluded data from predominantly low- andmiddle-income countries (22 countries inthree different regions). An additional strengthis the homogeneity of the design and data col-lection across countries through a standardizedform and training for data collection.

Conclusions

Tis study is the first to investigate the rela-tionship between air pollution and adversepregnancy outcomes using WHOGS datafrom mostly low- and middle-income coun-tries from around the world. We found no

association between PM2.5levels and pretermbirth, but higher PM2.5levels were associatedwith a higher risk of LBW. In rapidly devel-oping countries, such as China, the highestlevels of air pollution may be of concern forboth preterm birth and LBW.

REFERENCES

Bell ML, Ebisu K, Belanger K. 2007. Ambient air pollution and lowbirth weight in Connecticut and Massachusetts. EnvironHealth Perspect 115:11181124; doi:10.1289/ehp.9759.

Brauer M, Lencar C, Tamburic L, Koehoorn M, Demers P,Karr C. 2008. A cohort study of traffic-related air pollu-

tion impacts on b irth outcomes. Environ Health Perspect116:680686; doi:10.1289/ehp.10952.

Brunekreef B, Holgate ST. 2002. Air pollution and health.Lancet 360:12331242.Casanueva E, Ripoll C, Tolentino M, Morales RM, Pfeffer F,

Vilchis P, et al. 2005. Vitamin C supplementation to preventpremature rupture of the chorioamniotic membranes: arandomized trial. Am J Clin Nutr 81:859863.

Central Intelligence Agency. 2007. The World Factbook.Available: https://www.cia.gov/library/publications/the-world-factbook/index.html [accessed 14 July 2010].

Chang HH, Reich BJ, Miranda ML. 2012. Time-to-event analysisof fine particle air pollution and preterm birth: results fromNorth Carolina, 20012005. Am J Epidemiol 175:9198.

Cohen AJ, Anderson HR, Ostro B, Pandey KD, Krzyzanowski M,Knzli N, et al. 2005. The global burden of disease due tooutdoor air pollution. J Toxicol Env Heal A 68:13011307.

Dadvand P, Parker J, Bell ML, Bonzini M, Brauer M,Darrow LA, et al. 2013. Maternal exposure to particulateair pollution and term birth weight: a multi-country evalu-ation of effect and heterogeneity. Environ Health Perspect121:367373; doi:10.1289/ehp.1205575.

Darrow LA, Klein M, Flanders WD, Waller LA, Correa A,Marcus M, et al. 2009. Ambient air pollution and pretermbirth a time-series analysis. Epidemiology 20:689698.

Diner DJ, Beckert JC, Reilly TH, Bruegge CJ, Conel JE,Kahn RA, et al. 1998. Multi-angle imaging spectroradiom-eter (MISR)instrument description and experiment over-

view. Trans Geosci Remote Sensing 36:10721087.Gehring U, Wijga AH, Fischer P, de Jongste JC, Kerkhof M,

Koppelman GH, et al. 2011. Traffic-related air pollution,preterm birth and term birth weight in the Piama birthcohort study. Environ Res 111:125135.

Glinianaia SV, Rankin J, Bell R, Pless-Mulloli T, Howel D. 2004.Particulate air pollution and fetal health: a systematicreview of the epidemiologic evidence. Epidemiology15:3645.

Gouveia N, Bremner SA, Novaes HMD. 2004. Associationbetween ambient air pollution and birth weight in SoPaulo, Brazil. J Epidemiol Community Health 58:1117.

Huynh M, Woodruff TJ, Parker JD, Schoendorf KC. 2006.Relationships between air pollution and preterm birth inCalifornia. Paediatr Perinat Epidemiol 20:454461.

Jiang LL, Zhang YH, Song GX, Chen GH, Chen BH, Zhao NQ, et al.2007. A time series analysis of outdoor air pollution and pre-

term birth in Shanghai, China. Biomed Environ Sci 20:426431.Junger WL, de Leon AP. 2007. Air pollution and low birth

weight in the city of Rio de Janeiro, Brazil, 2002 [inPortuguese]. Cad Saude Publica 23:S588S598.

Kampa M, Castanas E. 2008. Human health effects of air pollu-tion. Environ Pollut 151:362367.

Kloog I, Melly SJ, Ridgway WL, Coull BA, Schwartz J. 2012.Using new satellite based exposure methods to study theassociation between pregnancy PM2.5exposure, pre-mature birth and birth weight in Massachusetts. EnvironHealth 11:40; doi:10.1186/1476-069X-11-40.

Kramer MS. 2003. The epidemiology of adverse pregnancy out-comes: an overview. J Nutr 133:1592S1596S.

Lacasana M, Esplugues A, Ballester F. 2005. Exposure to ambi-ent air pollution and prenatal and early childhood healtheffects. Eur J Epidemiol 20:183199.

Lee SJ, Serre ML, van Donkelaar A, Martin RV, Burnett RT,Jerrett M. 2012. Comparison of geostatistical interpola-

tion and remote sensing techniq ues f or estimatin g lo ng-term expos ure to ambie nt PM2.5 concentrations across

the cont inent al Unite d Stat es. Envir on Healt h P erspe ct120:17271732; doi:10.1289/ehp.1205006.Levy RC, Remer LA, Mattoo S, Vermote EF, Kaufman YJ. 2007.

Second-generation operational algorithm: retrieval ofaerosol properties over land from inversion of moderateresolution imaging spectroradiometer spectral reflectance.J Geophys Res-Atmos 112:16; doi:10.1029/2006JD007811.

Liang KY, Zeger SL. 1986. Longitudinal data-analysis using gen-eralized linear-models. Biometrika 73:1322.

Longo S, Bollani L, Decembrino L, Di Comite A, Angelini M,Stronati M. 2013. Short-term and long-term sequelae inintrauterine growth retardation (IUGR). J Matern FetalNeonatal Med 26:222225.

Maisonet M, Correa A, Misra D, Jaakkola JJ. 2004. A reviewof the literature on the effects of ambient air pollution onfetal growth. Environ Res 95:106115.

Miller KA, Siscovick DS, Sheppard L, Shepherd K, Sullivan JH,Anderson GL, et al. 2007. Long-term exposure to air pol-lution and incidence of cardiovascular events in women.

N Engl J Med 356:447458.Morello-Frosch R, Jesdale BM, Sadd JL, Pastor M. 2010.Ambient air pollution exposure and full-term birth weight inCalifornia. Environ Health 9:44; doi:10.1186/1476-069X-9-44.

ONeill MS, Kinney PL, Cohen AJ. 2008. Environmental equityin air quality management: local and international impli-cations for human health and climate change. J ToxicolEnviron Health A 71:570577.

Park SK, ONeill MS, Wright RO, Hu H, Vokonas PS, Sparrow D,et al. 2006. Hfe genotype, particulate air pollution, andheart rate variability: a gene-environment interaction.Circulation 114:27982805.

Parker JD, Woodruff TJ, Basu R, Schoendorf KC. 2005. Air pol-lution and birth weight among term infants in California.Pediatrics 115:121128.

Ritz B, Wilhelm M. 2008. Ambient air pollution and adversebirth outcomes: methodologic issues in an emerging field.Basic Clin Pharmacol Toxicol 102:182190.

Ritz B, Wilhelm M, Hoggatt KJ, Ghosh JKC. 2007. Ambient airpollution and preterm birth in the environment and preg-nancy outcomes study at the University of California, LosAngeles. Am J Epidemiol 166:10451052.

Rogers LK, Velten M. 2011. Maternal inflammation, growthretardation, and preterm birth: insights into adult cardio-vascular disease. Life Sci 89:417421.

Rossner P, Tabashidze N, Dostal M, Novakova Z, Chvatalova I,

Spatova M, et al. 2011. Genetic, biochemical, and envi-ronmental factors associated with pregnancy outcomesin newborns from the Czech Republic. Environ HealthPerspect 119:265271; doi:10.1289/ehp.1002470.

Rudra CB, Williams MA, Sheppard L, Koenig JQ, Schiff MA.2011. Ambient carbon monoxide and fine particulatematter in relation to preeclampsia and preterm deliveryin western Washington State. Environ Health Perspect119:886892; doi:10.1289/ehp.1002947.

Sarnat JA, Brown KW, Schwartz J, Coull BA, Koutrakis P. 2005.Ambient gas concentrations and personal particulate mat-

ter exposuresimplicat ions for studying the health effe ctsof particles. Epidemiology 16:385395.

Seo JH, Ha EH, Kim OJ, Kim BM, Park HS, Leem JH, et al. 2007.Environmental health surveillance of low birth weightin Seoul using air monitoring and birth data [in Korean].J Prev Med Public Health 40:363370.

Seo JH, Leem JH, Ha EH, Kim OJ, Kim BM, Lee JY, et al. 2010.Population-attributable risk of low birthweight related toPM10pollution in seven Korean cities. Paediatr PerinatEpidemiol 24:140148.

Shah A, Faundes A, Machoki M, Bataglia V, Amokrane F,Donner A, et al. 2008. Methodological considerationsin implementing the WHO Global Survey for MonitoringMaternal and Perinatal Health. Bull World Health Organ86:126131.

Slama R, Darrow L, Parker J, Woodruff T, Strickland M,Nieuwenhuijsen M, et al . 2008. Meeting report:Atmospheric pollution and human reproduction. Reportof the Munich international workshop. Environ HealthPerspect 116:791798; doi:10.1289/ehp.11074.

rm RJ, Binkova B, Dejmek J, Bobak M. 2005. Ambient air pol-lution and pregnancy outcomes: a review of the literature.Environ Health Perspect 113:375382; doi:10.1289/ehp.6362.

Suh YJ, Kim H, Seo JH, Park H, Kim YJ, Hong YC, et al. 2009.Different effects of PM10 exposure on preterm birth bygestational period estimated from time-dependent survival

analyses. Int Arch Occup Environ Health 82:613621.United Nations UniversityWorld Institute for DevelopmentEconomics Research. 2008. World Income InequalityDatabase V2.0c May 2008. Available: http://www.wider.unu.edu/research/Database/en_GB/database/ [accessed23 March 2010].

van Donkelaar A, Martin RV, Brauer M, Kahn R, Levy R,Verduzco C, et al. 2010. Global estimates of exposure tofine particulate matter concentrations from satellite basedaerosol optical depth. Environ Health Perspect 118:847855;doi:10.1289/ehp.0901623.

Wilhelm M, Ghosh JK, Su J, Cockburn M, Jerrett M, Ritz B.2012. Traffic-related air toxics and term low birth weightin Los Angeles County, California. Environ Health Perspect120:132138; doi:10.1289/ehp.1103408.

Woodruff TJ, Parker JD, Darrow LA, Slama R, Bell ML, Choi H,et al. 2009. Methodological issues in studies of air pollutionand reproductive health. Environ Res 109:311320.

World Bank. 2009. World DataBank. Available: http://ddp-ext.

worldbank.org/ [accessed 30 March 2010].Wu J, Ren CZ, Delfino RJ, Chung J, Wilhelm M, Ritz B. 2009.Association between local traffic-generated air pollu-

tion and preeclamp sia an d pre term delivery in t he So uthCoast Air Basin of California. Environ Health Perspect117:17731779; doi:10.1289/ehp.0800334.

Wu J, Wilhelm M, Chung J, Ritz B. 2011. Comparing expo-sure assessment methods for traffic-related air pollutionin an adverse pregnancy outcome study. Environ Res111:685692.

Zhao QG, Liang ZJ, Tao SJ, Zhu JA, Du YK. 2011. Effects of airpollution on neonatal prematurity in Guangzhou of China:a time-series study. Environ Health 10:2; doi:10.1186/1476-069X-10-2.


7/7

VOLUME121 |NUMBER6 |June 2014 Environmental Health Perspectives A 151

All EHPcontent is accessible to individuals with disabilities. A fully accessible (Section 508compliant)

HTML version of this article is available at http://dx.doi.org/10.1289/ehp.122-A151. Erratum

Erratum: Outdoor Air Pollution, Preterm Birth, and Low Birth Weight: Analysis of the World Health OrganizationGlobal Survey on Maternal and Perinatal Health

In Figure 1 of the article Outdoor Air Pollution, Preterm Birth, and Low Birth Weight: Analysis of the World Health OrganizationGlobal Survey on Maternal and Perinatal Health by Fleischer et al. [Environ Health Perspect 122:425430 (2014); http://dx.doi.org/10.1289/ehp.1306837], clinics in Algeria were omitted from the map of Africa (center). In addition, clinics in other countries that

were excluded from the analysis as a result of data quality issues should not have been included in the map. The corrected figure appears

below.

0 10 20 30 40 50 60 70 80 90

PM2.5(g/m3)

Figure 1. Map showing estimated PM2.5levels in 50-kmradius buffers around clinics in 22 countries, 20012006.

The authors regret the error.
http://dx.doi.org/10.1289/ehp.1306837http://dx.doi.org/10.1289/ehp.1306837http://dx.doi.org/10.1289/ehp.1306837http://dx.doi.org/10.1289/ehp.1306837http://dx.doi.org/10.1289/ehp.1306837http://dx.doi.org/10.1289/ehp.1306837

ehp.1306837 outdoor air pollution, preterm birth, and low birth weight

Documents