seasonal variation in fetal growth: accounting for sociodemographic, biological, and environmental...

mctia

Research www.AJOG.org

OBSTETRICS

Seasonal variation in fetal growth: accounting forsociodemographic, biological, and environmental exposuresGavin Pereira, MAppStat; Angus Cook, PhD; Fatima Haggar, MPH; Carol Bower, PhD; Natasha Nassar, PhD

OBJECTIVE: We sought to investigate seasonal variation in fetalgrowth, accounting for important sociodemographic, biological, andenvironmental exposures.

STUDY DESIGN: Records of births 1998 through 2006 in Perth, WesternAustralia were obtained (N � 147,357). We investigated small for gesta-tional age and sex and the proportion of optimal birthweight (POBW) in rela-tion to seasonal exposures (season, temperature, sunlight) by trimester ofpregnancy. Adjustment was made for a wide range of risk factors.

RESULTS: The POBW for neonates with third trimesters predominantly
in summer was 0.18% (0.00-0.36%) lower than for those in winter.
exposures. Am J Obstet Gynecol 2012;206:74.e1-7.

cagbbnbvvvfl

catp

birthweight, failure tdoi: 10.1016/j.ajog.2011.07.038

74.e1 American Journal of Obstetrics & Gynecology JANUARY 2012

POBW decreased by 0.14% (0.01-0.27%) per interquartile range in-crease in third-trimester temperature (9.15°C). An interquartile rangeincrease in temperature over pregnancy (0.73°C) was associated withan odds ratio of 1.02 (95% confidence interval, 1.00–1.05) for smallfor gestational age and sex.

CONCLUSION: Reduced fetal growth was associated with elevated am-bient temperatures throughout and late in pregnancy, independently ofair pollution and other risk factors.

Key words: air pollution, fetal growth, pregnancy, temperature,
weather
Cite this article as: Pereira G, Cook A, Haggar F, et al. Seasonal variation in fetal growth: accounting for sociodemographic, biological, and environmental

B irthweight is highly indicative of in-fant mortality and morbidity, and

ay play a role in adverse health out-omes later in life.1 As a full range of an-hropometric measurements during thentrauterine period is rarely available on

population basis, birthweight is typi-

From the School of Population Health, TheUniversity of Western Australia (Mr Pereira,Dr Cook, and Ms Haggar), Crawley, and theTelethon Institute for Child HealthResearch, Centre for Child Health Research,(Mr Pereira and Drs Bower and Nassar), TheUniversity of Western Australia, Subiaco,and the Cooperative Research Centre forAsthma (Mr Pereira and Dr Cook) andPerinatal Research, Kolling Institute ofMedical Research, University of Sydney (DrNassar), Sydney, New South Wales,Australia.

Received Feb. 10, 2011; revised May 15,2011; accepted July 25, 2011.

The authors report no conflict of interest.

Reprints: Gavin Pereira, MAppStat, TelethonInstitute for Child Health Research, Centre forChild Health Research, University of WesternAustralia, 100 Roberts Rd., Subiaco, WA 6008Australia. [email protected].

0002-9378/$36.00© 2012 Mosby, Inc. All rights reserved.

ally used to retrospectively assess theppropriateness of fetal growth in a sin-le metric. Fetal growth and subsequentirthweight is associated with a range ofiological, socioeconomic, dietary, eth-ic, and parental factors.2 Reducedirthweight has been observed with ad-erse environmental exposures3 and ele-ated maternal blood pressure causingasoconstriction and reducing bloodow to the fetus.4 Birthweight may be af-

fected by certain environmental expo-sures, including prenatal exposure tomaternal smoking, lead, chemical agents(eg, pesticides), environmental tobaccosmoke, outdoor air pollution, and drink-ing water contaminants.5 A number ofpredictive factors relevant to birthweightmay also be modified by seasonality,6 in-luding environmental/climatic vari-bles (temperature,7 sunlight,6 exposureo ambient air pollutants8), maternal ex-osures/behaviors (nutritional intake,9

infection,10 physical activity,11 environ-mental tobacco smoke12), and biologicalprocesses (pregnancy-induced hyper-tension4). It is important to recognizethe importance of the seasonal variationin fetal growth because some of the po-tential causes listed above may be modi-fiable. Moreover, in studies that estimate

o fully adjust for

these factors may result in confoundedresults. For example, in the case of expo-sures to air pollution, cyclical fluctua-tions and photochemical reactions influ-ence the levels of criteria air pollutants inthe atmosphere and seasonal effects onfetal growth must be considered instudies.13

Published reports of the observed an-nual pattern of seasonal variation in fetalgrowth are inconsistent. Various studiesin developed countries have linked awide range of seasonal factors to lowerbirthweights.14-18 Difference in latitudehas been offered as an explanation forsuch discrepancy, and a vitamin D hy-pothesis followed as a consequence.6,18

However, latitude is closely related to cli-mate and therefore may be a surrogate forweather conditions, such as temperature.This demonstrates the need to investigatetemperature and sunlight exposure, whichare more specific correlates of climate thanlatitude and can be attributed to a biologi-cally plausible mechanism.

The objective of this study was toinvestigate fetal growth in relation toseasonal exposures and identify thevulnerable stages of pregnancy, after ac-counting for a range of important so-ciodemographic, biological, and envi-
ronmental exposures.
mailto:[email protected]

gwlsauTo

n(omtailciiaptt

ynec

www.AJOG.org Obstetrics Research

MATERIALS AND METHODSStudy population and settingThe study population included all single-ton live births in metropolitan Perth,Western Australia, in the period from 1998through 2006. Records were obtainedfrom the Midwives Notification System,which contains birth notifications com-pleted at the time of separation for everystill born or live born baby of �400 gbirthweight and/or �20 weeks’ gestationin Western Australia. The collection in-cludes information on maternal demo-

TABLE 1Maternal and infant characteristicsbirths in metropolitan Perth, 1998 t

Maternal and infant characteristics

All births...................................................................................................................

Maternal..........................................................................................................

Age, y.................................................................................................

�20.................................................................................................

20-24.................................................................................................

25-29.................................................................................................

30-34.................................................................................................

�35..........................................................................................................

Parity.................................................................................................

First birth.................................................................................................

Second birth.................................................................................................

Third birth.................................................................................................

�Fourth birth..........................................................................................................

Aboriginal or TSI..........................................................................................................

Married or de facto..........................................................................................................

Medical condition/pregnancy complication..........................................................................................................

Smoked during pregnancy..........................................................................................................

SEIFA index, mean (SD)...................................................................................................................

Infant, n (%)..........................................................................................................

Male..........................................................................................................

Female..........................................................................................................

Preterm birth (�37 wk)..........................................................................................................

SGA..........................................................................................................

POBW (%) mean (SD)..........................................................................................................

Birthweight, g...................................................................................................................

POBW is unitless as it is ratio of 2 weights.POBW, proportion of optimal birthweight; SEIFA, socioeconomTorres Strait Islander.a Preexisting or gestational diabetes, preeclampsia.

Pereira. Seasonal variation in fetal growth. Am J Obstet G

graphic characteristics, maternal health,

pregnancy, labor, delivery, and infantoutcomes.

Perth is the capital of Western Australiaand lies at 31 degrees latitude south of theequator. Being in the southern hemi-sphere, winter occurs in June through Au-gust and summer in December throughFebruary. The weather in Perth is very dif-ferent to northern hemisphere weatherpatterns; summer and winter tempera-tures average around 30°C and 18°C, re-spectively. The population is geographi-cally concentrated in the capital, Perth,

singletonough 2006

n (%)

147,357..................................................................................................................

..................................................................................................................

..................................................................................................................

6813 (5)..................................................................................................................

21,727 (15)..................................................................................................................

41,863 (28)..................................................................................................................

49,001 (33)..................................................................................................................

27,953 (19)..................................................................................................................

..................................................................................................................

62,713 (43)..................................................................................................................

50,308 (34)..................................................................................................................

21,919 (15)..................................................................................................................

12,417 (8)..................................................................................................................

4100 (3)..................................................................................................................

134,141 (91)..................................................................................................................

13,098 (9)..................................................................................................................

24,442 (17)..................................................................................................................

1015 (87)..................................................................................................................

..................................................................................................................

75,221 (51)..................................................................................................................

72,134 (49)..................................................................................................................

7479 (5)..................................................................................................................

12,031 (8)..................................................................................................................

99 (12)..................................................................................................................

3409 (499)..................................................................................................................

ex for areas; SGA, small for gestational age and sex; TSI,

ol 2012.

with almost 75% of the state’s total popu- w

JANUARY 2012 Ameri

lation of approximately 2 million (2004)living in the metropolitan area.19

Description of outcome andexplanatory factorsThe primary outcome was fetal growth:small for gestational age/sex (SGA) andthe proportion of optimal birthweight(POBW). SGA was calculated based on theAustralian national birthweight percen-tiles20 whereby infants with a weight forestational age �10th centile for their sexere defined as SGA. POBW was calcu-

ated using the ratio of each individual ob-erved birthweight to optimal birthweightnd expressed as a percentage. POBW isnitless because it is a ratio of 2 weights.his model for optimal birthweight, devel-ped by Blair et al,21 accounts for constitu-

tional growth potential by controlling forgestational age, maternal height, birth or-der, and infant sex and was developed us-ing an independent birth cohort free ofknown recorded risk factors. We investi-gated only births of at least 33 weeks ofcompleted gestation because POBW is po-tentially less accurate before this time.21

Seasonal and environmental data wereobtained from the Australian Bureau ofMeteorology and the Western AustralianDepartment of Environment, respectively.Seasonal exposures were defined as dailymaximum temperature (°C) and hours ofsunlight and based on data from the Perthmetropolitan area site (location no.009225) and airport site (location no.009021). Daily metropolitan levels of crite-ria air pollutants, including particulatematter with aerodynamic diameter �2.5�m (PM2.5) and �10 �m (PM10), ozone,

itric oxide (NO), nitrogen dioxideNO2), and carbon monoxide (CO) werebtained from measurements taken at 10etropolitan air pollutant monitoring sta-

ions. Environmental measurements wereveraged across all 10 monitoring stationsn the study region to obtain citywide pol-utant levels. Averaging was applied be-ause not all of these pollutants were mon-tored at every site, which made itnappropriate to use the nearest monitorpproach that has been applied in other airollution analyses. Daily levels of pollu-ion, temperature, and sunlight were ob-ained for each day of pregnancy for each

ofhr

.........

.........

.........

.........

.........

.........

.........

.........

.........

.........

.........

.........

.........

.........

.........a

.........

.........

.........

.........

.........

.........

.........

.........

.........

.........

ic ind

oman in the cohort and then averaged

can Journal of Obstetrics & Gynecology 74.e2

t

mtprsj

cnvAldr

ai

siiwp

at

ynec

Research Obstetrics www.AJOG.org

over the entire duration pregnancyand over each trimester of pregnancyseparately.

As trimesters of pregnancy typically donot overlap exactly with climatic sea-sons, we used temperature to assist in amore appropriate definition of season offirst, second, and third trimesters. Theaverage temperature in third trimesterwas lowest for August through Octoberbirths and highest for February throughApril births, so we defined births duringthese periods as having winter and sum-mer third trimesters, respectively. Thisdefinition is intuitively more sensitive toearly and late seasonal changes. Simi-larly, season of first trimester and secondtrimester were defined using weeks 13and 26 of gestation.

Statistical analyses and adjustmentDescriptive statistics for maternal andinfant characteristics and measurementsof seasonal and criteria pollutants werecalculated. The association of SGA andPOBW was investigated in relation toseasonal exposures (season, tempera-ture, sunlight) for each trimester of preg-nancy and season using logistic and mul-tiple linear regression, respectively,using SAS 9.1 software (SAS InstituteInc, Cary, NC). We presented results forinterquartile range (IQR) (75th–25thpercentile point) increases in tempera-ture, sunlight, and air pollutants. TheIQR represents the range of the middle50% of exposures. Exposures were di-vided by the IQR to gauge the effect of anincrease in exposure from typically lowo typically high levels.

In multivariate models adjustment wasade for maternal condition/complica-

ion (gestational and preexisting diabetes,reeclampsia, indigenous status, mar-ied/de facto relationship status, andmoking during pregnancy). We also ad-usted for criteria air pollutants (PM2.5,

PM10, ozone, NO, NO2, and CO) and so-ioeconomic status using the socioeco-omic index for areas advantage-disad-antage score obtained from theustralian Bureau of Statistics. Due to the

arge number of comparisons, we con-ucted multivariate analyses only for crite-
ia air pollutants and meteorological vari-
74.e3 American Journal of Obstetrics & Gynecolog

bles statistically significant at the 5% leveln univariate analyses.

Assessment of multicollinearityThe presence of multicollinearity was as-sessed by calculating variance inflation fac-tors (VIFs) for effect estimates obtainedfrom both multiple22 and logistic23 regres-ion models. The square root of the VIF isnterpreted as the factor by which the SE isnflated compared to the situation inhich the explanatory variables are com-letely uncorrelated.

Ethics approvalApproval for this study was granted by theUniversityofWesternAustralia(project:RA/4/1/2006) and the Department of HealthWestern Australian Human Research EthicsCommittee (project: 2008/7).

RESULTSMaternal and infant characteristicsThere were 147,357 singleton live births inthe metropolitan Perth area from 1998through 2006 (Table 1). The most preva-lent groups in the study were women aged30-34 years (33%), those having a firstbirth(43%),andwomeninamarriedordefacto relationship (91%). A total of 4100women (3%) were of Aboriginal ethnicityand 24,442 women (17%) reportedlysmoked during pregnancy. Approximately5% of births were preterm. A total of13,098 women (9%) experienced pre-eclampsia or had preexisting or gestational

TABLE 2Overall mean and percentiles of enpregnancy in metropolitan Perth, 1

Variable Mean

Maximum daily temperature, °C 25.03...................................................................................................................

Sunlight, h 8.79...................................................................................................................

PM2.5, �g/m3 8.09...................................................................................................................

PM10, �g/m3 17.36...................................................................................................................

O3, pphm 1.95...................................................................................................................

NO2, pphm 0.64...................................................................................................................

NO, pphm 0.64...................................................................................................................

CO, ppm 0.34...................................................................................................................

CO, carbon monoxide; NO, nitric oxide; NO2, nitrogen dioxide; O�2.5 �m; PM10, particulate matter with aerodynamic diamet


diabetes. The mean birthweight was 3409

y JANUARY 2012

(SD499)gandatotalof12,031births(8%)were classified as SGA.

Distribution of environmentalexposuresThe mean exposures over pregnancy were25.1°C for maximum ambient temperatureand 8.8 hours for sunlight (Table 2). ThePearson correlation between first-trimestertemperature and third-trimester tempera-ture was negative and very high (r � –0.97).Correlations between temperature and airpollutants were low to moderate in magni-tude, with all �0.65 in absolute value (Table3).WeobservedveryhighcorrelationsofCOwith the oxides of nitrogen: NO (r � 0.92)and NO2 (r � 0.82). High correlations werelsoobservedbetweenthe2measuresofpar-iculate matter, PM2.5 and PM10 (r � 0.71)

and between the 2 oxides of nitrogen, NOand NO2 (r � 0.84).

Seasonal variation in fetal growthA summer third trimester was associatedwith a significant decrease in POBW of0.18% (95% confidence interval [CI],0.00 – 0.36%) relative to a winter thirdtrimester (Table 4). Effect sizes were rel-atively smaller for other periods and nostatistically significant effects were ob-served using the classification of SGA. Insummary we observed lower fetal growthfor pregnancies with third trimesterspredominantly in summer, but the ex-tent of this decline in weight was not suf-ficient to produce detectable elevations

onmental exposures duringthrough 2006

h 25th Median 75th 90th

32 24.70 25.10 25.42 25.66..................................................................................................................

49 8.67 8.81 8.95 9.07..................................................................................................................

56 7.76 8.01 8.28 8.70..................................................................................................................

12 16.59 17.26 17.91 18.86..................................................................................................................

88 1.92 1.95 1.98 2.01..................................................................................................................

60 0.61 0.63 0.66 0.69..................................................................................................................

55 0.57 0.61 0.66 0.82..................................................................................................................

29 0.31 0.34 0.36 0.39..................................................................................................................

one; PM2.5, particulate matter with aerodynamic diameter10 �m.

ol 2012.

vir998

10t

24..........

8..........

7..........

16..........

1..........

0..........

0..........

0..........

3, ozer �

in the risk of SGA.

ynec


Effects of temperature andsunlight on fetal growthEffect sizes observed for sunlight expo-sure were small and statistically nonsig-nificant. Relatively larger effect sizes andstatistical significance were limited to

TABLE 3Pearson correlations among weekllevels of environmental variables

Variable Temperature Sunlight P

Temperature 0.74 0...................................................................................................................

Sunlight 0...................................................................................................................

PM2.5...................................................................................................................

PM10...................................................................................................................

O3...................................................................................................................

CO...................................................................................................................

NO...................................................................................................................

CO, carbon monoxide; NO, nitrogen oxide; NO2, nitrogen didiameter �2.5 �m; PM10, particulate matter with aerodynam


TABLE 4Seasonal effects on fetal growth bfor births in Perth metropolitan are

Variable

POBW %

Unadjusted(95% CI)

Season of trimester 1..........................................................................................................

Winter Referent..........................................................................................................

Spring 0.01 (–0.18..........................................................................................................

Summer 0.01 (–0.18..........................................................................................................

Autumn –0.13 (–0.32...................................................................................................................


Winter Referent..........................................................................................................

Spring 0.08 (–0.10..........................................................................................................

Summer 0.13 (–0.07..........................................................................................................

Autumn 0.08 (–0.10...................................................................................................................


Winter Referent..........................................................................................................

Spring –0.03 (–0.21..........................................................................................................

Summer –0.18 (–0.37..........................................................................................................

Autumn –0.07 (–0.26...................................................................................................................

Adjustment variables: socioeconomic status, medical conditioit is ratio of 2 weights.CI, confidence interval; OR, odds ratio; POBW, proportion of oa Statistically significant, alpha � 0.05.

Pereira. Seasonal variation in fetal growth. Am J Obstet Gynec

ambient maximum daily temperatureexposure. An IQR increase in tempera-ture over whole pregnancy (0.73°C) wasassociated with an odds ratio of 1.02(95% confidence interval [CI], 1.00 –1.05) for SGA. POBW decreased by

veraged

5 PM10 O3 CO NO NO2

0.64 0.14 –0.51 –0.51 –0.42..................................................................................................................

0.51 0.16 –0.50 –0.44 –0.44..................................................................................................................

0.71 –0.20 0.43 0.34 0.31..................................................................................................................

0.06 –0.12 –0.14 –0.18..................................................................................................................

–0.52 –0.59 v0.52..................................................................................................................

0.92 0.82..................................................................................................................

0.84..................................................................................................................

; O3, oxide; PM2.5, particulate matter with aerodynamicameter �10 �m.

ol 2012.

ming of pregnancy1998 through 2006

a Adjusted Beta(95% CI)

.........................................................................................................................

Referent.........................................................................................................................

0.19) 0.00 (–0.19 to 0.18).........................................................................................................................

0.19) 0.01 (–0.18 to 0.19).........................................................................................................................

0.05) –0.13 (–0.31 to 0.05).........................................................................................................................

.........................................................................................................................

Referent.........................................................................................................................

0.27) 0.07 (–0.10 to 0.26).........................................................................................................................

0.30) 0.08 (–0.10 to 0.27).........................................................................................................................

0.26) 0.07 (–0.11 to 0.26).........................................................................................................................

.........................................................................................................................

Referent.........................................................................................................................

0.16) –0.02 (–0.20 to 0.17).........................................................................................................................

0.00)a –0.18 (–0.36 to 0.00)a.........................................................................................................................

0.11) –0.06 (–0.24 to 0.12).........................................................................................................................

gnancy complication, Aboriginal status, married or de facto relati

l birthweight; SGA, small for gestational age.

ol 2012.

JANUARY 2012 Ameri

0.14% (0.01-0.27%) for an IQR increase(9.15°C) in third-trimester temperature(Table 5).

In first trimester, an IQR increase intemperature (9.28°C) was associatedwith an increase in POBW of 0.13%(95% CI, 0.00 – 0.26%). However, astemperature in first trimester was nega-tively very highly correlated with tem-perature in third trimester (r � – 0.97),we conducted a post hoc analysis usingan extra sum of squares F test. This fur-ther analysis indicated that third-trimes-ter temperature explained a significantamount of variation in POBW beyondthat explained by first-trimester temper-ature (P � .0446). However, the con-verse did not apply: first-trimester tem-perature did not explain a significantamount of variation beyond that ex-plained by third-trimester temperature(P � .1317). Therefore, it is likely thatthe effect observed in first trimester is an

GA

nadjusted OR95% CI)

Adjusted OR(95% CI)

..................................................................................................................

eferent Referent..................................................................................................................

.01 (0.96–1.07) 1.02 (0.97–1.08)..................................................................................................................

.99 (0.94–1.05) 1.00 (0.95–1.05)..................................................................................................................

.99 (0.94–1.05) 1.00 (0.94–1.05)..................................................................................................................

..................................................................................................................


.00 (0.94–1.05) 1.00 (0.94–1.05)..................................................................................................................

.01 (0.96–1.06) 1.02 (0.96–1.07)..................................................................................................................

.99 (0.94–1.05) 0.99 (0.94–1.05)..................................................................................................................

..................................................................................................................


.99 (0.94–1.04) 0.98 (0.93–1.03)..................................................................................................................

.99 (0.94–1.04) 0.98 (0.93–1.03)..................................................................................................................

.00 (0.95–1.05) 0.99 (0.94–1.04)..................................................................................................................

ip status, smoking during pregnancy. POBW is unitless as

y a

M2.

.27

.........

.27

.........

.........

.........

.........

.........

.........

oxideic di

y tia,

S

Bet U(

......... .........

R......... .........

to 1......... .........

to 0......... .........

to 0......... .........

......... .........

R......... .........

to 1......... .........

to 1......... .........

to 0......... .........

......... .........

R......... .........

to 0......... .........

to 0......... .........

to 1......... .........

n/pre onsh

ptima


iawiawwpceMcta

ynec


“inverse” proxy for that observed inthird trimester.

Assessment of confoundingby air pollutantsThe only pollutants associated with SGAand POBW in unadjusted models wereCO and PM10 but these effects dimin-ished to negligible levels after adjustmentfor temperature and other factors thataffect fetal growth. After adjustment, anIQR increase in whole pregnancy expo-sure to CO was associated with a negligi-ble 1% (95% CI, –2 to 4%) decreasedodds of SGA. The VIFs for temperature,CO, and all other risk factors were�1.10. An IQR increase in third-trimes-ter PM10 was associated with a negligible0.01% (95% CI, – 0.14 to 0.17%) in-crease in POBW. The VIFs for tempera-ture and PM10 in this model were both3.50, indicating a moderate level of mul-ticollinearity well below the commonlyused threshold of 10. The VIFs for theother risk factors were all �1.10, indicat-ing that the level of multicollinearity wasnegligible for these risk factors. There-fore, there was no significant evidencethat associations observed for tempera-

TABLE 5Effects of ambient temperature andarea, 1998 through 2006, before an

Variable IQR

POBW %

Unadjus(95% CI

Temperature, °C..........................................................................................................

Trimester 1 9.28 0.13 (0..........................................................................................................

Trimester 2 9.11 –0.05 (–..........................................................................................................

Trimester 3 9.15 –0.15 (–..........................................................................................................

Pregnancy 0.73 –0.02 (–...................................................................................................................

Sunlight, h..........................................................................................................

Trimester 1 3.61 0.07 (–..........................................................................................................

Trimester 2 3.59 –0.08 (–..........................................................................................................

Trimester 3 3.50 –0.08 (–..........................................................................................................

Pregnancy 0.29 –0.02 (–...................................................................................................................

Adjustment variables: socioeconomic status, medical conditioit is ratio of 2 weights.CI, confidence interval; IQR, interquartile range; OR, odds ratia Adjusted for particulate matter with aerodynamic diameter �

pregnancy; c Statistically significant, alpha � 0.05.


ture were confounded by the air pollut-


ants or other risk factors investigated inthis study.

COMMENTWe observed that fetal growth, as char-acterized by POBW, was lowest for sum-mer third trimesters. A more detailedevaluation of temperature and sunlightexposure revealed that fetal growth maybe adversely affected by higher tempera-tures sustained over pregnancy and wereassociated with a 2% (95% CI, 0 –5%)increase in risk of growth restriction (asdefined using SGA classification). Therewas also indication that elevated temper-atures in third trimester predicted de-clines in fetal growth, with a decrease inPOBW of 0.14% (0.01-0.27%) per quar-tile increase in temperature (9.15°C).These results suggest that heat effectsthroughout and potentially late in preg-nancy may constrain fetal growth inde-pendently of ambient air pollutants andother risk factors, particularly in areassystematically subjected to high seasonalpeaks. Effect sizes observed for sunlightexposure were small and statisticallynonsignificant, and therefore not sup-

unlight on fetal growth for births in Pfter adjustment for criteria air pollu

Beta Adjusted Beta(95% CI)

.........................................................................................................................

0.26)c 0.13 (0.00–0.26)c.........................................................................................................................

to 0.08) –0.04 (–0.17 to 0.10).........................................................................................................................

to –0.02)c –0.14 (–0.27 to –0.01)a,c

.........................................................................................................................

to 0.07) 0.05 (–0.04 to 0.14).........................................................................................................................

.........................................................................................................................

to 0.19) 0.07 (–0.06 to 0.19).........................................................................................................................

to 0.04) –0.08 (–0.20 to 0.04).........................................................................................................................

to 0.05) –0.07 (–0.20 to 0.05).........................................................................................................................

to 0.07) –0.02 (–0.10 to 0.06).........................................................................................................................

gnancy complication, Aboriginal status, married or de facto relati

BW, proportion of optimal birthweight; SGA, small for gestational

m, only pollutant in third trimester associated with POBW; b Adju

ol 2012.

portive of a vitamin D hypothesis. How- t

y JANUARY 2012

ever, the levels of vitamin D may havebeen consistently high among our studypopulation due to sunshine in Perth.

Consideration of climatic differencesis important because it is debatablewhether it is the pollutant or the col-linear meteorological condition that is agreater contributor to the likelihood ofthe adverse health outcome. Other au-thors have suggested that much of theseasonal variation in mortality and mor-bidity thought due to particulate mattermight be explained by meteorologicaland climatic information.24 Our find-ngs were consistent with this hypothesisnd highlight the strength of our models,hich were specifically designed to take

nto account any confounding effects ofmbient air pollutants. Interestingly,hen levels of ambient air pollutantsere included in the models with tem-erature, the temperature effects onlyhanged slightly but the air pollutantffects diminished to negligible levels.

oreover, the presence of strong multi-ollinearity between temperature andhe air pollutant would inflate the vari-nce of the effect estimates to the extent

th metropolitann

SGA

Unadjusted OR(95% CI)

Adjusted OR(95% CI)

..................................................................................................................

1.02 (0.98–1.06) 1.02 (0.98–1.06)..................................................................................................................

1.01 (0.97–1.05) 1.00 (0.97–1.04)..................................................................................................................

1.00 (0.96–1.03) 0.99 (0.95–1.03)..................................................................................................................

1.04 (1.01–1.07)c 1.02 (1.00–1.05)b,c

..................................................................................................................

..................................................................................................................

1.02 (0.98–1.05) 1.02 (0.98–1.05)..................................................................................................................

1.00 (0.97–1.04) 1.01 (0.96–1.03)..................................................................................................................

0.98 (0.95–1.02) 0.98 (0.95–1.02)..................................................................................................................

1.01 (0.98–1.03) 1.01 (0.98–1.03)..................................................................................................................

ip status, smoking during pregnancy. POBW is unitless as

and sex.

for carbon monoxide, only pollutant associated with SGA over

s erd a tio

ted)

......... .........

.00–......... .........

0.18......... .........

0.29......... .........

0.11......... .........

......... .........

0.06......... .........

0.20......... .........

0.20......... .........

0.10......... .........

n/pre onsh

o; PO age

10 � sted

hat one would not typically observe sta-

nes

iwtsriebawbtAwaws

msmc

ttwtsntao


tistically significant individual effects.However, we did observe such effects forelevated temperature exposure. Very lowVIFs indicated that it was highly unlikelythat the effect of elevated temperatureover whole pregnancy on the risk of SGAwas confounded by exposure to CO. Themoderate size VIFs for elevated PM10

(VIF � 3.50) and temperature (VIF �3.50) in third trimester indicated thatthere were no obvious signs of con-founding. Nonetheless, the possibility ofconfounding by PM10 in this model can-

ot be practicably ruled out due to thelement of subjectivity in interpretinguch moderate size VIFs.

Study location is an important factorn interpreting our results in contextith past studies, particularly in relation

o the mean ambient temperature expo-ure, which varies across different studyegions. Similar findings were reportedn an ecological study conducted in east-rn Australia that found a spring peak inirthweight, consistent with our result ofwinter third-trimester peak in birth-eight.17 Most previous studies haveeen conducted in cooler climates thanhose experienced in Western Australia.

study from Turkey, where subjectsere exposed to lower ambient temper-

tures, observed an increase in birth-eight with low ambient temperature in

econd trimester.14 These results suggesta detrimental effect of cooler tempera-tures in midpregnancy. A time seriesanalysis adopted by Murray et al9 in 2000in Northern Ireland reported similarfindings.

Although we observed a detrimentaleffect of higher temperatures in the thirdtrimester (and a possibly spurious nega-tive effect of cooler temperature in firsttrimester), a study conducted in Scot-land by Lawlor et al (2005) reported as-sociations in these trimesters but withopposite direction of effects.15 The op-posite direction of effect may be ex-plained by their study setting, wherelower temperature may be more relevantthan ambient hyperthermal stress in acooler climate.15 Their study used daily

inimum temperatures to derive expo-ures (which may be a potentially better

etric to capture hypothermal stress in a
ooler climate) compared to our study
hat used daily maxima (potentially bet-er to capture hyperthermal stress in aarmer climate). However, it is possible

hat their first-trimester effect was alsopurious, given that they observed a highegative correlation between tempera-

ure in the middle of the first trimesternd temperature in the middle of the sec-nd trimester (r � – 0.79).Based on the total evidence to date it

remains unclear whether second or thirdtrimester is the most relevant period forthese exposures. One of the strengths ofour study was that we were able to adjustfor ambient air pollution, which exhibitstemporal variation, as well as adjust for awide range of other maternal and so-ciodemographic risk factors, includingsocioeconomic status, complications ofpregnancy, and smoking during preg-nancy. Correlations between fetalgrowth and temperature or air pollutionare attenuated when these exposures areaveraged over pregnancy because preg-nancies occur at different times of theyear. Nonetheless, there remains a strongunderlying temporal correlation be-tween air pollution and temperature.Therefore, it is likely that the size of theeffect estimates were attenuated by theadjustment for air pollutants. Further in-accuracies in exposure estimation mayhave arisen because the subjects onlyspent part of their time outdoors andmaternal temperature may be regulatedby heating and air conditioning in theindoor environment.

A limitation of our study was that notall pollutants were measured at the airmonitoring stations, and as a result weused averages across monitoring stationsrather than adopt a more accuratemethod for measurement.25,26 Althoughit is likely that this approach adequatelyaccounted for temporal variations in airpollutant levels, it would not account forspatial variation in these levels. As withtemperature estimates, the recordedoutdoor levels of the air pollutants mayhave differed from indoor levels. How-ever, it is not anticipated that such vari-ability would have resulted in differentialexposure misclassification and hencewould not have led to significant bias in
the final risk estimates.
JANUARY 2012 Ameri

Implications● Effects of restricted fetal growth ob-

served in air pollution studies may bedue to a correlated temperature effectrather than an air pollution effect.

● Future studies on fetal growth out-comes should account for such sea-sonal variation.

ConclusionOur study confirmed the hypothesis ofseasonality in fetal growth. The inconsis-tency between our results and those inprevious studies suggests that the effectsof seasonal exposures may be geograph-ically specific. Elevated ambient temper-atures throughout and potentially late inpregnancy were associated with reduc-tions in fetal growth independently ofother risk factors and irrespective of thepresence of an air pollution effect. Theeffect of ambient air pollution (CO andPM10) fell to negligible levels after con-sideration of temperature. Ambient airpollution effects previously reportedmay be artefacts of an association withtemperature. f

REFERENCES1. Barker DJP. Adult consequences of fetalgrowth restriction. Clin Obstet Gynecol 2006;49:270-83.2. Wilcox AJ. Fertility and pregnancy: an epide-miologic perspective. New York: Oxford Univer-sity Press; 2010.3. Perera FP, Rauh V, Whyatt RM, et al. Molec-ular evidence of an interaction between prenatalenvironmental exposures and birth outcomes ina multiethnic population. Environ Health Per-spect 2004;112:626-30.4. Rothenberg SJ, Kondrashov V, Manalo M, etal. Increases in hypertension and blood pres-sure during pregnancy with increased bonelead levels. Am J Epidemiol 2002;156:1079-87.5. Wigle DT, Arbuckle TE, Turner MC, et al. Ep-idemiologic evidence of relationships betweenreproductive and child health outcomes and en-vironmental chemical contaminants. J ToxicolEnviron Health B Crit Rev 2008;11:373-517.6. Chodick G, Flash S, Deoitch Y, Shalev V.Seasonality in birth weight: review of global pat-terns and potential causes. Hum Biol 2009;81:463-77.7. Wells JCK. Thermal environment and humanbirth weight. J Theor Biol 2002;214:413-25.8. Slama R, Darrow L, Parker J, et al. Meetingreport: atmospheric pollution and human repro-duction. Environ Health Perspect 2008;116:791-8.9. Murray LJ, O’Reilly DP, Betts N, Patterson
CC, Davey Smith G, Evans AE. Season and


outdoor ambient temperature: effects on birthweight. Obstet Gynecol 2000;96:689-95.10. Robinson JS, Moore VM, Owens JA,McMillen IC. Origins of fetal growth restriction.Eur J Obstet Gynecol Reprod Biol 2000;92:13-9.11. Pivarnik JM. Potential effects of maternalphysical activity on birth weight: brief review.Med Sci Sports Exerc 1998;30:400.12. Ward C, Lewis S, Coleman T. Prevalence ofmaternal smoking and environmental tobaccosmoke exposure during pregnancy and impacton birth weight: retrospective study using mil-lennium cohort. BMC Public Health 2007;7:81-5.13. Pereira G, Nassar N, Bower C, Weinstein P,Cook A. Residential exposure to traffic emis-sions and adverse pregnancy outcomes.Surveys and Perspectives Integrating Environ-ment and Society (S.A.P.I.EN.S) 2010;3:1-24.Available at: http://sapiens.revues.org/966#quotation. Accessed Sept. 15, 2011.14. Elter K, Ay E, Uyar E, Kavak ZN. Exposureto low outdoor temperature in the midtrimesteris associated with low birth weight. Aust N Z JObstet Gynaecol 2004;44:553-7.15. Lawlor DA, Leon DA, Davey Smith G. The
association of ambient outdoor temperature

throughout pregnancy and offspring birthweight:findings from the Aberdeen children of the 1950scohort. BJOG 2005;112:647-57.16. McGrath J, Barnett A, Eyles D, Burne T,Pedersen CB, Mortensen PB. The impact ofnonlinear exposure-risk relationships on sea-sonal time-series data: modeling Danish neona-tal birth anthropometric data. BMC Med ResMethodol 2007;7:45.17. McGrath JJ, Barnett AG, Eyles DW. The as-sociation between birth weight, season of birthand latitude. Ann Hum Biol 2005;32:547-59.18. Waldie KE, Poulton R, Kirk IJ, Silva PA. Theeffects of pre- and post-natal sunlight exposureon human growth: evidence from the southernhemisphere. Early Hum Dev 2000;60:35-42.19. Australian Bureau of Statistics. Australiandemographic statistics 3101.0. Canberra: Aus-tralian Bureau of Statistics; 2010.20. Roberts CL, Lancaster PA. Australian na-tional birthweight percentiles by gestationalage. Med J Aust 1999;170:114-8.21. Blair EM, Liu Y, de Klerk NH, Lawrence DM.Optimal fetal growth for the Caucasian single-ton and assessment of appropriateness of fetalgrowth: an analysis of a total population perina-
tal database. BMC Pediatr 2005;5:13.
y JANUARY 2012

22. Hocking R. Methods and applications of lin-ear models: regression and the analysis of vari-ance, 2nd ed. New Jersey: Wiley; 2003:166.23. Menard S. An introduction to logistic re-gression diagnostics. In: Loughton CD, San-toyo D, eds. Applied logistic regression analy-sis. Sage university series on quantitativeapplications in the social sciences, 07-106.Thousand Oaks, CA: Sage; 2002:76. Availableat: http://dionysus.psych.wisc.edu/coursewebsites/Psy710/Readings/SAGE_AppliedLo-gisticRegressionAnalysis.pdf. Accessed Sept.15, 2011.24. Cox LH. Statistical issues in the study of airpollution involving airborne particulate matter.Environmetrics 2000;11:611-26.25. Briggs DJ, de Hoogh C, Gulliver J, et al. Aregression-based method for mapping traffic-related air pollution: application and testing infour contrasting urban environments. Sci TotalEnviron 2000;253:151-67.26. Pereira G, Cook A. Time series modelingof background air pollution for health studies.In: Bridgeman H, ed. 19th International cleanair and environment conference. [CD-ROM].
CASANZ, Perth, Australia, 2009.
http://sapiens.revues.org/966#quotation

http://dionysus.psych.wisc.edu/coursewebsites/Psy710/Readings/SAGE_AppliedLo-gisticRegressionAnalysis.pdf

http://sapiens.revues.org/966#quotation



seasonal variation in fetal growth: accounting for sociodemographic, biological, and environmental...

Documents