traffic emissions are associated with reduced fetal growth in areas of perth, western australia: an...

2011 vol. 35 no. 5 AUSTRALIAN AND NEW ZEALAND JOURNAL OF PUBLIC HEALTH 451© 2011 The Authors. ANZJPH © 2011 Public Health Association of Australia

Traffic emissions contribute greatly to

anthropogenic ambient air pollution

in many urbanised locations and

its association with adverse pregnancy

outcomes is increasingly investigated.1

Intuitively, traffic emission concentrations

are higher closer to roads.2-4 However, much

of the past research that has investigated

ambient air pollution in relation to

pregnancy outcomes has used wide-area

measurements obtained from air monitoring

stations, which may account for the some

of the inconsistency among results.5-8 Few

studies have investigated the association

between such localised variation in traffic

emissions and intrauterine growth (IUG)9-

11 as historically studies have focused

on respiratory outcomes such as asthma

exacerbation.12 None of these fetal growth

studies have explicitly accounted for the

Traffic emissions are associated with reduced

fetal growth in areas of Perth, Western Australia:

an application of the AusRoads dispersion model

Gavin Pereira Centre for the Built Enviironment and Health, School of Population Health, The University of Western Australia; Cooperative Research Centre for Asthma, New South Wales; Telethon Institute for Child Health Research, Centre for Child Health Research, The University of Western Australia

Natasha NassarTelethon Institute for Child Health Research, Centre for Child Health Research, The University of Western Australia; Perinatal Research, Kolling Institute of Medical Research, Sydney University, New South Wales

Angus CookSchool of Population Health, The University of Western Australia; Cooperative Research Centre for Asthma, New South Wales

Carol BowerTelethon Institute for Child Health Research, Centre for Child Health Research, The University of Western Australia

Submitted: November 2010 Revision requested: April 2011 Accepted: May 2011Correspondence to: Gavin Pereira, School of Population Health, M707, The University of Western Australia, 35 Stirling Highway, Crawley WA 6009; e-mail: [email protected]

Abstract

Background: This study, in a region with

relatively low industrial activity, used a

highly specific marker for traffic emissions,

accounted for the inherent fetal growth

potential, and used complete record

linkage of births, midwife notifications,

deaths, hospital morbidity and birth defect

records.

Methods: Clinical records were obtained

for pregnancies between 2000 and 2006

in three areas of Perth, Western Australia

(n=3,501). We used carbon monoxide

as a marker for locally derived traffic

emissions, and assessed exposure using

the AusRoads dispersion model. Fetal

growth was characterised by proportion of

optimal birth weight and investigated using

multivariate mixed-effects regression.

Results: Exposure in the third trimester

was associated with a -0.49% (sd=0.23%)

change in proportion of optimal birth

weight per 10 µg/m3 increase in locally

derived traffic emissions. However, this

result was confined to one of the three

study areas due to elevated exposure

misclassification among women in the

other two areas. Among this group, a

neonate who would have otherwise

attained an optimal birth weight of

3.5 kg would be expected to be born

58 g lighter for an interquartile increase

in third trimester exposure, which was

approximately half of the effect observed

for maternal smoking during pregnancy.

Conclusion: We observed an association

between maternal exposure to traffic

emissions and reduced fetal growth.

This effect was supported by sensitivity

analyses but only observed in one of the

three study areas. Further studies are

required to corroborate our results.

Key words: Intrauterine growth, traffic,

motor vehicle emissions, fetal growth

Aust NZ J Public Health. 2011;35:451-8

doi: 10.1111/j.1753-6405.2011.00760.x

constitutional growth potential of the fetus,

and all applied a land-use regression model

for exposure that included non-traff ic

predictor variables, which thereby diluted

specificity.

In Perth, Western Australia, carbon

monoxide (CO) is a suitable marker for

traffic emissions as 80% of atmospheric

CO can be attributed to motor vehicle

emissions. 13 Although past s tudies

have used background air monitoring

station concentrations of CO to represent

exposure,10,14,15 which is suitable for the

assessment of risks associated with traffic-

related wide-area background air pollution,

it is an inappropriate marker for local traffic

emissions about the residential address. In

an Australian context, the relatively recent

development of the AusRoads dispersion

model allows the calculation of maximum

Article Fetal Exposure

452 AUSTRALIAN AND NEW ZEALAND JOURNAL OF PUBLIC HEALTH 2011 vol. 35 no. 5© 2011 The Authors. ANZJPH © 2011 Public Health Association of Australia

hourly CO concentrations, specific to the maternal residential

address. Dispersion models can directly characterise the localised

spatial variability of emission concentrations, and has been

recently applied to investigate the association between motor

vehicle emissions and pre-eclampsia and pre-term delivery.16

The advantage of using CO over other candidate markers for traffic emissions (such as nitric oxide, nitrogen dioxide, particulate matter and polycyclic aromatic hydrocarbons) is that it has greater specificity when dispersion modelling is applied. Nitric oxide is highly reactive, making modelling difficult, while nitrogen dioxide is soluble and therefore possibly less suitable for exposure assessment during rainy seasons. Other sources of particulate matter, such as bushfires, controlled burns, wind-blown dust, pollen, sea salt and wood heater emissions reduce its specificity as a traffic-exposure. Polycyclic aromatic hydrocarbons are highly toxic but are also highly volatile. Conversely, CO is relatively inert, and therefore a more accurate marker for motor vehicle traffic.

IUG is influenced by a variety of pathological determinants (such as maternal infection, diabetes and the presence of birth defects) and perinatal factors (such as gestational age, gender, multiplicity of birth, parity and ethnicity).17-21 One of the main debates in the assessment of IUG is the use of an appropriate measure of growth. To date, low birth weight (LBW) and small for gestational age (SGA) have been the most commonly applied measures, but these have been considered to be inappropriate indicators of adverse IUG.22,23 The 2,500 g cut-off for LBW does not account for neonates who are constitutionally small. Even within an entirely healthy population the distributional criterion (usually 10%) used to define SGA would result in the erroneous identification of growth restricted neonates. In response to these issues, proportion of optimal birth weight (POBW), a ratio of the observed birth weight to the expected individual optimal birth weight, was proposed by Blair et al. 2005.24 This ratio, which accounts for constitutional growth potential, also reflects IUG on a continuum, rather than as a binary outcome.

The objective of this study was to investigate the relationship between IUG and personal residential exposure to motor vehicle traffic emissions during pregnancy, using a more appropriate measure of IUG and a well established method for the ascertainment of exposure.

MethodsStudy design and setting

A retrospective cohort study was conducted in Perth, Western Australia for births to women resident in the study region between 1 January 2000 and 31 December 2006 (n=27,441).

The study region was defined as the south-west metropolitan area of Perth, Western Australia and was selected because it represented a residential area traversed by a combination of both high and low volume roads, thereby providing a good degree of exposure contrast. There is minimal exposure to industrial emissions. The region also spans low to high socio-economic areas. In 2006, the population of the region was 269,734 (2006 Census of Population and Housing). Socioeconomic status was assigned

using the advantage-disadvantage score of the Australian Bureau of Statistics (ABS) Socioeconomic Index for Areas. These scores were classified into tertiles and assigned according to Census collection district based on maternal residence at the time of birth. Census collection districts contain, on average, 225 dwellings.

Data sourcesMidwife notifications, birth registrations, death registrations,

birth defect registrations and hospital morbidity records were

obtained from the Western Australian Department of Health. These

clinical records provided for complete coverage of births within

the state. The midwife notifications contain birth notifications

completed by the midwife in attendance, the medical officer, or

the first qualified midwife or medical officer in their absence. It

contains a completed record for every baby born, either stillborn

or live born of 400 g or more birth weight and/or 20 weeks or

more gestation, occurring in WA. The collection includes data on

maternal demographic characteristics, maternal health, pregnancy,

labour, delivery and neonatal outcomes.

The Western Australian Department of Health obtains data on

birth defects from the WA Birth Defects Registry, a comprehensive

collection of structural or functional abnormalities that are present

from conception or occur before the end of pregnancy, and

diagnosed by six years of age.

It was computationally infeasible to obtain accurate exposure

estimates for the entire study region due to the large total length

of major roads in the region and the long duration of the study

period. Therefore, three smaller study areas were defined within

the study region, corresponding to each socioeconomic tertile. A

study area was defined for each socioeconomic tertile by selecting

the largest conjoint area that intersected a main road (>25,000

vehicles/day). This strategy may be defined as geographic cluster

sampling – study areas were maximised, using the distribution

of main roads and socioeconomic status as constraints. This

approach maximised sample size and resulted in three study areas

corresponding to low, mid and high socioeconomic status.

Exclusion criteriaOnly neonates whose mothers resided inside one of the three

areas were included in the study (n=3,501). Neonates with a

known pathological determinant of birth weight were excluded.

Specifically, the exclusion criteria were for neonates: multiple

births, births before 37 weeks of gestation, diagnosis of a birth

defect (n=515); and for mothers: those with pre-existing or

gestational diabetes, and mothers of an Aboriginal or Torres

Strait Islander ethnic origin (n=299). Stillborn neonates were

excluded because their growth was more likely influenced by

other pathological causes (n=23). Neonates were also excluded if

their maternal records indicated complications of pregnancy such

as threatened abortion, threatened preterm labour, urinary tract

infections, pre-eclampsia, placenta praevia/abruption, and pre-

labour rupture of membranes (n=1,435). As some neonates met

multiple exclusion criteria, there were 1,637 neonates excluded in

Pereira et al. Article


total. These exclusions made the study areas more homogeneous

in terms of their socioeconomic status and hence the study areas

were referred to as ‘A’, ‘B’ and ‘C’.

Exposure assessmentThe aim of the exposure assessment was to characterise the

spatially localised variation in traffic emissions (trafficLCO

). We

computed 24-hour CO concentrations based on the location of

maternal residence recorded at the time of birth. The duration

and timing of pregnancy was based on date of delivery minus

gestational age. Exposures at various time points in pregnancy

were determined. These included trimester exposure, whole

pregnancy exposure, and pre-pregnancy exposure defined as

exposure in the 12 weeks prior to the last menstrual period.

Exposure levels for trafficLCO

were determined for each hour

of each day of pregnancy by applying AusRoads, a line source

Gaussian plume dispersion model using identical algorithms to

those developed for the US CALINE-4 model.25,26 CALINE-4 has

been validated for a variety of traffic emissions,26-28 including CO29

and has been applied recently to a pregnancy outcome.16 AusRoads

was selected over CALINE-4 as it allowed a larger number of

road links and receptor locations, and accepted meteorological

information in an Australian (AusPlume) format. Daily traffic

counts for each road were obtained from Main Roads WA for

the period 1999-2006 and converted into hourly values using

factors developed for Perth by the Department of Environment

and Conservation.

We obtained CO emission factors (g/km) for each vehicle class

and fuel type from the National Pollution Inventory.30 These

emission factors were weighted by the proportion of the fleet that

use the fuel type,31 and by the proportion of vehicle kilometres

travelled among the Perth vehicle fleet.32 Hourly wind speeds,

wind directions and temperatures were obtained from the closest

monitoring site (site no. 9172) from the Bureau of Meteorology

for the period 1999-2006. Hourly mixing depths and Pasquill

atmospheric stability classes were obtained from the Department

of Environment Western Australia for the study region for the

same period.

IUGFetal growth was defined using POBW, a ratio of the observed

birth weight to the individual optimal birth weight.24 Optimal fetal

growth is most likely achieved in the absence of maternal or fetal

pathology, or exposures that can pathologically affect fetal growth.

Blair et al (2005) proposed a model for optimal birth weight by

using non-pathological determinants of growth as predictors of

birth weight after restricting births for which there existed no

recorded pathological determinants. The model for optimal birth

weight which takes into account gestational age, maternal height,

birth order and neonate gender as predictors was independently

developed using Western Australian births in 1998-2002. The

POBW measure has been applied in a range of applications

investigating infant and child health and development outcomes

such as childhood leukemia,33 non-central nervous system

tumours,34 children’s numeracy and literacy,35 delayed language

development,36 birth defects37 and mental health.38

Statistical methods and adjustmentA mixed-effects multivariable regression model was applied

to investigate the effect of trafficLCO

on fetal growth. This model

accounted for correlation of POBW within families and analyses

were performed using Proc Mixed, SAS v9.1.39 We adjusted

for known risk factors and confounders including maternal

smoking during pregnancy, marital/de facto status and distance

to industrial areas. As CO levels at the residential location are

temporally correlated with risk factors that also change with

season or meteorological conditions, we also adjusted for city-wide

background pollution such as background levels of particulate

matter (PM2.5

, PM10

), nitric oxide, nitrogen dioxide, tropospheric

ozone, and CO; and season of birth. The mean daily background

concentrations averaged across each pregnancy period was used as

for adjustment. Separate analyses were conducted for each study

area and timing of pregnancy. Interactions between exposure and

maternal smoking during pregnancy (binary classification) were

also investigated.

Deviations from the optimal birth due to maternal smoking

during pregnancy and for an interquartile increase in exposure

were compared. In this analysis we assessed exposure during the

third trimester, when the fetal growth rate is highest. We selected

four optimal birth weights 500 g apart, starting with the cut-off

for LBW (2,500 g).

AssumptionsAll analyses were based on the following assumptions: (i)

the impact of women moving to different residential addresses

during pregnancy was minimal, (ii) there was minimal outcome

misclassification resulting from occupational and other exposures

away from the home, and (iii) effect sizes were not affected by

a small group of statistically influential neonates. To ascertain

the validity of these assumptions we performed three sensitivity

analyses on all cohorts where a statistically significant effect was

observed. To check assumption (i) we used electoral roll records

to ascertain whether a mother moved during pregnancy, removed

these records, and repeated the analyses in order to quantify the

magnitude of the influence of such exposure misclassification on

effect estimates. We restricted analysis to those that did not move

rather than compare (or re-calculate) exposures to those that did

move because electoral roll records are updated infrequently. In

the case of assumption (ii) although it is generally considered that

pregnant women spend more time at home during the latter stages

of pregnancy,40 we compared results with the effect sizes for third

trimester exposure among neonates whose mothers listed variants

of ‘home duties’ as their occupation. Finally, to check assumption

(iii) we used the dfβ statistic which is a measure of the influence

of a particular neonate on the overall effect estimate.41 We defined

highly influential neonates as those having a dfβCO

more than 1.5

times the inter-quartile range below the first quartile or above the

third quartile, then removed these observations and repeated the

Fetal Exposure Traffic emissions associated with reduced fetal growth in WA


The confidence intervals for the effect estimates resulting from

each of the three sensitivity analyses were plotted along with that

corresponding to the original findings for comparison.

Ethics approval for this study was obtained from the University

of Western Australia (Project: RA/4/1/2006) and the Department

of Health Western Australian Human Research Ethics Committee

(Project: 2008/7).

ResultsThere were 3,501 neonates born between January 2000 and

December 2006 in the three study areas. Application of the

participant exclusion criteria resulted in a final sample size of

1,864 neonates. The sample size in areas A and B were 826 and 673

respectively. Although the sampling strategy aimed at maximising

the sample size area C had a low sample size of 365 neonates.

Table 1 presents a summary of population characteristics.

Generally, neonate characteristics were similar across the study

areas. However, maternal smoking during pregnancy was higher in

area A than the other areas (n=196, 24%, p<0.001). The proportion

of women not in a relationship was also higher in area A than the

other areas (n=93, 11%, p<0.001). The mean POBW score was also

lowest in study area A (mean=100%, sd=12%, p<0.001). The mean

distance of maternal residence to an industrial area was greatest

in area C (mean=4,591 m, sd=1,123 m, p<0.001). Although the

estimated average exposure to trafficLCO

over the duration of

pregnancy was similar for areas A and C, it was considerably lower

in area B (mean=27 µg/m3, sd=20 µg/m3, p<0.001). This can be

partially explained by the marginally lower levels of traffic and

the orientation of the road. The major road in area B was nearly

perpendicular to the main roads in the other areas, resulting in a

different road-wind angle. Moreover, the major roads in the other

areas were much closer to the wind angle. The CALINE algorithm

tends to over-predict concentrations under these conditions.42 As

this resulted in differential exposure misclassification we focussed

on the results for area B only.

There was a low proportion (7-12%) of background exposure to

traffic-related CO attributable to trafficLCO

exposure from the main

road in the study areas. The background levels of pollutants were

similar across the three areas because the concentrations across

all stations were averaged over the pregnancy period in order to

ensure sufficient distinction between exposure from ‘local’ and

‘area-wide’ levels.

Table 1: Summary of maternal, neonate and environmental characteristics by study area.

Study area

A (n=826)

n (%)

B (n=673)

n (%)

C (n=365)

n (%)

Demographic characteristics

Maternal

Maternal height (cm) (mean, SD)

164 (7) 164 (7) 165 (7)

Smoke during pregnancy

Yes 196 (24) 59 (9) 19 (5)

No 630 (76) 614 (91) 346 (95)

Married/de-facto

Yes 733 (89) 632 (94) 348 (95)

No 93 (11) 41 (6) 17 (5)

Neonate

POBW (%) (mean, SD) 100 (12) 102 (11) 102 (12)

Gestation (wks) (mean, SD)

39 (1) 39 (1) 39 (1)

Neonate gender

Male 416 (51) 329 (49) 196 (54)

Female 410 (50) 344 (51) 169 (46)

Season of birth

Summer 196 (24) 165 (25) 88 (24)

Autumn 183 (22) 156 (23) 93 (25)

Winter 216 (26) 169 (25) 101 (28)

Spring 231 (28) 183 (27) 83 (23)

Environmental characteristics

Motor vehicle CO during preg. (µg/m3) (mean, SD)

45 (34) 27 (20) 46 (42)

Distance to industry (m) (mean, SD)

1295 (383) 1234 (555) 4591 (1123)

Background pollution during preg. (µg/m3) (mean, SD)

CO 379 (37) 378 (37) 377 (34)

NO 7.73 (1.18) 7.66 (1.15) 7.65 (1.08)

NO2 12.08 (0.66) 12.02 (0.64) 12.08 (0.64)

O3 38.32 (1.02) 38.34 (1.00) 38.38 (0.98)

PM10 17.09 (0.78) 17.05 (0.78) 17.13 (0.80)

PM2.5 8.00 (0.32) 8.00 (0.32) 8.00 (0.32)

Table 2: Adjusted percentage change in POBW (β) for a 10 µg/m3 increase in trafficLCO by exposure period for study area B.

Unadjusted Adjusted

Exposure period n β SE p n β SE p

Pregnancy 541 -0.40 0.24 0.100 529 -0.42 0.25 0.086

Pre-pregnancy 473 -0.35 0.22 0.108 461 -0.38 0.24 0.110

Trimester 1 490 -0.18 0.22 0.407 478 -0.24 0.24 0.313

Trimester 2 518 -0.05 0.21 0.813 506 -0.08 0.22 0.719

Trimester 3 541 -0.53 0.22 0.015 529 -0.49 0.23 0.033

analysis to quantify their influence on the effect estimates. As

restriction to sub-populations naturally results in wider confidence

intervals, unadjusted effect sizes were compared to maximise

the precision of the estimates resulting from sensitivity analyses.



For area B, the percentage change in POBW for a 10 µg/m3

increase in trafficLCO

is shown in Table 2. POBW decreased by

0.42% (sd=0.25) with every 10 µg/m3 increase in trafficLCO

over

the full pregnancy. There appeared to be a stronger negative effect

in POBW with increasing exposure in third trimester. Exposure

in the third trimester was associated with a -0.49% (sd=0.23%)

change in POBW score. There was strong correlation between

pre-pregnancy and third trimester exposures (Pearson correlation

0.74, p<0.0001). Moreover, trafficLCO

exposures over all pregnancy

periods were highly correlated, which weakened the distinction

between the trimester specific results (Table 3). Whole pregnancy

trafficLCO

exposure was not correlated with any of the background

pollutants over the same period.

No significant interactive effect was observed between maternal

smoking during pregnancy and third trimester trafficLCO

exposure.

The influence of exposure on the resultant birth weight depends

on the optimal rate of growth in utero. We selected four optimal birth

weights at 500 g intervals starting with the LBW cut-off weight of

2500 g. We looked at deviations from these optimal birth weights

due to maternal smoking during pregnancy, and separately due to

the effect of an inter-quartile increase in third trimester trafficLCO

(Table 4). The inter-quartile range (IQR) for exposure to trafficLCO

during pregnancy for all study areas was 31 µg/m3 (lower quartile

17 µg/m3, upper quartile 48 µg/m3). Effect estimates from the

univariate models (Table 2) were applied to obtain deviations from

the optimal birth weight as the multivariate effect estimates were

adjusted for maternal smoking status. This decision minimally

affected the results as the unadjusted and adjusted effect sizes

were similar in magnitude. A neonate who would have otherwise

attained an optimal birth weight of 3.5 kg would be expected to be

born 58 g lighter for an IQR increase in third trimester trafficLCO

.

By comparison, the effect of maternal smoking during pregnancy

would result in a 127 g reduction from this optimal birth weight.

Sensitivity analyses were performed to assess the robustness of

the effect observed for third trimester exposure. Unadjusted effect

sizes and their 95% confidence intervals corresponding to a 10

µg/m3 increase in third trimester trafficLCO

were calculated (Figure

1) and we found considerable overlap among these confidence

intervals. Analysis of each of the pre-specified assumptions

revealed that restriction of analysis to neonates whose mothers did

not change residential address during the study period resulted in

Figure 1: 95% confidence intervals for sensitivity analyses: Unadjusted effects for a 10µg/m3 increase in trafficLCO on POBW(%) among the mid-socioeconomic cohort.

all – no restriction (N=673)

(i) – restriction to infants whose mothers did not move during pregnancy (N=405)

(ii) – restriction to infants whose mothers spent more time at home (N=68)

(iii) – restriction by removal of statistically influential infants (N=606)

Table 1: Summary of maternal, neonate and environmental characteristics by study area

Study area

A (n=826) B (n=673) C (n=365)

N (%) N (%) N (%)

Demographic characteristics

Maternal

Maternal height (cm) (mean, SD) 164(7) 164(7) 165(7)

Smoke during pregnancy

Yes 196 (24) 59 (9) 19(5)

No 630 (76) 614 (91) 346 (95)

Married/de-facto

Yes 733 (89) 632 (94) 348 (95)

No 93 (11) 41 (6) 17 (5)

Neonate

Figure 1: 95% confidence intervals for sensitivity analyses: Unadjusted effects for a 10 µg/m3 increase in traffic LCO on POBW (%) among the mid-socioeconomic cohort.

all no restriction (n=673).

(i) restriction to infants whose mothers did not move during pregnancy (n=405).

(ii) restriction to infants whose mothers spent more time at home (n=68).

(iii) restriction by removal of statistically influential infants (n=606).

Table 3: Correlations between trafficLCO concentrations with p-values in parentheses.

Pregnancy Pre-pregnancy Trimester 1 Trimester 2 Trimester 3Pregnancy 0.91 0.92 0.91 0.90

Pre-pregnancy (<0.0001) 0.83 0.67 0.79

Trimester 1 (<0.0001) (<0.0001) 0.81 0.67

Trimester 2 (<0.0001) (<0.0001) (<0.0001) 0.80

Trimester 3 (<0.0001) (<0.0001) (<0.0001) (<0.0001)

Table 4: Comparison between the effect of maternal smoking during pregnancy and an IQR increase (31 µg/m3) in third trimester trafficLCO on birth weight (bw).

Mother smoked during pregnancy IQR increase in 3rd trimester trafficLCO

Optimal bw (g) Reduction (g) Consequential bw (g) Reduction (g) Consequential bw (g)

2,500 -91 2,409 -41 2,459

3,000 -109 2,891 -50 2,950

3,500 -127 3,373 -58 3,442

4,000 -145 3,855 -66 3,934



a weaker effect (compared with the original analysis); a 10 µg/m3

increase in third trimester trafficLCO

was associated with a change in

POBW of -0.35% (-0.92%, 0.22%). Limiting analysis to neonates

of mothers who spent more of their time at home resulted in a

stronger effect than that originally observed: a reduction in POBW

of -0.71% (-1.65%, 0.23%). The exclusion of highly influential

observations also led to a stronger effect than that originally

observed: a decline in POBW of -0.76% (-1.31%, -0.21%).

DiscussionThis is the first study to investigate the association between local

traffic emissions and IUG, accounting for constitutional growth

potential. We observed an association between third trimester

exposure and IUG among a population which was free of the other

major determinants of fetal growth. A reduction in POBW of 0.5%

per 10 µg/m3 trafficLCO

was observed. In many urban settings, the

difference in exposure between women living close to a major road

and those who live much further from the road will be significantly

greater than 10 µg/m3. Therefore, we investigated the influence of

an interquartile range increase in trafficLCO

exposure and detected a

reduction in birth weight that equated to approximately half of that

observed for maternal smoking during pregnancy.

The study was intentionally designed to obtain effect estimates

independent of other known risk factors, which increased

homogeneity between the three areas and potentially biased the

study population towards those with higher socio-economic

characteristics. This issue was partially addressed by targeting the

selection of three geographically defined socioeconomic groups.

Exposure is potentially correlated with risk factors that also change

with proximity to major roads, including socioeconomic status

and proximity to sources of industrial pollution. To minimise the

influence of such factors, we stratified analyses by socioeconomic

status and adjusted for distance to industrial sources of air pollution.

A significant effect was only observed for the cohort in area B.

Among a population from which those with known risk factors

have been removed, one would not necessarily expect to observe

an effect in area A, the low socioeconomic area. We also note that

the prevailing wind direction was closer to parallel the major roads

in these two areas. The CALINE algorithm tends to over-predict

concentrations in this situation and hence a degree of exposure

misclassification would have resulted. A limitation of the study

was that the sample size in the high socioeconomic area (C) was

approximately half that of the other two groups. This resulted in a

considerable loss of statistical power in the analysis conducted for

this group. A Canadian study has demonstrated that higher family

income and employment are associated with fewer hours spent at

home, indicating that personal residential exposure may be less

relevant for the high-socioeconomic group. Lack of detection of an

effect may also indicate a lower penetration factor of ambient air

into the homes among this group. The non-detection of an effect in

the low socioeconomic area (A) may be explained by our inability

to exclude factors not documented on clinical records. For example

alcohol consumption, substance abuse during pregnancy, passive

smoking and undiagnosed maternal diabetes all affect fetal growth43

and are potentially more prevalent among this group. Moreover, it is

possible that traffic emissions may have contributed to a condition

used as an exclusion criterion in our study. For example, exposure

to traffic emissions may have contributed to stillbirth or the presence

of a birth defect while dependently or independently restricting

fetal growth. By excluding such cases, those most susceptible to

environmental insults were removed at the expense of obtaining

independent effect estimates. We acted on the assumption that

this situation was unlikely and was outweighed by the benefits of

excluding such records.

To summarise, a large degree of the socioeconomic differences

across the three areas would have been removed through the

exclusion criteria. Therefore, the fact that an effect was not observed

in two of the three study areas was more likely due to unavoidable

exposure misclassification in these two areas. The other possibility

is that effects were not observed in some of the areas due to the

lack of the existence of an underlying true association. However,

the effect that was observed in one of the areas may be evidence

against this hypothesis.

It should be noted that the high correlation among the exposures

across the periods in pregnancy implies that the results shown in

Table 2 are necessarily dependent. For example, the pre-pregnancy

period and third trimester are likely to have a large degree of seasonal

overlap. Such correlation may have resulted in the subsequent

inflation of the pre-pregnancy effect estimate. A consequence of the

high correlations among exposures across the periods in pregnancy

is that the observed significant effect for third trimester exposure

may be spurious. However, it may cautiously be interpreted as a true

effect based on the combination of biological knowledge (elevated

vulnerability of the fetus in relation to IUG), the fact that it was the

only statistically significant association observed, and the fact that

it had the highest effect size.

Although we specifically investigated the association between

IUG and CO as a marker for locally derived traffic emissions,

findings from ambient air pollution studies are also relevant. In

a review of the effects of air pollution on fetal growth, all three

studies that assessed CO reported positive findings, two of which

reported associations in the third trimester.6 An adverse association

between third trimester CO and low birth weight was observed in a

US study that addressed the local heterogeneity of the pollutant.44 An

earlier study revealed a link between low birth weight and distance

weighted traffic density was observed in the same geographical

area.45 In a study conducted in California, a 22% increased risk

of low birth weight was observed among neonates exposed to

ambient third trimester CO greater than 5.5 ppm compared with

the lowest exposure group.46 A 22 g decrease in birth weight per

1.4 ppm change in first trimester ambient CO was observed among

an earlier Californian birth cohort.47 Among a Canadian cohort, an

increased risk of being born small for gestational age and sex was

observed for all trimester exposures in a region where the mean CO



level was 1.1 ppm, lower than that of the Californian studies.48 The

observed increase in risk peaked for exposure in the ninth month of

pregnancy at 23% per 1 ppm of ambient CO because the ability of

CO to diffuse through the placenta increases with placental blood

flow and hence gestational age. An observational study conducted

in France found that birth weight decreased with maternally expired

CO greater than 5 ppm, but did not investigate the relationship

below this level.49 However in a separate study, no reduction in birth

weight was observed after pregnant rats were exposed to higher

levels (30-90 ppm) of CO.50 We observed an adverse association

between fetal growth at much lower levels, possibly explained by

CO being a marker for other traffic emissions or a mixture thereof.

This possibility is reinforced by the non-detection of an effect with

background CO and also because the levels of exposure were a

small fraction of the background CO levels. Moreover, the trafficLCO

levels were much lower for the area in which we observed an effect

than the other areas in which we did not detect an effect. A recent

animal study detected an association between reduced fetal weight

and traffic emissions, excluding CO.51

ConclusionOur results, in context with the previous epidemiological

literature, indicate that an adverse association between local

residential traffic emissions and IUG was observed in a population

from which other major determinants of IUG had been deliberately

excluded. For an interquartile increase in exposure, we detected a

reduction in birth weight that equates to approximately half of that

observed for maternal smoking during pregnancy. The stability

of this effect was confirmed by sensitivity analyses. As this was

an observational study, we can not exclude the possibility that

the association observed was due to unknown factors uniquely

correlated with this cohort. Therefore, further research is required

to corroborate our results.

References1. Pereira G, Nassar N, Bower C, Weinstein P, Cook A. Residential exposure to

traffic emissions and adverse pregnancy outcomes. Surveys and Perspectives Integrating Environment & Society. 2010;3(1):41-64.

2. Gilbert NL, Woodhouse S, Stieb DM, Brook JR. Ambient nitrogen dioxide and distance from a major highway. Sci Total Environ. 2003;312(1-3):43-6.

3. Janssen NAH, van Vliet PHN, Aarts F, Harssema H, Brunekreef B. Assessment of exposure to traffic related air pollution of children attending schools near motorways. Atmospheric Environment. 2001;35(22):3875-84.

4. Zhu Y, Hinds WC, Kim S, Sioutas C. Concentration and Size Distribution of Ultrafine Particles Near a Major Highway. J Air Waste Manag Assoc. 2002;52(9):1-2.

5. Glinianaia SV, Rankin J, Bell R, Pless-Mulloli T, Howel D. Particulate air pollution and fetal health: a systematic review of the epidemiologic evidence. Epidemiology. 2004;15(1):36-45.

6. Maisonet M, Correa A, Misra D, Jaakkola JJK. A review of the literature on the effects of ambient air pollution on fetal growth. Environ Res. 2004;95(1):106-15.

7. Slama R, Darrow L, Parker J, Woodruff TJ, Strickland M, Nieuwenhuijsen M, et al. Meeting report: Atmospheric pollution and human reproduction. Environ Health Perspect. 2008;116(6):791-8.

8. Sram RJ, Binkova B, Dejmek J, Bobak M. Ambient air pollution and pregnancy outcomes: a review of the literature. Environ Health Perspect. 2005;113(4):375-82.

9. Aguilera I, Guxens M, Garcia-Esteban R, Corbella T, Nieuwenhuijsen MJ, Foradada CM, et al. Association between gis-based exposure to urban air

pollution during pregnancy and birth weight in the INMA Sabadell cohort. Environ Health Perspect. 2009;117(8):1322-7.

10. Brauer M, Lencar C, Tamburic L, Koehoorn M, Demers P, Karr C. A cohort study of traffic-related air pollution impacts on birth outcomes. Environ Health Perspect. 2008;116(5):680-6.

11. Slama R, Morgenstern V, Cyrys J, Zutavern A, Herbarth O, Wichmann HE, et al. Traffic-related atmospheric pollutants levels during pregnancy and offspring’s term birth weight: a study relying on a land-use regression exposure model. Environ Health Perspect. 2007;115(9):1283-92.

12. Pereira G, Cook A, De Vos AJBM, Holman D. A case-crossover analysis of traffic-related air pollution and emergency department presentations for asthma in Perth, Western Australia. Med J Aust. 2010;193(9):511-4.

13. Environmental Protection Authority. State of the Environment Report: Western Australia. Perth (AUST): State Government of Western Australia; 2007

14. Alderman BW. Maternal exposure to neighborhood carbon monoxide and risk of low infant birth weight. Public Health Rep. 1987;102(4):410.

15. Bell ML, Ebisu K, Belanger K. Ambient air pollution and low birth weight in Connecticut and Massachusetts. Environl Health Perspect. 2007;115(7):1118-24.

16. Wu J, Ren C, Delfino RJ, Chung J, Wilhelm M, Ritz B. Association between local traffic-generated air pollution and preeclampsia and preterm delivery in the South Coast Air Basin of California. Environ Health Perspect. 2009;117(11):1773-9.

17. Maulik DEV. Fetal growth restriction: the etiology. Clin Obstet Gynecol. 2006;49(2):228.

18. Bernstein PS. Etiologies of fetal growth restriction. Clin Obstet Gynecol. 1997;40(4):723.

19. Skjaerven R, Skjærven. Birthweight by gestational age in Norway. Acta Obstet Gynecol Scand. 2000;79(6):440.

20. Voorhorst FJ. Maternal characteristics and expected birth weight. Eur J Obstet Gynecol Reprod Biol. 1993;50(2):115.

21. Blair E. Why do Aboriginal newborns weigh less? Determinants of birthweight for gestation. J Paediatr Child Health. 1996;32(6):498.

22. Gardosi J. Customised antenatal growth charts. Lancet. 1992;339(8788):283.23. Sanderson D, Wilcox M, Johnson I. The individualised birth weight ratio: a new

method of identifying intrauterine growth retardation. BJOG. 1994;101(4):310-4.24. Blair E, Liu Y, de Klerk N, Lawrence D. Optimal fetal growth for the Caucasian

singleton and assessment of appropriateness of fetal growth: an analysis of a total population perinatal database. BMC Pediatr. 2005;5(1):13.

25. Benson P. CALINE4: A Dispersion Model for Predicting Air Pollutant Concentrations near Roadways. Sacramento (CA): California Department of Transportation; 1989.

26. Benson P. A review of the development and application of the Caline3 and Caline4 models. Atmospheric Environment Part B. Urban Atmosphere. 1992;26(3):379–90.

27. Gramotnev G. Determination of average emission factors for vehicles on a busy road. Atmospheric Environment. 2003;37(4):465.

28. Levitin J, Levitin. Evaluation of the CALINE4 and CAR-FMI models against measurements near a major road. Atmospheric Environment. 2005;39(25):4439.

29. Broderick BM, Budd U, Misstear BD, Ceburnis D, Jennings SG. Validation of CALINE4 modelling for carbon monoxide concentrations under free-flowing and congested traffic conditions in Ireland. International Journal of Environment and Pollution. 2005;24(1):104.

30. Environment Australia. Emission Estimation Technique Manual for Aggregated Emissions from Motor Vehicles. Canberra (AUST): National Pollutant Inventory, Department of Sustainability, Environment, Water, Population and Communities; 2000.

31. Australian Bureau of Statistics. 9309.0 – Motor Vehicle Census, Australia, 31 March 2008. Canberra (AUST): ABS; 2009.

32. Department of Environment and Conservation. Clean-Run – Perth’s Vehicle Fleet [Fact Sheet]. Perth (AUST): State Government of WesternAustralia; 2008.

33. Milne E, Laurvick CL, Blair E, Bower C, de Klerk N. Fetal Growth and Acute Childhood Leukemia: Looking Beyond Birth Weight. Am J Epidemiol. 2007;166(2):151-9.

34. Laurvick CL, Milne E, Blair E, de Klerk N, Charles AK, Bower C. Fetal growth and the risk of childhood non-CNS solid tumours in Western Australia. Br J Cancer. 2008;99(1):179-81.

35. Malacova E, Li J, Blair E, Leonard H, de Klerk N, Stanley F. Association of Birth Outcomes and Maternal, School, and Neighborhood Characteristics with Subsequent Numeracy Achievement. Am J Epidemiol. 2008;168(1):21-9.

36. Zubrick SR, Taylor CL, Rice ML, Slegers DW. Late language emergence at 24 months: an epidemiological study of prevalence, predictors, and covariates. Journal of Speech, Language and Hearing Research. 2007;50(6):1562-92.

37. Nassar N, Abeywardana P, Barker A, Bower C. Parental occupational exposure to potential endocrine disrupting chemicals and risk of hypospadias in infants. Occup Environ Med. 2009;67(9):585-9.

38. Oddy WH, Kendall GE, Li J, Jacoby P, Robinson M, de Klerk NH, et al. The Long-Term Effects of Breastfeeding on Child and Adolescent Mental Health: A Pregnancy Cohort Study Followed for 14 Years. J Pediatr. 2009;156(4):568-74.



39. SAS: statistical software [computer program]. Service Pack 4. Cary (NC): SAS Institute; 2003.

40. Nethery E, Brauer M, Janssen P. Time-activity patterns of pregnant women and changes during the course of pregnancy. J Expos Sci Environ Epidemiol. 2008;19(3):317-24.

41. Belsley DA, Kuh AE, Welsch RE. Regression Diagnostics. New York (NY): Wiley; 1980.

42. Sivacoumar R, Thanasekaran K. Comparison and performance evaluation of models used for vehicular pollution prediction. Journal of Environmental Engineering (New York, N.Y.). 2001;127(6):524.

43. Ergaz Z, Avgil M, Ornoy A. Intrauterine growth restriction--etiology and consequences: What do we know about the human situation and experimental animal models? Reprod Toxicol. 2005;20(3):301-22.

44. Wilhelm M, Ritz B. Local variations in CO and particulate air pollution and adverse birth outcomes in Los Angeles County, California, USA. Environ Health Perspect. 2005;113(9):1212-21.

45. Wilhelm M, Ritz B. Residential proximity to traffic and adverse birth outcomes in Los Angeles County. Environ Health Perspect. 2003;111:207-16.

46. Ritz B, Yu F. The effect of ambient carbon monoxide on low birth weight among children born in southern California between 1989 and 1993. Environ Health Perspect. 1999;107(1):17.

47. Salam MT, Millstein J, Yu-Fen L, Lurmann FW, Margolis HG, Gilliland FD. Birth outcomes and prenatal exposure to ozone, carbon monoxide, and particulate matter: Results from the Children’s Health Study. Environ Health Perspect. 2005;113(11):1638.

48. Liu S, Krewski D, Shi Y, Chen Y, Burnett RT. Association between maternal exposure to ambient air pollutants during pregnancy and fetal growth restriction. J Expo Sci Environ Epidemiol. 2007;17(5):426-32.

49. Gomez C, Berlin I, Marquis P, Delcroix M. Expired air carbon monoxide concentration in mothers and their spouses above 5 ppm is associated with decreased fetal growth. Prev Med. 2005;40(1):10-5.

50. Garvey DJ. Chronic low level maternal carbon monoxide exposure and fetal growth and development. Biol Reprod. 1978;19(1):8.

51. Silva IRR, Lichtenfels AJFC, Pereira LAA, Saldiva PHN. Effects of ambient levels of air pollution generated by traffic on birth and placental weights in mice. Fertil Steril. 2008;90(5):1921-4.

Safe and HealthierFood for AllAustralians

Registration Now Open

Following the success of the inaugural FoodFutures Conference in 2010, the PHAA invitesyou to participate in the 2011 conference.We aim to continue the theme of ‘joined up

food policy’, recognising that food production,food access and healthy eating are strongly

linked and need overarching policy andleadership to achieve better health

outcomes.

ood utures

Public Health Association Australia2nd National Food Futures Conference

22-23 November 2011 Grand Chancellor, Hobart

traffic emissions are associated with reduced fetal growth in areas of perth, western australia: an...

Documents