Epidemiological studies addressing the relation between ambient air pollution and fetal development are accumulating worldwide. Studies conducted in China,¹ the Czech Republic,^2,3 Korea,⁴ the UK,⁵ Brazil,⁶ and North America^7,8,9,10,11 have examined the link between ambient air pollution levels during pregnancy and reduction in birth weight or intrauterine growth retardation (IUGR). There is evidence to suggest that air pollutant exposure during pregnancy has an adverse impact on birth weight, although the findings are inconsistent; the effect of individual pollutants and the period(s) during pregnancy when pollutant levels are likely to have most impact on birth weight is not clear.

It is useful, then, to examine the impact of ambient air pollution and birth weight in a variety of different sites to clarify the nature of the relation. We examined this relation in Sydney as there is complete, routinely collected data available. In the present study we evaluated the effect of prenatal exposure (in early, mid, and late pregnancy) to five common urban air pollutants: particulate matter (PM₁₀ and PM_2.5), carbon monoxide (CO), nitrogen dioxide (NO₂), and ozone (O₃) from routine air monitoring in the Sydney metropolitan area on birth weight.

METHODS

Information on all births in metropolitan Sydney between 1 January 1998 and 31 December 2000 was obtained from the Midwives Data Collection (MDC) at the New South Wales Department of Health. The MDC is a population based surveillance system covering all live births and stillbirths of at least 20 weeks gestation and at least 400 grams birth weight. Birth data includes maternal demographic factors (age, smoking status, country of birth, postcode of residence at time of delivery), pregnancy factors (date of the last menstrual period, gestational hypertension and diabetes, parity, time of first antenatal visit to a healthcare provider), details about the delivery (type of delivery), and infant factors (birth weight, gestational age). Other known causes of low birth weight including multiple births, hypertension of pregnancy, and gestational diabetes were excluded.

We obtained daily air pollution data from the New South Wales Environment Protection Authority (EPA) and daily meteorological data from the Australian Bureau of Meteorology. Air pollution data included concentrations of PM₁₀, PM_2.5, NO₂, O₃, and CO recorded at 14 monitoring stations in metropolitan Sydney. Monitoring stations were excluded if less than 80% of readings were available for each pollutant. Pollutant concentrations were analysed as continuous variables.

For each birth, exposure to each air pollutant during gestation was estimated by calculating the average of each pollutant over 30 days (last month) and 90 days (third trimester) before birth, the mid 90 days of gestation (second trimester), and 90 days (first trimester) after the estimated date of conception. Pollutant concentrations from monitoring stations were averaged to provide an estimate for the whole of metropolitan Sydney, using an approach similar to the APHEA2 (Short-term Effects of Air Pollution and Health: A European Approach) studies.^12–14 The process was repeated matching air pollution concentrations from each eligible monitoring station and births to mothers residing in postcodes within 5 km of the monitoring station.

The following covariates were included in regression models: sex of child, maternal age (in year groupings), gestational age (only included in linear regression models), maternal smoking (yes/no), gestational age at first antenatal visit (⩽20 weeks or >20 weeks), maternal indigenous status (whether mother identifies as being Aboriginal or Torres Strait Islander), whether first pregnancy, season of birth, and socioeconomic status (SES). SES was measured using the Index of Relative Socioeconomic Disadvantage (IRSD) of postcode of maternal residence. The Australian Bureau of Statistics (ABS) constructs the IRSD to classify geographical areas on the basis of social and economic information collected in the population census. Each postcode in New South Wales (population of approximately 1000–5000 people) is assigned an IRSD index. We then ranked the postcodes and divided the list into quintiles.

A variable “small for gestational age” (SGA) was calculated based on the Australian national birth weight centile for gestational age 1991–94.¹⁵ SGA was defined as greater than two standard deviations below the mean birth weight according to gestational age as SGA. SGA was analysed as a categorical variable in logistic regression models using the SAS System for Windows v8.02. Linear regression models were also developed using birth weight as a continuous variable. Following the development of the basic model, air pollutants were then added to the model to determine the association between ambient air pollutants and birth weight. Single and multi-pollutant pollutant models were assessed and we investigated interactions between air pollution variables and covariates. The impact of pollutant exposures in other pregnancy periods on key findings was also analysed.

RESULTS

There were 138 056 singleton births in Sydney between 1998 and 2000; 9.7% of babies (13 402) were classified as SGA. Mean (SD) birth weight for babies born in Sydney between 1998 and 2000 was 3418 (531) grams. After adjusting for other maternal and infant characteristics, SGA was significantly associated with maternal smoking, gestational age greater than 20 weeks at first antenatal visit, and first pregnancy (table 1). All SES variables exhibit a statistically significant association with SGA. Compared to the high SES category, there was increasing risk of SGA with decreasing SES.

Pollutant concentrations in Sydney from 1 April 1997 to 31 December 2000 are presented in table 2. A correlation matrix of air pollution variables is presented in table 3 to show correlations between pollutants. Correlation coefficients above 0.8 were observed between PM₁₀ and PM_2.5. Correlation coefficients were also calculated to determine the correlations between air monitoring stations. Correlation coefficients between the 14 monitoring stations ranged from 0.68 to 0.85 for CO. Correlation coefficients between the 14 monitoring stations ranged from 0.46 to 0.85 for NO₂; from 0.53 to 0.94 for O₃; from 0.67 to 0.91 for PM₁₀; and from 0.66 to 0.93 for PM_2.5. All pollutants differed according to season (p < 0.01) and post hoc analysis revealed that all seasons were different to all others (p < 0.05). Concentrations of CO, NO₂, and PM_2.5 were highest in winter and lowest in summer. Concentrations of PM₁₀ and O₃ were highest in summer and lowest in winter.

Citywide average air pollutant concentrations in the last month, third trimester, and first trimester of pregnancy had no statistically significant effect on SGA after adjusting for infant and maternal characteristics (table 4). Concentrations of PM₁₀, PM_2.5, and O₃ in the second trimester of pregnancy had a small but statistically significant adverse effect on SGA (OR 1.01, 95% CI 1.00 to 1.04; OR 1.03, 95% CI 1.01 to 1.05; and OR 1.01, 95% CI 1.00 to 1.01, respectively).

Eleven metropolitan air monitoring stations were eligible for inclusion in the analysis of babies born to women residing within 5 km of an air monitoring station (table 4). There were 51 460 eligible births in 1998–2000. Of these, 5985 (11.6%) were classified as SGA. NO₂ concentrations in the second and third trimesters and one month before birth had a statistically significant adverse effect on risk of SGA (ORs between 1.07 and 1.14) (table 4). The NO₂ effects were robust to controlling for exposures in other pregnancy periods (ORs between 1.1 (95% CI 1.0 to 1.2) and 1.1 (95% CI 1.0 to 1.2) for second and third trimesters respectively). Exposure to PM₁₀ in the second trimester of pregnancy had a small but statistically significant adverse effect on risk of SGA (OR 1.02, 95% CI 1.01 to 1.03) (table 4), although these findings were not robust to analysis when controlling for exposures to PM₁₀ in other periods of pregnancy (OR 1.01, 95% CI 0.99 to 1.02).

In multivariate linear regression models, the effect of pollutant concentrations on birth weight (in grams) was examined for all births in Sydney during the study period and for babies born to women residing within 5 km of an air monitoring station (table 5). Citywide average levels of PM₁₀ and PM_2.5 in the second trimester and last month of pregnancy had a small but statistically significant adverse effect on birth weight. When analysing only babies born to women residing within 5 km of an air monitoring station, we observed a small adverse effect of PM₁₀ exposure during the second trimester on birth weight (regression coefficient −4.3, 95% CI −5.8 to −2.8). Thus, for every 1 μg/m³ increase in 24 hour average PM₁₀ during the second trimester of pregnancy, a 4 gram reduction in birth weight was estimated. This finding persisted after controlling for exposures in other pregnancy periods and analysis in multi-pollutant models (table 6).

Citywide average levels of CO in the last month of pregnancy had a statistically significant adverse effect on birth weight, whereas the 5 km results reveal an adverse effect of CO exposure on birth weight for the second and third trimesters of pregnancy. Linear regression coefficients for the 5 km analysis ranged from −23 (95% CI −44.6 to −1.2) to −29 (95% CI −51.0 to −6.8); thus, for every 1 part per million increase in 8 hour average CO during pregnancy, a 23–29 gram reduction in birth weight was estimated. Analysis of second trimester CO findings in two and three pollutant models and controlling for exposures in other pregnancy periods are presented in table 6. CO findings in the second trimester did not remain statistically significant when analysed in multi-pollutant models.

Citywide average levels of NO₂ in the first and third trimesters of pregnancy resulted in a small but statistically significant adverse effect on birth weight, whereas the 5 km results reveal an adverse effect of NO₂ exposure on birth weight for all periods of pregnancy. Linear regression coefficients for the 5 km analysis ranged from −20 (95% CI −27.8 to −11.5) to −34 (95% CI −43.4 to −24.3). NO₂ findings in the second trimester remained when analysed in two pollutant models, but did not remain when analysed in the four pollutant model (table 6).

DISCUSSION

We examined associations between birth weight and exposure to air pollution concentrations at various stages of pregnancy across all of Sydney and for those infants born to women residing within 5 km of air monitoring stations. We showed a small effect of PM₁₀ concentration in the second trimester of pregnancy, on birth weight in both linear and logistic regression models. We also showed that CO and NO₂ concentrations in the second and third trimesters of pregnancy (most pronounced in the second trimester), had a statistically significant effect on birth weight. The effect of NO₂ concentrations was only seen when examining births to women residing within 5 km of an air monitoring station. The effect of CO concentrations was only seen in the linear regression analysis and was most pronounced when examining births to women residing within 5 km of an air monitoring station.

There are a number of strengths to our study. We were able to investigate the association between low birth weight and pollutant concentrations for a large number of births in metropolitan Sydney between 1998 and 2000, and we were able to test this association using information on infants born to women residing within 5 km of an air monitoring station, again for a large number of births. We were also able to analyse the effect of five common air pollutants (PM₁₀, PM_2.5, CO, NO₂, and O₃), as these pollutants are routinely monitored at many air monitoring stations in metropolitan Sydney. One strength of this study is that we have been able to control for maternal smoking, indigenous status, gestational age at first antenatal visit, and parity among other potential confounders.

Although our analysis controlled for a number of important potential confounders, we did not have the information to adjust for some known risk factors for low birth weight; for example, maternal nutrition, maternal occupation, or pre-pregnancy maternal weight. Also, residual confounding by SES cannot be ruled out. Our SES estimate is a group level variable assigned to postcode of maternal residence at time of birth.

The most important source of bias in our study is due to measurement of exposure. By using air pollutant concentrations averaged across Sydney, we assume that ambient pollutant concentrations represent an individual’s actual exposure to pollutants. This assumption does not account for time-activity patterns that may mediate exposure such as commuting habits, place or type of work, or time spent outdoors.

The use of citywide average exposure does not account for variations in pollutant concentrations across Sydney or even within the 5 km zone around an air monitoring station. It is likely that pollutants, particularly primary pollutants such as nitrogen dioxide and carbon monoxide, are not homogeneously distributed across Sydney. The distribution of primary pollutants will largely depend on the presence of combustion sources such as roads or industry. Thus we also examined the impact of pollutant concentrations only for births to women residing within 5 km of an air monitoring station to attempt to provide a better estimate of exposure for pollutants with large geographic variability. Analysis of the effect of CO and NO₂ shows a clear pattern of increased effect when examining only births to women residing within 5 km of an air monitoring station. Ozone and PM_2.5 are secondary pollutants and thus should be more homogeneously distributed throughout metropolitan Sydney. The effects of concentrations of PM_2.5 and O₃ change little when analysing only births to women residing within 5 km of an air monitoring station.

Our findings of an adverse effect of particulates in the second trimester and of NO₂ and CO at various stages of pregnancy on birth weight have been replicated in previous studies. In a meta-analysis of studies examining the impact of particulate exposure on birth weight, Glinianaia and colleagues¹⁶ concluded that in those studies where an effect of particulates in birth outcomes has been shown, there is considerable variability in the stage of pregnancy that the impact occurs.¹⁶ The same can be said of those studies examining the impact of other pollutants on birth weight.

Chen and colleagues⁹ examined the impact of ambient PM₁₀ and other pollutant concentrations on birth weight in Northern Nevada (USA) between 1991 and 1999, and found that exposure to PM₁₀ in the third trimester of pregnancy was negatively associated with birth weight in Nevada.⁹ Gouveia and colleagues⁶ examined the impact of O₃, CO, and PM₁₀ concentrations on birth weight in Sao Paolo, Brazil in 1997, and found that first trimester exposure to CO and PM₁₀ had an adverse affect on birth weight. No association between O₃ concentrations in any trimester and birth weight was observed.

IUGR, defined as birth weight below the 10th centile of recorded birth weight, was associated with exposure to PM_2.5 and PM₁₀ concentrations in the first trimester of pregnancy in Northern Bohemia.³ In the Czech Republic, IUGR was not associated with exposure to sulphur dioxide (SO₂) and total suspended particulate (TSP) concentrations during pregnancy.² Bobak² also reported that exposure to SO₂, TSP, and oxides of nitrogen (NO_x) during pregnancy had no affect on risk of low birth weight, after adjusting for gestational age (as well as other maternal characteristics). Ha and colleagues⁴ examined the effect of exposure to CO, NO₂, TSP, and SO₂ in the first trimester of pregnancy on low birth weight in Seoul, South Korea. For each inter-quartile increase in pollutant the relative risk of low birth weight was 1.08 for CO, 1.07 for NO₂, 1.06 for SO₂, and 1.04 for TSPs.⁴ No effect of exposure to these pollutants in the third trimester of pregnancy was noted.

Liu and colleagues¹¹ examined the effect of SO₂, NO₂, CO, and O₃ concentrations on IUGR, low birth weight, and preterm birth in Vancouver, Canada. Low birth weight was associated with exposure to SO₂ during the first month of pregnancy (OR 1.1 for a 5 ppb increase). IUGR was associated with exposure to SO₂ (OR 1.1 for a 5 ppb increase), NO₂ (OR 1.1 for a 10 ppb increase), and CO (OR 1.1 for a 1 ppm increase) in the first month of pregnancy.

Last trimester exposure to CO in Los Angeles, USA was reported to result in an increased risk of low birth weight in term births after adjusting for maternal characteristics including commuting habits (OR = 1.22).⁷ Maisonet and colleagues¹⁰ reported an increased risk of term low birth weight due to exposure to CO and SO₂ in all trimesters of pregnancy in six cities in North Eastern United States. No effect of exposure to PM₁₀ was noted.¹⁰ Concentrations of TSP and SO₂ in the last trimester of pregnancy were associated with low birth weight in four residential areas of Beijing, China.¹

In a recent study in Poland, Jedrychowski and colleagues¹⁷ monitored individual exposure to fine particles over 48 hours during the second trimester of pregnancy of 362 women who gave birth between 34 and 43 weeks gestation. PM_2.5 exposures in this study averaged 43 μg/m³ (with a range of 10–147 μg/m³), thus were much higher than those observed in the present study. The authors observed an association between birth weight, birth length, and head circumference, and PM_2.5 exposure in the second trimester of pregnancy. A reduction in birth weight at an increased exposure from 10 to 50 μg/m³ of 140 grams was estimated.

The hypothesised effect of air pollutants on reproductive health relate to a decreased in utero oxygen supply, resulting from a reduction of oxygen carrying capacity induced by air pollution.⁴ Another possibility is that the production of free radicals induced by air pollution might cause an inflammatory response, increasing blood viscosity.^18,19 Suboptimal placenta perfusion from blood viscosity changes may cause adverse pregnancy outcomes, including low birth weight and preterm birth.²⁰ The possible biological mechanisms involved in the reduction in birth weight associated with maternal exposure to air pollution are likely to vary according to the timing of this exposure. The implantation of the fetus and the formation of the placenta occur during the first trimester, while weight gain occurs predominantly during the third trimester. In the first trimester genetic mutations are generally considered the most important cause of placental abnormalities, and in the second and third trimesters complex vascular alterations are considered to be the main cause of placental abnormalities and consequent fetal growth retardation.²¹ Pollutants are recognised as being able to have an effect on both dimensions.²¹ The effect of air pollutant exposure during pregnancy on birth weight has a plausible biological basis; however, the reported studies fail to show consistency in pollutants and periods during pregnancy where an effect occurs.

The lack of consistency in findings may be due to difficulties distinguishing between pollutants. In this study, when examining multiple pollutant models it appears that NO₂ is the most important pollutant, despite the fact that NO₂ levels in Sydney are well below the national standard. The lack of consistency in findings among countries may also be accounted for by differences in pollutant levels and mix. Air quality in the Sydney metropolitan area is generally good. Concentrations of pollutants measured at air monitoring sites in Sydney are typically well below standards, although seasonal conditions can cause the occasional exceedence of the national air quality standards.²² In Sydney, motor vehicles are a major source of air pollutants, although solid fuel heating for domestic purposes in winter and occasional bush fires in summer also add to the fine particle concentrations.²² Given the number of comparisons made in this study, the positive findings may be spurious. It is important, then, to corroborate these finding with future research.

Future research should involve the formation of individual testable hypotheses to avoid the problem of multiple tests. Research should also focus on obtaining higher quality exposure data, for example, modelling air pollution concentrations to smaller geographical areas using emission inventory, traffic density, and meteorological data in the presence of improved time activity data.

In conclusion, we observed that CO and NO₂ concentrations, particularly in the second trimester of pregnancy, have an adverse effect on birth weight. We also observed that PM_2.5 and PM₁₀ concentrations in the second trimester of pregnancy have a small adverse effect on birth weight. While the number of studies in this area is accumulating, a consistent pattern of pollutants and exposure times is not emerging. Further research is required to corroborate our findings.