Effects of fine particulate matter and its constituents on low birth weight among full-term infants in California

Environmental Research 128 (2014) 42–51

Contents lists available at ScienceDirect

Environmental Research journal homepage: www.elsevier.com/locate/envres

Effects of fine particulate matter and its constituents on low birth weight among full-term infants in California Rupa Basu a,n, Maria Harris b, Lillian Sie c, Brian Malig a, Rachel Broadwin a, Rochelle Green a a

California Office of Environmental Health Hazard Assessment, Air Pollution Epidemiology Section, Oakland, CA, USA School of Public Health, Boston University, Boston, MA, USA c School of Public Health, University of California, Berkeley, CA, USA b

art ic l e i nf o

a b s t r a c t

Article history: Received 7 May 2013 Received in revised form 21 October 2013 Accepted 24 October 2013 Available online 17 December 2013

Relationships between prenatal exposure to fine particles (PM2.5) and birth weight have been observed previously. Few studies have investigated specific constituents of PM2.5, which may identify sources and major contributors of risk. We examined the effects of trimester and full gestational prenatal exposures to PM2.5 mass and 23 PM2.5 constituents on birth weight among 646,296 term births in California between 2000 and 2006. We used linear and logistic regression models to assess associations between exposures and birth weight and risk of low birth weight (LBW; o2500 g), respectively. Models were adjusted for individual demographic characteristics, apparent temperature, month and year of birth, region, and socioeconomic indicators. Higher full gestational exposures to PM2.5 mass and several PM2.5 constituents were significantly associated with reductions in term birth weight. The largest reductions in birth weight were associated with exposure to vanadium, sulfur, sulfate, iron, elemental carbon, titanium, manganese, bromine, ammonium, zinc, and copper. Several of these PM2.5 constituents were associated with increased risk of term LBW. Reductions in birth weight were generally larger among younger mothers and varied by race/ethnicity. Exposure to specific constituents of PM2.5, especially traffic-related particles, sulfur constituents, and metals, were associated with decreased birth weight in California. & 2013 Elsevier Inc. All rights reserved.

Keywords: Particulate matter Constituents Low birth weight Full-term infants California Retrospective cohort

1. Introduction Low birth weight (LBW; o2500 g) is a risk factor for greater infant mortality (Hauck et al., 2011; Kochanek et al., 2012; Mathews and MacDorman, 2007) and morbidity throughout adulthood (Johnson and Schoeni, 2011). Although some risk factors for LBW have been studied, the etiology remains largely unknown, and the rate of LBW continues to rise (Donahue et al., 2010). Term LBW babies have been shown to have challenges that are lifelong, ranging from respiratory difficulties throughout childhood (Caudri et al., 2007) and psychological distress in adulthood (Wiles et al., 2005). To date, many studies of fine particulate matter (PM2.5) and LBW have been conducted (Shah and Balkhair, 2011), but the effect of specific PM2.5 constituents on birth outcomes remains largely unstudied. Examining specific constituents of PM2.5 rather than total PM2.5 mass may more accurately capture the specific risk factors for reduced birth weight, and may help elucidate the biological mechanisms involved. To date, only a few n Correspondence to: Air Pollution Epidemiology Section, California Office of Environmental Health Hazard Assessment, 1515 Clay St., 16th floor, Oakland, California 94612, USA. Fax: þ 1 510 622 3210. E-mail address: [email protected] (R. Basu).

0013-9351/$ - see front matter & 2013 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.envres.2013.10.008

studies have evaluated the effects of PM2.5 mass and some PM2.5 constituents on birth weight using data from Connecticut and Massachusetts (Bell et al., 2012, 2010), Atlanta, Georgia (Darrow et al., 2011), Los Angeles, California (Wilhelm et al., 2012), and most recently, the Northeast and mid-Atlantic regions (Ebisu and Bell, 2012). Studies in California are warranted, since the chemical constituents, sources, and levels of particulate matter differ between the western and eastern coasts of the US (Blanchard, 2003). Furthermore, California has a number of air pollution monitors located throughout the state that measure pollutant concentrations for a population diverse in race/ethnicity, socioeconomic status, and age, providing the opportunity to study impacts on potentially vulnerable subgroups. A few previous studies reported associations between PM2.5 and birth weight among full-term infants in California even after controlling for carbon monoxide (CO), although PM2.5 constituents were not examined (Morello-Frosch et al., 2010; Parker et al., 2005; Wilhelm and Ritz, 2005). A recent study conducted in Los Angeles, California, by Wilhelm et al. found PM2.5, nitric oxide, nitrogen dioxide (NO2), and polycyclic aromatic hydrocarbons (PAHs) to be associated with term LBW (Wilhelm et al., 2012). Only a few PM2.5 constituents were considered in the Los Angeles metropolitan area. More studies are needed in

R. Basu et al. / Environmental Research 128 (2014) 42–51

California specifically to examine PM2.5 constituents associated with LBW so that compounds and major sources with particularly strong impacts on birth weight can be identified. In the current study, we examined exposures to PM2.5 and several PM2.5 constituents on LBW among full-term infants in California over a seven-year period. We stratified by season of births, and also identified vulnerable subgroups by exploring possible modification of effects by maternal characteristics.

2. Methods 2.1. Study population Our study population consisted of infants born between 2000 and 2006 to mothers residing in California, with birth records accessed from the Natality Database supplied by the California Department of Public Health (California Department of Health Services, 2006). We included singleton live full-term births with gestational ages from 37 through 44 weeks with data available for birth date, birth weight, gestational age, infant sex, maternal ethnicity, maternal educational attainment, maternal age and maternal residential zip code. Births with maternal age 449 years, unreasonable gestational ages ( 4 44 weeks) or unreasonable combinations of gestational age and birth weights were excluded (Alexander et al., 1996). 2.2. Particulate matter data and exposure assessment We accessed data on ambient PM2.5 mass and concentrations of PM2.5 constituents from US EPA monitors in eight sites in California that supplied data during our study period (Los Angeles, Riverside, El Cajon, San Jose, Simi Valley, Bakersfield, Fresno, and Sacramento) (California Air Resources Board, 2011). Data were available from 2000 through 2006, with varying start dates by site (ranging from 2000 to 2002) and some gaps in operation. Monitoring frequency was every 3 or 6 days. PM2.5 constituents were included in the analysis if the constituent was monitored continuously throughout the study period and was detected at levels above the laboratory detection limit for at least 30% of sampling days. We used Arc GIS (Version 9.2) to calculate the distance between the geocoded monitor location and the population-weighted centroid of the 2000 US Census zip code tabulation area associated with the maternal residential zip code reported in the birth record. Zip Code Tabulation Areas are area units developed by the US Census Bureau to represent the geographic boundaries of US Postal Service zip codes (Grubesic and Matisziw, 2006). All births with valid maternal zip codes located within 20 km of a monitor were assigned exposures from that monitor, while all other births were excluded. If there was more than one monitor within 20 km, we assigned exposures using the closest monitor. We calculated exposure to PM2.5 constituents for the full gestational period of each infant, defined as the period between conception (estimated as 2 weeks past the last menstrual period) and birth date. We estimated the date of last menstrual period by subtracting total gestational days from each birth's reported date of birth. We estimated exposures for each trimester of pregnancy, defining the first trimester as the gestational weeks three to 13 post-last menstrual period, the second trimester as the 14–26th weeks and the third trimester as the 27th week through birth. We calculated trimester exposures as the average of weekly mean exposures for all weeks in the trimester if mean data were available for at least 75% of the weeks in that trimester. Weekly exposures were calculated as the mean of all readings for a given PM2.5 constituent during each gestational week, provided there was at least one reading during that week. Weekly means for partial weeks occurring at the end of a pregnancy were included in the third trimester mean if the partial week included at least 4 days. Full pregnancy exposure was estimated as the mean of all three trimesters, with births missing exposure measurements for one or more trimesters excluded from the analysis. Calculating trimester and full pregnancy exposures using weekly means reduces bias that may result from variation in sampling frequency across monitors (Bell et al., 2010). 2.3. Covariates We used demographic covariates from the birth certificate database including maternal race/ethnicity (White: non-Hispanic White, Black: non-Hispanic Black, Hispanic: Hispanic of any race, Asian: non-Hispanic Asian), maternal age (less than 20, 20–24, 25–34, 35–39, greater than 39 years) and maternal educational attainment (fewer than 12 years, 12 years, 13–15 years, 16 years or greater). Additional covariates included: weeks of gestational age, used to account for birth weight differences attributable to variation in gestational age; month of birth, used to control for seasonal patterns in birth weight; year of birth; and infant sex. We calculated average apparent temperature for each trimester and full gestational

43

period using data from centrally located temperature monitors using techniques similar to those used for PM2.5 and PM2.5 constituents. Temperature and relative humidity readings used to calculate apparent temperature originated from three databases: the US Environmental Protection Agency Data Mart (U.S. EPA, 2009), the National Climatic Data Center (NCDC, 2012), and the California Irrigation Management and Information System (Office of Water Use Efficiency, 2009). 2.4. Statistical analyses We conducted linear regression analyses relating birth weight to continuous measures of full gestational and trimester exposures to total PM2.5 mass and PM2.5 constituents, with a separate model used for each exposure variable. All models were adjusted for maternal race/ethnicity, maternal age, maternal educational attainment, weeks of gestational age, month and year of birth, infant sex, and gestational apparent temperature exposure. We also conducted analyses stratified by season of birth (warm, May 1–October 31; or cool, November 1–April 30) and several maternal factors (age, education, and race/ethnicity) to examine whether associations differed across levels of these covariates. To account for potential confounding by regional factors, we adjusted for region: Northern California/ Central Valley (monitors located in San Jose, Sacramento, Fresno, and Bakersfield) and Southern California (monitors located in Simi Valley, Los Angeles, Riverside, and El Cajon). We used percentage groupings for unemployment (0–85%, 86–90%, 91–92.5%, 92.6–95%, 96% and higher), non-White race/ethnicity (10–19%, 20–39%, 40–59%, 60–79%, 80% and higher), and home ownership (under 20%, 20–39%, 40–59%, 60–79%, 80% and higher) by zip code tabulation area provided by the 2000 US Census to adjust for community-level race and socioeconomic status. We also conducted sensitivity analyses limiting our study population to subjects residing within a 5 km and 10 km radius of each monitor. Furthermore, we conducted analyses stratified by maternal age, maternal race/ethnicity, and maternal education and were examined to investigate possible effect modification by these factors. SAS statistical software was used to conduct all analyses (SAS Institute Inc., 2008). Results are presented as the change in birth weight associated with each interquartile range increase in trimester and full gestational exposures for total PM2.5 mass and each constituent. We also conducted logistic regression analyses relating LBW (birth weight less than 2500 g) and full gestational and trimester exposures to total PM2.5 mass and PM2.5 constituents. Results are presented as percent change ((odds ratio  1)  100) in risk of LBW per interquartile increase in trimester and full gestational exposures for total PM2.5 mass and each constituent. Statistical significance of effect modification was assessed using models with an interaction term between the pollutant and the stratification variable. As a sensitivity analysis, we added a squared term for the PM2.5 and each PM2.5 constituent in our model as a statistical test of linearity. For models improved by the squared term, we used generalized additive models replacing the linear constituent exposure term with a cubic regression spline term with four degrees of freedom to better visualize the relationship.

3. Results There were a total of 3,673,224 live, singleton births in California from 2000 through 2006. After excluding the births with missing data on our variables of interest, maternal age greater than 49 years, gestational age greater than 44 weeks or unreasonable combinations of age/birth weight, 3,221,706 births remained. Of these births, 2,930,300 were full-term (from 37 through 44 weeks of gestation), but only 1,078,166 births occurred in the counties included in our study to mothers residing within 20 km of a monitor. Our final study population consisted of 646,296 singleton full-term births from eight counties with full gestational exposure information available for PM2.5 and PM2.5 constituents. Fig. 1 depicts a map of California with the location of the PM2.5 monitors used in our analysis of eight sites. As shown in Table 1, the majority of mothers were Hispanic, and aged 25–34 years at the time of giving birth. About 32.5% of mothers had completed less than a high school education. A total of 2.4% of infants were found to be term LBW. In addition, infants were slightly more likely to be male, had an average weight of 3401 g, and a mean gestation of 39.3 weeks. Total PM2.5 mass and 23 PM2.5 constituents had sufficient data to be included in the final analysis (means and standard deviations by site listed in Table 2). In Table 3, the correlation between PM2.5 and each PM2.5 constituent are listed, as well as the correlations between the gaseous

44

R. Basu et al. / Environmental Research 128 (2014) 42–51

pollutants. Because the correlation between some of the pollutants were relatively high, we did not examine multi-pollutant models in this study.

Fig. 1. Map of California with location of PM2.5 monitors in eight counties, 1999– 2006.

In Table 4, the change in birth weight per interquartile increase in exposure (from the linear regression models) and percent increase in risk of LBW per interquartile increase in exposure (from the logistic regression models) for PM2.5 and PM2.5 constituents are listed. Higher full gestational exposures of PM2.5 mass and nearly all PM2.5 constituents appeared to be associated with a reduction in term birth weight (linear model), with many also associated with increased risk of term LBW (logistic model). The largest reductions in birth weight were associated with exposure to vanadium, sulfur, sulfate, iron, titanium, elemental carbon, manganese, bromine, ammonium, zinc and copper. The PM2.5 constituents with significant changes in birth weight for all exposure windows are shown by trimester in Fig. 2. There was no consistent pattern in the comparative strength of the associations between exposures to PM2.5 or PM2.5 constituents and reduced birth weight in the first, second, or third trimester. Full pregnancy exposure was generally the best predictor of reduced birth weight based on model fit determined by Akaike's Information Criterion. Full pregnancy results were also more consistent and had higher effect estimates. Nearly all interactions between PM2.5 or PM2.5 constituents and maternal age, education, and race/ethnicity were statistically significant (data not shown). The significant findings for PM2.5 and PM2.5 constituents and risk of birth weight by race/ethnic group are summarized in Fig. 3. Black and Hispanic mothers had greater risks associated with PM2.5 total mass exposure than White and Asian mothers, although the confidence intervals overlapped between these values. The associations were stronger for Whites for several PM2.5 constituents, but generally varied by maternal race/ethnic group. Younger mothers (20–24 years) generally displayed greater reductions in birth weight associated with exposure to PM2.5 constituents (Fig. 4), although these differences were not statistically different. Maternal education showed no consistent pattern (not shown), and no stable seasonal patterns

Table 1 Summary of study population characteristics for eight California sites examined. Population characteristics

ZCTAs Births Mean birth weight in g (SD) Infant weight under 2500 g (%) Mother's age (%) Age o 20 Age 20–24 Age 25–34 Age 35–39 Age 439 Mother's race/ethnicity (%) White (non-Hispanic) Black (non-Hispanic) Hispanic Asian (non-Hispanic) Other non-White Mother's education (%) Less than high school High school graduate Some college College graduate or beyond Infant sex (%) Male Mean ZCTA % home ownership (SD) Mean ZCTA % employment (SD) Mean ZCTA % non-White (SD)

All sites

Northern region

Southern region

Bakersfield

Fresno

Sacramento

San Jose

El Cajon

Los Angeles

Riverside

Simi Valley

335 646,296 3401.4 (466.8) 2.4

11 21,224 3430.2 (473.4) 2.3

19 45,200 3403.3 (466.0) 2.4

45 88,566 3460.9 (479.9) 2.0

46 76,385 3397.6 (463.4) 2.3

24 54,281 3436.4 (470.7) 2.1

129 234,442 3369.6 (461.8) 2.7

33 92,026 3403.0 (462.3) 2.3

28 34,125 3394.3 (456.1) 2.2

10.0 23.7 51.0 12.3 2.9

13.6 30.6 46.4 7.5 1.9

13.6 29.0 46.3 9.0 2.1

9.3 25.5 51.3 11.2 2.7

5.5 15.1 58.7 17.0 3.6

9.8 25.1 50.9 11.6 2.6

10.6 23.5 49.8 12.9 3.2

11.9 27.9 48.9 9.3 2.1

5.8 16.0 56.2 17.4 4.6

23.8 7.3 54.9 12.7 1.4

34.9 6.3 54.6 3.4 0.9

28.4 6.7 52.8 6.0 6.0

47.2 9.9 27.7 11.6 3.6

24.3 2.2 33.8 38.8 0.9

25.9 9.3 52.3 11.2 1.3

11.6 8.1 70.1 9.9 0.3

19.9 8.1 66.3 5.1 0.6

39.0 3.0 44.4 13.2 0.4

32.5 26.4 19.0 22.1

31.5 34.0 21.2 13.2

31.4 29.4 22.5 16.7

21.9 31.3 23.3 23.5

18.8 19.2 18.5 43.6

26.9 28.8 22.0 22.2

42.8 23.8 15.3 18.0

35.3 32.1 20.1 12.6

24.4 20.2 20.0 35.4

50.9 50.7 (18.7)

50.3 61.4 (13.0)

50.9 53.6 (14.4)

50.7 58.2 (14.5)

51.4 59.4 (16.6)

51.0 48.9 (18.9)

50.8 37.6 (16.3)

50.7 62.5 (12.7)

50.8 60.9 (14.1)

92.0 (3.5) 67.2 (24.8)

90.2 (4.2) 52.5 (22.1)

89.8 (4.7) 60.1 (21.1)

93.3 (2.9) 43.5 (20.2)

96.2 (1.4) 61.5 (19.8)

93.2 (2.8) 63.6 (23.2)

90.4 (2.9) 84.5 (20.5)

91.7 (2.2) 67.3 (13.6)

94.0 (2.0) 46.5 (18.6)

ZCTA ¼ zip code tabulation area; SD ¼standard deviation.

Table 2 Means (SD) of study population gestational exposures to PM2.5 mass and PM2.5 constituents (mg/m3), 2000–2006. Constituent name

SO24  S Ti NO3 V Zn Apparent temperature (1F)

Southern region

Whole cohort

Bakersfield

Fresno

18.7 0.03 2.3 0.0044 0.08 0.07 0.009 1.1 0.16 0.004 0.0033 0.0033 6.2 0.12 0.14 0.19 0.09 0.33 2.6

(5.0) (0.02) (1.2) (0.0011) (0.04) (0.06) (0.005) (0.4) (0.07) (0.002) (0.0015) (0.0040) (1.3) (0.08) (0.05) (0.09) (0.08) (0.14) (1.2)

21.1 0.07 2.5 0.0056 0.09 0.03 0.013 1.0 0.15 0.003 0.0024 0.0010 6.9 0.15 0.25 0.08 0.10 0.21 1.8

(3.4) (0.02) (0.6) (0.0007) (0.01) (0.02) (0.006) (0.1) (0.02) (0.001) (0.0005) (0.0002) (0.9) (0.03) (0.05) (0.03) (0.03) (0.09) (0.1)

20.1 0.04 2.0 0.0041 0.05 0.05 0.006 0.9 0.11 0.004 0.0020 0.0035 7.6 0.13 0.16 0.10 0.09 0.23 1.6

0.83 0.010 5.5 0.0043 0.015 61.8

(0.33) (0.004) (2.7) (0.0023) (0.008) (4.0)

0.63 0.010 6.8 0.0016 0.011 64.0

(0.05) (0.002) (1.8) (0.0009) (0.002) (4.6)

0.55 0.007 5.5 0.0013 0.012 62.2

Sacramento

San Jose

El Cajon

Los Angeles

Riverside

Simi Valley

(5.4) (0.02) (0.5) (0.0005) (0.01) (0.03) (0.002) (0.2) (0.01) (0.001) (0.0005) (0.0016) (2.1) (0.03) (0.04) (0.04) (0.03) (0.12) (0.1)

14.6 0.03 0.9 0.0028 0.04 0.04 0.006 0.8 0.08 0.003 0.0017 0.0039 6.5 0.11 0.11 0.14 0.08 0.25 1.2

(3.4) (0.01) (0.2) (0.0004) (0.01) (0.02) (0.002) (0.2) (0.02) (0.001) (0.0005) (0.0051) (1.7) (0.03) (0.04) (0.05) (0.03) (0.10) (0.1)

14.2 0.03 0.9 0.0034 0.07 0.18 0.005 1.0 0.12 0.004 0.0022 0.0076 5.5 0.08 0.12 0.35 0.06 0.50 1.4

(2.9) (0.01) (0.2) (0.0004) (0.03) (0.07) (0.001) (0.3) (0.04) (0.001) (0.0012) (0.0087) (1.1) (0.01) (0.06) (0.12) (0.02) (0.17) (0.2)

14.5 0.03 1.6 0.0040 0.05 0.08 0.007 0.8 0.10 0.004 0.0026 0.0019 5.3 0.08 0.11 0.25 0.05 0.44 2.7

(1.6) (0.01) (0.3) (0.0005) (0.01) (0.03) (0.001) (0.2) (0.02) (0.001) (0.0004) (0.0003) (0.8) (0.01) (0.03) (0.06) (0.02) (0.11) (0.5)

20.4 0.03 3.0 0.0050 0.07 0.06 0.014 1.5 0.23 0.005 0.0048 0.0029 6.3 0.15 0.13 0.19 0.12 0.33 3.8

(2.5) (0.01) (0.7) (0.0006) (0.02) (0.03) (0.004) (0.2) (0.04) (0.001) (0.0010) (0.0005) (0.6) (0.12) (0.05) (0.04) (0.12) (0.09) (0.8)

25.3 0.05 3.7 0.0055 0.16 0.07 0.008 1.3 0.17 0.006 0.0036 0.0022 6.5 0.11 0.19 0.13 0.07 0.29 2.9

(3.6) (0.01) (0.7) (0.0007) (0.03) (0.04) (0.002) (0.1) (0.02) (0.001) (0.0007) (0.0004) (0.8) (0.03) (0.03) (0.03) (0.03) (0.09) (0.5)

13.1 0.02 1.5 0.0039 0.05 0.03 0.005 0.7 0.08 0.002 0.0014 0.0017 4.6 0.07 0.11 0.19 0.04 0.31 2.5

(1.6) (0.01) (0.3) (0.0007) (0.01) (0.01) (0.001) (0.1) (0.01) (0.001) (0.0003) (0.0003) (0.5) (0.03) (0.02) (0.07) (0.04) (0.09) (0.5)

(0.03) (0.002) (1.6) (0.0004) (0.003) (4.5)

0.39 0.005 2.5 0.0014 0.006 59.2

(0.04) (0.002) (0.8) (0.0005) (0.001) (3.8)

0.47 0.007 2.8 0.0019 0.008 57.0

(0.05) (0.003) (0.7) (0.0005) (0.002) (2.6)

0.84 0.007 3.5 0.0041 0.007 61.8

(0.15) (0.002) (0.7) (0.0006) (0.001) (3.3)

1.13 0.014 6.6 0.0066 0.020 63.3

(0.22) (0.003) (1.5) (0.0010) (0.004) (2.6)

0.96 0.011 9.5 0.0054 0.025 64.2

(0.16) (0.003) (1.8) (0.0011) (0.006) (4.0)

0.81 0.006 3.2 0.0041 0.004 61.0

(0.17) (0.002) (0.8) (0.0008) (0.001) (2.5)

R. Basu et al. / Environmental Research 128 (2014) 42–51

Total PM2.5 Al NH4þ Br Ca Cl Cu EC Fe Pb Mn Ni OC K Si Na Kþ Na þ

Northern region

SD ¼ standard deviation.

45

46

Table 3 Correlation coefficients between gestational exposures to PM2.5 mass, constituents, temperature, and community-level socioeconomic status variables for the study population, 2000–2006. PM2.5

Al

NH4 þ

Br

Ca

Cl

Cu

EC

Fe

Pb

Mn

Ni

OC

K

PM2.5 Al NH4 þ Br Ca Cl Cu EC Fe Pb Mn Ni OC K Si Na Kþ Na þ SO24  S Ti NO3 V Zn Apparent temperature (1F) % ZCTA home ownership % ZCTA non White % ZCTA employed

1 0.43 0.84 0.79 0.62 0.01 0.38 0.67 0.54 0.73 0.48 0.09 0.69 0.32 0.48 0.26 0.26 0.06 0.48

0.43 1 0.25 0.31 0.57 0.08 0.11 0.01 0.12 0.27 0.07 0.15 0.23 0.24 0.84 0.12 0.16 0.07 0.04

0.84 0.25 1 0.90 0.55  0.22 0.50 0.73 0.65 0.63 0.58  0.22 0.31 0.21 0.27  0.27 0.19  0.18 0.78

0.79 0.31 0.90 1 0.55  0.15 0.52 0.68 0.65 0.61 0.58  0.23 0.29 0.20 0.38  0.21 0.17  0.10 0.73

0.62 0.57 0.55 0.55 1 0.24 0.14 0.39 0.47 0.63 0.42 0.05 0.16 0.09 0.72 0.05 0.01 0.13 0.24

0.01 0.08 0.22 0.15 0.24 1 0.12 0.08 0.03 0.20 0.06 0.39 0.02 0.16 0.16 0.66 0.14 0.64 0.23

0.38 0.11 0.50 0.52 0.14  0.12 1 0.62 0.72 0.51 0.60  0.16 0.14 0.68 0.16  0.18 0.69  0.04 0.59

0.67 0.01 0.73 0.68 0.39  0.08 0.62 1 0.86 0.64 0.72  0.04 0.42 0.21 0.13  0.11 0.23  0.11 0.62

0.54 0.12 0.65 0.65 0.47  0.03 0.72 0.86 1 0.62 0.88 0.00 0.21 0.24 0.36  0.06 0.23 0.04 0.67

0.73 0.27 0.63 0.61 0.63 0.20 0.51 0.64 0.62 1 0.58 0.07 0.41 0.48 0.38 0.01 0.45 0.15 0.49

0.48 0.07 0.58 0.58 0.42  0.06 0.60 0.72 0.88 0.58 1 0.08 0.19 0.14 0.29 0.04 0.14 0.21 0.71

 0.09 0.15  0.22  0.23 0.05 0.39  0.16  0.04 0.00  0.07 0.08 1 0.03  0.06 0.11 0.38  0.02 0.50  0.16

0.69 0.23 0.31 0.29 0.16 0.02 0.14 0.42 0.21 0.41 0.19 0.03 1 0.24 0.30 0.27 0.18 0.05 0.01

0.32 0.24 0.21 0.20 0.09 0.16 0.68 0.21 0.24 0.48 0.14  0.06 0.24 1 0.15  0.09 0.98 0.09 0.24

0.51 0.58 0.92 0.43 0.81 0.25 0.08 0.26 0.32

0.01 0.20 0.38 0.01 0.22 0.31 0.17 0.06 0.12

0.76 0.64 0.96 0.74 0.87 0.45  0.19 0.38  0.37

0.75 0.64 0.86 0.69 0.79 0.46  0.18 0.37  0.35

0.31 0.50 0.66 0.32 0.67 0.28 0.12 0.16 0.08

0.20 0.01 0.11 0.23 0.06 0.33 0.10 0.01 0.30

0.57 0.75 0.42 0.63 0.48 0.33  0.39 0.41  0.36

0.56 0.74 0.68 0.68 0.82 0.11  0.33 0.47  0.31

0.65 0.89 0.58 0.73 0.75 0.29  0.39 0.52  0.35

0.52 0.69 0.67 0.50 0.75 0.19 0.17 0.35 0.23

0.71 0.84 0.50 0.70 0.67 0.34  0.38 0.49  0.33

 0.19 0.04  0.16  0.19  0.12  0.15 0.05  0.04 0.17

0.01 0.29 0.44 0.10 0.36 0.14 0.03 0.03 0.19

0.26 0.41 0.24 0.26 0.21 0.22  0.16 0.16  0.20

Variable

Si

Na

Kþ

Na þ

SO42 

S

Ti

NO3 

V

Zn

Apparent temperature (1F)

% ZCTA home ownership

% ZCTA non-White

% ZCTA employed

PM2.5 Al NH4 þ Br Ca Cl Cu EC Fe Pb Mn Ni OC K Si Na Kþ Na þ SO24  S Ti NO3 V Zn Apparent temperature (1F) % ZCTA home ownership % ZCTA non White % ZCTA employed

0.48 0.84 0.27 0.38 0.72 0.16 0.16 0.13 0.36 0.38 0.29 0.11 0.30 0.15 1 0.14 0.04 0.14 0.00

 0.26  0.12  0.27  0.21  0.05 0.66  0.18  0.11  0.06  0.01 0.04 0.38  0.27  0.09  0.14 1  0.04 0.73 0.02

0.26 0.16 0.19 0.17 0.01 0.14 0.69 0.23 0.23 0.45 0.14 0.02 0.18 0.98 0.04 0.04 1 0.11 0.26

 0.06 0.07  0.18  0.10 0.13 0.64  0.04  0.11 0.04 0.15 0.21 0.50  0.05 0.09 0.14 0.73 0.11 1 0.08

0.48 0.04 0.78 0.73 0.24 0.23 0.59 0.62 0.67 0.49 0.71 0.16 0.01 0.24 0.00 0.02 0.26 0.08 1

0.51 0.01 0.76 0.75 0.31  0.20 0.57 0.56 0.65 0.52 0.71  0.19 0.01 0.26 0.08 0.01 0.26 0.11 0.98

0.58 0.20 0.64 0.64 0.50 0.01 0.75 0.74 0.89 0.69 0.84 0.04 0.29 0.41 0.43 0.08 0.40 0.17 0.67

0.92 0.38 0.96 0.86 0.66  0.11 0.42 0.68 0.58 0.67 0.50  0.16 0.44 0.24 0.40  0.31 0.20  0.14 0.60

0.43 0.01 0.74 0.69 0.32 0.23 0.63 0.68 0.73 0.50 0.70 0.19 0.10 0.26 0.01 0.01 0.28 0.00 0.92

0.81 0.22 0.87 0.79 0.67  0.06 0.48 0.82 0.75 0.75 0.67  0.12 0.36 0.21 0.33  0.27 0.18  0.14 0.63

0.25 0.31 0.45 0.46 0.28  0.33 0.33 0.11 0.29 0.19 0.34  0.15  0.14 0.22 0.29  0.12 0.18 0.00 0.58

 0.08 0.17  0.19  0.18 0.12 0.10  0.39  0.33  0.39  0.17  0.38 0.05  0.03  0.16 0.11  0.02  0.18  0.03  0.38

0.26  0.06 0.38 0.37 0.16 0.01 0.41 0.47 0.52 0.35 0.49  0.04 0.03 0.16 0.03 0.06 0.17 0.08 0.48

 0.32  0.12  0.37  0.35  0.08 0.30  0.36  0.31  0.35  0.23  0.33 0.17  0.19  0.20  0.15 0.31  0.17 0.21  0.33

0.08 0.43 0.40 0.01 0.33 0.29 0.11 0.03 0.15

0.01  0.08  0.31 0.01  0.27  0.12  0.02 0.06 0.31

0.26 0.40 0.20 0.28 0.18 0.18 0.18 0.17 0.17

0.11 0.17  0.14 0.00  0.14 0.00  0.03 0.08 0.21

0.98 0.67 0.60 0.92 0.63 0.58 0.38 0.48 0.33

1 0.65 0.60 0.90 0.63 0.64  0.35 0.47  0.33

0.65 1 0.60 0.64 0.69 0.32 0.33 0.45 0.32

0.60 0.60 1 0.57 0.87 0.34  0.09 0.31  0.34

0.90 0.64 0.57 1 0.66 0.53 0.37 0.49 0.31

0.63 0.69 0.87 0.66 1 0.29  0.19 0.41  0.34

0.64 0.32 0.34 0.53 0.29 1  0.14 0.21  0.29

 0.35  0.33  0.09  0.37  0.19  0.14 1  0.45 0.51

0.47 0.45 0.31 0.49 0.41 0.21  0.45 1  0.65

 0.33  0.32  0.34  0.31  0.34  0.29 0.51  0.65 1

Bold ¼ correlation 0.80 or above; ZCTA ¼ zip code tabulation area.

R. Basu et al. / Environmental Research 128 (2014) 42–51

Variable

R. Basu et al. / Environmental Research 128 (2014) 42–51

47

Table 4 Effect estimates for PM2.5 mass and constituents on birth weight and risk of low birth weight births. Constituent IQR (lg/ m3)

Continuous birth weight Change in birth weight (95% CI) per IQR increase in constituent

Change in birth weight (95% CI) per IQR increase in constituent, also adjusted for ZCTA SES variables

Percent change in odds of Percent change in odds of LBW (95% CI) per LBW (95% CI) per IQR increase IQR increase in constituent, also adjusted for in constituent ZCTA SES variables

PM2.5 Al NH4 þ Br Ca Cl Cu EC Fe Pb Mn Ni OC K Si Na Kþ Na þ SO42  S Ti NO3  V Zn

6 5  12 13 1 3 10 16  21 3 14 1 2 2 1 3 2 3 22 29 15 8 32 10

7 3  11  12 1 3 7  13 18 2  11 1 1 1 3 2 1 0 17 24  12 9 26 9

2 1 4 5 1 1 5 7 11 2 9 1 0 1 1 1 1 1 10 12 7 2 18 4

7.563 0.022 1.901 0.002 0.050 0.062 0.007 0.574 0.115 0.002 0.003 0.001 1.415 0.050 0.082 0.093 0.047 0.168 1.937 0.566 0.007 4.068 0.005 0.014

( 9 to 4) (3–7) ( 14 to 9) ( 16 to 10) (0–3) ( 5 to 1) ( 12 to 8) ( 19 to 14) ( 24 to  18) ( 5 to 1) ( 17 to  12) ( 2 to 1) ( 4 to 0) ( 3 to 1) ( 4 to 1) ( 5 to 2) ( 3 to 1) ( 5 to 1) ( 25 to  18) ( 33 to  25) ( 17 to  13) ( 10 to 6) ( 38 to  27) ( 12 to 7)

Binary low birth weight (LBW)

( 9 to  4) (0–5) ( 14 to  8) ( 14 to  9) ( 3 to 1) ( 4 to  1) ( 9 to  5) ( 16 to  10) ( 21 to 15) ( 4 to 1) ( 13 to  8) ( 2 to  1) ( 2 to 1) ( 2 to  1) ( 5 to 0) ( 4 to 0) ( 2 to 0) ( 3 to 2) ( 21 to 13) ( 28 to  19) ( 14 to  10) ( 11 to 6) ( 32 to  21) ( 12 to 7)

( 1 to 5) ( 4 to 2) (0–8) (1–9) ( 3 to 2) ( 1 to 4) (2–8) (3–11) (7–16) ( 1 to 5) (5–14) (0–1) ( 3 to 2) (0–2) ( 2 to 5) ( 2 to 4) (0–2) ( 2 to 4) (4–15) (6–19) (3–11) ( 1 to 6) (10–28) (0–8)

1 1 3 3 1 1 2 2 6 0 4 1 2 0 2 0 0 1 3 5 3 2 9 2

( 2 to ( 2 to ( 1 to ( 1 to ( 2 to ( 2 to ( 1 to ( 3 to (2–11) ( 3 to (0–9) (0–1) ( 5 to ( 1 to ( 1 to ( 3 to ( 1 to ( 4 to ( 2 to ( 1 to ( 1 to ( 2 to (0–18) ( 2 to

5) 4) 7) 7) 4) 4) 5) 6) 4)

1) 1) 6) 3) 1) 3) 9) 12) 7) 6) 6)

LBW¼ low birth weight; IQR¼ interquartile range; ZCTA ¼ zip code tabulation area; SES=socioeconomic status; all models adjusted by maternal ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic, non-Hispanic Asian), maternal age (less than 20, 20–24, 25–34, 35–39, greater than 39 years), maternal educational attainment (less than high school, high school graduate, some college, college graduate), weeks of gestational age, month of birth, year of birth, infant sex, region (north, south), and gestational apparent temperature; models with SES variables also adjusted by ZCTA-level home ownership (under 20%, 20–39%, 40–59%, 60–79%, 80% and higher), employment (0–85%, 86–90%, 91–92.5%, 92.6–95%, 96% and higher), and non-White (10–19%, 20–39%, 40–59%, 60–79%, 80% and higher).

emerged when results were stratified by warm or cold season of birth and significant associations persisted (not shown). Effect estimates were generally attenuated somewhat following adjustment for indicators of community-level race/ethnicity and socioeconomic status, but reductions in birth weight remained stable (Table 4). Among the species with pronounced associations, vanadium did not show significantly better fits with squared terms when compared to linear models. Additional species, namely Fe, Ti, Ec, and Mn, did not show much visual deviation from linearity, especially when the focus was placed on data between the 10th and 90th percentiles, with mean absolute differences between linear modeled and spline modeled birth weight for these species ranging from 1.7 to 2.9 g. For other strongly associated species, the spline model suggested underprediction of birth weight by the linear model at moderate levels of the PM2.5 constituent (sulfur, sulfate, bromine) or non-monotonic relationships (ammonium, copper, zinc), with mean absolute differences ranging from 4.3 to 7.3 g. Linear models were generally sufficient fits for binary low birth weight, except for sulfur and sulfate, which were almost linear, and zinc, which showed curvature at both extremes. In our sensitivity analyses for buffer size, we found that a 10 km buffer produced results similar to those of a 20 km buffer analysis although confidence intervals were wider (not shown). We also found that a five kilometer buffer included only 1791 cases of term low birth weight, and yielded unstable estimates and very wide confidence intervals.

4. Discussion Our results suggest an association between PM2.5 mass and several PM2.5 constituent exposures and lower birth weight among infants born at full term in California. Among the constituents we examined, sulfur, sulfate, vanadium, iron, manganese, bromine,

ammonium, zinc and copper were found to be most strongly associated with reductions in birth weight. These constituents are often markers of traffic pollutants, industrial sources, oil combustion, or alloy production. A number of constituents were found to have a greater impact on birth weight than PM2.5 total mass, indicating that the observed associations between PM2.5 constituents and birth weight were not merely a result of correlations with PM2.5. We observed some evidence of possible non-linearity in PM2.5 constituent–birth weight relationships, though our analyses may be particularly sensitive due to large sample size. Correlational relationships with other constituents or other sources of confounding not accounted for in this analysis may have some influence on the dose-response curves observed. Future studies should investigate possible non-linear relationships in further detail, accounting for these and other factors. We observed disparities in effect magnitude by maternal age and maternal race/ethnic group, both of which are known to be associated with socioeconomic status and increased susceptibility (Blumenshine et al., 2010; Lhila and Long, 2012; O'Neill et al., 2003). Younger mothers had greater risks associated with PM2.5 mass and nearly all PM2.5 constituents. Asians, Blacks, and Hispanics, compared to Whites, exhibited smaller birth weight reductions for most PM2.5 constituents. However, Black and Hispanic mothers had greater reductions in birth weight with increases in exposure to total PM2.5 mass compared to White and Asian mothers. Black mothers have been previously reported to have greater reductions in birth weight or risk of term LBW from PM2.5 (Salihu et al., 2011) and PM2.5 constituents (Bell et al., 2010; Darrow et al., 2011) compared to White mothers. Compared to White mothers, Hispanic mothers are typically younger, have lower socioeconomic status, have completed less education, and are less likely to have access to prenatal care (McDonald et al., 2008). They also have been found to have better nutrition, and thus, generally greater birth weight (McDonald et al., 2008).

48

R. Basu et al. / Environmental Research 128 (2014) 42–51

Fig. 2. Reduction in birth weight per IQR increase in gestational exposure and trimester exposures to PM2.5 and PM2.5 constituents. Figure includes all species with statistically significant reductions in birth weight across all exposure periods (F, full pregnancy; 1, 1st trimester; 2, 2nd trimester; 3, 3rd trimester) based on 95% confidence interval excluding zero. Models are adjusted for maternal ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic, non-Hispanic Asian), maternal age (less than 20, 20–24, 25–34, 35–39, greater than 39 years), maternal educational attainment (less than high school, high school graduate, some college, college graduate), weeks of gestational age, month of birth, year of birth, infant sex, region (north, south), and apparent temperature.

Our results are consistent with a recent investigation that reported metals and elemental carbon (EC) to be associated with LBW (Darrow et al., 2011) and another study that found road dust and related constituents (silicon, calcium, iron, manganese, titanium, and aluminum), motor vehicle-related constituents (zinc, bromide, copper, lead, potassium, and EC), and oil-combustionassociated elements (vanadium and nickel) to be associated with LBW (Bell et al., 2010). Other studies have found that a specific type of road dust, resuspended brake wear, is often a source for aluminum, copper, iron, magnesium, manganese, potassium, sulfur, titanium, vanadium, zinc, organic carbon (OC) and EC (Garg et al., 2000; Schauer et al., 2006). Tires have been found to be a source for aluminum, copper, iron, magnesium, manganese, potassium, sodium, zinc, OC and EC (Schauer et al., 2006). Several investigators have examined PM2.5 constituents and sources in relation to acute morbidity and mortality. A time-series study of mortality and PM2.5 components in nine counties in California found significant associations during the cool season between cardiovascular mortality and several components, including sulfates, copper, iron, and manganese, although it did not find an effect from nickel (Ostro et al., 2007). Another recent time-series study of mortality and PM2.5 constituents in two US cities (Zhou et al., 2011) also found effects of several pollutants, such as aluminum, zinc, iron, potassium, sulfur, silicon, EC and nickel during the cool season in Seattle, after adjusting for gaseous pollutants including NO2. In Detroit, some pollutants, including sulfur, were generally more strongly associated with mortality outcomes during the warm season. In a study conducted in New York City that followed young children for respiratory

symptoms from age 3 months through 24 months, increases in ambient nickel and vanadium concentrations were associated with increased probability of wheeze after adjusting for other pollutants, including NO2 (Patel et al., 2009). Increases in EC were associated with cough during the cold season. A study of cardiovascular and respiratory hospital admissions associated with PM2.5 chemical components in 106 US counties found that the effect of PM2.5 was significantly modified by the fraction of vanadium, elemental carbon, or nickel in the PM2.5 mass (Bell et al., 2009). Biological mechanisms explaining the impacts of PM2.5 on birth outcomes are not well understood, and mechanisms implicating specific constituents are still in the early stages of investigation. Particulate matter may affect fetal weight by impacting the general cardiovascular and respiratory health of the mother, as it has been shown to influence hemodynamic behavior, blood coagulation, and heart rate variability (Glinianaia et al., 2004). Pulmonary inflammation in the mother can reduce the levels of oxygen available to the fetus, and placental inflammation can impair gas and nutrient exchange (Lee et al., 2011). More specifically, transition metals and traffic-related PAHs bound to particles may cause oxidative stress and result in DNA damage to the fetus itself (Kannan et al., 2006). Metals can also accumulate in fetal tissues and may influence fetal growth at high levels of exposure (Domingo, 1994). Nutritional status of the mother may influence how these effects are manifested (Kannan et al., 2006). Although we were able to examine a large number of births throughout California, we were limited to the variables provided by the birth certificate database. For example, we did not have

R. Basu et al. / Environmental Research 128 (2014) 42–51

49

Fig. 3. Reductions in birth weight per IQR increase in gestational exposure to PM2.5 constituents, stratified by mother's race/ethnicity. Figure includes all species with statistically significant reductions in birth weight associated with full gestational exposure among entire study population (based on 95% confidence interval excluding zero). Models are adjusted for maternal age (less than 20, 20–24, 25–34, 35–39, greater than 39 years), maternal educational attainment (less than high school, high school graduate, some college, college graduate), weeks of gestational age, month of birth, year of birth, infant sex, region (north, south), and apparent temperature.

information on maternal or passive smoking, which have been shown to be associated with LBW (Lupo et al., 2010; Thurston et al., 2008). However, previous investigations did not find smoking to be a confounder for PM2.5 or CO and post-neonatal respiratory mortality (Darrow et al., 2006), or for PM2.5 and LBW specifically in California (Basu et al., 2003). Other covariates that we could not consider included maternal alcohol consumption, parity, and delivery method. Although these variables may be associated with LBW, we do not suspect that they would be related PM2.5 or PM2.5 constituents, and thus, would not be confounders. We were only able to consider maternal zip code at the time of giving birth, rather than maternal residence during the entire pregnancy. Investigations in other groups of pregnant women have reported that 15–30% of mothers changed residences during their pregnancy, but the impact on observed associations with health effects may be relatively low, as exposure classification often does not change substantially following a move (Chen et al., 2010; Wilhelm et al., 2012). We assigned exposure to all subjects living within 20 km of a central monitor, based on the zip code tabulation area of residence. Basu et al. (2004) compared exposure estimates from neighborhood PM2.5 monitors within five miles of maternal residences and exposure estimates from county-level monitors without any specified distance in their study of LBW, and found exposure measurements from the county-level monitors to yield the strongest associations, suggesting that exposure assessment for PM2.5 mass based on a central monitor can be relatively accurate over a large area. However, a 20 km radius may cover too large of an area to allow for accurate exposure of PM2.5 constituents. If certain PM2.5

constituents have a greater degree of spatial heterogeneity than total PM2.5 mass, it is possible that associations with those constituents may be underestimated due to higher misclassification of exposure over a 20 km radius. Also, since zip codes and zip code tabulation areas are not identical, using zip code tabulation arealevel census demographics may introduce a spatial mismatch, potentially increasing exposure misclassification (Grubesic and Matisziw, 2006). Furthermore, monitors are primarily designed to measure pollution in urban areas that are often heavily populated and polluted compared to non-urban or rural areas. In a few cases, however, monitors are actually located outside of the city. San Francisco, for example, does not have a pollution monitor located within the city, but rather in a nearby county. Since our study population was limited to those living within 20 km of a monitor, our study results may be more generalizable to populations living in urban environments. Using eight distinct monitoring areas in our study may have induced spatial correlation in our observations and introduced the possibility of confounding by area-level factors such as socioeconomic status. We attempted to alleviate the influence of any spatial correlation by adjusting region and for area-level socioeconomic predictors. However, there is a possibility of residual confounding by area-level factors if these adjustments were not sufficient. By limiting our study population to full-term births, we alleviated the fixed cohort bias that is often present in birth outcome studies using various gestational lengths (Strand et al., 2011). This study adds to the literature of particulate matter and adverse birth outcomes in general, a research need documented by

50

R. Basu et al. / Environmental Research 128 (2014) 42–51

Fig. 4. Reductions in birth weight per IQR increase in gestational exposure to PM2.5 constituents, stratified by mother's age (in years). Figure includes all species with statistically significant reductions in birth weight associated with full gestational exposure among entire study population (based on 95% confidence interval excluding zero). Models are adjusted for adjusted for maternal ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic, non-Hispanic Asian), maternal educational attainment (less than high school, high school graduate, some college, college graduate), weeks of gestational age, month of birth, year of birth, infant sex, region (north, south), and apparent temperature.

the US EPA (US EPA, 2005). It is one of very few studies to date to examine a full range of PM2.5 constituents and LBW, and the first in California. It considers possible effect modification by individual maternal demographics as well as confounding by communitylevel socioeconomic status. Exposure to specific constituents of PM2.5, especially metals, was associated with an increased risk of LBW in California. Since the elemental composition of PM2.5 differs by region, studies of PM2.5 constituents and adverse birth outcomes in other locales are warranted.

Disclaimer The opinions expressed in this paper are those of the authors, and do not represent those of the California Environmental Protection Agency or the Office of Environmental Health Hazard Assessment. The authors declare they have no competing financial interests. References Alexander, G.R., et al., 1996. A United States national reference for fetal growth. Obstet. Gynecol. 87, 163–168. ArcGIS, Version 9.2, Environmental Systems Research Institute, Redlands, CA. Basu, R., et al., 2003. Does Maternal Smoking Confound the Relationship Between PM2.5 and Birthweight? (Oral Presentation). American Public Health Association annual meeting, San Francisco. Basu, R., et al., 2004. Comparing exposure metrics in the relationship between PM2.5 and birth weight in California. J. Expo. Anal. Environ. Epidemiol. 14, 391–396.

Bell, M.L., et al., 2012. Relationship between birth weight and exposure to airborne fine particulate potassium and titanium during gestation. Environ. Res.. Bell, M.L., et al., 2010. Prenatal exposure to fine particulate matter and birth weight: variations by particulate constituents and sources. Epidemiology 21, 884–891. Bell, M.L., et al., 2009. Emergency hospital admissions for cardiovascular diseases and ambient levels of carbon monoxide: results for 126 United States urban counties, 1999–2005. Circulation 120, 949–955. Blanchard, C., 2003. Spatial and temporal characterization of particulate matter. In: McMurry, P.H., et al. (Eds.), Particulate Matter Science for Policy Makers: A NARSTO Assessment. Cambridge University Press, Cambridge. (Chapter 6). Blumenshine, P., et al., 2010. Socioeconomic disparities in adverse birth outcomes: a systematic review. Am. J. Prev. Med. 39, 263–272. California Air Resources Board, 2011. Air Quality Data Section, Planning and Technical Support Division. California Department of Health Services, 2006. Birth Statistical Master Files, 1999– 2006 Office of Vital Statistics, Sacramento, CA. Caudri, D., et al., 2007. Respiratory symptoms in the first 7 years of life and birth weight at term: the PIAMA Birth Cohort. Am. J. Respir. Crit. Care Med. 175, 1078–1085. Chen, L., et al., 2010. Residential mobility during pregnancy and the potential for ambient air pollution exposure misclassification. Environ. Res. 110, 162–168. Darrow, L.A., et al., 2011. Ambient air pollution and birth weight in full-term infants in Atlanta, 1994–2004. Environ. Health Perspect. 119, 731–737. Darrow, L.A., et al., 2006. Maternal smoking as a confounder in studies of air pollution and infant mortality. Epidemiology 17, 592–593. Domingo, J.L., 1994. Metal-induced developmental toxicity in mammals: a review. J. Toxicol. Environ. Health 42, 123–141. Donahue, S.M., et al., 2010. Trends in birth weight and gestational length among singleton term births in the United States. Obstet. Gynecol. 115, 357–364. Ebisu, K., Bell, M.L., 2012. Airborne PM2.5 chemical components and low birth weight in the northeastern and mid-Atlantic regions of the United States. Environ. Health Perspect. 120, 1746–1752. Garg, B.D., et al., 2000. Brake wear particulate matter emissions. Environ. Sci. Technol. 34, 4463–4469. Glinianaia, S.V., et al., 2004. Particulate air pollution and fetal health: a systematic review of the epidemiologic evidence. Epidemiology 15, 36–45.

R. Basu et al. / Environmental Research 128 (2014) 42–51

Grubesic, T.H., Matisziw, T.C., 2006. On the use of ZIP codes and ZIP code tabulation areas (ZCTAs) for the spatial analysis of epidemiological data. Int. J. Health Geogr. 5, 58. Hauck, F.R., et al., 2011. Racial and ethnic disparities in infant mortality. Semin. Perinatol. 35, 209–220. Johnson, R.C., Schoeni, R.F., 2011. Early-life origins of adult disease: national longitudinal population-based study of the United States. Am. J. Public Health 101, 2317–2324. Kannan, S., et al., 2006. Exposures to airborne particulate matter and adverse perinatal outcomes: a biologically plausible mechanistic framework for exploring potential effect modification by nutrition. Environ. Health Perspect. 114, 1636–1642. Kochanek, K.D., et al., 2012. Annual summary of vital statistics: 2009. Pediatrics 129, 338–348. Lee, P.C., et al., 2011. Particulate air pollution exposure and C-reactive protein during early pregnancy. Epidemiology 22, 524–531. Lhila, A., Long, S., 2012. What is driving the black–white difference in low birthweight in the US? Health Econ. 21, 301–315. Lupo, P.J., et al., 2010. Differences in exposure assignment between conception and delivery: the impact of maternal mobility. Paediatr. Perinat. Epidemiol. 24, 200–208. Mathews, T.J., MacDorman, M.F., 2007. Infant mortality statistics from the 2004 period linked birth/infant death data set. Natl. Vital. Stat. Rep. 55, 1–32. McDonald, J.A., et al., 2008. Reproductive health of the rapidly growing Hispanic population: data from the Pregnancy Risk Assessment Monitoring System, 2002. Matern. Child. Health J. 12, 342–356. Morello-Frosch, R., et al., 2010. Ambient air pollution exposure and full-term birth weight in California. Environ. Health 9, 44. National Oceanic and Atmospheric Administration (NOAA), 2012. National Climatic Data Center. Climate Resources. O'Neill, M.S., et al., 2003. Health, wealth, and air pollution: advancing theory and methods. Environ. Health Perspect. 111, 1861–1870. Office of Water Use Efficiency, 2009. C. D. o. W. R., California Irrigation Management System (CIMIS). Ostro, B., et al., 2007. The effects of components of fine particulate air pollution on mortality in California: results from CALFINE. Environ. Health Perspect. 115, 13–19.

51

Parker, J.D., et al., 2005. Air pollution and birth weight among term infants in California. Pediatrics 115, 121–128. Patel, M.M., et al., 2009. Ambient metals, elemental carbon, and wheeze and cough in New York City children through 24 months of age. Am. J. Respir. Crit. Care Med. 180, 1107–1113. Salihu, H.M., et al., 2011. Particulate pollutants and racial/ethnic disparity in fetoinfant morbidity outcomes. Matern. Child Health J.. SAS Institute Inc., 2008. SAS Statistical Software. Cary, NC. Schauer, J.J., et al., 2006. Characterization of metals emitted from motor vehicles. Res. Rep. Health Eff. Inst., 1–76. (Discussion 77–88). Shah, P.S., Balkhair, T., 2011. Air pollution and birth outcomes: a systematic review. Environ. Int. 37, 498–516. Strand, L.B., et al., 2011. Methodological challenges when estimating the effects of season and seasonal exposures on birth outcomes. BMC Med. Res. Methodol. 11, 49. Thurston, R.C., et al., 2008. Hot flashes and subclinical cardiovascular disease: findings from the Study of Women's Health Across the Nation Heart Study. Circulation 118, 1234–1240. U.S. EPA, 2005. Review of the National Ambient Air Quality Standards for Particulate Matter: Policy Assessment of Scientific and Technical Information, OAQPS Staff Paper. Office of Air Quality Planning and Standards. Research Triangle Park, NC. EPA-452/R-05-005. U.S. EPA, 2009. Air Quality System Data Mart. Wiles, N.J., et al., 2005. Birth weight and psychological distress at age 45–51 years: results from the Aberdeen Children of the 1950s cohort study. Br. J. Psych. 187, 21–28. Wilhelm, M., et al., 2012. Traffic-related air toxics and term low birth weight in Los Angeles County, California. Environ. Health Perspect. 120, 132–138. Wilhelm, M., Ritz, B., 2005. Local variations in CO and particulate air pollution and adverse birth outcomes in Los Angeles County, California, USA. Environ. Health Perspect. 113, 1212–1221. Zhou, J., et al., 2011. Time-series analysis of mortality effects of fine particulate matter components in detroit and Seattle. Environ. Health Perspect. 119, 461–466.