Environmental Research 140 (2015) 414–420

Contents lists available at ScienceDirect

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

Fine particulate matter and the risk of autism spectrum disorder Evelyn O. Talbott a,n, Vincent C. Arena b, Judith R. Rager a, Jane E. Clougherty c, Drew R. Michanowicz c, Ravi K. Sharma d, Shaina L. Stacy c a

Department of Epidemiology, University of Pittsburgh Graduate School of Public Health, 130 De Soto Street, Pittsburgh, PA 15261, USA Department of Biostatistics, University of Pittsburgh Graduate School of Public Health, 130 De Soto Street, Pittsburgh, PA 15261, USA c Department of Environmental and Occupational Health, University of Pittsburgh Graduate School of Public Health, Bridgeside Point I, 100 Technology Drive, Pittsburgh, PA 15219, USA d Department of Behavioral and Community Health Sciences, University of Pittsburgh Graduate School of Public Health, 130 De Soto Street, Pittsburgh, PA 15261, USA b

art ic l e i nf o

a b s t r a c t

Article history: Received 3 March 2015 Received in revised form 24 April 2015 Accepted 28 April 2015

The causes of autism spectrum disorder (ASD) are not well known. Recent investigations have suggested that air pollution, including PM2.5, may play a role in the onset of this condition. The objective of the present work was to investigate the association between prenatal and early childhood exposure to fine particulate matter (PM2.5) and risk for childhood ASD. A population-based case-control study was conducted in children born between January 1, 2005 and December 31, 2009 in six counties in Southwestern Pennsylvania. ASD cases were recruited from specialty autism clinics, local pediatric practices, and school-based special needs services. ASD cases were children who scored 15 or above on the Social Communication Questionnaire (SCQ) and had written documentation of an ASD diagnosis. Controls were children without ASD recruited from a random sample of births from the Pennsylvania state birth registry and frequency matched to cases on birth year, gender, and race. A total of 217 cases and 226 controls were interviewed. A land use regression (LUR) model was used to create person- and time-specific PM2.5 estimates for individual (pre-pregnancy, trimesters one through three, pregnancy, years one and two of life) and cumulative (starting from pre-pregnancy) key developmental time periods. Logistic regression was used to investigate the association between estimated exposure to PM2.5 during key developmental time periods and risk of ASD, adjusting for mother's age, education, race, and smoking. Adjusted odds ratios (AOR) were elevated for specific pregnancy and postnatal intervals (pre-pregnancy, pregnancy, and year one), and postnatal year two was significant, (AOR¼ 1.45, 95% CI¼ 1.01–2.08). We also examined the effect of cumulative pregnancy periods; noting that starting with pre-pregnancy through pregnancy, the adjusted odds ratios are in the 1.46–1.51 range and significant for pre-pregnancy through year 2 (OR ¼1.51, 95% CI¼ 1.01–2.26). Our data indicate that both prenatal and postnatal exposures to PM2.5 are associated with increased risk of ASD. Future research should include multiple pollutant models and the elucidation of the biological mechanism for PM2.5 and ASD. & 2015 Elsevier Inc. All rights reserved.

Keywords: Autism spectrum disorder Case control study Fine particulate matter Geographic information systems Land use regression Odds ratio

1. Introduction Autism spectrum disorders (ASD) constitute a major public health problem, affecting approximately one in every 68 children and their families (CDC, 2014). ASDs are brain development disorders usually diagnosed in childhood and characterized by impaired social interaction and communication, and by restricted and repetitive behaviors (APA, 2000). Given that ASDs are lifelong n

Corresponding author. Fax: þ1 412 624 7397. E-mail addresses: [email protected] (E.O. Talbott), [email protected] (V.C. Arena), [email protected] (J.R. Rager), [email protected] (J.E. Clougherty), [email protected] (D.R. Michanowicz), [email protected] (R.K. Sharma), [email protected] (S.L. Stacy). http://dx.doi.org/10.1016/j.envres.2015.04.021 0013-9351/& 2015 Elsevier Inc. All rights reserved.

conditions for which there is no cure and for which treatment options are limited, there is a need to identify modifiable risk factors for these disorders. A recent comprehensive review of environmental chemical exposures and ASD was completed by Kalkbrenner et al. (2014a, 2014b). In this review, she and her colleagues identified studies of autism and estimates of exposure to tobacco, air pollutants, volatile organic compounds and solvents, metals as well as pesticides and organic endocrine disrupting compounds such as flame retardants, phthalates and bisphenol A. Several recent studies have focused on the association between exposure to air pollution (traffic related and from industry) and autism. Volk et al. (2011, 2013) examined the relationship between traffic-related air pollution and autism in California in two

E.O. Talbott et al. / Environmental Research 140 (2015) 414–420

population based case-control studies from the Childhood Autism Risks from Genetics and the Environment (CHARGE) study. The first study population (Volk et al., 2011) consisted of 304 autism cases and 259 typically developing controls. In addition to the address on the birth certificate, a detailed residential history from three months before conception until the current address was obtained by personal interview. They found that children who lived within 309 m (  1000 feet) of a freeway (birth address) had 1.86 the odds (95% CI ¼1.04–3.45) of having autism as children who lived 41149 m from a freeway, after adjusting for smoking during pregnancy, gender, ethnicity, education of parents and maternal age. The second study (Volk et al., 2013) used modelbased estimates of traffic related air pollution and EPA air quality system data for PM2.5, PM10, ozone, and nitrogen dioxide. Children with autism were more likely to have lived at residences that had the highest quartile of exposure to traffic-related air pollutants during gestation (OR ¼1.98, 95% CI ¼1.20–3.31) and during the first year of life (OR ¼3.10, 95% CI ¼ 1.76–5.57) compared with control children. For PM2.5 exposure during gestation, the adjusted odds ratio (AOR) was 2.08 (95% CI ¼1.93–2.25) and for the first year of life the AOR was 2.12 (95% CI ¼1.45–3.10) per 8.7 mg/m3. Other investigations of air pollution and autism have found an increased risk for autism associated with exposure to PM2.5 (Becerra et al., 2013; Raz et al., 2014) and PM10 (Kalkbrenner, 2015). Becerra estimated exposures with data from air monitoring stations and a land use regression model in a California study. She reported a 15% increase in the odds of autism (OR ¼1.15, 95% CI ¼1.06–1.24) per 4.68 mg/m3 increase in PM2.5 exposure during pregnancy. Kalkbrenner used birth address to assign exposure to PM10 to each child by a geostatistical interpolation method using daily concentrations for air pollution monitors. She identified 334 children with ASD in the San Francisco Bay area and 645 cases in Northern California born between 1994 and 2000 and compared with a random sample of children born in the same counties and years. The adjusted OR was significant for the third trimester only (AOR ¼1.36, 95% CI ¼1.13–1.63). Raz et al. identified 245 cases of ASD and 1522 frequency matched controls born 1990–2002 to women in the Nurses' Health Study II. Monthly averages of PM2.5 and PM10–2.5 predicted from a spatiotemporal model for the continental US were linked to residential address at birth. PM2.5 was associated with an increased odds of ASD during pregnancy of 1.57 (95% CI ¼1.22–2.03) per interquartile range (IQR) (4.42 μg/m3) increase of PM2.5. The strongest association was seen during the third trimester (OR ¼1.42, 95% CI ¼1.09–1.86 per IQR increase in PM2.5) than other trimesters. Although the time periods of the investigations, timing of the exposure periods and exposure assessments themselves are varied, the results of these studies suggest a potential role for air pollution, including PM2.5, in the development of autism. We sought to explore the association between ASD and exposure to PM2.5 during critical prenatal and postnatal periods in a case-control study in the greater Pittsburgh, Pennsylvania region. We used data from systematic air monitoring and land use regression models to develop PM2.5 exposure measures. This provided more spatially resolved exposure estimates for the entire study area than the use of only monitor-based values. We obtained residential addresses from personal interviews for the entire period of fetal and postnatal development. Thus, we were able to compute person-specific exposures to PM2.5 based on geospatial and time-specific exposure estimates during critical developmental periods for each study participant. 2. Materials and methods This study was approved by the University of Pittsburgh Institutional Review Board (IRB number (PRO10010240).

415

2.1. Study population We conducted a population-based case-control study of autism spectrum disorder (ASD) in southwestern Pennsylvania. Cases and controls for this study were children who were born between January 1, 2005 and December 31, 2009 in Allegheny, Armstrong, Beaver, Butler, Washington, or Westmoreland County and were currently residing in the six-county area. Children were not included in the study if the child was adopted, parents were not English speaking, or neither parent was available for interview. Cases of ASD were recruited from specialty autism clinics and treatment centers, from local pediatric and family medicine practices, and through the Intermediate Units of the Pennsylvania School System who provide services to special needs children. We conducted an extensive outreach and recruitment program to identify children with ASD in the six-county area of our study. A case of ASD was defined as any child who 1) scored a 15 or above on the Social Communication Questionnaire (SCQ), a screen for the presence of autistic features and 2) had written documentation, including the Autism Diagnostic Observation Schedule (ADOS) or other diagnostic test results, of a diagnosis of an ASD from a child psychologist or psychiatrist. Controls were recruited from a random selection of births from the Pennsylvania Department of Health (PADOH) state birth registry files for 2005 to 2009 in the six-county area and were frequency matched to the cases on year of birth, gender, and race. We recruited through a direct letter appeal signed by the Pennsylvania Secretary of Health. After we obtained informed consent, parents were screened for inclusion criteria and administered the SCQ for their child. Children with an SCQ Z 15 or with a reported diagnosis of ASD were not included as controls. For each of the cases and controls, a personal interview with the mother was conducted by trained interviewers using a structured questionnaire, adapted from the CDC's Study to Explore Early Development (SEED). The questionnaire included parental demographic and socioeconomic information, maternal and paternal occupational history, family history of ASD, smoking history, maternal reproductive and pregnancy history, child's medical history, and a detailed residential history. Data was obtained on all residential addresses and the corresponding start and end dates that the mother/child lived at those addresses from three months prior to last menstrual period (LMP) until the child's second birthday. A total of 217 cases and 226 controls met the criteria and were interviewed for the study. 2.2. PM2.5 measurements Regulatory air pollution monitoring networks are generally sparsely distributed, and therefore do not typically capture a representative range of exposures across urban environments (e.g., near heavily trafficked roads). Accordingly, we designed a Pittsburgh-area multi-pollutant air monitoring campaign, detailed in Shmool et al. (2014). 2.3. Merged-seasons LUR model Land use regression (LUR) models were constructed using manual forward step-wise linear regression, following methods adapted from Clougherty et al. (2013). Temporal variability was captured using a regional background site upwind of the study area, away from local sources. Comparable models were found to best explain spatial variance in PM2.5 in both seasons (Tunno et al., in press), suggesting consistency in source impacts year-round. Therefore, to increase sample size for final model fits, and to produce a single spatially-refined surface for temporal extrapolation, we constructed a merged-seasons model by combining both

416

E.O. Talbott et al. / Environmental Research 140 (2015) 414–420

seasonal datasets (n ¼74 measures). To account for repeated measures in modeling across seasons, a random intercept with an independent unstructured covariance was used (p¼ 0.003) in a mixed model with restricted maximum likelihood estimation. 2.4. Temporal model extrapolation To extrapolate the spatial estimates provided by the LUR over the length of study—and to enable exposure estimation at various time scales (e.g., weekly, or averaged to a specific trimester)—we used regional daily PM2.5 measures from EPA's regulatory Air Quality System (AQS), as in Ross et al. (2013). We then examined (1) temporal trends in PM2.5 data from each regulatory monitor across the six-county region, to establish the presence of a reasonably common regional trend observed across monitors, which coincided with regional background measures (two summer; two winter seasons) obtained during our dedicated sampling campaigns; (2) data quality (e.g., comparable sampling method, agreement between co-located monitors, non-systematic missingness); (3) representativeness of a greater regional trend of Southwestern PA from 2000-present; and (4) interpretability. A single AQS monitor (Thermo Scientific TEOM single point monitor), located outside of the urban core of Pittsburgh in a mixed commercial/residential area (FRM-Lawrenceville-88101-1), was co-located with multiple PM2.5 monitors (e.g., FRM filterbased, FEM continuous Met One BAM), reducing uncertainty in imputing missing values, and demonstrated a representative temporal trend for the region. There were only 176 missing days over 11 years (2003–2013). Daily measures from the AQS site was matched and averaged to our weekly sampling sessions, then substituted for the regional background measures in previously-built seasonal and mergedseasons LURs. All prior explanatory variables were retained (p o0.05), and thus we did not reconstruct LURs, and observed that the existing geographic covariates captured the spatial variation in intraurban PM. To corroborate the merged summer/winter model, an arithmetic mean concentration value was calculated from the central AQS site data corresponding to the 12 sampling weeks (six summer/six winter) (global mean ¼12.0 mg/m3). The original model (Table 1) was applied to the six-county area (  7320 km2), creating predictions at 100 m2 grid cells. To reduce edge effects and to avoid inflation (overestimation) when extrapolating outside the sampling domain, we constrained the upper bound of concentrations attributable to SO2 emissions at the highest value observed within our original domain.

Table 1 Combined season mixed-effects model predicting PM2.5. Covariates predicting PM2.5

B (p-Value)

Seq. R2

RMSE

Intercept Lawrenceville reference PM SO2 emissions 300 m Signaled intersections 750 m

 2.48 1.17 (o .0001) 5.0  10  3 ( o .0001) 0.08 (.003)

– 0.70 0.74 0.77

– 1.85 1.70 1.63

where PMi equals the predicted PM concentration at location i, and PMm equals the six-county-wide mean concentration. Therefore, [PMi/PMm] equates to the spatial concentration ratio from the 2012 annual PM model. Spatiotemporal specific concentration estimates were given by Eq. (2)

PMij = PMj * ⎡⎣PMi /PMμ ⎤⎦

(2)

where PMij equals the PM2.5 concentration at location i at time j (24-h concentrations) from the FRM monitor, to calculate daily PM2.5 exposure estimates for all residences and time periods. 2.6. Exposure assignments The calculations above were applied, accounting for changes in residence. For each child, average and cumulative exposure estimates were computed for key developmental time periods of the three months prior to LMP, the trimesters of pregnancy (LMP-12 weeks, 13–24 weeks, 25þ weeks) and first and second year of life as well as for combinations of those time periods (pre-pregnancy through Trimester 1, pre-pregnancy through Trimester 2, prepregnancy through pregnancy, pre-pregnancy through 1st year of life, pre-pregnancy through 2nd year of life, trimester one through two, pregnancy through 1st year of life, pregnancy through 2nd year of life, and first two years of life). 2.7. Statistical analyses We used multiple logistic regression to examine the association between exposure to PM2.5 and risk of autism spectrum disorder. We performed separate logistic regression models for exposure during each of a number of critical prenatal and postnatal time periods. We calculated ORs and 95% confidence intervals per interquartile range (IQR) increase in PM2.5 exposure based on the pregnancy time period IQR for controls. Analyses were adjusted for maternal age, maternal education (college graduate/not college graduate), maternal race (nonwhite/white), and maternal smoking during pregnancy (yes/no).

2.5. PM2.5 exposure predictions Residence-specific estimates of outdoor PM2.5 concentrations were assigned to each home, based on residential histories obtained via parental interview. Data was obtained on the corresponding start and end dates that the mother/child lived at the address from three months prior to pregnancy until the child's second birthday. Each address was geocoded to a latitude–longitude coordinate using ArcGIS 10.1 (ESRI Inc., Redlands, CA) and checked manually. When an address could not be successfully geocoded in ArcGIS, other methods were used, including MapQuest Latitude/Longitude Finder (http://developer.mapquest.com/ web/tools/lat-long-finder). To calculate time- and location-specific PM2.5 estimates, a concentration ratio was applied to each member of the cohort at both the point residence, and averaged within a 300 m buffer, given by Eq. (1)

PMi /PMμ

(1)

3. Results Of 217 cases and 226 controls interviewed for the study, six cases and seven controls were excluded because at least one of their residences was outside of the six-county area and therefore estimated exposures for PM2.5 were not available. A total of 211 cases and 219 controls were included in the statistical analyses. Table 2 presents characteristics of the 211 ASD cases and 219 controls. The majority of both cases (60.7%) and controls (59.4%) were born in Allegheny County, the most populated of the six counties. Mothers of ASD cases were younger overall than mothers of controls, with a mean (SD) maternal age of 30.4 (5.4) compared to 31.8 (4.6). However, although not statistically significant, case mothers were slightly more likely than control mothers to be age 40 or older (4.7% compared to 3.7%). The mean (SD) age of the case fathers was 32.7 (6.0) and the mean age of the control fathers was 33.7 (5.6). Compared to mothers of controls, mothers of ASD cases

E.O. Talbott et al. / Environmental Research 140 (2015) 414–420

Table 2 Demographic and prenatal characteristics of ASD cases and controls. Characteristics

ASD cases (n¼ 211) N (%)

Child sex Male 164 (77.7) Female 47 (22.3) Year of birth 2005 41 (19.4) 2006 58 (27.5) 2007 46 (21.8) 2008 37 (17.5) 2009 29 (13.7) County at birth Allegheny 128 (60.7) Armstrong 7 (3.3) Beaver 5 (2.4) Butler 17 (8.1) Washington 22 (10.4) Westmoreland 32 (15.2) Maternal age o 25 29 (13.7) 25–29 64 (30.3) 30–34 66 (31.3) 35–39 42 (19.9) 40þ 10 ( 4.7) Maternal age – mean(SD) 30.4 (5.4) a Paternal age – mean(SD) 32.7 (6.0) Maternal race White 188 (89.1) Black 15 (7.1) Other/more than one race 8 (3.7) Maternal education at time of birth High school graduate or less 29 (13.7) Some college, technical or associates 65 (30.8) degree College graduate or advanced degree 117 (55.4) Maternal smoking during pregnancy or in 3 months prior to pregnancy Yes 52 (24.6) No 159 (75.4) Multiple birth 18 (8.5) Low birth weight birth ( o 2500 g)b 23 (11.0) (from birth certificate) 31 (15.2) Preterm birth ( o 37 weeks)c (from birth certificate) a b c

Controls (n¼219) N (%)

Table 3 Descriptive statistics of PM2.5 (μg/m3) estimated exposure in ASD cases (n¼211) and controls (n¼ 219) during specific prenatal and postnatal time periods. Time period

Mean SD

25th percentile

Median 75th percentile

Case Control Case Control Case Control Case Control Case Control Case

15.3 15.1 15.3 15.0 14.8 14.6 14.9 14.7 15.0 14.8 15.1

3.3 3.1 3.7 3.4 3.4 3.4 3.4 3.3 1.9 1.8 1.4

12.8 12.5 12.5 12.4 12.3 12.2 12.5 12.2 13.4 13.3 14.5

14.1 14.1 13.4 13.8 13.4 13.2 13.5 13.3 15.2 14.8 15.1

17.4 17.0 18.5 17.2 17.0 16.7 16.8 16.8 16.2 16.2 15.7

Control Case Control Year 2 Case Control Year 1–2 Case Control Pregnancy through Case year 1 Control Pregnancy through Case year 2 Control Pre-pregnancy Case through year 1 Control Pre-pregnancy Case through year 2 Control

14.8 14.4 14.1 13.6 13.3 14.0 13.7 14.6

1.4 1.5 1.6 1.6 1.6 1.5 1.5 1.5

14.3 13.1 12.8 12.2 11.9 12.5 12.3 13.8

15.0 14.6 14.4 13.6 13.1 14.2 14.1 14.8

15.6 15.3 15.2 14.8 14.5 14.9 14.7 15.6

14.4 14.3

1.5 13.0 1.4 13.3

14.5 14.4

15.4 15.2

14.0 14.7

1.5 12.6 1.4 14.1

14.2 14.9

15.0 15.5

14.5 14.3

1.4 13.7 1.4 13.4

14.7 14.5

15.3 15.1

14.1

1.4 13.1

14.2

15.0

Pre-pregnancy 169 (77.2) 50 (22.8)

Trimester 1

50 42 46 39 42

(22.8) (19.2) (21.0) (17.8) (19.2)

Trimester 2

130 2 18 19 22 28

(59.4) (0.9) (8.2) (8.7) (10.0) (12.8)

Pre-pregnancy through pregnancy

11 53 91 56 8 31.8 33.7

(5.0) (24.2) (41.6) (25.6) ( 3.7) (4.6) (5.6)

212 (96.8) 4 (1.8) 3 (1.4) 17 ( 7.7) 29 (13.2)

417

Trimester 3 Pregnancy

Year 1

173 (79.0)

23 196 8 9

(10.5) (89.5) (3.7) (4.1)

20 (9.3)

Missing for 2 cases, 3 controls. Missing for 1 case. Missing for 7 cases, 3 controls.

were significantly less likely to have a college degree (55.4% of cases and 79.0% of controls.) Cigarette smoking at any time from three months prior to pregnancy until birth was significantly higher among mothers of ASD cases (24.6%) compared to mothers of controls (10.5%). ASD cases were more likely to be born low birth weight (o 2500 g) and/or preterm ( o37 weeks) and were more likely to be a multiple birth compared to controls. Shown inTable 3 are descriptive statistics for PM2.5 estimated exposures during the prenatal and postnatal time periods. Mean (SD) PM2.5 estimated exposure during pregnancy was 15.0 (1.9) μg/m3 for cases and 14.8 (1.8) for controls. The IQR was 2.80 μg/m3 for cases and 2.84 μg/m3 for controls. Tables 4–5 present the adjusted (AOR) and unadjusted odds ratios of ASD for a 2.84 μg/m3 (IQR) increase in PM2.5 during specific developmental periods. The odds ratios are elevated for all individual time periods of pre-pregnancy, pregnancy, year one and year two (Table 4), with significant odds ratios for year 2 (AOR ¼1.45, 95% CI ¼1.01–2.08, p ¼0.042). Table 5 and Fig. 1(a) and (b) show pregnancy periods aggregating each time period from pre-pregnancy onward to the next interval both adjusted and

unadjusted for covariates. This was done to examine the effect of the cumulative average exposure of PM2.5 beginning before pregnancy. We note that starting with the pre-pregnancy through pregnancy interval, the adjusted odds ratios are in the 1.46–1.51 range with confidence intervals very close or at significance. Significant odds ratios with confidence limits that do not include one are seen for year two (OR ¼1.45, 95% CI ¼1.01–2.08) and for prepregnancy through year two (OR¼ 1.51, 95% CI ¼ 1.01–2.26). We also examined the cumulative pregnancy periods starting from trimester one, rather than from pre-pregnancy, through year two. The odds ratios for all these periods were elevated, but none were statistically significant (not shown).

4. Discussion In our case-control study of ASD conducted in Southwestern Pennsylvania, exposure to PM2.5 during the period from three months before pregnancy through year two as well as year two alone were associated with increased risk of childhood ASD. Our results showed an approximate 50% increase in risk of ASD associated with an average cumulative exposure from three months before the pregnancy through age two of the child (p ¼0.046). Our effect estimates for PM2.5 were elevated but did not reach significance for individual pregnancy periods of T1, T2, or T3 or for the total pregnancy alone (OR¼ 1.20, 95% CI ¼0.88–1.63). This suggests that the pre-pregnancy exposure to the mother may have some importance. Our results are generally consistent with previous studies of PM2.5 and autism. Becerra et al. (2013) in a population based study in Los Angeles assigned exposure to PM2.5 based on air monitoring stations and land use regression models. She found a 15% increase

418

E.O. Talbott et al. / Environmental Research 140 (2015) 414–420

Table 4 Odds ratios of ASD per 2.84 μg/m3 (IQR) increase in average exposure to PM2.5 (μg/m3) during prenatal and postnatal periods. Time period

Pre-pregnancy Trimester 1 Trimester 2 Trimester 3 Pregnancy Year 1 Year 2 a

Adjusted odds ratios (AOR)a

Unadjusted odds ratios (OR) OR

Lower 95% CI

Upper 95% CI

p-Value

OR

Lower 95% CI

Upper 95% CI

p-Value

1.07 1.06 1.04 1.07 1.23 1.33 1.41

0.91 0.91 0.89 0.91 0.92 0.94 1.00

1.27 1.23 1.23 1.26 1.64 1.88 1.98

0.402 0.484 0.592 0.411 0.172 0.106 0.053

1.13 1.07 1.04 1.04 1.20 1.37 1.45

0.94 0.91 0.88 0.88 0.88 0.95 1.01

1.35 1.25 1.22 1.24 1.63 1.97 2.08

0.182 0.416 0.674 0.620 0.251 0.088 0.042

Adjusted for college education, smoking, race, and mom's age.

in odds of primary autism for a 4.68 mg/m3 increase in PM2.5 (OR ¼1.15, 95% CI ¼ 1.06–1.24), adjusting for ozone. Becerra did not however have access to complete residential history during pregnancy and relied on residence on the birth certificate to estimate exposure. Volk et al. (2013) assigned air quality data from monitoring stations located within 50 km of each residence using USEPA air quality system data for 1997–2009. Volk utilized the mother's address from the birth certificate and address report from a residential history questionnaire to estimate PM2.5 for each trimester of pregnancy and during the first year of life. Volk noted an approximate two-fold increase in odds of primary autism per 8.7 mg/m3 increase in PM2.5 for the total pregnancy (AOR ¼2.08, 95% CI ¼1.93–2.25) and for year one (AOR ¼2.12, 95% CI ¼1.45– 3.10). Similar to our study, the magnitude of the associations were higher during late gestation and early postnatal life. Raz et al. (2014) found an effect for overall pregnancy but did not note a difference nine months after birth. In these three studies, risk estimates are higher than those for our study due to higher interquartile ranges. We chose to use maternal age as the adjustment variable instead of paternal age, which has also been cited as a risk factor for ASD. In our dataset, maternal and paternal ages were highly correlated with each other. Further, maternal age was more statistically significant than paternal age in logistic regression models of ASD risk, with PM2.5 level and mother's race, education, and smoking as covariates. A recent Swedish study by Idring et al. (2014) considered parental age and risk of autism in 4746 ASD cases and 412,557 controls born between 1984 and 2003. Comparatively speaking, the risk of offspring ASD was greater for a given maternal age than for the same paternal age. Advancing maternal age increased the risk of ASD regardless of paternal age, whereas advancing paternal age linearly increased risk of offspring ASD only in mothers less than 35 years. A strength of our study was that we were able to evaluate exposure during specific individual as well as cumulative pregnancy/postnatal intervals, basing average exposure estimates on daily PM2.5 levels. In addition, an LUR model, developed using Pittsburgh-area measurements of PM2.5, provided more spatially resolved exposure estimates for the entire study area than the use

of monitor-based values alone could offer. LUR thus enabled us to make better estimates of PM2.5 for areas that did not have monitors close by, and this most likely resulted in more accurate exposure assignments. Our study did not take into account simultaneous exposures to other pollutants and, as a PM2.5 estimate was calculated for the residential location only, assumed that there was little mobility outside of the residence of the mother during this time (i.e., commuting to work, etc.).With respect to mobility outside of the home, Nethery et al. (2009) studied changes in location based activity in a small sample of 62 pregnant women mostly in their second (N ¼62) and third trimester (N ¼54) of pregnancy in 2005–2006 in a metropolitan area of Vancouver, British Columbia. Women were recruited through word of mouth, at prenatal yoga, fitness or prenatal classes. Overall, 91% of the Nethery pregnancy cohort (129 samples and 62 women) worked full or part time sometime during their pregnancy compared to 76% of a larger CHAPS (Canadian Human Activity Patterns Survey) female population sample (n¼ 103). Moreover, the pregnancy cohort worked on average 4.2 (3.6–4.7) hours per day with 0.9 h (0.7–1.0) spent in the car each day. Their data showed that there was a significant trend of increased time at home by trimester of pregnancy and decreased time at work: 5.57 h (4.44–6.70) in the first trimester (N ¼11), 4.26 h (3.46–5.06) in the second trimester (N ¼62), and 3.67 h (2.76–4.58) in the third trimester (N ¼ 54). They believed this underscored the importance of considering increased exposures in or near the home and reduced emphasis on work-place exposures. Upon examination of the maternal work history during pregnancy for the cases and controls in our study, we found that 65.4% (138) of 211 case mothers and 70.6% (154) of 218 control mothers worked 10 or more hours outside the home at some time during their pregnancy. The majority of mothers worked full-time with 52.2% of all case mothers and 56.4% of all control mothers reporting that they worked 30 or more hours outside the home. The mean hours worked was similar between case (23.7 719.3 h per week) and control mothers (25.4 718.1 h per day). The straight line distances were also calculated between the residence and the job location. We focused on first job and not all jobs. However, 94.5% of working mothers had only one job during the pregnancy

Table 5 Odds ratios of ASD per 2.84 μg/m3 (IQR) increase in average exposure to PM2.5 (μg/m3) during cumulative developmental periods beginning with pre-pregnancy. Time period

Pre-pregnancy through Trimester 1 Pre-pregnancy through Trimester 2 Pre-pregnancy through pregnancy Pre-pregnancy through year 1 Pre-pregnancy through year 2 a

Adjusted odds ratios (AOR)a

Unadjusted odds ratios (OR) OR

Lower 95% CI

Upper 95% CI

p-Value

OR

Lower 95% CI

Upper 95% CI

p-Value

1.13 1.21 1.42 1.42 1.46

0.90 0.90 0.97 0.97 0.99

1.42 1.63 2.08 2.09 2.15

0.282 0.203 0.075 0.074 0.057

1.20 1.29 1.46 1.47 1.51

0.95 0.94 0.98 0.98 1.01

1.52 1.76 2.19 2.21 2.26

0.132 0.113 0.066 0.062 0.046

Adjusted for college education, smoking, race, and mom's age.

E.O. Talbott et al. / Environmental Research 140 (2015) 414–420

419

Fig. 1. (a) Unadjusted odds ratios and 95% CI per 2.84 μg/m3 (IQR) increase in average exposure to PM2.5 during cumulative developmental periods beginning with prepregnancy. (b) Adjusted odds ratios and 95% CI per 2.84 μg/m3 (IQR) increase in average exposure to PM2.5 during cumulative developmental periods beginning with prepregnancy.

period. In addition, about 85% of these mothers did not change residences during pregnancy. A total of 269 participants (125 cases and 144 controls) had valid occupational address information. ArcMap 10.1 was used to calculate the straight line distance between each mother's first residence and the first job. There was no difference in the mean distance between work and home for case (9.0777.10 miles) and control mothers (8.977 7.73). The distributions of estimated work-home distances were also similar between cases and controls (19% and 19% o3 miles, 18% and 12% 3–4.9 miles, 23% and 36% 5–9.9 miles, and 39% and 33% Z10 miles, respectively). Although the time spent at a workplace outside of the home during pregnancy and the distributions of distances to work were similar for case and control mothers, there is a potential for misclassification bias of PM2.5 exposure due to differential PM2.5 exposures outside of the home. 5. Conclusions Our data indicates that both prenatal and postnatal exposure to

PM2.5 play a role in incurring increased risk of ASD. From a historical perspective, air pollution levels have been declining since the 1990s; however, we know from our research that pockets of increased levels of air toxicants remain throughout our region and others. Confirmation of these findings would be important to determine their consistency across other regions with similar exposures and within different populations. In addition future research should include the consideration of multiple pollutant models, and the elucidation of the biological mechanisms for PM2.5 and ASD.

Conflict of interest The authors have declared no conflicts of interest. Funding source This study was supported by the Heinz Endowments.

420

E.O. Talbott et al. / Environmental Research 140 (2015) 414–420

Acknowledgments This study was supported by the Heinz Endowments (Grant number C1627) and approved by the University of Pittsburgh Institutional Review Board (IRB number PRO10010240).

References: APA, American Psychiatric Association, 2000. Diagnostic and Statistical Manual of Mental Disorders DSM-IV-TR. APA, Washington, DC. Becerra, T.A., Wilhelm, M., Olsen, J., Cockburn, M., Ritz, B., 2013. Ambient air pollution and autism in Los Angeles County, California. Child. Health 121 (3), 380–386. CDC, Centers for Disease Control. 2014. Prevalence of Autism Spectrum Disorder Among Children Aged 8 Years—Autism and Developmental Disabilities Monitoring Network, 11 Sites, United States, 2010. Morbidity and Mortality Weekly Report Surveillance Summaries. 63(SS02): 1–21. Clougherty, J.E., Kheirbek, I., Eisl, H.M., Ross, Z., Pezeshki, G., Gorczynski, J.E., Johnson, S., Markowitz, S., Kass, D., Matte, T., 2013. Intra-urban spatial variability in wintertime street-level concentrations of multiple combustion-related air pollutants: the New York City Community Air Survey (NYCCAS). J. Expo. Sci. Environ. Epidemiol. 23, 232–240. Idring, S., Magnusson, C., Lundberg, M., Ek, M., Rai, D., Svensson, A.C., Dalman, C., Karlsson, H., Lee, B.K., 2014. Parental age and the risk of autism spectrum disorders: findings from a Swedish population-based cohort. Int. J. Epidemiol. 43 (1), 107–115.

Kalkbrenner, A.E., Schmidt, R.J., Penlesky, A.C., 2014a. Environmental chemical exposures and autism spectrum disorders: a review of the epidemiological evidence. Curr. Probl. Pediatr. Adolesc. Health Care 44 (10), 277–318. Kalkbrenner, A.E., Windham, G.C., Serre, M.L., Akita, Y., Wang, X., Hoffman, K., Thayer, B.P., Daniels, J.L., 2014b. Particulate matter exposure, prenatal and postnatal windows of susceptibility, and autism spectrum disorders. Epidemiology 26 (1), 30–42. Nethery, E., Brauer, M., Janssen, P., 2009. Time-activity patterns of pregnant women and changes during the course of pregnancy. J. Expo. Sci. Environ. Epidemiol. 19, 317–324. Raz, R., Roberts, A.L., Lyall, K., Hart, J.E., Just, A.C., Laden, F., Weisskopf, M.G., 2014. Autism spectrum disorder and particulate matter air pollution before, during, and after pregnancy: a nested-case-control analysis within the Nurses’ Health Study II Cohort. EHP 123 (3), 264–270. http://dx.doi.org/10.1289/ehp.1408133. Ross, Z., Ito, K., Johnson, S., Yee, M., Pezeshki, G., Clougherty, J.E., Savitz, D., Matte, T., 2013. Spatial and temporal estimation of air pollutants in New York City: exposure assignment for use in a birth outcome study. Environ. Health 12, 51. Shmool, J.L., Michanowicz, D.R., Cambal, L., Tunno, B., Howell, J., Gillooly, S., Roper, C., Tripathy, S., Chubb, L.G., Eisl, H.M., 2014. Saturation sampling for spatial variation in multiple air pollutants across an inversion-prone metropolitan area of complex terrain. Environ. Health 13, 28. Volk, H.E., Hertz-Picciotto, I., Delwiche, L., Lurmann, F., McConnell, R., 2011. Residential proximity to freeways and autism in the CHARGE study. EHP 119 (6), 873–877. Volk, H.E., Lurmann, F., Penfold, B., Hertz-Picciotto, I., McConnell, R., 2013. Traffic related air pollution, particulate matter, and autism. JAMA Psychiatry 70 (1), 71–77.

Fine particulate matter and the risk of autism spectrum disorder.

The causes of autism spectrum disorder (ASD) are not well known. Recent investigations have suggested that air pollution, including PM2.5, may play a ...
489KB Sizes 0 Downloads 10 Views