Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2024 Feb 27.
Published in final edited form as: Environ Sci Technol. 2022 Dec 22;57(1):405–414. doi: 10.1021/acs.est.2c05197

The Associations of Autism Spectrum Disorder with PM2.5 Components: A Comparative Study Using Two Different Exposure Models

Md Mostafijur Rahman 1, Sarah A Carter 2, Jane C Lin 2, Ting Chow 2, Xin Yu 3, Mayra P Martinez 2, Zhanghua Chen 1, Jiu-Chiuan Chen 1, Daniel Rud 1, Juan P Lewinger 1, Aaron van Donkelaar 4, Randall V Martin 4, Sandrah Proctor Eckel 1, Joel Schwartz 5,6, Fred Lurmann 7, Michael J Kleeman 8, Rob McConnell 1,$, Anny H Xiang 2,*,$
PMCID: PMC10898516  NIHMSID: NIHMS1966493  PMID: 36548990

Abstract

This retrospective cohort study examined associations of autism spectrum disorder (ASD) with prenatal exposure to major fine particulate matter (PM2.5) components estimated using two independent exposure models. The cohort included 318,750 mother-child pairs with singleton deliveries in Kaiser Permanente Southern California hospitals from 2001-2014 and followed until age five. ASD cases during follow-up (N=4559) were identified by ICD codes. Prenatal exposures to PM2.5, elemental (EC) and black carbon (BC), organic matter (OM), nitrate (NO3−), and sulfate (SO42−) were constructed using (i) a source-oriented chemical transport model and (ii) a hybrid model. Exposures were assigned to each maternal address during the entire pregnancy, first, second, and third trimester. In single-pollutant models, ASD was associated with pregnancy-average PM2.5, EC/BC, OM, and SO42− exposures from both exposure models, after adjustment for covariates. The direction of effect estimates was consistent for EC/BC and OM, and least consistent for NO3−. EC/BC, OM and SO42− were generally robust to adjustment for other components and for PM2.5. EC/BC and OM effect estimates were generally larger and more consistent in the first and second trimester and SO42− in the third trimester. Future PM2.5 composition health effects studies might consider using multiple exposure models and a weight of evidence approach when interpreting effect estimates.

Keywords: PM2.5, PM2.5 chemical components, autism spectrum disorders, prenatal exposures, exposure models

Graphical Abstract

graphic file with name nihms-1966493-f0001.jpg

INTRODUCTION

Autism spectrum disorder (ASD) is a complex neurodevelopmental disorder characterized by social communication impairments, sensory disturbances and repetitive behaviors with restricted interests; approximately one-third have intellectual disability 1-4. ASD imposes substantial lifetime social and economic costs on affected families and communities 5. In the United States, lifetime cost of supporting an individual with ASD with intellectual disability is an estimated $2.4 million 6. Early etiologic studies on ASD focused on the role of genetic risk factors because there is high heritability. However, only 20% of diagnoses are due to spontaneous single gene or chromosomal mutations 7,8; the remaining causes are likely multifactorial.

A growing number of epidemiological studies have reported associations between prenatal exposure to particulate matter (PM) with aerodynamic diameter < 2.5 μm (PM2.5) and increased risk for ASD 9-11. In all prior studies on this relationship, PM2.5 has been considered as a homogenous pollutant, but in reality PM2.5 is comprised of a heterogeneous mixture of solid and liquid particles with varying chemical composition that reflects sources of particles and may determine toxicity 12-16. Elemental carbon/black carbon (EC/BC), organic matter (OM), nitrate (NO3−), and sulfate (SO42−) are the major components 17. Understanding the effects of different chemical components of PM2.5 on ASD risk could lead 1) to better understanding of mechanisms underlying PM effects on the brain and 2) to better prevention strategies, improved health impact assessments, and potentially to source-specific ambient air quality standards.

In recent years, exposure assessment methods have been developed to characterize the PM2.5 composition at fine spatial and temporal resolution. Several models for estimating PM2.5 composition have been developed and applied to studying mortality but results have not been consistent 14,18-22. This inconsistency might be explained in part by methodological differences in exposure assessment methods 23-27. In the current study, we used exposure component estimates from two independent exposure models: (i) a source-oriented chemical transport model (CTM) and (ii) a hybrid model that uses a chemical transport model, satellite observations, and ground-based measurements. The aim of this study was to assess the association of ASD with prenatal exposure to PM2.5 and its major components, including EC/BC, OM, NO3−, and SO42−, at residences in a large population-based pregnancy cohort. In addition, we assessed the consistency of associations with components estimated with each of the two modeling approaches.

MATERIALS AND METHODS

Study Population

This population-based retrospective pregnancy cohort study included mother-child pairs of singleton deliveries at Kaiser Permanente Southern California (KPSC) hospitals between January 1, 2001 and December 31, 2014. KPSC is a large integrated healthcare system with over 4.5 million members across Southern California. KPSC membership is diverse and broadly representative of the region’s sociodemographic characteristics 28. Maternal social and demographic characteristics, pregnancy health information, and maternal residential address history were extracted from KPSC’s well-established, integrated electronic medical records (EMR) system. Maternal addresses during pregnancy were geocoded using ArcGIS, and geocodes were assessed for exposure assignment suitability 29. Addresses based only on street name, 5-digit postal code, locality, or administrative unit were considered too uncertain to be geolocated into the correct grid used for exposure assignments.

Singleton births with KPSC membership at age 1 (n=370,723) were eligible to be included in this study. Children were routinely screened for potential ASD risk starting at age 18 months during regular well-child visits at KPSC. A total of 51,973 births was excluded due to 1) missing gender, maternal race/ethnicity and age at delivery, implausible age of delivery or birth weight (n=666); 20) maternal age at delivery (n=159); 3) incomplete maternal residential address history in pregnancy or geocodes not suitable for exposure assignment (n=51,148). The final data analysis included 318,750 mother-child pairs with complete data on residential estimates of PM2.5 composition exposures. Derivation of study sample size is shown in Figure S1 in the supplement.

Both KPSC and University of Southern California Institutional Review Boards approved this study with waiver of individual subject consent.

Outcome ASD

The outcome was ASD diagnosis before age 5. Children were followed from birth through the EMR until clinical diagnosis of ASD, loss to follow-up, or age 5, whichever came first. ASD diagnosis was identified by International Classification of Diseases (ICD) - 9 codes 299.0, 299.1, 299.8, 299.9 for EMR records before October 1, 2015 (date of KPSC implementation of ICD-10 codes) or ICD-10 codes F84.0, F84.3, F84.5, F84.8, F84.9 for EMR records after October 1, 2015. Codes from at least two separate visits were required to establish an ASD diagnosis, as described previously 30-33.

Exposures to PM2.5 and Components

Air pollution exposure estimation was conducted using two methods: i) Source-Oriented Chemical Transport Model (SO-CTM) developed by University of California Davis/California Institute of Technology (UCD/CIT); and ii) hybrid model that integrates CTM outputs, satellite observations, and ground-based measurements developed by Atmospheric Composition Analysis Group now at Washington University in St. Louis (WUACAG). We estimated ASD associations with PM2.5 and four major PM2.5 chemical components (EC and conceptually equivalent BC; OM, the mass of oxygen, hydrogen, and nitrogen together with organic carbon 34; NO3−; and SO42− that are available from both exposure models.

Monthly estimates of PM2.5, BC, OM, NO3−, and SO42− with a 1 km spatial resolution were obtained from the WUACAG hybrid model (version V4.NA.02) 35. This modelling framework integrates satellite observations of aerosol optical depth from multiple satellite products (MISR, MODIS Dark Target, MODIS and SeaWiFS, Deep Blue, and MODIS MAIAC) and PM2.5 simulated by GEOS-Chem (http://geos-chem.org) chemical transport model with 12.5 km resolution to estimate ground-level mass concentrations of PM2.5. Ground level observations of PM2.5 were then incorporated via geographically weighted regression to produce final PM2.5 surfaces for North America between 2000 and 2016 at 1 km × 1 km resolution. Later, GEOS-Chem chemical transport model simulation was used to partition this PM2.5 into seven PM2.5 chemical components (i.e., BC, NO3−, OM, ammonium [NH4+], SO42−, dust, and sea-salt). These PM2.5 components were then statistically fused into corresponding ground-level measurements, to produce a spatially complete representation over North America for the study period. The model performances of monthly estimates over the United States assessed by 10-fold cross validations were highest for SO42− (R2 = 0.90, bias= 0.03 μg/m3, root mean standard deviation =0.3 μg/m3), followed by NO3− (R2 = 0.78, bias= 0.01 μg/m3, RMSD= 0.3 μg/m3), BC (R2 = 0.68, bias= 0.01 μg/m3, RMSD= 0.1 μg/m3), OM (R2 = 0.55, bias= −0.01 μg/m3, RMSD= 0.6 μg/m3).

BC was available as EC from the SO-CTM model. Monthly estimates of PM2.5, EC, OM, NO3−, and SO42− with a 4 km spatial resolution were obtained from SO-CTM model for the time between 2000 and 2014. This model was developed for the California region only. Calculated meteorological fields and emissions estimates for different sources were used to predict airborne PM concentrations. Using the extensive emissions inventory in California, the model calculations track the mass and number concentrations of PM components in particle diameters ranging from 0.01 to 10 μm through calculations that describe emissions, transport, diffusion, deposition, coagulation, gas- and particle phase chemistry, and gas-to-particle conversion 36,37. Good correlations between predictions and measurements (r > 0.8) were demonstrated for many of the PM2.5 species at most of the monitoring stations, particularly for the monthly, seasonal, and annual averages. Monthly SO-CTM predicted PM2.5 EC, OM, NO3−, and SO42− was correlated with measurements with r = 0.96 (bias= −0.05 μg/m3, root mean squared error, RMSE= 0.17 μg/m3), 0.97 (bias= 0.11 μg/m3, RMSE= 0.46 μg/m3), 0.75 (bias= −1.24 μg/m3, RMSE= 2.16 μg/m3), and 0.67 (bias= −0.81 μg/m3, RMSE= 1.75 μg/m3), respectively, in the Los Angeles Basin.

Exposures to PM2.5 and these selected components were assigned to maternal address during the entire pregnancy, first trimester, second trimester, and third trimester. Monthly exposure estimates that did not correspond exactly to a trimester were assigned proportionally based on overlap of the trimesters. Exposures were also time-weighted to account for changes of maternal addresses during pregnancy.

Covariates

Covariates were selected a priori based on past literature on air pollution exposures and ASD 3,30,38, including child sex, maternal parity, maternal self-reported education and race/ethnicity, maternal history of comorbidity [>=1 diagnosis of heart, lung, kidney, or liver disease; cancer], maternal age at delivery, median family household income in census tract of residence, birth year, and an indicator variable for season (Dry= April-October; Wet= November-March)]. Birth year was included as a non-linear term with 4 degrees of freedom to adjust for the non-linear relationship between birth year and ASD. Maternal pre-pregnancy obesity (BMI ≥ 30 kg/m2) and diabetes during pregnancy were also included as covariates, as both were shown to be risk factors for ASD in our study cohort 31.

Statistical Analyses

The associations of ASD with PM2.5 and its major components were evaluated using Cox-proportional hazard models (HR) and 95% confidence intervals (CI). We first fitted single pollutant models. The HRs of the associations were scaled to the interquartile range (IQR) increase in concentration of PM2.5 and of each component of PM2.5 during the entire pregnancy, so that the population HRs for each pollutant were for conceptually similar pollutant increments. Children from families with more than one ASD child were included in the study sample. Standard errors were estimated using robust sandwich estimators to control for potential correlation for families. Timing of the exposure and associated windows of vulnerability are important issues for air pollution neuro-epidemiology, because they have potential to guide preventive interventions. We previously reported that increased prenatal PM2.5 exposure during the first two trimesters (up to 27 gestational weeks) of pregnancy was associated with subsequent risk of ASD in childhood 39. Susceptible windows of exposure to PM2.5 components may be different from those for PM2.5. Therefore, we fitted models with trimester specific average exposures of each component. To evaluate the independence of PM2.5 and component associations, we adjusted the single component models for the total PM2.5. So that estimates were comparable across trimesters and entire pregnancy, the trimester-specific estimates were also scaled to the entire pregnancy interquartile range (IQR) increase in concentration for PM2.5 and each component of PM2.5. We also subtracted each component’s mass separately from PM2.5 mass (denoted as ‘remainder PM2.5’) and included the ‘remainder PM2.5’ in the model. Because the results of adjustment for total PM2.5 and for remainder PM2.5 were very similar, for parsimony we have shown only the adjustment for remainder PM2.5, because unlike PM2.5, the remainder PM2.5 does not include the component. We assessed the consistency of the direction and magnitude of associations between ASD and specific PM2.5 components across both exposure modeling strategies. We also examined correlations of components in each exposure model; in exploratory analyses we ran multi-component models. All models were adjusted for the covariates described above. The proportional hazards assumption of the Cox proportional hazard model was assessed using the Schoenfeld residual plot. No clear non-random patterns against follow-up time were observed.

All statistical analyses were performed in R Statistical Software (v3.5.2; R Core Team 2021).

RESULTS

Participant demographics are shown in Table 1. Among the cohort, 4559 (1.4%) were diagnosed with ASD before age 5. Boys were over 4 times more likely to have ASD (n=3703) than girls (n=856). Children diagnosed with ASD were more likely to have older, nulliparous mothers with maternal comorbidities, pre-pregnancy diabetes, and pre-pregnancy obesity than children who were not diagnosed with ASD.

Table 1.

Characteristics of children, with and without autism spectrum disorder (ASD)

Children, No. (%) or median (interquartile range)
Characteristics Overall
(n =318 750)		 With ASD
(n= 4559)		 Without ASD
(n= 314 191)
Sex; N (%)
 Male 163 181 (51.2) 3703 (81.2) 159 428 (50.7)
 Female 155 569 (49.8) 856 (18.8) 154 763 (49.3)
Follow-up year after birth, median [IQR*], years 4.0 [4.0, 4.0.] 3.0 [2.3, 3.7] 4.0 [4.0, 4.0]
Maternal age at delivery, median [IQR*], years 30.4 [26.3, 34.3] 31.3 [27.5, 35.2] 30.4 [26.2, 34.3]
Parity; N (%)
 0 111 981 (35.1) 1844 (40.4) 110 137 (35.1)
 1 104 561 (32.8) 1495 (32.8) 103 066 (32.8)
 >2 84 176 (26.4) 903 (19.8) 83 273 (26.5)
 Unknown 18 032 (5.7) 317 (7.0) 17 715 (5.6)
Maternal Education; N (%)
 High school or lower 112 096 (35.2) 1335 (29.3) 110 761 (35.3)
 Some college 94 524 (29.7) 1477 (32.4) 93 047 (29.6)
 College graduate or higher 109 087 (34.2) 1713 (37.6) 107 374 (34.2)
 Unknown 3043 (1.0) 43 (0.7) 3009 (1.0)
Household annual incomea; N (%)
 <$30,000 24 027 (7.5) 325 (7.1) 23 710 (7.5)
 $30,000-$49,999 100 575 (31.6) 1436 (31.5) 99 139 (31.6)
 $50,000-$69,999 98 015 (30.7) 1415 (31.0) 96 593 (30.7)
 $70,000-$89,999 55 611 (17.4) 801 (17.5) 54 816 (17.4)
 > $90,000 40 512 (12.7) 582 (12.8) 39 933 (12.7)
Race/ethnicity; N (%)
 Non-Hispanic white 81 050 (25.4) 956 (21.0) 80 094 (25.5)
 Non-Hispanic black 29 773 (9.3) 477 (9.8) 29 326 (9.3)
 Hispanic 161 414 (50.6) 2300 (50.4) 159 114 (50.6)
 Asian/Pacific Islander 39 974 (12.5) 744 (16.3) 39 230 (12.5)
 Other 6539 (2.1) 112 (2.5) 6427 (2.0)
Any history of maternal comorbidityb; N (%) 46 717 (14.6) 839 (18.4) 45 878 (14.6)
Pre-pregnancy diabetesc; N (%) 10 248 (3.2) 242 (5.3) 10 006 (3.2)
Pre-pregnancy obesityd; N (%) 53 354 (16.7) 1049 (23.0) 52 305 (16.6)
Year of birth, N (%)
 2001-2007 152 750 (47.9) 1802 (39.5) 164 198 (52.2)
 2008-2014 166 000 (52.1) 2757 (60.5) 149 993 (47.2)
Open in a new tab
*

Abbreviations: IQR, interquartile range.

a

Census tract level median household income.

b

>=1 diagnosis of heart, lung, kidney, or liver disease; cancer.

c

Type I and Type II diabetes diagnosed before pregnancy.

d

Pre-pregnancy BMI>=30

The relative contribution of components to the total PM2.5 mass during pregnancy is shown in Figure 2 in the supplement. These contributions varied between the two models. For the SO-CTM model, the four major components (excluding “other”) accounted for 60% of PM2.5 mass. For the hybrid model, these components accounted for 83.5%. OM accounted for 41.3% of the hybrid model predicted PM2.5 but only for 17.8% of the SO-CTM PM2.5. BC accounted for 12.2% of hybrid predicted PM2.5; EC for 4.6% of SO-CTM PM2.5. In contrast, the contributions of NO3− and SO42− to the hybrid PM2.5 were a little smaller than the proportion contributed to the SO-CTM PM2.5.

The estimated mean (14.2) and IQR (5.6) μg/m3 of the predicted SO-CTM PM2.5 differed from the hybrid model (15.2; 3.7 μg/m3). (See Figure 1). The greatest discrepancy between the models was for carbon, for which the EC mean (0.7 μg/m3) and IQR (0.4 μg/m3) were about half as large as for BC (mean 1.9 (0.8) μg/m3). EC and BC represent similar components; there is no universally agreed conversion, but the two-fold difference probably represents both differences between mass of EC and BC, and differences between modeling approaches. SO-CTM OM mean 2.5 (IQR 1.4) μg/m3 also differed markedly from hybrid OM, 6.3 (2.0) μg/m3; SO-CTM NO3− mean 3.6 (IQR 2.1) μg/m3 differed a little from hybrid NO3−, 3.1 (1.2) μg/m3; and SO-CTM SO42− mean 1.7 (IQR 0.5) μg/m3 differed minimally from hybrid SO42− (1.5 (0.5) μg/m3. The exposure distribution during each trimester window for PM2.5 and its major components was similar to the pregnancy average distribution (Supplementary Figures 3A-3C).

Figure 1.

Figure 1.

Open in a new tab

The distribution of PM2.5 and its major components during entire-pregnancy based on A) source-oriented chemical transport model and B) hybrid model, during the study period from 2001 – 2014.

In each exposure modeling approach, pregnancy-average PM2.5 was highly correlated with each of its components, highest with NO3− (R=0.87) and lowest with SO42− (0.69; Table 2). Between the components, correlations were low to moderate within each exposure model, with the exception of SO-CTM EC and OM (R=0.85). Between exposure models, the correlation for PM2.5 was 0.80. Components from the two models were moderately correlated (EC with BC 0.61, OM 0.63, SO42− 0.46), with the exception of NO3− (0.78). The patterns of correlations for PM2.5 and its major components in each trimester were similar to those observed during the entire pregnancy (Supplementary Tables 1-3).

Table 2.

Pearson correlation matrix of entire-pregnancy estimates of PM2.5, EC/BC, OM, NO3−, and SO42− from a source-oriented chemical transport model and a hybrid model.

UCD/CIT SO-CTM Model WUACAG Hybrid Model
PM2.5 EC OM NO3− SO42− PM2.5 BC OM NO3− SO42−
UCD/CIT SO-CTM Model PM2.5 1.00
EC 0.75 1.00 graphic file with name nihms-1966493-t0003.jpg
OM 0.74 0.85 1.00
NO3− 0.87 0.55 0.44 1.00
SO42− 0.69 0.40 0.31 0.47 1.00
 
WUACAG Hybrid Model PM2.5 0.80 0.68 0.67 0.67 0.46 1.00
BC 0.56 0.61 0.56 0.43 0.26 0.78 1.00
OM 0.57 0.59 0.63 0.35 0.32 0.81 0.48 1.00
NO3− 0.82 0.65 0.59 0.78 0.46 0.91 0.61 0.62 1.00
SO42− 0.52 0.34 0.27 0.43 0.46 0.59 0.17 0.47 0.59 1.00
Open in a new tab

The pregnancy-average ASD HR point estimates scaled per IQR increase corresponding to each component are shown in Table 3. Adverse associations with OM and SO42− were observed using either exposure model and effect estimates across models were generally similar after adjustment for the remainder PM2.5. The EC HR 1.12 (95% CI 1.07, 1.18) using the SO-CTM model was larger than the equivalent BC (1.06; 95% CI 1.02, 1.10), using the hybrid exposure model, but CI overlapped. EC effect estimates were robust to co-adjustment, BC was attenuated by adjustment for remainder PM2.5. NO3− was inversely associated with ASD in models adjusted for remainder PM2.5 and this was statistically significant for the SO-CTM model.

Table 3.

Associations of ASD with entire-pregnancy average exposures to EC/BC, OM, NO3−, and SO42− from a source-oriented chemical transport model and a hybrid model.

UCD/CIT SO-CTM Model WUACAG Hybrid Model
Primary
Exposure
of Interest		 Adjusted
Pollutant (s)		 HR (95% CI) Primary
Exposure
of Interest		 Adjusted
Pollutant (s)		 HR (95% CI)
EC Single pollutant 1.12 (1.07, 1.18) BC Single pollutant 1.06 (1.02, 1.10)
Remainder PM2.5 1.15 (1.09, 1.23) Remainder PM2.5 1.03 (0.98, 1.09)
OM+NO3−+SO42− 1.13 (1.06, 1.21) OM+NO3−+SO42− 1.04 (0.98, 1.09)
OM Single pollutant 1.09 (1.04, 1.15) OM Single pollutant 1.08 (1.03, 1.13)
Remainder PM2.5 1.09 (1.03, 1.16) Remainder PM2.5 1.08 (1.02, 1.15)
EC+NO3−+SO42− 1.02 (0.95, 1.10) BC+NO3−+SO42− 1.05 (0.99, 1.13)
NO3− Single pollutant 0.85 (0.79, 0.93) NO3− Single pollutant 1.05 (1.00, 1.09)
Remainder PM2.5 0.76 (0.70, 0.84) Remainder PM2.5 0.98 (0.92, 1.05)
EC+OM+SO42− 0.79 (0.72, 0.86) BC+OM+SO42− 0.97 (0.91, 1.03)
SO42− Single pollutant 1.05 (1.00, 1.11) SO42− Single pollutant 1.08 (1.02, 1.14)
Remainder PM2.5 1.04 (0.99, 1.10) Remainder PM2.5 1.05 (0.99, 1.12)
EC+OM+NO3− 1.05 (0.99, 1.10) EC+OC+NO3− 1.06 (0.99, 1.13)
Open in a new tab
*

All the models were adjusted for child sex, maternal race/ethnicity, maternal age at delivery, parity, education, maternal comorbidities, household income (census tract level), birth year (non-linear), season (wet/dry), and pre-pregnancy diabetes. The hazard ratios were scaled to the inter-quartile (IQR) increase in concentration of each air pollutant during pregnancy. Based on UCD/CIT SO-CTM, the IQRs (μg/m3) for PM2.5, EC, OM, NO3−, and SO42− during pregnancy were 5.56, 0.38, 1.38, 2.07, and 0.52, respectively. Based on WUACAG Hybrid Model, the IQRs (μg/m3) for PM2.5, BC, OM, NO3−, and SO42− during pregnancy were 3.73, 0.84, 1.98, 1.15, and 0.50, respectively.

In exploratory models, associations of ASD with each PM component adjusted for all other components simultaneously generally were similar to associations adjusted for remainder PM2.5, with some exceptions (Table 3). To understand the marked attenuation of the OM SO-CTM multicomponent-adjusted exposure estimate, compared with the model adjusted for remainder PM2.5, we examined which of the three co-pollutant components was responsible for the attenuation of the OM effect, by running 2-component models with OM. Adjustment for NO3− or SO42− did not change the OM effect estimate; adjustment for EC reduced the OM HR to 0.99 (95% CI 0.92. 1.07), suggesting that it may be an artifact of the high correlation of EC with OM (0.85 from Table 2).

Associations of ASD with EC/BC and OM across exposure modeling approaches in single pollutant models were similar to the pregnancy-average associations in the first and second trimester and were generally robust to adjustment for remainder PM2.5 and other components (Supplement Tables 4-6). As during the entire pregnancy, OM was highly correlated with EC and adjustment of the OM effect estimate for EC and other components resulted in marked attenuation of the HR in the first (to HR 1.02; 95% CI 0.96, 1.08) and second trimester (to HR 1.00; 95% CI 0.94, 1.07). Associations with SO42− were stronger in the third trimester, moreso for the hybrid model (HR 1.10; 95% CI 1.04, 1.16 in single pollutant model robust to co-pollutant adjustment) than for the SO-CTM exposure effect estimate.

In single-pollutant models, increased exposures to PM2.5 during the entire pregnancy were associated with increased ASD risk (Supplementary Table 7). In two-pollutant models the PM2.5 effect estimates were markedly attenuated by adjustment for OM using either exposure model; there was an inverse association of PM2.5 with ASD in models adjusted for EC (HR 0.92; 95% CI 0.83, 1.02) but not for BC (HR 1.05; 95% CI 0.98, 1.12). Effect estimates were similar or larger than in single pollutant models after adjusting for other components. ASD associations with single-pollutant PM2.5 were also consistently positive using both exposure modeling approaches in each trimester, and associations were statistically significant except for the third trimester SO-CTM model. As for the pregnancy average exposures, PM2.5 effect estimates were attenuated by adjustment for components, especially by EC or OM and in the SO-CTM models.

DISCUSSION

Average exposure to PM2.5 component EC/BC, OM and SO42− during pregnancy in a large population-based cohort was associated with small increases in ASD risk using both the SO-CTM and hybrid exposure models, although estimates from some health models were attenuated by adjustment for remainder PM2.5. Associations of prenatal PM2.5 itself with ASD were consistently positive using both exposure modeling approaches in single pollutant models, but were markedly attenuated in models adjusted for EC and OM. Associations using different modeling approaches were generally similar, at least in direction. A notable exception was NO3−, which had very different ASD associations depending on the modeling approach. Trimester-specific analysis revealed that EC/BC and OM exposure in the first and second trimester had significant associations with ASD; SO42− showed stronger associations with ASD in the third trimester.

Our results suggest that associations of PM2.5 with ASD previously observed in this cohort may be explained by components rather than by total PM2.5 mass 30,39. PM2.5 was highly correlated across exposure modeling approaches and showed the most consistent positive associations in single pollutant models. However, PM2.5 associations were markedly attenuated by adjustment for components, in particular by EC and OM in the SO-CTM model and by BC or OM in the pregnancy average and first and second trimester models. These components were only moderately correlated with PM2.5 (from Table 2) and the large data set provides more confidence that PM2.5 effects were confounded by these components and that attenuation of PM2.5 effect estimates was not due to high correlation.

There were differences between the SO-CTM and hybrid modeling approaches that could have affected the health associations in both magnitude and direction. For example, the SO-CTM model was developed for California only, whereas the hybrid model was developed for all North America. The state-specific model might be considered superior for this application, because the information available in California may better capture the intra-regional variability of the PM2.5 components. However, the SO-CTM model provide estimates at 4km spatial resolution. Primary EC and BC, in particular, have steep spatial gradients from major roadway sources, so exposure misclassification might be less in the hybrid model with 1km resolution. In contrast, SO42−, which occurs as a result of secondary formation from precursor sulfur dioxide, has comparatively smoother spatial variation than other PM2.5 components, so the spatial resolution should have little impact on the accuracy of the exposure assignment from each model. OM includes both primary and secondary particles, including emissions from vehicular and industrial fuel combustion and natural sources, and secondary organic carbon produced by photochemical reactions of gaseous precursors in the atmosphere. A limitation of the current study is that only primary OM from the SO-CTM was used in the exposure analysis because predictions of secondary OM were judged to be uncertain. This likely explains part of the markedly larger estimated exposure from the hybrid model than from the SO-CTM model. Thus, each exposure model has qualitative strengths and weaknesses, which are difficult to assess quantitatively.

In spite of the differences between exposure models, we found generally consistent results using both models, at least in the direction of the association of components with ASD. An exception was NO3− (which was positively associated in the hybrid model and inversely associated in the SO-CTM model). NO3− prediction is particularly challenging because it is largely secondary aerosol, originating from the atmospheric oxidation of NOx (nitrogen dioxide, NO2 and nitrous oxide), and in addition is semi-volatile; it is possible that the two measurement sites in southern California are insufficient to support accurate NO3− predictions in the hybrid modeling system, and/or some systematic error in the SO-CTM predicted NO3− may have led to bias. Mechanisms for NO3− associations with autism (either protective or increased risk) have not been published in peer-reviewed literature. Further investigation is needed to assess the reasons for this divergence in exposure estimates between the two models.

The general consistency of associations with other components estimated using different exposure methods increases the level of confidence in the observed ASD associations. Using different modeling approaches also helps prevent over- or under-interpretation of the importance of associations based on a single exposure models. For example, the pregnancy average HR for EC (1.12; CI 1.07, 1.18) from Table 3, which was not attenuated by adjustment for other components and remainder PM2.5, might be interpreted as robust evidence for a causal effect; however, if only the BC hybrid model were used, the conclusion might be that effects were confounded by the remainder PM2.5 (with a reduction of HR from 1.06 (1.02, 1.10) in the single pollutant BC model HR to 1.03 (0.98, 1.09) after adjusting for remainder PM2.5. A few previous studies of other outcomes reported some heterogeneity in component effect estimates in terms of statistical significance, magnitude, and direction of association, depending on the exposure assessment method used 23-27. Modeling components at local spatial scale for epidemiological studies is relatively new compared to well validated models of PM2.5. Future PM2.5 composition health effects studies might consider using multiple exposure models and a weight of evidence approach when interpreting effect estimates of associations.

There have been few studies on the developmental neurotoxicity of PM2.5 components 40,41, compared with a large emerging body of work examining PM2.5 mass 9-11, and none to our knowledge has examined the association with ASD risk in children. There has been limited prior epidemiological study of neurotoxicity of EC/BC exposure, and results were not consistent 42-46. Exposure to prenatal EC/BC or PM2.5 absorbance (a proxy for EC) was associated with increased hyperactivity/inattention among adolescents 44 and worse memory in urban children 42, no association with childhood cognitive and psychomotor development 43, and with verbal IQ in minimally adjusted, but not in fully adjusted models 45. Rodent studies of EC neurotoxicity (largely from diesel exhaust particles) have also been inconclusive 47,48,49,50. Mechanisms for effects are not clear, although several studies reported high oxidative potential of EC particles 51-53. EC/BC effects may also be a proxy for other co-emissions, such as semi-volatile organic compounds and polycyclic aromatic hydrocarbon, that are adsorbed onto the EC core 54-58. To the best of our knowledge, neurodevelopmental effects of prenatal OM exposure has not been examined either in epidemiological or animal studies. The pregnancy average and third trimester association with SO42− is also novel. There is little SO42− in Southern California; emissions from ships in the Long Beach/Los Angeles port complex are a major source.

Our study has several strengths. First, it leveraged a large, diverse pregnancy cohort with standard diagnostic criteria for ASD. The KPSC cohort is representative of the Southern California population 59, and thus results are relevant to similar populations across the United States; the regulated air pollutants in the region (with the exception of SO42−) encompass most of the entire range and mixture across the U.S. Exposure was assigned from validated prediction models, accounting for change of address during pregnancy. ASD associations with PM components were adjusted for key confounders such as maternal pre-pregnancy health status, season of conception, and year of birth, obtained from the high quality KPSC EMR. Finally, the multi-pollutant modelling strategies accounted for effects of the remainder PM2.5. We acknowledge some limitations. Mother’s time-activity patterns were not available for this analysis. Knowledge of the accuracy and uncertainty of the exposure model estimates of PM chemical components is limited by the much smaller observational database than is available for PM2.5 mass.

In summary, prenatal exposure to some PM2.5 component EC/BC, OM, and to a lesser extent SO42−, were associated with increased ASD risk. However, the strength and statistical significance of some effect estimates differed between exposure models. The results of this study are consistent with the emerging literature indicating that different exposure assessment models may be responsible for some of the heterogeneity in effect estimates across studies using different PM composition exposure models. Unlike PM2.5, which has been studied in epidemiological studies for decades, models for most PM2.5 components for use in epidemiological studies are only recently available. Better understanding of the role of PM2.5 components and their sources could lead to targeted regulatory interventions to reduce the health effects of particulate air pollution. Eventually, component-specific air pollution control policies merit consideration in regulatory strategies for reducing adverse effects of PM2.5 60.

Supplementary Material

Supplementary document

SYNOPSIS.

This study provided novel evidence of ASD risk resulting from prenatal exposure to fine particle components.

ACKNOWLEDGEMENTS

The authors thank the patients of Kaiser Permanente for helping us improve care through the use of information collected through our electronic health record systems, and the Kaiser Permanente and the Utility for Care Data Analysis (UCDA) team within Kaiser Permanente for creating the GEMS Datamart with consolidated addresses histories available to facilitate our research.

FUNDING

This research was supported by National Institutes of Environmental Health Sciences (R01 ES029963 (Xiang, McConnell); R56ES028121 (Xiang); P30ES007048 (McConnell); Simms/Mann Chair in Neurogenetics (PL); partly supported by Kaiser Permanente Southern California Direct Community Benefit Funds. Joel Schwartz was supported by EPA grant RD-8358720. Randall Martin acknowledges funding from NASA HAQAST (80NSSC21K0508). The funding agencies had no role in the design or conduct of the study; in the analysis or interpretation of the data; or in the preparation, review, or approval of the manuscript.

Footnotes

CONFLICT OF INTERESTS

The authors declare they have no actual or potential competing interests. Joel Schwartz declares that he has testified on behalf of the U.S. Department of Justice in a case involving a Clean Air Act violation. Frederick Lurmann is employed by Sonoma Technology, Inc., Petaluma, CA

Supporting Information

The supporting information is available online.

Correlation between PM2.5, EC/BC, OM, NO3−, and SO42− in each trimester, based on a SO-CTM and a hybrid model (Tables S1-S3); Associations of ASD with EC/BC, OM, NO3−, and SO42− from a SO-CTM and a hybrid model (Tables S4-S5); Associations of ASD with PM2.5 from a SO-CTM and a hybrid model during entire pregnancy and each trimester (Table S7); Derivation of sample (Figure S1); The relative contribution of each component to the total PM2.5 mass during entire pregnancy (Figure S2); The distribution of PM2.5, EC/BC, OM, NO3−, and SO42− in each trimester, based on a SO-CTM and a hybrid model (Figure S3).

ETHICAL APPROVAL

Both KPSC and University of Southern California Institutional Review Boards approved this study.

REFERENCES

  • (1).Zhang X; Lv CC; Tian J; Miao RJ; Xi W; Hertz-Picciotto I; Qi L Prenatal and Perinatal Risk Factors for Autism in China. J. Autism Dev. Disord 2010, 40 (11), 1311–1321. 10.1007/s10803-010-0992-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (2).Elsabbagh M. Linking Risk Factors and Outcomes in Autism Spectrum Disorder: Is There Evidence for Resilience? BMJ 2020, 368. 10.1136/bmj.l6880. [DOI] [PubMed] [Google Scholar]
  • (3).Pagalan L; Bickford C; Weikum W; Lanphear B; Brauer M; Lanphear N; Hanley GE; Oberlander TF; Winters M Association of Prenatal Exposure to Air Pollution with Autism Spectrum Disorder. JAMA Pediatr. 2019, 173 (1), 86–92. 10.1001/jamapediatrics.2018.3101. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (4).Xu G; Strathearn L; Liu B; O’Brien M; Kopelman TG; Zhu J; Snetselaar LG; Bao W Prevalence and Treatment Patterns of Autism Spectrum Disorder in the United States, 2016. JAMA Pediatr. 2019, 173 (2), 153–159. 10.1001/jamapediatrics.2018.4208. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (5).Werling DM; Geschwind DH Sex Differences in Autism Spectrum Disorders. Curr. Opin. Neurol 2013, 26 (2), 146–153. 10.1097/WCO.0b013e32835ee548. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (6).Buescher AVS; Cidav Z; Knapp M; Mandell DS Costs of Autism Spectrum Disorders in the United Kingdom and the United States. JAMA Pediatr. 2014, 168 (8), 721–728. 10.1001/jamapediatrics.2014.210. [DOI] [PubMed] [Google Scholar]
  • (7).Bai D; Yip BHK; Windham GC; Sourander A; Francis R; Yoffe R; Glasson E; Mahjani B; Suominen A; Leonard H; Gissler M; Buxbaum JD; Wong K; Schendel D; Kodesh A; Breshnahan M; Levine SZ; Parner ET; Hansen SN; Hultman C; Reichenberg A; Sandin S Association of Genetic and Environmental Factors With Autism in a 5-Country Cohort. JAMA Psychiatry 2019, 76 (10), 1035–1043. 10.1001/jamapsychiatry.2019.1411. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (8).Rylaarsdam L; Guemez-Gamboa A Genetic Causes and Modifiers of Autism Spectrum Disorder. Front. Cell. Neurosci 2019, 13 (August), 1–15. 10.3389/fncel.2019.00385. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (9).Chun HK; Leung C; Wen SW; McDonald J; Shin HH Maternal Exposure to Air Pollution and Risk of Autism in Children: A Systematic Review and Meta-Analysis. Environ. Pollut 2020, 256, 113307. 10.1016/j.envpol.2019.113307. [DOI] [PubMed] [Google Scholar]
  • (10).Lam J; Sutton P; Kalkbrenner A; Windham G; Halladay A; Koustas E; Lawler C; Davidson L; Daniels N; Newschaffer C; Woodruff T A Systematic Review and Meta-Analysis of Multiple Airborne Pollutants and Autism Spectrum Disorder. PLoS One 2016, 11 (9), 1–27. 10.1371/journal.pone.0161851. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (11).Lin L-Z; Zhan X-L; Jin C-Y; Liang J-H; Jing J; Dong G-H The Epidemiological Evidence Linking Exposure to Ambient Particulate Matter with Neurodevelopmental Disorders: A Systematic Review and Meta-Analysis. Environ. Res 2022, 209, 112876. https://doi.org/ 10.1016/j.envres.2022.112876. [DOI] [PubMed] [Google Scholar]
  • (12).National Research Council. Research Priorities for Airborne Particulate Matter: IV. Continuing Research Progress; The National Academies Press: Washington, DC, 2004. 10.17226/10957. [DOI] [Google Scholar]
  • (13).Fong KC; Di Q; Kloog I; Laden F; Coull BA; Koutrakis P; Schwartz JD Relative Toxicities of Major Particulate Matter Constituents on Birthweight in Massachusetts. Environ. Epidemiol 2019, 3 (3), e047. 10.1097/ee9.0000000000000047. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (14).Thurston GD; Burnett RT; Turner MC; Shi Y; Krewski D; Lall R; Ito K; Jerrett M; Gapstur SM; Diver WR; Iii CAP Ischemic Heart Disease Mortality and Long-Term Exposure to Source-Related Components of U.S. Fine Particle Air Pollution. Environ. Health Perspect 2016, 124 (6), 785–795. 10.1289/ehp.1509777. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (15).Adams K; Greenbaum DS; Shaikh R; van Erp AM; Russell AG Particulate Matter Components, Sources, and Health: Systematic Approaches to Testing Effects. J. Air Waste Manag. Assoc 2015, 65 (5), 544–558. 10.1080/10962247.2014.1001884. [DOI] [PubMed] [Google Scholar]
  • (16).Rahman MM; Begum BA; Hopke PK; Nahar K; Newman J; Thurston GD Cardiovascular Morbidity and Mortality Associations with Biomass- and Fossil-Fuel-Combustion Fine-Particulate-Matter Exposures in Dhaka, Bangladesh. Int. J. Epidemiol 2021, 1–12. 10.1093/ije/dyab037. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (17).Dominici F; Wang Y; Correia AW; Ezzati M; Pope CA 3rd; Dockery DW Chemical Composition of Fine Particulate Matter and Life Expectancy: In 95 US Counties Between 2002 and 2007. Epidemiology 2015, 26 (4), 556–564. 10.1097/EDE.0000000000000297. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (18).Ostro B; Hu J; Goldberg D; Reynolds P; Hertz A; Bernstein L; Kleeman MJ Associations of Mortality with Long-Term Exposures to Fine and Ultrafine Particles, Species and Sources: Results from the California Teachers Study Cohort. Environ. Health Perspect 2015, 123 (6), 549–556. 10.1289/ehp.1408565. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (19).Rob B; Gerard H; Ole R-N; Massimo S; Jovanovic AZ; Gudrun W; Barbara H; Kathrin W; Evangelia S; H. FP; J. NM; W. XW; Klea K; Konstantina D; Alessandro M; Erkki V; Timo L; Tarja Y-T; Bente O; E. SP; Per N; Ulf DF; L. PN; Claes-Göran Ö; Laura F; Johanna P; Michal K; Göran P; Thorup EK; Kim O; Mette S; Marloes E; H. PP; Kees M; Meng W; Bas B-MH; Dorothea S; Ursula K; Joachim H; Kees de H.; Timothy K; Annette P; Regina H; Hans C; Gabriele N; Andrea J; Alex I; Ming-Yi T; Emmanuel S; M. P-HN; Christian S; S. RM; Alice V; Françoise C-C; Christophe D; Fulvio R; Carlotta S; Claudia G; Enrica M; Andrea R; Giulia C; Chiara B; Francesco F; Michail K; Antonia T; Menno K; Aleksandra J; M. KI; Jaakko K; S. SR; Paolo V; Bert B Natural-Cause Mortality and Long-Term Exposure to Particle Components: An Analysis of 19 European Cohorts within the Multi-Center ESCAPE Project. Environ. Health Perspect 2015, 123 (6), 525–533. 10.1289/ehp.1408095. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (20).Wang Y; Xiao S; Zhang Y; Chang H; Martin RV; Van Donkelaar A; Gaskins A; Liu Y; Liu P; Shi L Long-Term Exposure to PM2.5 Major Components and Mortality in the Southeastern United States. Environ. Int 2022, 158, 106969. https://doi.org/ 10.1016/j.envint.2021.106969. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (21).Wang Y; Shi L; Lee M; Liu P; Di Q; Zanobetti A; Schwartz JD Long-Term Exposure to PM2.5 and Mortality Among Older Adults in the Southeastern US. Epidemiology 2017, 28 (2), 207–214. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (22).Yeonseung C; Francesca D; Yun W; A. CB; L BM Associations between Long-Term Exposure to Chemical Constituents of Fine Particulate Matter (PM2.5) and Mortality in Medicare Enrollees in the Eastern United States. Environ. Health Perspect 2015, 123 (5), 467–474. 10.1289/ehp.1307549. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (23).Gariazzo C; Carlino G; Silibello C; Tinarelli G; Renzi M; Finardi S; Pepe N; Barbero D; Radice P; Marinaccio A; Forastiere F; Michelozzi P; Viegi G; Stafoggia M; Carla A; Paola A; Stefania A; Sandra B; Lucia B; Michela B; Sergio B; Laura B; Serena B; Giuseppe B; Simone B; Giuseppe C; Giuseppe C; Achille C; Antonio C; Annamaria C; de’ Donato Francesca; Salvatore F; Sandro F; Francesco F; Claudia G; Claudio G; Paolo GR; Stefania LG; Gaetano L; Sara M; Alessandro M; Paola M; Enrica M; Antonino M; Alessandro N; Marta O; Federica P; Nicola P; Paola R; Andrea R; Matteo R; Salvatore S; Matteo S; Camillo S; Roberto S; Massimo S; Gianni T; Francesco U; Giovanni V Impact of Different Exposure Models and Spatial Resolution on the Long-Term Effects of Air Pollution. Environ. Res 2021, 192. 10.1016/j.envres.2020.110351. [DOI] [PubMed] [Google Scholar]
  • (24).Bauwelinck M; Chen J; de Hoogh K; Katsouyanni K; Rodopoulou S; Samoli E; Andersen ZJ; Atkinson R; Casas L; Deboosere P; Demoury C; Janssen N; Klompmaker JO; Lefebvre W; Mehta AJ; Nawrot TS; Oftedal B; Renzi M; Stafoggia M; Strak M; Vandenheede H; Vanpoucke C; Van Nieuwenhuyse A; Vienneau D; Brunekreef B; Hoek G Variability in the Association between Long-Term Exposure to Ambient Air Pollution and Mortality by Exposure Assessment Method and Covariate Adjustment: A Census-Based Country-Wide Cohort Study. Sci. Total Environ 2022, 804, 150091. 10.1016/j.scitotenv.2021.150091. [DOI] [PubMed] [Google Scholar]
  • (25).Klompmaker JO; Janssen N; Andersen ZJ; Atkinson R; Bauwelinck M; Chen J; de Hoogh K; Houthuijs D; Katsouyanni K; Marra M; Oftedal B; Rodopoulou S; Samoli E; Stafoggia M; Strak M; Swart W; Wesseling J; Vienneau D; Brunekreef B; Hoek G Comparison of Associations between Mortality and Air Pollution Exposure Estimated with a Hybrid, a Land-Use Regression and a Dispersion Model. Environ. Int 2021, 146, 106306. 10.1016/j.envint.2020.106306. [DOI] [PubMed] [Google Scholar]
  • (26).Chen J; Rodopoulou S; de Hoogh K; Strak M; Andersen ZJ; Atkinson R; Bauwelinck M; Bellander T; Brandt J; Cesaroni G; Concin H; Fecht D; Forastiere F; Gulliver J; Hertel O; Hoffmann B; Hvidtfeldt UA; Janssen NAH; Jöckel KH; Jørgensen J; Katsouyanni K; Ketzel M; Klompmaker JO; Lager A; Leander K; Liu S; Ljungman P; Macdonald CJ; Magnusson PKE; Mehta A; Nagel G; Oftedal B; Pershagen G; Peters A; Raaschou-Nielsen O; Renzi M; Rizzuto D; Samoli E; van der Schouw YT; Schramm S; Schwarze P; Sigsgaard T; Sørensen M; Stafoggia M; Tjønneland A; Vienneau D; Weinmayr G; Wolf K; Brunekreef B; Hoek G Long-Term Exposure to Fine Particle Elemental Components and Natural and Cause-Specific Mortality—a Pooled Analysis of Eight European Cohorts within the ELAPSE Project. Environ. Health Perspect 2021, 129 (4), 1–12. 10.1289/EHP8368. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (27).Jerrett M; Turner MC; Beckerman BS; Pope CA; van Donkelaar A; Martin RV; Serre M; Crouse D; Gapstur SM; Krewski D; Diver WR; Coogan PF; Thurston GD; Burnett RT Comparing the Health Effects of Ambient Particulate Matter Estimated Using Ground-Based versus Remote Sensing Exposure Estimates. Environ. Health Perspect 2017, 125 (4), 552–559. 10.1289/EHP575. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (28).Koebnick C; Langer-Gould AM; Gould MK; Chao CR; Iyer RL; Smith N; Chen W; Jacobsen SJ Sociodemographic Characteristics of Members of a Large, Integrated Health Care System: Comparison with US Census Bureau Data. Perm. J 2012, 16 (3), 37–41. 10.7812/tpp/12-031. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (29).ArcGIS. The geocoding process. 2020. https://desktop.arcgis.com/en/arcmap/latest/manage-data/geocoding/the-geocoding-process.htm (accessed Nov 13, 2021).
  • (30).Jo H; Eckel SP; Wang X; Chen JC; Cockburn M; Martinez MP; Chow T; Molshatzki N; Lurmann FW; Funk WE; Xiang AH; McConnell R Sex-Specific Associations of Autism Spectrum Disorder with Residential Air Pollution Exposure in a Large Southern California Pregnancy Cohort. Environ. Pollut 2019, 254, 113010. 10.1016/j.envpol.2019.113010. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (31).Xiang AH; Wang X; Martinez MP; Walthall JC; Curry ES; Page K; Buchanan TA; Coleman KJ; Getahun D Association of Maternal Diabetes with Autism in Offspring. JAMA - J. Am. Med. Assoc 2015, 313 (14), 1425–1434. 10.1001/jama.2015.2707. [DOI] [PubMed] [Google Scholar]
  • (32).Xiang AH; Wang X; Martinez MP; Page K; Buchanan TA; Feldman RK Maternal Type 1 Diabetes and Risk of Autism in Offspring. JAMA 2018, 320 (1), 89–91. 10.1001/jama.2018.7614. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (33).Coleman KJ; Lutsky MA; Yau V; Qian Y; Pomichowski ME; Crawford PM; Lynch FL; Madden JM; Owen-Smith A; Pearson JA; Pearson KA; Rusinak D; Quinn VP; Croen LA Validation of Autism Spectrum Disorder Diagnoses in Large Healthcare Systems with Electronic Medical Records. J. Autism Dev. Disord 2015, 45 (7), 1989–1996. 10.1007/s10803-015-2358-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (34).Philip S; Martin RV; Pierce JR; Jimenez JL; Zhang Q; Canagaratna MR; Spracklen DV; Nowlan CR; Lamsal LN; Cooper MJ; Krotkov NA Spatially and Seasonally Resolved Estimate of the Ratio of Organic Mass to Organic Carbon. Atmos. Environ 2014, 87, 34–40. https://doi.org/ 10.1016/j.atmosenv.2013.11.065. [DOI] [Google Scholar]
  • (35).Van Donkelaar A; Martin RV; Li C; Burnett RT Regional Estimates of Chemical Composition of Fine Particulate Matter Using a Combined Geoscience-Statistical Method with Information from Satellites, Models, and Monitors. Environ. Sci. Technol 2019, 53 (5), 2595–2611. 10.1021/acs.est.8b06392. [DOI] [PubMed] [Google Scholar]
  • (36).Hu J; Zhang H; Chen S-H; Wiedinmyer C; Vandenberghe F; Ying Q; Kleeman MJ Predicting Primary PM2.5 and PM0.1 Trace Composition for Epidemiological Studies in California. Environ. Sci. Technol 2014, 48 (9), 4971–4979. 10.1021/es404809j. [DOI] [PubMed] [Google Scholar]
  • (37).Hu J; Zhang H; Chen S; Ying Q; Wiedinmyer C; Vandenberghe F; Kleeman MJ Identifying PM2.5 and PM0.1 Sources for Epidemiological Studies in California. Environ. Sci. Technol 2014, 48 (9), 4980–4990. 10.1021/es404810z. [DOI] [PubMed] [Google Scholar]
  • (38).Ritz B; Liew Z; Yan Q; Cuia X; Virk J; Ketzel M; Raaschou-Nielsen O Air Pollution and Autism in Denmark. Environ. Epidemiol 2018, 2 (4), e028. 10.1097/ee9.0000000000000028. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (39).Rahman MM; Shu YH; Chow T; Lurmann FW; Yu X; Martinez MP; Carter SA; Eckel SP; Chen JC; Chen Z; Levitt P; Schwartz J; McConnell R; Xiang AH Prenatal Exposure to Air Pollution and Autism Spectrum Disorder: Sensitive Windows of Exposure and Sex Differences. Environ. Health Perspect 2022, 130 (1), 17008. 10.1289/EHP9509. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (40).Lubczyńska MJ; Sunyer J; Tiemeier H; Porta D; Kasper-Sonnenberg M; Jaddoe VWV; Basagaña X; Dalmau-Bueno A; Forastiere F; Wittsiepe J; Hoffmann B; Nieuwenhuijsen M; Hoek G; de Hoogh K; Brunekreef B; Guxens M Exposure to Elemental Composition of Outdoor PM(2.5) at Birth and Cognitive and Psychomotor Function in Childhood in Four European Birth Cohorts. Environ. Int 2017, 109, 170–180. 10.1016/j.envint.2017.09.015. [DOI] [PubMed] [Google Scholar]
  • (41).Peng C; Cayir A; Sanchez-Guerra M; Di Q; Wilson A; Zhong J; Kosheleva A; Trevisi L; Colicino E; Brennan K; Dereix AE; Dai L; Coull BA; Vokonas P; Schwartz J; Baccarelli AA Associations of Annual Ambient Fine Particulate Matter Mass and Components with Mitochondrial DNA Abundance. Epidemiology 2017, 28 (6), 763–770. 10.1097/EDE.0000000000000717. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (42).Cowell WJ; Bellinger DC; Coull BA; Gennings C; Wright RO; Wright RJ Associations between Prenatal Exposure to Black Carbon and Memory Domains in Urban Children: Modification by Sex and Prenatal Stress. PLoS One 2015, 10 (11), e0142492. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (43).Guxens M; Garcia-Esteban R; Giorgis-Allemand L; Forns J; Badaloni C; Ballester F; Beelen R; Cesaroni G; Chatzi L; De Agostini M; De Nazelle A; Eeftens M; Fernandez MF; Fernández-Somoano A; Forastiere F; Gehring U; Ghassabian A; Heude B; Jaddoe VWV; Klümper C; Kogevinas M; Krämer U; Larroque B; Lertxundi A; Lertxuni N; Murcia M; Navel V; Nieuwenhuijsen M; Porta D; Ramos R; Roumeliotaki T; Slama R; Sørensen M; Stephanou EG; Sugiri D; Tardón A; Tiemeier H; Tiesler CMT; Verhulst FC; Vrijkotte T; Wilhelm M; Brunekreef B; Pershagen G; Sunyer J Air Pollution during Pregnancy and Childhood Cognitive and Psychomotor Development: Six European Birth Cohorts. Epidemiology 2014, 25 (5), 636–647. 10.1097/EDE.0000000000000133. [DOI] [PubMed] [Google Scholar]
  • (44).Fuertes E; Standl M; Forns J; Berdel D; Garcia-Aymerich J; Markevych I; Schulte-Koerne G; Sugiri D; Schikowski T; Tiesler CMT; Heinrich J Traffic-Related Air Pollution and Hyperactivity/Inattention, Dyslexia and Dyscalculia in Adolescents of the German GINIplus and LISAplus Birth Cohorts. Environ. Int 2016, 97, 85–92. https://doi.org/ 10.1016/j.envint.2016.10.017. [DOI] [PubMed] [Google Scholar]
  • (45).H. HM; R. GD; L. R-SS; J. MS; Antonella Z; A. CB; D. SJ; Alexandros G; Itai K; Petros K; C. BD; F. WR; K. SS; Emily O Prenatal and Childhood Traffic-Related Pollution Exposure and Childhood Cognition in the Project Viva Cohort (Massachusetts, USA). Environ. Health Perspect 2015, 123 (10), 1072–1078. 10.1289/ehp.1408803. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (46).Suglia SF; Gryparis A; Wright RO; Schwartz J; Wright RJ Association of Black Carbon with Cognition among Children in a Prospective Birth Cohort Study. Am. J. Epidemiol 2008, 167 (3), 280–286. 10.1093/aje/kwm308. [DOI] [PubMed] [Google Scholar]
  • (47).Yokota S; Mizuo K; Moriya N; Oshio S; Sugawara I; Takeda K Effect of Prenatal Exposure to Diesel Exhaust on Dopaminergic System in Mice. Neurosci. Lett 2009, 449 (1), 38–41. https://doi.org/ 10.1016/j.neulet.2008.09.085. [DOI] [PubMed] [Google Scholar]
  • (48).Suzuki T; Oshio S; Iwata M; Saburi H; Odagiri T; Udagawa T; Sugawara I; Umezawa M; Takeda K In Utero Exposure to a Low Concentration of Diesel Exhaust Affects Spontaneous Locomotor Activity and Monoaminergic System in Male Mice. Part. Fibre Toxicol 2010, 7, 7. 10.1186/1743-8977-7-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (49).Morris-Schaffer K; Merrill AK; Wong C; Jew K; Sobolewski M; Cory-Slechta DA Limited Developmental Neurotoxicity from Neonatal Inhalation Exposure to Diesel Exhaust Particles in C57BL/6 Mice. Part. Fibre Toxicol 2019, 16 (1), 1. 10.1186/s12989-018-0287-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (50).Morris-Schaffer K; Merrill A; Jew K; Wong C; Conrad K; Harvey K; Marvin E; Sobolewski M; Oberdörster G; Elder A; Cory-Slechta DA Effects of Neonatal Inhalation Exposure to Ultrafine Carbon Particles on Pathology and Behavioral Outcomes in C57BL/6J Mice. Part. Fibre Toxicol 2019, 16 (1), 10. 10.1186/s12989-019-0293-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (51).Antiñolo M; Willis MD; Zhou S; Abbatt JPD Connecting the Oxidation of Soot to Its Redox Cycling Abilities. Nat. Commun 2015, 6 (1), 6812. 10.1038/ncomms7812. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (52).McWhinney RD; Badali K; Liggio J; Li S-M; Abbatt JPD Filterable Redox Cycling Activity: A Comparison between Diesel Exhaust Particles and Secondary Organic Aerosol Constituents. Environ. Sci. Technol 2013, 47 (7), 3362–3369. 10.1021/es304676x. [DOI] [PubMed] [Google Scholar]
  • (53).Li Q; Shang J; Zhu T Physicochemical Characteristics and Toxic Effects of Ozone-Oxidized Black Carbon Particles. Atmos. Environ 2013, 81, 68–75. https://doi.org/ 10.1016/j.atmosenv.2013.08.043. [DOI] [Google Scholar]
  • (54).Grahame TJ; Klemm R; Schlesinger RB Public Health and Components of Particulate Matter: The Changing Assessment of Black Carbon. J. Air Waste Manag. Assoc 2014, 64 (6), 620–660. 10.1080/10962247.2014.912692. [DOI] [PubMed] [Google Scholar]
  • (55).Basagaña X; Esnaola M; Rivas I; Amato F; Alvarez-Pedrerol M; Forns J; López-Vicente M; Pujol J; Nieuwenhuijsen M; Querol X; Sunyer J Neurodevelopmental Deceleration by Urban Fine Particles from Different Emission Sources: A Longitudinal Observational Study. Environ. Health Perspect 2016, 124 (10), 1630–1636. 10.1289/EHP209. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (56).Li J; Wang Y; Steenland K; Liu P; van Donkelaar A; Martin RV; Chang HH; Caudle WM; Schwartz J; Koutrakis P; Shi L Long-Term Effects of PM2.5 Components on Incident Dementia in the Northeastern United States. Innov. 2022, 3 (2), 100208. 10.1016/j.xinn.2022.100208. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (57).Hopkins LE; Laing EA; Peake JL; Uyeminami D; Mack SM; Li X; Smiley-Jewell S; Pinkerton KE Repeated Iron-Soot Exposure and Nose-to-Brain Transport of Inhaled Ultrafine Particles. Toxicol. Pathol 2018, 46 (1), 75–84. 10.1177/0192623317729222. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (58).Fang T; Zeng L; Gao D; Verma V; Stefaniak AB; Weber RJ Ambient Size Distributions and Lung Deposition of Aerosol Dithiothreitol-Measured Oxidative Potential: Contrast between Soluble and Insoluble Particles. Environ. Sci. Technol 2017, 51 (12), 6802–6811. 10.1021/acs.est.7b01536. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (59).Koebnick C; Langer-Gould AM; Gould MK; Chao CR; Iyer RL; Smith N; Chen W; Jacobsen SJ Sociodemographic Characteristics of Members of a Large, Integrated Health Care System: Comparison with US Census Bureau Data. Perm. J 2012, 16 (3), 37–41. 10.7812/tpp/12-031. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (60).Thurston G; Awe Y; Ostro B; Sánchez-triana E Are All Air Pollution Particles Equal? How Constituents and Sources of Fine Air Pollution Particles (PM2.5) Affect Health; World Bank, Washington, DC., 2021. https://openknowledge.worldbank.org/handle/10986/36269 (accessed 2022-06-06) [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary document
NIHMS1966493-supplement-Supplementary_document.docx (450KB, docx)

RESOURCES