University of Southern California Dissertations and Theses

Early life air pollution exposure, gestational diabetes mellitus, and autism spectrum disorder

(USC Thesis Other)

Early life air pollution exposure, gestational diabetes mellitus, and autism spectrum disorder

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

EARLY LIFE AIR POLLUTION EXPOSURE, GESTATIONAL DIABETES MELLITUS,
AND AUTISM SPECTRUM DISORDER

by
Heejoo Jo

A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(EPIDEMIOLOGY)

August 2019

Copyright 2019 Heejoo Jo

ii
TABLE OF CONTENTS

LIST OF TABLES .......................................................................................................... iv
LIST OF FIGURES......................................................................................................... vi
ABSTRACT .................................................................................................................. vii
OVERVIEW OF DISSERTATION ................................................................................... 1
CHAPTER 1. UNCERTAINTIES IN THE ETIOLOGIC RELATIONSHIP OF AIR
POLLUTION TO AUTISM: CHALLENGES AND OPPORTUNITIES ............................. 3
ABSTRACT ............................................................................................................................ 3
SIGNIFICANCE...................................................................................................................... 4
Prenatal and early-life exposure to air pollution: Current evidence for the development of
ASD in children .................................................................................................................. 5
Potential biological mechanisms underlying air pollution-ASD associations........................ 6
Current methods used in study of relationship between air pollution and ASD ................... 6
RESEARCH NEEDS AND FUTURE DIRECTIONS ............................................................... 7
Hypothesis-driven studies .................................................................................................. 7
Improved exposure assessment ......................................................................................... 8
Assessing multiple risk factors and their interactions .......................................................... 9
Electronic medical records (EMR) .....................................................................................10
Neuroimaging studies to probe pathways ..........................................................................11
CONCLUSIONS ....................................................................................................................11
PURPOSE OF THE DISSERTATION ...................................................................................12
REFERENCES .....................................................................................................................14
CHAPTER 2. SEX-SPECIFIC ASSOCIATIONS OF AUTISM SPECTRUM DISODER
WITH RESIDENTIAL AIR POLLUTION EXPOSURE IN A LARGE SOUTHERN
CALIFORNIA PREGNANCY COHORT ........................................................................ 25
ABSTRACT ...........................................................................................................................25
INTRODUCTION ...................................................................................................................27
METHODS ............................................................................................................................28
Study design and population .............................................................................................28
Outcome data on ASD ......................................................................................................29
Exposure assessment .......................................................................................................29
Covariates .........................................................................................................................30
Statistical analyses ............................................................................................................31
RESULTS .............................................................................................................................33
DISCUSSION ........................................................................................................................35
CONCLUSIONS ....................................................................................................................40
REFERENCES .....................................................................................................................41
CHAPTER 3. ASSOCIATIONS OF GESTATIONAL DIABETES MELLITUS WITH
RESIDENTIAL AIR POLLUTION EXPOSURE IN A LARGE SOUTHERN CALIFORNIA
PREGNANCY COHORT ............................................................................................... 54
ABSTRACT ...........................................................................................................................54
INTRODUCTION ...................................................................................................................56
METHODS ............................................................................................................................57
Study design and population .............................................................................................57
iii
Outcome data on GDM .....................................................................................................58
Exposure assessment .......................................................................................................59
Covariates .........................................................................................................................60
Statistical analyses ............................................................................................................60
RESULTS .............................................................................................................................62
DISCUSSION ........................................................................................................................65
CONCLUSIONS ....................................................................................................................68
REFERENCES .....................................................................................................................69
CHAPTER 4. GESTATIONAL DIABETES MELLITUS, PRENATAL AIR POLLUTION
EXPOSURE, AND AUTISM SPECTRUM DISORDER ................................................. 79
ABSTRACT ...........................................................................................................................79
INTRODUCTION ...................................................................................................................81
METHODS ............................................................................................................................82
Study design and population .............................................................................................82
Outcome data on ASD ......................................................................................................83
Exposure assessment .......................................................................................................84
Maternal diabetes ..............................................................................................................85
Covariates .........................................................................................................................86
Statistical analyses ............................................................................................................86
RESULTS .............................................................................................................................89
DISCUSSION ........................................................................................................................92
CONCLUSIONS ....................................................................................................................97
REFERENCES .....................................................................................................................98
CHAPTER 5. CONCLUSIONS ................................................................................... 118
SUMMARY OF MAJOR FINDINGS ..................................................................................... 118
CONCLUSIONS AND PUBLIC HEALTH IMPLICATIONS ................................................... 119
FUTURE DIRECTIONS ....................................................................................................... 121
REFERENCES ................................................................................................................... 123

iv
LIST OF TABLES

Table 1.1. Summary of current evidence on prenatal and early-life air pollution exposure
and development of ASD in children ............................................................................. 20

Table 2.1. Characteristics of the cohort by ASD status ................................................. 47

Table 2.2. Adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) for ASD
associated with each pollutant birth address exposure ................................................. 49

Table S2.1. Pearson partial correlations of PM2.5 across exposure windows ................ 51

Table S2.2. Adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) for ASD
joint associations with child sex and quartiles of first trimester PM2.5 exposure ............. 52

Table 3.1. Demographic characteristics by GDM for singleton deliveries in 1999-2009 73

Table 3.2. Pearson partial correlations of pollutants during preconception and first
trimester ........................................................................................................................ 74

Table 3.3. Adjusted odds ratios (ORs) and 95% confidence intervals (CIs) for
association between GDM ≥13 weeks' gestation and each pollutant (single-pollutant and
mutually adjusted) ......................................................................................................... 75

Table 4.1. Associations of maternal and child characteristics with risk of ASD ........... 103

Table 4.2. Minimally adjusted and fully adjusted hazard ratios (HRs) with 95%
confidence intervals (CIs) for associations of categories of maternal diabetes with risk of
ASD ............................................................................................................................. 104

Table 4.3. Minimally adjusted and fully adjusted hazard ratios (HRs) with 95%
confidence intervals (CIs) for the associations of each pollutant birth address exposure
with risk of ASD ........................................................................................................... 105

Table 4.4. Maternal diabetes-specific adjusted hazard ratios (HRs) and 95% confidence
intervals (CIs) for O3 per 15.7 ppb within each exposure window associated with risk of
ASD ............................................................................................................................. 106

Table 4.5. Adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) for joint
associations of exposures to maternal diabetes and tertiles of O3 with risk of ASD .... 107

Table S4.1. Exposure distributions during pregnancy by maternal diabetes status in
1999-2009 ................................................................................................................... 109

Table S4.2. P-values of global interaction between each pollutant exposure and the 4-
level categorical variable for maternal diabetes during pregnancy .............................. 110
v

Table S4.3. Pearson partial correlations of pollutants during first trimester and first year
of life exposure windows ............................................................................................. 111

Table S4.4. Entire cohort with no random effect: maternal diabetes-specific adjusted
hazard ratios (HRs) and 95% confidence intervals (CIs) for ASD in associations with
each pollutant within each exposure window .............................................................. 116

Table S4.5. Restricted to one child per family: maternal diabetes-specific adjusted
hazard ratios (HRs) and 95% confidence intervals (CIs) for ASD in association with
each pollutant within each exposure window .............................................................. 117

vi
LIST OF FIGURES

Figure 2.1. Crude incidence rate of ASD by birth year .................................................. 46

Figure 2.2. Distribution of pollutant concentrations during pregnancy across birth year
1999-2009 ..................................................................................................................... 48

Figure 2.3. Child sex-specific hazard ratios (HRs) and 95% confidence intervals (CIs)
for ASD per 6.5 μg/m3 PM2.5 during each exposure period ........................................... 50

Figure S2.1. Kaiser Permanente Southern California (KPSC) medical center service
areas ............................................................................................................................. 53

Figure S3.1. Kaiser Permanente Southern California (KPSC) medical center service
areas ............................................................................................................................. 76

Figure S3.2. Distribution of pollutant concentrations during preconception and first
trimester across birth year 1999-2009 ........................................................................... 77

Figure 4.1. Adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) for joint
associations of exposures to GDM diagnosis at <24 weeks’ gestation and tertiles of O3
with risk of ASD relative to no diabetes and low tertiles of O3 exposure. ..................... 108

Figure S4.1. Kaiser Permanente Southern California (KPSC) medical center service
areas ........................................................................................................................... 112

Figure S4.2. Crude incidence rate of ASD by birth year .............................................. 113

Figure S4.3. Distribution of pollutant concentrations during pregnancy across birth year
1999-2009 ................................................................................................................... 114

Figure S4.4. Distribution of 1999-2009 mean pollutant concentrations across KPSC
medical center service areas ....................................................................................... 115

vii
ABSTRACT
In the United States (U.S.), autism spectrum disorder (ASD) prevalence has increased
over several decades among young children, disproportionately affecting more boys
than girls.
The effects of air pollution and gestational diabetes mellitus (GDM) have been
hypothesized to increase the risk of ASD through common biologic mechanisms
including oxidative stress and inflammation that could alter fetal brain development.
Particulate matter <2.5 in aerodynamic diameter (PM2.5) has shown relatively consistent
adverse effects on ASD; however, associations between air pollution and GDM have
been inconsistent. Because GDM is also hypothesized to be a risk factor for ASD, it is
important to evaluate these associations together, which no studies to date have done
for ASD. This dissertation aimed to address this critical gap in the current literature on
environmental epidemiology of ASD by evaluating the role of air pollution in the etiology
of GDM and ASD. Based on animal studies showing male-specific effects of particles on
autism-like behaviors, results suggested that boys may be at a higher risk for ASD than
girls if exposed to higher first trimester PM2.5 exposure. In examining the associations
between pollutants and GDM, maternal exposure to nitrogen dioxide before conception
was associated with increased risk of GDM. Furthermore, GDM diagnosed earlier in
pregnancy may increase children’s susceptibility to prenatal and early life ozone-
associated ASD risk. Further studies are needed to better understand the biological
basis and timing of susceptibility during neurodevelopment. Air pollution and GDM are
common risk factors amenable to intervention targeting public policy and behavior.
1
OVERVIEW OF DISSERTATION
Recent epidemiologic studies have suggested that exposures to ambient air pollution
increase the risk of autism spectrum disorder (ASD) in young children. However, there
is a lack of consensus on the specific developmental time period during which exposure
to ambient air pollution will increase vulnerability to ASD. Additionally, few studies have
assessed joint effects of more than one risk factor (e.g. air pollution and GDM), which is
important in a multifactorial disease such as ASD that likely has multiple causal
pathways. In Chapter 1, I provided a commentary on the current evidence of the
associations between air pollution and risk of ASD and identified research needs to
improve the understanding of the causal relationship. Based on Chapter 1, I identified
the following gaps in knowledge to clarify the relationship between air pollution and
ASD:
1. During which time periods during pregnancy and in early postnatal life are
children most susceptible to air pollution producing ASD?
2. Are the effects of air pollution greater in boys than girls?
3. What about synergistic effects of air pollution with other common risk factors of
ASD, such as maternal diabetes during pregnancy?
Based on recent evidence from animal studies, my first paper (Chapter 2) tested the
hypothesis that the risk of ASD associated with PM2.5 exposure during early
development would be greater in boys than girls in a large, population-based birth
cohort. In the second paper (Chapter 3), I examined the associations between maternal
pollutant exposure during 12 weeks before last menstrual period date and during the
first trimester with risk of subsequent onset of gestational diabetes mellitus (GDM).
2
Based on recent epidemiologic studies that found associations between maternal
diabetes and ASD risk, and on potential common biological pathways of diabetes and
air pollution effects, the third paper (Chapter 4) assessed the joint effects of pollutant
exposures and maternal diabetes during pregnancy. Multiple pollutant exposure
windows were examined in this dissertation, including 12 weeks before last menstrual
period date, each trimester of pregnancy, and the first year of child’s life to identify
developmental windows of vulnerability.
In Chapter 5, I summarize the major findings and public health implications from
this dissertation, and discuss remaining questions that need further investigation in
future studies.

3
CHAPTER 1. UNCERTAINTIES IN THE ETIOLOGIC RELATIONSHIP OF AIR
POLLUTION TO AUTISM: CHALLENGES AND OPPORTUNITIES

ABSTRACT
Prevalence of autism spectrum disorder (ASD) has increased in past decades. Recent
studies have suggested air pollution increases the risk of ASD. However, it is largely
unknown during which times exposure to air pollution has the largest magnitude of
effect during the developmental period. Although there are multiple potential causal
pathways increasing ASD risk, most epidemiological studies have examined single risk
factors rather than examining how different risk factors interact. We summarize the
current evidence on the associations between air pollution and risk of ASD and identify
research needs to improve our understanding of the causal relationship. We also
discuss opportunities to examine combinations of risk factors and improve exposure
assessment. Associations between particulate matter less than 2.5 in aerodynamic
diameter and ASD have been consistently reported; however, most studies have used
parent-reported residential history or birth certificate address to assign air pollution
exposure. Future studies may consider using prospectively collected residential history
to improve trimester-specific exposure assessment, and examine joint effects of air
pollution and other risk factors to better understand the etiologies of this complex
disease. Identifying modifiable risk factors may reduce the incidence of ASD. More
research is needed to develop standardized methods in exposure and outcome
assessment for increasing comparability of results across studies of air pollution effects
4
on ASD. Neuroimaging studies and high quality electronic medical records may be
useful in addressing these issues and in clarifying biologic relationships.

SIGNIFICANCE
Autism spectrum disorder (ASD) is a complex disease syndrome with early childhood
onset characterized by impairments in social interactions, communication, and
restricted, repetitive, and stereotyped patterns of behavior (1). Prevalence estimates of
ASD have increased markedly from 1 in 294 children aged 3–10 years in 1996 to its
current estimated prevalence of 1 in 59 children by age 8 years (2). The cause of this
dramatic increase in prevalence remains largely unknown, although changes in
diagnostic practice resulting in better ascertainment of ASD may account for some of
the rise (3). Environmental factors likely contribute to the epidemic, perhaps in
combination with known genetic risks (4, 5). While several environmental risk factors
have been examined, including pesticides and household products, studies on
particulate air pollution have relatively consistently reported associations with increases
in the risk of autism (4). Previous studies have mostly evaluated single exposures in
isolation, not taking into account the potential multiple, complex pathways that
characterize ASD risk. In this first chapter of my dissertation, we summarize the current
evidence on the association between air pollution and risk of ASD. We identify research
needs that would improve our understanding of the causal relationship between
prenatal and early-life exposure to air pollution and subsequent development of ASD.
We identify opportunities to: 1) conduct more rigorous, hypothesis-driven studies, 2)
improve air pollution exposure assessment, and 3) examine multiple risk factors of ASD.
5

Prenatal and early-life exposure to air pollution: Current evidence for the development
of ASD in children
Recent research has identified air pollution as a risk factor for autism (4). There have
been 16 published studies (Table 1.1), most finding increased risk of ASD associated
with exposure to particulate matter less than 2.5 and 10 µm in aerodynamic diameter
(PM2.5 and PM10, respectively) or with nitrogen dioxide (NO2) (6-16). Studies have also
found positive associations with ozone (O3) (6, 7), carbon monoxide and sulfur dioxide
(SO2) (7, 14), near-roadway air pollution (NRAP) mixture (8, 17), and with diverse
hazardous air pollutants compiled by the United States Environmental Protection
Agency (18-21). Specifically, recent epidemiologic studies have found associations
between ASD and ambient PM2.5 during gestation and the first year of life (22). The
heterogeneity of exposures implicated in ASD risk in different studies poses a challenge
to a coherent interpretation of causal relationships, and suggests that multiple
mechanisms of action may be involved. Studies with contradictory results are also
evident. A recent large study of four European population-based birth/child cohorts
reported no association of autistic traits with prenatal exposure to air pollutants (NO2
and PM2.5) (23), and two studies from Sweden also reported no associations of autistic
traits and ASD with either pre- or postnatal traffic-related air pollutants (PM10 and NOx)
(24, 25). Moreover, a recent study from Iran reported no associations of ASD risk with
prenatal exposure to PM10 and SO2 (26), and a study from Ohio also reported no
associations of ASD risk with PM2.5 and O3 (15).

6
Potential biological mechanisms underlying air pollution-ASD associations
Both genetic and biomarker studies have suggested that children with ASD have
increased systemic levels of pro-inflammatory cytokines, and significantly increased
susceptibility to oxidative stress from impaired redox homeostasis (27). Urban PM and
diesel exhaust, a prototype near-roadway pollutant, have also been shown to induce
inflammatory responses and oxidative stress (4). Recent research has indicated that
intrauterine exposure to gestational diabetes mellitus (GDM) might also have adverse
effects on fetal brain development, including oxidative stress, chronic inflammation, and
hypoxia, thereby increasing the risk of ASD (4). The overlapped timing of both pollutant-
and GDM-induced systemic oxidative stress and inflammation may potentially affect
common biological pathways and thus, further increase the susceptibility to ASD in
children (28, 29).

Current methods used in study of relationship between air pollution and ASD
Existing investigations of environmental influences on ASD include the reliance on self-
report of outcome or reliance on ASD symptoms in the limited number of prospective
cohort studies (19, 23). One promising approach is a new generation of efficient,
prospective studies of high-risk siblings of children with ASD, who have higher rates of
the disease (30). In these studies, information on exposure and covariates are collected
prospectively and diagnosis can be clinically verified. However, the studies are not
population-based and the sample size is too small to be able to examine unusual
exposures or multifactorial causes of ASD. Several case-control studies have used
rigorous clinical examinations to verify ASD diagnosis, but there is potential for selection
7
bias based on the willingness of subjects to participate, especially among population-
based controls selected from birth records (8, 9, 11, 15, 17).

RESEARCH NEEDS AND FUTURE DIRECTIONS
Hypothesis-driven studies
Given the inconsistent results in exposure associations across studies, more
hypothesis-driven research is needed. Specific hypotheses can be generated based on
previous biological and animal toxicological studies as well as epidemiologic studies
incorporating complementary information designed to identify common biological
pathways. For example, in animal studies benzo(a)pyrene, a polycyclic aromatic
hydrocarbon present in high concentration near large sources of gasoline combustion
such as major roadways, reduced expression of the established functional ASD
candidate gene, MET (31). In the CHARGE study a larger risk of NRAP-associated ASD
was observed in children with the high-risk MET genotype (32). Additionally, recent
animal studies found that early life exposure to ambient particles caused autism-like
behaviors only in males. However, there has been little research on differences in
gender-specific effects on ASD in humans.
Existing paradigms for the exploration of genetic contributions to ASD are well
suited to exploring the combined effects of genetic and multiple environmental risk
factors in large data sets. Multi-stage designs in which hypothesis-generating analyses
are followed by studies of subjects selected jointly based on varying exposure and/or
outcome profiles could result in rapid progress in our understanding of the causes and
the biology of ASD.
8

Improved exposure assessment
A majority of studies have assigned trimester-specific air pollution exposure to a birth
certificate address (6, 9, 15, 19, 20, 23), or to residential addresses retrospectively
collected from a parent (8, 11, 26). In case-control studies, exposure to other maternal
and/or child risk factors during the critical gestational and early life etiologic periods has
often been based on retrospectively collected parental self-report. Air pollutant exposure
assessment may have less potential for systematic bias than other risk factors, since
exposure assignment is commonly based on residential address history, which is less
prone to recall failures. However, 9-34% of mothers move during pregnancy (33, 34),
and poorly recalled dates of move likely result in misclassification of trimester-specific
exposures. In large data sets, assigning exposures across pregnancy to a birth
certificate address will misclassify exposure of movers, and this misclassification is
likely to be worse in earlier trimesters when mothers are more likely to move (9).
To avoid potential misclassification of trimester-specific air pollutant exposure,
prospectively collected residential addresses is needed to identify potential dates of
move and to assign accurate air pollution concentrations to specific developmental
windows of vulnerability during pregnancy. This could provide insight into the
mechanism of air pollution effects, based on the known trajectory of neurodevelopment
across pregnancy.

9
Assessing multiple risk factors and their interactions
Uncertainties inherent to current approaches to the study of environmental risk factors
include the likely role of multiple factors that in combination cause or contribute to the
development of complex diseases like ASD. In contrast to recent studies examining the
effects of variation across the genome, most environmental studies have focused on
associations of single risk factors, in part because the ability to study combinations of
factors that interact requires large study populations. Synergistic effects of prenatal air
pollution exposure with other common risk factors could potentially explain a substantial
proportion of ASD. For example, since diabetes and GDM among women of
reproductive age have been increasing during the same period in which the prevalence
of ASD has been increasing, these conditions could help explain the trajectory of the
ASD epidemic (35, 36). Levels of regulated regional air pollutants have decreased
across the U.S. over the same period during which ASD prevalence has increased.
However, if increasing rates of GDM potentiate the effects of air pollution on ASD,
together they could begin to explain substantial proportions of the epidemic.
Possible synergistic effects of air pollution with other maternal co-morbid risk
factors during pregnancy could also contribute to high prevalence of ASD. For example,
biological mechanisms that increase oxidative stress and inflammation such as
maternal infections (i.e., rubella and cytomegalovirus), obesity, depression, alcohol or
drug use, and pregnancy complications (e.g. premature birth) have been previously
identified as risk factors of ASD (3). Moreover, modifiable protective factors such as folic
acid intake during first month of pregnancy have been shown to decrease ASD risk
among children or their mothers with a common variant of the
10
methylenetetrahydrofolate reductase gene, MTHFR 677 C>T (37). A recent study found
a significant interaction between folic acid intake and NO2, suggesting that higher folic
acid intake during the first trimester may reduce ASD risk in children born to mothers
exposed to higher NO2 levels (38).

Electronic medical records (EMR)
One approach that could address the gaps and uncertainties in our understanding of the
relationship between air pollution, including air pollution in combination with other risk
factors, and ASD risk, would be to utilize EMR data from health maintenance
organizations (HMOs) with stable, population-based membership. For example, Kaiser
Permanente has pioneered the use of EMR that could be used to improve the accuracy
of trimester-specific and early-life air pollution exposure assignment to residential
addresses that are updated prospectively at the time of clinical contacts that are
frequent during pregnancy and after birth. Certain Kaiser regional groups are
population-based. For example, Kaiser Permanente Southern California (KPSC)
represents 16% of working families in the region (39). Standardized and validated
clinical diagnostic algorithms, implemented across providers at KPSC would also help
improve outcome assessment (40).
The KPSC EMR database can be used to construct a large, population-based
pregnancy cohort. This database, which contains rich clinical information on maternal
co-morbidities and other covariates that share common biologically plausible pathways
with air pollution, can be leveraged to examine synergistic effects. Untargeted analyses
11
using bioinformatics approaches to explore the EMR data could also be investigated to
generate new hypotheses.

Neuroimaging studies to probe pathways
New neuroimaging tools could be used to examine the structural neurobiological
sequelae of environmental exposures in selected genotypes. Recent toxicological
studies have shown that air pollution causes structural changes in the brain that could
be imaged in children. Previous studies in a mouse model observed that concentrated
ultrafine particles (less than 0.1 µm in diameter) that are present in high concentrations
in fresh exhaust emissions from both diesel and gasoline combustion near major
roadways, caused largely male-specific reductions in the size of the corpus callosum
and associated hypomyelination, aberrant white matter development and
ventriculomegaly (41). Moreover, a recent study of children from New York City
exposed to high in-utero and early-life polyaromatic hydrocarbons found white matter
surface reduction in the left hemisphere and in the dorsal prefrontal regions bilaterally
compared to children with low in-utero and early-life exposures (42).

CONCLUSIONS
In summary, the uncertainties and challenges that remain in examining the relationship
between air pollution and ASD offer an opportunity to 1) improve air pollution exposure
assessment, 2) assess the joint effects of multiple risk factors, and 3) utilize new study
designs that would provide further insight into causal pathways and biological
mechanisms influencing ASD risk. More research is needed to develop standardized
12
methods in exposure and outcome assessments for increasing comparability of results
across studies of air pollution effects. Identifying modifiable environmental factors would
provide an opportunity to intervene and reduce the incidence of ASD.

PURPOSE OF THE DISSERTATION
Given the limited number of population-based cohort studies examining the effects of
regional air pollution and joint effects of combinations of exposures on ASD risk in the
current literature, we used a large, population-based cohort constructed from the KPSC
EMR database that had high quality information on various covariates to examine our
study aims. In Aim 1, we hypothesized that boys would be at a higher risk of ASD
associated with PM2.5 exposure during early development compared to girls, based on
recent evidence from animal studies showing that early life exposure to ambient PM2.5
caused autism-like behaviors only in males. Based on this hypothesis, we evaluated
whether the association between child sex and PM2.5 differed by child sex using the
large KPSC cohort. In Aim 2, we examined the associations between maternal pollutant
exposures and GDM by defining the temporal relationship between pollutant exposures
both before and after conception, and subsequent development of GDM based on the
GDM diagnosis date from KPSC EMR. We accounted for potential confounding by
exposure windows and co-pollutants by mutually adjusting for both preconception and
first trimester exposures of each pollutant, and controlling for each co-pollutant from
both exposure windows. In Aim 3, we evaluated whether pollutant exposures interact
with maternal diabetes status during pregnancy to impact child’s ASD risk. Based on
recent epidemiologic studies that found associations between maternal diabetes and
13
ASD risk, we hypothesized that the effect of pollutant exposures on ASD were greater
among children born to mothers diagnosed with diabetes during pregnancy compared to
mothers without diabetes.

14
REFERENCES

1. Association AP. Diagnostic and statistical manual of mental disorders, 4th
edition, text revision (DSM-IV-TR). Washington, DC: American Psychiatric
Publishing; 2000.
2. Baio J, Wiggins L, Christensen DL, et al. Prevalence of Autism Spectrum
Disorder Among Children Aged 8 Years - Autism and Developmental Disabilities
Monitoring Network, 11 Sites, United States, 2014. MMWR Surveill Summ
2018;67(6):1-23.
3. Ornoy A, Weinstein-Fudim L, Ergaz Z. Prenatal factors associated with autism
spectrum disorder (ASD). Reprod Toxicol 2015;56:155-69.
4. Lyall K, Schmidt RJ, Hertz-Picciotto I. Maternal lifestyle and environmental risk
factors for autism spectrum disorders. Int J Epidemiol 2014;43(2):443-64.
5. Tchaconas A, Adesman A. Autism spectrum disorders: a pediatric overview and
update. Curr Opin Pediatr 2013;25(1):130-44.
6. Becerra TA, Wilhelm M, Olsen J, et al. Ambient air pollution and autism in Los
Angeles county, California. Environ Health Perspect 2013;121(3):380-6.
7. Jung CR, Lin YT, Hwang BF. Air pollution and newly diagnostic autism spectrum
disorders: a population-based cohort study in Taiwan. PLoS One
2013;8(9):e75510.
8. Volk HE, Lurmann F, Penfold B, et al. Traffic-related air pollution, particulate
matter, and autism. JAMA Psychiatry 2013;70(1):71-7.
15
9. Kalkbrenner AE, Windham GC, Serre ML, et al. Particulate matter exposure,
prenatal and postnatal windows of susceptibility, and autism spectrum disorders.
Epidemiology 2015;26(1):30-42.
10. Raz R, Roberts AL, Lyall K, et al. 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. Environ Health Perspect
2015;123(3):264-70.
11. Talbott EO, Marshall LP, Rager JR, et al. Air toxics and the risk of autism
spectrum disorder: the results of a population based case-control study in
southwestern Pennsylvania. Environ Health 2015;14:80.
12. Chen G, Jin Z, Li S, et al. Early life exposure to particulate matter air pollution
(PM. Environ Int 2018;121(Pt 2):1121-7.
13. Pagalan L, Bickford C, Weikum W, et al. Association of Prenatal Exposure to Air
Pollution With Autism Spectrum Disorder. JAMA Pediatr 2018.
14. Ritz B, Liew Z, Yan Q, et al. Air pollution and autism in Denmark. Environmental
Epidemiology 2018;2(e028).
15. Kaufman JA, Wright JM, Rice G, et al. Ambient ozone and fine particulate matter
exposures and autism spectrum disorder in metropolitan Cincinnati, Ohio.
Environ Res 2019;171:218-27.
16. Raz R, Levine H, Pinto O, et al. Traffic-Related Air Pollution and Autism
Spectrum Disorder: A Population-Based Nested Case-Control Study in Israel.
Am J Epidemiol 2018;187(4):717-25.
16
17. Volk HE, Hertz-Picciotto I, Delwiche L, et al. Residential proximity to freeways
and autism in the CHARGE study. Environ Health Perspect 2011;119(6):873-7.
18. Kalkbrenner AE, Daniels JL, Chen JC, et al. Perinatal exposure to hazardous air
pollutants and autism spectrum disorders at age 8. Epidemiology
2010;21(5):631-41.
19. Roberts AL, Lyall K, Hart JE, et al. Perinatal air pollutant exposures and autism
spectrum disorder in the children of Nurses' Health Study II participants. Environ
Health Perspect 2013;121(8):978-84.
20. von Ehrenstein OS, Aralis H, Cockburn M, et al. In utero exposure to toxic air
pollutants and risk of childhood autism. Epidemiology 2014;25(6):851-8.
21. Windham GC, Zhang L, Gunier R, et al. Autism spectrum disorders in relation to
distribution of hazardous air pollutants in the san francisco bay area. Environ
Health Perspect 2006;114(9):1438-44.
22. Lyall K, Croen L, Daniels J, et al. The Changing Epidemiology of Autism
Spectrum Disorders. Annu Rev Public Health 2016.
23. Guxens M, Ghassabian A, Gong T, et al. Air Pollution Exposure during
Pregnancy and Childhood Autistic Traits in Four European Population-Based
Cohort Studies: The ESCAPE Project. Environ Health Perspect 2015.
24. Gong T, Almqvist C, Bölte S, et al. Exposure to air pollution from traffic and
neurodevelopmental disorders in Swedish twins. Twin Res Hum Genet
2014;17(6):553-62.
17
25. Gong T, Dalman C, Wicks S, et al. Perinatal Exposure to Traffic-Related Air
Pollution and Autism Spectrum Disorders. Environ Health Perspect
2017;125(1):119-26.
26. Yousefian F, Mahvi AH, Yunesian M, et al. Long-term exposure to ambient air
pollution and autism spectrum disorder in children: A case-control study in
Tehran, Iran. Sci Total Environ 2018;643:1216-22.
27. Rossignol DA, Frye RE. A review of research trends in physiological
abnormalities in autism spectrum disorders: immune dysregulation, inflammation,
oxidative stress, mitochondrial dysfunction and environmental toxicant
exposures. Mol Psychiatry 2012;17(4):389-401.
28. Bowers K, Zhang C. Maternal diabetes and autism spectrum disorders in the
offspring: a review of epidemiological evidence and potential biologic
mechanisms. N A J Med Sci 2011;4(4):217-21.
29. Estes ML, McAllister AK. Maternal immune activation: Implications for
neuropsychiatric disorders. Science 2016;353(6301):772-7.
30. Szatmari P, Chawarska K, Dawson G, et al. Prospective Longitudinal Studies of
Infant Siblings of Children With Autism: Lessons Learned and Future Directions.
J Am Acad Child Adolesc Psychiatry 2016;55(3):179-87.
31. Sheng L, Ding X, Ferguson M, et al. Prenatal polycyclic aromatic hydrocarbon
exposure leads to behavioral deficits and downregulation of receptor tyrosine
kinase, MET. Toxicol Sci 2010;118(2):625-34.
18
32. Volk HE, Kerin T, Lurmann F, et al. Autism spectrum disorder: interaction of air
pollution with the MET receptor tyrosine kinase gene. Epidemiology
2014;25(1):44-7.
33. Bell ML, Belanger K. Review of research on residential mobility during
pregnancy: consequences for assessment of prenatal environmental exposures.
J Expo Sci Environ Epidemiol 2012;22(5):429-38.
34. Chen L, Bell EM, Caton AR, et al. Residential mobility during pregnancy and the
potential for ambient air pollution exposure misclassification. Environ Res
2010;110(2):162-8.
35. Albrecht SS, Kuklina EV, Bansil P, et al. Diabetes trends among delivery
hospitalizations in the U.S., 1994-2004. Diabetes Care 2010;33(4):768-73.
36. Ferrara A. Increasing prevalence of gestational diabetes mellitus: a public health
perspective. Diabetes Care 2007;30 Suppl 2:S141-6.
37. Schmidt RJ, Tancredi DJ, Ozonoff S, et al. Maternal periconceptional folic acid
intake and risk of autism spectrum disorders and developmental delay in the
CHARGE (CHildhood Autism Risks from Genetics and Environment) case-
control study. Am J Clin Nutr 2012;96(1):80-9.
38. Goodrich AJ, Volk HE, Tancredi DJ, et al. Joint effects of prenatal air pollutant
exposure and maternal folic acid supplementation on risk of autism spectrum
disorder. Autism Res 2018;11(1):69-80.
39. Koebnick C, Langer-Gould AM, Gould MK, et al. Sociodemographic
characteristics of members of a large, integrated health care system: comparison
with US Census Bureau data. Perm J 2012;16(3):37-41.
19
40. Coleman KJ, Lutsky MA, Yau V, et al. Validation of Autism Spectrum Disorder
Diagnoses in Large Healthcare Systems with Electronic Medical Records. J
Autism Dev Disord 2015;45(7):1989-96.
41. Allen JL, Oberdorster G, Morris-Schaffer K, et al. Developmental neurotoxicity of
inhaled ambient ultrafine particle air pollution: Parallels with neuropathological
and behavioral features of autism and other neurodevelopmental disorders.
Neurotoxicology 2015.
42. Peterson BS, Rauh VA, Bansal R, et al. Effects of prenatal exposure to air
pollutants (polycyclic aromatic hydrocarbons) on the development of brain white
matter, cognition, and behavior in later childhood. JAMA Psychiatry
2015;72(6):531-40.

20
Table 1.1. Summary of current evidence on prenatal and early-life air pollution exposure and development of ASD in
children
Authors Study Design /
Population
Exposure
Assessment
ASD Definition Confounders Results Limitations
Particles, Nitrogen Oxides, Ozone, Near-Roadway Air Pollution
Becerra et
al., 2013
Nested case-
control (7603
cases, 76,782
population
controls with no
known ASD
diagnosis)
Daily NO2, PM2.5,
and PM10 levels
estimated at
geocoded birth
addresses based on
monitoring data and
land use regression;
1995-2006 CA birth
certificates linked to
CA DDS; LA county
Autistic disorder
diagnosis (severe
autism) at 3-5
years of age
identified through
CA DDS (birth
years 1995-2006)
Maternal: age,
education,
race/ethnicity, parity,
insurance type

Child: birthplace,
gestational age at
delivery, birth year
Associations with ASD
risk:

From monitoring data:
ozone and PM2.5
mutually adjusted

From LUR: NO and NO2
Limited comparability
across studies due to
exposure estimates in
different units
Chen et al.
2018 (China)
Case-control
(124 cases, 1240
controls)
Daily PM1, PM2.5,
and PM10 at
residential address
during first three
years of life using
satellite remote
sensing data 2005-
2016
ASD cases
identified based on
parent and teacher
questionnaires,
and validated by
pediatricians for
children born after
2005
Birth weight,
gestational weeks,
disease history, trauma
history, maternal age,
familial mental health
history, marital status,
parental relationship,
parenting, income,
parents’ education
level, smoking status;
sensitivity analyses
controlled for potential
spatial clustering and
neighborhood-level
SES
Associations with ASD
risk:

PM1, PM2.5, and PM10
during first three year of
life; higher risk in the
second and third year
of life compared to first
year of life
Only postnatal
exposure windows;
limited comparability to
US studies due to
higher air pollution
levels; potential ASD
misclassification due to
lack of diagnosis date;
analysis adjusted for
potential mediators
(i.e. birth weight,
gestational weeks)
Gong et al.,
2014
(Sweden)
Nested case-
control (3426
twins)
Residence time-
weighted PM10 and
NOx at residential
addresses during
pregnancy, first year,
and ninth year using
dispersion models
and controlling for
season
Telephone
interviews (A-TAC)
for ASD and
ADHD symptom-
based score for
twins aged 9-12
years; birth years
1992-2000
Maternal: age,
smoking, marital status
at birth year, parity,
parental education,
family income,
neighborhood
deprivation at birth year

Child: gender
No associations between
pre- and post-natal air
pollution from road traffic
(PM10, NOx) and ASD,
ADHD
Exposure time at ninth
year may have
occurred after onset of
ASD or ADHD
Gong et al.,
2017
(Sweden)
Case-control
(5136 cases,
18,237 controls)
PM10, NOx from
road traffic estimated
at residential
addresses during
pregnancy and first
ASD cases
identified from
multiple health
registers in
Stockholm
Conditioned on
municipality and
calendar year of birth;
maternal marital status,
parental education and
No associations between
early life air pollution from
road traffic (PM10, NOx)
and ASD; inverse
association among
Limited comparability
to US studies due to
lower air pollution
levels
21
year using
dispersion models
employment,
disposable household
income, neighborhood
deprivation; child
gender, birth month,
birth order
children of mothers who
moved during pregnancy
Guxens et
al., 2015
(Europe)
Retrospective
cohort (8079
children from
ESCAPE)
NO2, PM2.5, PM10
estimated at birth
address by LUR
models based on
monitoring
campaigns between
2008-2011
Symptom-based
scores for autistic
traits at 4-10 years
of age
Maternal: age, prenatal
smoking, education,
parity, prepregnancy
BMI, height

Child: gender, season
at birth, urbanicity at
birth address, age at
autistic trait
assessment and
evaluator
No associations between
NO2, PM2.5 and PM10 and
autistic traits in children
aged 4-10 years
Limited comparability
across studies that
used clinical ASD
diagnosis; different
autistic traits
assessments used
across the 4 cohorts
(e.g. CBCL, SRS,
CAST)
Jung et al.,
2013
(Taiwan)
Cohort (49,073
children)

Inverse distance
weighting method
used for yearly
average postnatal
exposure of ozone,
CO, NO2, SO2, and
PM10 before newly
diagnostic ASD
(preceding 1-4
years)
Two consensus
diagnosis of ASD
using ICD-9 codes
recorded in the
Taiwan National
Insurance
Research
Database for
children aged <3
years in
2000 through 2010
(birth years 1997-
2000)
Municipal level of SES
(annual average
income)
Child: age, gender,
preterm birth,
comorbidities (e.g.
anxiety, bipolar
disorder, intellectual
disability)

Associations with ASD
risk:

Increased levels of ozone,
CO, NO2, and SO2 (multi-
pollutant models)
Limited comparability
across studies due to
lack of important
confounders (e.g.
parental age &
education) and using
postnatal air pollution
exposure
Kalbrenner
et al., 2015
Nested case-
control (NC: 645
cases, 12,434
controls; CA: 334
cases, 2232
controls)
Estimated daily
PM10 levels from
regulatory monitors
at birth address
using a geostatistical
interpolation method
Cases recorded in
autism
surveillance
systems with birth
records (CDC
ADDM) at age 8;
birth years 1994,
1996, 1998, and
2000
Year, state, maternal
education & age,
race/ethnicity,
neighborhood
urbanization,
household income,
week of birth (to
account for season)
Third trimester-specific
association between PM10
and ASD risk; null first
trimester-specific
association after adjusting
for third trimester
exposure

Lack of important
confounders (e.g.
prenatal smoking and
child sex)
Kaufman et
al. 2019
Case-control
(428 cases, 6420
frequency-
matched
controls)
US EPA-modeled
daily PM2.5 and
ozone estimates
based on mother’s
census tract of
ASD cases
identified from
Cincinnati
Children’s Hospital
EMR based on
Birth year, maternal
education, birth
spacing, maternal
prepregnancy BMI,
month of conception;
No associations between
trimester-specific and
early life (PM2.5 and
ozone) and ASD in single-
pollutant and multi-
Only pairwise
adjustments of multiple
exposure windows
(e.g. first and second
trimesters); lack of
22
residence at birth
using
ICD-9 diagnostic
codes for children
born between
2006-2010
other pollutant and
other exposure
windows; sensitivity
analyses included other
covariates
pollutant & multi-window
models
data on other
pollutants
Pagalan et
al., 2018
(Canada)
Cohort (132,256
births, 1307 ASD
cases)
Monthly mean
estimates of PM2.5,
NO, and NO2 during
pregnancy (also by
trimester) using land
use regression
models
ASD diagnosis by
age 5 based on
ADIR and ADOS
for children born
between 2004-
2009 with follow-
up to 2014
Child sex, birth month,
birth year, maternal
age, maternal
birthplace,
neighborhood-level
urbanicity, income
band
NO associations with
increased ASD risk during
pregnancy; males at
higher NO-associated
ASD risk than females (p
for interaction not
significant);
Lacked important
confounders (e.g.
maternal race/ethnicity,
education); only single-
pollutant models
examined
Raz et al.,
2015
Nested case-
control (245
cases, 1522
controls without
ASD)
Monthly averages of
PM2.5, and PM10
predicted from a
spatiotemporal
model and linked to
residential
addresses (mailing
address from
biennial NHS II
questionnaire)
Maternal report of
ASD diagnosis
(validated for 50
cases) for children
born between
1990-2002
Maternal & paternal
ages at birth, median
census tract income in
birth year

Child: birth year &
month, sex
PM2.5 associations with
increased ASD risk among
non-movers; stronger third
trimester risk after
mutually adjusting for all
trimesters
Lacked exact dates of
moves; small sample
size; lacked important
confounders (e.g.
maternal education,
smoking)
Raz et al.
2018 (Israel)
Nested case-
control (2098
cases, 54,191
controls)
Weekly NO2
estimates based on
birth address during
pregnancy to 9
months after birth
ASD cases based
on national claims
data; children born
between 2005-
2009
Birth year, birth month,
population group,
paternal age, census
tract SES; sensitivity
analyses included other
covariates
Mutually adjusting for two
exposure windows, NO2
associations during
pregnancy were negative
while during 9 months
after pregnancy was
positive
Lack of information on
other potentially
confounding pollutants;
potentially biased
estimates from
mutually adjusting for
highly correlated
exposure windows
Ritz et al.
2018
Case-control
(15,387 cases,
68,139
population
controls matched
by birth year &
sex)
AirGIS dispersion-
modeled for PM2.5,
PM10, NO2, SO2
based on maternal
residential address
from Danish
population registry
during 9 months
before pregnancy,
each trimester of
pregnancy, 9 months
after birth
ASD diagnosis
based on ICD-10
diagnostic codes
from hospital &
outpatient
admissions data;
children born
between 1989-
2013
Parental age,
neighborhood
socioeconomic
indicators (employment
& housing), and
maternal smoking
Mutually adjusting for all
exposure periods,
9 months after pregnancy:
NO2, PM2.5, PM10, SO2
associations; stronger
associations in more
recent years (2000–2013)
and in larger cities
compared with provincial
towns/rural counties. For
particles and NO2,
associations were only
specific to autism and
Asperger diagnoses.
Lack of information on
individual-level SES
confounders such as
education; potentially
biased estimates from
mutually adjusting for
highly correlated
exposure windows
23
Talbott et al.,
2015
Case-control
(211 cases, 219
population-based
controls)
LUR model used for
person-and time-
specific PM2.5 (pre-
pregnancy,
trimesters, 1-2 years
after birth,
cumulative) and
assigned to
questionnaire-based
residential and work
history; multi-
pollutant air
monitoring campaign
ASD cases based
on SCQ and
ADOS for birth
years 2005-2009
in Pennsylvania
Maternal age,
education, race,
smoking
PM2.5 exposure at year 2
and cumulative exposure
from pre-pregnancy to
year 2 associated with
increased ASD risk
Lacked important
confounders (e.g.
income, child sex);
small sample size;
lower IQR than other
studies; lack of multi-
pollutant and multi-
window models
Volk et al.,
2011
Case-control
(304 cases, 259
typically
developing
population
controls)
Distance to freeway
and major road
estimated at time of
delivery based on
residential history
ASD diagnosis
from ADOS and
ADI-R for birth
years 1997-2006
in CHARGE Study
(CA)
Maternal: age,
smoking, parental
education

Child: sex,
race/ethnicity
Higher ASD risk for those
living < 309m from nearest
freeway at time of delivery
and during third trimester;
no association with living
near major roads
Global measure of
exposure (no individual
pollutants); small
sample size
Volk et al.,
2013
Case-control
(279 cases, 245
typically
developing
population
controls)
Traffic-related air
pollution (TRP), NO2,
PM2.5, PM10
estimated at
residential address
during pregnancy &
1 year after birth
using CALINE4
dispersion model
and regional air
monitors
ASD diagnosis
from ADOS and
ADI-R for birth
years 1997-2006
in CHARGE Study
(CA)
Maternal: age,
smoking, parental
education

Child: sex,
race/ethnicity
Higher ASD risk
associated with highest
quartile TRP exposure
both during pregnancy &
1
st
year of life, gestational
NO2, PM2.5, PM10
exposures; all trimesters
associated
Small sample size;
correlations across
pollutants
Yousefian et
al. 2018
(Iran)
Case-control
(134 cases, 388
controls without
ASD)
Annual mean
estimates of PM10
and SO2 using LUR
models; benzene,
toluene,
ethylbenzene, p-
xylene, o-xylene, m-
xylene (BTEX), total
BTEX
ASD cases
identified from
three specialty
centers for ASD
for children born
between 2004-
2012
Maternal age, maternal
and paternal education,
cousin marriage,
maternal smoking
during pregnancy, birth
order, gestational age,
multiplicity, maternal
and paternal disease
No associations between
annual ambient pollutants
and ASD risk
Small sample size;
exposure based on
mean annual estimates
from 2010 only; lacked
information on income;
all covariate
information based on
parental
questionnaires
Hazardous Air Pollutants (Metals, Solvents, PAHs)
Windham et
al., 2006
(both
particles and
Nested case-
control (284
cases, 657
population
US EPA-modeled
estimates in second
year of life assigned
to Census tract of
DSM-IV-R criteria
based ASD
diagnosis from
DDS and Kaiser
Maternal age,
education, child race
Higher risk for ASD
associated with top
quartile of metal
exposures (collectively &
2nd year of life
exposure estimates
used instead of
exposure during
24
HAPs) controls with no
know
developmental
disability)
home (San
Francisco); used
National-Scale Air
Toxics Assessment
(NATA) 1996
Permanente for
birth year 1994
individually for cadmium,
mercury, and nickel),
chlorinated solvents, and
diesel particles
pregnancy
Kalkbrenner
et al., 2010
US EPA-modeled
estimates in North
Carolina and West
Virginia using
National-Scale Air
Toxics Assessment
(NATA) 1996

DSM-IV-R criteria
based ASD
diagnosis for birth
years 1992, 1994,
and 1996
Maternal: age,
education, smoking,
marital status, Census
tract SES variables

Child: race
Higher ASD risk
associated with quinoline
and styrene; no
association with diesel
particles
Control group not likely
represents exposure of
population; exposure
estimates 0-4 years
after birth; limited
comparability
Roberts et
al., 2013
Cohort (325
cases, 22,101
population
controls without
ASD)
US EPA-modeled
estimates for year
closest to birth year
assigned to Census
tract of birth home
(across USA); NATA
1990, 1996, 1999,
2002
Maternal report of
ASD diagnosis for
birth years 1987-
2002 (NHS II)
Maternal education &
age; Census tract
income & % college
educated; child sex,
birth year, HAP year
Higher ASD risk
associated with top
quintile of metal
exposures; dose-response
observed
Exposure estimates
ranged from 0-3 years
before or after delivery;
parental-report of ASD
Talbott et al.,
2015
Case-control
(217 cases, 224
frequency-
matched
controls; 216
cases, 4971 birth
certificate
controls)

NATA 2005 annual
average by census
tract in Pennsylvania
ASD cases based
on SCQ score and
written
documentation
(e.g. ADOS) for
birth years 2005-
2009
Maternal: age,
education, race,
smoking,

Child: birth year, sex
Higher ASD risk
associated with higher
levels of styrene and
chromium
Exposure estimates at
group level and ranged
from 0-4 years after
birth
von
Ehrenstein et
al., 2014
Cohort (768
cases, 147,954
population
controls)
Monthly average
estimates assigned
to residence during
pregnancy in a 5km
buffer around air-
toxics monitoring
stations in LA
County
Autistic disorder
diagnosis (severe
autism) at 3-5
years of age
identified through
CA DDS (birth
years 1995-2006)
Maternal: age,
race/ethnicity, nativity,
education, insurance
type, birth place, parity

Child: sex, birth year
Higher ASD risk per IQR
associated with 1,3-
butadiene, meta/para-
xylene, lead,
perchloroethylene, and
formaldeyhyde
Lacked information on
maternal smoking;
limited comparability
due to very young
population of children

25
CHAPTER 2. SEX-SPECIFIC ASSOCIATIONS OF AUTISM SPECTRUM DISODER
WITH RESIDENTIAL AIR POLLUTION EXPOSURE IN A LARGE SOUTHERN
CALIFORNIA PREGNANCY COHORT

ABSTRACT
Autism spectrum disorder (ASD) affects more boys than girls. Recent animal studies
found that early life exposure to ambient particles caused autism-like behaviors only in
males. However, there has been little study of sex-specificity of effects on ASD in
humans. We evaluated ASD risk associated with prenatal and first year of life
exposures to particulate matter less than 2.5 μm in aerodynamic diameter (PM2.5) by
child sex. This retrospective cohort study included 246,420 singleton children born in
Kaiser Permanente Southern California (KPSC) hospitals between 1999 and 2009. The
cohort was followed from birth through age five to identify 2471 ASD cases from the
electronic medical record. Ambient PM2.5 and other regional air pollution measurements
(PM less than 10 μm, ozone, nitrogen dioxide) from regulatory air monitoring stations
were spatially interpolated to estimate exposure during each trimester and first year of
life at each geocoded birth address. Hazard ratios (HRs) were estimated using Cox
regression models to adjust for birth year, KPSC medical center service areas, and
relevant maternal and child characteristics. Adjusted HRs per 6.5 μg/m
3
PM2.5 were
elevated during entire pregnancy [1.17 (95% confidence interval (CI), 1.04-1.33)]; first
trimester [1.10 (95% CI, 1.02-1.19)]; third trimester [1.08 (1.00-1.18)]; and first year of
life [1.21 (95% CI, 1.05-1.40)]. Only the first trimester association remained robust to
adjustment for other exposure windows, and was specific to boys only (HR = 1.18; 95%
26
CI, 1.08-1.27); there was no association in girls (HR = 0.90; 95% CI, 0.76-1.07;
interaction p-value 0.03). There were no statistically significant associations with other
pollutants. PM2.5-associated ASD risk was stronger in boys, consistent with findings
from recent animal studies. Further studies are needed to better understand these sex
differences that may provide a model for study of environmental effects on sexually
dimorphic neurodevelopmental outcomes.

27
INTRODUCTION
Autism spectrum disorder (ASD) prevalence increased dramatically during 2000-2012
from 0.7% to 1.7% in the United States (1). The increase can only partly be explained
by better ascertainment (2); causes of ASD, in prenatal and early postnatal life, are
likely multi-factorial (3). A particularly strong risk factor is male sex. ASD is about four
times more likely to occur in boys than in girls (1), for reasons that are not well
understood.
Recent epidemiological studies cited in a recent review have found associations
between ASD and ambient particulate matter < 2.5 μm in aerodynamic diameter (PM2.5)
during gestation and the first year of life (4). Exposure to particles causes
neuroinflammatory effects, perhaps mediated by particle-induced oxidative stress and
systemic inflammation (5-8). Recent toxicological studies showed that PM2.5 exposure
during gestation and early life causes ASD-related behaviors in male mice only,
including communication deficits, poor social interaction and novelty avoidance; and
causes neuroinflammation, including microglial cell activation (9, 10).
Two previous epidemiologic studies found stronger associations of ASD with
PM2.5 in boys than girls, but differences were not statistically significant (11, 12). In
contrast, two other recently published studies reported similar associations of ASD with
PM2.5 in boys and in girls (13, 14). Many studies have had small sample sizes and
hence were underpowered to identify differences in particle-associated risk by sex or
other factors.
Another uncertainty in epidemiologic studies to date is the vulnerable
developmental window of exposure, which has not been extensively examined with
28
mutual adjustment for multiple exposure windows. Based on a recent review, previous
studies generally found stronger ASD-PM2.5 associations during the third trimester
and/or first year of life than earlier in pregnancy (15).
Based on the evidence from animal toxicological studies, we hypothesized that
boys would have a higher risk of ASD associated with PM2.5 exposure during early
development than girls. In a large, population-based pregnancy cohort from Kaiser
Permanente Southern California (KPSC), we examined sex-specific associations of
ASD with prenatal and first year of life PM2.5 and other regional pollutant exposures.

METHODS
Study design and population
This retrospective cohort study included mother-child pairs with singleton deliveries in
KPSC hospitals between January 1, 1999, and December 31, 2009) in 14
geographically located service areas across Southern California (Fig. S2.1).
Residential addresses extracted from birth certificate records were linked by a unique
KPSC membership identifier. The primary analysis included 246,420 mother-child
pairs with children still enrolled as KPSC plan members at age one year, as previously
described (16), after excluding children with birth certificate addresses outside
Southern California (n = 636) or addresses that could not be accurately geocoded (n =
4406). Follow-up was accrued until the first occurrence of 1) clinical diagnosis of ASD;
2) last date of continuous KPSC plan membership; 3) death from any cause; or 4) age
five. Children were censored at age five to ensure the same follow-up time for the
entire cohort, regardless of birth date. Thus, the youngest children, born in 2009, were
29
followed through 2014. Both outcome and covariate data were extracted from the
KPSC electronic medical records (EMR), as previously described (16). Both the KPSC
and the University of Southern California Institutional Review Boards approved this
study.

Outcome data on ASD
KPSC neurodevelopment screening procedures included an abbreviated Checklist for
Autism in Toddlers (CHAT) (17) administered at 18- and 24-month well child visits.
Children failing the screening were referred to a pediatric developmental specialist for
further evaluation and ASD diagnosis (16, 18). The presence or absence of ASD
during follow-up was identified by ICD-9 codes 299.x or equivalent KPSC codes from
the EMR from at least two separate visits, an approach validated previously (18, 19).

Exposure assessment
Birth certificate residential addresses were geocoded using MapMarker USA Version
28.0.0.11. Exposure metrics at each geocoded address included concentrations of
regional pollutants PM2.5, PM≤10 μm in aerodynamic diameter (PM10), nitrogen dioxide
(NO2), and ozone (O3). Monthly averages for each pollutant between 1998-2009 were
calculated from data collected by the EPA regulatory air quality monitoring network.
Exposure at each address was assigned based on the monthly inverse distance-
squared weighted average from up to four closest monitoring stations within 50 km for
each pollutant. For geocoded address locations within 0.25 km of a monitor, data only
from that monitoring station were used. Although the distance-weighted approach has
30
limited accuracy in areas with sparse monitoring networks, performance is acceptable in
Southern California due to the dense geographical network of historical measurements
covering the region. In a previous Southern California study evaluating this method
using leave-one-out validation for monthly exposure prediction, the coefficients of
determination (r
2
) were 0.76, 0.73, 0.53, and 0.46 for O3, NO2, PM2.5 and PM10,
respectively, with lower R
2
values for PM attributed to the local (primary emission) dust
component that is not regional (20). Bias was less than 1 ppb or 1 µg/m
3
for the
gaseous and particulate pollutants, respectively. Each address was assigned the
monthly average of the 24-hour concentrations of PM2.5, PM10, and NO2. For O3, the
monthly average of daily maximum 8-hour concentrations was estimated. Based on the
mother’s last menstrual period, averages of the monthly concentrations were calculated
during each trimester, the entire pregnancy and the first year of life. First trimester
exposure was defined as 0-12 weeks, second trimester as 13-26 weeks, and third
trimester as 27 weeks to birth. The monthly average exposures during months
overlapping these different time periods were weighted by the number of days in each
period.

Covariates
Potential confounders chosen a priori, based on previous associations with ASD in this
cohort (16), included child sex, maternal race/ethnicity, maternal age at delivery, parity,
education, maternal history of comorbidity [≥1 diagnosis of heart, lung, kidney, or liver
disease; cancer], and median family household income in the census tract of residence.
An indicator variable was created for missing covariates (parity [n = 4,125], education [n
31
= 2,222], and household income [n = 1,850]). To account for temporal changes in ASD
incidence rate and pollution levels, we adjusted for birth year, and to account for broad
geographical characteristics associated with ASD, we adjusted for the 14 KPSC service
areas.
Additional pregnancy-related covariates potentially in PM2.5-ASD causal
pathways included maternal pre-eclampsia/eclampsia and diabetes (none, gestational
diabetes mellitus, pre-existing type 2 diabetes), preterm birth (< 37 weeks vs. ≥ 37
weeks), and congenital anomalies.

Statistical analyses
Partial Pearson correlations were calculated between pollutants within each exposure
window and across multiple exposure windows, adjusted for birth year and KPSC
service areas. Cox proportional hazards models were used to estimate the ASD hazard
ratios (HRs) associated in separate models with each pollutant exposure, adjusting for
potential confounders and for potential correlation due to multiple siblings born to the
same mother. Additionally adjusting for covariates potentially on the causal pathway did
not change estimated effects by >10%, so these variables were not included in the final
models. Restricted cubic splines identified no evidence of non-linear associations of
ASD with pollutants, so pollutants were treated as continuous variables and modeled
linearly. Because the analysis of pollutant effects on ASD were adjusted for year and
service areas, we scaled each HR to be representative of exposure contrasts both
within-service area and within-year. For each pollutant, this effect estimate was scaled
to the difference between the 95
th
and the 5
th
percentile of the distribution of deviations
32
of each child’s pregnancy exposure from the average for children born in the same
service area in the same year. Deviations were calculated as each residential pollutant
exposure value minus the within-service area, within-year mean exposure. For example,
for each of the 14 service areas and 11 years (154 in total) the average PM2.5
residential exposure and the deviations of individual PM2.5 from this average were
calculated. The 95
th
percentile (3.0 µg/m
3
) minus the 5
th
percentile (-3.5 µg/m
3
) of PM2.5
deviation distributions resulted in the within-service area and within-year scale of 6.5
µg/m
3
for PM2.5. The same procedure was used to calculate the within-service area,
within-year scales for other pollutants: 16.1 µg/m
3
for PM10, 10.4 ppb for NO2, and 15.7
ppb for O3. For each pollutant found to be associated with ASD, we conducted
sensitivity analyses mutually adjusted for multiple exposure windows, including entire
pregnancy, each trimester during pregnancy and first year of life in a single model. To
assess the hypothesis that effects of PM2.5 would be larger in boys, we tested for
exposure interactions with sex. We also stratified by sex and estimated the HRs
separately for boys and girls. For exposure windows for which significant interactions
with sex were observed, we also examined the joint effects by fitting a model with an 8-
category exposure (sex by pollutant quartiles), using the first quartile of exposure in girls
as reference.
Two-sided statistical tests were conducted at an alpha level of 0.05, and
precision was measured using 95% confidence intervals (CIs). Data analyses were
conducted using SAS 9.4 (SAS Institute, Inc, Cary, NC) and R, version 3.0.2 (64
bit).

33
RESULTS
During the follow-up period, 2471 children were diagnosed with ASD in the cohort.
Unadjusted annual ASD incidence rates (per 1,000 person-years) increased with birth
year from 1.93 in 1999 to 4.03 in 2009 (Fig. 2.1), a period during which national
prevalence rates of ASD were also increasing (2).
Children with ASD were substantially more likely to be boys (n = 2,030) than girls
(n = 441) (Table 2.1). Mothers of cases were more likely to be older, to have more than
a high school education, and to have a history of other comorbidities. Mothers of cases
were older at delivery (31.0 years; standard deviation (SD) 5.7) than mothers of children
without ASD (29.7 years, SD 5.8).
Mean levels of PM2.5, PM10, NO2, and O3 during entire pregnancy across all
years were 17.9 µg/m
3
, 38.1 µg/m
3
, 25.1 parts per billion (ppb), and 41.6 ppb,
respectively. However, mean levels of both PM2.5 and NO2 decreased across years from
1999-2009 (Fig. 2.2). PM10 levels fluctuated across time, potentially reflecting varying
levels of precipitation across the years. O3 levels remained relatively stable across
years.
Mean levels of pollutants also varied between KPSC service areas, with highest
levels of mean 1999-2009 PM2.5, PM10 and NO2 in Ontario (21.6 µg/m
3
, 49.7 µg/m
3
, 31.2
ppb, respectively), and lowest levels of PM2.5, PM10, and NO2 in San Diego (13.1 µg/m
3
,
30.7 µg/m
3
, 18.0 ppb). Highest mean levels of O3 were in Moreno Valley (52.5 ppb), and
lowest mean levels of O3 were in Downey (31.8 ppb).
An increase in ASD risk was associated with PM2.5 exposure during the entire
pregnancy (HR = 1.17 per 6.5 μg/m
3
, 95% CI: 1.04, 1.33), the first trimester (HR = 1.10,
34
95% CI: 1.02, 1.19) and the third trimester (HR = 1.08, 95% CI: 1.00, 1.18; Table 2.2) in
models adjusted for potential confounders. Increased risk of ASD was also associated
with first year of life PM2.5 exposure (HR = 1.21, 95% CI: 1.05, 1.40). There were
generally weak associations of ASD with other pollutants, none of which was statistically
significant.
Based on moderate partial correlations of PM2.5 exposures across trimesters
(R=0.33-0.54) and between trimester-specific exposures and first year of life (R=0.66-
0.67; Table S2.1), we fit a multiple exposure window model that co-adjusted all 3
trimester-specific and first year of life PM2.5 exposures. The ASD risk associated with
PM2.5 during the first trimester was still elevated, but the estimate was less precise (HR
= 1.09 per 6.5 μg/m
3
, 95% CI: 0.99-1.19; p = 0.07) after adjusting for exposures in other
time windows, which were markedly attenuated (results not shown).
The interaction between PM2.5 exposure during first trimester and child sex was
statistically significant (p = 0.03) in a model adjusted for potential confounders.
Increased ASD risk was associated with PM2.5 in boys (HR = 1.18 per 6.5 μg/m
3
, 95%
CI: 1.08, 1.27; Fig. 2.3). No associations were observed among girls (HR = 0.90 per 6.5
μg/m
3
, 95% CI: 0.76, 1.07). The adjusted sex-specific PM2.5 effect estimates were
generally larger in boys during the other exposure windows (except for the second
trimester), but none of these statistical interactions was significant. To compare the
relative effects of sex and first trimester PM2.5, we fit a model with an 8-category
exposure (sex by quartiles of PM2.5). There was almost a 4-fold increased risk of ASD
among boys in the lowest quartile of exposure as compared to girls with low exposure
(Table S2.2). For boys, there was a further 1.30-fold increased HR with an exposure-
35
response relationship across four quartiles of first trimester exposure across all years
(from <14.0 μg/m
3
, 14.0-<17.3 μg/m
3
, 17.3-<21.6 μg/m
3
, and ≥21.6 μg/m
3
, respectively).
There was little evidence for an exposure gradient in risk in girls.

DISCUSSION
In this large cohort study, increased risk of ASD was associated with maternal exposure
to PM2.5 during pregnancy, specific to the first and third trimester, and during the first
year of the child’s life. Only first trimester exposure was robust to adjustment for other
windows of PM2.5 exposure. Associations were generally stronger for boys than for girls
and this difference was statistically significant for first trimester exposure. These
findings were consistent with animal studies showing that early-life PM2.5 exposure
caused an autism-like behavioral phenotype exclusively in males (9, 10). These
hypothesis-driven results suggest that targeted investigations of early-life exposure to
airborne particles provide a novel model that could be used in future hypothesis-
generating studies examining multiple risk factors to better understand the
environmental origin of sex-related differences in the pathogenesis of ASD.
Oxidative stress and associated systemic inflammation during pregnancy that
have adverse affects on the developing fetal brain may contribute to the pathogenesis
of ASD (21). The microglial cell response to inflammatory insults during embryonic brain
development may lead to synapse dysfunction involved in the development of ASD (22).
Inflammatory insults such as PM2.5 exposure may have larger effects in males, because
they have significantly more microglia than females during gestational and early
postnatal periods (23, 24). This potentially explains the higher male-prevalence of ASD
36
diagnosed during early life, and the susceptibility of boys to PM2.5 (23). Other potential
contributing explanations include greater fetal testosterone during male brain
development that leads to more activated microglia compared to females and therefore
to greater vulnerability to PM or other insults that increase risk of ASD (25, 26).
Newborn males also have lower glutathione and sulfate levels, which are involved in
detoxifying oxidant environmental insults such as PM2.5, compared to newborn females
(27). In male mice, an autism-like behavioral phenotype was caused by PM2.5 exposure
during early postnatal life (equivalent to the human third trimester) (9, 10), in contrast to
our findings of more robust ASD associations with first trimester exposure. Given these
male-specific results across developmental windows, experimental studies designed to
specify more precisely the window(s) of vulnerability to these sexually dimorphic effects
are needed.
A few previous epidemiological studies have assessed sex differences in PM2.5-
ASD susceptibility. In the Nurses Health Study, stronger ASD associations with PM2.5
during pregnancy among boys (odds ratio (OR) 1.73 per 4.42 μg/m
3
, 95% CI: 1.29,
2.31), than among girls (OR 1.12, 95% CI: 0.59, 2.12; interaction p-value = 0.17);
however, the power of the study was limited by a small number of ASD cases in girls (n
= 23) (11). A Danish study with the largest number of ASD cases in both boys and girls
to date (n = 11,853; n = 3,534, respectively) reported that the association of PM2.5
exposure during 9 months after birth with ASD was slightly stronger among boys (OR
1.07 per 3.61 μg/m
3
, 95% CI: 1.01, 1.13) than among girls (OR 1.02 per 3.61 μg/m
3
,
95% CI: 0.92, 1.12) (12). In contrast, a study from Canada reported similar PM2.5
pregnancy exposure associations with ASD among boys (OR 1.04 per 1.5 μg/m
3
, 95%
37
CI: 0.98, 1.10) and among girls (OR 1.03 per 1.5 μg/m
3
, 95% CI: 0.90, 1.18) (13). This
study had a large number of ASD cases in girls (n = 216). A recent study in Ohio also
reported similar patterns of PM2.5 associations during pregnancy and early life when
restricted to boys and in the entire study population; this study did not report ORs
among girls (ASD cases = 78) (14). Our study included a large number of ASD cases in
girls (n = 441) and had 80% power to detect an effect estimate as small as 1.175 in girls
under a two-sided hypothesis test with α = 0.05. Because ASD is a rare disease, a large
sample size is required to assess interactions of sex with modifiable ASD risk factors
such as PM2.5.
Large population studies provide an opportunity to study the role of multifactorial
causal pathways that might identify substantial attributable proportions of ASD. It is
unlikely that PM2.5 by itself contributed substantially to the epidemic of ASD reflected in
the KPSC temporal trends, because ambient concentrations of PM2.5 were decreasing
during this same period (Figure 2). However, in the context of synergistic effects with
another risk factor such as maternal obesity (28) or gestational diabetes mellitus (29),
which were increasing during this period, PM2.5 could contribute to the epidemic even in
the context of stable or decreasing regional PM2.5 levels over the period of increasing
ASD rates.
Identifying the window of vulnerability to PM2.5 during pregnancy can provide
clues to the pathogenesis of ASD, based on brain development during this window, with
potential implications for prevention. We observed associations of increased ASD risk
with several exposure windows in early life, but the first trimester was most robust to co-
adjustment for other exposure windows. Previous epidemiological studies have
38
observed ASD-PM2.5 associations mainly during the third trimester and first year of life
(11, 12, 30-32). Three recently published studies mutually adjusted for PM2.5 exposure
across trimesters and/or early life (11-13). Also, one recently published study that found
null associations of PM2.5 with ASD mutually adjusted for pairwise combinations of
trimester-specific and early life exposures, (14). The date of last menstrual period
determined from self-report at first prenatal visit or from gestational age determined by
ultrasound in the KPSC EMR likely more accurately specified the window of exposure
than most previous studies that calculated trimesters based on self-reported gestational
age at birth and child’s birth date (11, 12, 14, 30). Another study that found null ASD
associations with trimester-specific PM2.5 exposure reported an increased risk of ASD
with cumulative exposure from 3 months before pregnancy to the second year of life
(33).
A recent meta-analysis of ASD-PM2.5 associations during pregnancy published in
observational studies reported a pooled odds ratio (OR) of 2.32 per 10 μg/m
3
(4). We
observed substantially smaller estimates of risk. For example, the first trimester HR 1.10
per 6.5 μg/m
3
was equivalent to HR 1.16 per 10 μg/m
3
, and the first trimester HR 1.18
per 6.5 μg/m
3
for males was equivalent to 1.29 per 10 μg/m
3
. However, the pooled
estimate was based on only 3 studies, which had substantial heterogeneity, with a
range of ORs from 1.15 to 2.77 per 10 μg/m
3
, that could not be explained by differences
in study design and methods. Although our ASD-PM2.5 associations were in the same
direction as in these other studies, unexplained differences in magnitude of the
associations could potentially be resolved with additional studies. Additionally,
39
comparing results across studies with different designs and measures of association
remains a challenge.
We found no associations of ASD with other pollutants. Our results are
consistent with the human epidemiological and animal toxicological evidence that has
largely converged on the neurotoxicological effects of PM2.5, although there is some
uncertainty to the epidemiological associations (4, 9, 10). A cohort study from Europe
reported no PM2.5 association with autistic behavioral traits (34). Various other studies
have examined associations with other pollutants (PM10, O3, NO2, and sulfur dioxide),
with indicators of exposure to the near-roadway mixture, and with hazardous air
pollutants, including gaseous, traffic-related air pollutants, metals and other volatile
organic compounds (4, 12, 13, 32, 35, 36). However, associations of ASD with these
pollutants have not been consistent.
This study has several strengths. It is one of the few cohort studies to examine
risk of ASD associated with air pollution. Selection bias was unlikely to have influenced
the results, because the cohort did not self-select and also had a high annual average
95% retention through age 5 after cohort enrollment (16). The large study sample
provided statistical power to assess interaction of PM2.5 with sex. Unlike many health
systems that have only recently adopted an EMR, the KPSC EMR system has
developed since its implementation in the early 1990’s. Incident ASD ascertainment
occurred in a clinical setting using standardized diagnostic algorithms developed
through the EMR (19). The analysis controlled for various individual-level covariates that
were available through the EMR, such as gestational diabetes, that are difficult to obtain
without standardized procedures in a single healthcare system. The KPSC membership
40
comprised approximately 16% of the census reference population, which represented
all age, racial/ethnic, and socioeconomic groups, and results are generalizable to the
population of working Southern California families (37). Because Southern California
has large pollutant exposure gradients representative of ranges and extremes across
the U.S., the results are broadly relevant to the U.S. and other countries with similar
exposures. There were some limitations. Although there were large exposure gradients
in the cohort, exposure concentrations estimated at the birth address were used as a
proxy for personal exposure. Residential mobility during pregnancy and the first year of
life, and time of mother or child away from home would have resulted in exposure
measurement error. During pregnancy 9-34% of mothers move, resulting in
corresponding changes in exposure that may not be recalled well (38-40). If the effect of
this bias were non-differential with respect to the outcome, then the true effect of
exposure could likely have been larger than we observed (41).

CONCLUSIONS
We conclude that boys may be at increased risk of ASD induced by maternal PM2.5
exposure during the first trimester. This study provides potential clues to the reasons for
the sex-specificity of ASD, effects that were predictable based on animal studies.
Further toxicological study is needed to better understand the biological basis and
timing of susceptibility during neurodevelopment and of sex differences in
neurobehavioral effects of particle exposures.

41
REFERENCES

1. Baio J, Wiggins L, Christensen DL, et al. Prevalence of Autism Spectrum
Disorder Among Children Aged 8 Years - Autism and Developmental Disabilities
Monitoring Network, 11 Sites, United States, 2014. MMWR Surveill Summ
2018;67(6):1-23.
2. Baio J. Prevalence of autism spectrum disorder among children aged 8 years -
Autism and Developmental Disabilities Monitoring Network, 11 sites, United
States, 2010. MMWR Surveill Summ 2014;63(2):1-21.
3. Lyall K, Schmidt RJ, Hertz-Picciotto I. Maternal lifestyle and environmental risk
factors for autism spectrum disorders. Int J Epidemiol 2014;43(2):443-64.
4. Lam J, Sutton P, Kalkbrenner A, et al. A Systematic Review and Meta-Analysis
of Multiple Airborne Pollutants and Autism Spectrum Disorder. PLoS One
2016;11(9):e0161851.
5. Campbell A, Araujo JA, Li H, et al. Particulate matter induced enhancement of
inflammatory markers in the brains of apolipoprotein E knockout mice. J Nanosci
Nanotechnol 2009;9(8):5099-104.
6. Campbell A, Daher N, Solaimani P, et al. Human brain derived cells respond in a
type-specific manner after exposure to urban particulate matter (PM). Toxicol In
Vitro 2014;28(7):1290-5.
7. Rossignol DA, Frye RE. Evidence linking oxidative stress, mitochondrial
dysfunction, and inflammation in the brain of individuals with autism. Front
Physiol 2014;5:150.
42
8. Block ML, Calderón-Garcidueñas L. Air pollution: mechanisms of
neuroinflammation and CNS disease. Trends Neurosci 2009;32(9):506-16.
9. Li K, Li L, Cui B, et al. Early postnatal exposure to airborne fine particulate matter
induces autism-like phenotypes in male rats. Toxicol Sci 2017.
10. Church JS, Tijerina PB, Emerson FJ, et al. Perinatal exposure to concentrated
ambient particulates results in autism-like behavioral deficits in adult mice.
Neurotoxicology 2017.
11. Raz R, Roberts AL, Lyall K, et al. 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. Environ Health Perspect
2015;123(3):264-70.
12. Ritz B, Liew Z, Yan Q, et al. Air pollution and autism in Denmark. Environmental
Epidemiology 2018;2(e028).
13. Pagalan L, Bickford C, Weikum W, et al. Association of Prenatal Exposure to Air
Pollution With Autism Spectrum Disorder. JAMA Pediatr 2018.
14. Kaufman JA, Wright JM, Rice G, et al. Ambient ozone and fine particulate matter
exposures and autism spectrum disorder in metropolitan Cincinnati, Ohio.
Environ Res 2019;171:218-27.
15. Flores-Pajot MC, Ofner M, Do MT, et al. Childhood autism spectrum disorders
and exposure to nitrogen dioxide, and particulate matter air pollution: A review
and meta-analysis. Environ Res 2016;151:763-76.
16. Xiang AH, Wang X, Martinez MP, et al. Association of maternal diabetes with
autism in offspring. JAMA 2015;313(14):1425-34.
43
17. Baron-Cohen S, Wheelwright S, Cox A, et al. Early identification of autism by the
CHecklist for Autism in Toddlers (CHAT). J R Soc Med 2000;93(10):521-5.
18. Xiang AH, Wang X, Martinez MP, et al. Maternal Type 1 Diabetes and Risk of
Autism in Offspring. JAMA 2018;320(1):89-91.
19. Coleman KJ, Lutsky MA, Yau V, et al. Validation of Autism Spectrum Disorder
Diagnoses in Large Healthcare Systems with Electronic Medical Records. J
Autism Dev Disord 2015;45(7):1989-96.
20. Eckel SP, Cockburn M, Shu YH, et al. Air pollution affects lung cancer survival.
Thorax 2016;71(10):891-8.
21. Block ML, Elder A, Auten RL, et al. The outdoor air pollution and brain health
workshop. Neurotoxicology 2012;33(5):972-84.
22. Hammond TR, Robinton D, Stevens B. Microglia and the Brain: Complementary
Partners in Development and Disease. Annu Rev Cell Dev Biol 2018.
23. Hanamsagar R, Bilbo SD. Sex differences in neurodevelopmental and
neurodegenerative disorders: Focus on microglial function and
neuroinflammation during development. J Steroid Biochem Mol Biol
2016;160:127-33.
24. Bolton JL, Marinero S, Hassanzadeh T, et al. Gestational Exposure to Air
Pollution Alters Cortical Volume, Microglial Morphology, and Microglia-Neuron
Interactions in a Sex-Specific Manner. Front Synaptic Neurosci 2017;9:10.
25. McCarthy MM. Sex differences in the developing brain as a source of inherent
risk. Dialogues in Clinical Neuroscience 2016;18(4):361-72.
44
26. Baron-Cohen S, Lombardo MV, Auyeung B, et al. Why are autism spectrum
conditions more prevalent in males? PLoS Biol 2011;9(6):e1001081.
27. Kern JKG, D. A.; Homme, K. G.; King, P. G.; Bjørklund, G.; Chirumbolo, S.;
Geier, M. R. Developmental neurotoxicants and the vulnerable male brain: a
systematic review of suspected neurotoxicants that disproportionally affect
males. Acta Neurobiol Exp (Wars) 2017;77(4):269-96.
28. Flegal KM, Carroll MD, Kit BK, et al. Prevalence of obesity and trends in the
distribution of body mass index among US adults, 1999-2010. JAMA
2012;307(5):491-7.
29. Bardenheier BH, Imperatore G, Gilboa SM, et al. Trends in Gestational Diabetes
Among Hospital Deliveries in 19 U.S. States, 2000-2010. Am J Prev Med
2015;49(1):12-9.
30. Becerra TA, Wilhelm M, Olsen J, et al. Ambient air pollution and autism in Los
Angeles county, California. Environ Health Perspect 2013;121(3):380-6.
31. Volk HE, Lurmann F, Penfold B, et al. Traffic-related air pollution, particulate
matter, and autism. JAMA Psychiatry 2013;70(1):71-7.
32. Chen G, Jin Z, Li S, et al. Early life exposure to particulate matter air pollution
(PM. Environ Int 2018;121(Pt 2):1121-7.
33. Talbott EO, Arena VC, Rager JR, et al. Fine particulate matter and the risk of
autism spectrum disorder. Environ Res 2015;140:414-20.
34. Guxens M, Ghassabian A, Gong T, et al. Air Pollution Exposure during
Pregnancy and Childhood Autistic Traits in Four European Population-Based
Cohort Studies: The ESCAPE Project. Environ Health Perspect 2015.
45
35. Yousefian F, Mahvi AH, Yunesian M, et al. Long-term exposure to ambient air
pollution and autism spectrum disorder in children: A case-control study in
Tehran, Iran. Sci Total Environ 2018;643:1216-22.
36. Raz R, Levine H, Pinto O, et al. Traffic Related Air Pollution and Autism
Spectrum Disorder: A Population Based Nested Case-Control Study in Israel. Am
J Epidemiol 2017.
37. Koebnick C, Langer-Gould AM, Gould MK, et al. Sociodemographic
characteristics of members of a large, integrated health care system: comparison
with US Census Bureau data. Perm J 2012;16(3):37-41.
38. Chen Z, Salam MT, Toledo-Corral C, et al. Ambient Air Pollutants Have Adverse
Effects on Insulin and Glucose Homeostasis in Mexican Americans. Diabetes
Care 2016;39(4):547-54.
39. Chen L, Bell EM, Caton AR, et al. Residential mobility during pregnancy and the
potential for ambient air pollution exposure misclassification. Environ Res
2010;110(2):162-8.
40. Bell ML, Belanger K. Review of research on residential mobility during
pregnancy: consequences for assessment of prenatal environmental exposures.
J Expo Sci Environ Epidemiol 2012;22(5):429-38.
41. Rothman K, Greenland S, Lash TL. Modern Epidemiology, 3rd Edition.
Philadelphia, PA: Lippincott Williams & Wilkins.; 2008.

46
Figure 2.1. Crude incidence rate of ASD by birth year

Note: based on 2,471 children diagnosed with ASD during follow-up through age 5 in the cohort of 246,420 children.

1.93
2.51
2.48
2.93
3.30
3.60
3.06
3.43
3.28
3.76
4.03
0.00
0.50
1.00
1.50
2.00
2.50
3.00
3.50
4.00
4.50
1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
Incidence Rate of ASD per 1,000 Person-Years
Birth Year
47
Table 2.1. Characteristics of the cohort by ASD status

ASD No ASD
Characteristics n (column %)
a,b
n (column %)
a,b

Child
Male 2030 (82.2) 124,278 (50.9)
Female 441 (17.8) 119,671 (49.1)
Maternal

Parity

0 1123 (45.5) 95,841 (39.3)
1 802 (32.5) 77,692 (31.9)
≥2 511 (20.7) 66,326 (27.2)
Unknown 35 (1.4) 4090 (1.7)
Education

High school or lower 811 (32.8) 100,959 (41.4)
Some college 761 (30.8) 67,968 (27.9)
College graduate or higher 888 (35.9) 72,811 (29.9)
Unknown 11 (0.5) 2211 (0.9)
Household annual income
c

<$30,000 223 (9.0) 19,846 (8.1)
$30,000-$49,999 845 (34.2) 81,793 (33.5)
$50,000-$69,999 810 (32.8) 78,116 (32.0)
$70,000-$89,999 365 (14.8) 39,393 (16.2)
≥$90,000 228 (9.2) 24,801 (10.2)
Race/ethnicity

Non-Hispanic white 569 (23.0) 62,205 (25.5)
Non-Hispanic black 266 (10.8) 23,589 (9.7)
Hispanic 1216 (49.2) 124,907 (51.2)
Asian/Pacific Islander 374 (15.1) 29,400 (12.1)
Other 46 (1.9) 3848 (1.6)
History of comorbidity
d
301 (12.2) 21,753 (8.9)
a
Based on 2471 ASD cases and 243,949 without ASD
b
All characteristics were significantly different based on the χ
2
–squared test for proportions and analysis
of variance for means (p<0.001)
c
Based on census tract median
d
≥1 diagnosis of heart, lung, kidney, or liver disease; cancer in mothers
48
Figure 2.2. Distribution of pollutant concentrations during pregnancy across birth year 1999-2009

49
Table 2.2. Adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) for ASD associated with each pollutant birth
address exposure
PM2.5 PM10 NO2 O3
Exposure window HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI)
Entire pregnancy 1.17 (1.04-1.33) 1.01 (0.89-1.15) 1.05 (0.91-1.20) 1.10 (0.95-1.26)
First trimester 1.10 (1.02-1.19) 1.00 (0.92-1.10) 1.03 (0.95-1.11) 0.97 (0.91-1.02)
Second trimester 1.06 (0.97-1.14) 1.02 (0.93-1.12) 1.01 (0.93-1.09) 1.03 (0.97-1.10)
Third trimester 1.08 (1.00-1.18) 1.00 (0.91-1.10) 1.02 (0.94-1.11) 1.05 (0.99-1.11)
First year of life 1.21 (1.05-1.40) 1.06 (0.92-1.22) 1.12 (0.97-1.30) 0.94 (0.80-1.11)
Separate models were estimated for each time window, adjusted for birth year, KPSC medical center service areas, maternal age, parity, maternal
race/ethnicity, maternal education, census tract median household income, maternal history of comorbidities before pregnancy (≥1 diagnosis of
heart, lung, kidney, liver disease or cancer), child sex, and family specified as a random effect. Hazard ratios were scaled per 6.5 μg/m
3
PM2.5; per
16.1 μg/m
3
PM10; per 10.4 ppb NO2; and per 15.7 ppb O3.
ASD = autism spectrum disorder; PM2.5 = particulate matter <2.5 μm in aerodynamic diameter; PM10 particulate matter <10 μm, NO2 = nitrogen
dioxide; O3 = ozone)

50
Figure 2.3. Child sex-specific hazard ratios (HRs) and 95% confidence intervals (CIs) for ASD per 6.5 μg/m3 PM2.5 during
each exposure period

Note: 441 ASD cases among 120,112 girls; 2,030 ASD cases among 126,308 boys. Separate models were estimated for each time window, and
estimates were adjusted for birth year, KPSC medical center service areas, maternal age, parity, maternal race/ethnicity, maternal education,
census tract median household income, maternal history of comorbidities before pregnancy, and family specified as a random effect. Circles
represent the HRs and whiskers represent the 95% CIs. *P for interaction with sex < 0.05.

51
Table S2.1. Pearson partial correlations of PM2.5 across exposure windows

Entire
pregnancy
First
trimester
Second
trimester
Third
trimester
First year of
life
Entire
pregnancy
1.00 0.78 0.87 0.75 0.83
First
trimester

1.00 0.54 0.33 0.66
Second
trimester

1.00 0.53 0.67
Third
trimester

1.00 0.67
First year of
life
1.00
Note: All correlations were adjusted for birth year and KPSC medical center service areas and were statistically significant (p < 0.05).

52
Table S2.2. Adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) for ASD joint associations with child sex and
quartiles of first trimester PM2.5 exposure
Girls Boys
First trimester PM2.5 exposure HR (95% CI)
a
HR (95% CI)
a

1
st
quartile (<14.0 μg/m
3
) Reference 3.85 (3.14-4.71)
2
nd
quartile (14.0-<17.3 μg/m
3
) 0.93 (0.72-1.21) 4.27 (3.47-5.24)
3
rd
quartile (17.3-<21.6 μg/m
3
) 0.86 (0.65-1.12) 4.57 (3.67-5.67)
4
th
quartile (≥21.6 μg/m
3
) 0.85 (0.62-1.16) 5.00 (3.96-6.32)
a
Adjusted for birth year, KPSC medical center service areas, maternal age, parity, maternal race/ethnicity, maternal education, household income,
maternal history of comorbidities before pregnancy, and family specified as a random effect

53
Figure S2.1. Kaiser Permanente Southern California (KPSC) medical center service areas

Note: Black dots indicate air quality monitoring stations that had data for one or more pollutants during the study period. Service areas contributing
to the analysis are shown in the legend. KPSC covers Imperial, Kern, Los Angeles, Orange, Riverside, San Bernardino, San Diego, San Luis
Obispo, Santa Barbara, and Ventura counties. Of the18 geographically located KPSC medical center service areas across Southern California, 4
were excluded (Kern, Antelope Valley, West Ventura, and Coachella Valley) because there were few cases of ASD.
54
CHAPTER 3. ASSOCIATIONS OF GESTATIONAL DIABETES MELLITUS WITH
RESIDENTIAL AIR POLLUTION EXPOSURE IN A LARGE SOUTHERN CALIFORNIA
PREGNANCY COHORT

ABSTRACT
Studies of effects of air pollution on gestational diabetes mellitus (GDM) have not been
consistent, and there has been little investigation of effects of exposure preceding
pregnancy. In previous studies, the temporal relationship between exposure and GDM
onset has been difficult to establish. Data from pregnancies between 1999 and 2009
were available in a population-based integrated healthcare system (n=239,574).
Exposures to nitrogen dioxide (NO2), ambient particulate matter (PM) ≤2.5 μm in
aerodynamic diameter (PM2.5) and ≤10 μm (PM10), and ozone (O3), were interpolated
from regulatory air monitoring stations to each geocoded birth address during 12 weeks
before conception and the first trimester of pregnancy. Odds ratios (ORs) of GDM
diagnosed after 12 weeks gestation in association with pollutant exposure were
estimated using generalized estimating equation models adjusted for birth year, medical
center service areas, maternal age, and race/ethnicity. In single-pollutant models,
preconception NO2 was associated with increased risk of GDM (OR = 1.11 per 10.4
ppb, 95% confidence interval [CI]: 1.08, 1.14). First trimester NO2 was associated with
GDM, but this was not statistically significant (RR = 1.02 per 10.4 ppb, 95% CI: 1.00,
1.05, P=0.09). Preconception NO2 associations were robust in models adjusted for first
trimester NO2 with another co-pollutant from both exposure windows. In single-pollutant
models, preconception PM2.5 and PM10 associations were associated with increased risk

55
of GDM (RR = 1.04 per 6.5 μg/m
3
,

95% CI: 1.01, 1.06; RR = 1.03 per 16.1 µg/m
3
, 95%
CI: 1.00, 1.06, P=0.04, respectively), but these effect estimates were not robust to
adjustment for other pollutants. In single-pollutant models, preconception and first
trimester O3 were associated with reduced risk of GDM (RR = 0.93 per 15.7 ppb, 95%
CI: 0.91, 0.95; RR = 0.95 per 15.7 ppb, 95% CI: 0.93, 0.97), associations that were
robust to adjustment for co-pollutants. Maternal exposure to NO2 during the
preconception trimester may increase risk of GDM.

56
INTRODUCTION
Rates of diabetes, including gestational diabetes mellitus (GDM), among women
of reproductive age increased during years 2000-2010 from 3.71 to 5.77 per 100
deliveries in the United States (1). Pregnancy is a vulnerable period when women are
naturally in an insulin-resistant state (2), but emerging evidence indicates the period
before conception may also be a critical time-window during which to reduce harmful
exposures (3). Preconception health care focused on improving lifestyle choices (e.g.
diet, folic acid supplement, weight loss) and reducing adverse risk factors before
pregnancy among women of reproductive age has been shown to prevent pregnancy
and delivery complications, such as GDM, and adverse birth outcomes (4). There has
been limited study of effects of other environmental exposures during the preconception
period on risk of GDM.
Recent epidemiological studies have found associations of air pollution with type
2 diabetes, insulin resistance and glucose homeostasis and diabetes-related mortality in
adults (5-7). There has been limited study of effects of maternal exposure to air pollution
on the development of GDM, and results have not been consistent (8-18). A few studies
have found that increased risk of GDM was associated with nitrogen dioxide (NO2) (10,
19), particulate matter ≤2.5 μm in aerodynamic diameter (PM2.5) (9, 12, 14, 15), and PM
<10 μm (PM10) (15). Key uncertainties in previous studies include lack of accurate
information on the date of GDM diagnosis; thus, it was not possible to determine if
exposure preceded the development of GDM, and some pre-gestational diabetes may
have been misclassified as GDM. Moreover, previous studies have generally not
mutually adjusted exposure associations for other pollutants or for different windows of

57
exposure (for example preconception and first trimester exposures). To date, only two
studies have examined effects of preconception exposures (10, 12). These two studies
assigned exposure to either the delivery hospital region or to city/township, rather than
to the mother’s residence.
To address gaps in our understanding of effects of air pollution on GDM, we
investigated the association between maternal residential exposure to regional air
pollution and the development of GDM during pregnancy in a large population-based
pregnancy cohort in an integrated healthcare system. We defined GDM diagnosis and
gestational age of GDM diagnosis based on laboratory data and corresponding date of
diagnosis recorded in a comprehensive electronic medical record system (EMR). We
were able to establish clear temporal relationships by assessing pollutant exposures
during preconception and first trimester and examining associations with subsequent
development of GDM at or after 13 weeks of gestation. Independent associations of
each air pollutant studied and the exposure window of effect (preconception or first
trimester) were assessed through multivariate adjusted analyses.

METHODS
Study design and population
This population-based retrospective cohort study included women who gave
birth to singleton children between January 1, 1999, and December 31, 2009 at
Kaiser Permanente Southern California (KPSC) hospitals. KPSC covers Imperial,
Kern, Los Angeles, Orange, Riverside, San Bernardino, San Diego, San Luis

58
Obispo, Santa Barbara, and Ventura counties, with 14 medical center service areas
(Figure S3.1). Women with residential addresses at the time of child’s birth outside
Southern California (n = 636) or addresses that could not be accurately geocoded (n
= 4,406) were excluded. Residential addresses at the time of birth were extracted
from birth certificate records, which were linked by a unique KPSC membership
identifier. The primary exposure windows included 12-weeks preconception and first
trimester. To assure that exposure occurred before the onset of disease, women
with pre-existing diabetes (n = 4,093) or a GDM diagnosis before 13 weeks’
gestation (n = 2,761) were excluded, leaving a total of 239,574 women included in
the primary analyses. Both outcome and covariate data were extracted from the
KPSC EMR, as previously described (20). This study was approved by the
Institutional Review Boards of University of Southern California and KPSC.

Outcome data on GDM
KPSC follows the American College of Obstetricians and Gynecologists
guidelines for GDM screening (21). Diagnosis of GDM was based on laboratory
values confirming a plasma glucose level of 200mg/dL or higher on the glucose
challenge test or at least 2 plasma glucose values meeting or exceeding the
following values on the 100-g or 75-g oral glucose tolerance test: fasting, 95 mg/dL;
1 hour, 180 mg/dL; 2 hours, 155 mg/dL; and 3 hours, 140 mg/dL, as previously
described (2, 20). Gestational age at GDM diagnosis was calculated using the date
of the first glucose test result that met the GDM diagnosis criteria, date of delivery,
and gestational age at delivery available in the electronic medical record.

59

Exposure assessment
Ambient exposures to regional air pollutant exposures were estimated at
residential addresses recorded on birth certificates. These birth addresses were
geocoded using MapMarker USA Version 28.0.0.11.
Exposures at each geocoded address included PM2.5, PM10, NO2, and ozone
(O3). Monthly averages for each pollutant between 1998 and 2009 were obtained from
data compiled from the EPA regional air quality monitoring network across Southern
California. To estimate the exposure at the residential location, we used the inverse
distance-weighted monthly average from four closest monitoring stations within 50 km,
except for geocoded locations within 0.25 km of a monitor, for which only data from the
nearest monitoring station were used. Although the distance-weighted approach has
limited accuracy in areas with sparse monitoring networks, performance is acceptable in
Southern California due to the dense geographical network of historical measurements
covering the region. In a previous Southern California study evaluating this method
using leave-one-out validation for monthly exposure prediction, the coefficients of
determination (R
2
) were 0.76, 0.73, 0.53, and 0.46 for O3, NO2, PM2.5 and PM10,
respectively, with lower R
2
values for PM attributed to the local (primary emission) dust
component that is not regional (22). Bias was less than 1 ppb or 1 µg/m
3
for the
gaseous and particulate pollutants, respectively. Each address was assigned the
monthly average of the 24-hour concentrations of PM2.5, PM10, and NO2. For O3, the
monthly average of daily maximum 8-hour concentrations was estimated. Averages of
the monthly concentrations within 12 weeks before conception and first trimester of

60
pregnancy (0-12 weeks) were then aggregated from these monthly estimates, with each
specific time window determined based on the last menstrual period date. For months
overlapping different exposure windows (e.g. preconception and first trimester), the
exposure was assigned proportional to the number of days in each window.

Covariates
Potential confounding variables chosen a priori, based on previous associations
with GDM (5, 6), included birth year, maternal age at delivery (continuous), and self-
reported race/ethnicity. To control for spatial confounding, KPSC medical center service
areas were adjusted as proxies for geographical characteristics associated with GDM.
Other covariates available in the EMR included parity, education [high school or lower,
some college, college graduate or higher] and median family household income in the
census tract of residence. Additional pregnancy-related covariates that may be in the
causal pathways included maternal pre-pregnancy body mass index (BMI) that was
categorized as underweight, normal, overweight and obese. This covariate was
routinely recorded in the EMR starting in late 2006 and was available for 72,044 of the
239,574 total pregnancies. An indicator variable was created for each missing value for
each covariate (parity [n = 3,956], education [n = 2,131], household income [n = 1,819],
except BMI.

Statistical analyses
Maternal characteristics were compared between women with GDM diagnosed at
≥ 13 weeks gestation (n = 18,244) and women without GDM (n = 221,330). Partial

61
Pearson correlation coefficients were calculated between regional pollutant exposures
during 12 weeks before conception and the first trimester, adjusting for birth year and
KPSC medical center service areas. Restricted cubic splines identified no evidence of
non-linear associations of GDM with pollutants. Therefore, each pollutant was treated
as a continuous variable and modeled linearly. Generalized Estimating Equations
models with the logit function and binomial distribution were used to estimate the odds
ratios (ORs) for GDM associated with each pollutant exposure, adjusting for potential
confounders. To account for within-cluster correlation for women with more than one
singleton pregnancy during the study period, we used an exchangeable covariance
structure.
Potential confounding due to temporal changes in rates of GDM and of pollution
levels was addressed by adding calendar birth year as a continuous covariate. We
controlled for broad geographic characteristics associated with GDM by adjusting for 14
KPSC medical center service areas. Because the analysis of estimated GDM effects of
each pollutant was adjusted for year and for service areas, we scaled each RR to be
representative of exposure contrasts both within-service area and within-year. For each
pollutant, this effect estimate was scaled to the difference between the 95
th
and the 5
th

percentile of the distribution of deviations of each child’s pregnancy exposure from the
average for children born in the same service area in the same year. Deviations were
calculated as each residential pollutant exposure value minus the within-service area,
within-year mean exposure. For example, for each of the 14 service areas and 11 years
(154 in total) the average PM2.5 residential exposure and the deviations of individual
PM2.5 from this average were calculated. The 95
th
percentile (3.0 µg/m
3
) minus the 5
th

62
percentile (-3.5 µg/m
3
) of PM2.5 deviation distributions resulted in the within-service area
and within-year scale of 6.5 µg/m
3
for PM2.5. The same procedure was used to calculate
the within-service area, within-year scales for other pollutants: 16.1 µg/m
3
for PM10, 10.4
ppb for NO2, and 15.7 ppb for O3.
Additionally adjusting for maternal education, median family household income,
parity, and season did not change estimates by more than 10%, and thus were not
included as confounders in the final models. Maternal pre-pregnancy BMI (that may also
be on the causal pathway) was also not included in the final models because this
covariate was only recorded for the subset of pregnancies after 2006. In a sensitivity
analysis restricted to the subset of pregnancies since 2006 with data for pre-pregnancy
BMI, adjustment for BMI did not change the estimates of effect by >10%. In order to
assess potential confounding by exposure during pre-exposure and first trimester,
models mutually adjusting for both preconception and first trimester exposures of each
pollutant and in addition for each of the co-pollutants in both exposure windows were
fitted.
Two-sided statistical tests were conducted at the alpha level of 0.05, and
precision was measured using 95% confidence intervals (CIs). Data analyses were
conducted using SAS 9.4 (SAS Institute, Inc, Cary, NC).

RESULTS
In this study, 18,244 (7.7%) women had a GDM diagnosis ≥13 weeks’ gestation,
and 221,330 women did not have GDM. Women with GDM were more likely to be
multiparous; to be Asian/Pacific Islander; and to be overweight or obese before

63
pregnancy compared to women without GDM (Table 3.1). Women with GDM were older
at delivery (32.4 years; standard deviation (SD) 5.4) than women without GDM (29.4
years; SD 5.8). Proportions of maternal education and household income levels were
similar among women with and without GDM.
Mean levels of both PM2.5 and NO2 during preconception and during the first
trimester decreased across birth years from 1999 to 2009 (Figure S3.2). PM10 exposure
estimates fluctuated across time while mean O3 levels remained relatively stable across
years. Mean concentrations of pollutants also varied between KPSC medical center
service areas, with highest mean PM2.5 and PM10 levels across years in Ontario (21.9
µg/m
3
, 50.4 µg/m
3
), and lowest PM2.5 and PM10 levels in San Diego (13.4 µg/m
3
, 31.0
µg/m
3
); results not shown. Highest mean levels of NO2 were in LA (32.3 ppb), and
lowest in Irvine (17.2 ppb). Highest mean O3 levels were in Moreno Valley (51.8 ppb),
and lowest mean levels were in Downey (31.4 ppb). Rates of GDM also varied by birth
year from 7.6% to 10.7%; and by KPSC service areas, with the lowest rate of GDM in
Moreno Valley (2.0%) and the highest rate of GDM in San Diego (14.6%).
Adjusting for year and KPSC service areas, the partial correlations between
pollutants were positive, except for O3, which was negatively correlated with both NO2
and PM2.5 (Table 3.2). The partial correlations across preconception and first trimester
for each pollutant ranged from low to moderately positive (R = 0.54-0.58), with O3
having the smallest positive correlation (R = 0.17) between exposure windows.
Maternal preconception exposure to NO2 was associated with increased risk of
GDM diagnosed >13 weeks (OR = 1.11 per 10.4 ppb, 95% CI: 1.08, 1.14). (See Table
3.3). The effect estimate for first trimester exposure was substantially weaker (OR =

64
1.02 per 10.4 ppb, 95% CI: 1.00, 1.05) and not statistically significant (p=0.09). We fitted
models that mutually adjusted for preconception and first trimester NO2 exposure and
for a co-pollutant in both exposure windows (“mutually adjusted” in Table 3.3). The
preconception NO2 effect estimate was robust to adjustment for PM2.5 or for PM10. For
example, the OR was 1.10 per 10.4 ppb, 95% CI: 1.06, 1.14), after adjustment for NO2
in the first trimester and for PM2.5 preconception and first trimester exposure. Mutually
adjusting for O3 attenuated this effect estimate (OR = 1.04 per 10.4 ppb, 95% CI: 1.01,
1.08). First trimester NO2 association remained null in models mutually adjusted for
preconception NO2 and a co-pollutant in both exposure windows. Preconception PM2.5
exposure was associated with GDM (OR = 1.04 per 6.5 µg/m
3
, 95% CI: 1.01, 1.06) but
the effect was markedly reduced by adjustment for first trimester PM2.5 and either NO2
or O3 in both exposure windows. First trimester PM2.5 exposure was associated with a
reduction in risk for GDM (OR = 0.98 per 6.5 µg/m
3
, 95% CI: 0.95, 1.00; p=0.07), and
this association was statistically significant in models including preconception PM2.5
exposure and either NO2 or PM10, but not O3, in both exposure periods. The
preconception PM10 exposure association (OR = 1.03 per 16.1 µg/m
3
, 95% CI: 1.00,
1.06; p=0.04) was also attenuated by co-adjustment for first trimester exposure and any
other pollutant. Preconception O3 exposure was associated with decreased risk of GDM
(OR = 0.93 per 15.7 ppb, 95% CI: 0.91, 0.95). Sensitivity analyses that included either
season as an adjustment variable resulted in nearly identical pattern of associations
(data not shown).

65
DISCUSSION
In this large retrospective cohort study, increased risk of GDM was associated
with exposure to NO2 during the 12 weeks before conception, an association that was
robust to adjustment for multiple covariates, including co-exposure to NO2 during the
first trimester and models including another pollutant exposure during both
preconception and first trimester. The study was novel in examining effects of
preconception exposures and in adjusting for first trimester and other pollutant
exposures. A key strength of the study, compared to several previous studies, was to
refine the temporal relationship between pollutant exposures both before and after
conception with subsequent development of GDM. This was possible because
laboratory measurements and dates from the KPSC EMR were used to specify the date
of diagnosis.
Preconception substance abuse and exposures to radiation and chemicals such
as organic solvents have been associated with adverse birth outcomes (3). Less is
known about environmental effects during preconception on pregnancy complications,
including GDM. Although positive associations between air pollution, mainly PM2.5, NO2
or NOX, and type 2 diabetes have been consistently reported, GDM associations have
been inconsistent across only a small number of studies, and there has been little study
to date of GDM associations with preconception pollutant exposures (5, 6).
GDM likely shares pathways for development in common with type 2 diabetes,
since both are characterized by insulin resistance, and women with GDM are at
increased risk of developing type 2 diabetes after pregnancy (23). PM2.5 causes
diabetes in animal models (24), but the effects of NO2 exposure are less studied. NO2

66
causes oxidative stress and increased levels of proinflammatory cytokines (25) that also
characterize GDM (26). NO2 can also be a surrogate for the mixture of near-roadway air
pollution that results in increased local concentrations of NO2, or for other regional
pollutant mixtures correlated with NO2 that may explain the observed associations of
NO2 with GDM (27). However, our findings indicate that the NO2 associations were not
explained by PM2.5 exposure.
Only two studies have assessed pollutant exposures before conception (10, 12).
Our findings are consistent with those of a hospital-based cohort study in the US that
reported positive associations between GDM diagnosis and preconception NOx (10).
However, unlike our study, that study also identified an association with first trimester
NOx. A case-control study from Taiwan reported no associations of GDM with
preconception NO2 exposure (12). Other studies have reported null trimester-specific
NO2 exposure associations with (11, 12). Early gestational exposures have also been
associated with protective GDM associations with NO2 (15); positive associations with
NOx (16); and positive associations with NO (19). Reasons for these different results
between studies merit further investigation.
Why we observed consistently negative O3 associations with GDM is not clear,
as protective effects are not biologically plausible. However, if exposures to O3 were
causing fetal loss in early pregnancy before a GDM diagnosis could have occurred, then
we might see protective effects of O3 on GDM (15). One recent epidemiologic study
reported that exposures to O3 and PM2.5 during pregnancy were associated with
increased risk of fetal loss (28). Residual confounding by protective factors such as
healthy diet and physical activity could have explained protective associations, if these

67
protective factors were correlated with ozone exposure. However, preconception and
first trimester O3 exposures had very small positive correlations (R = 0.05 and 0.03,
respectively) with maternal pre-pregnancy BMI, a proxy for diet and physical activity that
was available in the EMR, and adjusting for BMI did not change our effect estimates.
There were several strengths to this study. All women followed during pregnancy
in the KPSC system administrative database were included in this study; therefore,
selection bias was unlikely to have influenced these results. The large, well-
characterized population with residential exposure for multiple criteria air pollutants was
another strength. Exposure assessment based on the child’s birth residential address
rather than delivery hospital referral region (10) or city/township (12, 19), used in some
previous studies, likely reduced exposure misclassification in our study. Our analysis
controlled for individual-level confounders available through the KPSC EMR, such as
maternal education and medically relevant covariates, which are generally not available
outside a single healthcare system. The KPSC membership comprised approximately
16% of the census reference population, with eligibility largely based on employment, so
the findings are generalizable to the working population of Southern California and
probably to other similar populations across the country (29). There were also some
limitations to the study, including the use of exposure at the child’s birth address as a
proxy for personal exposure. Measurement error could have occurred from not taking
into account residential mobility during pregnancy, and time spent away from home. If
the effect of this bias were non-differential with respect to the outcome, then the true
effect of exposure may have been larger than we observed (30).

68
CONCLUSIONS
NO2 exposure in the 12 weeks prior to conception was associated with an increased
risk of development of GDM. The findings have potentially large public health
implications, because GDM increases the risk of subsequent development of maternal
type 2 diabetes, and of childhood obesity and its consequences; and NO2 exposure can
be reduced with regulatory intervention.

69
REFERENCES

1. Bardenheier BH, Imperatore G, Gilboa SM, et al. Trends in Gestational Diabetes
Among Hospital Deliveries in 19 U.S. States, 2000-2010. Am J Prev Med
2015;49(1):12-9.
2. American Diabetes Association. Gestational diabetes mellitus [electronic article].
Diabetes Care.
3. Lassi ZS, Imam AM, Dean SV, et al. Preconception care: caffeine, smoking,
alcohol, drugs and other environmental chemical/radiation exposure. Reprod
Health 2014;11 Suppl 3:S6.
4. Johnson K, Posner SF, Biermann J, et al. Recommendations to improve
preconception health and health care--United States. A report of the
CDC/ATSDR Preconception Care Work Group and the Select Panel on
Preconception Care. MMWR Recomm Rep 2006;55(RR-6):1-23.
5. Thiering E, Heinrich J. Epidemiology of air pollution and diabetes. Trends
Endocrinol Metab 2015;26(7):384-94.
6. Eze IC, Hemkens LG, Bucher HC, et al. Association between ambient air
pollution and diabetes mellitus in Europe and North America: systematic review
and meta-analysis. Environ Health Perspect 2015;123(5):381-9.
7. Chen Z, Salam MT, Toledo-Corral C, et al. Ambient Air Pollutants Have Adverse
Effects on Insulin and Glucose Homeostasis in Mexican Americans. Diabetes
Care 2016;39(4):547-54.

70
8. Baio J. Prevalence of autism spectrum disorder among children aged 8 years -
Autism and Developmental Disabilities Monitoring Network, 11 sites, United
States, 2010. MMWR Surveill Summ 2014;63(2):1-21.
9. Hu H, Ha S, Henderson BH, et al. Association of Atmospheric Particulate Matter
and Ozone with Gestational Diabetes Mellitus. Environ Health Perspect
2015;123(9):853-9.
10. Robledo CA, Mendola P, Yeung E, et al. Preconception and early pregnancy air
pollution exposures and risk of gestational diabetes mellitus. Environ Res
2015;137:316-22.
11. Pedersen M, Olsen SF, Halldorsson TI, et al. Gestational diabetes mellitus and
exposure to ambient air pollution and road traffic noise: A cohort study. Environ
Int 2017;108:253-60.
12. Shen HN, Hua SY, Chiu CT, et al. Maternal Exposure to Air Pollutants and Risk
of Gestational Diabetes Mellitus in Taiwan. Int J Environ Res Public Health
2017;14(12).
13. Fleisch AF, Gold DR, Rifas-Shiman SL, et al. Air pollution exposure and
abnormal glucose tolerance during pregnancy: the project Viva cohort. Environ
Health Perspect 2014;122(4):378-83.
14. Fleisch AF, Kloog I, Luttmann-Gibson H, et al. Air pollution exposure and
gestational diabetes mellitus among pregnant women in Massachusetts: a cohort
study. Environ Health 2016;15:40.
15. Padula AM, Yang W, Lurmann FW, et al. Prenatal exposure to air pollution,
maternal diabetes and preterm birth. Environ Res 2018;170:160-7.

71
16. Malmqvist E, Jakobsson K, Tinnerberg H, et al. Gestational diabetes and
preeclampsia in association with air pollution at levels below current air quality
guidelines. Environ Health Perspect 2013;121(4):488-93.
17. Yorifuji T, Naruse H, Kashima S, et al. Residential proximity to major roads and
obstetrical complications. Sci Total Environ 2015;508:188-92.
18. van den Hooven EH, Jaddoe VW, de Kluizenaar Y, et al. Residential traffic
exposure and pregnancy-related outcomes: a prospective birth cohort study.
Environ Health 2009;8:59.
19. Pan SC, Huang CC, Lin SJ, et al. Gestational diabetes mellitus was related to
ambient air pollutant nitric oxide during early gestation. Environ Res
2017;158:318-23.
20. Xiang AH, Wang X, Martinez MP, et al. Association of maternal diabetes with
autism in offspring. JAMA 2015;313(14):1425-34.
21. Committee on Practice Bulletins-Obstetrics. ACOG Practice Bulletin No. 190:
Gestational Diabetes Mellitus. Obstet Gynecol 2018;131(2):e49-e64.
22. Eckel SP, Cockburn M, Shu YH, et al. Air pollution affects lung cancer survival.
Thorax 2016;71(10):891-8.
23. Xiang AH, Li BH, Black MH, et al. Racial and ethnic disparities in diabetes risk
after gestational diabetes mellitus. Diabetologia 2011;54(12):3016-21.
24. Rao X, Montresor-Lopez J, Puett R, et al. Ambient air pollution: an emerging risk
factor for diabetes mellitus. Curr Diab Rep 2015;15(6):603.
25. Lodovici M, Bigagli E. Oxidative stress and air pollution exposure. J Toxicol
2011;2011:487074.

72
26. Bowers K, Zhang C. Maternal diabetes and autism spectrum disorders in the
offspring: a review of epidemiological evidence and potential biologic
mechanisms. N A J Med Sci 2011;4(4):217-21.
27. Hoek G, Beelen R, de Hoogh K, et al. A review of land-use regression models to
assess spatial variation of outdoor air pollution. Atmospheric Environment
2008;42(33):7561-78.
28. Ha S, Sundaram R, Buck Louis GM, et al. Ambient air pollution and the risk
of pregnancy loss: a prospective cohort study. Fertil Steril 2018;109(1):148-53.
29. Koebnick C, Langer-Gould AM, Gould MK, et al. Sociodemographic
characteristics of members of a large, integrated health care system: comparison
with US Census Bureau data. Perm J 2012;16(3):37-41.
30. Rothman K, Greenland S, Lash TL. Modern Epidemiology, 3rd Edition.
Philadelphia, PA: Lippincott Williams & Wilkins.; 2008.

73
Table 3.1. Demographic characteristics by GDM for singleton deliveries in 1999-2009
GDM
(n=18,244)

No GDM
(n=221,330)
n (row %)
a
n (row %)
a

Age (yrs) 32.4 (5.4) 29.4 (5.8)
Parity
0 6486 (6.8) 88309 (93.2)
1 5326 (7.0) 70964 (93.0)
≥ 2 6107 (9.5) 58426 (90.5)
Missing 325 (8.2) 3631 (91.8)
Race/ethnicity
Non-Hispanic white 3461 (5.6) 58077 (94.4)
Non-Hispanic black 1289 (5.5) 22028 (94.5)
Hispanic 9753 (8.0) 112620 (92.0)
Asian/Pacific Islander 3451 (12.1) 25136 (87.9)
Other 290 (7.7) 3469 (92.3)
Education
High school or lower 7706 (7.8) 91397 (92.2)
Some College 5009 (7.5) 61716 (92.5)
College graduate or higher 5377 (7.5) 66238 (92.5)
Missing 152 (7.1) 1979 (92.9)
Household annual income
b

< $30,000 1376 (7.1) 18044 (92.9)
$30,000-$49,999 6318 (7.9) 73942 (92.1)
$50,000-$69,999 5872 (7.7) 70911 (92.3)
$70,000-$89,999 2871 (7.4) 35876 (92.6)
≥ $90,000 1807 (7.4) 22557 (92.6)
Prepregnancy body mass index
c

Underweight 61 (3.2) 1832 (96.8)
Normal 1761 (5.5) 30465 (94.5)
Overweight 1828 (8.9) 18780 (91.1)
Obese 2308 (13.3) 15009 (86.7)
Missing 12286 (7.3) 155244 (92.7)
a
Defined as GDM diagnosed at ≥ 13 weeks’ gestation
b
Based on census tract median

c
Information available starting in late 2006

74
Table 3.2. Pearson partial correlations of pollutants during preconception and first trimester

Preconception First trimester

PM2.5 PM10 NO2 O3 PM2.5 PM10 NO2 O3
Preconception
PM2.5 1.00 0.65 0.60 -0.13 0.54 0.33 0.44 -0.30
PM10 1.00 0.48 0.16 0.49 0.55 0.46 -0.33
NO2 1.00 -0.42 0.28 0.18 0.58 -0.37
O3 1.00 0.24 0.26 0.06 0.17
First trimester
PM2.5 1.00 0.66 0.63 -0.14
PM10 1.00 0.49 0.15
NO2 1.00 -0.42
O3 1.00
Note: All correlations were adjusted for birth year and KPSC medical center service areas and were statistically significant (p < 0.0001).
Gray cells indicate correlations between pollutants across exposure windows during preconception and first trimester.

75
Table 3.3. Adjusted odds ratios (ORs) and 95% confidence intervals (CIs) for association between GDM ≥13 weeks'
gestation and each pollutant (single-pollutant and mutually adjusted)
Model Pollutant Preconception First trimester
OR
a, b
(95% CI) OR
a, b
(95% CI)
Single-pollutant
c
NO2 1.11 (1.08-1.14) 1.02 (1.00-1.05)
Mutually adjusted
d
Adjusted for PM2.5 1.10 (1.06-1.14) 1.02 (0.99-1.06)
Adjusted for PM10 1.11 (1.07-1.15) 1.00 (0.97-1.04)
Adjusted for O3 1.04 (1.01-1.08) 1.01 (0.97-1.04)
Single-pollutant
c
PM2.5 1.04 (1.01-1.06) 0.98 (0.95-1.00)
Mutually adjusted
d
Adjusted for PM10 1.04 (1.01-1.08) 0.96 (0.93-0.99)
Adjusted for NO2 1.00 (0.97-1.03) 0.96 (0.93-1.00)
Adjusted for O3 1.01 (0.98-1.04) 0.99 (0.96-1.02)
Single-pollutant
c
PM10 1.03 (1.00-1.06) 0.98 (0.95-1.01)
Mutually adjusted
d
Adjusted for PM2.5 1.02 (0.98-1.06) 0.99 (0.96-1.03)
Adjusted for NO2 0.99 (0.96-1.03) 0.99 (0.95-1.02)
Adjusted for O3 1.02 (0.98-1.05) 1.02 (0.99-1.06)
Single-pollutant
c
O3 0.93 (0.91-0.95) 0.95 (0.93-0.97)
Mutually adjusted
d
Adjusted for PM2.5 0.94 (0.92-0.96) 0.96 (0.94-0.97)
Adjusted for PM10 0.93 (0.91-0.95) 0.96 (0.94-0.98)
Adjusted for NO2 0.94 (0.92-0.97) 0.97 (0.95-0.99)
a
Adjusted for family correlation, birth year, KPSC medical center service areas, maternal age, race/ethnicity
b
NO2 per 10.4 ppb, PM2.5 per 6.5 μg/m
3
,

PM10 per 16.1 μg/m
3
,

O3 per 15.7 ppb
c
Odds ratios for each pollutant preconception or first trimester exposure alone
d
Odds ratios mutually adjusted for preconception and first trimester exposures of the pollutant plus a co-pollutant from both preconception and first
trimester exposure windows. For example, the estimate for preconception NO2 exposure mutually adjusted for PM2.5 included first trimester NO2 +
preconception PM2.5 + first trimester PM2.5 exposures.
Bold=statistically significant at p<0.05

76
Figure S3.1. Kaiser Permanente Southern California (KPSC) medical center service areas

Note: Black dots indicate air quality monitoring stations that had data for one or more pollutants during the study period. Service areas contributing
to the analysis are shown in the legend. KPSC covers Imperial, Kern, Los Angeles, Orange, Riverside, San Bernardino, San Diego, San Luis
Obispo, Santa Barbara, and Ventura counties. Of the18 geographically located KPSC medical center service areas across Southern California, 4
were excluded (Kern, Antelope Valley, West Ventura, and Coachella Valley) because there were few cases of ASD.

77

Figure S3.2. Distribution of pollutant concentrations during preconception and first trimester across birth year 1999-2009

78

79
CHAPTER 4. GESTATIONAL DIABETES MELLITUS, PRENATAL AIR POLLUTION
EXPOSURE, AND AUTISM SPECTRUM DISORDER

ABSTRACT
Ambient air pollution and maternal diabetes may affect common biological pathways
that underlie adverse effects on neurodevelopment. However, no study has examined
the joint effects of maternal diabetes and air pollution on autism spectrum disorder
(ASD). We evaluated whether prenatal and early-life exposure to air pollution interact
with maternal diabetes status to affect child’s ASD risk. This retrospective cohort study
included 246,420 singleton children born in Kaiser Permanente Southern California
(KPSC) hospitals between 1999 and 2009. Children were followed from birth until age 5,
during which, 2,471 ASD cases were diagnosed in the electronic medical record.
Ambient ozone (O3), particulate matter <2.5 μm and <10 μm in aerodynamic diameter,
and nitrogen dioxide were estimated based on regulatory air monitoring station
measurements interpolated to estimate exposures during preconception, each trimester
of pregnancy, and first year of life at each child’s birth address. Hazard ratios (HRs)
were estimated using Cox regression models to adjust for birth year, KPSC medical
center service areas, and relevant maternal and child characteristics. For each
exposure window, interactions between pollutants and 4-category maternal diabetes
variable (none, GDM ≥24 weeks’ gestation, GDM <24 weeks’ gestation, and pre-
existing type 2 diabetes) were tested. For an exposure window with statistically
significant global interaction between pollutant and diabetes, pollutant-associated HRs
were estimated separately for each category of maternal diabetes status. There were

80
associations of PM2.5 with preconception, entire pregnancy, first and third trimesters,
and first year of life, but no associations with other pollutants. There were, however,
statistically significant interactions between maternal diabetes and O3 during the first
trimester (p=0.047) and first year of life (p=0.007). Increased risk of ASD was
associated with O3 among mothers with GDM <24 weeks’ gestation [adjusted HRs 1.50
per 15.7 ppb

O3 (95% CI: 1.08-2.09)] during the first trimester. No statistically significant
O3 associations with ASD were observed in offspring of pregnancies without diabetes,
GDM ≥24 weeks’ gestation, or pre-gestational diabetes. No interactions were observed
with other pollutants. GDM with onset early in pregnancy may increase children’s
susceptibility to prenatal O3-associated ASD risk. These novel findings merit further
investigation.

81
INTRODUCTION
Autism spectrum disorder (ASD) prevalence increased dramatically during 2000-
2012 from 0.7% to 1.7% in the United States (1). The increase can only partly be
explained by better ascertainment (2). Emerging evidence from both human and animal
studies has identified ambient particles and other criteria pollutants as potentially
modifiable risk factors for ASD (3-5). ASD has also been associated with a broad
spectrum of conditions characterized by maternal immune activation (6, 7), including
maternal diabetes during pregnancy (8). With the rising prevalence of maternal diabetes
during the same period as increases in ASD prevalence (9), a few studies have
consistently found positive associations with maternal diabetes, especially gestational
diabetes mellitus (GDM) (10-14).
Particulate air pollution has shown relatively consistent associations with ASD
(3). The influence of particulate matter < 2.5 μm in aerodynamic diameter (PM2.5) on
ASD risk may vary by exposure time windows, including some studies reporting
associations with exposure during the first and third trimesters of pregnancy, and others
showing associations with exposure during the first year of life (15, 16). Fewer studies
have found prenatal and postnatal NO2 associations with ASD risk (16). Relatively less
studied is ozone (O3). Two studies found O3 exposures during the second and third
trimesters, and early postnatal period, respectively, were associated with ASD (17, 18).
Causes of ASD in prenatal and early postnatal life are likely multi-factorial (19).
Yet few epidemiologic studies have evaluated effects of common exposures like air
pollution and maternal diabetes to see if they have independent or synergistic effects.
Exposure to O3, for example, has neuroinflammatory effects, perhaps mediated by O3-

82
induced oxidative stress and systemic inflammation (20-23). Maternal hyperglycemia
during pregnancy also causes systemic oxidative stress and chronic inflammation,
resulting in the associated inflammatory intrauterine environment that may have
adverse effects on fetal brain development (11). Thus, the overlapping timing of
inflammation and systemic oxidative stress induced by both O3 and diabetes may
potentially affect common biological pathways that could further increase the
susceptibility to ASD in children (6, 11).
One of the main challenges in examining the effects of combinations of risk
factors is acquiring a large enough study population. Additionally, we can better assess
the heterogeneity of exposure effect between individuals in a large study population.
Based on recent evidence of neurotoxic effects of air pollution and of maternal diabetes,
we hypothesized that the children born to mothers with diabetes during pregnancy
would have a higher risk of ASD in early childhood associated with pollutant exposures
than mothers without diabetes. In a large, population-based pregnancy cohort from
Kaiser Permanente Southern California (KPSC), we examined diabetes-specific
associations of ASD with prenatal and first year of life regional pollutant exposures.

METHODS
Study design and population
This retrospective cohort study included mother-child pairs with singleton deliveries in
KPSC hospitals between January 1, 1999, and December 31, 2009 in 14 service
areas located across Southern California (Figure S4.1). The residential addresses
extracted from birth certificate records were linked by a unique KPSC membership

83
identifier. The primary analysis included 246,420 mother-child pairs with children
enrolled as KPSC plan members at age one year, as previously described (24), after
excluding children with birth certificate addresses outside Southern California (n =
636) or addresses that could not be accurately geocoded (n = 4406), because of an
address missing or not matchable to a U.S. postal service address. Follow-up was
accrued until the first occurrence of 1) clinical diagnosis of ASD; 2) last date of
continuous KPSC plan membership; 3) death from any cause; or 4) age five. Children
were censored at age five to ensure the same follow-up time for the entire cohort,
regardless of birth date. Thus, the youngest children, born in 2009, were followed
through 2014. Both outcome and covariate data were extracted from the KPSC
electronic medical records (EMR). Both the KPSC and the University of Southern
California Institutional Review Boards approved this study.

Outcome data on ASD
KPSC neurodevelopment screening procedures included an abbreviated Checklist
for Autism in Toddlers (CHAT) (25) administered at 18- and 24-month well child
visits. Children failing the screening were referred to a pediatric developmental
specialist for further evaluation and ASD diagnosis (14, 24). The presence or
absence of ASD during follow-up was identified by International Classification of
Diseases, Ninth Revision (ICD-9) codes 299.x or equivalent KPSC codes from the
EMR from at least two separate visits, an approach validated previously (14, 26).

84
Exposure assessment
Birth certificate residential addresses were geocoded using MapMarker USA
Version 28.0.0.11. Exposure metrics at each geocoded address included regional
O3, PM2.5, PM ≤10 μm in diameter (PM10), and nitrogen dioxide (NO2). Monthly
averages for each pollutant between 1998-2009 were obtained from data compiled
from the EPA regional air quality monitoring network. Exposure at each address was
assigned based on the monthly inverse distance-squared weighted average from
these regional monitoring stations. For geocoded address locations within 0.25 km
of a monitor, data only from that monitoring station were used. Although the
distance-weighted approach has limited accuracy in areas with sparse monitoring
networks, performance is acceptable in Southern California due to the dense
geographical network of historical measurements covering the region. In a previous
Southern California study evaluating this method using leave-one-out validation for
monthly exposure prediction, the coefficients of determination (r
2
) were 0.76, 0.73,
0.53, and 0.46 for ozone, NO2, PM2.5 and PM10, respectively, with lower R
2
values
for PM10 attributed to the local (primary emission) dust component that is not
regional (27). Bias was less than 1 ppb or 1 µg/m
3
for the gaseous and particulate
pollutants, respectively. Each address was assigned the monthly average of the 24-
hour concentrations of PM2.5, PM10, and NO2. For O3, the monthly average of daily
maximum 8-hour concentrations was estimated. Based on the mother’s last
menstrual period, averages of the monthly concentrations were calculated during
preconception, each trimester, the entire pregnancy and the first year of life.
Preconception exposure was defined as 12 weeks before mother’s last menstrual

85
period date. First trimester exposure was defined as 0-12 weeks, second trimester
as 13-26 weeks, and third trimester as 27 weeks to birth. The monthly average
exposures during months overlapping these different time periods were weighted by
the number of days in each period.

Maternal diabetes
As previously described (24), maternal diabetes during index pregnancy was
categorized based on ICD-9 codes, anti-diabetic medication use, and glucose
values from 1-hour 50-g glucose challenge tests and/or oral glucose tolerance tests
administered during pregnancy: no diabetes; GDM; and type 2 diabetes before
pregnancy. KPSC follows the American College of Obstetricians and Gynecologists
guidelines for screening for GDM (28). Diagnosis of GDM was based on laboratory
values confirming a plasma glucose level of 200mg/dL or higher on the glucose
challenge test or at least 2 plasma glucose values meeting or exceeding the
following values on the 100-g or 75-g oral glucose tolerance test: fasting, 95 mg/dL;
1 hour, 180 mg/dL; 2 hours, 155 mg/dL; and 3 hours, 140 mg/dL (24, 29).
Gestational age at GDM diagnosis was calculated using the date of the first glucose
test result that met the GDM diagnosis criteria, date of delivery, and gestational age
at delivery available in the electronic medical record. Based on previous study that
found higher risk of ASD among mothers with GDM diagnosed earlier in pregnancy
(24) and routine screening of GDM starting at 24 weeks’ gestation, GDM exposure
was further categorized as diagnosis before or after 24 weeks’ gestation.

86
Covariates
Potential confounders chosen a priori, based on previous associations with ASD in this
cohort (24), included sex of the child and maternal age at delivery, parity, education,
self-reported race/ethnicity, history of comorbidity [≥1 diagnosis of heart, lung, kidney, or
liver disease; cancer], and median family household income in the census tract of
residence. A missing category was used for categorical covariates with missing data
(parity [n = 4,125], education [n = 2,222], and household income [n = 1,850]). To
account for temporal changes in ASD incidence rate and pollution levels, we adjusted
for birth year, and to account for broad geographical characteristics associated with
ASD, we adjusted for the 14 KPSC service areas. Additional pregnancy-related
covariates included maternal pre-eclampsia/eclampsia, preterm birth (< 37 weeks vs. ≥
37 weeks), and congenital anomalies.

Statistical analyses
Pearson Partial correlations were calculated between pollutants within each exposure
window and across multiple exposure windows, adjusted for birth year and KPSC
service areas. Cox proportional hazards models were used to estimate the ASD hazard
ratios (HRs) associated in separate models with each pollutant exposure, adjusting for
potential confounders and for potential correlation due to multiple siblings born to the
same mother. Additionally, adjusting for covariates potentially on the causal pathway
(maternal pre-eclampsia/eclampsia, preterm birth and congenital anomalies) did not
change estimated effects by >10%, so these variables were not included in the final
models. Restricted cubic splines identified no evidence of non-linear associations of

87
ASD with pollutants, so pollutants were treated as continuous variables and modeled
linearly. Both Goodness-of-Fit Test and Extended Cox models identified no violations of
proportional hazards assumptions for all covariates, and HRs for pollutant associations
were in the same direction across all follow-up time periods (1-2, >2-3, >3-4, >4-5 years
of child’s age).
All models were adjusted for calendar birth year as a continuous covariate and
for the 14 KPSC medical center service areas, as well as maternal age, parity, maternal
race/ethnicity, maternal education, census tract median household income, maternal
history of comorbidities before pregnancy, and child sex. Because the analysis of
pollutant effects on ASD were adjusted for year and service areas, we scaled each HR
to be representative of exposure contrasts both within-service area and within-year.
Pollutant estimates were scaled to the difference of the 95
th
to the 5
th
percentile of the
deviations of each pregnancy exposure from the average within the same service area
in the same year. Deviations were calculated as each residential pollutant exposure
value minus the within-service area, within-year mean exposure. For example, for each
of the 14 service areas and 11 years (154 in total) the average PM2.5 residential
exposure and the deviations of individual PM2.5 from this average were calculated. The
95
th
percentile (3.0 µg/m
3
) minus the 5
th
percentile (-3.5 µg/m
3
) of PM2.5 deviation
distributions resulted in the within-service area and within-year scale of 6.5 µg/m
3
for
PM2.5. The same procedure was used to calculate the within-service area, within-year
scales for other pollutants: 16.1 µg/m
3
for PM10, 10.4 ppb for NO2, and 15.7 ppb for O3.
HRs were also estimated for ASD association with a 4-level categorical variable
for maternal diabetes (none (reference), GDM diagnosed <24 weeks’ gestation, GDM

88
diagnosed ≥24 weeks’ gestation, and pre-existing type 2 diabetes). To evaluate the
hypothesis that the effects of air pollution on ASD risk would be greater among mothers
with diabetes during pregnancy compared to mothers without diabetes, we tested for
global interactions between this 4-level categorical variable for maternal diabetes and
each pollutant separately in each exposure window (preconception, entire pregnancy,
each trimester, and first year of life).
For pollutants and exposure windows for which the global interaction was
statistically significant, we estimated separate HRs for the effect of each pollutant
exposure on ASD risk among mothers with no diabetes, GDM diagnosed <24 weeks’
gestation, GDM diagnosed ≥24 weeks’ gestation, and pre-existing type 2 diabetes. We
also examined the joint effects of maternal diabetes and pollutant exposures on ASD
risk by fitting a model with the 12-category exposure (4-category diabetes by pollutant
tertiles), using the lowest tertile of pollutant exposure in mothers without diabetes as
reference. To evaluate the exposure-response associations of these joint exposures on
ASD risk, we tested for a global linear trend of HRs from the lowest tertile to the highest
tertile of pollutant exposure separately for mothers with no diabetes, GDM diagnosed
<24 weeks’ gestation, GDM diagnosed ≥24 weeks’ gestation, and pre-existing type 2
diabetes. This was done by treating the categorical pollutant tertiles (coded as 1, 2, 3)
as continuous variables and adjusting for potential confounders, 4-level categorical
variable for maternal diabetes, and an interaction term between continuous pollutant
tertile variable and 4-category diabetes.
Two-sided statistical tests were conducted at an alpha level of 0.05, and
precision was measured using 95% confidence intervals (CIs). Data analyses were

89
conducted using SAS 9.4 (SAS Institute, Inc, Cary, NC) and R, version 3.0.2 (64
bit).

RESULTS
Unadjusted annual ASD incidence rates (per 1,000 child-years) increased with birth
year from 1.93 in 1999 to 4.03 in 2009 (Figure S4.2) (15), a period during which national
prevalence rates of ASD were also increasing (2). There were 2471 children diagnosed
with ASD in the cohort.
Levels of O3, PM2.5, PM10, and NO2, during the averaged across the entire 9
months of pregnancy during the entire 2000-2009 period were 17.9 micrograms per
meter-cubed (µg/m
3
), 38.1 µg/m
3
, 25.1 parts per billion (ppb), and 41.6 ppb,
respectively. O3 levels remained relatively stable across years (Figure S4.3) (15).
However, mean levels of both PM2.5 and NO2 decreased across years from 1999-2009.
PM10 levels fluctuated across time, potentially reflecting variable precipitation across the
years.
Levels of pollutants during the entire pregnancy averaged across the 1999-2009
period also varied between KPSC service areas (Figure S4.4). Highest mean levels of
O3 were in Moreno Valley (52.5 ppb), and lowest mean levels of O3 were in Downey
(31.8 ppb). Highest levels of mean PM2.5, PM10 and NO2 were in Ontario (21.6 µg/m
3
,
49.7 µg/m
3
, 31.2 ppb, respectively), and lowest levels of PM2.5, PM10, and NO2 were in
San Diego (13.1 µg/m
3
, 30.7 µg/m
3
, 18.0 ppb).

90
Levels of PM2.5, PM10, NO2, and O3 during pregnancy averaged across the entire
study were similar for mothers with no diabetes, GDM diagnosed <24 weeks’ gestation,
GDM diagnosed ≥24 weeks’ gestation, and pre-existing type 2 diabetes (Table S4.1).
Table 4.1 showed the associations of ASD risk with maternal and child
characteristics, adjusting for birth year and KPSC service areas. Older maternal age,
being first born, higher maternal education, history of maternal comorbidity, <$30,000
median family household income in the census tract of residence, and being a boy were
associated with increased risk of ASD. In contrast, mothers who were multiparous and
residing in a census tract with >$50,000 median household income were less likely to
give birth to child with ASD.
Children born to mothers with pre-existing type 2 diabetes were at significantly
increased risk of ASD compared to mothers without diabetes, after adjusting for birth
year and KPSC service areas (HR = 1.60, 95% CI: 1.26, 2.16; Table 4.2). This
association was modestly attenuated after further adjusting for other confounders (HR =
1.45, 95% CI: 1.11, 1.91). Children born to mothers with GDM diagnosed before 24
weeks’ gestation were also at an increased risk of ASD; however, this positive
association was attenuated after further adjusting for other confounders (HR = 1.24,
95% CI: 0.95, 1.62). In contrast, there was little association between GDM diagnosed
after 24 weeks and risk of ASD (HR = 0.92, 95% CI: 0.77, 1.09).
The third trimester O3 association with ASD was weakly positive (HR = 1.05, 95%
CI: 0.99, 1.11; P=0.11), after adjusting for confounders (Table 4.3). In other exposure
windows, there were weak associations of ASD with O3, which were not statistically
significant. There were associations with PM2.5 exposure, including during

91
preconception (HR = 1.11 per 6.5 μg/m
3
, 95% CI: 1.03, 1.20), the entire pregnancy (HR
= 1.17 per 6.5 μg/m
3
, 95% CI: 1.04, 1.33), the first trimester (HR = 1.10, 95% CI: 1.02,
1.19) and the third trimester (HR = 1.08, 95% CI: 1.00, 1.18). Increased risk of ASD was
also associated with first year of life PM2.5 exposure (HR = 1.21, 95% CI: 1.05, 1.40).
None of the associations of ASD with either PM10 or NO2 were statistically significant.
During the first trimester and first year of life, the global interactions between
continuous O3 exposure and the 4-level categorical variable for maternal diabetes (no
diabetes, GDM diagnosed at <24 weeks and ≥24 weeks, and pre-existing diabetes)
were statistically significant (p = 0.047; p = 0.007; Table S4.2). No other pollutants in
any exposure window had statistically significant global interactions with the 4-level
categorical variable for maternal diabetes (p > 0.05; Table S4.4 and Table S4.5).
For all O3 exposure windows, we estimated separate HRs associated with the
risk of ASD among mothers with no diabetes, GDM diagnosed at <24 weeks and ≥24
weeks, and pre-existing diabetes). Among mothers with GDM <24 weeks’ gestation,
increased ASD risk was associated with residential O3 exposure during entire
pregnancy and during the first trimester (HR = 1.96 per 15.7 ppb, 95% CI: 1.24, 3.11
and HR = 1.50 per 15.7 ppb, 95% CI: 1.08, 2.09, respectively; Table 4.4). A large
increased risk was also associated with the child’s O3 exposure during the first year of
life (HR = 2.01 per 15.7 ppb, 95% CI: 0.67, 6.07); however, this association was not
statistically significant. No statistically significant O3 associations within each exposure
window were observed among mothers without diabetes, with GDM ≥24 weeks’
gestation, or with pre-existing type 2 diabetes. In sensitivity analyses, we co-adjusted
these first trimester and first year of life O3 associations for NO2 exposure, which was

92
negatively correlated with O3 (first trimester R= -0.42, first year of life R=-0.49; Table
S4.3). The estimates of O3 effect were not substantially different after adjustment
(results not shown). Other pollutants had weak associations with O3 during these
exposure windows.
To further evaluate the joint effects of maternal diabetes and O3 on ASD risk, we
fit a model with a 12-category variable combining the four maternal diabetes categories
(no diabetes, GDM diagnosed at <24 weeks and ≥24 weeks, and pre-existing diabetes)
by tertiles of O3, using the lowest tertile of O3 exposure in mothers without diabetes as
the referent group (Table 4.5), with tertiles classified based on first trimester O3
exposure averaged across all years (<37.7 ppb, 37.7-<44.3 ppb, and ≥44.3 ppb). As
shown in Table 4.5 and Figure 4.1, we identified statistically significant trends in O3
effect estimates among mothers with GDM <24 weeks’ gestation during the first
trimester (p for trend = 0.03) and during the first year of life (p for trend = 0.01). In the
first trimester, the middle and high tertile associations with ASD were 1.10 (95% CI:
0.69, 1.74) and 1.71 (95% CI: 1.12, 2.60) compared with the lowest tertile; in the first
year of life, the effect estimates for O3 with ASD were 1.65 (95% CI: 1.08, 2.51) and
1.68 (95% CI: 1.01, 2.81) compared with the lowest tertile. ASD was associated with
pre-pregnancy diabetes, but there was little evidence for an exposure gradient in risk.

DISCUSSION
In this first population-based cohort study examining the joint effects of air pollution and
maternal diabetes during pregnancy on the risk of ASD, we observed associations of
ASD with PM2.5 during multiple prenatal and early postnatal exposure windows and

93
with pre-existing maternal type 2 diabetes, which we have reported previously (14, 15).
Although no associations of ASD with O3 by itself were observed, there was substantial
heterogeneity of effects by categories of diabetes/GDM during the first trimester and
first year of life. Prenatal exposure to O3 during the first trimester was associated with
increased risk of ASD among mothers with GDM diagnosis before 24 weeks’ gestation.
In models examining effects of O3 tertiles, there were statistically significant
associations among mothers diagnosed before 24 weeks both in the first trimester and
first year of life. We did not find any heterogeneity in associations of ASD with other
pollutants by maternal diabetes status. Because we observed no main effects of O3 on
ASD, caution is warranted in the interpretation of the significance of these diabetes
category-specific effects.
Only two epidemiologic studies have reported associations of O3 with increased
risk of ASD (17, 18), according to a recent review (30). One study reported ASD
associations with O3 during pregnancy (odds ratio (OR) = 1.06 per 11.54 ppb, 95% CI:
1.01, 1.12) (17), but not during the first trimester (OR = 1.00 per 11.54 ppb, 95% CI:
0.97, 1.03). Small increases in ASD risk were associated with second and third
trimester exposures (OR = 1.02 per 11.54 ppb, 95% CI: 1.00, 1.05; OR = 1.04 per 11.54
ppb, 95% CI: 1.01, 1.06). A study from Taiwan also reported ASD association with
postnatal O3 exposure in the year preceding newly diagnosed ASD cases under 3 years
of age (HR = 1.59, 95% CI: 1.42-1.78) (18). Neither of these studies examined
interactions of O3 with diabetes.
In contrast to these early studies, our results suggest that O3 by itself may not be
causing ASD, but that first trimester and first year of life O3 exposure-associated ASD

94
requires a second “hit” from GDM during a critical early window of brain development
during gestation for ASD to occur (6). Diabetes prior to pregnancy was associated with
ASD, but O3 was not associated with increased ASD risk in mothers with diabetes prior
to pregnancy. Since the KPSC EMR does not have information on the actual date of
pre-existing type 2 diabetes diagnosis, it is possible that that the degree of glycemic
control during the first trimester of pregnancy may vary in these mothers with pre-
pregnancy diabetes and thus, contribute to our null findings (24). Thus, the effects of O3
may be greater among children born to mothers with possibly undetected GDM in early
pregnancy since they are already at a greater risk for ASD due to other diabetes-related
pregnancy complications (31).
A limited number of animal studies has examined adverse neurobehavioral
effects after gestational exposure to O3 (32-35). Autism-related behavior deficits,
including reduced social interaction and increased repetitive behavior, have been
reported after early gestational O3 exposure (33). O3-induced oxidative stress and
associated systemic inflammation during pregnancy that have adverse affects on the
developing fetal brain may contribute to the pathogenesis of ASD (36). Moreover,
maternal immune activation-related pregnancy complications such as hyperglycemia
during pregnancy share common biological pathways with O3 (6, 7), which causes
chronic inflammation and systemic oxidative stress, and the associated inflammatory
intrauterine environment that could adversely affect neurodevelopment (10, 11).
Consequently, exposure to simultaneous “hits” of inflammation and systemic oxidative
stress from early gestation GDM and O3 may further increase children’s susceptibility for
ASD in early childhood (6, 11, 22, 37). These synergistic biologic effects of O3 and GDM

95
may together explain our findings of O3-associated increased risk of ASD among
children born to mothers with GDM diagnosed before 24 weeks of pregnancy.
An important consideration is that our results were based on multiple
comparisons that increased the chance of incorrectly rejecting a null hypothesis (i.e.
Type 1 error). Our findings were not based on an a priori hypothesis that diabetes would
specifically augment O
3
effects on ASD but not other pollutants. If we adjusted for
multiple comparisons of the 20 interactions in total across 4 pollutants and 5 exposure
windows (Table S2; excluding entire pregnancy), then none of these interactions would
have been statistically significant at p = 0.05/20 = 0.0025. Thus, our results should be
interpreted as hypothesis-generating, although the interaction of O3 x type of diabetes
during the first year of life (p = 0.007) would have been close to significant.
Other studies have identified synergistic effects of air pollution. Lower folic acid
intake, male sex, and high-risk genes have been shown to have synergistic interactions
with NO-, NO2-, and O3-associated ASD, for example (38-41). We previously reported
that first trimester PM2.5-associated ASD risk was stronger in boys in this cohort (HR =
1.18 per 6.5 μg/m
3
; 95% CI, 1.08-1.27) compared to girls (HR = 0.90 per 6.5 μg/m
3
;
95% CI, 0.76-1.07) (15). Studies with a large enough sample size, like the KPSC cohort
resource, are needed to assess interactions of environmental and other risk factors that
may explain the etiology of complex diseases like autism.
This study has several strengths. The large study sample provided statistical
power to assess an interaction between pollutants and maternal diabetes. Selection
bias was unlikely to have influenced the results, because the cohort did not self-select
and also had a high annual average 95% retention through age 5 after cohort

96
enrollment (24). Unlike many health systems that have only recently adopted an EMR,
the KPSC EMR system has been refined and improved since its implementation in the
early 1990’s. Incident ASD ascertainment occurred in a clinical setting using
standardized diagnostic algorithms developed through the EMR (26). Chances of GDM
misclassification were reduced since we had information on gestational age at diagnosis
from the EMR. The analysis controlled for various individual-level covariates that were
available through the EMR, such as maternal education, that are difficult to obtain
without standardized procedures in a single healthcare system. We also fully accounted
for temporal trends of ASD and pollutants across study period, as well as for any
temporal trends in ascertainment of ASD by adjusting for birth year in our analyses. The
KPSC membership comprised approximately 16% of the census reference population,
so results are generalizable to the population of working Southern California families
(42). Because Southern California has large pollutant exposure gradients representative
of ranges and extremes across the U.S., the results are broadly relevant to the U.S. and
other countries with similar exposures.
There were some limitations. Perhaps most important, our findings should be
considered to be hypothesis generating and may not reflect a true biological interaction
between GDM and O3, as noted above. Exposure concentrations estimated at the birth
address were used as a proxy for personal exposure. Residential mobility during
pregnancy and the first year of life, and time of mother or child away from home would
have resulted in exposure measurement error. If the effect of this bias were non-
differential with respect to the outcome, then the true effect of exposure could have
been larger than we observed (43).

97

CONCLUSIONS
GDM diagnosed before the end of the second trimester of pregnancy may increase
children’s susceptibility to prenatal and early life O3-associated ASD risk. However,
these hypothesis-generating findings need to be replicated in future studies with large
study populations. Such studies may begin to explain the multifactorial etiology of
diseases such as ASD. Further toxicological study is also needed to better understand
the biological basis and timing of susceptibility during neurodevelopment and of
differences in neurobehavioral effects of O3 exposures by maternal diabetes.

98
REFERENCES

1. Baio J, Wiggins L, Christensen DL, et al. Prevalence of Autism Spectrum
Disorder Among Children Aged 8 Years - Autism and Developmental Disabilities
Monitoring Network, 11 Sites, United States, 2014. MMWR Surveill Summ
2018;67(6):1-23.
2. Baio J. Prevalence of autism spectrum disorder among children aged 8 years -
Autism and Developmental Disabilities Monitoring Network, 11 sites, United
States, 2010. MMWR Surveill Summ 2014;63(2):1-21.
3. Lam J, Sutton P, Kalkbrenner A, et al. A Systematic Review and Meta-Analysis
of Multiple Airborne Pollutants and Autism Spectrum Disorder. PLoS One
2016;11(9):e0161851.
4. Church JS, Tijerina PB, Emerson FJ, et al. Perinatal exposure to concentrated
ambient particulates results in autism-like behavioral deficits in adult mice.
Neurotoxicology 2017.
5. Li K, Li L, Cui B, et al. Early postnatal exposure to airborne fine particulate matter
induces autism-like phenotypes in male rats. Toxicol Sci 2017.
6. Estes ML, McAllister AK. Maternal immune activation: Implications for
neuropsychiatric disorders. Science 2016;353(6301):772-7.
7. Malkova NV, Yu CZ, Hsiao EY, et al. Maternal immune activation yields offspring
displaying mouse versions of the three core symptoms of autism. Brain Behav
Immun 2012;26(4):607-16.

99
8. Alderete TL, Chen Z, Toledo-Corral CM, et al. Ambient and Traffic-Related Air
Pollution Exposures as Novel Risk Factors for Metabolic Dysfunction and Type 2
Diabetes. Curr Epidemiol Rep 2018;5(2):79-91.
9. Albrecht SS, Kuklina EV, Bansil P, et al. Diabetes trends among delivery
hospitalizations in the U.S., 1994-2004. Diabetes Care 2010;33(4):768-73.
10. Wan H, Zhang C, Li H, et al. Association of maternal diabetes with autism
spectrum disorders in offspring: A systemic review and meta-analysis. Medicine
(Baltimore) 2018;97(2):e9438.
11. Bowers K, Zhang C. Maternal diabetes and autism spectrum disorders in the
offspring: a review of epidemiological evidence and potential biologic
mechanisms. N A J Med Sci 2011;4(4):217-21.
12. Xu G, Jing J, Bowers K, et al. Maternal diabetes and the risk of autism spectrum
disorders in the offspring: a systematic review and meta-analysis. J Autism Dev
Disord 2014;44(4):766-75.
13. Xiang AH, Black MH, Li BH, et al. Racial and ethnic disparities in extremes of
fetal growth after gestational diabetes mellitus. Diabetologia 2015;58(2):272-81.
14. Xiang AH, Wang X, Martinez MP, et al. Maternal Type 1 Diabetes and Risk of
Autism in Offspring. JAMA 2018;320(1):89-91.
15. Jo H, Eckel SP, Chen J, et al. Sex-specific Associations of Autism Spectrum
Disorder with Residential Air Pollution Exposure. Unpublished manuscript 2018.
16. Flores-Pajot MC, Ofner M, Do MT, et al. Childhood autism spectrum disorders
and exposure to nitrogen dioxide, and particulate matter air pollution: A review
and meta-analysis. Environ Res 2016;151:763-76.

100
17. Becerra TA, Wilhelm M, Olsen J, et al. Ambient air pollution and autism in Los
Angeles county, California. Environ Health Perspect 2013;121(3):380-6.
18. Jung CR, Lin YT, Hwang BF. Air pollution and newly diagnostic autism spectrum
disorders: a population-based cohort study in Taiwan. PLoS One
2013;8(9):e75510.
19. Lyall K, Schmidt RJ, Hertz-Picciotto I. Maternal lifestyle and environmental risk
factors for autism spectrum disorders. Int J Epidemiol 2014;43(2):443-64.
20. Campbell A, Araujo JA, Li H, et al. Particulate matter induced enhancement of
inflammatory markers in the brains of apolipoprotein E knockout mice. J Nanosci
Nanotechnol 2009;9(8):5099-104.
21. Campbell A, Daher N, Solaimani P, et al. Human brain derived cells respond in a
type-specific manner after exposure to urban particulate matter (PM). Toxicol In
Vitro 2014;28(7):1290-5.
22. Rossignol DA, Frye RE. Evidence linking oxidative stress, mitochondrial
dysfunction, and inflammation in the brain of individuals with autism. Front
Physiol 2014;5:150.
23. Block ML, Calderón-Garcidueñas L. Air pollution: mechanisms of
neuroinflammation and CNS disease. Trends Neurosci 2009;32(9):506-16.
24. Xiang AH, Wang X, Martinez MP, et al. Association of maternal diabetes with
autism in offspring. JAMA 2015;313(14):1425-34.
25. Baron-Cohen S, Wheelwright S, Cox A, et al. Early identification of autism by the
CHecklist for Autism in Toddlers (CHAT). J R Soc Med 2000;93(10):521-5.

101
26. Coleman KJ, Lutsky MA, Yau V, et al. Validation of Autism Spectrum Disorder
Diagnoses in Large Healthcare Systems with Electronic Medical Records. J
Autism Dev Disord 2015;45(7):1989-96.
27. Eckel SP, Cockburn M, Shu YH, et al. Air pollution affects lung cancer survival.
Thorax 2016;71(10):891-8.
28. Committee on Practice Bulletins-Obstetrics. ACOG Practice Bulletin No. 190:
Gestational Diabetes Mellitus. Obstet Gynecol 2018;131(2):e49-e64.
29. American Diabetes Association. Gestational diabetes mellitus [electronic article].
Diabetes Care.
30. Zhao T, Markevych I, Romanos M, et al. Ambient ozone exposure and mental
health: A systematic review of epidemiological studies. Environ Res
2018;165:459-72.
31. Ornoy A, Weinstein-Fudim L, Ergaz Z. Prenatal factors associated with autism
spectrum disorder (ASD). Reprod Toxicol 2015;56:155-69.
32. Sorace A, de Acetis L, Alleva E, et al. Prolonged exposure to low doses of
ozone: short- and long-term changes in behavioral performance in mice. Environ
Res 2001;85(2):122-34.
33. Petruzzi S, Fiore M, Dell'Omo G, et al. Medium and long-term behavioral effects
in mice of extended gestational exposure to ozone. Neurotoxicol Teratol
1995;17(4):463-70.
34. Kavlock RJ, Meyer E, Grabowski CT. Studies on the developmental toxicity of
ozone: postnatal effects. Toxicol Lett 1980;5(1):3-9.

102
35. Bignami G, Musi B, Dell'Omo G, et al. Limited effects of ozone exposure during
pregnancy on physical and neurobehavioral development of CD-1 mice. Toxicol
Appl Pharmacol 1994;129(2):264-71.
36. Block ML, Elder A, Auten RL, et al. The outdoor air pollution and brain health
workshop. Neurotoxicology 2012;33(5):972-84.
37. Rajagopalan S, Brook RD. Air pollution and type 2 diabetes: mechanistic
insights. Diabetes 2012;61(12):3037-45.
38. Goodrich AJ, Volk HE, Tancredi DJ, et al. Joint effects of prenatal air pollutant
exposure and maternal folic acid supplementation on risk of autism spectrum
disorder. Autism Res 2018;11(1):69-80.
39. Pagalan L, Bickford C, Weikum W, et al. Association of Prenatal Exposure to Air
Pollution With Autism Spectrum Disorder. JAMA Pediatr 2018.
40. Volk HE, Kerin T, Lurmann F, et al. Autism spectrum disorder: interaction of air
pollution with the MET receptor tyrosine kinase gene. Epidemiology
2014;25(1):44-7.
41. Kim D, Volk H, Girirajan S, et al. The joint effect of air pollution exposure and
copy number variation on risk for autism. Autism Res 2017;10(9):1470-80.
42. Koebnick C, Langer-Gould AM, Gould MK, et al. Sociodemographic
characteristics of members of a large, integrated health care system: comparison
with US Census Bureau data. Perm J 2012;16(3):37-41.
43. Rothman K, Greenland S, Lash TL. Modern Epidemiology, 3rd Edition.
Philadelphia, PA: Lippincott Williams & Wilkins.; 2008.

103

Table 4.1. Associations of maternal and child characteristics with risk of ASD
Characteristics No. With ASD/Total Hazard Ratio (95% CI)
a

Maternal
Age (per year) 1.03 (1.02-1.04)
Parity
b

0
1,123 / 96,964 1.00 (Reference)
1
802 / 78,494 0.87 (0.79-0.95)
≥2
511 / 66,837 0.65 (0.58-0.73)
Education
b

High school or lower
811 / 101,770 1.00 (Reference)
Some college
761 / 68,729 1.33 (1.19-1.48)
College graduate or
higher
888 / 73,699 1.31 (1.17-1.46)
Household annual
income
b,c

<$30,000
223 / 20,069 1.32 (1.12-1.57)
$30,000-$49,999
845 / 82,638 1.00 (Reference)
$50,000-$69,999
810 / 78,926 0.91 (0.82-1.02)
$70,000-$89,999
365 / 39,758 0.77 (0.67-0.89)
≥$90,000
228 / 25,029 0.70 (0.59-0.82)
Race/ethnicity

Non-Hispanic white
569 / 62,774 1.00 (Reference)
Non-Hispanic black
266 / 23,855 1.09 (0.92-1.30)
Hispanic
1,216 / 126,123 0.95 (0.85-1.07)
Asian/Pacific Islander
374 / 29,774 1.13 (0.98-1.31)
Other
46 / 3,894 1.26 (0.90-1.74)
History of comorbidity
d

No
2,170 / 224,366 1.00 (Reference)
Yes
301 / 22,054 1.32 (1.15-1.51)
Child
Female 441 / 120,112 1.00 (Reference)
Male 2,030 / 126,308 4.77 (4.28-5.32)
a
Specified family as a random effect, and birth year and KPSC medical center service areas were
adjusted for as covariates.
b
Hazard ratios not reported for children with missing data categories in these variables.
c
Based on census tract median
d
≥1 diagnosis of heart, lung, kidney, or liver disease; cancer in mothers

104
Table 4.2. Minimally adjusted and fully adjusted hazard ratios (HRs) with 95%
confidence intervals (CIs) for associations of categories of maternal diabetes with risk of
ASD
Diabetes during
pregnancy
No. With ASD/Total Minimally adjusted
a

HR (95% CI)
Fully adjusted
b

HR (95% CI)
No diabetes 2,167 / 221,330 1.00 (Reference) 1.00 (Reference)
GDM ≥24 weeks’
gestation
160 / 16,112 1.01 (0.85-1.20) 0.92 (0.77-1.09)
GDM <24 weeks’
gestation
73 / 4,893 1.40 (1.08-1.81) 1.24 (0.95-1.62)
Pre-existing type 2
diabetes
71 / 4,085 1.65 (1.26-2.16) 1.45 (1.11-1.91)
a
Models for minimally adjusted hazard ratios specified family as a random effect, and birth year and
KPSC medical center service areas were adjusted for as covariates.
b
Models for fully adjusted hazard ratios were adjusted for birth year, KPSC medical center service areas,
maternal age, parity, maternal race/ethnicity, maternal education, census tract median household income,
maternal history of comorbidities before pregnancy (≥1 diagnosis of heart, lung, kidney, liver disease or
cancer), child sex, and family specified as a random effect.
105
Table 4.3. Minimally adjusted and fully adjusted hazard ratios (HRs) with 95%
confidence intervals (CIs) for the associations of each pollutant birth address exposure
with risk of ASD
Pollutant exposure window Minimally adjusted
a

HR (95% CI)
Fully adjusted
b

HR (95% CI)
O3
Preconception 0.99 (0.93-1.04) 0.98 (0.93-1.04)
Entire pregnancy 1.13 (0.99-1.28) 1.10 (0.95-1.26)
First trimester 0.97 (0.92-1.03) 0.97 (0.91-1.02)
Second trimester 1.04 (0.98-1.10) 1.03 (0.97-1.10)
Third trimester 1.05 (1.00-1.11) 1.05 (0.99-1.11)
First year of life 1.00 (0.87-1.16) 0.94 (0.80-1.11)
PM2.5
Preconception 1.13 (1.06-1.22) 1.11 (1.03-1.20)
Entire pregnancy 1.18 (1.06-1.32) 1.17 (1.04-1.33)
First trimester 1.11 (1.03-1.20) 1.10 (1.02-1.19)
Second trimester 1.07 (0.99-1.15) 1.06 (0.97-1.14)
Third trimester 1.09 (1.01-1.18) 1.08 (1.00-1.18)
First year of life 1.22 (1.07-1.39) 1.21 (1.05-1.40)
PM10
Preconception 1.06 (0.97-1.15) 1.05 (0.96-1.14)
Entire pregnancy 1.03 (0.91-1.16) 1.01 (0.89-1.15)
First trimester 1.02 (0.93-1.11) 1.00 (0.92-1.10)
Second trimester 1.03 (0.94-1.12) 1.02 (0.93-1.12)
Third trimester 1.01 (0.93-1.11) 1.00 (0.91-1.10)
First year of life 1.06 (0.94-1.21) 1.06 (0.92-1.22)
NO2
Preconception 1.09 (1.01-1.18) 1.07 (0.99-1.17)
Entire pregnancy 1.08 (0.96-1.22) 1.05 (0.91-1.20)
First trimester 1.04 (0.97-1.13) 1.03 (0.95-1.11)
Second trimester 1.02 (0.95-1.10) 1.01 (0.93-1.09)
Third trimester 1.04 (0.96-1.12) 1.02 (0.94-1.11)
First year of life 1.15 (1.00-1.31) 1.12 (0.97-1.30)
Note: Separate models were estimated for each time window. Hazard ratios were scaled per 6.5 μg/m
3

PM2.5; per 16.1 μg/m
3
PM10; per 10.4 ppb NO2; and per 15.7 ppb O3.
a
Models for minimally adjusted hazard ratios specified family as a random effect, and birth year and
KPSC medical center service areas were adjusted for as covariates.
b
Models for fully adjusted hazard ratios were adjusted for adjusted for birth year, KPSC medical center
service areas, maternal age, parity, maternal race/ethnicity, maternal education, census tract median
household income, maternal history of comorbidities before pregnancy (≥1 diagnosis of heart, lung,
kidney, liver disease or cancer), child sex, and family specified as a random effect.

106
Table 4.4. Maternal diabetes-specific adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) for O3 per 15.7
ppb within each exposure window associated with risk of ASD
No diabetes Gestational diabetes mellitus Pre-existing type
2 diabetes
P for interaction
Diagnosis at ≥24
weeks’ gestation
Diagnosis at <24
weeks’ gestation

No. With ASD/Total 2,167 / 221,330 160 / 16,112 73 / 4,893 71 / 4,085

O3 exposure window HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI)

Preconception 0.98 (0.92-1.03) 0.95 (0.77-1.17) 1.14 (0.76-1.72) 1.09 (0.85-1.38) 0.707
Pregnancy 1.10 (0.95-1.27) 0.83 (0.53-1.29) 1.96 (1.24-3.11) 1.14 (0.53-2.46) 0.067
First trimester 0.95 (0.90-1.01) 0.90 (0.69-1.18) 1.50 (1.08-2.09) 1.07 (0.69-1.64) 0.047
*

Second trimester 1.04 (0.98-1.11) 0.86 (0.72-1.03) 1.29 (0.95-1.76) 1.05 (0.66-1.65) 0.165
Third trimester 1.06 (0.99-1.12) 1.03 (0.78-1.35) 0.97 (0.73-1.30) 0.99 (0.73-1.35) 0.847
First year of life 0.93 (0.78-1.10) 0.72 (0.50-1.02) 2.01 (0.67-6.07) 1.17 (0.63-2.17) 0.007
*

Note: Separate interaction models were estimated for each exposure window,

which adjusted for birth year, KPSC medical center service areas,
maternal age, parity, maternal race/ethnicity, maternal education, census tract median household income, maternal history of comorbidities before
pregnancy (≥1 diagnosis of heart, lung, kidney, liver disease or cancer), child sex, and family specified as a random effect. *Global P for
interaction of each exposure window with 4-level categorical variable of maternal diabetes <0.05.

107
Table 4.5. Adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) for joint associations of exposures to
maternal diabetes and tertiles of O3 with risk of ASD
No diabetes Gestational diabetes mellitus Pre-existing type 2
diabetes
Diagnosis at ≥24
weeks’ gestation
Diagnosis at <24
weeks’ gestation
a

HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI)
First trimester
Low tertile
(<37.7 ppb)
Reference 0.94 (0.70-1.26) 0.87 (0.53-1.44) 1.54 (0.97-2.44)
Middle tertile
(37.7-<44.3 ppb)
0.97 (0.87-1.09) 1.01 (0.75-1.34) 1.10 (0.69-1.74) 1.37 (0.87-2.17)
High tertile
(≥44.3 ppb)
0.94 (0.84-1.06) 0.71 (0.50-1.00) 1.71 (1.12-2.60) 1.33 (0.81-2.18)
First year of life
Low tertile
(<37.7 ppb)
Reference 1.07 (0.81-1.41) 0.87 (0.54-1.39) 1.41 (0.91-2.20)
Middle tertile
(37.7-<44.3 ppb)
1.21 (1.06-1.38) 1.10 (0.82-1.48) 1.65 (1.08-2.51) 1.60 (0.98-2.60)
High tertile
(≥44.3 ppb)
0.97 (0.80-1.18) 0.69 (0.46-1.05) 1.68 (1.01-2.81) 1.66 (1.01-2.75)
a
P for linear trend of HRs among mothers with GDM diagnosis at <24 weeks’ gestation <0.05 during first trimester and first year of life
Note: Models were adjusted for birth year, KPSC medical center service areas, maternal age, parity, maternal race/ethnicity, maternal education,
household income, maternal history of comorbidities before pregnancy, and family specified as a random effect.

108
Figure 4.1. Adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) for joint associations of exposures to GDM
diagnosis at <24 weeks’ gestation and tertiles of O3 with risk of ASD relative to no diabetes and low tertiles of O3
exposure.

Note: Separate models were estimated for O3 exposures during the first trimester and first year of life, and estimates were
adjusted for birth year, KPSC medical center service areas, maternal age, parity, maternal race/ethnicity, maternal
education, census tract median household income, maternal history of comorbidities before pregnancy, and family
specified as a random effect. Circles represent the HRs and whiskers represent the 95% CIs. Each O3 tertile group is
labeled as low, middle, high on the x-axis. P for linear trend of HRs among mothers with GDM diagnosis at <24 weeks’
gestation <0.05.

109
Table S4.1. Exposure distributions during pregnancy by maternal diabetes status in 1999-2009
No diabetes
a
Gestational diabetes
mellitus ≥24 weeks
a

Gestational diabetes
mellitus <24 weeks
a

Pre-existing type 2
diabetes
a

O3 (ppb)
Mean (SD) 41.6 (8.1) 41.7 (8.1) 40.8 (7.6) 41.8 (7.9)
Median 40.9 41.1 40.4 41.3
PM2.5 (µg/m
3
)
Mean (SD) 17.9 (4.7) 17.7 (4.6) 17.5 (4.3) 16.8 (4.2)
Median 17.2 17.0 16.7 16.0
PM10 (µg/m
3
)
Mean (SD) 38.1 (9.1) 38.2 (9.3) 37.2 (8.6) 36.2 (8.2)
Median 36.5 36.5 35.6 35.0
NO2 (ppb)
Mean (SD) 25.2 (7.0) 24.8 (6.9) 25.0 (6.9) 24.2 (7.1)
Median 24.7 24.2 24.4 23.5
a
Based on 221,330 without diabetes, 16,112 with GDM ≥24 weeks, 4,893 with GDM <24 weeks, and 4,085 with pre-existing type 2 diabetes
SD: standard deviation; IQR = interquartile range; Q1-Q3: first quartile to third quartile range

110
Table S4.2. P-values of global interaction between each pollutant exposure and the 4-level categorical variable for
maternal diabetes during pregnancy
Preconception Entire
pregnancy
First trimester Second
trimester
Third trimester First year of
life
PM2.5 0.979 0.937 0.366 0.700 0.749 0.744
PM10 0.752 0.660 0.529 0.929 0.155 0.092
NO2 0.716 0.870 0.187 0.650 0.758 0.706
O3 0.707 0.067 0.047 0.165 0.847 0.007
Note: Test for significance at p <0.05. Significant p-values are bolded. Separate interaction models were estimated for each exposure window that
adjusted for birth year, KPSC medical center service areas, maternal age, parity, maternal race/ethnicity, maternal education, household income,
maternal history of comorbidities before pregnancy, child sex, and specified family as random effect. Maternal diabetes was categorized as none,
GDM diagnosed at ≥24 weeks’ gestation, GDM diagnosed at <24 weeks’ gestation, and pre-existing type 2 diabetes. Preconception exposure
window calculated as 12 weeks before mother’s last menstrual period date.

111
Table S4.3. Pearson partial correlations of pollutants during first trimester and first year of life exposure windows

First trimester
a
First year of life
a

PM2.5 PM10 NO2 O3 PM2.5 PM10 NO2 O3
First trimester
PM2.5 1.00 0.66 0.63 -0.14 0.66 0.46 0.44 -0.04
PM10 1.00 0.49 0.15 0.53 0.63 0.39 -0.05
NO2 1.00 -0.42 0.51 0.41 0.69 -0.29
O3 1.00 -0.06 -0.05 -0.18 0.31
First year of life
PM2.5 1.00 0.72 0.69 -0.18
PM10 1.00 0.55 -0.10
NO2 1.00 -0.49
O3 1.00
a
First trimester and first year of life exposure windows had ASD associations with interactions of O3 and 4-level categorical variable for maternal
diabetes (from Table S4.2).
Note: Maternal diabetes was categorized as none, GDM <24 weeks, GDM >= 24 weeks, and pre-existing type 2 diabetes. All correlations were
adjusted for birth year and KPSC medical center service areas and were statistically significant (p < 0.0001). Gray cells indicate correlations
between pollutants across exposure windows during first trimester and first year of life.

112
Figure S4.1. Kaiser Permanente Southern California (KPSC) medical center service areas

Note: KPSC service areas contributing to the analysis shown in the legend. Black dots indicate air quality monitoring stations that had data for one
or more pollutants during the study period. KPSC covers Imperial, Kern, Los Angeles, Orange, Riverside, San Bernardino, San Diego, San Luis
Obispo, Santa Barbara, and Ventura counties. Of the18 geographically located KPSC medical center service areas across Southern California, 4
were excluded (Kern, Antelope Valley, West Ventura, and Coachella Valley) because there were few cases of ASD.

113

Figure S4.2. Crude incidence rate of ASD by birth year

1.93
2.51
2.48
2.93
3.30
3.60
3.06
3.43
3.28
3.76
4.03
0.00
0.50
1.00
1.50
2.00
2.50
3.00
3.50
4.00
4.50
1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
Incidence Rate of ASD per 1,000 Person-Years
Birth Year

114
Figure S4.3. Distribution of pollutant concentrations during pregnancy across birth year 1999-2009

115
Figure S4.4. Distribution of 1999-2009 mean pollutant concentrations across KPSC medical center service areas

116

Table S4.4. Entire cohort with no random effect: maternal diabetes-specific adjusted hazard ratios (HRs) and 95%
confidence intervals (CIs) for ASD in associations with each pollutant within each exposure window
No diabetes Gestational diabetes mellitus Pre-existing type
2 diabetes
P for interaction
Diagnosis at ≥24
weeks’ gestation
Diagnosis at <24
weeks’ gestation

No. with ASD / Total 2,167 / 221,330 160 / 16,112 73 / 4,893 71 / 4,085
Exposure window HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI)
PM2.5
Preconception 1.10 (1.03-1.18) 1.14 (0.93-1.38) 1.08 (0.79-1.47) 1.17 (0.83-1.64) 0.97
Pregnancy 1.15 (1.03-1.28) 1.21 (0.94-1.56) 1.10 (0.75-1.63) 1.03 (0.68-1.56) 0.92
First trimester 1.10 (1.03-1.19) 1.12 (0.92-1.38) 0.93 (0.67-1.30) 0.88 (0.62-1.26) 0.48
Second trimester 1.04 (0.96-1.12) 1.18 (0.95-1.45) 1.05 (0.76-1.46) 0.93 (0.66-1.33) 0.62
Third trimester 1.07 (0.99-1.15) 1.02 (0.83-1.26) 1.14 (0.84-1.55) 1.16 (0.84-1.62) 0.89
First year of life 1.17 (1.02-1.33) 1.30 (0.99-1.70) 0.99 (0.64-1.52) 1.22 (0.79-1.88) 0.72
PM10
Preconception 1.03 (0.94-1.12) 1.00 (0.77-1.29) 1.07 (0.72-1.58) 1.26 (0.84-1.88) 0.80
Pregnancy 1.00 (0.89-1.13) 0.92 (0.66-1.28) 0.87 (0.52-1.46) 1.24 (0.74-2.09) 0.75
First trimester 1.00 (0.92-1.09) 0.97 (0.75-1.26) 0.87 (0.57-1.33) 1.33 (0.89-1.98) 0.49
Second trimester 1.02 (0.93-1.11) 1.02 (0.78-1.33) 1.03 (0.68-1.56) 0.86 (0.54-1.36) 0.91
Third trimester 1.00 (0.92-1.09) 0.85 (0.65-1.12) 0.83 (0.54-1.28) 1.34 (0.89-2.03) 0.26
First year of life 1.02 (0.90-1.17) 1.00 (0.71-1.42) 0.69 (0.39-1.25) 1.73 (1.02-2.93) 0.13
NO2
Preconception 1.07 (0.99-1.15) 1.05 (0.86-1.29) 0.89 (0.65-1.22) 1.01 (0.73-1.40) 0.72
Pregnancy 1.02 (0.91-1.16) 1.12 (0.87-1.46) 0.95 (0.65-1.37) 0.94 (0.65-1.38) 0.80
First trimester 1.02 (0.95-1.10) 1.12 (0.91-1.37) 0.76 (0.56-1.04) 0.96 (0.70-1.31) 0.21
Second trimester 1.00 (0.93-1.08) 1.14 (0.93-1.40) 1.04 (0.77-1.39) 0.92 (0.68-1.26) 0.59
Third trimester 1.01 (0.94-1.09) 0.96 (0.78-1.19) 1.12 (0.83-1.50) 0.98 (0.72-1.35) 0.87
First year of life 1.09 (0.95-1.24) 1.12 (0.85-1.49) 0.82 (0.54-1.26) 1.07 (0.72-1.60) 0.61
Note: Among entire cohort, separate interaction models were estimated for each exposure window,

which adjusted for birth year, KPSC medical
center service areas, maternal age, parity, maternal race/ethnicity, maternal education, household income, maternal history of comorbidities
before pregnancy, child sex (no random effect for family). Preconception exposure window calculated as 12 weeks before mother’s last menstrual
period date.

117
Table S4.5. Restricted to one child per family: maternal diabetes-specific adjusted hazard ratios (HRs) and 95%
confidence intervals (CIs) for ASD in association with each pollutant within each exposure window
No diabetes Gestational diabetes mellitus Pre-existing type
2 diabetes
P for interaction
Diagnosis at ≥24
weeks’ gestation
Diagnosis at <24
weeks’ gestation

No. with ASD / Total 1,693 / 173,227 124 / 12,967 62 / 3,990 63 / 3,264
Exposure window HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI)
PM2.5
Preconception 1.13 (1.04-1.22) 1.07 (0.86-1.35) 1.13 (0.81-1.58) 1.31 (0.93-1.86) 0.81
Pregnancy 1.19 (1.05-1.35) 1.21 (0.91-1.62) 1.06 (0.69-1.62) 1.14 (0.74-1.77) 0.95
First trimester 1.13 (1.04-1.22) 1.11 (0.88-1.40) 0.92 (0.64-1.31) 0.93 (0.64-1.36) 0.52
Second trimester 1.05 (0.96-1.15) 1.19 (0.93-1.50) 0.95 (0.66-1.37) 0.99 (0.68-1.45) 0.70
Third trimester 1.08 (0.99-1.19) 1.00 (0.79-1.28) 1.14 (0.82-1.59) 1.26 (0.89-1.77) 0.75
First year of life 1.23 (1.06-1.43) 1.36 (1.00-1.85) 1.01 (0.64-1.61) 1.44 (0.92-2.27) 0.63
PM10
Preconception 1.06 (0.96-1.17) 0.93 (0.69-1.26) 1.14 (0.75-1.74) 1.32 (0.87-2.02) 0.59
Pregnancy 1.00 (0.87-1.15) 0.89 (0.61-1.30) 0.83 (0.47-1.48) 1.34 (0.77-2.33) 0.58
First trimester 0.99 (0.90-1.09) 0.96 (0.71-1.30) 0.86 (0.54-1.36) 1.32 (0.86-2.02) 0.54
Second trimester 1.00 (0.90-1.11) 0.98 (0.72-1.33) 0.94 (0.59-1.50) 0.94 (0.58-1.53) 0.99
Third trimester 1.03 (0.94-1.14) 0.83 (0.60-1.13) 0.88 (0.55-1.41) 1.46 (0.94-2.24) 0.18
First year of life 1.07 (0.92-1.24) 1.03 (0.70-1.53) 0.73 (0.38-1.38) 1.99 (1.14-3.45) 0.09
NO2
Preconception 1.08 (0.99-1.18) 1.03 (0.82-1.30) 0.87 (0.62-1.22) 1.11 (0.79-1.55) 0.65
Pregnancy 1.07 (0.93-1.23) 1.16 (0.87-1.56) 0.91 (0.61-1.36) 1.01 (0.68-1.51) 0.76
First trimester 1.05 (0.96-1.14) 1.12 (0.89-1.41) 0.73 (0.52-1.01) 1.03 (0.74-1.43) 0.17
Second trimester 1.01 (0.93-.1.10) 1.16 (0.92-1.46) 0.98 (0.71-1.35) 0.94 (0.67-1.30) 0.66
Third trimester 1.03 (0.94-1.12) 0.98 (0.78-1.25) 1.09 (0.80-1.50) 1.01 (0.73-1.41) 0.96
First year of life 1.15 (0.99-1.33) 1.17 (0.85-1.62) 0.79 (0.50-1.25) 1.22 (0.80-1.86) 0.43
Note: Among subset of families with one child only, separate interaction models were estimated for each exposure window,

which adjusted for
birth year, KPSC medical center service areas, maternal age, parity, maternal race/ethnicity, maternal education, census tract median household
income, maternal history of comorbidities before pregnancy (≥1 diagnosis of heart, lung, kidney, liver disease or cancer), and child sex.
Preconception exposure window calculated as 12 weeks before mother’s last menstrual period date.

118
CHAPTER 5. CONCLUSIONS
SUMMARY OF MAJOR FINDINGS
Autism spectrum disorder (ASD) is a complex disorder, disproportionately impacting
more boys than girls, with likely numerous causal pathways involving genetics,
environment, pregnancy, and lifestyle factors. Most environmental studies, however,
have focused on associations of single risk factors, partly because they did not have the
adequate sample size to study combinations of factors. Additionally, few studies have
examined child sex-specific effects of environmental exposures. This dissertation
addressed these important gaps in environmental neuroepidemiology literature that
were previously discussed in Chapter 1, using electronic medical records data from a
large Kaiser Permanente Southern California (KPSC) pregnancy cohort. The overall
objective of this dissertation was to clarify the causal relationships between air pollution,
gestational diabetes mellitus (GDM), and ASD.
In Aim 1, we evaluated interactions between child sex and PM2.5 on ASD risk
motivated by findings from recent animal studies (1, 2). We observed increased ASD
risk associated with exposure to PM2.5 during several exposure windows including the
entire pregnancy, first trimester, third trimester, and first year of life. Only during the first
trimester did the PM2.5 association remain robust after mutually adjusting for other
exposure windows, and was specific to boys only. There was a small negative
association in girls that was not statistically significant. We also did not observe any
associations with other pollutants.
In Aim 2, we evaluated the effects of preconception and first trimester air
pollution exposures on GDM. In single-pollutant models, we observed that the

119
preconception NO2 was associated with increased risk of GDM, and this association
remained robust to adjustment for first trimester NO2 with another co-pollutant from both
exposure windows. These findings suggested that maternal NO2 exposure prior to
conception may increase the risk of development of GDM.
In Aim 3, we evaluated whether the effect of air pollution exposures on early life
ASD risk varied by maternal diabetes status. We observed statistically significant
interactions between maternal diabetes and O3 during the first trimester and first year of
life. O3 exposure during the first trimester was associated with increased risk of ASD
only among mothers with GDM diagnosed <24 weeks’ gestation. No statistically
significant O3 associations of ASD were observed in offspring of pregnancies without
diabetes, GDM diagnosed ≥24 weeks’ gestation, or pre-gestational diabetes. We also
did not observe statistically significant interactions with other pollutants. These findings
suggest that GDM diagnosed earlier in pregnancy may increase a child’s susceptibility
to prenatal O3-associated ASD risk.

CONCLUSIONS AND PUBLIC HEALTH IMPLICATIONS
Findings from this dissertation were based on specific hypotheses that were generated
from previous toxicological and epidemiologic studies designed to identify ASD risk
factors involved in common biological pathways. We addressed important knowledge
gaps in the literature by identifying vulnerable developmental windows of exposure to
ambient air pollutants, and by examining two common risk factors working in
conjunction to affect ASD risk. Our sex-specific findings from Aim 1 suggest that there
may be sex-specific biologic or genetic differences that exacerbate ASD risk among

120
boys, and also indicate that further study is needed to better understand the
environmental influences on sex-related differences in ASD.
This dissertation also addressed an important gap in the literature regarding
preconception effects of environmental exposures on GDM in Aim 2. Our findings
suggest that reducing NO2 exposure before conception may reduce GDM risk. The
public health implications from these results are significant, since GDM increases the
risk of subsequent development of maternal type 2 diabetes after pregnancy, and may
also increase the risk of childhood obesity.
Findings from Aim 3 suggest that developing GDM before the end of the second
trimester of pregnancy may increase children’s susceptibility to prenatal and early life
O3-associated ASD risk. However, these findings should be interpreted with caution
since we did not observe any main effects of O3 on ASD risk, and they need to be
replicated in future studies. Further toxicological studies are needed to better
understand the biological pathways for O3 effects on ASD risk during specific
neurodevelopmental periods of vulnerability. Additionally, such studies can provide
more insight into the differences in neurobehavioral effects of exposures to O3 by
maternal diabetes.
The findings from this dissertation may provide support for interventions that
target modifiable environmental factors such as air pollution to reduce the incidence of
ASD. Since ASD is a multifactorial disease, intervening for any modifiable risk factor
that is part of a causal combination of factors necessary for disease development would
prevent the disease. These findings also highlight the importance of receiving

121
preconception care that may prevent pregnancy complications such as GDM, which
may further reduce ASD incidence.

FUTURE DIRECTIONS
Our large pregnancy cohort from KPSC provided a unique opportunity to assess the
synergistic effects of air pollution exposure with child sex and maternal diabetes that
could potentially provide more insight into the changing trajectory of ASD incidence over
time. However, as discussed in Chapter 1, the development of standardized methods in
exposure and outcome assessments for increasing the comparability of results across
epidemiologic studies evaluating the effects of air pollution on ASD remains a
challenge. Although there were large exposure gradients in our cohort, we did not have
the resources to improve the exposure assessment by using prospectively collected
residential address history and thus, used the child’s birth address to estimate regional
air pollution exposure concentrations to use as a proxy for personal maternal exposure.
To improve trimester-specific exposure assessment, future air pollution studies may
consider using prospectively collected residential address history to account for mobility
during pregnancy and early life. Taking into account exposures during the time the
mother or child spent at locations away from home would also reduce exposure
measurement error.
Furthermore, more studies are needed to assess the joint effects of air pollution
in parallel with other risk factors to better understand the complex causal pathways
affecting ASD risk. As discussed in Chapter 1, future studies using EMR data from
KPSC or other health care organizations will enable researchers to construct large

122
population-based cohorts with residential addresses that are updated prospectively at
the time of the clinical contacts, and standardized and validated clinical diagnostic
algorithms to accurately assess ASD. Moreover, exploratory hypothesis-generating
analyses may be conducted using bioinformatics approaches for clinical information
available in the EMR. For example,
The biologic relationship of air pollution exposure to GDM and ASD remains
poorly understood. Neuroimaging tools could be used in future studies to examine the
structural changes in the brain that are caused by environmental stressors associated
with GDM and ASD. Additionally, identification of biological markers related to oxidative
stress or systemic inflammation in future studies may clarify the understanding of
mechanisms underlying air pollution-induced ASD effects.
This dissertation provided preliminary data for a recent R01 application (scored
at 6
th
percentile) that, if funded, will investigate the following aims:
1. Effects of prenatal and early life exposure to PM0.1 and PM2.5 and
composition on ASD in an expanded KPSC cohort with prospectively
assessed address history, including maternal workplace in a sample.
2. Effect modification by indicators of maternal immune activation, including
prepregnancy obesity, infection, diabetes, asthma,
preeclampsia/eclampsia.
3. PM effects on early life biological markers of oxidative stress, and their
association with early life neurodevelopmental phenotype, using a novel
procedure for assessing the adductome in dried blood spots archived at
birth for all children born in California.

123
REFERENCES

1. Church JS, Tijerina PB, Emerson FJ, et al. Perinatal exposure to concentrated
ambient particulates results in autism-like behavioral deficits in adult mice.
Neurotoxicology 2017.
2. Li K, Li L, Cui B, et al. Early postnatal exposure to airborne fine particulate matter
induces autism-like phenotypes in male rats. Toxicol Sci 2017.

Abstract (if available)

Abstract In the United States (U.S.), autism spectrum disorder (ASD) prevalence has increased over several decades among young children, disproportionately affecting more boys than girls. The effects of air pollution and gestational diabetes mellitus (GDM) have been hypothesized to increase the risk of ASD through common biologic mechanisms including oxidative stress and inflammation that could alter fetal brain development. Particulate matter <2.5 in aerodynamic diameter (PM₂.₅) has shown relatively consistent adverse effects on ASD

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Evaluation of new methods for estimating exposure to traffic-related pollution and early health effects for large population epidemiological studies

PDF

Prenatal environmental exposures and fetal growth in the MADRES cohort

PDF

Associations between ambient air pollution and hypertensive disorders of pregnancy

PDF

Assessing the impact of air pollution on adverse birth outcomes in a low resource setting

PDF

Examining exposure to extreme heat and air pollution and its effects on all-cause, cardiovascular, and respiratory mortality in California: effect modification by the social deprivation index

PDF

Ambient air pollution and lung function in children

PDF

Air pollution, mitochondrial function, and growth in children

PDF

Sensor-based mobile health approaches for personal air pollution and pediatric asthma studies

PDF

Air pollution and childhood obesity

PDF

Personal exposure to particulate matter PM2.5 sources during pregnancy and birthweight

PDF

The development of type 2 diabetes mellitus (type 2 DM) in California twins

PDF

Disparities in exposure to traffic-related pollution sources by self-identified and ancestral Hispanic descent in participants of the USC Children’s Health Study

PDF

Assessment of the mortality burden associated with ambient air pollution in rural and urban areas of India

PDF

Spatial analysis of PM₂.₅ air pollution in association with hospital admissions in California

PDF

Susceptibility to autism likelihood associated with prenatal ambient air pollution exposures: identifying vulnerable populations

PDF

Street connectivity and childhood obesity: a longitudinal, multilevel analysis

PDF

Prenatal air pollution exposure, newborn DNA methylation, and childhood respiratory health

PDF

Effectiveness of individual- and household-level protective actions in reducing symptoms associated with hydrogen sulfide chronic low-level exposure

PDF

Association of traffic-related air pollution and lens opacities in the Los Angeles Latino Eye Study

PDF

Spatial modeling of non-tailpipe emissions and its association with children's lung function

Asset Metadata

Creator Jo, Heejoo (author)

Core Title Early life air pollution exposure, gestational diabetes mellitus, and autism spectrum disorder

School Keck School of Medicine

Degree Doctor of Philosophy

Degree Program Epidemiology

Publication Date 06/17/2019

Defense Date 04/11/2019

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag Air pollution,autism,gestational diabetes mellitus,OAI-PMH Harvest

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor McConnell, Rob (committee chair), Chen, Jiu-Chiuan (committee member), Cockburn, Myles (committee member), Habre, Rima (committee member), Xiang, Anny (committee member)

Creator Email heejoojo@gmail.com,heejoojo@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-176553

Unique identifier UC11660411

Identifier etd-JoHeejoo-7495.pdf (filename),usctheses-c89-176553 (legacy record id)

Legacy Identifier etd-JoHeejoo-7495.pdf

Dmrecord 176553

Document Type Dissertation

Format application/pdf (imt)

Rights Jo, Heejoo

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA