University of Southern California Dissertations and Theses

Air pollution and childhood obesity

(USC Thesis Other)

Air pollution and childhood obesity

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

AIR POLLUTION AND CHILDHOOD OBESITY

by

Jeniffer S. Kim, MPH

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

December 2019

Copyright 2019 Jeniffer S. Kim

ii

Dedication

To my mother, Diane Kim, and my late father, Steve Kim:
Without your sacrifice and perseverance, I would not be where I am today. Thank you for
supporting my dreams and guiding me in the right direction. This work is dedicated to you both.

iii

Acknowledgements

I would like to express my sincerest gratitude to all my committee members. To my
mentor, Dr. Frank Gilliland, thank you for your guidance and support over the years. The
countless meetings, progress reviews, and feedback have been invaluable. Thank you for always
challenging me and giving me the freedom to become an independent researcher. To my other
committee members, Drs. Kiros Berhane, Sandrah Eckel and Michael Goran, many thanks for
your support, encouragement, and guidance.
I would also like to acknowledge the help and support of many colleagues who provided
countless hours of review and feedback. My deepest gratitude to Drs. Tanya Alderete, Zhanghua
Chen, Claudia Toledo-Corral and Robert Urman for going beyond your post-doctoral duties by
helping me in so many ways. I would also like to thank Ed Rappaport, Dr. Rima Habre, Fred
Lurmann, and Monica Solares for their dedication to the Children’s Health Study and Meta-AIR
Study. I would also like to express my appreciation for all study participants and staff; without
you, this dissertation would not have been be possible.
Finally, I would like to thank my family and friends who have encouraged and supported
me during my studies. To my dear husband, Jay Chang, thank you for allowing me to pursue my
dreams, and I promise this will be my last degree. To my sister, Irene Kim, thank you for being
my biggest cheerleader. And lastly to my mother, Diane Kim, you are the reason why I have
achieved success in my life. With the upmost gratitude to all, I thank each and everyone one of
you.

iv

Table of Contents
Dedication .......................................................................................................................... ii
Acknowledgements .......................................................................................................... iii
List of Tables .................................................................................................................... vi
List of Figures ................................................................................................................... ix
Dissertation Abstract ........................................................................................................ x
Chapter 1: Background and Review ............................................................................... 1
1.1 Childhood Obesity ..................................................................................................... 1
1.1.1 Prevalence and trends ....................................................................................... 1
1.1.2 Childhood body mass index trajectory and obesity ......................................... 1
1.1.3 Health consequences of childhood obesity ...................................................... 3
1.2 Risk Factors for Childhood Obesity .......................................................................... 5
1.2.1 Diet and physical activity ................................................................................ 5
1.2.2 Genetics and other risk factors ......................................................................... 5
1.3 Air Pollution Exposure Assessment .......................................................................... 6
1.3.1 Direct and indirect measurements ................................................................... 6
1.3.2 Measurement error ......................................................................................... 10
1.4 Air Pollution, Obesity and Cardiometabolic Risk Factors ...................................... 12
1.4.1 Early life air pollution exposure and childhood obesity ............................... 13
1.4.2 Air pollution and cardiometabolic risk factors ............................................... 14
1.5 Biological Mechanisms of Air Pollution on Obesity ............................................... 16
1.6 Dissertation Goals.................................................................................................... 18
1.7 References ............................................................................................................... 20
1.8 Figures ..................................................................................................................... 27
Chapter 2: Longitudinal Associations of In Utero and Early Life Near-Roadway Air
Pollution with Trajectories of Childhood Body Mass Index ....................................... 29
2.1 Abstract .................................................................................................................... 30
2.2 Introduction ............................................................................................................. 31
2.3 Materials and Methods ............................................................................................ 33
2.4 Results ..................................................................................................................... 38
2.5 Discussion ................................................................................................................ 41
2.6 References ............................................................................................................... 45
2.7 Tables....................................................................................................................... 48
2.8 Figures ..................................................................................................................... 53
2.9 Supplemental Tables................................................................................................ 55
Chapter 3: Early Life Near-Roadway Air Pollution, Genetic Susceptibility, and Childhood
Body Mass Index Growth: Analysis of a Genetic Risk Score-Environment Interaction
........................................................................................................................................... 66
3.1 Abstract .................................................................................................................... 66
3.2 Introduction ............................................................................................................. 68
3.3 Materials and Methods ............................................................................................ 69
3.4 Results ..................................................................................................................... 75

v

3.5 Discussion ................................................................................................................ 80
3.6 References ............................................................................................................... 83
3.7 Tables....................................................................................................................... 85
3.8 Figures ..................................................................................................................... 96
3.9 Supplemental Tables................................................................................................ 97
Chapter 4: Associations of Air Pollution, Obesity and Cardiometabolic Health in Young
Adults: The Meta-AIR Study ....................................................................................... 111
4.1 Abstract .................................................................................................................. 111
4.2 Introduction ........................................................................................................... 113
4.3 Materials and Methods .......................................................................................... 114
4.4 Results ................................................................................................................... 121
4.5 Discussion .............................................................................................................. 127
4.6 References ............................................................................................................. 132
4.7 Tables..................................................................................................................... 135
4.8 Figures ................................................................................................................... 142
4.9 Supplemental Tables.............................................................................................. 144
Chapter 5: Summary and Future Research ............................................................... 155
5.1 Summary ................................................................................................................ 155
5.2 Future Direction .................................................................................................... 158
5.3 References ............................................................................................................ 162

vi

List of Tables
Chapter 2
Table 2.1: Baseline characteristics and age-adjusted associations with BMI growth 48
and BMI at age 10 years in children enrolled in the longitudinal Children’s Health Study.
Table 2.2 Residential NRAP exposures from freeway and non-freeway sources for 50
in utero, first year of life, and childhood periods in children in the CHS.
Table 2.3 Effects of in utero/first year of life and childhood freeway NOx exposures 51
on 4-year childhood BMI trajectories in CHS children.
Table 2.4 Effects of in utero/first year of life and childhood non-freeway NOx 52
exposures on 4-year childhood BMI trajectories in children in CHS.
Supplemental Table 2.1 Pearson Correlation Coefficients Between In Utero/ 55
First Year of Life and Childhood NRAP Exposure in Non-Movers

and Movers

in
Children in the Children’s Health Study.
Supplemental Table 2.2 Baseline Characteristics of Movers and Non-Movers 56
With Early Life NRAP Exposures and Who Were Enrolled in the Longitudinal
Children’s Health Study.
Supplemental Table 2.3 Effects of In Utero/First Year of Life and Childhood 58
Near-Road Total NOx Exposure on 4-Year Childhood BMI Trajectories.
Supplemental Table 2.4 Effects of In Utero/First Year of Life Near-Road Freeway 59
NOx on 4-Year Childhood BMI Trajectories for Male and Female Children.
Supplemental Table 2.5 Effects of In Utero/First Year of Life Near-Road Freeway 60
NOx on 4-Year Childhood BMI Trajectories for Non-Hispanic White/Hispanic Children.
Supplemental Table 2.6 Effects of In Utero/First Year of Life Near-Road Freeway 61
NOx on 4-Year Childhood BMI Trajectories for Baseline Overweight/Obese and
Normal BMI Children.
Supplemental Table 2.7 Effects of In Utero/First Year of Life Near-Road Freeway 62
NOx on 4-year Childhood BMI Trajectories Adjusting for Birth Weight and
Gestational Age.
Supplemental Table 2.8 Effects of In Utero/First Year of Life and Childhood 63
Ambient PM2.5 Exposure on 4-Year Childhood BMI Trajectories.
Supplemental Table 2.9 Independent Effects of In Utero or First Year of Life 64
Near-Road Freeway NOx Exposure on 4-Year Childhood BMI Trajectories in
Movers and Non-Movers.
Supplemental Table 2.10 Effects of Childhood, Near-Road Non-Freeway NOx 65
Exposure on 4-year Childhood BMI Trajectories for Children Enrolled in the
Children’s Health Study.

Chapter 3
Table 3.1 List of Studies and SNPs Used to Create the Obesity-Related 85
Genetic Risk Score (GRS)
Table 3.2 Baseline Characteristics of Children Enrolled in the Longitudinal Children's 86
Health Study
Table 3.3 Residential NRAP exposures from freeway and non-freeway sources for 88
in utero, first year of life, and childhood periods in children in the CHS.
Table 3.4a Associations of Obesity-Related GRS and Childhood BMI Trajectory 89

vii

Table 3.4b Associations of Obesity-Related GRS and Childhood BMI Trajectory in 90
Participants with Early Life NRAP Exposure
Table 3.5 Comparison of Associations of In Utero Freeway and Non-Freeway NOx 91
on Childhood BMI Trajectory with Chapter 2’s Sample versus GRS Sample
Table 3.6 Comparison of Associations of First Year of Life Freeway and Non-Freeway 92
NOx on Childhood BMI Trajectory with Chapter 2’s Sample versus GRS Sample
Table 3.6a Comparison of Associations of First Year of Life Freeway NOx 93
on Childhood BMI Trajectory with Chapter 2’s Sample versus GRS Sample by
Asthma Status and by Race/Ethnicity
Table 3.7 Associations of In Utero NRAP Exposures, Unweighted GRSxE and 94
Childhood BMI Trajectory in CHS Children
Table 3.8 Associations First Year of Life NRAP Exposures, Unweighted GRSxE and 95
Childhood BMI Trajectory in CHS Children
Supplemental Table 3.1 Pearson correlation coefficients between in utero/first year 97
of life and childhood NRAP exposure in non-movers and movers in children in the
Children’s Health Study.
Supplemental Table 3.2 Distribution of Unweighted and Weighted GRS Quintiles in 98
Participants with Early Life NRAP Exposures
Supplemental Table 3.3. Associations of Obesity-Related GRS and Childhood BMI 99
by Sex

Supplemental Table 3.4 Associations of Obesity-Related GRS and Childhood BMI 100
by Race/Ethnicity

Supplemental Table 3.5 Associations of In Utero NRAP Exposures, Weighted GRSxE 101
and Childhood BMI Trajectory in CHS Children
Supplemental Table 3.6 Associations of First Year of Life NRAP Exposures, Weighted 102
GRSxE and Childhood BMI Trajectory in CHS Children
Supplemental Table 3.7 Associations of In Utero NRAP, GRSxE 103
(Unweighted- Quintiles), and Childhood BMI
Supplemental Table 3.8 Associations of First Year of Life NRAP, GRSxE 105
(Unweighted-Quintiles), and Childhood BMI
Supplemental Table 3.9 Associations of In Utero NRAP, GRSxE (Weighted- Quintiles) 107
and Childhood BMI
Supplemental Table 3.10 Associations of First Year of Life NRAP, GRSxE 109
(Weighted-Quintiles), and Childhood BMI

Chapter 4
Table 4.1 Sociodemographic Characteristics by Obesity Status of 158 Participants 135
Enrolled in the Meta-AIR Study from 2014-2018.
Table 4.2 Obesity and Cardiometabolic Measures by Obesity Status in 158 Participants 136
Enrolled in the Meta-AIR Study from 2014-2018.
Table 4.3 Short- and Long-Term Regional Ambient and Near-Roadway Air Pollution 138
Exposures Among 158 Participants Enrolled in the Meta-AIR Study from 2014-2018.
Table 4.4 Associations of Short- and Long-Term Ambient Air Pollution Exposures with 139
Obesity Measures in 158 Participants Enrolled in the Meta-AIR Study from 2014-2018.
Table 4.5 Associations of Short-Term Ambient Air Pollution Exposures with 140
Cardiometabolic Measures in 158 Participants Enrolled in the Meta-AIR Study

viii

from 2014-2018.
Table 4.6 Associations of Long-Term Regional Ambient Air Pollution Exposures 141
with Cardiometabolic Measures in 158 Participants Enrolled in the Meta-AIR Study
from 2014-2018.
Supplemental Table 4.1 Historic Regional Ambient and Near-Roadway Air Pollution 144
Exposures Among 158 Participants Enrolled in the Meta-AIR Study from 2014-2018.
Supplemental Table 4.2 Spearman Correlations Between Prior 1-Month and 1-Year 145
Average Air Pollution Exposures with Historic Air Pollution Exposures.
Supplemental Table 4.3 Spearman Correlations Between Short-Term Ambient and 146
Near-Roadway Air Pollutants.
Supplemental Table 4.4 Spearman Correlations Between Long-Term Ambient and 147
Near-Roadway Air Pollutants.
Supplemental Table 4.5 Associations of Short-Term Ozone (O3) Exposures with 148
Liver Fat (HFF) and Lipid Measures in Multipollutant Models.
Supplemental Table 4.6 Associations of Long-Term NO2 Exposures with Lipid 149
Measures in Multipollutant Models.
Supplemental Table 4.7 Association of Long-Term PM2.5 Exposures with 150
Insulin AUC in Multipollutant Models.
Supplemental Table 4.8 Associations of Short-Term NRAP Exposures with Obesity 151
and Cardiometabolic Measures in 158 Participants Enrolled in the Meta-AIR Study
from 2014-2018.
Supplemental Table 4.9 Associations of Long-Term NRAP Exposures with Obesity 153
and Cardiometabolic Measures in 158 Participants Enrolled in the Meta-AIR Study
from 2014-2018.

ix

List of Figures

Chapter 1
Figure 1.1 Trends in U.S. Childhood Obesity: 1999-2015 27
Figure 1.2 Mechanistic Framework: How Air Pollution May Lead to Increased Risk 28
of Obesity.

Chapter 2
Figure 2.1 Map of Children’s Health Study communities. 53
Figure 2.2 Flow chart of children enrolled in the Children’s Health Study from 54
2002-2003 included and excluded from the current analysis

Chapter 3
Figure 3.1 Flow Chart of Children Enrolled in the Children’s Health Study from 96
2002-2003 Included and Excluded from the Current Analysis

Chapter 4
Figure 4.1 Meta-AIR Study

Flow 142
Figure 4.2 Associations of Prior 1-Year

NO2 Exposures and Lipid Metabolism Measures 143
by Obesity Status in 158 Participants Enrolled in the Meta-AIR Study from 2014-2018.

x

Dissertation Abstract
Childhood obesity continues to be a serious public health problem worldwide. In the
United States, prevalence of obesity has been increasing at an alarming rate over the past 40
years and has somewhat tapered since early 2000s (1). Global prevalence of childhood obesity
has also steadily increased with nearly a quarter of children being overweight or obese in
developed countries (2). The childhood obesity epidemic has grave public health implications
including a rise in obesity-related health burdens. Obese children are at a greater risk for
developing health complications including insulin resistance, type 2 diabetes, metabolic
syndrome and cardiovascular disease (3-5).
Despite decades of diet and physical activity interventions, prevalence of childhood
obesity has yet to decline suggesting that other factors may play an important role in obesity
prevention. Emerging studies show the importance of environmental pollutants like air pollution
as obesogens, chemicals that disrupt metabolic processes; however, few epidemiological studies
have investigated the longitudinal association of air pollution and childhood obesity.
Additionally, genetic susceptibility to obesity has been shown in previous studies as obesity
tracks within families (6). Taken together, synergistic associations of genetic susceptibility for
obesity and air pollution exposures with increased risk for obesity may exist.
The purpose of this dissertation is to explore the effects of air pollution on childhood
obesity by leveraging data from the Children’s Health Study (CHS) and the newly completed
Metabolic and Asthma Incidence Research (Meta-AIR) Study. In chapter 1, a general overview
is provided of the epidemiology of childhood obesity and associated risk factors. Then a
summary on air pollution exposure assessment is given looking at direct and indirect measures of
air pollution as well as potential types of measurement error. Next an overview of the literature

xi

on the associations of early life air pollution and childhood obesity as well as associations of air
pollution and metabolic risk factors was conducted. Moreover, a review of potential biological
mechanisms by which air pollution may affect obesity and metabolic risk factors was conducted.
Finally, the dissertation goals are outlined at the end of chapter 1.
In chapter 2, the longitudinal association of early life near-roadway air pollution (NRAP)
exposure with trajectories of childhood body mass index (BMI) in the CHS were examined.
Linear mixed effects models were used to estimate the effect of in utero or first year of life near-
road nitrogen oxide (NOx) exposure on childhood BMI growth and attained BMI at age 10 years.
NRAP consists of nitrogen oxides (NOx) from freeway and non-freeway sources. This statistical
model was unique because childhood NRAP exposures were also incorporated in the model.
Statistically significant associations were found with increased first year of life freeway NOx and
childhood BMI trajectory. These findings suggest that elevated early life NRAP exposures may
contribute to increased obesity risk in children. This work was published in Environmental
Health in September 2018 (7).
In chapter 3, the second project builds on the first project’s longitudinal model by
assessing the association of an obesity related genetic risk score (GRS) and its interaction with
early life NRAP exposures through a GRS-environment (GRSxE) interaction on childhood BMI
trajectory. First, a genome-wide association studies (GWAS) catalog derived GRS was created
using 67 single nucleotide polymorphisms (SNPs) that were associated with obesity in studies
with adults and children. Then the association of the obesity related GRS (unweighted, weighted,
and quintiles) on childhood BMI trajectory without any air pollution measures was examined.
Lastly, the full model with early life NRAP exposures, GRS, and the GRSxE interaction on
childhood BMI trajectory was explored. Both unweighted and weighted GRS were statistically,

xii

significantly associated with childhood BMI growth as well as attained BMI at age 10 years. No
significant associations with GRSxE on childhood BMI trajectory were found.
In chapter 4, further exploration of the effects of air pollution on obesity was conducted
by evaluating clinical data collected from the Meta-AIR Study. The Meta-AIR Study examined
the associations of short-term (prior 1-month) and long-term (prior 1-year) average air pollution
exposures on obesity and cardiometabolic risk factors in a subset of CHS participants. Between
2014-2018, participants underwent a study visit where measures of adiposity, glucose
metabolism and lipid metabolism were assessed. Effects of both ambient and NRAP exposures
on these various clinical outcomes were examined. Linear regression models showed that long-
term nitrogen dioxide (NO2) exposure was associated with higher levels of total cholesterol and
low-density lipoprotein cholesterol. These effects of long-term NO2 on fasting lipid measures
were more pronounced in obese subjects compared to non-obese subjects. This work was
accepted for publication by Environment International on September 10, 2019.
Lastly in chapter 5, summaries of the main findings of each project, the general
implications of this dissertation work, and suggestions for future research are provided.
Collectively, these projects provide much needed evidence for the association of air pollution
and childhood obesity by looking at effects of early life exposures and GRSxE interactions on
childhood BMI growth as well as short- and long-term exposures on obesity measures in young
adults. The following chapters will provide additional information on childhood obesity and air
pollution as well as more details of each of the projects presented.

xiii

Abstract References
1. Statistics NCfH. Health, United States, 2011: With Special Feature on Socioeconomic
Status and Health Hyattsville, MD, 2012,
2. Ng M, Fleming T, Robinson M, et al. Global, regional, and national prevalence of
overweight and obesity in children and adults during 1980-2013: a systematic analysis for
the Global Burden of Disease Study 2013. Lancet 2014;384(9945):766-81.
3. Weiss R, Dziura J, Burgert TS, et al. Obesity and the metabolic syndrome in children and
adolescents. N Engl J Med 2004;350(23):2362-74.
4. Hannon TS, Rao G, Arslanian SA. Childhood obesity and type 2 diabetes mellitus.
Pediatrics 2005;116(2):473-80.
5. Freedman DS, Mei Z, Srinivasan SR, et al. Cardiovascular risk factors and excess
adiposity among overweight children and adolescents: the Bogalusa Heart Study. J
Pediatr 2007;150(1):12-7.e2.
6. Silventoinen K, Rokholm B, Kaprio J, et al. The genetic and environmental influences on
childhood obesity: a systematic review of twin and adoption studies. Int J Obes (Lond)
2010;34(1):29-40.
7. Kim JS, Alderete TL, Chen Z, et al. Longitudinal associations of in utero and early life
near-roadway air pollution with trajectories of childhood body mass index. Environ
Health 2018;17(1):64.

1

Chapter 1: Background and Review
1.1 Childhood Obesity
1.1.1 Prevalence and trends
In 2012, approximately 12.7 million or 16.9% of children and adolescents between 2-19
years of age were obese and nearly a third of United States (U.S.) children were either
overweight or obese (1). Recently, the National Center for Health Statistics (NCHS) reported a
rise in childhood obesity prevalence with 18.5% in 2015-2016 versus 16.9% in 2011-2012 (2).
Though childhood obesity prevalence seem to have stabilized in the mid-2000s, a significant
increasing trend can be seen from 1999-2000 through 2015-2016 in Figure 1.1 (2). Additionally,
prevalence increases with age where 13.9% of preschool aged children (2-5 years), 18.4% of
school aged children (6-11 years), and 20.3% of adolescents (12-19 years) are obese. There was
no significant difference between boys and girls in the overall prevalence or by age group in the
recent NCHS report. Across all ages, Hispanic and non-Hispanic Black children have the highest
burden of obesity compared to non-Hispanic White and Asian children. For example, between
2013-2016 the prevalence of obesity was 23.6% in Hispanics and 20.4% in non-Hispanic Blacks
compared to 14.7% in Non-Hispanic Whites and 9.8% in Non-Hispanic Asians (3). High rates of
obesity place youth at a greater risk for a myriad of health consequences, which in turn leads to
acute and chronic health conditions later in life since obese children are more likely to become
obese adults (4, 5).

1.1.2 Childhood body mass index trajectory and obesity
In the U.S., the Centers for Disease Control (CDC) assesses overweight and obesity
status in children 2-19 years by age- and sex-specific growth curves. A child’s height and weight

2

are used to calculate body mass index (BMI) where BMI=weight/height
2
(kilograms/meter
2
)

and
then expressed as a percentile using the CDC growth chart which interprets a child’s BMI
relative to a reference group of children of the same age and sex (6). Childhood overweight and
obesity are defined as ≥85
th
BMI percentile (BMI %tile) and ≥95
th
BMI %tile, respectively. BMI
does have its limitations as it is not a direct measure of adiposity however in a research setting
where thousands of children’s measurements are needed, height and weight are the most feasible,
standardized measure.
Childhood BMI trajectories are important as these may predict future obesity risk. As
such, one study reported four types of BMI %tile growth patterns (high rising, median stable,
low to high, low stable) in a cohort of children followed from 24 months to 13 years of age,
where two classes (high rising and low to high) with steeper BMI %tile trajectories resulted in
risk of overweight/obese by age 13 years (7). Moreover, percent weight gained during infancy (0
to 15 months) predicted high-rising BMI trajectory across childhood independent of birth weight
(7). Similarly, in Australian children, differing early life growth trajectories (low, intermediate,
high, or accelerating) between birth to 3.5 years were examined. Compared to the intermediate
group, the high and accelerating groups were associated with increased body size that resulted in
increased odds of overweight/obesity at age 9 years (odds ratio (OR) high: 4.3, 95% confidence
interval (95% CI): 2.5,7.3; OR accelerating:15.4, 95% CI: 5.2,45.1) (8). These studies show that
childhood BMI trajectories can predict obesity in later years and may be used as an indicator of
future obesity risk. Obesity that tracks through childhood and into adolescence has also been
shown follow suit into adulthood. Recent simulations indicate increased risk of adult obesity for
overweight children where over half of today’s U.S. children will be obese by age 35 (9). This
draws much concern as there are various health consequences associated with obesity.

3

1.1.3 Health consequences of childhood obesity
Obese children are at a greater risk of detrimental health effects which include and are
not limited to insulin resistance (10-12), type 2 diabetes (T2DM) (13-16), cardiovascular disease
(CVD) (4, 17), and premature death (18). Insulin resistance is one of the most common
metabolic complications observed among obese children (10-12). Such metabolic consequences
of overweight and obesity disproportionally affect Hispanic/Latino children. For example, one
study found that over a fourth of overweight Latino children with a family history of T2DM had
impaired glucose tolerance and already showed poor beta-cell function as young as 8 years of
age (12). These metabolic changes increase the risk for T2DM since decreased insulin sensitivity
causes pancreatic beta-cells to produce more insulin to maintain normal blood glucose levels.
Specifically, with prolonged insulin resistance, many children suffer from prediabetes (fasting
glucose of between 100-125 mg/dL) which then may progress to beta-cell fatigue and
development of T2DM (13, 19).
While T2DM was once an adult disease, it is now being diagnosed among obese children
(14). Although T2DM prevalence amongst children is still low at 0.46 per 1000 children 10-19
years of age, prevalence in children increased by 35% from 2001-2009 (15), and researchers
estimate about a third of adolescent T2DM cases were undiagnosed according to National Health
and Nutrition Examination Survey (NHANES) data from 1999-2010 (20). Children who develop
T2DM at a young age also have increased risk for complications related to T2DM, such as
kidney failure, stroke, heart attack, and even sudden death (13). As rates of childhood T2DM are
on the rise, the SEARCH study projects a four-fold increase in the number of children with
T2DM over the next forty years (16).

4

Along with insulin resistance and T2DM, several studies have shown increased risk of
cardiovascular related outcomes associated with excess adiposity in children. Childhood obesity
has been associated with development of atherosclerosis with greater arterial stiffness, high
carotid artery intima-media thickness in adulthood, and calcification of the coronary artery (21-
23). Others have shown cardiovascular risk factors were associated with high adiposity which
included elevated risk for high blood pressure, higher levels of triglycerides, higher levels of
low-density lipoprotein cholesterol and lower levels of high-density lipoprotein (HDL)
cholesterol (4, 17). Dyslipidemia, elevated cholesterol or lipids in the blood, often affects
overweight/obese children and adolescents and is a known risk factor for CVD. Between 2011-
2014, NHANES data showed that 1 in 5 children and adolescents between 6-19 years of age had
an abnormal cholesterol measure that was either high total cholesterol, low HDL-cholesterol, or
high non-HDL-cholesterol with adolescents having a greater prevalence than children (24). This
prevalence increased substantially amongst overweight or obese children/adolescents with 22.3%
and 43.3%, respectively, compared to 13.8% prevalence of abnormal cholesterol measures in
normal weight children. As CVD is the leading cause of death among U.S. adults, it is worrisome
that children as young as 6 years have high blood cholesterol levels.
Childhood obesity not only has numerous health consequences but severe economic
implications as well. Increased rates of overweight and obesity contribute to declines in quality
of life and also have profound economic and social costs (25). Increased healthcare costs are
attributed to increased cases of obesity associated comorbidities such as diabetes, heart disease,
stroke, and cancer. If obesity trends continue, by 2030 the U.S. will spend $48-66 billion per
year treating obesity related cases (26). With mounting health implications and economic costs,

5

it is imperative to identify underlying causes of obesity for intervention and prevention efforts
which in turn will save lives and have tremendous economic benefits.

1.2 Risk Factors for Childhood Obesity
1.2.1 Diet and physical activity
Great efforts were made in the 1990s to understand the underlying causes of rapidly
growing number of obese children and adults. Traditionally, obesity has been described as an
energy imbalance, which favors caloric intake over energy expenditure. Therefore, physical
activity and diet have been identified as major risk factors for childhood obesity. Some
researchers, like Andersen et al., believe it is the inactivity that arises from sedentary behaviors
like television watching that contributes to childhood obesity (27). Others have looked at the
quantity and intensity of physical activity or lack thereof as contributing to childhood obesity
(28). Beyond this, others further speculate that specific foods, such as sugary beverages, have
uniquely contributed to increased rates of childhood obesity (29, 30). Since preventative
measures focusing on diet and physical activity have yet to curb the obesity epidemic, other risk
factors should be considered.

1.2.2 Genetics and other risk factors
Other than diet and physical activity, genetics also contributes to childhood obesity risk
as suggested by studies showing the hereditary nature of obesity (31). In a systematic review of
twin and adoption studies, genetic factors were found to strongly influence BMI (32). BMI
correlations of monozygotic twins were higher than correlations than dizygotic twins; similarly,
in adoption studies, adoptees and adoptive parents had a weaker correlation with BMI than with

6

biological children. Genome wide associations studies (GWAS) have reported several loci
associated with BMI and obesity (33-36). In past decade, single nucleotide polymorphisms
(SNP) within specific genes have been identified to be associated with BMI and obesity. Two of
these genes are fat mass and obesity associated gene (FTO) (37, 38) and melanocortin 4 receptor
gene (MC4R) (39, 40). Since then, several other genes have been identified (33, 36). Though
BMI has been shown to be highly heritable ranging from ~0.4-0.7 (41, 42), the majority of
genetic liability for obesity remains unknown as only 1.45% of the variance in BMI has been
explained by gene variants (35). As such, other factors like gene-environment interactions may
be contributing to increased obesity risk.
Aside from genetics, other findings suggest that high pre-pregnancy, maternal BMI (43,
44), excessive gestational weight gain (43, 45), and maternal smoking during pregnancy (46) are
additional risk factors for childhood obesity. After birth, small for gestational age infants are
more likely to be become overweight compared to infants born with average weight for
gestational age due to catch-up growth (47). As these studies suggest a multitude of factors may
contribute to childhood obesity, other studies have begun to examine the effects of
environmental exposures like air pollution on obesity risk.

1.3 Air Pollution Exposure Assessment
1.3.1 Direct and indirect measurements
Assessment of air pollution exposure is a critical component of epidemiological studies
that evaluates the effect of air pollution on human health. Generally, there are two approaches to
air pollution assessment that use direct or indirect measurements (48). Direct measurements are
personal measures that use a monitoring device that is carried or worn by a subject, and these

7

monitors can actively or passively sample pollutants. Active samplers pump air through a
collector or sensor where some can provide real time data on measured pollutants throughout the
day. Though more accurate for estimating acute exposures, active samplers have limitations as
they tend to be expensive, bulky, noisy, and labor intensive as they require frequent calibration
(48). Passive samplers are inexpensive, small, quiet, and easy to use however require longer
sampling periods to collect enough pollutants for analysis; however, passive samplers tend to be
less accurate than active monitors and are highly dependent on diffusion factors (48, 49). Though
active and passive personal monitors are accurate means of short-term exposure assessment,
personal monitoring tends to be very costly and logistically difficult for large epidemiological
studies, especially when long-term exposure estimates are of interest (50).
Long-term air pollution exposure assessment remains challenging; however, great
advancements in indirect measures of air pollution assessment have been made. Early air
pollution studies characterized exposures by assigning average concentrations from central
monitoring sites within a city (51, 52). These air pollution concentrations reflect mean exposures
for a group of individuals rather than individualized exposures. There are six criteria pollutants
(particulate matter (PM), ozone (O3), nitrogen dioxide (NO2), sulfur dioxide (SO2), carbon
monoxide (CO) and lead) that are monitored by the U.S. Environmental Protection Agency
where a network of outdoor monitors quantitatively tracks pollutants by mass in micrograms or
nanograms per meter cubed (µg/m
3
or ng/m
3
) or by fractional volume in parts per million or
billion (ppm or ppb respectively). Over the years, more advanced models have been developed to
include proximity models, interpolation methods, land use regression models, and dispersion
models (53, 54).

8

Previous methods of population level exposures were simple assignments of city mean
exposures. Though not an accurate assessment of individual air pollution exposures, associations
with health outcomes may be found if air pollution variability is large across cities (52). Aside
from the lack of individual accuracy, population level exposure means also overlook variations
within a city. In order to assess inner city variations with individualized exposures, proximity
methods were developed to measure the proximity or “closeness” of an individual to an emission
source as a proxy of air pollution exposure where traffic counts and distance to roads are often
used (53). Proximity measures are then combined with traffic density and road type to further
classify exposures. This method is often used in exploratory studies where more advanced
methods of exposure assessment are not feasible. Proximity methods should be used with caution
as they are not a quantitative measure of pollutants but rather surrogate measures of exposure.
Interpolation models implement geostatistical techniques that use data from various
monitoring stations to produce individualized exposure estimates in unsampled locations. For
instance, data from multiple monitoring stations is used, adding weight to stations that are closer
in distance to the location of interest (i.e. residential address) (53, 54). Several interpolation
methods have been used in studies including the Kriging method (55, 56) and inverse distance
weighting (57). Although, interpolation methods provide better exposures estimates than
proximity models as real pollution measures are being used, interpolation methods still are
lacking in accuracy. These methods do not always incorporate physical terrain, meteorological
and traffic variations and algorithms do not hold when monitoring sites are sparse (53).
Therefore, interpolation methods like kriging may not be the most effective methods in capturing
local variation in air pollution; other methods like land use regression models may perform better
(58).

9

Land use regression (LUR) models go further by adding characteristics like land use and
traffic volume via geographic information systems (GIS) to predict pollution concentrations (59).
Least squares regression models are built using measured concentrations of pollutants and a set
of geographic (land use) predictors like traffic patterns, geographic density, land use and other
variables like altitude to estimate exposures in non-measured locations (i.e. residential address of
study participants). In general, because LUR models are built specific to the region’s land use
characteristics, it makes it difficult to transfer a LUR model to use in a different region (54).
Another limitation to LUR models is the lack of temporal variation. However, spatiotemporal
models have now been developed to incorporate temporal predictors in the model.
Dispersion models go even further in the modeling process by incorporating physical and
chemical processes in the atmosphere. These models consider chemical and physical properties
of the pollutants, traffic volume, emissions, topography and meteorology to account for
transformation of gases or particles in the atmosphere to predict ground level concentrations that
humans are exposed to (53). There are several types of dispersion models one being the
California line-source dispersion model (CALINE4) which is commonly used to estimate traffic-
related/near-roadway exposures. CALINE4 uses meteorology, road geometry, traffic volume and
vehicle emissions rates as model inputs and then predicts NOx exposures from freeway and non-
freeway roads (60). Dispersion models are advantageous in that they can incorporate both spatial
and temporal variation of air pollution across a study area without the need of dense network of
monitors; nevertheless, limitations include extensive information on model inputs (data that may
or not be readily available), limitations on emissions data, and high level of GIS expertise.
Given the limitations of a single method, studies have begun to use hybrid models that
incorporate multiple methods. Some studies have combined personal monitoring as well as a

10

form of indirect method. Others have combined satellite monitoring for surface particulate matter
≤ 2.5µm in diameter (PM2.5) and NO2 concentrations by using measured column concentrations
with chemical transport models (61, 62). The use of new technologies including GPS, smart
phones and sensors may offer new ways to monitor and track individualized long-term air
pollution exposures.
These indirect measures however only reflect concentrations of outdoor air pollution and
not indoors. Indoor air pollution can vary by an individual’s lifestyle and microenvironment
where open or closed windows/doors, air flow through the residence/workplace, any sources of
combustion (gas stoves, fireplaces, candles, etc.) can influence indoor air pollution levels. One
study found that polycyclic aromatic hydrocarbon (PAH) levels, a marker of traffic pollution,
tend to be higher in outdoor environments than indoor environments, however indoor PAH levels
are heavily influenced by other sources like environmental tobacco smoke (ETS), wood burning
(use of fireplace), or cooking sources (gas stove) (63). Furthermore, the Relationship of Indoor,
Outdoor and Personal Air (RIOPA) study collected indoor, outdoor and personal PM2.5 samples
in non-smoking homes across several U.S. cities and showed outdoor concentration of PM2.5 was
the leading predictor and main contributor to indoor and personal exposures (64). Therefore,
indoor air pollution may vary across residences differ due to personal lifestyle, but also may be
influenced by outdoor pollution levels.

1.3.2 Measurement error
Measurement error is a major potential challenge in epidemiological studies of air
pollution. It is difficult to assess exact, personal exposures as it takes place across various
locations over the life course. Since air pollution exposures are continuous, numeric measures,

11

random or systematic errors may exist. A systematic measurement error is an error that effects
measures by a unit of measure or a given proportion of measure whereas a random measurement
error is one that effects measures randomly with a random unit of measure (65). For instance, a
calibration error for an active monitor would underestimate all particulate matter (PM) measures
by 2.0 μg/m
3
resulting in a systematic shift across all PM measures however this shift is unlikely
to bias association estimates. Systematic errors that result in an incorrect estimate of the
association between an exposure and outcome are the ones that are problematic. A random error
in the same active monitor would yield measures of PM that are underestimated or overestimated
from the true measure without any distinct pattern so errors of PM measures would be at random.
When measurement error occurs, there are two types of measurement error models:
Berkson and classical (65-67). Classical measurement error arises when an individual’s exposure
is measured, and this measurement varies around a “true” value. Berkson error occurs when a
group’s exposure is assigned to an individual (rather than having an exposure for each individual
like the classical error model) where the “true” value will vary around this group exposure.
Classic errors will attenuate linear regression estimates or bias the estimates toward the null;
whereas Berkson errors lead to no bias in linear regression estimates however may lead to less
precise estimates (large confidence intervals) (65, 67). As simply put by Heid et al., “… classical
error is rather related to the measurement process, whereas Berkson error is often a matter of
defining the exposure…”. Exposure measurement errors can also arise from spatial and temporal
uncertainties that are inherent in predicted exposure models.

12

1.4 Air Pollution, Obesity and Cardiometabolic Risk Factors
Energy imbalance and genetic susceptibility were attributed to the high prevalence of
obesity; however, these risk factors alone cannot explain the nearly tripling of obesity rates in
children. There is substantial evidence which shows the detrimental effects of air pollution on
health, and new research suggests that near-roadway air pollution (NRAP) or traffic-related air
pollution (TRAP) exposure may also be contributing to obesity epidemic (51, 68). Jerrett and
colleagues reported that traffic proximity and density around a child’s home was positively
associated with attained BMI at age 18 years in a cohort of Southern California children (69). In
addition, traffic density and non-freeway NOx were associated with attained BMI at age 10 years
and BMI growth over 4 years in a second cohort of Southern California children (70). Children
in the highest 10% TRAP exposure group showed a 0.39 BMI unit increase in attained BMI at 10
years of age as well as a 13.6% increase in average annual BMI growth when compared to
children in the lowest 10% exposure group. Furthermore, researchers found synergistic effects of
NRAP and second-hand smoke exposure on attained BMI at age 18 years (71). Children with
high NRAP exposure and history of secondhand smoke (SHS) exposure showed 2.15 kg/m
2

higher attained BMI than children with low NRAP exposure and no SHS (95% confidence
interval (CI): 1.52-2.77). While these studies show NRAP exposure during childhood is strongly
linked with increased BMI, it is unclear if early life exposures such as in utero also contribute to
increased obesity risk.
Though the literature in NRAP/TRAP and childhood obesity is limited, other health
outcomes have been shown to be associated with NRAP. Increased risk of diabetes with
exposures to NO2 and PM2.5, components of NRAP, have been shown (72, 73). Furthermore,
both short- and long-term exposures to air pollutants increased the risk of insulin resistance in

13

children (74, 75). Not only does air pollution compromise glucose metabolism, increasing risk
for T2DM, increased exposure to air pollution may also perturb lipid metabolism (76-79).
Compared to adults, children are more vulnerable to the effects of air pollution and exposures
can take place as early as fetal development. In utero exposure to traffic pollutants like black
carbon (BC) and PM2.5 were associated with lower fetal growth and rapid weight gain during the
first 6 months of life (80). Additionally, gestational exposure to higher PM2.5, particulate matter
≤ 10µm in diameter (PM10), CO and/or NO2 resulted in increased risk for preterm birth (81-84),
low birth weight (83, 85) and small for gestational age (85).

1.4.1 Early life air pollution exposure and childhood obesity
Early life NRAP exposures may significantly alter the life course as in utero and first
year of life are critical periods for development (86). Although some studies have shown
increased childhood NRAP exposures result in higher childhood BMI (69-71), epidemiological
studies looking at the longitudinal relationships between early life periods of NRAP exposure
and childhood obesity are limited. In one study, increased prenatal polycyclic aromatic
hydrocarbon exposure, a marker for NRAP, was associated with higher BMI z-scores at 5 and 7
years in African-American and Hispanic children (87). More recently, data from the Boston
Birth Cohort showed increased PM2.5 exposure during in utero and first, two years of life
increased the risk of obesity in children aged 2-9 years (88). Similarly, increased prenatal PM2.5
exposure was strongly associated with increased BMI-z score and fat mass in boys and waist to
height ratio in girls at age 4 years (89). Collectively, these studies suggest an association between
early life NRAP and childhood obesity, however, this association has not been entirely consistent
as a two European studies recently reported null associations (90, 91).

14

Several animal models report negative effects of increased early life air pollution
exposures on obesity. In mice, in utero exposure to diesel exhaust predisposed offspring to
higher weight gain when fed a high fat diet compared to those offspring exposed to filtered air
and given a high fat diet (92). Significant associations have also been noted between PM2.5
exposure and increased total abdominal fat in mice (93). Moreover, rats exposed to unfiltered
Beijing NRAP prenatally and continuously after birth had significantly higher fat mass at 8
weeks compared to rats exposed to filtered air (94).

1.4.2 Air pollution and cardiometabolic risk factors
There are numerous health effects associated with childhood obesity—insulin resistance
being a common metabolic consequence seen in obese children. To understand how air pollution
may affect obesity, effects on cardiometabolic risk factors such as fasting glucose, fasting
insulin, insulin resistance, and fasting lipid profiles are also warranted. Long term exposure to
TRAP in 10 year old German children was associated with greater insulin resistance quantified
by homeostasis model assessment (HOMA-IR) (74). For every 2-standard deviation (SD)
increase in NO2 and PM10, insulin resistance increased by 17.0% and 18.7% respectively (74). In
another study, significant associations with acute exposures (past week) to PM10 and insulin
resistance were reported in children aged 10-18 years (75). Additionally, many animal studies
report associations with air pollution and increased risk for insulin resistance (93, 95, 96).
Together, these studies suggest that increased exposure to air pollution independently contributes
to increased risk for obesity and type 2 diabetes. Not only does increased air pollution exposure
affect glucose metabolism, it can also affect lipid metabolism as well. In a cohort of healthy,
Irani adolescents aged 10-18 years, low air quality was positively correlated with total

15

cholesterol, low density lipoprotein–cholesterol, and triglycerides and negatively correlated
with high density lipoprotein–cholesterol (76). Other adult studies have shown increased
exposure to PM10 (77-79) and NO2 (79) are associated with elevated lipid levels.
There may be underlying long-term and short-term effects of NRAP as well as ambient
air pollution, but few studies have truly defined short-, long-term air pollution exposures with
various obesity and cardiometabolic phenotypes. For example, one study used land use
regression models to evaluate the association of NO2, PM10, PM2.5 and PM2.5 absorbance and
insulin resistance in German children 10 years of age using birth addresses (74). Higher levels of
air pollution were associated with higher insulin resistance in children; yet, it is unclear if these
exposures were truly long term since children could have moved since their birth address or if
higher insulin resistance was due to more recent air pollution exposures. In contrast, prior week
exposure to ambient PM10 was significantly associated with insulin resistance using the
homeostatic model assessment (HOMA-IR) in Iranian children 10-18 years of age (75). More
recently, one-year average NRAP and ambient air pollution exposures were associated with
adverse effects on glucose metabolism in overweight/obese African American and Latino youth
(97). Another reported one-year average exposure of ambient NO2 and PM2.5 were associated
with longitudinal changes in insulin sensitivity and beta-cell function where higher levels
ambient air pollution were associated with a faster decline in insulin sensitivity and disposition
index in overweight and obese Latino children 8-15 years of age (98). Collectively these studies
show increased air pollution exposure may alter metabolic processes in children and adolescents,
however potential biological pathways still need further investigation.

16

1.5 Biological Mechanisms of Air Pollution on Obesity
Though exact mechanisms remain uncertain, animal models have proposed potential
biological pathways underlying the association of increased air pollution and increased risk of
obesity and cardiometabolic dysfunction. Oxidative stress and inflammatory pathways are the
most commonly proposed mechanisms. For example, inhaled air pollution induces inflammation
in the lungs which may lead to a spill-over effect of pro-inflammatory cytokines and chemokines
to other tissues in the body (95, 99-102). Similarly, PM components have been found to activate
toll-like receptors (TLRs) which play a key role in our innate immune system (103). A recent
animal study showed TLR4 deficiency attenuates reactive oxygen species and reduces
inflammatory markers in TLR4 deficient mice providing further evidence that TLR activation
may induce systemic oxidative stress and inflammation (104).
These pathways affect the lungs through inhalation, but also may promote inflammation
in adipose tissue (94, 105). A study in China showed that mice prenatally and postnatally
exposed to unfiltered Beijing air showed histological evidence of lung inflammation with
significantly higher levels of malondialdehyde, a marker of oxidative stress, as well as systemic
inflammation from higher levels of proinflammatory chemokines and cytokines and lower anti-
inflammatory cytokines (94). Figure 1.2 shows the mechanistic framework proposed by these
Chinese researchers. They suggest that inhalation of increased PM2.5 prompted inflammation and
lipid oxidation in the lungs followed by a spillover effect to the entire body system leading to
metabolic dysfunction and weight gain. In a different study, male mice were exposed to PM2.5
for 24 weeks, and these mice exhibited insulin resistance and increase adipose tissue
macrophages in visceral fat (105).

17

Other studies have reported air pollution effects on the brain which may potentially alter
appetite regulation and in turn increase susceptibility for weight-gain. A 30-day exposure to
mixed vehicle emissions in mice resulted in a significant increase in blood brain barrier
permeability and increased oxidized low-density lipoprotein signaling in the brain (106). In
another study, mice exposed to diesel exhaust (DE) during in utero exhibited higher levels of
fetal inflammatory cytokines in the brain, increased microglial activation as adults, as well as
displayed increased anxiety as adults compared to mice parentally exposed to filtered air (92).
DE exposed mice also gained more weight on a high fat diet compared those with filtered air and
high fat diet which may be a result from the observed increase in anxiety in those mice exposed
to DE.
Additionally, some studies have reported detrimental effects of sensitive periods of
development like, in utero, where air pollution exposure that may be passed on from mother to
developing fetus. Higher TRAP exposure during pregnancy was associated with higher levels of
leptin and adiponectin in cord blood, and cord blood adipokine levels were associated with
increased change in weight in female infants from birth to 6 months (107). Furthermore, another
study assessed the effects of increased NO2 and PM2.5 exposures during pregnancy and showed
similar results of elevated levels of leptin and adiponectin in cord blood (108). These increased
levels of cord blood leptin and adiponectin are concerning as these two cytokines are secreted by
adipocytes and have been previously been shown to be correlated with birthweight and neonatal
adiposity (109). These studies have uncovered some potential mechanisms underlying the
association of NRAP exposures with obesity and that effects of air pollution exposure can be
seen as early as the in utero period. Given the evidence from animal studies and a handful of

18

epidemiological studies, additional studies are warranted to validate the effects of air pollution
on childhood obesity and cardiometabolic risk factors.

1.6 Dissertation Goals
Recent studies have shown that increased in utero and early life NRAP exposure is
associated with increased childhood BMI and risk for obesity, however these studies lacked
longitudinal data where effects of early life NRAP can be seen on childhood BMI trajectory. No
studies to our knowledge have explored the impact of a genetic risk score and early life air
pollution gene-environment interactions on childhood BMI growth. There are also very few
studies that evaluate short- and long-term air pollution exposures and its association with various
obesity/cardiometabolic phenotypes in young adults. Therefore, the goal of this dissertation is to
address gaps in the literature through three unique projects:
1. The first project examines the association of early life NRAP exposures (in utero and
first year of life) with childhood BMI trajectory. Given the lack of data on
longitudinal effects of early life air pollution exposure on children’s growth, data
from the Children’s Health Study (CHS) with annual measures of BMI and lifetime
NRAP exposures were used to examine this association.
2. The second project examines if the association between early life NRAP exposure
and childhood BMI trajectory is modified by genetic predisposition to obesity using a
genetic risk score. The association of genetic risk score and childhood BMI trajectory
was examined as well as associations with genetic risk score-environment interactions
and childhood BMI growth. Projects one and two (chapters 2 and 3 respectively) used

19

BMI as a surrogate measure of obesity and did not assess metabolic health as this data
was not available in the CHS.
3. For this reason, the third project examines the short- and long-term consequences of
increased air pollution exposures on adiposity measures and cardiometabolic risk
factors in a subgroup of CHS participants enrolled in the Metabolic and Asthma
Incidence Research (Meta-AIR) Study. The Meta-AIR study collected clinical
measures of obesity and cardiometabolic risk factors along with short-term (prior 1-
month) and long-term (prior 1-year) average ambient and NRAP exposures to
examine these associations.

20

1.7 References
1. Ogden CL, Carroll MD, Kit BK, et al. Prevalence of childhood and adult obesity in the
United States, 2011-2012. JAMA 2014;311(8):806-14.
2. Hales CM, Carroll MD, Fryar CD, et al. Prevalence of Obesity Among Adults and Youth:
United States, 2015-2016. NCHS Data Brief 2017(288):1-8.
3. Ogden CL, Fryar CD, Hales CM, et al. Differences in Obesity Prevalence by
Demographics and Urbanization in US Children and Adolescents, 2013-2016. JAMA
2018;319(23):2410-8.
4. Freedman DS, Mei Z, Srinivasan SR, et al. Cardiovascular risk factors and excess
adiposity among overweight children and adolescents: the Bogalusa Heart Study. J
Pediatr 2007;150(1):12-7.e2.
5. Singh AS, Mulder C, Twisk JW, et al. Tracking of childhood overweight into adulthood:
a systematic review of the literature. Obes Rev 2008;9(5):474-88.
6. Kuczmarski RJ, Ogden CL, Guo SS, et al. 2000 CDC Growth Charts for the United
States: methods and development. Vital Health Stat 11 2002(246):1-190.
7. Bichteler A, Gershoff ET. Identification of Children's BMI Trajectories and Prediction
from Weight Gain in Infancy. Obesity (Silver Spring) 2018;26(6):1050-6.
8. Giles LC, Whitrow MJ, Davies MJ, et al. Growth trajectories in early childhood, their
relationship with antenatal and postnatal factors, and development of obesity by age 9
years: results from an Australian birth cohort study. Int J Obes (Lond) 2015;39(7):1049-
56.
9. Ward ZJ, Long MW, Resch SC, et al. Simulation of Growth Trajectories of Childhood
Obesity into Adulthood. N Engl J Med 2017;377(22):2145-53.
10. Weiss R, Kaufman FR. Metabolic complications of childhood obesity: identifying and
mitigating the risk. Diabetes Care 2008;31 Suppl 2:S310-6.
11. Caprio S, Hyman LD, Limb C, et al. Central adiposity and its metabolic correlates in
obese adolescent girls. Am J Physiol 1995;269(1 Pt 1):E118-26.
12. Goran MI, Bergman RN, Avila Q, et al. Impaired glucose tolerance and reduced beta-cell
function in overweight Latino children with a positive family history for type 2 diabetes.
J Clin Endocrinol Metab 2004;89(1):207-12.
13. Hannon TS, Rao G, Arslanian SA. Childhood obesity and type 2 diabetes mellitus.
Pediatrics 2005;116(2):473-80.
14. Liu LL, Lawrence JM, Davis C, et al. Prevalence of overweight and obesity in youth with
diabetes in USA: the SEARCH for Diabetes in Youth study. Pediatr Diabetes
2010;11(1):4-11.
15. Dabelea D, Mayer-Davis EJ, Saydah S, et al. Prevalence of type 1 and type 2 diabetes
among children and adolescents from 2001 to 2009. JAMA 2014;311(17):1778-86.
16. Imperatore G, Boyle JP, Thompson TJ, et al. Projections of type 1 and type 2 diabetes
burden in the U.S. population aged <20 years through 2050: dynamic modeling of
incidence, mortality, and population growth. Diabetes Care 2012;35(12):2515-20.
17. Juonala M, Magnussen CG, Berenson GS, et al. Childhood adiposity, adult adiposity, and
cardiovascular risk factors. N Engl J Med 2011;365(20):1876-85.
18. Franks PW, Hanson RL, Knowler WC, et al. Childhood obesity, other cardiovascular risk
factors, and premature death. N Engl J Med 2010;362(6):485-93.

21

19. Buchanan TA. Pancreatic beta-cell loss and preservation in type 2 diabetes. Clin Ther
2003;25 Suppl B:B32-46.
20. Demmer RT, Zuk AM, Rosenbaum M, et al. Prevalence of diagnosed and undiagnosed
type 2 diabetes mellitus among US adolescents: results from the continuous NHANES,
1999-2010. Am J Epidemiol 2013;178(7):1106-13.
21. Magnussen CG, Venn A, Thomson R, et al. The association of pediatric low- and high-
density lipoprotein cholesterol dyslipidemia classifications and change in dyslipidemia
status with carotid intima-media thickness in adulthood evidence from the cardiovascular
risk in Young Finns study, the Bogalusa Heart study, and the CDAH (Childhood
Determinants of Adult Health) study. J Am Coll Cardiol 2009;53(10):860-9.
22. Cote AT, Phillips AA, Harris KC, et al. Obesity and arterial stiffness in children:
systematic review and meta-analysis. Arterioscler Thromb Vasc Biol 2015;35(4):1038-
44.
23. Mahoney LT, Burns TL, Stanford W, et al. Coronary risk factors measured in childhood
and young adult life are associated with coronary artery calcification in young adults: the
Muscatine Study. J Am Coll Cardiol 1996;27(2):277-84.
24. Nguyen D, Kit B, Carroll M. Abnormal Cholesterol Among Children and Adolescents in
the United States, 2011-2014. NCHS Data Brief 2015(228):1-8.
25. Finkelstein EA, Graham WC, Malhotra R. Lifetime direct medical costs of childhood
obesity. Pediatrics 2014;133(5):854-62.
26. Wang YC, McPherson K, Marsh T, et al. Health and economic burden of the projected
obesity trends in the USA and the UK. Lancet 2011;378(9793):815-25.
27. Andersen RE, Crespo CJ, Bartlett SJ, et al. Relationship of physical activity and
television watching with body weight and level of fatness among children: results from
the Third National Health and Nutrition Examination Survey. JAMA 1998;279(12):938-
42.
28. Trost SG, Kerr LM, Ward DS, et al. Physical activity and determinants of physical
activity in obese and non-obese children. Int J Obes Relat Metab Disord 2001;25(6):822-
9.
29. Ludwig DS, Peterson KE, Gortmaker SL. Relation between consumption of sugar-
sweetened drinks and childhood obesity: a prospective, observational analysis. Lancet
2001;357(9255):505-8.
30. Davis JN, Whaley SE, Goran MI. Effects of breastfeeding and low sugar-sweetened
beverage intake on obesity prevalence in Hispanic toddlers. Am J Clin Nutr 2012;95(1):3-
8.
31. Maes HH, Neale MC, Eaves LJ. Genetic and environmental factors in relative body
weight and human adiposity. Behav Genet 1997;27(4):325-51.
32. Silventoinen K, Rokholm B, Kaprio J, et al. The genetic and environmental influences on
childhood obesity: a systematic review of twin and adoption studies. Int J Obes (Lond)
2010;34(1):29-40.
33. Bradfield JP, Taal HR, Timpson NJ, et al. A genome-wide association meta-analysis
identifies new childhood obesity loci. Nat Genet 2012;44(5):526-31.
34. Frayling TM, Timpson NJ, Weedon MN, et al. A common variant in the FTO gene is
associated with body mass index and predisposes to childhood and adult obesity. Science
2007;316(5826):889-94.

22

35. Speliotes EK, Willer CJ, Berndt SI, et al. Association analyses of 249,796 individuals
reveal 18 new loci associated with body mass index. Nat Genet 2010;42(11):937-48.
36. Zhao J, Bradfield JP, Li M, et al. The role of obesity-associated loci identified in
genome-wide association studies in the determination of pediatric BMI. Obesity (Silver
Spring) 2009;17(12):2254-7.
37. Loos RJ, Bouchard C. FTO: the first gene contributing to common forms of human
obesity. Obes Rev 2008;9(3):246-50.
38. Cecil JE, Tavendale R, Watt P, et al. An obesity-associated FTO gene variant and
increased energy intake in children. N Engl J Med 2008;359(24):2558-66.
39. Loos RJ, Lindgren CM, Li S, et al. Common variants near MC4R are associated with fat
mass, weight and risk of obesity. Nat Genet 2008;40(6):768-75.
40. Farooqi IS, Yeo GS, Keogh JM, et al. Dominant and recessive inheritance of morbid
obesity associated with melanocortin 4 receptor deficiency. J Clin Invest
2000;106(2):271-9.
41. Turula M, Kaprio J, Rissanen A, et al. Body weight in the Finnish Twin Cohort. Diabetes
Res Clin Pract 1990;10 Suppl 1:S33-6.
42. Wardle J, Carnell S, Haworth CM, et al. Evidence for a strong genetic influence on
childhood adiposity despite the force of the obesogenic environment. Am J Clin Nutr
2008;87(2):398-404.
43. Fuemmeler BF, Wang L, Iversen ES, et al. Association between Prepregnancy Body
Mass Index and Gestational Weight Gain with Size, Tempo, and Velocity of Infant
Growth: Analysis of the Newborn Epigenetic Study Cohort. Child Obes 2016;12(3):210-
8.
44. Yu Z, Han S, Zhu J, et al. Pre-pregnancy body mass index in relation to infant birth
weight and offspring overweight/obesity: a systematic review and meta-analysis. PLoS
One 2013;8(4):e61627.
45. Sridhar SB, Darbinian J, Ehrlich SF, et al. Maternal gestational weight gain and offspring
risk for childhood overweight or obesity. Am J Obstet Gynecol 2014;211(3):259 e1-8.
46. Oken E, Levitan EB, Gillman MW. Maternal smoking during pregnancy and child
overweight: systematic review and meta-analysis. Int J Obes (Lond) 2008;32(2):201-10.
47. Stettler N, Zemel BS, Kumanyika S, et al. Infant weight gain and childhood overweight
status in a multicenter, cohort study. Pediatrics 2002;109(2):194-9.
48. Watson AY, Bates RR, Kennedy D. Air Pollution, the Automobile, and Public Health.
Washington (DC), 1988.
49. Zou B, Wilson JG, Zhan FB, et al. Air pollution exposure assessment methods utilized in
epidemiological studies. J Environ Monit 2009;11(3):475-90.
50. Larkin A, Hystad P. Towards Personal Exposures: How Technology Is Changing Air
Pollution and Health Research. Curr Environ Health Rep 2017;4(4):463-71.
51. Dockery DW, Pope CA, Xu X, et al. An association between air pollution and mortality
in six U.S. cities. N Engl J Med 1993;329(24):1753-9.
52. Pope CA, Burnett RT, Thun MJ, et al. Lung cancer, cardiopulmonary mortality, and
long-term exposure to fine particulate air pollution. JAMA 2002;287(9):1132-41.
53. Jerrett M, Arain A, Kanaroglou P, et al. A review and evaluation of intraurban air
pollution exposure models. J Expo Anal Environ Epidemiol 2005;15(2):185-204.
54. Hoek G. Methods for Assessing Long-Term Exposures to Outdoor Air Pollutants. Curr
Environ Health Rep 2017;4(4):450-62.

23

55. Carrat F, Valleron AJ. Epidemiologic mapping using the "kriging" method: application to
an influenza-like illness epidemic in France. Am J Epidemiol 1992;135(11):1293-300.
56. Jerrett M, Burnett RT, Ma R, et al. Spatial analysis of air pollution and mortality in Los
Angeles. Epidemiology 2005;16(6):727-36.
57. Beelen R, Hoek G, van den Brandt PA, et al. Long-term effects of traffic-related air
pollution on mortality in a Dutch cohort (NLCS-AIR study). Environ Health Perspect
2008;116(2):196-202.
58. Pouliou T, Kanaroglou PS, Elliott SJ, et al. Assessing the health impacts of air pollution:
a re-analysis of the Hamilton children's cohort data using a spatial analytic approach. Int
J Environ Health Res 2008;18(1):17-35.
59. Briggs DJ, Collins S, Elliott P, et al. Mapping urban air pollution using gis: a regression-
based approach. International Journal of Geographical Information Science,
1997;11(7):699-718.
60. Benson P. CALINE4 – A Dispersion Model for Predicting Air Pollutant Concentrations
Near Roadways. 1989;Report No. FHWA/CA/TL-84/15.(Prepared by the California
Department of Transportation, Sacramento, CA.).
61. Crouse DL, Peters PA, van Donkelaar A, et al. Risk of nonaccidental and cardiovascular
mortality in relation to long-term exposure to low concentrations of fine particulate
matter: a Canadian national-level cohort study. Environ Health Perspect
2012;120(5):708-14.
62. Anderson HR, Butland BK, van Donkelaar A, et al. Satellite-based estimates of ambient
air pollution and global variations in childhood asthma prevalence. Environ Health
Perspect 2012;120(9):1333-9.
63. Delgado-Saborit JM, Stark C, Harrison RM. Carcinogenic potential, levels and sources of
polycyclic aromatic hydrocarbon mixtures in indoor and outdoor environments and their
implications for air quality standards. Environ Int 2011;37(2):383-92.
64. Meng QY, Spector D, Colome S, et al. Determinants of Indoor and Personal Exposure to
PM(2.5) of Indoor and Outdoor Origin during the RIOPA Study. Atmos Environ (1994)
2009;43(36):5750-8.
65. Armstrong BG. Effect of measurement error on epidemiological studies of environmental
and occupational exposures. Occup Environ Med 1998;55(10):651-6.
66. Sheppard L, Burnett RT, Szpiro AA, et al. Confounding and exposure measurement error
in air pollution epidemiology. Air Qual Atmos Health 2012;5(2):203-16.
67. Heid IM, Kuchenhoff H, Miles J, et al. Two dimensions of measurement error: classical
and Berkson error in residential radon exposure assessment. J Expo Anal Environ
Epidemiol 2004;14(5):365-77.
68. Gauderman WJ, Avol E, Gilliland F, et al. The effect of air pollution on lung
development from 10 to 18 years of age. N Engl J Med 2004;351(11):1057-67.
69. Jerrett M, McConnell R, Chang CC, et al. Automobile traffic around the home and
attained body mass index: a longitudinal cohort study of children aged 10-18 years. Prev
Med 2010;50 Suppl 1:S50-8.
70. Jerrett M, McConnell R, Wolch J, et al. Traffic-related air pollution and obesity
formation in children: a longitudinal, multilevel analysis. Environ Health 2014;13:49.
71. McConnell R, Shen E, Gilliland FD, et al. A longitudinal cohort study of body mass
index and childhood exposure to secondhand tobacco smoke and air pollution: the

24

Southern California Children's Health Study. Environ Health Perspect 2015;123(4):360-
6.
72. Brook RD, Jerrett M, Brook JR, et al. The relationship between diabetes mellitus and
traffic-related air pollution. J Occup Environ Med 2008;50(1):32-8.
73. Kramer U, Herder C, Sugiri D, et al. Traffic-related air pollution and incident type 2
diabetes: results from the SALIA cohort study. Environ Health Perspect
2010;118(9):1273-9.
74. Thiering E, Cyrys J, Kratzsch J, et al. Long-term exposure to traffic-related air pollution
and insulin resistance in children: results from the GINIplus and LISAplus birth cohorts.
Diabetologia 2013;56(8):1696-704.
75. Kelishadi R, Mirghaffari N, Poursafa P, et al. Lifestyle and environmental factors
associated with inflammation, oxidative stress and insulin resistance in children.
Atherosclerosis 2009;203(1):311-9.
76. Poursafa P, Mansourian M, Motlagh ME, et al. Is air quality index associated with
cardiometabolic risk factors in adolescents? The CASPIAN-III Study. Environ Res
2014;134:105-9.
77. Chuang KJ, Yan YH, Cheng TJ. Effect of air pollution on blood pressure, blood lipids,
and blood sugar: a population-based approach. J Occup Environ Med 2010;52(3):258-62.
78. Shanley RP, Hayes RB, Cromar KR, et al. Particulate Air Pollution and Clinical
Cardiovascular Disease Risk Factors. Epidemiology 2016;27(2):291-8.
79. Cai Y, Hansell AL, Blangiardo M, et al. Long-term exposure to road traffic noise,
ambient air pollution, and cardiovascular risk factors in the HUNT and lifelines cohorts.
Eur Heart J 2017;38(29):2290-6.
80. Fleisch AF, Rifas-Shiman SL, Koutrakis P, et al. Prenatal exposure to traffic pollution:
associations with reduced fetal growth and rapid infant weight gain. Epidemiology
2015;26(1):43-50.
81. Llop S, Ballester F, Estarlich M, et al. Preterm birth and exposure to air pollutants during
pregnancy. Environ Res 2010;110(8):778-85.
82. Padula AM, Mortimer KM, Tager IB, et al. Traffic-related air pollution and risk of
preterm birth in the San Joaquin Valley of California. Ann Epidemiol 2014;24(12):888-
95e4.
83. Stieb DM, Chen L, Eshoul M, et al. Ambient air pollution, birth weight and preterm birth:
a systematic review and meta-analysis. Environ Res 2012;117:100-11.
84. Ritz B, Yu F, Chapa G, et al. Effect of air pollution on preterm birth among children born
in Southern California between 1989 and 1993. Epidemiology 2000;11(5):502-11.
85. Hyder A, Lee HJ, Ebisu K, et al. PM2.5 exposure and birth outcomes: use of satellite-
and monitor-based data. Epidemiology 2014;25(1):58-67.
86. Dietz WH. Critical periods in childhood for the development of obesity. Am J Clin Nutr
1994;59(5):955-9.
87. Rundle A, Hoepner L, Hassoun A, et al. Association of childhood obesity with maternal
exposure to ambient air polycyclic aromatic hydrocarbons during pregnancy. Am J
Epidemiol 2012;175(11):1163-72.
88. Mao G, Nachman RM, Sun Q, et al. Individual and Joint Effects of Early-Life Ambient
PM2.5 Exposure and Maternal Pre-Pregnancy Obesity on Childhood Overweight or
Obesity. Environ Health Perspect 2016.

25

89. Chiu YM, Hsu HL, Wilson A, et al. Prenatal particulate air pollution exposure and body
composition in urban preschool children: Examining sensitive windows and sex-specific
associations. Environ Res 2017;158:798-805.
90. Fioravanti S, Cesaroni G, Badaloni C, et al. Traffic-related air pollution and childhood
obesity in an Italian birth cohort. Environ Res 2018;160:479-86.
91. Frondelius K, Oudin A, Malmqvist E. Traffic-Related Air Pollution and Child BMI-A
Study of Prenatal Exposure to Nitrogen Oxides and Body Mass Index in Children at the
Age of Four Years in Malmo, Sweden. Int J Environ Res Public Health 2018;15(10).
92. Bolton JL, Smith SH, Huff NC, et al. Prenatal air pollution exposure induces
neuroinflammation and predisposes offspring to weight gain in adulthood in a sex-
specific manner. FASEB J 2012;26(11):4743-54.
93. Xu X, Yavar Z, Verdin M, et al. Effect of early particulate air pollution exposure on
obesity in mice: role of p47phox. Arterioscler Thromb Vasc Biol 2010;30(12):2518-27.
94. Wei Y, Zhang JJ, Li Z, et al. Chronic exposure to air pollution particles increases the risk
of obesity and metabolic syndrome: findings from a natural experiment in Beijing.
FASEB J 2016.
95. Liu C, Xu X, Bai Y, et al. Air pollution-mediated susceptibility to inflammation and
insulin resistance: influence of CCR2 pathways in mice. Environ Health Perspect
2014;122(1):17-26.
96. Yan YH, Chou CC, Lee CT, et al. Enhanced insulin resistance in diet-induced obese rats
exposed to fine particles by instillation. Inhal Toxicol 2011;23(9):507-19.
97. Toledo-Corral CM, Alderete TL, Habre R, et al. Effects of air pollution exposure on
glucose metabolism in Los Angeles minority children. Pediatr Obes 2018;13(1):54-62.
98. Alderete TL, Habre R, Toledo-Corral CM, et al. Longitudinal Associations Between
Ambient Air Pollution With Insulin Sensitivity, beta-Cell Function, and Adiposity in Los
Angeles Latino Children. Diabetes 2017;66(7):1789-96.
99. Tamagawa E, Bai N, Morimoto K, et al. Particulate matter exposure induces persistent
lung inflammation and endothelial dysfunction. Am J Physiol Lung Cell Mol Physiol
2008;295(1):L79-85.
100. van Eeden SF, Tan WC, Suwa T, et al. Cytokines involved in the systemic inflammatory
response induced by exposure to particulate matter air pollutants (PM(10)). Am J Respir
Crit Care Med 2001;164(5):826-30.
101. Fry RC, Rager JE, Zhou H, et al. Individuals with increased inflammatory response to
ozone demonstrate muted signaling of immune cell trafficking pathways. Respir Res
2012;13:89.
102. Happo MS, Salonen RO, Halinen AI, et al. Dose and time dependency of inflammatory
responses in the mouse lung to urban air coarse, fine, and ultrafine particles from six
European cities. Inhal Toxicol 2007;19(3):227-46.
103. Rajagopalan S, Brook RD. Air pollution and type 2 diabetes: mechanistic insights.
Diabetes 2012;61(12):3037-45.
104. Pushpakumar S, Ren L, Kundu S, et al. Toll-like Receptor 4 Deficiency Reduces
Oxidative Stress and Macrophage Mediated Inflammation in Hypertensive Kidney. Sci
Rep 2017;7(1):6349.
105. Sun Q, Yue P, Deiuliis JA, et al. Ambient air pollution exaggerates adipose inflammation
and insulin resistance in a mouse model of diet-induced obesity. Circulation
2009;119(4):538-46.

26

106. Suwannasual U, Lucero J, McDonald JD, et al. Exposure to traffic-generated air
pollutants mediates alterations in brain microvascular integrity in wildtype mice on a
high-fat diet. Environ Res 2018;160:449-61.
107. Alderete TL, Song AY, Bastain T, et al. Prenatal traffic-related air pollution exposures,
cord blood adipokines and infant weight. Pediatr Obes 2018;13(6):348-56.
108. Lavigne E, Ashley-Martin J, Dodds L, et al. Air Pollution Exposure During Pregnancy
and Fetal Markers of Metabolic Function: The MIREC Study. Am J Epidemiol 2016.
109. Tsai PJ, Yu CH, Hsu SP, et al. Cord plasma concentrations of adiponectin and leptin in
healthy term neonates: positive correlation with birthweight and neonatal adiposity. Clin
Endocrinol (Oxf) 2004;61(1):88-93.

27

1.8 Figures
Figure 1.1 Trends* in U.S. Childhood Obesity: 1999-2015.

*Source: NCHS, National Health and Nutrition Examination Survey data from 1999-2016

13.9
15.4
17.1
15.4
16.8 16.9 16.9
17.2
18.5
0
5
10
15
20
Obesity Prevalence (%)
NHANES Survey Years

28

Figure 1.2 Mechanistic Framework: How Air Pollution May Lead to Increased Risk of Obesity.

Source: Wei et al. FASEB J. 2016 Jun; 30(6): 2115–2122.

29

Chapter 2: Longitudinal Associations of In Utero and Early Life Near-
Roadway Air Pollution with Trajectories of Childhood Body Mass Index

Authors: Jeniffer S. Kim
1
, Tanya L. Alderete
1
, Zhanghua Chen
1
, Fred Lurmann
2
, Ed
Rappaport
1
, Rima Habre
1
, Kiros Berhane
1
, Frank D. Gilliland
1

Institutional Affiliation:
1
Department of Preventive Medicine, Division of Environmental
Health, Keck School of Medicine, University of Southern California, Los Angeles, CA, USA;
2
Sonoma Technology Inc., Petaluma, CA, USA.

Published in Environmental Health 14 September 2018; 17(1):64. PMID:30213262.

30

2.1 Abstract
Evidence suggests that childhood near-roadway air pollution (NRAP) exposures
contribute to increased body mass index (BMI); however, effects of NRAP exposure during the
vulnerable periods including in utero and first year of life have yet to be established. In this
study, we examined whether exposure to elevated concentrations of NRAP during in
utero and/or first year of life increase childhood BMI growth. Participants in the Children’s
Health Study enrolled from 2002-2003 with annual visits over a four-year period and who
changed residences before study entry were included (n=2,318). Annual height and weight were
measured and lifetime residential NRAP exposures including in utero and first year of life
periods were estimated by nitrogen oxides (NOx) using the California line-source dispersion
model. Linear mixed effects models assessed in utero or first year near-road freeway and non-
freeway NOx exposures and BMI growth after adjusting for age, sex, race/ethnicity, parental
education, Spanish questionnaire, and later childhood near-road NOx exposure. A two-standard
deviation difference in first year of life near-road freeway NOx exposure was associated with a
0.1 kg/m
2
(95% confidence interval (CI): 0.03, 0.2) faster increase in BMI growth per year and a
0.5 kg/m
2
(95% CI: 0.02, 0.9) higher attained BMI at age 10 years. Higher exposure to early life
NRAP increased the rate of change of childhood BMI and resulted in a higher attained BMI at
age 10 years that were independent of later childhood exposures. These findings suggest that
elevated early life NRAP exposures contribute to increased obesity risk in children.

31

2.2 Introduction
In the United States, approximately 32% of children 2 to 19 years were overweight or
obese in 2011-2012 (1). High prevalence of childhood obesity present significant clinical and
public health problems since obese children are more likely to become obese adults (2, 3) and are
at a greater risk for developing type 2 diabetes and cardiovascular disease (4-7). Despite decades
of diet and physical activity interventions, the prevalence of childhood obesity remains high (8).
Previous studies have shown that increased near-roadway air pollution (NRAP) exposure (9, 10)
and increased traffic density (9, 11) during childhood contributes to increased obesity risk in
children. These findings suggest that modifiable environmental factors such as air pollution
exposures may be contributing to the obesity epidemic (9, 12).
Longitudinal studies have shown that increased mid-childhood NRAP exposure and
traffic density are associated with substantially increased body mass index (BMI) growth and a
higher level of attained BMI at ages 18 (11) and 10 years (9). Furthermore, increased childhood
NRAP exposures, secondhand smoke (SHS), and maternal smoking during pregnancy were
associated with increased BMI growth and a higher BMI at age 18 years (10). To date, few
studies have examined the effects of in utero and early childhood NRAP exposure on childhood
BMI growth. Beyond mid-childhood exposures, early life periods like in utero and first year of
life represent critical windows of air pollution exposure that may significantly alter childhood
growth trajectories. One epidemiological study found that increased in utero ambient polycyclic
aromatic hydrocarbon exposure, a marker for NRAP, was associated with higher BMI z-scores at
age 5 and 7 years (13). Additionally, recent data from the Boston Birth Cohort showed that in
utero and early life exposures to ambient, fine particulate matter ≤ 2.5 µm in diameter (PM2.5)
were significantly associated with increased risk of childhood overweight or obesity in children

32

2-9 years of age (14). Although these studies suggest that elevated NRAP exposures during early
life and childhood may increase future obesity risk, studies have not been entirely consistent as a
European birth cohort recently reported no association of first four years of life NRAP exposure
and childhood obesity at 4 years and 8 years of age (15).
Despite mixed findings, in utero and first year of life are periods of rapid growth that are
highly susceptible to environmental influences (16). Lasting effects of in utero and early life
environmental exposures on childhood growth trajectories have been shown, and these early life
stages are critical periods for the development of obesity (16, 17). For example, past studies have
reported associations of increased in utero air pollution exposure and restricted fetal growth
resulting in low birth weight (LBW) in full term babies (18-20). Consequently, term babies with
LBW have been shown to have postnatal catch-up growth which in turn has been associated with
higher weight gain through infancy and childhood (21, 22). Because in utero and first year of life
are important developmental periods that influence growth, increased NRAP exposure during
these critical periods may be contributing to future obesity risk through altered growth
trajectories resulting in faster childhood BMI growth. The objective of this study was to examine
the relationships between in utero and first year of life NRAP exposures with longitudinal
measurements of BMI in a subset of Southern California children enrolled in the Children’s
Health Study (CHS). We hypothesized that higher in utero and/or first year of life NRAP
exposure is associated with a faster increase in BMI growth over a 4-year follow up period and a
higher attained BMI at age 10 years, and that these early life NRAP effects are independent of
NRAP exposures later in childhood.

33

2.3 Materials and Methods
Study Design
From 2002-2003, a cohort of kindergarten and first grade children were recruited from
45 public schools across 13 Southern California communities (Figure 2.1). Details of CHS
recruitment methods have been previously presented (23). Informed consents were obtained from
parents and assents from children. This study was reviewed and approved by the Institutional
Review Board at the University of Southern California.
Height and weight were measured at baseline and at each subsequent, annual school visit.
A trained technician measured height to the nearest centimeter and weight to the nearest pound
(0.45 kilograms) of each child without shoes and with daily calibrations of the weight scale. BMI
was calculated by weight/height
2
(kilograms/meters
2
) and overweight and obese were defined
using the Centers for Disease Control and Prevention (CDC) age-and sex-specific growth charts
(24). Parents completed baseline and yearly follow-up questionnaires pertaining to
sociodemographic factors, characteristics of the home, and other covariates that were explored as
potential confounders. These included age, sex, race/ethnicity, self-reported premature birth,
maternal smoking during pregnancy, SHS, lifetime history of asthma, parental education (marker
for socioeconomic status), if baseline questionnaire was completed in Spanish (marker for recent
immigration status), and child’s participation in team sports during the past year. A subset of
children had birth weight and gestational age at birth which were extracted from the California
vital statistics records. These data were obtained using the CDC LINKPLUS program which
matched CHS children to the state’s vital records database using several variables including
child’s name, sex, birthdate, as well as, mother’s name, birthdate, zip code, and father’s name.

34

NRAP Exposures
Lifetime residential history was collected from all CHS participants at study enrollment
and at each annual CHS visit via questionnaires, including move in and move out dates for each
residence. NRAP exposures were estimated based on street level geocoded, residential locations
from in utero to the most current follow-up date, which gave estimates of NRAP exposure for in
utero, first year of life, and childhood periods. For in utero exposures, street address level data
from study questionnaires (74.8%) and birth certificates (25.2%) were used. Residential
addresses were uploaded to the ESRI geocoding database and software (ESRI Inc., Redlands, CA
http://www.esri.com), geocoded to street level using the software, and assigned latitude and
longitude coordinates. Google Earth and the Texas A&M geocoder (25) were also used to assign
coordinates for a small number of problematic residential addresses.
The California line-source dispersion model (CALINE4) was used to estimate
concentrations of traffic-related nitrogen oxides (NOx) for freeway and non-freeway roads using
EMFAC2011 (for 1994-1999) and EMFAC2014 (for >1999) vehicle emission rates, traffic
volume, road geometry and meteorological conditions, including wind speed and direction,
pollution mixing heights, and atmospheric stability (26). Our roadway information is classified
according to Feature Class Codes (FCC) which includes 1) primary highways with limited access
(freeways), 2) primary roads without limited access, 3) secondary and connecting roads, and 4)
local, neighborhood, and rural roads. Annual average traffic volumes from imbedded loop
sensors provide 100% coverage for freeways and 95% coverage for state highways (Class 2).
The freeways have data for each link between interchanges whereas data for Class 2 roads is
often extrapolated for longer distances to provide full coverage for all links. About 5% of
smaller roads have traffic volume measurements and because these measurements are often on

35

higher volume roads within their class, we assign the 25
th
percentile of annual traffic volume to
all roads in the corresponding class. Exposure to NRAP was modeled using monthly average
CALINE4 estimated NOx concentrations from freeway and non-freeway sources within 5 km of
the residential coordinates during the in utero, first year of life, and mid-childhood periods. In
utero near-road NOx exposure was defined as the nine-month average exposure prior to birth and
first year of life near-road NOx exposure was defined as the twelve-month average exposure after
birth. Lastly, to account for childhood near-road NOx exposures beyond the early life periods,
childhood NRAP exposure was calculated as the average near-road NOx exposure from 13
months of age through the 4-year study follow up period. Traffic pollutants are a complex
mixture of gases and particles that include NOx, carbon monoxide, elemental carbon, particulate
matter, organic compounds, and polycyclic aromatic hydrocarbons amongst others. These NRAP
exposures reflect increases in local vehicle emissions beyond background ambient levels.
Therefore, NRAP was modeled using near-road NOx from freeway and non-freeway roadways as
a marker for traffic pollution as this measure is highly correlated with other pollutants estimated
by CALINE4.

Statistical Methods
Linear mixed effects models (27, 28) were fitted to estimate longitudinal relationships
between BMI trajectory and in utero and first year of life NRAP exposures. We examined the
associations of in utero or first year of life near-road NOx exposure from freeway and non-
freeway roads with 1) the rate of change in BMI during four years of study follow up and 2) the
attained BMI level at age 10 years. Due to the high correlation of in utero and first year of life

36

near-road NOx exposures (correlation r= 0.8 for freeway, r=0.93 for non-freeway), these two
early life periods were analyzed in separate models.
Specifically, the following multi-level mixed effects model was used in the analysis. Let
repeated measures of BMI (Ycij) with c, i, and j representing the study community, individual,
and year of BMI measurement, respectively. Then,
Level 1: Y
cij
= a
ci
+ b
ci
(t
cij
− C) + γ
1
(E
Fij
− E
Fi
) + γ
2
W
ij
+ε
cij

Level 2a (level): a
ci
= a
c
+ α
1
E
Ui
+ α
2
E
Fi
+ α
3
Z
i
+ δ
ci

Level 2b (growth): b
ci
= β
0
+ β
1
E
Ui
+ β
2
E
Fi
+ β
3
Z
i
+ σ
ci

Level 3a: a
c
= α
0
+ ε
c

In Level 1, 𝑡 𝑐𝑖𝑗
is the age of participants at each visit centered by age C (10 years), and γ
1

represents cross-sectional association between year to year fluctuations of near-road NOx with
follow up BMI measure at each study visit. 𝐸 𝐹𝑖𝑗 reflects average near-road NOx exposure for the
time between each subsequent follow up visit and 𝐸 𝐹𝑖
is the average childhood near-road NOx
exposure from 13 months of age till last height/weight measure in 2006-2007 school year.
Importantly, in this analysis we wanted to elucidate associations of in utero or first year of life
near-road NOx exposures (E
Ui
) with BMI growth independent of childhood near-road NOx
exposures. Therefore, the mixed model also includes average childhood near-road NOx
exposures (𝐸 𝐹𝑖
) while accounting for yearly deviations of near-road NOx during this follow-up
period (E
Fij
− E
Fi
).
In levels 2a and 2b, α
1
and β
1
correspond to estimated effects of in utero near-road NOx
exposure (or first year of life) on attained BMI level at age 10 years and the growth of BMI
during the follow-up period, respectively. Whereas, α
2
and β
2
correspond to estimated effects of
childhood near-road NOx exposure on attained BMI level at age 10 years and the growth of BMI

37

during the childhood period, respectively. Furthermore, α
3
Z
i
denotes adjustment factors for
time-independent covariates at BMI level at age 10, β
3
Z
i
are adjustment factors for BMI growth
during follow-up, ε
cij
, ε
c
, δ
ci
and σ
ci
reflect error terms at each level of model to account for the
random variations across communities and internal correlations of repeated measures from each
subject over time. In addition, nonlinearity of BMI growth trajectory was tested by looking at the
slope of BMI growth over time from baseline age 6.5 years to 10 years. We found no deviations
from a linear trend as this is a relatively short period of growth ~3.5 years, and children in this
age range have not yet reached puberty where growth tends to be nonlinear. Therefore, linear
BMI growth was considered in the final analysis.
The current analysis included 2,318 children who: 1) completed a baseline questionnaire,
2) had at least two measures of BMI across the 4-year study follow-up period, 3) had NRAP
exposure data for the time windows of interest, and 4) moved homes prior to study enrollment to
avoid collinearity between NRAP exposures during the in utero or first year of life periods with
later childhood exposures (Figure 2.2; Supplemental Table 2.1). Children who lived in the
same residence from birth through follow-up had a high correlation of NRAP exposures across
each time period, therefore all analyses were restricted to “Movers”. “Movers” were subjects
who had a change in address between the in utero period and study enrollment that resulted in a
move farther than or equal to 500 meters. This selection method allowed us to look at later
NRAP exposures in childhood in the same model as the early life exposures, in utero and first
year of life. In the “Movers” group, NRAP exposure periods had lower correlations, which
allowed us to use the modeling framework described.
Mixed effects models were used to examine age-adjusted associations between baseline
characteristics and BMI growth as well as attained BMI at age 10 years. These characteristics

38

were identified as confounders and included in the final model if they resulted in a ≥ 10% change
in the effect estimate for BMI growth or attained BMI at age 10 years. Of the identified potential
confounders, age, sex, race/ethnicity, parental education, and Spanish baseline questionnaire
were included as confounders in the final model. Furthermore, effect modification by sex and
race/ethnicity was tested using interaction terms in the full model. Effect estimates of NRAP
exposure on BMI growth as well as attained BMI level at age 10 years are reported for a two-
standard deviation (SD) difference in near-road freeway or non-freeway NOx exposures during
each exposure window. Statistical significance was based on a two-sided p<0.05. All analyses
were performed in SAS, version 9.4 (SAS, Institute, Cary, NC).

2.4 Results
Baseline characteristics are reported in Table 2.1. At study entry, the mean age was 6.5
years (SD: 0.7) and 50.6% were male. Approximately 29% of children were overweight or obese
using CDC growth chart cutoffs (24) with a mean BMI percentile of 60.5 (SD: 30.1) at study
entry. Children were predominately Hispanic (56%) or Non-Hispanic White (33%), where 22%
of the parents completed the baseline questionnaire in Spanish, which was used a marker of
recent immigration status. On average, more than half of the children included in this study had
parents with an education level above high school. SHS exposure was relatively low where 7%
of mothers smoked during pregnancy and 5% of children lived in homes where someone smoked
daily in the presence of the child. At the end of the study follow up period, the mean age of
children was 9.5 years (SD: 1.2).
Age-adjusted associations between each baseline characteristic and BMI growth
trajectory and attained BMI at age 10 years are also shown in Table 2.1. Briefly, BMI growth

39

through the study period was associated with baseline overweight/obesity status, sex,
race/ethnicity, parental education, Spanish questionnaire, residential SHS as well as participating
in an organized team sport (p<0.05 for each characteristic). BMI at age 10 years was associated
with baseline overweight/obesity status, race/ethnicity, parental education, Spanish
questionnaire, residential SHS, and participating in an organized team sport (p<0.05 for each
characteristic). The correlation between in utero and childhood freeway NRAP exposures was
0.35 (p <0.0001) and the correlation between the first year of life and childhood freeway NRAP
exposures was 0.58 (p <0.0001) (Supplemental Table 2.1).
Because our analysis was limited to movers only, comparison of baseline characteristics
of movers and non-movers can be found in Supplemental Table 2.2. Movers were slightly older
at baseline (6.5 years, SD=0.7) compared to non-movers (6.3 years, SD=0,7). Racial/ethnics
groups differed between movers and non-movers with more Hispanic children in the non-movers
group, 63% versus 56% in movers. Movers has more parents with an above high school
education (62%) compared to non-movers (51%) and non-movers had more parents who filled
out the baseline questionnaire in Spanish (37%) than movers (22%). Movers had a higher
participation in organized team sports (42%) than non-movers (34%). Movers and non-movers
did not differ in obesity status, sex, self-reported premature birth, maternal smoking during
pregnancy, residential second-hand smoke, and lifetime history of asthma.
Residential NRAP exposures measured in near-road NOx for freeway and non-freeway
sources for in utero, first year of life and childhood periods are described in Table 2.2. Mean
near-road NOx exposures from freeway roadways during in utero, first year of life, and
childhood were 16.7 parts per billion (ppb) (SD: 20.1), 16.2 ppb (SD: 19.5), and 15.1 ppb (SD:
18.9), and mean near-road NOx exposures from non-freeway roadways during in utero, first year

40

of life, and childhood were 10.3 ppb (SD: 7.4), 9.3 ppb (SD: 6.7), and 6.2 ppb (SD: 4.7),
respectively.

Associations of early life NRAP exposures with childhood BMI
First year of life exposures to NRAP from freeway roads were positively associated with
BMI at age 10 years and BMI growth during study follow up and these associations were
independent of mid-childhood NRAP exposures (Table 2.3). For first year of life (model 2), a
39.1 ppb difference in near-road freeway NOx exposure was significantly associated with a 0.1
kg/m
2
(95% confidence interval (CI): 0.03, 0.2) faster increase in BMI per year resulting in a 0.5
kg/m
2
(95% CI: 0.02, 0.9) higher BMI at age 10 years. For in utero (model 1), a 40.1 ppb
difference in near-road freeway NOx exposure was associated with a 0.05 kg/m
2
(95% CI: -0.02,
0.1) faster increase in BMI per year and a 0.1 kg/m
2
(95% CI: -0.3, 0.5) higher BMI at age 10
years; however, these estimates did not reach statistical significance after adjusting for
confounders (Table 2.3). In contrast, non-freeway NRAP exposures during mid-childhood were
associated with BMI at age 10 years and BMI growth while early life non-freeway NRAP
exposures showed no significant association (Table 2.4). Additionally, near-road total NOx
exposures were similar in magnitude to that of near-road freeway NOx exposures
(Supplemental Table 2.3).
Based on tests for interaction and the analysis stratified by effect modifiers, there was
little evidence to support differences in effects of early life freeway NRAP exposure by sex
(males versus females), race/ethnicity (Non-Hispanic Whites versus Hispanics), and baseline
overweight/obese status (overweight/obese versus normal BMI) (Supplemental Table 2.4-2.6).
In a subsample of children who had complete data of birth weight and gestational age (n=2,129),

41

birth weight and gestational age did not significantly change the effects of in utero or first year
of life near-road freeway NOx exposure on BMI growth and BMI at age 10 years (Supplemental
Table 2.7). We also explored effects of in utero and first year of life ambient PM2.5 exposures on
childhood BMI trajectory however we did not see any significant associations (Supplemental
Table 2.8). We conducted further sensitivity analysis comparing independent contributions of in
utero or first year of life near-road freeway NOx exposures without adjustments of mid-
childhood exposures on BMI trajectory for movers and non-movers to assess possibly of
selection bias (Supplemental Table 2.9). We found similar growth trajectories amongst movers
and non-movers when looking at in utero or first year of life NRAP exposures.

2.5 Discussion
In our study population of school-aged children in Southern California, higher first year
of life NRAP from freeway sources were associated with faster increases in BMI during
childhood after adjusting for confounders such as age, sex, race/ethnicity, parental education,
Spanish questionnaire, and mid-childhood NRAP exposures. These longitudinal associations
were independent of mid-childhood NRAP exposures and resulted in significant differences in
BMI at age 10 years, suggesting that early life NRAP exposures may represent important
windows of exposure that increase risk for developing childhood obesity. Previous studies of this
cohort have shown that increased mid-childhood NRAP was associated with increased BMI
growth where children in the highest 10% of non-freeway NOx exposure showed a 0.39 kg/m
2

higher BMI at age 10 years and a 0.087 kg/m
2
faster increase in BMI per year when compared to
those in the lowest 10% of exposure using a similar modeling approach (9). We found consistent
results as those reported in Jerrett et al 2014 using our childhood, non-freeway NRAP exposure

42

in our subset of children with longitudinal BMI measures, movers only, as well as movers with
in utero NRAP exposures, and movers with first year of life NRAP exposures suggesting that the
NRAP association were not the result of selecting movers for this study (Supplemental Table
2.10).
Collectively, these results suggest that specific components of NRAP exposure in
freeway versus non-freeway mixtures may independently contribute to obesity risk during
different periods of early life and childhood. Vehicle exhaust is the main contributor to NRAP
and chemical composition of freeway and non-freeway roads have shown to be different due to
differences in vehicle types and vehicle volume (29). For example, heavy duty diesel trucks with
compression ignition engines travel most densely on freeways compared to non-freeway roads
particularly in Southern California, and diesel truck emissions also differ from spark ignition
engine emissions which are primarily gasoline derived passenger vehicles. Furthermore, a study
in Texas also showed that three different road types had notable differences in chemical
composition due to the varying vehicle types and emissions (30, 31). Due to differences in total
volume and emissions from diesel engine versus gasoline engine vehicles, chemical composition
downwind of freeway and non-freeway roadways is expected to differ.
Our findings build on previous work in animal studies that have reported associations of
early life air pollution exposures and obesity (32, 33). In mice, in utero exposure to diesel
exhaust predisposed offspring to higher weight gain when fed a high fat diet compared to those
offspring exposed to filtered air and given a high fat diet (32). Additionally, Sprague Dawley rats
exposed to unfiltered Beijing air pollution prenatally and continuously after birth had
significantly higher fat mass at 8 weeks (33). Together, results from our study coupled with
animal models suggest that early life exposures may represent a critical window of exposure

43

where increased NRAP may result in increased risk for higher childhood BMI trajectories, which
in turn may lead to childhood obesity.
The biological mechanisms linking air pollution exposure with increased childhood BMI
remain uncertain. However, animal models suggest inflammatory pathways where increased air
pollution exposures have been shown to result in higher levels of circulating proinflammatory
cytokines, inflammation in the lungs, lower levels of anti-inflammatory cytokines as well as
adipose tissue inflammation (33-35). Exposures may also have effects on the brain via
neuroinflammation where increased in utero air pollution exposure was shown to stimulate
appetite or anxiety induced over-eating in mice (32). Additionally, animal models suggest that
increased polycyclic aromatic hydrocarbon exposure, a byproduct of traffic combustion, may
increase white adipose tissue accumulation and inhibit lipolysis (36, 37). These animal studies
have uncovered some of the potential mechanisms underlying the associations between increased
in utero and early life NRAP exposures with obesity. Our study builds on this work by showing
longitudinal associations between early life NRAP exposure and BMI growth in children 6-10
years of age.
Despite the strengths of this study, dietary data was unavailable and residual confounding
may have occurred since poor diet, such as increased sugar-sweetened beverage consumption, is
associated with increased risk of childhood obesity (38, 39) and NRAP exposure through lower
socioeconomic position (40). However, the models adjusted for important covariates that may be
related to these factors and it is unlikely that results are fully explained by residual confounding.
The current study used residential based estimates of NRAP exposure with near-road NOx
concentrations as a marker of NRAP exposures. While exposure misclassification may have
occurred, misclassification should be random amongst subjects therefore biasing estimates

44

toward the null. This study was also limited to BMI from annual height and weight measures
since direct measures of adiposity were not performed (41). Lastly, since in utero and first year
of life NRAP exposures were highly correlated, we are unable to conclude the relative
contribution of each exposure window (i.e., in utero versus first year of life) to increased
childhood BMI.
Our results show that increased first year of life near-road freeway NOx exposures are
associated with increased velocity of childhood BMI growth trajectory and higher attained BMI
at 10 years and remained robust after controlling for multiple confounders as well as childhood
near-road freeway NOx exposures. Furthermore, increased childhood near-roadway exposures
from non-freeway sources were associated with increased BMI growth and higher BMI at 10
years, consistent with our past findings (cite). These results, along with other epidemiological
and animal studies, implicate environmental exposures such as NRAP as potential risk factors for
higher childhood BMI growth and higher attained BMI at age 10 years. These findings have
significant public health relevance for intervention since the number of children living near
freeways and busy roadways is large and continues to increase (42). Additional epidemiological
and experimental studies are warranted to examine the mechanisms by which early life NRAP
exposures may impact early growth trajectories in children.

45

2.6 References:
1. Ogden CL, Carroll MD, Kit BK, et al. Prevalence of childhood and adult obesity in the
United States, 2011-2012. JAMA 2014;311(8):806-14.
2. Singh AS, Mulder C, Twisk JW, et al. Tracking of childhood overweight into adulthood:
a systematic review of the literature. Obes Rev 2008;9(5):474-88.
3. Ward ZJ, Long MW, Resch SC, et al. Simulation of Growth Trajectories of Childhood
Obesity into Adulthood. N Engl J Med 2017;377(22):2145-53.
4. Weiss R, Kaufman FR. Metabolic complications of childhood obesity: identifying and
mitigating the risk. Diabetes Care 2008;31 Suppl 2:S310-6.
5. Franks PW, Hanson RL, Knowler WC, et al. Childhood obesity, other cardiovascular risk
factors, and premature death. N Engl J Med 2010;362(6):485-93.
6. Freedman DS, Mei Z, Srinivasan SR, et al. Cardiovascular risk factors and excess
adiposity among overweight children and adolescents: the Bogalusa Heart Study. J
Pediatr 2007;150(1):12-7.e2.
7. Hannon TS, Rao G, Arslanian SA. Childhood obesity and type 2 diabetes mellitus.
Pediatrics 2005;116(2):473-80.
8. Ogden CL, Carroll MD, Fryar CD, et al. Prevalence of Obesity Among Adults and
Youth: United States, 2011-2014. NCHS Data Brief 2015(219):1-8.
9. Jerrett M, McConnell R, Wolch J, et al. Traffic-related air pollution and obesity
formation in children: a longitudinal, multilevel analysis. Environ Health 2014;13:49.
10. McConnell R, Shen E, Gilliland FD, et al. A longitudinal cohort study of body mass
index and childhood exposure to secondhand tobacco smoke and air pollution: the
Southern California Children's Health Study. Environ Health Perspect 2015;123(4):360-
6.
11. Jerrett M, McConnell R, Chang CC, et al. Automobile traffic around the home and
attained body mass index: a longitudinal cohort study of children aged 10-18 years. Prev
Med 2010;50 Suppl 1:S50-8.
12. Janesick AS, Blumberg B. Obesogens: an emerging threat to public health. Am J Obstet
Gynecol 2016;214(5):559-65.
13. Rundle A, Hoepner L, Hassoun A, et al. Association of childhood obesity with maternal
exposure to ambient air polycyclic aromatic hydrocarbons during pregnancy. Am J
Epidemiol 2012;175(11):1163-72.
14. Mao G, Nachman RM, Sun Q, et al. Individual and Joint Effects of Early-Life Ambient
PM2.5 Exposure and Maternal Pre-Pregnancy Obesity on Childhood Overweight or
Obesity. Environ Health Perspect 2016.
15. Fioravanti S, Cesaroni G, Badaloni C, et al. Traffic-related air pollution and childhood
obesity in an Italian birth cohort. Environ Res 2018;160:479-86.
16. Adair LS. Child and adolescent obesity: epidemiology and developmental perspectives.
Physiol Behav 2008;94(1):8-16.
17. Dietz WH. Critical periods in childhood for the development of obesity. Am J Clin Nutr
1994;59(5):955-9.
18. Dadvand P, Parker J, Bell ML, et al. Maternal exposure to particulate air pollution and
term birth weight: a multi-country evaluation of effect and heterogeneity. Environ Health
Perspect 2013;121(3):267-373.

46

19. Ghosh JK, Wilhelm M, Su J, et al. Assessing the influence of traffic-related air pollution
on risk of term low birth weight on the basis of land-use-based regression models and
measures of air toxics. Am J Epidemiol 2012;175(12):1262-74.
20. Pedersen M, Giorgis-Allemand L, Bernard C, et al. Ambient air pollution and low
birthweight: a European cohort study (ESCAPE). Lancet Respir Med 2013;1(9):695-704.
21. Ong KK, Ahmed ML, Emmett PM, et al. Association between postnatal catch-up growth
and obesity in childhood: prospective cohort study. BMJ 2000;320(7240):967-71.
22. Fleisch AF, Rifas-Shiman SL, Koutrakis P, et al. Prenatal exposure to traffic pollution:
associations with reduced fetal growth and rapid infant weight gain. Epidemiology
2015;26(1):43-50.
23. McConnell R, Berhane K, Yao L, et al. Traffic, susceptibility, and childhood asthma.
Environ Health Perspect 2006;114(5):766-72.
24. Kuczmarski RJ, Ogden CL, Guo SS, et al. 2000 CDC Growth Charts for the United
States: methods and development. Vital Health Stat 11 2002(246):1-190.
25. Goldberg D. Texas A&M University Geoservices. Available online at
http://geoservices.tamu.edu. 2016. (Accessed June 1 2016).
26. Benson P. A review of the development and application of the CALINE3 and 4 models.
Atmospheric Environment Part B Urban Atmosphere 1992;26(3):379-90.
27. Diggle P, Heagerty P, Liang K-Y, et al. Analysis of longitudinal data. 2002. Oxford
Statistical Science Series 2002.
28. Berhane K, Gauderman W, Stram D, et al. Statistical issues in studies of the long term
effects of air pollution: The Southern California Children's Health Study. Stat Sci
2004;19(3):414-49.
29. Fujita EM, Zielinska B, Campbell DE, et al. Variations in speciated emissions from
spark-ignition and compression-ignition motor vehicles in California's south coast air
basin. J Air Waste Manag Assoc 2007;57(6):705-20.
30. Clements AL, Jia YL, Denbleyker A, et al. Air pollutant concentrations near three Texas
roadways, part II: Chemical characterization and transformation of pollutants.
Atmospheric Environment 2009;43(30):4523-34.
31. Zhu YF, Pudota J, Collins D, et al. Air pollutant concentrations near three Texas
roadways, Part I: Ultrafine particles. Atmospheric Environment 2009;43(30):4513-22.
32. Bolton JL, Smith SH, Huff NC, et al. Prenatal air pollution exposure induces
neuroinflammation and predisposes offspring to weight gain in adulthood in a sex-
specific manner. FASEB J 2012;26(11):4743-54.
33. Wei Y, Zhang JJ, Li Z, et al. Chronic exposure to air pollution particles increases the risk
of obesity and metabolic syndrome: findings from a natural experiment in Beijing.
FASEB J 2016.
34. Sun Q, Yue P, Deiuliis JA, et al. Ambient air pollution exaggerates adipose inflammation
and insulin resistance in a mouse model of diet-induced obesity. Circulation
2009;119(4):538-46.
35. Liu C, Xu X, Bai Y, et al. Air pollution-mediated susceptibility to inflammation and
insulin resistance: influence of CCR2 pathways in mice. Environ Health Perspect
2014;122(1):17-26.
36. Irigaray P, Ogier V, Jacquenet S, et al. Benzo[a]pyrene impairs beta-adrenergic
stimulation of adipose tissue lipolysis and causes weight gain in mice. A novel molecular
mechanism of toxicity for a common food pollutant. FEBS J 2006;273(7):1362-72.

47

37. Ravindra KS, R.; Van Griekenb, R. Atmospheric polycyclic aromatic hydrocarbons:
Source attribution, emission factors and regulation. Atmospheric Environment
2008;42(13):2895-921.
38. Ludwig DS, Peterson KE, Gortmaker SL. Relation between consumption of sugar-
sweetened drinks and childhood obesity: a prospective, observational analysis. Lancet
2001;357(9255):505-8.
39. Davis JN, Whaley SE, Goran MI. Effects of breastfeeding and low sugar-sweetened
beverage intake on obesity prevalence in Hispanic toddlers. Am J Clin Nutr 2012;95(1):3-
8.
40. Green RS, Smorodinsky S, Kim JJ, et al. Proximity of California public schools to busy
roads. Environ Health Perspect 2004;112(1):61-6.
41. Freedman DS, Wang J, Maynard LM, et al. Relation of BMI to fat and fat-free mass
among children and adolescents. Int J Obes (Lond) 2005;29(1):1-8.
42. Greenbaum D. Traffic- Related Air Pollution: A Critical Review of the Literature on
Emissions, Exposure, and Health Effects. Health Effects Institute, 2009.

48

2.7 Tables

Table 2.1 Baseline characteristics and age-adjusted associations with BMI growth and BMI at
age 10 years in children enrolled in the longitudinal Children’s Health Study
a

Characteristic n (%)
b

Associations with
BMI growth
c

(95% CI)
Associations with
BMI at Age 10
years
c
(95% CI)
Overweight/Obesity Status
d

Normal 1651 (71.2)

reference reference
Overweight 320 (13.8)

0.4 (0.3, 0.5) 4.1 (3.8, 4.5)
Obese 347 (15.0)

0.7 (0.7,0.8) 9.0 (8.6, 9.3)
Sex

Female 1145 (49.4)

reference reference
Male 1173 (50.6)

0.06 (0.004, 0.1) 0.3 (-0.06, 0.7)
Race/Ethnicity

White 771 (33.3)

reference reference
Hispanic 1290 (55.7)

0.3 (0.2, 0.4) 1.9 (1.6, 2.3)
Black 72 (3.1)

0.2 (-0.02, 0.3) 1.2 (0.1, 2.2)
Asian/Pacific Islander 67 (2.9)

0.2 (-0.006, 0.4) 0.9 (-0.2, 2.0)
Other 114 (4.9)

0.2 (0.04, 0.3) 0.7 (-0.2, 1.5)
Parental Education

Less than high school 410 (18.4)

reference reference
High school 430 (19.3)

-0.2 (-0.3, -0.06) -1.0 (-1.6, -0.5)
Above high school 1389 (62.3)

-0.3 (-0.4, -0.2) -2.0 (-2.4, -1.4)
Spanish Questionnaire
e

No 1818 (78.4)

reference reference
Yes 500 (21.6)

0.2 (0.2, 0.3) 1.7 (1.3, 2.1)
Self-Reported Premature
Birth

No 2003 (88.9)

reference reference
Yes 251 (11.1)

-0.01 (-0.1, 0.08) -0.6 (-1.1, 0.02)
Maternal Smoking During
Pregnancy

No 2083 (92.7)

reference reference
Yes 164 (7.3)

-0.01 (-0.1, 0.1) -0.5 (-1.2, 0.2)
Residential Second-Hand
Smoke
f

No 2110 (93.2)

reference reference
Yes, child is home 110 (4.9)

0.1 (0.02, 0.2) 0.3 (0.07, 0.6)
Yes, child is not home 44 (1.9)

0.1 (-0.04, 0.2) 0.2 (-0.2, 0.5)

49

Lifetime History of Asthma
No 1928 (85.1)

reference reference
Yes 338 (14.9)

-0.04 (-0.1, 0.008) 0.09 (-0.08, 0.2)
Organized Team Sport
g

No 1141 (57.8)

reference reference
Yes
832 (42.2)

-0.08 (-0.1, -0.04)
-0.2 (-0.3, -0.09)
a
This analysis includes a subset of the Children’s Health Study participants who had available NRAP exposure data
for in utero or first year of life periods, at least two measures of BMI during study follow up period, completed a
baseline questionnaire, and had moved homes at least once before study enrollment.
b
First observation of participant with NRAP exposures (n=2,318); variable denominators may differ due to missing
values.
c
Age-adjusted association for each characteristic with BMI growth over study follow up period and attained BMI at
age 10 years.
d
Overweight/ Obesity status: Normal is < 85
th
percentile of age-, sex-specific BMI using 2000 CDC growth chart,
overweight is 85-95th percentile of age-, sex-specific BMI, obese is ≥ 95th percentile of age-, sex- specific BMI.
e
Spanish Questionnaire is if parent filled out baseline questionnaire in Spanish and serves as a surrogate measure for
recent immigration.
f
Residential second-hand smoke is if anyone living in the child’s home smokes daily inside the home.
g
Organized team sport is if the child played outdoors in any organized team sport at least twice a week during the
past year.

50

Table 2.2 Residential NRAP exposures from freeway and non-freeway sources for in utero, first
year of life, and childhood periods in children in the CHS.
Exposure Period Mean ± SD Median IQR Range
Freeway NOx (ppb)
In utero 16.7 ± 20.1 10.8 4.4-22.1 0-233.3
First year of life 16.2 ± 19.5 10.4 4.1-21.9 0-304.2
Childhood 15.1 ± 18.9 9.6 4.6-18.1 0-351.0
Non-Freeway NOx (ppb)
In utero 10.3 ± 7.4 8.7 5.2-13.7 0.0003-74.0
First year of life 9.3 ± 6.7 7.9 4.6-12.2 0.0003-77.4
Childhood 6.2 ± 4.7 5.2 3.2-7.6 0.09-65.7

51

Table 2.3 Effects of in utero/first year of life and childhood freeway NOx exposures on 4-year
childhood BMI trajectories in CHS children.
Freeway NOx (ppb)
BMI Growth Per Year
a

Effect (95% CI)
BMI at Age 10 Years
a

Effect (95% CI)
Model 1
In utero (n=2,072) 0.05 (-0.02, 0.1) 0.1 (-0.3, 0.5)
Childhood -0.02 (-0.1, 0.05) 0.05 (-0.4, 0.5)
Model 2
First year of life (n=2,318) 0.1 (0.03, 0.2)* 0.5 (0.02, 0.9)*
Childhood -0.06 (-0.1, 0.02) -0.1 (-0.6, 0.3)
a
BMI growth (kg/m
2
) over study follow up and difference in attained BMI at age 10 years scaled to 2 standard
deviations of in utero freeway NOx exposure with 40.1 ppb, first year of life freeway NOx with 39.1 ppb, and
childhood freeway NOx with 37.8 ppb.
*p<0.05.

52

Table 2.4 Effects of in utero/first year of life and childhood non-freeway NOx exposures on 4-
year childhood BMI trajectories in children in CHS.
Non-Freeway NOx (ppb)
BMI Growth Per Year
a

Effect (95% CI)
BMI at Age 10 Years
a

Effect (95% CI)
Model 1
In utero (n=2,072) 0.03 (-0.05, 0.1) 0.1 (-0.3, 0.6)
Childhood 0.08 (-0.007, 0.2) 0.6 (0.08, 1.03)*
Model 2
First year of life (n=2,318) -0.02 (-0.1, 0.06) -0.07 (-0.5, 0.4)
Childhood 0.1 (0.01, 0.2)* 0.6 (0.1, 1.1)*
a
BMI growth (kg/m
2
) over study follow up and difference in attained BMI at age 10 years scaled to 2 standard
deviations of in utero non-freeway NOx with 14.7 ppb, first year of life non-freeway NOx with 18.7 ppb, and
childhood non-freeway NOx with 9.4 ppb.
*p<0.05.

53

2.8 Figures

Figure 2.1 Map of Children’s Health Study communities.

54

Figure 2.2 Flow chart of children enrolled in the Children’s Health Study from 2002-2003
included and excluded from the current analysis

In Utero NRAP
Participants with in utero
NRAP exposure measures
N=2,072
Children’s Health Study
(Cohort E)
Children with non-missing BMI,
age, sex, baseline questionnaire
N=5,337
Longitudinal BMI
Children with ≥ 2 BMI measures
N=4,400
Non-movers
Children who moved
residences < 500 meters
before study enrollment
N=976
Movers
Children who moved
residences ≥ 500 meters
before study enrollment
N=3,424
First Year of Life NRAP
Participants with first year
of life NRAP exposure
measures
N=2,318

55

2.9 Supplemental Tables

Supplemental Table 2.1 Pearson Correlation Coefficients Between In Utero/First Year of Life
and Childhood NRAP Exposure in Non-Movers

and Movers

in Children in the Children’s Health
Study.

Non-Movers
a
Movers
b
NRAP Exposure Period
(Freeway sources)
In Utero First Year of Life

In Utero First Year of Life
Childhood 0.97* 0.98*

0.35* 0.58*
NRAP Exposure Period
(Non-freeway sources)
In Utero First Year of Life In Utero First Year of Life
Childhood 0.98* 0.98* 0.57* 0.65*
a
Non-movers were children who did not have a change in address or moved less than 500 meters between in utero
period and study entry.
b
Movers were children who had a change in address between in utero period and CHS study entry that resulted in a
move greater than or equal to 500 meters.
*p<.0001

56

Supplemental Table 2.2 Baseline Characteristics of Movers and Non-Movers with Early Life
NRAP Exposures and Who Were Enrolled in the Longitudinal Children’s Health Study
a
.
Characteristic
Movers
b
Non-Movers
c
P-value
No. (%)
a

No. (%)
a

Obesity Status 0.1
Normal weight 1651 (71.2) 605 (67.5)
Overweight 320 (13.8)

142 (15.9)
Obese 347 (15.0)

149 (16.6)
Sex

0.3
Female 1145 (49.4)

462 (51.6)
Male 1173 (50.6)

434 (48.4)
Race/Ethnicity

<0.0001
White 771 (33.3)

223 (24.9)
Hispanic 1290 (55.7)

564 (63.0)
Black 72 (3.1)

27 (3.0)
Asian/Pacific Islander 67 (2.9)

34 (3.8)
Other 114 (4.9)

41 (4.6)
Parental Education

<0.0001
Less than high school 410 (18.4)

249 (30.4)
High school 430 (19.3)

149 (18.2)
Above high school 1389 (62.3)

420 (51.3)
Spanish Questionnaire
d

No 1818 (78.4)

562 (62.7) <0.0001
Yes 500 (21.6)

334 (37.3)
Self-reported premature birth

0.1
No 2003 (88.9)

763 (90.9)
Yes 251 (11.1)

76 (9.1)
Maternal smoking during
pregnancy

0.8
No 2083 (92.7)

770 (92.7)
Yes 164 (7.3)

61 (7.3)
Residential second hand smoke
e

0.9
No 2110 (93.2)

784 (92.8)
Yes, when child is home 110 (4.9)

41 (4.9)
Yes, when child is not home 44 (1.9)

20 (2.3)
Life-time history of asthma

0.1
No 1928 (85.1)

744 (87.1)
Yes 338 (14.9)

110 (12.9)
Organized team sport
f

<0.0001
No 1141 (57.8)

449 (66.3)
Yes 832 (42.2)

228 (33.7)
a
This analysis includes a subset of the Children’s Health Study participants who had available NRAP exposure data
for in utero or first year of life periods. Movers were defined as those who had a change in address between in utero

57

period and study entry that resulted in a move ≥ 500 meters. Non-movers were those who did not move > 500
meters before study entry.
b
First observation of participants with NRAP exposures in movers (n=2318); variable denominators may differ due
to missing values.
c
First observation of participants with NRAP exposures in non-movers (n=896); variable denominators may differ
due to missing values.
d
Spanish Questionnaire is if parent filled out baseline questionnaire in Spanish and serves as a surrogate measure
for recent immigration.
e
Residential second-hand smoke is if anyone living in the child’s home smokes daily inside the home.
f
Organized team sport is if the child played outdoors in any organized team sport at least twice a week during the
past year.

58

Supplemental Table 2.3 Effects of In Utero/First Year of Life and Childhood Near-Road Total
NOx Exposure on 4-Year Childhood BMI Trajectories.
Total NOx Exposure (ppb)
BMI Growth Per Year
b

Effect (95% CI)
BMI at Age 10 Years
b

Effect (95% CI)
In utero (n=2,072) 0.06 (-0.009, 0.1) 0.2 (-0.2, 0.6)
Childhood
-0.009 (-0.09, 0.07) 0.2 (-0.3, 0.6)
First year of life (n=2,318)

0.1 (0.03, 0.2)* 0.5 (0.02, 0.9)*
Childhood -0.04 (-0.1, 0.04) -0.05 (-0.5, 0.4)
a
Total NO x= Freeway + Non-freeway NO x
b
BMI growth and BMI at age 10 years scaled to 2 standard deviations of in utero total NOx exposure with 46.1 ppb,
first year of life total NOx with 44.9 ppb and childhood total NOx with 42.2 ppb. Models adjusted for age, sex,
race/ethnicity, parental education, and Spanish questionnaire.
*p<0.05.

59

Supplemental Table 2.4 Effects of In Utero/First Year of Life Near-Road Freeway NOx on 4-
Year Childhood BMI Trajectories for Male and Female Children.
Freeway NOx
Exposure (ppb)
BMI growth per year
a

Effect (95% CI)

BMI at age 10 years
a

Effect (95% CI)
Male Female

Male Female
In utero
b

0.08
(-0.01, 0.2)
0.009
(-0.09, 0.1)

0.4
(-0.2, 0.9)
-0.07
(-0.6, 0.5)
First year of life
c
0.1
(0.02, 0.2)*
0.07
(-0.02,0.2)

0.7
(0.07, 1.1)*
0.2
(-0.5, 0.8)
a
BMI growth and BMI at age 10 years scaled to 2 standard deviations of in utero near-road freeway NO x exposure
with 40.1ppb and first year of life near-road freeway NO x with 39.1 ppb. Models adjusted for age, race/ethnicity,
parental education, Spanish questionnaire, and childhood near-road freeway NOx exposure.
b
In utero model, males=1057, females=1014.
c
First year of life model, males=1173, females=1145.
*p<0.05
Interaction p-values for in utero: p interaction BMI growth=0.05, p interaction BMI at age 10=0.19
Interaction p-values for first year of life: p interaction BMI growth=0.062, p interaction BMI at age 10=0.25

60

Supplemental Table 2.5 Effects of In Utero/First Year of Life Near-Road Freeway NOx on 4-
Year Childhood BMI Trajectories for Non-Hispanic White/Hispanic Children.
Freeway NOx
Exposure (ppb)
BMI growth per year
a

Effect (95% CI)

BMI at age 10 years
a

Effect (95% CI)
Non-Hispanic
White
Hispanic

Non-Hispanic
White
Hispanic
In utero
b

0.09
(-0.03, 0.2)
0.01
(-0.07, 0.1)

0.5
(-0.2, 1.1)
-0.07
(-0.6, 0.5)
First year of life
c
0.1
(-0.04, 0.3)
0.08
(-0.009,0.2)

0.7
(-0.1, 1.4)
0.3
(-0.3, 0.9)
a
BMI growth and BMI at age 10 years scaled to 2 standard deviations of in utero near-road freeway NO x exposure
with 40.1 ppb and first year of life with 39.1 ppb. Models adjusted for age, sex, parental education, Spanish
questionnaire, and childhood near-road freeway NOx.
b
In utero model: White, n=695; Hispanic, n=1151.
c
First year of life: White, n=771; Hispanic n=1289.
Interaction p-values for in utero: p interaction BMI growth=0.078, p interaction BMI at age 10=0.35
Interaction p-values for first year of life: p interaction BMI growth=0.0.013, p interaction BMI at age 10=0.2

61

Supplemental Table 2.6 Effects of In Utero/First Year of Life Near-Road Freeway NOx on 4-
Year Childhood BMI Trajectories for Baseline Overweight/Obese and Normal BMI Children.
Freeway NOx
Exposure (ppb)
BMI growth per year
a

Effect (95% CI)

BMI at age 10 years
a

Effect (95% CI)
Baseline
Overweight or
Obese
b

Baseline
Normal BMI
c

Baseline
Overweight or
Obese
b

Baseline
Normal BMI
c

In utero
d

0.06
(-0.07, 0.2)
0.03
(-0.04, 0.1)

0.2
(-0.5, 0.9)
0.07
(-0.2, 0.3)
First year of life
e
0.07
(-0.06, 0.2)
0.08
(-0.001,0.2)

0.09
(-0.6, 0.8)
0.2
(-0.1, 0.5)
a
BMI growth and BMI at age 10 years scaled to 2 standard deviations of in utero near-road freeway NO x exposure
with 40.1 ppb and first year of life NO x with 39.1 ppb. Models adjusted for age, sex, race/ethnicity, parental
education, Spanish questionnaire.

b
Baseline overweight/obese= age-, sex-specific CDC BMI percentile ≥ 85.
c
Normal BMI= age-, sex-specific CDC BMI percentile <85.
d
In utero: overweight/obese, n=591; normal, n=1480.
e
First year of life: overweight/obese, n=667; normal, n=1651.
Interaction p-values for in utero: p interaction BMI growth<0.0001, p interaction BMI at age 10<0.0001.
Interaction p-values for first year of life: p interaction BMI growth<0.0001, p interaction BMI at age 10<0.0001.

62

Supplemental Table 2.7 Effects of In Utero/First Year of Life Near-Road Freeway NOx on 4-
year Childhood BMI Trajectories Adjusting for Birth Weight and Gestational Age.
Freeway NOx
Exposure (ppb)
BMI Growth Per Year
a

Effect (95% CI)
BMI at Age 10 Years
a

Effect (95% CI)
In utero (n=1,926) 0.06 (-0.01, 0.1) 0.2 (-0.2, 0.6)
First year of life (n=2,129)

0.1 (0.03, 0.2)* 0.5 (0.04, 0.9)*
a
BMI growth and BMI at age 10 years scaled to 2 standard deviations of in utero near-road freeway NO x exposure
with 40.1 ppb and first year of life NO x with 39.1 ppb. Models adjusted for age, sex, race/ethnicity, parental
education, Spanish questionnaire, childhood near-road freeway NO x exposure, birth weight (grams), and gestational
age (days).
*p<0.05.

63

Supplemental Table 2.8 Effects of In Utero/First Year of Life and Childhood Ambient PM2.5
Exposure on 4-Year Childhood BMI Trajectories.
Ambient PM2.5
Exposure (ug/m
3
)
BMI Growth Per Year
a

Effect (95% CI)
BMI at Age 10 Years
a

Effect (95% CI)
In utero (n=2,524)
-0.06 (-0.1, 0.02)
-0.6 (-1.1, -0.1)
Childhood
-0.03 (-0.1, 0.06)
0.3 (-0.1, 0.8)
First year of life (n=2,621)

0.03 (-0.1, 0.05) -0.5 (-0.9, -0.02)
Childhood -0.05 (-0.1, 0.04) 0.2 (-0.2, 0.7)
a
BMI growth and BMI at age 10 years scaled to 2 standard deviations of in utero ambient PM 2.5 exposure with 17.0
ug/m
3
, first year of life total NOx with 14.8 ug/m
3
and childhood with 9.9 ug/m
3
. Models adjusted for age, sex,
race/ethnicity, parental education, and Spanish questionnaire.

64

Supplemental Table 2.9 Independent Effects of In Utero or First Year of Life Near-Road
Freeway NOx Exposure on 4-Year Childhood BMI Trajectories in Movers and Non-Movers.
Freeway NOx
Exposure (ppb)
BMI growth per year
a

Effect (95% CI)

BMI at age 10 years
a

Effect (95% CI)
Movers Non-movers

Movers Non-movers
In utero only
b
0.04
(-0.03, 0.1)
0.05
(-0.05, 0.1)

0.2
(-0.2, 0.5)
0.7
(0.07, 1.3)*
First year of life only
c
0.07
(0.01, 0.1)*
0.04
(-0.06, 0.1)

0.4
(0.03, 0.7)*
0.6
(0.02, 1.2)*
a
BMI growth and BMI at age 10 years scaled to 2 standard deviations: for movers in utero freeway NOx exposure
with 40.1 ppb, first year of life freeway NOx with 39.1 ppb; for non-movers in utero freeway NOx exposure with
42.6 ppb, first year of life freeway NOx with 41.3 ppb. Models adjusted for age, sex, race/ethnicity, parental
education, and Spanish questionnaire.
b
In utero model, movers=2,072; non-movers=884.
c
First year of life movers=2,318; non-movers=896.
*p<0.05.

65

Supplemental Table 2.10 Effects of Childhood, Near-Road Non-Freeway NOx Exposure on 4-
year Childhood BMI Trajectories for Children Enrolled in the Children’s Health Study.
Childhood Non-freeway NOx Exposure (ppb)
BMI growth
Effect (95% CI)
BMI at age 10 years
Effect (95% CI)
Model 1: Subjects with 2+ BMI Measures 0.06 (0.02, 0.1)* 0.4 (0.2, 0.7)*
Model 2: Movers only 0.07 (0.02, 0.1)* 0.6 (0.3, 0.9)*
Model 3: Subjects with in utero exposures 0.07 (0.02, 0.1)* 0.5 (0.1, 0.8)*
Model 4: Subjects with first year of life exposures 0.07 (0.02, 0.1)* 0.4 (0.1, 0.8)*
a
BMI growth and BMI at age 10 years scaled to 2 standard deviations of childhood near-road non-freeway NOx
exposure with 9.4 ppb. Models adjusted for age, sex, race/ethnicity, parental education, and Spanish questionnaire.
*p<0.05.

66

Chapter 3: Early Life Near-Roadway Air Pollution, Genetic Susceptibility,
and Childhood Body Mass Index Growth: Analysis of a Genetic Risk Score-
Environment Interaction

3.1 Abstract
In chapter 2, we showed that higher first year of life freeway NOx exposure was
associated with a faster increase of childhood BMI per year resulting in higher attained BMI at
age 10 years independent of confounders and later childhood exposures. Using this longitudinal
mixed model, we evaluated the effect of a gene-environment interaction using an obesity related
genetic risk score (GRS) and early life exposures on 4-year childhood BMI trajectory. A subset
of Children’s Health Study participants with genetic data from chapter 2 was analyzed
(n=1,221). A GRS was created using 67 single nucleotide polymorphisms (SNPs) related to
obesity from the genome-wide association studies (GWAS) catalog that used studies from both
children and adults. First, longitudinal associations of BMI growth and attained BMI at age 10
years were evaluated with the GWAS derived GRS and covariates. Next, linear mixed effects
models then further assessed early life, in utero or first year of life, NRAP exposures (total,
freeway and non-freeway NOx), GRS-environment (GRSxE) interaction, and BMI growth after
adjusting for potential confounders and later childhood NRAP exposure. The GRS (alone
without any air pollution variables) was associated with attained BMI at age 10 years as well as
BMI growth after adjusting for confounders. For example, an increase of 1 risk allele in the GRS
was significantly associated with 0.006 kg/m
2
faster increase in BMI per year which resulted in a
0.07 kg/m
2
higher BMI at age 10 years. Similarly, comparing children with the highest 10% GRS
to those with the lowest 10% of GRS yielded a 0.27 kg/m
2
faster increase in BMI growth and a

67

2.6 kg/m
2
higher attained BMI at age 10 years, p<0.0001 and p=0.002 respectively. In the in
utero or first year of life NRAP models, no statistically significant associations with childhood
BMI trajectory and GRSxE were observed. This study suggests that known genetic determinants
that are related to obesity affect childhood BMI growth as well as attained BMI at age 10 years.
Further investigation of the effects of GRSxE on childhood BMI trajectory in a larger sample of
children is warranted.

68

3.2 Introduction
Low levels of physical activity and poor diet have been identified as major risk factors
for obesity. However, growing evidence suggests genetic susceptibility (1-3) and non-traditional
risk factors like air pollution (4-8) may play a role in the obesity epidemic. Genetic inheritance
of obesity has been well documented as it tracks within families (9, 10), and genetic factors
strongly influence body mass index (BMI) where high heritability for BMI has been reported
ranging from ~0.4-0.7 (11, 12). Amongst individuals, genetic effects may differ depending on
lifestyle and environmental factors due to gene-environmental interactions. For example, obesity
risk factors like diet have been previously shown to interact with genetic susceptibility that
results in increased obesity risk (13-16). Additionally, a recent study showed significant effects
of GxE interactions with a genetic score for BMI (derived from BMI-associated SNPs) and
several lifestyle factors (physical activity, alcohol consumption and socioeconomic status) on
BMI (17). With increasing evidence of early life air pollution exposures on obesity risk as well
as genetic susceptibility on obesity risk, studies evaluating the effect of the interaction of these
two obesity risk factors on childhood obesity are needed to inform prevention interventions.
Genome wide associations studies (GWAS) have reported several loci associated with
BMI and obesity (1, 2, 18, 19). One of the first common variants associated with BMI in adults
and children was found in the fat mass and obesity associated (FTO) gene with a 1.67 increase in
odds of obesity when comparing those with both risk alleles and those without a risk allele (2).
Some common variants in form of single nucleotide polymorphisms (SNP) associated with adult
BMI have also replicated in children. These SNPs are located in several different genes which
include and not limited to FTO, insulin induced gene 2 (INSIG2), melanocortin 4 receptor
(MC4R), transmembrane protein 18 (TMEM18), glucosamine-6-phosphate deaminase 2

69

(GNPDA2), neuronal growth regulator 1 (NEGR1), brain-derived neurotrophic factor (BDNF),
potassium channel tetramerization domain containing 15 (KCTD15), and 1q25 (19). Two new
childhood obesity loci were discovered by the Early Growth Genetics Consortium near
olfactomedin 4 (OLFM4) and within homeobox b5 (HOXB5) gene (18).
Several studies have assessed the association of childhood BMI growth and BMI/obesity
related genetic risk scores (GRS) (3, 20-22), however no studies to our knowledge have
investigated the effects of early life air pollution, GRS, and the interaction between a GRS and
early life air pollution exposures (GRSxE) on childhood BMI growth. Although BMI is a highly
heritable trait, the majority of genetic liability for obesity remains unknown as only 1.45% of the
variance in BMI has been explained by gene variants (1). There are likely other factors such as
GxE interactions that may contribute to the risk of obesity. In this study, we investigated the
association of early life NRAP exposure, obesity related GRS, its interaction (GRSxE) and
longitudinal BMI in children enrolled in the Children’s Health Study (CHS). There are several
aims of this study: first, we created a GWAS-derived GRS from obesity related SNPs. Second,
we assessed the longitudinal associations of the GRS with childhood BMI trajectory and whether
the associations were sex or race/ethnic specific. Third, we evaluated the association of early life
NRAP exposure, GRS, GRSxE interaction and childhood BMI trajectory.

3.3 Materials and Methods
Study Population and Potential Covariates
Cohort E of the CHS began recruitment in 2002-2003 school year where BMI measures
of height and weight were collected at baseline and each follow up school visit. Height was
measured to the nearest centimeter and weight to the nearest pound (0.45 kg) of each child

70

without shoes and with daily calibrations of the weight scale. BMI was calculated by
weight/height
2
(kilograms/meters
2
). Parents completed baseline and yearly follow-up
questionnaires relating to sociodemographic factors and other covariates that were explored as
potential confounders. These included age, sex, race/ethnicity, self-reported premature birth,
maternal smoking during pregnancy, residential second hand smoke (SHS), lifetime history of
asthma, parental education (marker for socioeconomic status), if baseline questionnaire was
completed in Spanish (marker for recent immigration status), and child’s participation in team
sports during the past year. Informed assents and consents were obtained from children and
parents. This study was reviewed and approved by the Institutional Review Board at the
University of Southern California.

Genotyping and Imputation
From 1996-2002, CHS participants gave DNA samples via buccal cells with a smaller
subsample with blood. Samples were genotyped by the University of Southern California
Epigenome Center using Illumina HumanHap550, HumanHap550-Duo and Human610-Quad
BeadChip microarrays. Because CHS GWAS was genotyped onto different chips (550k vs
610k), imputations were performed on those SNPs found on both chips and with a genotyping
rate ≥ 95%. SNPs were separately by ethnicity (non-Hispanic White or Hispanic) and those
SNPs failing the Hardy Weinberg equilibrium test were excluded. There were 1,901 Non-
Hispanic White (NHW) subjects and 475,294 SNPs and 1,944 Hispanic subjects and 473,532
SNPs across several cohorts of the CHS. SHAPEIT was used to phase the genotype data for each
chromosome before imputations were performed on Impute2 using the 1000 Genome Phase 3
reference panel.

71

Genetic Risk Score
A SNP list was created using key search terms of “obesity” in the GWAS catalog
(https://www.ebi.ac.uk/gwas/) (23). Seventy SNPs with genome wide significance (p<5x10
-8
)
were found that were significantly associated with obesity from 9 different GWAS studies in
adults and children (18, 24-31). Studies reported outcomes of obesity, obesity class I, II, III,
overweight, and extreme overweight. The SNP list was then further refined by looking at the
correlation amongst SNPs. Three SNPs were in high linkage disequilibrium with 3 other SNPs.
Highly correlated SNPs can simply act as a proxy for the other, so these three SNPs were
removed from the list. The final SNP list has 67 obesity related SNPs, and we extracted these 67
SNPs from individuals’ imputed data detailed above.
To create the genetic risk score, individual alleles were given a 0 or 1 (1 if risk allele, 0 if
not as reported in the GWAS catalog) which then assigns each allele pair a score between 0-2. If
no risk allele was reported, then the minor allele was used as the risk allele. The unweighted
GRS is the sum of all SNP scores by each participant with a higher GRS corresponding to a
greater number of total risk alleles. A weighted GRS was calculated by incorporating the natural-
logged published odds ratios to each allele pair score which gives more weight to those SNPs
that had greater odds of association with obesity. We created GRS quintiles where quintile 1 has
the 20% of participants with the lowest GRS scores (ie. least number of obesity related alleles)
and quintile 5 has 20% of participants with the highest GRS scores (the greatest number of
obesity related alleles). All levels of the quintile are compared to the reference group, the middle
quintile (quintile 3) to contrast mean risk versus high risk. Additionally, we made comparisons
between the highest 10% and lowest 10% of GRS.

72

NRAP Exposures
Lifetime residential history was collected from all CHS participants at study enrollment
and at each follow up CHS school visit. Exposures were estimated from street level geocoded,
residential addresses from in utero through the most current follow-up date. This provided
NRAP exposure estimates for in utero, first year of life and childhood periods. NRAP exposures
were estimated using the California line-source dispersion model (CALINE4). CALINE4
estimates concentrations of near-roadway nitrogen oxides (NOx) for freeway and non-freeway
roads using vehicle emissions, traffic volume, road geometry and meteorological conditions,
including wind speed and direction, pollution mixing heights, and atmospheric stability (32).
Nine months prior to birth was used as the average in utero NRAP and 12 months after birth was
used as the average first year of life NRAP. To account for childhood near-road NOx exposures
beyond the early life periods, childhood NRAP exposure was calculated as the average near-road
NOx exposure from 13 months of age through the 4-year study follow up period.

Statistical Methods
The longitudinal mixed effects model from chapter 2 was expanded to explore effects of
an obesity related GRS on the longitudinal association of BMI and early life NRAP exposures.
First, the association the unweighted or weighted GRS and GRS in quintiles will be evaluated in
the longitudinal model with BMI (Y
cij
) with the following:
Level 1: Y
cij
= a
ci
+ b
ci
(t
cij
− C) + γ
1
W
ij
+ε
cij

Level 2a (level): a
ci
= a
c
+ 𝛂 𝟏 𝐆𝐑𝐒 𝐢 + α
2
Z
i
+δ
ci

Level 2b (growth): b
ci
= β
0
+ 𝛃 𝟏 𝐆𝐑𝐒 𝐢 + β
2
Z
i
+ +σ
ci

Level 3a: a
c
= α
0
+ ε
c

73

In Level 1, 𝑡 𝑐𝑖𝑗
is the age of participants at each visit centered by attained age C (10 years) and
γ
1
W
ij
time dependent covariates. At level 2a and level 2b, 𝛂 𝟏 and 𝛃 𝟏 correspond to estimated
effects of GRS on attained BMI at 10 years and the growth of BMI per year, respectively.
Furthermore, α
2
Z
i
denotes adjustment factors at level, β
2
Z
i
are adjustment factors at growth,
ε
ij
, δ
ci
, σ
ci
reflect error terms at each level of the model and ε
c
accounts for the random effect of
community.
Next, we investigated the longitudinal association of early life NRAP exposures, GRSxE
and childhood BMI trajectory. This model builds on the previous longitudinal mixed model from
Chapter 2:
Level 1: Y
cij
= a
ci
+ b
ci
(t
cij
− C) + γ
1
(E
Fij
− E
Fi
) + γ
2
W
ij
+ε
cij

Level 2a (level): a
ci
= a
c
+ 𝛂 𝟏 (𝐄 𝐔𝐢
− 𝐄 𝐔
̅ ̅ ̅ ̅
) + α
2
E
Fi
+ α
3
Z
i
+ 𝛂 𝟒 (𝐆𝐑𝐒
𝐢 − 𝐆𝐑𝐒
̅ ̅ ̅ ̅ ̅ ̅
) +
𝛂 𝟓 (𝐄 𝐔𝐢
− 𝐄 𝐔
̅ ̅ ̅ ̅
) (𝐆𝐑𝐒
𝐢 − 𝐆𝐑𝐒
̅ ̅ ̅ ̅ ̅ ̅
) + δ
ci

Level 2b (growth): b
ci
= β
0
+ 𝛃 𝟏 (𝐄 𝐔𝐢
− 𝐄 𝐔
̅ ̅ ̅ ̅
) + β
2
E
Fi
+ β
3
Z
i
+ 𝛃 𝟒 (𝐆𝐑𝐒
𝐢 − 𝐆𝐑𝐒
̅ ̅ ̅ ̅ ̅ ̅
) +
𝛃 𝟓 (𝐄 𝐔𝐢
− 𝐄 𝐔
̅ ̅ ̅ ̅
) (𝐆𝐑𝐒
𝐢 − 𝐆𝐑𝐒
̅ ̅ ̅ ̅ ̅ ̅
) + σ
ci

Level 3a: a
c
= α
0
+ ε
c

Early life NRAP exposures and GRS were centered at the study mean value so the GRSxE is
more easily interpretable. In the fully combined model, 𝛂 𝟓 and 𝛃 𝟓 corresponds to effects of
GRSxE on attained BMI at age 10 years and BMI growth. Similarly, from Chapter 2,
γ1 represents cross-sectional association between year to year fluctuations of near-road NOx with
follow up BMI measure at each study visit. EFij reflects average near-road NOx exposure for the
time between each subsequent follow up visit and EFi is the average childhood near-road NOx
exposure from 13 months of age till last height/weight measure in 2006–2007 school year. The
model also includes average childhood near-road NOx exposures (EFi) while accounting for

74

yearly deviations of near-road NOx during this follow-up period (EFij − EFi) because we wanted to
elucidate associations of in utero or first year of life NRAP exposures (EUi) with BMI growth
independent of childhood NRAP exposures.
This analysis included 1,221 children with 1) a baseline CHS questionnaire, 2) at least
two measures of BMI across the 4-year follow up period, 3) NRAP data for in utero or first year
of life periods, 4) moved homes prior to study enrollment to avoid collinearity between NRAP
exposures during early life periods and later childhood exposures, 5) and finally genetic data for
the GRS (Figure 3.1). As previously described in chapter 2, this analysis included only
“movers” who are CHS participants who had a change in their residential address that resulted in
a move farther than or equal to 500 meters before study enrollment. Amongst “movers” NRAP
exposure periods (in utero vs childhood or first year of life vs childhood) had lower correlations,
which allowed us to use the modeling framework described (Supplemental Table 3.1).
Covariates were tested individually and those with 10% or greater change in effect
estimate of attained BMI or BMI growth were included in the final model. Of the identified
potential confounders, age, sex, race/ethnicity, parental education, and Spanish baseline
questionnaire were included in the final model. Effect estimates of the association between GRS
and BMI at level and growth reflect a 1 risk allele increase in the GRS. Effect modification by
sex and race/ethnicity were explored using interactions terms in the GRS-BMI growth model.
Effect estimates of NRAP exposure on BMI growth are scaled to two-standard deviations (SD)
of NRAP exposure during each exposure window. Statistical significance was based on a two-
sided p < 0.05. All analyses were performed in SAS, version 9.4 (SAS, Institute, Cary, NC).

75

3.4 Results
Details of the GWAS studies that were used to create the GRS for this analysis are
presented in Table 3.1. Briefly, of the 9 studies, discovery populations consisted of 4 studies in
adults only (24-26, 29), 2 studies in both adults and children (27, 28), and 3 studies in
exclusively in children (18, 30, 31). All GWAS studies reported associations with obesity
ranging from obese to extreme obesity. Characteristics of the CHS participants in this analysis
are shown in Table 3.2. At study entry, mean age was 6.5 years (SD=0.6) and 54% were male. In
this analysis, 15% of children were overweight and 13% were obese as defined by the age-, sex-
specific CDC growth charts (33). Participants were predominately Hispanic (62%) compared to
NHW (38%). Sixty three percent of the children had parents with an education above high school
and 23% completed the baseline questionnaire in Spanish, which was used a marker of recent
immigration status. SHS exposure was low where 7% of mothers smoked during pregnancy and
6% of children were exposed to residential SHS. Mean unweighted GRS was 65.1 (SD=7.6) with
a range of 44.2-88.1 and mean weighted GRS was 9.7 (SD=1.6) ranging from 5.6-14.3.
Distribution of quintiles of the weighted and unweighted GRS are shown in Supplemental
Table 3.2. At the end of the follow up period, children had a mean age of 10 years (SD=0.9).
Residential NRAP exposures measured in near-roadway NOx for in utero, first year of
life and childhood periods are shown in Table 3.3. Mean freeway NRAP during in utero, first
year of life, and childhood were 17.4 parts per billion (ppb) (SD= 21.3), 17.2 ppb (SD=21.0), and
16.3 ppb (SD=20.6), respectively; mean non-freeway NRAP during in utero, first year of life,
and childhood periods were 10.4 ppb (SD=7.4), 9.5 ppb (SD=7.0), and 6.2 ppb (SD=4.3),
respectively. Total NOx which is the combined NOx from non-freeway and freeway sources had

76

means 27.8 ppb (SD=24.0), 26.7 ppb (SD=23.5), 22.5 (SD=22.5) for in utero, first year of life,
and childhood periods, respectively.

Associations of Obesity-Related GRS and Childhood BMI Trajectory
Both unweighted and weighted GRS were statistically significantly associated with BMI
growth and attained BMI at age 10 years (Table 3.4a-b). Table 3.4a shows all movers who had
a data available for a GRS (n=1768) however sensitivity analysis was performed only including
participants with early life NRAP exposures as this will be our final analytical sample (n=1221,
Table 3.4b). A 1 risk allele increase in unweighted GRS was statistically, significantly
associated with a 0.006 kg/m
2
faster increase in BMI per year resulting in a 0.07 kg/m
2
higher
BMI at age 10 years after adjusting for confounders age, sex, race/ethnicity, parental education
and Spanish baseline questionnaire (Table 3.4b). Unweighted GRS quintiles show an increasing
trend for attained BMI at age 10 years and BMI growth with increasing quintiles; the highest
quintile, quintile 5, showed the strongest, statistically significant association with BMI growth
and attained BMI at age 10 years when compared to the mid quintile. Furthermore, children in
the 90
th
percentile of GRS (ie the highest 10%) compared to those children in the 10
th
percentile
(ie lowest 10%) yielded a 0.27 kg/m
2
faster increase in BMI growth and a 2.6 kg/m
2
higher
attained BMI at age 10 years, p<0.0001 and p=0.002 respectively. Similar trends were seen with
the weighted GRS and weighted quintiles (Tables 3.4b). We further investigated the association
between GRS and childhood BMI growth and if it were modified by sex (male vs female) or
race/ethnicity (NHW vs Hispanic) in Supplemental Table 3.3-3.4. We did not find any
statistically significant interactions for sex-GRS at level (p=0.07) or growth (p=0.97) or for
race/ethnicity-GRS at level (p=0.14) and growth (p=0.97).

77

After this initial analysis, we decided to move forward with the remaining analyses using
only the unweighted GRS. The weighted GRS puts weight into studies that may have been
derived from adult studies, where obesity classification is different in adults compared to
children. Adults have absolute BMI cut offs for differing obesity levels however in children the
CDC reference curve is used with BMI percentiles. Forty nine of the sixty seven SNPs used to
create our obesity related GRS came from one adult study (24) and because of this, the
unweighted GRS may be a more conservative approach to reflect genetic risk in children than the
weighted GRS. The effect of the unweighted GRS is also easier to interpret compared to the
weighted GRS. However, all analyses with the weighted GRS have been included as
supplemental materials.

Associations of Early Life NRAP Exposures and Childhood BMI Trajectory
Before moving onto the full GRSxE longitudinal mixed model, we wanted to verify
chapter 2’s findings in this subset of CHS participants. Of the 2,318 participants that contributed
to chapter 2’s analysis only 1,221 had genetic data available for first year of life exposures while
2,072 participants contributed to chapter 2’s in utero exposures only 1,099 were part of this GRS
analysis (Figure 3.1). Our GRS analysis sample differed from chapter 2’s sample by the
following characteristics: GRS sample only had NHW and Hispanic participants whereas the
previous chapter’s sample also had other racial/ethnic groups like Black, Asian/Pacific Islander,
and other races which accounted for 11% of the total sample; this GRS analysis had more life-
time history of asthma at 21% compared to the 15% in chapter 2’s sample. All other baseline
characteristics as well as NRAP exposures were similar across the two analytical samples.

78

Associations between early life NRAP exposures and childhood BMI growth from
chapter 2 results compared to this GRS sample are shown in Table 3.5-3.6. Briefly, in utero
freeway NOx had similar effects on BMI growth (0.05 vs 0.03 kg/m
2
increase per year) with
overlapping 95% confidence intervals (95% CI) and attenuated effects on BMI at age 10 years
with wider 95% CIs in the current GRS sample (Table 3.5). In utero non-freeway NOx, our
current GRS sample had attenuated effects on BMI growth and BMI at age 10 years with wider
95% CIs compared to previous chapter’s sample (Table 3.5).
First year of life freeway NOx in our GRS sample had attenuated effects on BMI growth
and attained BMI at age 10 years (Table 3.6). First year of life freeway NOx no longer was
statistically, significantly associated with childhood BMI trajectory (at level or growth), though
attenuated effects were in the same positive direction. First year of life non-freeway NOx
showed slightly attenuated effects on BMI growth and BMI at age 10 years with wider 95% CIs
compared to chapter 2’s analysis sample (Table 3.6). We further investigated potential reasons
why we see attenuated, non-significant associations in the first year of life freeway NOx models.
We compared effects by the two baseline characteristics that differed in this sample compared to
chapter 2’s sample: asthma status and race/ethnicity (Table 3.6a). Stratified analysis of
asthmatics vs non-asthmatics is shown in Table 3.6a. Though not statistically significant,
asthmatics had higher first year of life effects on both BMI growth and BMI at age 10 years.
However, this did not explain the attenuated effects of first year of life on BMI at age 10 years in
the GRS sample. We then compared effects by racial/ethnic group as the proportion of Hispanics
in the GRS sample was higher compared to the previous analysis sample (Table 3.6a). Stratified
analysis in the GRS sample by NHW vs Hispanic shows that the effect of first year of life
freeway NOx exposure on BMI growth in NHW were closer to chapter 2’s effect estimate

79

compared to Hispanic participants; moreover, effect of first year of life freeway NOx on attained
BMI at age 10 in NHW was similar to previous chapter’s reported effect whereas in Hispanics
the effect was in the negative direction. Aside from sample size differences, this may potentially
be the reason why we see decreased effects in our GRS sample compared to chapter 2’s reported
effects of first year of life on attained BMI at age 10 years as the large sample of Hispanics in
this current analysis may be driving the association towards the null.

Associations of Early Life NRAP exposures, GRSxE and Childhood BMI Trajectory
Associations of childhood BMI trajectory and GRSxE are shown in Tables 3.7-3.8. With
in utero NRAP exposures (total, freeway and non-freeway NOx), there were statistically
significant associations with BMI growth and attained BMI at age 10 years and the obesity
related, unweighted GRS across all models of NRAP exposures (all p<0.05, Table 3.7). For
example, given average in utero freeway NOx exposure, a 1 risk allele increase in the
unweighted GRS was associated with a 0.006 kg/m
2
faster BMI growth (p=0.02) resulting in a
0.07 kg/m
2
higher attained BMI at age 10 years (p<0.0001). Similar trends in associations were
seen with the weighted GRS with in utero NRAP exposures (Supplemental Table 3.5). There
were no significant associations with unweighted or weighted GRSxE with any in utero NRAP
exposures and BMI growth or attained BMI at age 10 years (Table 3.7, Supplemental Table
3.5).
Like in utero exposures, with first year of life NRAP exposures we saw statistically
significant associations with BMI growth and attained BMI at age 10 years and the obesity
related unweighted GRS across all models of NRAP exposures (all p<0.05, Table 3.8). Given
average first year of life freeway NOx exposure, a 1 risk allele increase in unweighted GRS was

80

statistically significantly associated with a 0.006 kg/m
2
faster increase in BMI per year (p=0.02)
resulting in a 0.07 kg/m
2
higher BMI at age 10 years (p<0.0001) (Table 3.8). Associations with
the weighted GRS and attained BMI at age 10 years across all models of first year of life NRAP
exposures showed similar trends as the unweighted GRS (Supplemental Table 3.6). There were
no significant associations with unweighted or weighted GRSxE with any first year of life NRAP
exposures and BMI growth or attained BMI at age 10 years (Table 3.8, Supplemental Table
3.6).
Associations of childhood BMI trajectory and GRSxE using quintiles of unweighted GRS
are shown in Supplemental Tables 3.7-3.8 and weighted GRS in Supplemental Tables 3.9-
3.10. Across all in utero NRAP exposures (total, freeway and non-freeway), the highest quintile
(quintile 5) of unweighted and weighted GRS showed the highest effect on BMI growth and
attained BMI when compared to the mean risk group (quintile 3), though we only see statistically
significant associations with the highest quintile in the in utero non-freeway NOx exposure
model (Supplemental Table 3.7). First year of life models with childhood BMI trajectory and
GRSxE using quintiles of unweighted GRS is shown in Supplemental Table 3.8. Similar trends
were seen in the first year of life NRAP exposures as the in utero NRAP models. Across both in
utero and first year of life NRAP exposure models, we did not see significant GRSxE quintile
associations with childhood BMI trajectory. Additionally, single SNP associations were explored
however there were no significant associations with childhood BMI trajectory.

3.5 Discussion

In this study, we further explored early life NRAP associations with childhood BMI
trajectory by looking at effects of a GWAS derived, obesity-related GRS and GRSxE interaction

81

in the model. Similar to previous studies (3, 20, 22) , we reported that an obesity-related GRS
was associated with childhood BMI growth in the CHS. Both unweighted and weighted GRS
were positively, significant associated with BMI growth through study follow up and attained
BMI at age 10 years (Table 3.4a-b). GRS in quintiles also reinforced the association showing
the highest quintile in both unweighted and weighted GRS was showed the largest effect on BMI
growth and BMI at age 10 years (across all quintiles) when compared to the mean risk group
(middle quintile).
Other studies have also shown associations of a GRS and childhood BMI growth. In
9,328 British and Australian children from respective birth cohorts, Avon Longitudinal Study of
Parents and Children (ALSPAC) and Western Australian Pregnancy Cohort (Raine) Study, a
GRS derived from 32 adult BMI associated SNPS was associated with BMI growth throughout
childhood (3). In another study with 5,906 Native Americans, 36 SNPs associated with
childhood BMI growth were defined and these SNPs were taken to create an allelic risk score
(ARS). The childhood-ARS was associated with rate of BMI change of 0.032 kg/m
2
per year per
risk allele resulting in a higher BMI at 19 years of 10.2 kg/m
2
when comparing individuals with
the highest number of risk alleles to those with the lowest number of risk alleles (22).
All our obesity related SNPs were derived from populations with European ancestry as
were most of the GRS and childhood BMI growth studies. Though over half of our analytic
sample was Hispanic (62%), we still saw significant associations with GRS and childhood BMI
trajectory. Perhaps an ethnic specific GRS may reflect more accurate associations with genetic
variants and childhood growth trajectory, however, other studies have used European ancestry
derived GRS and shown associations in non-European (white) populations. Recently, one study
used a GRS derived from SNPs discovered in European ancestry populations to evaluate

82

associations of GRS and childhood BMI in black, urban South African children (20). They
reported significant cross-sectional associations with BMI and GRS from 11-18 years. If more
GWAS studies were conducted in racially diverse samples, we may discover race-specific
variants that may better characterize the association between genetic variants and childhood
obesity risk.
This study has strengths where we had lifetime measures of NRAP exposure from in
utero through childhood, childhood BMI growth, as well as some genetic data in a subset of
children. Though we did not see any significant GRSxE interactions with in utero or first year of
life exposures, the GRS was positively, significantly associated with attained BMI at age 10
years across all, full GxE-early life NRAP exposure models. When compared to our findings in
chapter 2, first year of life NRAP exposures were no longer statistically significant therefore we
did not expect the GRSxE to hold in the full model. Unfortunately, we were limited in sample
size to those who had early life exposures as well as genetic data. We were also limited to the
data collected from CHS, so we did not have any robust measures of diet or physical activity in
this cohort of CHS.
We have conducted an association analysis that explored the gene-environment
interaction and its effects on childhood BMI trajectory in a well establish cohort of children with
lifetime air pollution exposures. Our study suggests that known genetic variants that are related
to obesity have observable effects on childhood BMI growth and potentially over the life course.
Genetic differences amongst individuals may prime some and not others to the deleterious
effects of an obesogenic environment; therefore, further investigation of the effects of gene-
environment interactions with early life exposures in a larger sample of children is warranted.

83

3.6 References

1. Speliotes EK, Willer CJ, Berndt SI, et al. Association analyses of 249,796 individuals
reveal 18 new loci associated with body mass index. Nat Genet 2010;42(11):937-48.
2. Frayling TM, Timpson NJ, Weedon MN, et al. A common variant in the FTO gene is
associated with body mass index and predisposes to childhood and adult obesity. Science
2007;316(5826):889-94.
3. Warrington NM, Howe LD, Wu YY, et al. Association of a body mass index genetic risk
score with growth throughout childhood and adolescence. PLoS One 2013;8(11):e79547.
4. Jerrett M, McConnell R, Chang CC, et al. Automobile traffic around the home and
attained body mass index: a longitudinal cohort study of children aged 10-18 years. Prev
Med 2010;50 Suppl 1:S50-8.
5. Jerrett M, McConnell R, Wolch J, et al. Traffic-related air pollution and obesity
formation in children: a longitudinal, multilevel analysis. Environ Health 2014;13:49.
6. Kim JS, Alderete TL, Chen Z, et al. Longitudinal associations of in utero and early life
near-roadway air pollution with trajectories of childhood body mass index. Environ
Health 2018;17(1):64.
7. Mao G, Nachman RM, Sun Q, et al. Individual and Joint Effects of Early-Life Ambient
PM2.5 Exposure and Maternal Pre-Pregnancy Obesity on Childhood Overweight or
Obesity. Environ Health Perspect 2016.
8. Rundle A, Hoepner L, Hassoun A, et al. Association of childhood obesity with maternal
exposure to ambient air polycyclic aromatic hydrocarbons during pregnancy. Am J
Epidemiol 2012;175(11):1163-72.
9. Maes HH, Neale MC, Eaves LJ. Genetic and environmental factors in relative body
weight and human adiposity. Behav Genet 1997;27(4):325-51.
10. Silventoinen K, Rokholm B, Kaprio J, et al. The genetic and environmental influences on
childhood obesity: a systematic review of twin and adoption studies. Int J Obes (Lond)
2010;34(1):29-40.
11. Turula M, Kaprio J, Rissanen A, et al. Body weight in the Finnish Twin Cohort. Diabetes
Res Clin Pract 1990;10 Suppl 1:S33-6.
12. Wardle J, Carnell S, Haworth CM, et al. Evidence for a strong genetic influence on
childhood adiposity despite the force of the obesogenic environment. Am J Clin Nutr
2008;87(2):398-404.
13. Qi Q, Chu AY, Kang JH, et al. Sugar-sweetened beverages and genetic risk of obesity. N
Engl J Med 2012;367(15):1387-96.
14. Brunkwall L, Chen Y, Hindy G, et al. Sugar-sweetened beverage consumption and
genetic predisposition to obesity in 2 Swedish cohorts. Am J Clin Nutr 2016;104(3):809-
15.
15. Qi Q, Chu AY, Kang JH, et al. Fried food consumption, genetic risk, and body mass
index: gene-diet interaction analysis in three US cohort studies. BMJ 2014;348:g1610.
16. Heianza Y, Qi L. Gene-Diet Interaction and Precision Nutrition in Obesity. Int J Mol Sci
2017;18(4).
17. Rask-Andersen M, Karlsson T, Ek WE, et al. Gene-environment interaction study for
BMI reveals interactions between genetic factors and physical activity, alcohol
consumption and socioeconomic status. PLoS Genet 2017;13(9):e1006977.

84

18. Bradfield JP, Taal HR, Timpson NJ, et al. A genome-wide association meta-analysis
identifies new childhood obesity loci. Nat Genet 2012;44(5):526-31.
19. Zhao J, Bradfield JP, Li M, et al. The role of obesity-associated loci identified in
genome-wide association studies in the determination of pediatric BMI. Obesity (Silver
Spring) 2009;17(12):2254-7.
20. Munthali RJ, Sahibdeen V, Kagura J, et al. Genetic risk score for adult body mass index
associations with childhood and adolescent weight gain in an African population. Genes
Nutr 2018;13:24.
21. Monnereau C, Vogelezang S, Kruithof CJ, et al. Associations of genetic risk scores based
on adult adiposity pathways with childhood growth and adiposity measures. BMC Genet
2016;17(1):120.
22. Hohenadel MG, Baier LJ, Piaggi P, et al. The impact of genetic variants on BMI increase
during childhood versus adulthood. Int J Obes (Lond) 2016;40(8):1301-9.
23. Welter D MJ, Morales J, Burdett T, Hall P, Junkins H, Klemm A, Flicek P, Manolio T,
Hindorff L, and Parkinson H. The NHGRI GWAS Catalog, a curated resource of SNP-
trait associations. Nucleic Acids Research 2014;Vol. 42 (Database issue): D1001-D1006.
24. Berndt SI, Gustafsson S, Magi R, et al. Genome-wide meta-analysis identifies 11 new
loci for anthropometric traits and provides insights into genetic architecture. Nat Genet
2013;45(5):501-12.
25. Wang K, Li WD, Zhang CK, et al. A genome-wide association study on obesity and
obesity-related traits. PLoS One 2011;6(4):e18939.
26. Cotsapas C, Speliotes EK, Hatoum IJ, et al. Common body mass index-associated
variants confer risk of extreme obesity. Hum Mol Genet 2009;18(18):3502-7.
27. Jiao H, Arner P, Hoffstedt J, et al. Genome wide association study identifies KCNMA1
contributing to human obesity. BMC Med Genomics 2011;4:51.
28. Meyre D, Delplanque J, Chevre JC, et al. Genome-wide association study for early-onset
and morbid adult obesity identifies three new risk loci in European populations. Nat
Genet 2009;41(2):157-9.
29. Paternoster L, Evans DM, Nohr EA, et al. Genome-wide population-based association
study of extremely overweight young adults--the GOYA study. PLoS One
2011;6(9):e24303.
30. Wheeler E, Huang N, Bochukova EG, et al. Genome-wide SNP and CNV analysis
identifies common and low-frequency variants associated with severe early-onset obesity.
Nat Genet 2013;45(5):513-7.
31. Scherag A, Dina C, Hinney A, et al. Two new Loci for body-weight regulation identified
in a joint analysis of genome-wide association studies for early-onset extreme obesity in
French and german study groups. PLoS Genet 2010;6(4):e1000916.
32. Benson P. CALINE4 – A Dispersion Model for Predicting Air Pollutant Concentrations
Near Roadways. 1989;Report No. FHWA/CA/TL-84/15.(Prepared by the California
Department of Transportation, Sacramento, CA.).
33. Kuczmarski RJ, Ogden CL, Guo SS, et al. 2000 CDC Growth Charts for the United
States: methods and development. Vital Health Stat 11 2002(246):1-190.

85

3.7 Tables
Table 3.1 List of Studies and SNPs Used to Create the Obesity-Related Genetic Risk Score (GRS)
Study
# SNPs
in GRS
SNPs Mapped Genes
Discovery
Population
Obesity Outcome
Berndt et al.
Nat Genet. 2013;
45(5):501-12.
49
rs10182181, rs2531995, rs2030323, rs2112347,
rs2307111, rs13078807, rs1516725, rs9816226,
rs887912, rs6731302, rs1421085, rs7185735,
rs8043757, rs10423928, rs1800437,
rs17024258, rs13130484, rs10938397,
rs11639988, rs12446632, rs12446554,
rs4735692, rs7989336, rs1412239, rs10968576,
rs7138803, rs10875976, rs8028313,
rs11152213, rs538656, rs10871777,
rs13041126, rs7531118, rs2568958, rs7141420,
rs2370983, rs9568867, rs7184597, rs11042023,
rs7503807, rs633715, rs7498665, rs2207139,
rs2206277, rs10189761, rs6711012, rs1514177,
rs1514174, rs17381664
ADCY3, ADCY9, BDNF, POC5,
CADM2, ETV5, FANCL, FTO,
GIPR, GNAT2, PRDX4P1-
PRKRIRP9, GPRC5B, HNF4G,
HS6ST3, LINGO2, BCDIN3D,
MAP2K5, MC4R, MRPS33P4,
NEGR1, NRXN3, ZNF646P1 -
LINC00558, RABEP2, TRIM66,
RPTOR, SEC16B, SH2B1,
TRNAI25, TFAP2B, TMEM18,
TNNI3K, ZZZ3

Adults (18+ years)
Overweight (BMI ≥25 kg/m
2
)
Obesity Class 1 (BMI ≥30)
Obesity Class 2 (BMI ≥35)
Obesity Class 3 (BMI ≥40)
Paternoster et al.
PLoS One.2011;
6(9):e24303.
2 rs7132908, rs9936385 FAIM2, FTO Adults (18+ years) BMI≥31.0 kg/m
2

Wang et al.
PLoS One. 2011;
6(4):e18939.
1 rs3751812 FTO
Adults (18+ years)
mostly women
BMI≥35 kg/m
2

Cotsapas et al.
Hum Mol Genet. 2009;
18(18):3502-7.
1 rs9941349 FTO Adults (18-75 years) BMI>33.3 kg/m
2

Jiao et al.
BMC Med Genomics. 2011;
4:51.
2 rs988712, rs2116830 BDNF, KCNMA1 Adults and children
BMI>40 kg/m
2
(adults),
BMI Z-score ≥3 SD (children)
Meyre et al.
Nat Genet. 2009;
41(2):157-9.
2 rs1424233, rs17782313 MAF, MC4R Adults and children
BMI ≥97th %tile (children)
BMI≥40 kg/m
2
(adults)
Wheeler et al.
Nat Genet. 2013;
45(5):513-7.
8
rs11208659, rs476828, rs1993709, rs3101336,
rs564343, rs1957894, rs11109072, rs12463617
LEPR, MC4R, RPL31P12 -
KRT8P21, GDI2P2 - RPL31P12,
PACS1, PRKCH, RMST,
TMEM18, FTO, MC4R, MSRA,
TNKS
Children (<10 years) Severe obesity (BMI≥3 SD)
Bradfield et al.
Nat Genet. 2012;
44(5):526-31.
2 rs9299, rs9568856
HOXB3, HOXB5,
HOXB-AS3,
ZNF646P1 - LINC00558

Children (<18 years) BMI ≥95th percentile
Scherag et al.
PLoS Genet. 2010;
6(4):e1000916.
3 rs1558902, rs17700144, rs17150703 FTO, MC4R, TNKS, MSRA Children (<18 years) BMI ≥97th percentile

86

Table 3.2 Baseline Characteristics of Children Enrolled in the Longitudinal Children's Health
Study
a

Characteristic Mean (SD) or n (%)
b

Age at Study Entry (years) [mean (SD)] 6.5 (0.6)
Age at End of Study (years) [mean (SD)] 10.0 (0.9)
Sex

Female 566 (46.4)
Male 655 (53.6)
Race/Ethnicity

White 466 (38.2)
Hispanic 755 (61.8)
Overweight/Obesity Status
c

Normal 883 (72.3)
Overweight 179 (14.7)
Obese 159 (13.0)
Parental Education

Less than high school 222 (18.7)
High school 218 (18.4)
Above high school 746 (62.9)
Spanish Questionnaire
d

No 944 (77.3)
Yes 277 (22.7)
Self-Reported Premature Birth

No 1065 (89.3)
Yes 127 (10.7)
Maternal Smoking During Pregnancy

No 1108 (93.2)
Yes 81 (6.8)
Residential Second-Hand Smoke
e
No 1128 (94.2)
Yes, child is home 48 (4.0)
Yes, child is not home 22 (1.8)
Life-Time History of Asthma

No 949 (79.5)
Yes 245 (20.5)
Organized Team Sport
f
No 511 (52.3)
Yes 467 (47.8)

Genetic Risk Score (GRS)
g
[mean (SD)] 65.1 (7.6)
Weighted GRS
h
[mean (SD)] 9.7 (1.6)

87

a
This analysis includes a subset of the Children’s Health Study participants who had available GWAS data, NRAP
exposure data for in utero or first year of life periods, at least two measures of BMI during study follow up period,
completed a baseline questionnaire, and had moved homes at least once before study enrollment.
b
First observation of participant with genetic risk score and NRAP exposures (n=1221); variable denominators may
differ due to missing values.

c
Overweight/ Obesity status: Normal is < 85
th
percentile of age-, sex-specific BMI using 2000 CDC growth chart,
overweight is 85-95th percentile of age-, sex-specific BMI, obese is ≥ 95th percentile of age-, sex- specific BMI.
d
Spanish Questionnaire is if parent filled out baseline questionnaire in Spanish and serves as a surrogate measure for
recent immigration.
e
Residential second-hand smoke is if anyone living in the child’s home smokes daily inside the home.
f
Organized team sport is if the child played outdoors in any organized team sport at least twice a week during the
past year.
g
Genetic Risk Score (GRS)- created from 67 GWAS derived, obesity-related SNPs where allele frequencies were
summed.
h
Weighted GRS- calculated by incorporating the natural-logged published odds ratios to each allele pair score.

88

Table 3.3 Residential NRAP exposures from freeway and non-freeway sources for in utero, first
year of life, and childhood periods in children in the CHS.
Exposure Period n Mean SD IQR
Freeway NOx (ppb)

In utero 1099 17.4 21.3 5.05-22.1
First year of life 1221 17.2 21.0 4.8-22.9
Childhood 1221 16.3 20.6 5.5-19.4
Non-Freeway NOx (ppb)
In utero 1099 10.4 7.4 5.3-13.7
First year of life 1221 9.5 7.0 4.8-12.2
Childhood 1221 6.2 4.3 3.4-7.4
Total NOx
a
(ppb)

In utero 1099 27.8 24.0 12.7-35.7
First year of life 1221 26.7 23.5 11.9-35.3
Childhood 1221 22.5 22.5 9.8-26.9
NRAP=near-roadway air pollution; SD=standard deviation; IQR=interquartile range; NOx=nitrogren oxides;
ppb=parts per billion
a
Total NOx= freeway + non-freeway NOx

89

Table 3.4a Associations
a
of Obesity-Related GRS
b
and Childhood BMI Trajectory

BMI Growth

BMI Age 10 Years
GRS (Unweighted) Estimate

SE p-value Estimate SE p-value
Continuous GRS
GRS 0.0045 0.0021 0.02

0.058 0.013 <0.001
Quintiles of GRS
Quintile 5
c
0.08 0.05 0.12 0.76 0.30 0.01
Quintile 4 0.016 0.05 0.75 0.22 0.30 0.46
Quintile 3 ref ref
Quintile 2 0.019 0.05 0.71 -0.23 0.30 0.46
Quintile 1 -0.03 0.05 0.53 -0.48 0.30 0.12
GRS (Weighted) Estimate SE p-value Estimate SE p-value
Continuous GRS
GRS 0.02 0.01 0.04 0.26 0.058 <0.0001
Quintiles of GRS
Quintile 5
c
0.07 0.05 0.17 0.93 0.30 0.002
Quintile 4 -0.037 0.05 0.46 0.08 0.30 0.8
Quintile 3 ref ref
Quintile 2 -0.002 0.05 0.97 0.07 0.30 0.82
Quintile 1 -0.02 0.05 0.64 -0.38 0.31 0.21
GRS= genetic risk score; BMI=body mass index; SE=standard error.
a
Models adjusting for age, sex, race/ethnicity, parental education and Spanish baseline questionnaire at level and
growth.
b
All CHS cohort E movers with a GRS (n=1,768).
c
Highest quintile with 20% of participants with the highest GRS score.

90

Table 3.4b Associations
a
of Obesity-Related GRS
b
and Childhood BMI Trajectory in
Participants with Early Life NRAP Exposure

BMI Growth

BMI Age 10 Years
GRS (Unweighted) Estimate

SE p-value

Estimate SE p-value
Continuous GRS
GRS 0.006 0.003 0.02 0.07 0.02 <.0001
Quintiles of GRS
Quintile 5
c
0.10 0.06 0.09 0.79 0.37 0.03
Quintile 4 0.06 0.06 0.36 0.26 0.37 0.49
Quintile 3 ref ref
Quintile 2 -0.007 0.06 0.91 -0.45 0.37 0.23
Quintile 1 -0.01 0.06 0.84 -0.51 0.38 0.17
GRS (Weighted) Estimate SE p-value Estimate SE p-value
Continuous GRS
GRS 0.022 0.012 0.07 0.27 0.073 0.0002
Quintiles of GRS
Quintile 5
c
0.09 0.06 0.15 1.03 0.37 0.006
Quintile 4 -0.002 0.061 0.97 0.06 0.37 0.87
Quintile 3 ref ref
Quintile 2 0.004 0.061 0.95 0.06 0.37 0.88
Quintile 1 -0.02 0.06 0.72 -0.48 0.37 0.20
GRS= genetic risk score; BMI=body mass index; SE=standard error.
a
Models adjusting for age, sex, race/ethnicity, parental education and Spanish baseline questionnaire at level and
growth.
b
All CHS cohort E participants with a GRS and in utero or first year of life NRAP exposure (n=1,221)
c
Highest quintile with 20% of participants with the highest GRS score.

91

Table 3.5 Comparison of Associations
a
of In Utero Freeway and Non-Freeway NOx on Childhood BMI Trajectory with Chapter 2’s
Sample versus GRS Sample

Freeway NOx

BMI Growth BMI at Age 10 Years

n Estimate SE 95% CI p-value Estimate SE 95% CI p-value
Chapter 2 Sample:
In utero NOx 2072 0.05 0.03 -0.02 0.11 0.18 0.13 0.20 -0.26 0.53 0.50
Childhood NOx

-0.02 0.04 -0.10 0.05 0.54 0.05 0.22 -0.38 0.49 0.81
GRS sample:

In utero NOx 1099 0.03 0.04 -0.05 0.12 0.47 -0.02 0.26 -0.54 0.50 0.94
Childhood NOx

-0.01 0.05 -0.11 0.09 0.82 0.08 0.31 -0.53 0.69 0.79

Non-Freeway NOx

BMI Growth BMI at Age 10 Years

n Estimate SE 95% CI p-value Estimate SE 95% CI p-value
Chapter 2 Sample:
In utero NOx 2072 0.02 0.04 -0.05 0.10 0.52 0.14 0.23 -0.31 0.59 0.54
Childhood NOx

0.08 0.04 -0.007 0.17 0.07 0.55 0.24 0.08 1.03 0.02
GRS sample:

In utero NOx 1099 -0.02 0.05 -0.12 0.08 0.67 0.04 0.30 -0.56 0.63 0.91
Childhood NOx

0.06 0.05 -0.05 0.17 0.26 0.36 0.32 -0.26 0.98 0.25
a
Models adjusting for age, sex, race/ethnicity, parental education and Spanish baseline questionnaire at level and growth.

92

Table 3.6. Comparison of Associations
a
of First Year of Life Freeway and Non-Freeway NOx on Childhood BMI Trajectory with
Chapter 2’s Sample versus GRS Sample

Freeway NOx

BMI Growth BMI at Age 10 Years
n Estimate SE 95% CI p-value Estimate SE 95% CI p-value
Chapter 2 Sample:
First year of life NOx 2318 0.1 0.04 0.03 0.17 0.006 0.45 0.22 0.02 0.88 0.04
Childhood NOx

-0.06 0.04 -0.14 0.02 0.12 -0.14 0.22 -0.58 0.30 0.53
GRS sample:

First year of life NOx 1221 0.04 0.05 -0.06 0.13 0.46 0.09 0.30 -0.50 0.68 0.77
Childhood NOx

-0.02 0.05 -0.12 0.08 0.69 0.11 0.30 -0.48 0.71 0.71

Non-Freeway NOx

BMI Growth BMI at Age 10 Years
n Estimate SE 95% CI p-value Estimate SE 95% CI p-value
Chapter 2 Sample:
First year of life NOx 2318 -0.02 0.04 -0.10 0.06 0.62 -0.07 0.24 -0.54 0.39 0.76
Childhood NOx

0.10 0.05 0.01 0.19 0.03 0.61 0.25 0.12 1.10 0.01
GRS sample:

First year of life NOx 1221 -0.06 0.05 -0.15 0.04 0.26 -0.18 0.31 -0.78 0.43 0.57
Childhood NOx 0.07 0.06 -0.04 0.17 0.23 0.46 0.32 -0.16 1.09 0.15
a
Models adjusting for age, sex, race/ethnicity, parental education and Spanish baseline questionnaire at level and growth.

93

Table 3.6a Comparison of Associations
a
of First Year of Life Freeway NOx on Childhood BMI Trajectory with Chapter 2’s Sample
versus GRS Sample by Asthma Status and by Race/Ethnicity

BMI Growth
a
BMI at Age 10 Years
a

Estimate SE 95% CI p-value Estimate SE 95% CI p-value
Chapter 2 Sample:
First year of life NOx 0.1 0.04 0.03 0.17 0.006 0.45 0.22 0.02 0.88 0.04
Childhood NOx -0.06 0.04 -0.14 0.02 0.12 -0.14 0.22 -0.58 0.30 0.53
GRS sample:

First year of life NOx 0.04 0.05 -0.06 0.13 0.46 0.09 0.30 -0.50 0.68 0.77
Childhood NOx -0.02 0.05 -0.12 0.08 0.69 0.11 0.30 -0.48 0.71 0.71
By Asthma Status Estimate SE 95% CI p-value Estimate SE 95% CI p-value
Non-asthmatics
First year of life NOx 0.023 0.054 -0.08 0.13 0.67 -0.06 0.32 -0.69 0.57 0.86
Childhood NOx -0.024 0.055 -0.13 0.08 0.66 0.16 0.31 -0.46 0.77 0.62
Asthmatics
First year of life NOx 0.082 0.11 -0.13 0.30 0.45 0.57 0.58 -0.57 1.70 0.33
Childhood NOx 0.035 0.14 -0.23 0.30 0.80 0.37 0.72 -1.05 1.78 0.61
By Race/Ethnicity Estimate SE 95% CI p-value Estimate SE 95% CI p-value
Non-Hispanic White
First year of life NOx 0.06 0.09 -0.12 0.25 0.51 0.37 0.53 -0.66 1.42 0.48
Childhood NOx 0.07 0.13 -0.19 0.33 0.61 0.46 0.76 -1.01 1.95 0.54
Hispanic
First year of life NOx 0.03 0.06 -0.09 0.14 0.66 -0.03 0.37 -0.75 0.71 0.95
Childhood NOx -0.02 0.06 -0.14 0.09 0.64 0.11 0.36 -0.59 0.81 0.75
Models adjusting for age, race/ethnicity, parental education and Spanish baseline questionnaire at level and growth (n=1221).
a
BMI growth (kg/m
2
) over study follow up and difference in attained BMI at age 10 years scaled to 2 standard deviations of of first year of life freeway NOx
with 42.0 ppb. Childhood NOx was also scaled to 2 SDs where childhood freeway NOx was 41.2 ppb.

94

Table 3.7 Associations of In Utero NRAP Exposures, Unweighted GRSxE and Childhood BMI Trajectory in CHS Children

Models adjusting for age, race/ethnicity, parental education and Spanish baseline questionnaire at level and growth (n=1099).
a
BMI growth (kg/m
2
) over study follow up and difference in attained BMI at age 10 years scaled to 2 standard deviations of mean centered in utero total NOx
with 47.9 ppb, freeway NOx with 42.6 ppb, and non-freeway NOx with 14.9 ppb. Childhood NOx was also scaled to 2 SDs where childhood total NOx was
45.1ppb, childhood freeway NOx was 41.2 ppb, and childhood non-freeway NOx was 8.7 ppb.
b
In utero NOx centered to study mean value where total NOx mean=27.8, freeway NOx mean=17.4, and non-freeway NOx mean=10.4.
c
GRS centered to study mean where the unweighted GRS mean= 65.1.

BMI Growth
a

BMI Age 10 Years
a

Estimate SE p-value

Estimate SE p-value
Total NOx
In utero NOx (“E”)
b
0.03 0.05 0.52

0.005 0.3 0.99
GRS
c
0.006 0.003 0.02

0.07 0.02 <.0001
GRSxE -0.0004 0.006 0.95

0.02 0.04 0.67
Childhood NOx -0.0006 0.052 0.99

0.15 0.31 0.63
Freeway NOx
In utero NOx (“E”)
b
0.03 0.04 0.50 -0.06 0.3 0.82
GRS
c
0.006 0.003 0.02 0.07 0.02 <.0001
GRSxE -0.0005 0.006 0.94 0.02 0.04 0.56
Childhood NOx -0.01 0.05 0.85 0.095 0.31 0.76
Non-Freeway NOx
In utero NOx (“E”)
b
-0.02 0.05 0.74 0.09 0.30 0.77
GRS
c
0.006 0.003 0.02 0.07 0.02 <.0001
GRSxE -0.0005 0.006 0.93 -0.005 0.03 0.87
Childhood NOx 0.056 0.054 0.30 0.29 0.31 0.35

95

Table 3.8 Associations First Year of Life NRAP Exposures, Unweighted GRSxE and Childhood BMI Trajectory in CHS Children

Models adjusting for age, race/ethnicity, parental education and Spanish baseline questionnaire at level and growth (n=1221).
a
BMI growth (kg/m
2
) over study follow up and difference in attained BMI at age 10 years scaled to 2 standard deviations of mean centered first year of life total
NOx with 47.0 ppb, freeway NOx with 42.0 ppb, and non-freeway NOx with 14.1 ppb. Childhood NOx was also scaled to 2 SDs where childhood total NOx was
45.1ppb, childhood freeway NOx was 41.2 ppb, and childhood non-freeway NOx was 8.7 ppb.
b
First year of life NOx centered to study mean value where total NOx mean=26.7, freeway NOx mean=17.2, and non-freeway NOx mean=9.5.
c
GRS centered to study mean where the unweighted GRS mean= 65.1.

BMI Growth
a

BMI Age 10 Years
a

Estimate SE p-value

Estimate SE p-value
Total NOx
First Year of Life NOx (“E”)
b
0.02 0.05 0.69 0.05 0.3 0.87
GRS
c
0.006 0.003 0.02 0.07 0.02 <.0001
GRSxE -0.002 0.005 0.72 0.01 0.03 0.75
Childhood NOx -0.008 0.053 0.87 0.14 0.32 0.66
Freeway NOx
First Year of Life NOx (“E”)
b
0.03 0.05 0.47 0.06 0.3 0.85
GRS
c
0.006 0.003 0.02 0.07 0.02 <.0001
GRSxE -0.002 0.005 0.67 0.01 0.03 0.74
Childhood NOx -0.021 0.051 0.68 0.08 0.31 0.79
Non-Freeway NOx
First Year of Life NOx (“E”)
b
-0.05 0.05 0.29 -0.12 0.31 0.70
GRS
c
0.006 0.003 0.02 0.07 0.02 <.0001
GRSxE 0.00004 0.005 0.99 0.005 0.033 0.89
Childhood NOx 0.06 0.06 0.27 0.38 0.32 0.23

96

3.8 Figures

Figure 3.1 Flow Chart of Children Enrolled in the Children’s Health Study from 2002-2003
Included and Excluded from the Current Analysis

Note: BMI, body mass index; NRAP, near-roadway air pollution; GRS, genetic risk score.

In Utero NRAP
Participants with in utero
NRAP exposure measures
N=2,072
Children’s Health Study
(Cohort E)
Children with non-missing BMI,
age, sex, baseline questionnaire
N=5,337
Longitudinal BMI
Children with ≥ 2 BMI measures
N=4,400
Non-movers
Children who moved
residences < 500 meters
before study enrollment
N=976
Movers
Children who moved
residences ≥ 500 meters
before study enrollment
N=3,424
First Year of Life NRAP
Participants with first year of life
NRAP exposure measures
N=2,318
GRS
Participants with genetic
data for GRS
N=1,099
GRS
Participants with genetic
data for GRS
N=1,221

Project 1
Project 2

97

3.9 Supplemental Tables

Supplemental Table 3.1 Pearson correlation coefficients between in utero/first year of life and
childhood NRAP exposure in non-movers

and movers

in children in the Children’s Health Study.
NRAP Exposure

Non-Movers
a
Movers
b
Freeway NOx In Utero
First Year of
Life

In Utero First Year of Life
Childhood 0.96* 0.97* 0.32* 0.60*
Non-freeway NOx
Childhood 0.97* 0.98* 0.58* 0.63*
Total NOx

Childhood 0.96* 0.97* 0.40* 0.63*
a
Non-movers were children who did not have a change in address or moved less than 500 meters between in utero
period and study entry (n=458).
b
Movers were children who had a change in address between in utero period and CHS study entry that resulted in a
move greater than or equal to 500 meters.
*p<.0001

98

Supplemental Table 3.2 Distribution of Unweighted and Weighted GRS Quintiles in
Participants with Early Life NRAP Exposures

GRS= genetic risk score; SD=standard deviation.
All CHS cohort E movers with a GRS (n=1,768).
a
Highest quintile with 20% of participants with the highest GRS score/greatest number of risk alleles.
b
Lowest quintile with 20% of participants with the lowest GRS score/least number of risk alleles.

GRS (unweighted) n Mean SD Range
Quintile 5
a
244 76.0 3.3 72.0-88.1
Quintile 4 247 69.3 1.4 67.02-71.98
Quintile 3 245 65.0 1.3 63.0-67.0
Quintile 2 243 60.8 1.3 58.11-62.99
Quintile 1
b
242 54.4 2.9 44.2-58.10
GRS (weighted)

Mean SD Range
Quintile 5
a
232 12.2 0.7 11.19-14.31
Quintile 4 249 10.6 0.3 10.09-11.18
Quintile 3 258 9.6 0.3 9.13-10.08
Quintile 2 246 8.7 0.2 8.26-9.12
Quintile 1
b
236 7.5 0.5 5.62-8.25

99

Supplemental Table 3.3 Associations
a
of Obesity-Related GRS and Childhood BMI by Sex
b

Males

Females

BMI Growth

BMI Age 10

BMI Growth

BMI Age 10

Estimate p-value

Estimate p-value

Estimate p-value

Estimate p-value
GRS (unweighted) 0.005 0.12

0.04 0.03

0.005 0.11

0.08 <.0001
GRS (weighted) 0.022 0.11 0.18 0.02 0.018 0.19 0.34 <.0001
Restricted to subjects with early life NRAP exposures
GRS (unweighted) 0.006 0.10 0.05 0.03 0.006 0.07 0.09 <.0001
GRS (weighted) 0.026 0.13 0.21 0.04 0.018 0.28 0.35 0.001
a
Models adjusting for age, race/ethnicity, parental education and Spanish baseline questionnaire at level and growth.
b
Unrestricted: P interaction Sex-GRS: p=0.07 (level), p=0.97 (growth); Restricted: P interaction Sex-GRS: p=0.14 (level), p=0.97 (growth).
Unrestricted (n=1768), restricted (n=1221).

100

Supplemental Table 3.4 Associations
a
of Obesity-Related GRS and Childhood BMI by Race/Ethnicity
b

Non-Hispanic White

Hispanic

BMI Growth

BMI Age 10

BMI Growth

BMI Age 10
GRS Estimate p-value

Estimate p-value

Estimate p-value

Estimate p-value
GRS (unweighted) 0.005 0.05

0.04 0.01

0.004 0.20

0.07 <.0001
GRS (weighted) 0.021 0.1

0.17 0.01

0.018 0.19

0.33 0.0002
Restricted to subjects with early life NRAP exposures
GRS (unweighted) 0.009 0.008 0.06 0.0013 0.004 0.23 0.07 0.001
GRS (weighted) 0.037 0.01 0.26 0.003 0.012 0.50 0.28 0.009
a
Models adjusting for age, race/ethnicity, parental education and Spanish baseline questionnaire at level and growth.
b
Unrestricted P interaction Race-GRS: p=0.2 (level), p=0.73 (growth); Restricted: P interaction Race-GRS: p=0.71 (level), p=0.43 (growth).
Unrestricted (n=1768), restricted (n=1221).

101

Supplemental Table 3.5 Associations of In Utero NRAP Exposures, Weighted GRSxE and Childhood BMI Trajectory in CHS
Children

Models adjusting for age, race/ethnicity, parental education and Spanish baseline questionnaire at level and growth (n=1099).
a
BMI growth (kg/m
2
) over study follow up and difference in attained BMI at age 10 years scaled to 2 standard deviations of in utero total NOx with 47.9
ppb, freeway NOx with 42.6 ppb, and non-freeway NOx with 14.9 ppb. Childhood NOx was also scaled to 2 SDs where childhood total NOx was 45.1ppb,
childhood freeway NOx was 41.2 ppb, and childhood non-freeway NOx was 8.7 ppb.
b
In utero NOx centered to study mean value where total NOx mean=27.8 , freeway NOx mean=17.4 , and non-freeway NOx mean=10.4 .
c
GRS centered to study mean where the weighted GRS mean= 9.7.

BMI Growth
a

BMI Age 10
a

Estimate SE p-value

Estimate SE p-value
Total NOx
In utero NOx (“E”)
b
0.029 0.04 0.52 0.039 0.27 0.89
GRS
c
0.023 0.012 0.07 0.28 0.08 0.0003
GRSxE -0.023 0.027 0.39 0.06 0.16 0.71
Childhood NOx 0.0001 0.052 1.00 0.14 0.31 0.66
Freeway NOx
In utero NOx (“E”)
b
0.03 0.04 0.49 -0.016 0.26 0.95
GRS
c
0.023 0.012 0.07 0.28 0.08 0.0002
GRSxE -0.022 0.027 0.42 0.075 0.16 0.65
Childhood NOx -0.010 0.051 0.84 0.071 0.31 0.82
Non-Freeway NOx
In utero NOx (“E”)
b
-0.02 0.05 0.69 0.07 0.30 0.81
GRS
c
0.023 0.012 0.06 0.28 0.08 0.0003
GRSxE -0.012 0.025 0.64 -0.013 0.16 0.93
Childhood NOx 0.062 0.054 0.25 0.34 0.32 0.28

102

Supplemental Table 3.6 Associations of First Year of Life NRAP Exposures, Weighted GRSxE and Childhood BMI Trajectory in
CHS Children

Models adjusting for age, race/ethnicity, parental education and Spanish baseline questionnaire at level and growth (n=1221).
a
BMI growth (kg/m
2
) over study follow up and difference in attained BMI at age 10 years scaled to 2 standard deviations of first year of life total NOx with 47.0
ppb, freeway NOx with 42.0 ppb, and non-freeway NOx with 14.1ppb. Childhood NOx was also scaled to 2 SDs where childhood total NOx was 45.1ppb,
childhood freeway NOx was 41.2 ppb, and childhood non-freeway NOx was 8.7 ppb.
b
First year of life NOx centered to study mean value where total NOx mean=26.7, freeway NOx mean=17.2, and non-freeway NOx mean=9.5.
c
GRS centered to study mean where the weighted GRS mean= 9.7.

BMI Growth
a

BMI Age 10
a

Estimate SE p-value

Estimate SE p-value
Total NOx
First Year of Life NOx (“E”)
b
0.02 0.05 0.68 0.07 0.31 0.81
GRS
c
0.021 0.012 0.07 0.28 0.07 0.0002
GRSxE -0.010 0.026 0.70 0.13 0.15 0.39
Childhood NOx -0.0093 0.053 0.86 0.10 0.32 0.75
Freeway NOx
First Year of Life NOx (“E”)
b
0.04 0.05 0.47 0.08 0.30 0.79
GRS
c
0.021 0.012 0.07 0.28 0.07 0.0002
GRSxE -0.012 0.026 0.64 0.13 0.15 0.41
Childhood NOx -0.022 0.051 0.66 0.03 0.31 0.91
Non-Freeway NOx
First Year of Life NOx (“E”)
b
-0.05 0.05 0.29 -0.11 0.31 0.72
GRS
c
0.022 0.012 0.07 0.27 0.07 0.0002
GRSxE 0.002 0.026 0.94 0.081 0.16 0.60
Childhood NOx 0.065 0.055 0.24 0.42 0.32 0.19

103

Supplemental Table 3.7 Associations of In Utero NRAP, GRSxE (Unweighted-Quintiles), and Childhood BMI
Total Nox

BMI Growth
a
BMI Age 10 Years
a

Estimate SE p-value Estimate SE p-value
In utero Total NOx -0.091 0.081 0.26 -0.072 0.477 0.88
Quintile 5 0.106 0.098
0.18
0.972 0.591
0.14

Quintile 4 -0.087 0.094 0.146 0.571
Quintile 3 ref ref
Quintile 2 -0.106 0.096 -0.471 0.579
Quintile 1 -0.086 0.103 -0.413 0.622
Q5xE 0.007 0.127
0.13
-0.222 0.761
0.93

Q4xE 0.267 0.118 0.478 0.708
Q3xE ref ref
Q2xE 0.191 0.122 0.095 0.729
Q1xE 0.145 0.139 -0.028 0.840
Childhood Total NOx -0.010 0.052 0.85 0.131 0.316 0.68
Freeway NOx
BMI Growth
a
BMI Age 10 Years
a

Estimate SE p-value Estimate SE p-value
In utero Freeway NOx -0.102 0.079 0.20 -0.323 0.460 0.48
Quintile 5 0.092 0.082
0.21
0.749 0.496
0.06

Quintile 4 -0.036 0.081 0.201 0.489
Quintile 3 ref ref
Quintile 2 -0.085 0.081 -0.623 0.494
Quintile 1 -0.079 0.086 -0.517 0.523
Q5xE 0.046 0.125
0.15
0.234 0.745
0.94

Q4xE 0.252 0.117 0.541 0.702
Q3xE ref ref
Q2xE 0.221 0.122 0.504 0.729
Q1xE 0.196 0.144 0.216 0.869
Childhood Freeway NOx -0.023 0.052 0.66 0.059 0.312 0.85

104

Models adjusting for age, race/ethnicity, parental education and Spanish baseline questionnaire at level and growth (n=1099).
a
BMI growth (kg/m
2
) over study follow up and difference in attained BMI at age 10 years scaled to 2 standard deviations of in utero total NOx with 47.9
ppb, freeway NOx with 42.6 ppb, and non-freeway NOx with 14.9 ppb. Childhood NOx was also scaled to 2 SDs where childhood total NOx was 45.1ppb,
childhood freeway NOx was 41.2 ppb, and childhood non-freeway NOx was 8.7 ppb.
*p<0.05

Non-Freeway NOx
BMI Growth
a
BMI Age 10
a

Estimate SE p-value Estimate SE p-value
In utero Non-Freeway NOx -0.018 0.096 0.85 0.911 0.584 0.12
Quintile 5 0.202 0.113
0.28
2.048* 0.691
0.04

Quintile 4 -0.026 0.109 0.603 0.660
Quintile 3 ref ref
Quintile 2 0.012 0.106 0.510 0.646
Quintile 1 0.040 0.113 0.303 0.690
Q5xE -0.128 0.133
0.33
-1.699 0.815

0.15
Q4xE 0.145 0.127 -0.194 0.769
Q3xE ref ref
Q2xE -0.004 0.120 -1.279 0.732
Q1xE -0.054 0.130 -0.998 0.801
Childhood Non-Freeway NOx 0.065 0.055 0.24 0.279 0.317 0.38

105

Supplemental Table 3.8 Associations of First Year of Life NRAP, GRSxE (Unweighted-Quintiles), and Childhood BMI
Total Nox

BMI Growth
a
BMI Age 10
a

Estimate SE p-value Estimate SE p-value
First Year of Life Total NOx -0.097 0.090 0.28 0.072 0.553 0.90
Unweighted GRS Quintile 5 0.038 0.092
0.28
0.658 0.551
0.13
Unweighted GRS Quintile 4 -0.050 0.096 0.275 0.582
Unweighted GRS Quintile 3 ref ref
Unweighted GRS Quintile 2 -0.075 0.094 -0.226 0.569
Unweighted GRS Quintile 1 -0.162 0.098 -0.793 0.600
Q5xE 0.111 0.120
0.37
0.206 0.716
0.85
Q4xE 0.186 0.128 -0.044 0.786
Q3xE ref ref
Q2xE 0.118 0.122 -0.394 0.744
Q1xE 0.266 0.137 0.498 0.841
Childhood Total NOx -0.022 0.054 0.68 0.100 0.322 0.76
Freeway NOx
BMI Growth
a
BMI Age 10
a

Estimate SE p-value Estimate SE p-value
First Year of Life Freeway NOx -0.108 0.089 0.22 -0.236 0.551 0.67
Unweighted GRS Quintile 5 0.045 0.078
0.17
0.535 0.470
0.03

Unweighted GRS Quintile 4 -0.030 0.082 0.143 0.496
Unweighted GRS Quintile 3 ref ref
Unweighted GRS Quintile 2 -0.064 0.080 -0.462 0.488
Unweighted GRS Quintile 1 -0.149 0.083 -0.920 0.505
Q5xE 0.142 0.119
0.15
0.603 0.708
0.71

Q4xE 0.214 0.132 0.277 0.808
Q3xE ref ref
Q2xE 0.140 0.123 0.041 0.752
Q1xE 0.349 0.141 1.039 0.862
Childhood Freeway NOx -0.041 0.052 0.43 0.002 0.313 0.99

106

Models adjusting for age, race/ethnicity, parental education and Spanish baseline questionnaire at level and growth (n=1221).
a
BMI growth (kg/m
2
) over study follow up and difference in attained BMI at age 10 years scaled to 2 standard deviations of of first year of life total NOx with
47.0 ppb, freeway NOx with 42.0 ppb, and non-freeway NOx with 14.1 ppb. Childhood NOx was also scaled to 2 SDs where childhood total NOx was 45.1ppb,
childhood freeway NOx was 41.2 ppb, and childhood non-freeway NOx was 8.7 ppb.
*p<0.05

Non-Freeway NOx
BMI Growth
a
BMI Age 10
a

Estimate SE p-value Estimate SE p-value
First Year of Life Non-Freeway NOx -0.050 0.096 0.60 0.755 0.583 0.20
Unweighted GRS Quintile 5 0.131 0.110
0.7

1.544* 0.667
0.16
Unweighted GRS Quintile 4 0.040 0.103 0.854 0.623
Unweighted GRS Quintile 3 ref ref
Unweighted GRS Quintile 2 -0.015 0.102 0.409 0.615
Unweighted GRS Quintile 1 0.029 0.108 0.249 0.659
Q5xE -0.041 0.133
0.96
-1.082 0.815
0.49
Q4xE 0.025 0.119 -0.845 0.724
Q3xE ref ref
Q2xE 0.016 0.118 -1.220 0.713
Q1xE -0.056 0.128 -1.063 0.785
Childhood Non-Freeway NOx 0.069 0.056 0.22 0.351 0.322 0.28

107

Supplemental Table 3.9 Associations of In Utero NRAP, GRSxE (Weighted-Quintiles), and Childhood BMI
Total Nox

BMI Growth
a
BMI Age 10
a

Estimate SE p-value Estimate SE p-value
In utero Total NOx 0.062 0.098 0.53 0.139 0.551 0.80
Weighted GRS Quintile 5 0.112 0.102
0.7
0.749 0.606
0.5
Weighted GRS Quintile 4 0.035 0.098 0.117 0.586
Weighted GRS Quintile 3 ref ref
Weighted GRS Quintile 2 -0.014 0.096 0.123 0.571
Weighted GRS Quintile 1 -0.023 0.100 -0.404 0.598
Q5xE -0.063 0.144
0.97
0.293 0.835
0.94
Q4xE -0.071 0.130 -0.016 0.763
Q3xE ref ref
Q2xE -0.003 0.124 -0.311 0.726
Q1xE -0.038 0.140 -0.334 0.821
Childhood Total NOx 0.004 0.053 0.94 0.159 0.317 0.61
Freeway NOx
BMI Growth
a
BMI Age 10
a

Estimate SE p-value Estimate SE p-value
In utero Freeway NOx 0.037 0.099 0.71 -0.115 0.547 0.83
Weighted GRS Quintile 5 0.095 0.085
0.58
0.728 0.508
0.23
Weighted GRS Quintile 4 0.004 0.083 -0.012 0.499
Weighted GRS Quintile 3 ref ref
Weighted GRS Quintile 2 -0.023 0.081 -0.019 0.483
Weighted GRS Quintile 1 -0.040 0.083 -0.526 0.500
Q5xE -0.044 0.145
0.99
0.490 0.838
0.93
Q4xE -0.022 0.131 0.319 0.763
Q3xE ref ref
Q2xE 0.025 0.125 -0.070 0.724
Q1xE -0.005 0.142 -0.139 0.824
Childhood Freeway NOx -0.008 0.052 0.88 0.087 0.312 0.78

108

Models adjusting for age, race/ethnicity, parental education and Spanish baseline questionnaire at level and growth (n=1099).
a
BMI growth (kg/m
2
) over study follow up and difference in attained BMI at age 10 years scaled to 2 standard deviations of in utero total NOx with 47.9
ppb, freeway NOx with 42.6 ppb, and non-freeway NOx with 14.9 ppb. Childhood NOx was also scaled to 2 SDs where childhood total NOx was 45.1ppb,
childhood freeway NOx was 41.2 ppb, and childhood non-freeway NOx was 8.7 ppb.
*p<0.05

Non-Freeway NOx
BMI Growth
a
BMI Age 10
a

Estimate SE p-value Estimate SE p-value
In utero Non-Freeway NOx 0.073 0.100 0.46 0.767 0.596 0.20
Weighted GRS Quintile 5 0.140 0.112
0.68
1.296 0.685
0.25
Weighted GRS Quintile 4 0.122 0.113 0.831 0.681
Weighted GRS Quintile 3 ref ref
Weighted GRS Quintile 2 0.048 0.105 0.528 0.637
Weighted GRS Quintile 1 0.028 0.112 -0.079 0.682
Q5xE -0.093 0.138
0.74
-0.564 0.839
0.71
Q4xE -0.192 0.138 -1.077 0.826
Q3xE ref ref
Q2xE -0.087 0.119 -0.872 0.723
Q1xE -0.108 0.134 -0.770 0.806
Childhood Non-Freeway NOx 0.065 0.055 0.24 0.342 0.318 0.28

109

Supplemental Table 3.10 Associations of First Year of Life NRAP, GRSxE (Weighted-Quintiles), and Childhood BMI
Total Nox

BMI Growth
a
BMI Age 10
a

Estimate SE p-value Estimate SE p-value
First Year of Life Total NOx 0.077 0.107 0.47 0.252 0.654 0.70
Weighted GRS Quintile 5 0.089 0.094
0.66
0.737 0.563
0.29
Weighted GRS Quintile 4 0.095 0.097 0.299 0.593
Weighted GRS Quintile 3 ref ref
Weighted GRS Quintile 2 0.044 0.095 0.420 0.577
Weighted GRS Quintile 1 -0.024 0.098 -0.460 0.594
Q5xE -0.003 0.132
0.61
0.486 0.782
0.55
Q4xE -0.170 0.137 -0.429 0.835
Q3xE ref ref
Q2xE -0.075 0.129 -0.618 0.786
Q1xE 0.004 0.143 -0.034 0.868
Childhood Total NOx -0.018 0.054 0.73 0.036 0.323 0.91
Freeway NOx
BMI Growth
a
BMI Age 10
a

Estimate SE p-value Estimate SE p-value
First Year of Life Freeway NOx 0.069 0.110 0.53 -0.053 0.677 0.94
Weighted GRS Quintile 5 0.090 0.080
0.59
0.746 0.482
0.09
Weighted GRS Quintile 4 0.044 0.083 0.020 0.506
Weighted GRS Quintile 3 ref ref
Weighted GRS Quintile 2 0.028 0.081 0.212 0.491
Weighted GRS Quintile 1 -0.037 0.082 -0.615 0.501
Q5xE -0.005 0.134
0.79
0.683 0.792
0.65
Q4xE -0.113 0.141 0.113 0.867
Q3xE ref ref
Q2xE -0.063 0.132 -0.324 0.807
Q1xE 0.046 0.147 0.390 0.896
Childhood Freeway NOx -0.032 0.052 0.54 -0.028 0.314 0.93

110

Models adjusting for age, race/ethnicity, parental education and Spanish baseline questionnaire at level and growth (n=1221).
a
BMI growth (kg/m
2
) over study follow up and difference in attained BMI at age 10 years scaled to 2 standard deviations of of first year of life total NOx with
47.0 ppb, freeway NOx with 42.0 ppb, and non-freeway NOx with 14.1 ppb. Childhood NOx was also scaled to 2 SDs where childhood total NOx was 45.1ppb,
childhood freeway NOx was 41.2 ppb, and childhood non-freeway NOx was 8.7 ppb.
*p<0.05

Non-Freeway NOx
BMI Growth
a
BMI Age 10
a

Estimate SE p-value Estimate SE p-value
First Year of Life Non-Freeway NOx 0.003 0.097 0.97 0.491 0.587 0.40
Weighted GRS Quintile 5 0.056 0.107
0.63
0.905 0.651
0.39
Weighted GRS Quintile 4 0.160 0.107 1.051 0.644
Weighted GRS Quintile 3 ref ref
Weighted GRS Quintile 2 0.030 0.099 0.531 0.597
Weighted GRS Quintile 1 0.035 0.106 0.091 0.644
Q5xE 0.051 0.135
0.23
0.185 0.818
0.22
Q4xE -0.246 0.131 -1.522 0.790
Q3xE ref ref
Q2xE -0.039 0.114 -0.712 0.689
Q1xE -0.086 0.130 -0.863 0.780
Childhood Non-Freeway NOx 0.072 0.056 0.20 0.434 0.323 0.18

111

Chapter 4: Associations of Air Pollution, Obesity and Cardiometabolic Health
in Young Adults: The Meta-AIR Study

4.1 Abstract
Growing evidence indicates exposure to air pollution contributes to obesity and
cardiometabolic disease risk in children and adults, however studies are lacking in young
adulthood, an important transitional period in the life course. The aim of this study was to
examine the associations of short- and long-term regional ambient and near-roadway air
pollution (NRAP) exposures on adiposity and cardiometabolic health in young adults aged 17-22
years. From 2014-2018, a subset of participants (n=158) were recruited from the Children’s
Health Study to participate in the Meta-AIR (Metabolic and Asthma Incidence Research) study
to assess obesity (body composition and abdominal adiposity) and cardiometabolic health (blood
pressure, fasting glucose, fasting insulin and lipid profiles) measures. Prior 1-month and 1-year
average air pollution exposures were calculated from residential addresses. This included
nitrogen dioxide (NO2), ozone (O3), particulate matter with aerodynamic diameter <10 μm
(PM10), particulate matter with aerodynamic diameter <2.5 μm (PM2.5)) and NRAP (freeway,
non-freeway, and total nitrogen oxides (NOx)) exposures. Linear regression models examined
associations of prior 1-month (short-term) and 1-year (long-term) air pollution exposures on
obesity and cardiometabolic factors adjusting for covariates and past childhood air pollution
exposures. A 1 standard deviation (SD) change in long-term NO2 exposure was associated with a
11.2 mg/dL higher level of total cholesterol (p=0.04) and 9.4 mg/dL higher level of low-density
lipoproteins (LDL)-cholesterol (p=0.04). Among obese participants, associations between long-
term NO2 and total cholesterol and LDL-cholesterol were 4.5 and 9 times larger than the

112

associations in non-obese participants (pinteraction=0.008 and 0.03, respectively). Additionally, we
observed a statistically significant association with increased short-term O3 exposure and higher
triglyceride and very-low-density lipoprotein (VLDL) cholesterol levels as well as lower high-
density lipoprotein (HDL) cholesterol levels. With adiposity measures, short-term O3 exposure
was associated with higher hepatic fat levels (p=0.02). Increased short-term NO2, PM10, and
PM2.5 were associated with increased systolic blood pressure (p=0.003, p=0.03, p=0.009,
respectively). Similarly, increased short-term PM2.5 exposure was associated with increased
diastolic blood pressure levels (p=0.003). Amongst glucose-related factors, long-term PM2.5
exposure was associated with higher levels of insulin area under the curve (p=0.03), and short-
term NO2 was associated with higher levels of glucose under the curve (p=0.04). There were no
other statistically significant associations with short- or long-term air pollutants and BMI, other
measures of adiposity, and cardiometabolic outcomes. Higher exposure to regional air pollutants,
namely prior 1-year average NO2, was associated with higher fasting serum lipid measures.
These associations were more pronounced in obese participants, suggesting obesity may
exacerbate the effects of air pollution exposure on lipids levels in young adults. Further studies in
young adults are warranted as our study suggests deleterious associations of both short- and
long-term air pollution exposures and lipid metabolism.

113

4.2 Introduction
New data shows significant increasing obesity trends in both youth and adults from 1999-
2016 (1). Early onset of detrimental health effects is a concern as obesity can influence the
development of type 2 diabetes (T2DM) (2-5) and cardiovascular disease (CVD) (6, 7) later in
life. As obese children are likely to become obese adults (8), it is important to characterize
obesity and cardiometabolic profiles in young adults as they are at the forefront of the obesity
epidemic and have greater risk to obesity-related health consequences. Apart from traditional
obesity risk factors of poor diet, low physical activity, and low socioeconomic status (SES),
epidemiological evidence has shown that air pollution may contribute to increased risk for
obesity (9-12) and cardiometabolic disease (13-17). Furthermore, several studies have shown
stronger associations with air pollution and adverse cardiovascular health in obese subjects
compared to normal weigh subjects suggesting that obesity status may exacerbate the effects of
air pollution (15, 18). Research on the effects of air pollution on obesity and cardiometabolic
outcomes focused on the young adulthood period, however, is lacking in the literature.
Our current study, the Meta-AIR (Metabolic and Asthma Incidence Research) study, is a
subset of young adults aged 17-22 years from the larger Southern California Children’s Health
Study (CHS). We examined associations of prior 1-month (short-term) and prior 1-year (long-
term) air pollution exposures on various indicators of obesity and cardiometabolic health.
Regional ambient pollutants explored include nitrogen dioxide (NO2), ozone (O3), particulate
matter with aerodynamic diameter <10 μm (PM10), and particulate matter with aerodynamic
diameter <2.5 μm (PM2.5), and near-roadway air pollution (NRAP) include freeway, non-
freeway, and total nitrogen oxides (NOx). The aim of this study was to determine if prior 1-
month or 1- year ambient and NRAP exposures are associated with obesity measures and

114

cardiometabolic outcomes in young adults. We hypothesized that increased exposure to prior 1-
month and 1-year ambient and NRAP will be associated with higher levels of adiposity measures
and adverse levels of cardiometabolic outcomes.

4.3 Material and Methods
Study Recruitment
The Meta-AIR study is a subset of young adults aged 17-22 years who were originally
part of the larger CHS. Details of the CHS have been described previously (19). Briefly, in 2002
a cohort of kindergarten and first grade children were recruited from public schools across
Southern California communities and followed through their high school years. Meta-AIR
subjects were selected based on their high school overweight or obese status in 2011-2012 of
CHS as well as predicted NOx exposures from their respective residential addresses in CHS
towns. Potential participants were oversampled from “low” and “high” predicted NOx exposures
to ensure maximum exposure contrast amongst study subjects within each CHS Southern
California community. This recruitment strategy allowed for a wide range of air pollution
exposures amongst potentially overweight and obese CHS young adults. Inclusion criteria
included age- and sex-specific BMI percentiles ≥ 85
th
percentile measured by CHS staff in
school year 2011-2012. Exclusion criteria were as follows: ineligible if using any medications
known to influence body composition and insulin action/secretion, any diagnosis of diseases that
may influence insulin or body composition including Type 1 and Type 2 diabetes, and any major
illness since birth. Eligible participants, who are now young adults, were then contacted and
invited to enroll in the Meta-AIR study between 2014-2018. Written informed assents and

115

consents were obtained from study participants. The Institutional Review Board at the University
of Southern California approved this study.

Study Design
The Meta-AIR study visit included several questionnaires as well as extensive phenotyping of
obesity and cardiometabolic outcomes conducted at the University of Southern California
Diabetes and Obesity Research Institute and the Clinical Trials Unit from 2014-2018. The study
visit flow is shown in Figure 4.1. In short, we administered questionnaires detailing
sociodemographic characteristics, parental health and education, smoking history including e-
cigarette use, self-reported physical activity, residential history, and 24-hour diet recalls (20).
The first dietary recall was completed in person at the study visit, and the second was conducted
by phone. A third phone recall was conducted if one of the first two recalls was either “more
than usual” or “less than usual” from what the participant usually consumes day to day. These
diet data were processed using the Nutrition Data System for Research (version 2014, University
of Minnesota). Details of adiposity and cardiometabolic measures obtained are below

Adiposity and Cardiometabolic Outcomes
Adiposity
Several anthropometric and body composition measures were taken to estimate adiposity: 1)
height and weight to determine body mass index (BMI) where BMI= weight/height
2
(kg/m
2
), 2)
dual-energy X-ray absorptiometry (DEXA) scan to determine total body fat percent, and 3) 3T
magnetic resonance imaging (MRI) abdominal scan to determine subcutaneous abdominal

116

adipose tissue (SAAT), visceral adipose tissue (VAT), and hepatic fat fraction (HFF). Obesity
was defined as BMI≥ 30 kg/m
2
and non-obesity defined as BMI <30.0 kg/m
2
.

Cardiometabolic outcomes: glucose and lipid metabolism
All participants underwent three measures of blood pressure where average of the measures was
then taken for average systolic and diastolic blood pressure. Following a minimum 10-hour fast,
a 2-hour oral glucose tolerance test (OGTT) was administered using a load of anhydrous glucose
dissolved in water for 1.75 grams per kilogram of body weight with a max dose of 75 grams. All
participants received the maximum glucose load. Blood glucose and insulin samples were
collected at fasting (pre-glucose load) and then post glucose challenge at 30-, 60-, 90-, and 120-
minutes. Glucose-related outcomes included fasting glucose, glucose at 120 min. (last draw of
OGTT), glucose area under the curve (AUC), fasting insulin, insulin AUC, homeostatic model
assessment for insulin resistance (HOMA-IR), and the Matsuda Index. Glucose and insulin AUC
were calculated using the trapezoidal method using all time points from the OGTT. HOMA-IR
gives estimates of insulin resistance (IR) from fasting insulin and glucose concentrations where
HOMA-IR =
𝑓𝑎𝑠𝑡𝑖𝑛𝑔 𝑔𝑙𝑢𝑐𝑜𝑠𝑒 ∗𝑓𝑎𝑠𝑡𝑖𝑛𝑔 𝑖𝑛𝑠𝑢𝑙𝑖𝑛 405
(21). The Matsuda Index gives an approximation of
whole-body insulin sensitivity using all times points from the OGTT were the ratio of plasma
glucose to insulin concentrations are calculated.
Matsuda index is defined as
1000
√ glucose
fasting
∗insulin
fasting
∗A∗B
such that
A=
(glucose fasting ∗15+glucose 30min∗30+glucose 60min∗30+glucose 90min∗30+glucose 120min∗15)
120
and
B=
(insulin fasting ∗15+insulin 30min∗30+insulin 60min∗30+insulin 90min∗30+insulin 120min∗15)
120
(22).

117

Fasting lipid-related outcomes included triglycerides, total cholesterol, high-density lipoprotein
(HDL), low-density lipoprotein (LDL), and very-low-density lipoprotein (VLDL) cholesterols.

Assays
Blood samples from the OGTT were collected in potassium oxalate, sodium fluoride 2mL tubes
and centrifuged for 15 minutes at 1500 RCF. These plasma samples were then assayed for
glucose concentration by hexokinase-mediated reaction assay run on Roche Covas C501.
Additional OGTT samples were collected in sodium heparin 2mL tubes for insulin and
centrifuged at 2500 RPM for 10 minutes. Plasma samples were stored at -80°C and later assayed
for insulin in duplicate by Human Insulin ELISA Kit (EZHI-14BK). Fasting blood for lipids was
collected in serum separator 4mL tube, inverted several times, placed in room temperature for 60
minutes for clotting and centrifuged at 2000 RPM for 10 minutes. Serum lipid samples were
stored at -80°C and later assayed in duplicate by Fujifilm Wako Diagnostics enzymatic assay.

Air Pollution Exposures
Residential history was collected from all Meta-AIR participants at their study visit including
move in and move out dates for each respective residence. Street level data from residential
addresses were geocoded using the Texas A&M geocoder (23) and assigned latitude and
longitude coordinates. Monthly air pollution data was averaged for prior 1-month and 1-year
regional ambient and near-roadway air pollution (NRAP) exposures to reflect short-term and
long-term exposures prior to each participant’s study visit. Exposures were weighted by time
spent at each different residential address by month since some of our Meta-AIR participants
were college students who lived between two residences during the year. In these instances,

118

short-term and long-term exposures prior to the study visit accounts for both college and parental
home residences using move in and move out months to appropriately weigh time spent at each
respective residence. Additionally, our analysis included historic air pollution exposures or
cumulative childhood exposures that were obtained from the parent study CHS, which account
for past exposures beyond our periods of interests: prior 1-month and 1-year. For regional
ambient pollutants, historic air pollution was defined as average childhood exposures for each
participant from birth through year 2011. For NRAP, historic air pollution exposure was defined
as average childhood exposures from CHS study entry (May 2003) through year 2011 where all
our study participants had NRAP data available.

Ambient Air Pollution Exposures
Regional exposures were obtained from ambient monitoring stations by downloading hourly air
quality data from the U.S. Environmental Protection Agency’s Air Quality System
(http://www.epa.gov/ttn/airs/airsaqs). Daily averages for four regional ambient air pollutants,
NO2, O3, PM10, and PM2.5, were calculated. For O3 only, levels were characterized as the eight-
hour average daily maximum concentrations. Air monitoring stations in California are spaced
20-30 kilometers (km) apart, which provides a good characterization of air pollution gradients
across the region. Gaseous pollutants like NO2 and O3 are measured by the Federal Reference
Method (FRM) monitors while particulates like PM10 and PM2.5 are measured through FRM and
Federal Equivalent Method (FEM) monitors. Monthly averages were calculated from daily data
using 75% completeness criteria. To calculate monthly ambient exposures, parcel level data was
used in the inverse distance-squared weighting algorithm which spatially interpolated air quality

119

data from up to four monitoring stations within a 50 km radius of the participant’s residence
(24).

Near-Roadway Air Pollution Exposures
NRAP exposures were estimated by the California Line Source Dispersion Model (CALINE4)
through detailed residential history where street-level residential addresses were geocoded to
parcel level. CALINE4 line-source dispersion model then estimated concentrations of near-road
NOx at each latitude and longitude for freeway and non-freeway roads using traffic emissions
(calculated within 5 km buffer of the residence), traffic volume, roadway geometry and
meteorological conditions including wind speed and direction, pollution mixing heights, and
atmospheric stability (25). Traffic counts and road geometry were obtained from Caltrans and
TeleAtlas/GDT, and average daily traffic volumes were assigned based on year. Monthly near-
road freeway, non-freeway and total NOx (sum of freeway and non-freeway) were then
calculated for 1-month (short-term) and 1-year (long-term) average NRAP exposures prior to the
study visit.

Statistical Methods
Physical and cardiometabolic characteristics of the cohort were compared by obesity status (non-
obese vs obese) using chi-square or t-tests. Non-obesity was defined as BMI <30.0 kg/m
2
and
obesity as BMI≥ 30.0 kg/m
2
.

All outcomes were assessed for normality and skewed measures
were log transformed to fit a normal distribution. Triglycerides, VLDL-cholesterol, fasting
insulin, HOMA-IR, Matsuda Index were log transformed to meet assumptions of the linear
regression. One subject was removed from this analysis due to undiagnosed diabetes; another

120

subject was removed from the glucose-related cardiometabolic analysis due to a high fasting
insulin that was greater than 4 standard deviations (SDs) above the mean.
Linear regression models were used to estimate effects of short-term (1-month) and long-
term (1-year) air pollution exposures prior to study visit on obesity and cardiometabolic
measures. Models were adjusted for age, sex, race/ethnicity, occupational status of participant
(SES surrogate), parental education (SES surrogate), self-reported exercise, current cigarette
smoking, e-cigarette use, body fat percent, diet (average total calories per day), season of study
visit (warm or cool), historic air pollution exposure, and baseline CHS town as a random effect.
Historic air pollution data, or cumulative childhood exposures, allows us to evaluate the effects
of short- and long-term exposures on obesity- and cardiometabolic-related outcomes independent
of past childhood air pollution exposures. Given we had this data available from the parent CHS
study, we included these historic exposures to be able to account for the more recent short-term
or long-term exposures of interest. Besides near-road non-freeway and total NOx, historic air
pollution exposures have low to median correlation with prior 1-month or 1-year average air
pollution exposures (all spearman correlation coefficient ≤0.7, Supplement Table 2). Association
estimates of air pollution exposure and obesity- and cardiometabolic-related outcomes are
reported for a 1 SD in air pollution exposure for prior 1-month and 1-year average regional
ambient and NRAP exposures. We also investigated whether the associations between air
pollution exposure and metabolic outcomes differed by sex, race/ethnicity and obesity status by
testing interaction terms in the full model. Additionally, we further explored associations with
multipollutant models with additional short-term pollutants in short-term associations as well as
additional long-term pollutants in long-term air pollution exposure associations. A two-sided p-

121

value < 0.05 was considered statistically significant for all models. All analyses were performed
in SAS, version 9.4 (SAS, Institute, Cary, NC).

4.4 Results
From 2014-2018, the Meta-AIR study enrolled 158 young adults who underwent
extensive obesity and cardiometabolic phenotyping. General study characteristics are presented
in Table 4.1. Briefly, mean age of participants was 19.7 years (SD=1.2, range=17.6-22.9). There
were slightly more males than females (52.5% vs 47.5%), and 60% of participants were
Hispanic, 28% were Non-Hispanic White and remaining 13% of participants were Asian,
African American or other/mixed races. Generally, participants were full-time college students,
students with part time/full time jobs, or working full time. Approximately 80% of participant’s
parents had education levels of high school graduation and beyond. About 6% of participants
were current smokers who have smoked 20 cigarettes or more in the past month, and e-cigarette
ever use was about 15% amongst study participants. Our participants consumed an average of
2050 kcal (SD=632) per day obtained from the dietary recalls. Sociodemographic characteristics
did not differ by obesity status (non-obese vs obese) across all variables (all p>0.1, Table 4.1).
Mean adiposity- and cardiometabolic-related outcomes amongst all Meta-AIR
participants as well as by obesity status are shown in Table 4.2. Of the 158 participants, 37%
were obese (n=59) with BMI ≥ 30 kg/m
2
, 47% were overweight (n=75) with 25 kg/m
2
≤ BMI<
30 kg/m
2
, and 15% had normal BMI (n=24) with BMI<25 kg/m
2
. Amongst all study participants,
mean BMI was 29.9 kg/m
2
(SD=5.1) and mean body fat percent was 34.9% (SD=8.5). As
expected, obesity-related measures (BMI, total body fat percent, SAAT, VAT, and HFF) were
higher in obese compared to non-obese participants, all p<0.0001 (Table 4.2). Cardiometabolic

122

measures were classified into two groups: lipid and glucose metabolism in addition to blood
pressure measures. Higher levels of systolic and diastolic blood pressure are seen in obese versus
non-obese participants (p=0.009 and 0.001). For lipid metabolism, means for fasting lipid
measures are presented in Table 4.2. Higher levels of triglycerides and VLDL-cholesterol are
seen in obese versus non-obese subjects (p=0.0004); furthermore, lower HDL-cholesterol levels
were seen in obese compared to non-obese subjects (p=0.006). Total cholesterol and LDL-
cholesterol levels were similar across non-obese and obese subjects (p=0.6 and 0.7, respectively).
Details of glucose metabolism measures are found in Table 4.2. Higher levels of glucose-related
metabolic measures, like fasting glucose and fasting insulin, are seen in obese compared to non-
obese participants, all p<0.01. Compared to the non-obese participants, obese participants show
early signs of insulin resistance with higher HOMA-IR (3.2 in obese vs 1.5 in non-obese) and
lower Matsuda index levels (3.9 in obese vs 7.3 in non-obese) (Table 4.2). In adults, the HOMA
cutoff point for IR is >2.5 (21); however studies in children and adolescents have proposed
higher cut off points >3.16 (26) and >4.0 (27). The Matsuda Index ≤2.5 has been proposed as the
cut off for IR (28).
Prior to study visit, 1-month and 1-year average regional ambient and NRAP exposures
are shown in Table 4.3, and historic exposures are show in Supplemental Table 4.1. To avoid
potential collinearity of 1-month vs historic or 1-year vs historic air pollution exposures in the
same model, Spearman correlations between 1-month vs historic exposures and 1-year vs
historic exposures are shown in Supplemental Table 4.2. All regional ambient models (NO2,
O3, PM10, and PM2.5) for 1-month and 1-year average air pollution exposures included the
historic exposures (all correlations<0.7). For NRAP models (freeway, non-freeway, total NOx),

123

freeway NOx was the only NRAP exposure that included the historic exposures in the 1-month
and 1-year NRAP models (correlation<0.7).

Associations of Short- and Long-Term Ambient Air Pollution and Obesity-Related Outcomes
Associations of short-term (1-month) and long-term (1-year) average ambient air pollution
exposures and obesity-related outcomes are shown in Table 4.4. All models reflect association
estimates for one obesity outcome and one short-term or long-term ambient air pollutant
adjusting for age, sex, race/ethnicity, occupational status of participant, parental education, self-
reported exercise, current cigarette smoking, e-cigarette use, total body fat % (not included in
SAAT, VAT, HFF models), diet, season of visit, and respective historic air pollution exposures.
In models pertaining to adiposity measures of total body fat percent and abdominal adiposity
(SAAT, VAT, and HFF), we found prior 1-month O3 exposure was statistically, significantly
associated with HFF, liver fat. A 1 SD (14.1 ppb) increase in prior 1-month O3 exposure was
associated with a 20% higher liver fat levels after adjusting for covariates (p=0.02) (Table 4.4).
We further explored the association between liver fat and short-term O3 exposure in
multipollutant models where short-term NO2 or short-term PM2.5 were added separately to the
model as these pollutants were not highly correlated with short-term O3 (Supplemental Table
4.3). Association estimates were slightly attenuated by adding short-term NO2 or PM2.5, however
associations remained statistically significant (Supplemental Table 4.5). Additional sensitivity
analysis was conducted by including SNP rs738409, a variant of the PNPLA3 gene in the
association of liver fat and short term O3 exposure. The effect estimate remained unchanged
however the association was marginally significant at p=0.07. The association between HFF and
short-term O3 was not modified by sex (male vs female), Hispanicity (non-hispanic white vs

124

124ispanic), or obesity status (obese vs non-obese) (all pinteraction>0.1). We did not find any other
statistically significant associations of BMI, total body fat percent, SAAT, or VAT and short- or
long-term ambient measures of NO2, O3, PM10, and PM2.5.

Associations of Short-Term Ambient Air Pollution and Cardiometabolic-Related Outcomes
Associations with short-term, prior 1-month, ambient pollutant exposures and cardiometabolic
measures are shown in Table 4.5. Increased short-term NO2, PM10, and PM2.5 were associated
with higher systolic blood pressure levels. A 1 SD (5.7 ppb) increase in prior 1-month NO2
exposure was associated with a 3.7 mm/Hg higher systolic blood pressure levels after adjusting
for covariates (p=0.003). Similarly, a 1 SD (9.7 μg/m
3
) increase in prior 1-month PM10 exposure
was associated with a 2.2 mm/Hg higher systolic blood pressure levels (p=0.03) and lastly, a 1
SD (4.3 μg/m
3
) increase in prior 1-month PM2.5 exposure was associated with a 2.9 mm/Hg
higher systolic blood pressure levels after adjusting for covariates (p=0.009). Diastolic blood
pressure was also associated with short-term PM2.5 (p=0.003).
Amongst lipid metabolism measures, statistically significant associations with higher
short-term O3 exposures and higher triglycerides, higher VLDL-cholesterol, and lower HDL-
cholesterol levels were found after adjusting for covariates (all p<0.05, Table 4.5). For example,
a 1 SD (14.1 ppb) increase in prior 1-month O3 exposure was associated with a 18% higher
triglyceride levels (p=0.04) and a 18% higher VLDL-cholesterol levels (p=0.04). Additionally, a
1 SD increase in prior 1-month O3 exposure was associated with a 3.02 mg/dL lower HDL levels
after adjusting for covariates (p=0.03). We further explored short-term O3 and lipid associations
in multipollutant models by adding short-term NO2 or short-term PM2.5 (Supplemental Table
4.5). In multipollutant models with short-term O3 and NO2, there was slight attenuation in

125

association estimates across triglycerides, HDL-cholesterol and VLDL-cholesterol but all
associations remained statistically significant (all p<0.05). With short-term O3 and PM2.5,
association estimates were also attenuated, however triglyceride and VLDL-cholesterol models
were no longer statistically significant (p=0.09 and p=0.09, respectively). Effect modification by
sex, Hispanicity, and obesity status for triglycerides, HDL-cholesterol, and VLDL-cholesterol
models were further explored however there were no statistically significant interactions were
found (all pinteraction>0.1). No other statistically significant associations were found with short-
term ambient pollution exposures, NO2, PM10, and PM2.5, and lipid metabolism measures.
Amongst glucose metabolism measures, higher short-term NO2 exposure was associated with
higher glucose AUC levels (p=0.04). Other glucose metabolism measures of fasting glucose,
glucose 120 min., fasting insulin, insulin AUC, HOMA-IR and Matsuda Index were not
associated with short-term ambient pollutants.

Associations of Long-Term Ambient Air Pollution and Cardiometabolic-Related Outcomes
Associations of long-term, prior 1-year, ambient pollution exposures and cardiometabolic
measures are shown in Table 4.6. Amongst lipid metabolism measures, higher long-term
ambient NO2 exposure was associated with higher fasting total cholesterol and LDL-cholesterol
levels in this cohort of young adults (Table 4.6). For example, a 1 SD (3.9 ppb) increase in 1-
year average NO2 exposure was associated with 11.3 mg/dL higher total cholesterol levels after
adjusting for covariates (p=0.04). Similarly, a 1 SD increase in 1-year NO2 exposure was
associated with a 9.4 mg/dL higher LDL-cholesterol levels (p=0.04). Additional exploration with
multipollutant models was completed by adding long-term O3 or long-term PM2.5.
(Supplemental Table 4.4). In multipollutant models with long-term NO2 and long-term O3,

126

associations were slightly attenuated in total cholesterol and LDL-cholesterol models with
marginal significance (p=0.05 and p=0.06, respectively) (Supplemental Table 4.6). In
multipollutant models with long-term NO2 and long-term PM2.5, associations strengthened when
adding long-term PM2.5 in total cholesterol and LDL-cholesterol models maintaining statistical
significance p<0.05 (Supplemental Table 4.6).
Associations of long-term NO2 exposure and total cholesterol and LDL-cholesterol were
further assessed for effect modification by sex, Hispanicity, and obesity status. There were no
statistically significant interactions of sex and hispancity (all p>0.1); however, the interaction for
obesity status (non-obese vs obese) and NO2 were statistically significant for total cholesterol
(p=0.008) and LDL-cholesterol (p=0.03). These results suggest differences in the effect of long-
term NO2 exposure on lipid levels by obesity status, so associations were stratified by obesity
status. In obese subjects, the association estimate of prior 1-year NO2 exposure on total
cholesterol and LDL-cholesterol were substantially higher compared to non-obese subjects
(Figure 4.2). The association estimates between prior 1-year NO2 exposure and total cholesterol
among obese participants were nearly 5-fold larger (21.4 mg/dL vs 4.7 mg/dL) than non-obese
participants. Likewise, the association estimates between prior 1-year NO2 exposure and LDL-
cholesterol among obese participants were 9-fold larger (19.9 mg/dL vs 2.2 mg/dL) than non-
obese participants. There were no statistically significant associations with long-term ambient
exposures and other lipid and blood pressure measures.
Amongst glucose metabolism measures, higher long-term PM2.5 exposure was associated
with higher insulin AUC such that a 1 SD (2.5 μg/m
3
) increase in prior 1-year PM2.5 exposure
was associated with 33.6 higher units of insulin AUC (p=0.03). Further exploration was
conducted with multipollutant models adding long-term NO2 or long-term O3 (Supplemental

127

Table 4.4). In the multipollutant model with long-term PM2.5 and NO2, the association with
insulin AUC was attenuated and was no longer statistically significant (p=0.09, Supplemental
Table 4.7). Similarly, the multipollutant model with long-term PM2.5 and O3, further attenuation
was found with loss in significance (p=0.17, Supplemental Table 4.7). There was no effect
modification with insulin AUC and long-term PM2.5 exposure by sex, Hispanicity, and obesity
status (pinteraction>0.1). There were no other statistically significant associations between long-
term ambient exposures, NO2, O3, PM10, and PM2.5, and glucose metabolism measures of fasting
glucose, glucose 120min., glucose AUC, fasting insulin, insulin AUC, HOMA-IR, and Matsuda
Index.

Associations of Long-Term and Short-Term NRAP and Obesity- and Cardiometabolic-Related
Outcomes
Like regional ambient air pollution exposures, associations of short-term and long-term NRAP
exposures with obesity- and cardiometabolic-related outcomes were explored (Supplemental
Table 4.8-4.9). There were no statistically significant associations with prior 1-month and 1-year
average NRAP exposures of non-freeway, freeway and total NOx with obesity- or
cardiometabolic-related outcomes.

4.5 Discussion
In the Meta-AIR study, we conducted a comprehensive analysis with short- and long-
term regional ambient and NRAP exposures (in both single- and multi-pollutant models) and
obesity- and cardiometabolic-related outcomes, and found associations in several outcomes
including lipid profiles, blood pressure, liver fat and glucose/insulin-related phenotypes. First,

128

compared to non-obese young adults, obese young adults in our study showed increased risk of
glucose and lipid metabolism dysfunction. Second, amongst short-term ambient exposures we
showed that higher short-term O3 exposure was associated with higher liver fat as well as higher
triglyceride, higher VLDL-cholesterol and lower HDL-cholesterol levels. Several measures of
short-term ambient pollutants were associated with blood pressure measures, and short-term NO2
exposure was associated with glucose AUC. Third, amongst long-term exposures, we found
higher long-term NO2 exposure was associated with higher levels of total cholesterol and LDL-
cholesterol and higher long-term PM2.5 exposure was associated with higher insulin AUC after
adjusting for covariates and past childhood air pollution exposures. Importantly, obesity status
modified the associations between long-term NO2 and total cholesterol and LDL-cholesterol
levels as we see multi-fold higher association estimates in obese compared to non-obese subjects.
Though we did not find statistically significant associations with short-term or long-term
ambient and NRAP exposures and BMI and some adiposity measures, we showed a positive
association where higher short-term (prior 1-month) O3 exposure was associated with higher
liver fat in young adults. Increasing incidence of non-alcoholic fatty liver disease (NAFLD), an
accumulation of liver fat, has been strongly liked to obesity, and many with NAFLD are obese
and insulin resistant (29). Additionally, there is some evidence of the role of air pollution on
NAFLD (30).
Our findings show that higher short-term O3 and higher long-term NO2 exposures may
increase risk of dyslipidemia in young adults. Though it is not clear why higher short-term O3
and higher long-term NO2 exposures affects different lipid types, perhaps some ambient
pollutants like O3 elicit acute or short-term effects on lipid profiles whereas NO2 exposures may
have more chronic or long-term effects on lipid profiles. Several reported associations support

129

our study findings, however most studies have focused effects of air pollution and lipid
abnormalities in adult and elderly populations (31-34). Similar to our findings, a recent study of
15,000 Chinese adults (aged 18-74 years) detected statistically significant associations with
increased long-term ambient air pollution exposures and altered lipid measures with stronger
associations in obese participants (34). Studies in young adults are lacking; however, one study
in youth has shown deleterious effects of poor air quality to elevated levels of total cholesterol
and triglycerides in Irani adolescents (35). Additionally, experimental model has shown that
increased air pollution can perturb lipid levels (36). With evidence from both epidemiological
studies and animal model, air pollution can affect lipid metabolism though exact biological
mechanisms remain uncertain. One proposed mechanism is through air pollution exposure
triggering an inflammatory response and inducing macrophage infiltration in adipose tissue (37).
Macrophage infiltration cues expression of proinflammatory cytokines inducing uncontrolled
lipolysis which leads to elevated levels of circulating non-esterified fatty acids (38). These fatty
acids are then transported to the liver for upregulation of triglycerides synthesis, VLDL
production and ultimately dyslipidemia.
Lastly, associations of air pollution and glucose metabolism in children and adolescents
has been shown previously (13, 39, 40); yet studies are limited in young adults. One study from
Southern California showed adverse effects of higher NO2 and PM2.5 exposures on insulin
sensitivity and beta-cell function in overweight and obese children (aged 8-15 years) after an
average 3 year follow up period (13). Though we did not find statistically significant
associations with NRAP exposures and glucose metabolism measures, higher long-term (prior 1-
year) PM2.5 exposure was associated with higher insulin AUC levels as well as higher short-erm
NO2 exposure was associated with higher glucose AUC levels. Several animal models have

130

proposed mechanisms in which air pollution may affect glucose metabolism (37, 41). Though
obesity and cardiometabolic risk factors like glucose- and lipid-related metabolism are closely
related, these studies show a multitude of pathways in which air pollution uniquely effects
different organ systems leading to various adverse health outcomes.
The Meta-AIR study has notable strengths. Unlike most air pollution studies, our study
had life-time residential history on our participants by which we were able to incorporate past,
childhood air pollution exposures. We used the well-established CHS to recruit our Meta-AIR
participants where we sampled across high and low air pollution exposures across CHS
communities to ensure a wide range of air pollution exposures in both regional ambient and
NRAP exposures. An extensive number of exposure metrics (both short- and long-term regional
and local traffic pollutants) alongside various measures of adiposity and cardiometabolic health
were carried out in this study. A rigorous collection of adiposity measures using gold-standards
such as DEXA and MRI and cardiometabolic-related outcomes were performed prospectively.
We collected many different measures of modifiable risk factors such as diet, physical activity,
current smoking and e-cigarette use, and non-modifiable risk factors age, race/ethnicity, sex,
occupational status of participant and parental education via questionnaires. Our air pollution
exposure estimates incorporated multiple residences as our study population included college-
aged subjects who resided at their parent’s residence as well as their school residence giving
appropriate weights to each respective residence.
Despite the strengths, there were limitations to this current study. First, we cannot draw
causal relationships between air pollution exposures and obesity- and cardiometabolic-related
outcomes as our study outcomes were only collected at one time point. We were limited in our
sample size as we only had 158 young adults in this study; therefore, various interactions tested

131

should be interpreted with caution. Given our study design, our participants were primed for
potentially adverse levels of obesity and cardiometabolic outcomes, however given the standard
deviations of our clinical measures, there was range of outcomes. Despite capturing multiple
residences, we were unable to incorporate other locations such as work locations that our
subjects may have frequently visited or indoor exposures. We also acknowledge our variables of
occupational status of the participant and parental education may not capture SES fully. Our
study findings may be only generalizable to young adults with similar demographic data
(primarily Hispanic or White), with similar range of air pollution exposures, and those with a
history of overweight/obesity during mid-teenage years. Additionally, markers of oxidation were
not available in this current study. Finally, our findings cannot determine the precise mechanism
behind the association of air pollution and obesity and cardiometabolic health, however our
results indicate potential pathways that may involve disrupted lipid metabolism, increased liver
fat and increased insulin production.
In conclusion, findings from the Meta-AIR study suggests that differential ambient
regional air pollution exposures, NO2, O3 and PM2.5, may contribute to poor cardiometabolic
health in young adults aged 17-22 years. Notably, the association between long-term NO2 and
fasting lipid measures may adversely affect obese young adults compared to non-obese young
adults. Differences in association by obesity status suggest that obese young adults may be more
susceptible to adverse effects of long-term air pollution exposure, and this may exacerbate
indicators of cardiometabolic health. Additional longitudinal studies in young adults are
warranted as to verify associations of air pollution and adverse obesity and cardiometabolic
outcomes.

132

4.6 References

1. Hales CM, Carroll MD, Fryar CD, et al. Prevalence of Obesity Among Adults and Youth:
United States, 2015-2016. NCHS Data Brief 2017(288):1-8.
2. Hannon TS, Rao G, Arslanian SA. Childhood obesity and type 2 diabetes mellitus.
Pediatrics 2005;116(2):473-80.
3. Buchanan TA. Pancreatic beta-cell loss and preservation in type 2 diabetes. Clin Ther
2003;25 Suppl B:B32-46.
4. Weiss R, Kaufman FR. Metabolic complications of childhood obesity: identifying and
mitigating the risk. Diabetes Care 2008;31 Suppl 2:S310-6.
5. Sinha R, Fisch G, Teague B, et al. Prevalence of impaired glucose tolerance among
children and adolescents with marked obesity. N Engl J Med 2002;346(11):802-10.
6. Juonala M, Magnussen CG, Berenson GS, et al. Childhood adiposity, adult adiposity, and
cardiovascular risk factors. N Engl J Med 2011;365(20):1876-85.
7. Freedman DS, Mei Z, Srinivasan SR, et al. Cardiovascular risk factors and excess
adiposity among overweight children and adolescents: the Bogalusa Heart Study. J
Pediatr 2007;150(1):12-7.e2.
8. Ward ZJ, Long MW, Resch SC, et al. Simulation of Growth Trajectories of Childhood
Obesity into Adulthood. N Engl J Med 2017;377(22):2145-53.
9. Jerrett M, McConnell R, Wolch J, et al. Traffic-related air pollution and obesity
formation in children: a longitudinal, multilevel analysis. Environ Health 2014;13:49.
10. McConnell R, Shen E, Gilliland FD, et al. A longitudinal cohort study of body mass
index and childhood exposure to secondhand tobacco smoke and air pollution: the
Southern California Children's Health Study. Environ Health Perspect 2015;123(4):360-
6.
11. Kim JS, Alderete TL, Chen Z, et al. Longitudinal associations of in utero and early life
near-roadway air pollution with trajectories of childhood body mass index. Environ
Health 2018;17(1):64.
12. Rundle A, Hoepner L, Hassoun A, et al. Association of childhood obesity with maternal
exposure to ambient air polycyclic aromatic hydrocarbons during pregnancy. Am J
Epidemiol 2012;175(11):1163-72.
13. Alderete TL, Habre R, Toledo-Corral CM, et al. Longitudinal Associations Between
Ambient Air Pollution With Insulin Sensitivity, beta-Cell Function, and Adiposity in Los
Angeles Latino Children. Diabetes 2017;66(7):1789-96.
14. Brook RD, Rajagopalan S, Pope CA, 3
rd
, et al. Particulate matter air pollution and
cardiovascular disease: An update to the scientific statement from the American Heart
Association. Circulation 2010;121(21):2331-78.
15. Qin XD, Qian Z, Vaughn MG, et al. Gender-specific differences of interaction between
obesity and air pollution on stroke and cardiovascular diseases in Chinese adults from a
high pollution range area: A large population based cross sectional study. Sci Total
Environ 2015;529:243-8.
16. Brook RD, Sun Z, Brook JR, et al. Extreme Air Pollution Conditions Adversely Affect
Blood Pressure and Insulin Resistance: The Air Pollution and Cardiometabolic Disease
Study. Hypertension 2016;67(1):77-85.

133

17. Li J, Zhou C, Xu H, et al. Ambient Air Pollution Is Associated With HDL (High-Density
Lipoprotein) Dysfunction in Healthy Adults. Arterioscler Thromb Vasc Biol
2019;39(3):513-22.
18. Weichenthal S, Hoppin JA, Reeves F. Obesity and the cardiovascular health effects of
fine particulate air pollution. Obesity (Silver Spring) 2014;22(7):1580-9.
19. Chen Z, Salam MT, Eckel SP, et al. Chronic effects of air pollution on respiratory health
in Southern California children: findings from the Southern California Children's Health
Study. J Thorac Dis 2015;7(1):46-58.
20. Hoffmann K, Boeing H, Dufour A, et al. Estimating the distribution of usual dietary
intake by short-term measurements. Eur J Clin Nutr 2002;56 Suppl 2:S53-62.
21. Matthews DR, Hosker JP, Rudenski AS, et al. Homeostasis model assessment: insulin
resistance and beta-cell function from fasting plasma glucose and insulin concentrations
in man. Diabetologia 1985;28(7):412-9.
22. Matsuda M, DeFronzo RA. Insulin sensitivity indices obtained from oral glucose
tolerance testing: comparison with the euglycemic insulin clamp. Diabetes Care
1999;22(9):1462-70.
23. Goldberg D. Texas A&M University Geoservices. Available online at
http://geoservices.tamu.edu. 2016. (Accessed June 1 2016).
24. Wong DW, Yuan L, Perlin SA. Comparison of spatial interpolation methods for the
estimation of air quality data. J Expo Anal Environ Epidemiol 2004;14(5):404-15.
25. Benson P. CALINE4 – A Dispersion Model for Predicting Air Pollutant Concentrations
Near Roadways. 1989;Report No. FHWA/CA/TL-84/15.(Prepared by the California
Department of Transportation, Sacramento, CA.).
26. Keskin M, Kurtoglu S, Kendirci M, et al. Homeostasis model assessment is more reliable
than the fasting glucose/insulin ratio and quantitative insulin sensitivity check index for
assessing insulin resistance among obese children and adolescents. Pediatrics
2005;115(4):e500-3.
27. Reinehr T, Andler W. Changes in the atherogenic risk factor profile according to degree
of weight loss. Arch Dis Child 2004;89(5):419-22.
28. Kernan WN, Inzucchi SE, Viscoli CM, et al. Pioglitazone improves insulin sensitivity
among nondiabetic patients with a recent transient ischemic attack or ischemic stroke.
Stroke 2003;34(6):1431-6.
29. Moore JB. Non-alcoholic fatty liver disease: the hepatic consequence of obesity and the
metabolic syndrome. Proc Nutr Soc 2010;69(2):211-20.
30. Tomaru M, Takano H, Inoue K, et al. Pulmonary exposure to diesel exhaust particles
enhances fatty change of the liver in obese diabetic mice. Int J Mol Med 2007;19(1):17-
22.
31. Wu XM, Broadwin R, Basu R, et al. Associations between fine particulate matter and
changes in lipids/lipoproteins among midlife women. Sci Total Environ 2019;654:1179-
86.
32. Bind MA, Peters A, Koutrakis P, et al. Quantile Regression Analysis of the Distributional
Effects of Air Pollution on Blood Pressure, Heart Rate Variability, Blood Lipids, and
Biomarkers of Inflammation in Elderly American Men: The Normative Aging Study.
Environ Health Perspect 2016;124(8):1189-98.
33. Chuang KJ, Yan YH, Cheng TJ. Effect of air pollution on blood pressure, blood lipids,
and blood sugar: a population-based approach. J Occup Environ Med 2010;52(3):258-62.

134

34. Yang BY, Bloom MS, Markevych I, et al. Exposure to ambient air pollution and blood
lipids in adults: The 33 Communities Chinese Health Study. Environ Int 2018;119:485-
92.
35. Poursafa P, Mansourian M, Motlagh ME, et al. Is air quality index associated with
cardiometabolic risk factors in adolescents? The CASPIAN-III Study. Environ Res
2014;134:105-9.
36. Takano H, Yanagisawa R, Inoue K, et al. Nitrogen dioxide air pollution near ambient
levels is an atherogenic risk primarily in obese subjects: a brief communication. Exp Biol
Med (Maywood) 2004;229(4):361-4.
37. Sun Q, Yue P, Deiuliis JA, et al. Ambient air pollution exaggerates adipose inflammation
and insulin resistance in a mouse model of diet-induced obesity. Circulation
2009;119(4):538-46.
38. Gutierrez DA, Puglisi MJ, Hasty AH. Impact of increased adipose tissue mass on
inflammation, insulin resistance, and dyslipidemia. Curr Diab Rep 2009;9(1):26-32.
39. Toledo-Corral CM, Alderete TL, Habre R, et al. Effects of air pollution exposure on
glucose metabolism in Los Angeles minority children. Pediatr Obes 2018;13(1):54-62.
40. Thiering E, Markevych I, Bruske I, et al. Associations of Residential Long-Term Air
Pollution Exposures and Satellite-Derived Greenness with Insulin Resistance in German
Adolescents. Environ Health Perspect 2016;124(8):1291-8.
41. Xu Z, Xu X, Zhong M, et al. Ambient particulate air pollution induces oxidative stress
and alterations of mitochondria and gene expression in brown and white adipose tissues.
Part Fibre Toxicol 2011;8:20.

135

4.7 Tables

Table 4.1 Sociodemographic Characteristics by Obesity Status of 158 Participants Enrolled in
the Meta-AIR Study from 2014-2018.
Total
a

Non-Obese
b
Obese
c
p-value
d
Sex n (%)

n (%) n (%) 0.24
Male 83 (52.5)

56 (56.6) 27 (45.8)

Female 75 (47.5)

43 (43.4) 32 (54.2)

Race/Ethnicity

0.11
White 44 (27.9)

33 (33.3) 11 (18.6)

Hispanic 94 (59.5)

53 (53.5) 41 (69.5)

Other
e
20 (12.7)

13 (13.1) 7 (11.9)

Occupational Status

0.44
Student only 53 (33.5)

33 (33.3) 20 (33.9)

Full or part time work only 32 (20.3)

22 (22.2) 10 (17.0)

Student + full/part time 65 (41.2)

41 (41.4) 24 (40.7)

Unemployed + other 8 (5.1)

3 (3.0) 5 (8.5)

Parental Education

Less than high school 31 (20.3)

20 (20.4) 11 (20.0) 0.28
High school graduate 24 (15.7)

12 (12.2) 12 (21.8)

Some college and beyond 98 (64.1)

66 (67.3) 32 (58.2)

Self-Reported Exercise

0.24
Yes 121 (76.5)

79 (79.8) 42 (71.2)

No 37 (23.4)

20 (20.2) 17 (28.9)

Current Smoker
f

0.49
Yes 9 (5.7)

7 (7.1) 2 (3.4)

No 149 (94.3)

92 (92.9) 57 (96.6)

E-cigarette Use

0.31
Ever 20 (15.4)

11 (12.8) 9 (20.5)

Never 110 (84.6)

75 (87.2) 35 (79.5)

a
Variable denominators may differ due to missing values.
b
Non-Obese= BMI<30.
c
Obese= BMI ≥ 30.
d
Chi-square (non-obese vs obese) p-value.
e
Other races= Asian (n=10), African-American (n=6), Other/Mixed Races=(n=4).
f
Current smoker= smoked more than 20 cigarettes (1 pack) in the past month.

136

Table 4.2 Obesity and Cardiometabolic Measures by Obesity Status in 158 Participants Enrolled in the Meta-AIR Study from 2014-
2018.
Total

Non-Obese
a
Obese
b
Obesity Measures: n Mean SD

n Mean SD n Mean SD
p-value
c

BMI (kg/m
2
) 158 29.9 5.1

99 26.7 2.1 59 35.1 4.2 <0.0001
Total Body fat (%) 158 34.9 8.5

99 31.4 7.8 59 40.9 6.0 <0.0001
Abdominal MRI Measures:

SAAT (L) 152 5.2 2.8

95 3.7 1.6 57 7.7 2.5 <0.0001
VAT (L) 152 1.4 1

95 0.9 0.6 57 2.1 1.2 <0.0001
HFF (%) 152 4.5 3.8

95 3.4 2.7 57 6.6 4.5 <0.0001
Cardiometabolic Measures: n Mean SD

n Mean SD n Mean SD p-value
c

Systolic Blood Pressure (mmHg) 158 115.1 11.2 99 113.3 10.8 59 118.2 11.5 0.009
Diastolic Blood Pressure (mmHg) 158 70.1 7.4 99 68.7 6.9 59 72.6 7.7 0.001
Lipid Metabolism n Mean SD n Mean SD n Mean SD p-value
c

Triglycerides (mg/dL) 145 85.8 47.7

91 75.2 43.0 54 103.5 50.1 0.0004
Total Cholesterol (mg/dL) 145 158.9 37.6

91 157.8 38.1 54 160.8 37.1 0.63
HDL-Cholesterol (mg/dL) 145 39.6 9.7

91 41.3 9.8 54 36.7 9.1 0.006
LDL-Cholesterol (mg/dL) 145 102.2 32.2

91 101.4 32.6 54 103.5 32.0 0.71
VLDL-Cholesterol (mg/dL) 145 17.2 9.5

91 15.0 8.6 54 20.7 10.0 0.0004
Glucose Metabolism n Mean SD n Mean SD n Mean SD p-value
c

Clinical fasting glucose (mg/dL) 154* 90.1 7.6 98 88.8 6.8 56 92.3 8.4 0.006
Glucose 120 min (mg/dL) 154* 120.6 27.0 98 115.6 27.5 56 129.4 24.1 0.002
Glucose AUC 154* 273.8 46.5 98 267.5 47.0 56 285.0 44.0 0.02
Fasting insulin (µU/mL) 141*
†
9.4 9.2 89 6.8 6.6 52 13.8 11.2 0.0001
Insulin AUC 141*
†
212.2 151.0 89 183.5 140.8 52 261.5 156.4 0.0003
HOMA-IR 141*
†
2.1 2.2 89 1.5 1.4 52 3.2 2.8 0.0001
Matsuda Index 141*
†
6.1 5.5 89 7.3 5.8 52 3.9 4.1 <0.0001

137

SD= standard deviation; BMI=body mass index; MRI=magnetic resonance imaging; SAAT= subcutaneous abdominal adipose tissue; VAT=visceral adipose
tissue; HFF= Hepatic (Liver) Fat Fraction; AUC= area under the curve; HOMA-IR= homeostatic model assessment-insulin resistance; HDL= high-density
lipoprotein; LDL= low-density lipoprotein; VLDL= very low-density lipoprotein.
a
Non-Obese= BMI<30 where 24 participants have normal BMI (BMI<25) and 75 participants were overweight (25≤BMI<30).
b
Obese= BMI ≥ 30.
c
T-test (non-obese vs obese) p-value.
*one diabetic subject was removed.
† one subject with high fasting insulin removed.

138

Table 4.3 Short- and Long-Term Regional Ambient and Near-Roadway Air Pollution Exposures
Among 158 Participants Enrolled in the Meta-AIR Study from 2014-2018.
1-Month (Short-term) Exposure
a
Mean SD IQR
Regional Ambient Air Pollutants

NO2 (ppb) 16.1 5.7 12.6-20.0
O3 (ppb) 48.8 14.1 38.5-58.1
PM10 (μg/m
3
) 30.3 9.7 22.8-36.7
PM2.5 (μg/m
3
) 12.4 4.3 9.2-15.2
Near-Roadway Air Pollutants

Freeway NOx (ppb) 5.6 6.4 2.1-6.3
Non-freeway NOx (ppb) 1.7 1.3 0.8-2.2
Total NOx (ppb) 6.9 6.9 3.0-8.0

1-Year (Long-term) Exposure
b

Mean SD IQR
Regional Ambient Air Pollutants

NO2 (ppb) 16.0 3.9 14.3-18.7
O3 (ppb) 48.7 6.5 42.2-53.3
PM10 (μg/m
3
) 30.9 7.9 26.1-35.0
PM2.5 (μg/m
3
) 12.4 2.5 10.3-14.5
Near-Roadway Air Pollutants

Freeway NOx (ppb) 6.0 6.3 2.0-7.9
Non-freeway NOx (ppb) 1.8 1.4 0.8-2.3
Total NOx (ppb) 7.3 6.9 3.3-9.3
SD=standard deviation; IQR=interquartile range; NO 2= nitrogen dioxide; O 3=ozone 8-hour maximum daily; PM 10=
particulate matter with aerodynamic diameter <10 μm; PM 2.5=particulate matter with aerodynamic diameter <2.5
μm; ppb=parts per billion; NO x=nitrogen oxides.
a
1-month average air pollution exposure prior to the Meta-AIR visit date.
b
1-year average air pollution exposure prior to the Meta-AIR visit date.

139

Table 4.4 Associations
a
of Short- and Long-Term Ambient Air Pollution Exposures with Obesity Measures in 158 Participants
Enrolled in the Meta-AIR Study from 2014-2018.
Short-Term Exposures
Obesity Measures
1-Month NO2 1-Month O3 1-Month PM10 1-Month PM2.5
Estimate
b
p-value Estimate
b
p-value Estimate
b
p-value Estimate
b
p-value
BMI 0.64 0.16 0.93 0.07 0.30 0.42 -0.02 0.96
Total Body Fat Percent -0.66 0.43 -0.02 0.98 -0.024 0.97 0.41 0.57
SAAT -0.005 0.63 0.008 0.52 0.004 0.70 0.007 0.47
VAT 0.002 0.64 0.007 0.19 0.003 0.50 0.003 0.51
HFF
†
0.11 0.14 0.18 0.02* 0.098 0.12 0.06 0.37
Long Term Exposures
Obesity Measures
1-Year NO2 1-Year O3 1-Year PM10 1-Year PM2.5
Estimate
b
p-value Estimate
b
p-value Estimate
b
p-value Estimate
b
p-value
BMI 0.34 0.44 -0.062 0.90 -0.52 0.22 -0.34 0.38
Total Body Fat Percent -1.49 0.06 -0.91 0.31 0.039 0.96 -0.24 0.74
SAAT -0.013 0.21 -0.004 0.73 0.006 0.58 0.001 0.94
VAT -0.007 0.10 -0.003 0.61 -0.003 0.43 -0.002 0.57
HFF
†
0.044 0.53 0.024 0.76 0.063 0.36 0.068 0.29
NO 2= nitrogen dioxide; O 3=ozone 8-hour maximum daily; PM 10= particulate matter with aerodynamic diameter <10 μm; PM 2.5=particulate matter with
aerodynamic diameter <2.5 μm; BMI=body mass index; SAAT= subcutaneous abdominal adipose tissue volume/ total abdominal volume; VAT=visceral adipose
tissue volume/ total abdominal volume; HFF= hepatic fat fraction.
a
Associations reflect change in outcome measure (association estimate (β)) scaled to 1 standard deviation of prior 1-month average ambient NO 2 with 5.7 ppb, O 3
with 14.1 ppb, PM 10 with 9.7 μg/m
3
, and PM 2.5 with 4.3 μg/m
3
OR prior 1-year average ambient NO 2 with 3.9 ppb, O 3 with 6.5 ppb, PM 10 with 7.9 μg/m
3
, and
PM 2.5 with 2.5 μg/m
3
.
b
Linear regression model was used to estimate the associations of prior 1-month or 1-year NO 2, O 3, PM 10, and PM 2.5 average exposures with obesity related
outcomes after adjusting for age, sex, race/ethnicity, occupational status of subject, parental education, self-reported exercise, current cigarette smoking, e-
cigarette use (ever/never), total body fat % (not included in SAAT, VAT, HFF models), diet (total calories), season (warm/cool), and historic air pollution
exposure.
†
log-transformed variable.
*p<0.05.

140

Table 4.5 Associations
a
of Short-Term Ambient Air Pollution Exposures with Cardiometabolic Measures in 158 Participants Enrolled
in the Meta-AIR Study from 2014-2018.
1-Month NO2 1-Month O3 1-Month PM10 1-Month PM2.5
Cardiometabolic
Measures
Estimate
b
p-value Estimate
b
p-value Estimate
b
p-value Estimate
b
p-value
Systolic Blood Pressure 3.66 0.003* 0.37 0.79 2.16 0.03* 2.85 0.009*
Diastolic Blood Pressure 1.79 0.06 0.62 0.57 1.41 0.07 2.46 0.003*
Lipid Metabolism Estimate
b
p-value Estimate
b
p-value Estimate
b
p-value Estimate
b
p-value
Triglycerides
†
0.059 0.40 0.17 0.04* 0.058 0.33 0.083 0.19
Total Cholesterol 5.98 0.27 -3.067 0.61 3.049 0.48 5.92 0.21
HDL-Cholesterol 0.29 0.82 -3.017 0.03* -0.41 0.69 0.43 0.69
LDL-Cholesterol 5.22 0.25 -3.36 0.51 2.43 0.51 3.63 0.36
VLDL-Cholesterol
†
0.059 0.40 0.17 0.04* 0.058 0.33 0.083 0.19
Glucose Metabolism Estimate
b
p-value Estimate
b
p-value Estimate
b
p-value Estimate
b
p-value
Fasting Glucose 0.19 0.85 -0.45 0.69 -0.51 0.52 -0.85 0.32
Glucose 120 min. 1.57 0.63 1.62 0.66 0.51 0.85 2.00 0.48
Glucose AUC 11.90 0.04* -0.73 0.91 2.78 0.55 5.72 0.26
Fasting Insulin
†
0.18 0.22 0.11 0.52 0.023 0.85 -0.19 0.14
Insulin AUC 30.1 0.08 6.02 0.76 12.99 0.35 22.21 0.14
HOMA-IR
†
0.18 0.23 0.01 0.56 0.017 0.89 -0.2 0.12
Matsuda Index
†
-0.15 0.13 -0.037 0.74 -0.037 0.65 0.059 0.51
NO 2= nitrogen dioxide; O 3=ozone 8-hour maximum daily; PM 10= particulate matter with aerodynamic diameter <10 μm; PM 2.5=particulate matter with
aerodynamic diameter <2.5 μm; HDL= high-density lipoprotein; LDL= low-density lipoprotein; VLDL= very low-density lipoprotein; AUC= area under the
curve; HOMA-IR= Homeostatic model assessment-insulin resistance.
a
Associations reflect change in outcome measure (effect estimate (β)) scaled to 1 standard deviation of prior 1-month average ambient NO 2 with 5.7 ppb, O 3 with
14.1 ppb, PM 10 with 9.7 μg/m
3
, and PM 2.5 with 4.3 μg/m
3
.
b
Linear regression model used to estimate the associations of prior 1-month NO 2, O 3, PM 10, and PM 2.5 exposures with glucose- and lipid-related measures after
adjusting for age, sex, race/ethnicity, occupational status of participant, parental education, self-reported exercise, current cigarette use, e-cigarette use
(ever/never), total body fat %, diet (total calories), season (warm/cool), and childhood air pollution exposure.
†
log-transformed variable.
*p<0.05.

141

Table 4.6 Associations
a
of Long-Term Regional Ambient Air Pollution Exposures with Cardiometabolic Measures in 158 Participants
Enrolled in the Meta-AIR Study from 2014-2018.
1-Year NO2 1-Year O3 1-Year PM10 1-Year PM2.5
Cardiometabolic
Measures
Estimate
b
p-value Estimate
b
p-value Estimate
b
p-value Estimate
b
p-value
Systolic Blood Pressure 2.20 0.08 0.17 0.91 -0.08 0.95 1.05 0.34
Diastolic Blood Pressure 1.74 0.07 -1.20 0.25 -0.76 0.39 1.31 0.11
Lipid Metabolism Estimate
b
p-value Estimate
b
p-value Estimate
b
p-value Estimate
b
p-value
Triglycerides
†
0.11 0.11 0.085 0.35 0.022 0.78 0.043 0.51
Total Cholesterol 11.25 0.04* 4.23 0.54 8.89 0.13 1.39 0.78
HDL-Cholesterol 0.12 0.93 -1.27 0.43 0.63 0.65 -0.72 0.53
LDL-Cholesterol 9.37 0.04* 3.98 0.49 7.48 0.14 0.82 0.84
VLDL-Cholesterol
†
0.11 0.11 0.085 0.35 0.022 0.78 0.043 0.51
Glucose Metabolism Estimate
b
p-value Estimate
b
p-value Estimate
b
p-value Estimate
b
p-value
Fasting Glucose -0.10 0.92 -0.81 0.51 -1.49 0.12 -0.031 0.97
Glucose 120 min. 1.09 0.74 2.80 0.50 -0.33 0.92 1.24 0.67
Glucose AUC 7.45 0.21 4.37 0.56 -2.36 0.69 5.52 0.29
Fasting Insulin
†
0.17 0.26 -0.049 0.79 -0.21 0.18 -0.13 0.34
Insulin AUC 29.93 0.09 37.50 0.08 20.51 0.27 33.57 0.03*
HOMA-IR
†
0.17 0.26 -0.054 0.78 -0.23 0.16 -0.12 0.36
Matsuda Index
†
-0.16 0.14 -0.064 0.63 0.064 0.57 -0.018 0.85
NO 2= nitrogen dioxide; O 3=ozone 8-hour maximum daily; PM 10= particulate matter with aerodynamic diameter <10 μm; PM 2.5=particulate matter with
aerodynamic diameter <2.5 μm; HDL= high-density lipoprotein; LDL= low-density lipoprotein; VLDL= very low-density lipoprotein; AUC= area under the
curve; HOMA-IR= homeostatic model assessment-insulin resistance.
a
Associations reflect change in outcome measure (effect estimate (β)) scaled to 1 standard deviation of prior 1-year average ambient NO 2 with 3.9 ppb, O 3 with
6.5 ppb, PM 10 with 7.9 μg/m
3
, and PM 2.5 with 2.5 μg/m
3
.
b
Linear regression model was used to estimate the associations of prior 1-year NO 2, O 3, PM 10, and PM 2.5 exposures with glucose- and lipid-related measures after
adjusting for age, sex, race/ethnicity, occupational status of participant, parental education, self-reported exercise, current cigarette use, e-cigarette use
(ever/never), total body fat %, diet (total calories), season (warm/cool), and historic air pollution exposure.
†
log-transformed variable.
*p<0.05.

142

4.8 Figures
Figure 4.1 Meta-AIR Study
a
Flow.

a
Meta-AIR subjects were recruited between 2014-2018 from Children’s Health Study to examine the effects of
short- and long-term ambient and near-roadway air pollution exposures on obesity and cardiometabolic health in
young adults.

143

Figure 4.2 Associations
a
of Prior 1-Year

NO2 Exposures and Lipid Metabolism Measures by
Obesity Status in 158 Participants Enrolled in the Meta-AIR Study from 2014-2018.

NO 2= nitrogen dioxide; HDL= high-density lipoprotein; LDL= low-density lipoprotein; VLDL= very low-density
lipoprotein; obese= BMI ≥ 30.0 kg/m
2
, non-obese= BMI< 30.0 kg/m
2
.

a
Associations reflect change in outcome measure (association estimate (β)) scaled to 1 standard deviation of prior 1-
year average ambient NO 2 with 3.9 ppb stratified by obesity status (non-obese vs obese).
b
Linear regression model was used to estimate the associations of 1-year NO 2 and lipid metabolism outcomes after
adjusting for age, sex, race/ethnicity, occupational status of subject, parental education, self-reported exercise,
current cigarette smoking, e-cigarette use, total body fat percent, diet, season, and historic air pollution exposure.
*p<0.05

144

4.9 Supplemental Tables

Supplemental Table 4.1 Historic Regional Ambient and Near-Roadway Air Pollution
Exposures Among 158 Participants Enrolled in the Meta-AIR Study from 2014-2018.
Historic Air Pollution Exposure Mean SD IQR
Ambient Air Pollutants
a

NO2 (ppb) 25.7 4.6 23.2-29.7
Ozone 8 Hr Max (ppb) 42.8 5.4 38.1-45.1
PM10 (μg/m
3
)
41.9 7.7 34.5-46.4
PM2.5 (μg/m
3
)
18.5 2.7 16.7-20.5

Near-Roadway Air Pollutants
b

Freeway NOx (ppb) 21.5 24.2 7.3-26.7
Non-freeway NOx (ppb) 5.2 3.5 2.6-7.2
Total NOx (ppb) 26.7 25.5 11.3-34.4
SD=standard deviation; IQR=interquartile range; NO 2= nitrogen dioxide; O 3=ozone 8-hour maximum daily; PM 10=
particulate matter with aerodynamic diameter <10 μm; PM 2.5=particulate matter with aerodynamic diameter <2.5
μm; ppb=parts per billion; NO x=nitrogen oxides; ppb=parts per billion.
a
Historic, ambient exposure is a life-time average of air pollution exposure from birth through year 2011.
b
Historic, near-roadway exposure is a life-time average of air pollution exposure from CHS study entry (5/2003)
through year 2011.

145

Supplemental Table 4.2 Spearman Correlations Between Prior 1-Month and 1-Year Average
Air Pollution Exposures with Historic Air Pollution Exposures.
Prior 1-Month vs Historic
Correlation
Coefficient
p-value
Ambient Pollutants
a

NO2 0.33 <0.0001
O3 0.23 0.006
PM10 0.44 <0.0001
PM2.5 0.49 <0.0001
Near-Roadway Pollutants
b

Freeway NOx 0.60 <0.0001
Non-Freeway NOx 0.71 <0.0001
Total NOx 0.72 <0.0001
Prior 1-Year vs Historic
Correlation
Coefficient
p-value
Ambient Pollutants
a

NO2 0.38 <0.0001
O3 0.67 <0.0001
PM10 0.46 <0.0001
PM2.5 0.54 <0.0001
Near-Roadway Pollutants
b

Freeway NOx 0.66 <0.0001
Non-Freeway NOx 0.76 <0.0001
Total NOx 0.81 <0.0001
NO 2= nitrogen dioxide; O 3=ozone 8-hour maximum daily; PM 10= particulate matter with aerodynamic diameter <10
μm; PM 2.5=particulate matter with aerodynamic diameter <2.5 μm; NO x=nitrogen oxides.
a
Historic exposure is a life-time average of air pollution exposure from birth through year 2011.
b
Historic exposure is a life-time average of air pollution exposure from CHS study entry (5/2003) through year
2011.

146

Supplemental Table 4.3 Spearman Correlations Between Short-Term Ambient and Near-
Roadway Air Pollutants.
Air Pollutants
1-Month Average Exposure
Freeway
NOx
Non-
freeway
NOx
Total
NOx
NO2 O3 PM10 PM2.5
1-Month Average
Exposure
Freeway NOx 1

Non-freeway NOx 0.18* 1

Total NOx 0.81* 0.52* 1

NO2 0.31* 0.17 0.39* 1

O3 -0.15 -0.32* -0.2* -0.35* 1

PM10 -0.11 -0.30* -0.15 0.13 0.44* 1
PM2.5 -0.04 -0.27* -0.1 0.34* 0.34* 0.68* 1
NO x= nitrogen oxides; NO 2= nitrogen dioxide; O 3=ozone 8-hour maximum daily; PM 10= particulate matter with
aerodynamic diameter <10 μm; PM 2.5=particulate matter with aerodynamic diameter <2.5 μm.
*p<0.05

147

Supplemental Table 4.4 Spearman Correlations Between Long-Term Ambient and Near-
Roadway Air Pollutants.
NO x= nitrogen oxides; NO 2= nitrogen dioxide; O 3=ozone 8-hour maximum daily; PM 10= particulate matter with
aerodynamic diameter <10 μm; PM 2.5=particulate matter with aerodynamic diameter <2.5 μm; *p<0.05

Air Pollutants
1-Year Average Exposure
Freeway
NOx
Non-
freeway
NOx
Total
NOx
NO2 O3 PM10 PM2.5
1-Year Average
Exposure
Freeway NOx 1

Non-freeway NOx 0.22* 1

Total NOx 0.82* 0.54* 1

NO2 0.32* 0.31* 0.46* 1

O3 -0.22* -0.56* -0.4* -0.13 1

PM10 -0.14 -0.31* -0.27* -0.13 0.58* 1
PM2.5 -0.06 -0.32* -0.13 0.32* 0.62* 0.5* 1

148

Supplemental Table 4.5 Associations
a
of Short-Term Ozone (O3) Exposures with Liver Fat (HFF) and Lipid Measures in
Multipollutant Models.
Short-term Exposures

1-Month O3
1-Month O3 +
1-Month NO2
1-Month O3 +
1-Month PM2.5
Obesity Measures: Estimate
b
p-value Estimate
b
p-value Estimate
b
p-value
HFF
†
0.18 0.02* 0.17 0.03* 0.17 0.048*
Cardiometabolic Measures:
Lipid Metabolism
Estimate
b
p-value Estimate
b
p-value Estimate
b
p-value
Triglycerides
†
0.17 0.04* 0.16 0.049* 0.14 0.09
HDL-Cholesterol -3.017 0.03* -3.03 0.03* -3.32 0.02*
VLDL-Cholesterol
†
0.17 0.04* 0.16 0.049* 0.14 0.09
O 3=ozone 8-hour maximum daily; NO 2= nitrogen dioxide; PM 2.5=particulate matter with aerodynamic diameter <2.5 μm; HFF=hepatic fat fraction; HDL=high
density lipoproteins, VLDL= very low-density lipoprotein.

a
Associations reflect change in outcome measure (effect estimate (β)) scaled to 1 standard deviation of prior 1-month average ambient O 3 with 14.1 ppb.
b
Linear regression model was used to estimate the associations of 1-month O 3 and HFF or lipid-related outcomes in multipollutant models after adjusting for
age, sex, race/ethnicity, occupational status of subject, parental education, self-reported exercise, current cigarette smoking, e-cigarette use (ever/never), total
body fat % (not included in HFF model), diet (total calories), season (warm/cool), and historic air pollution exposure.
†
log-transformed variable.
*p<0.05.

149

Supplemental Table 4.6 Associations
a
of Long-Term NO2 Exposures with Lipid Measures in Multipollutant Models.
Long-term Exposures 1-Year NO2
1-Year NO2 +
1-Year O3
1-Year NO2 +
1-Year PM2.5
Cardiometabolic Measures:
Lipid Metabolism
Estimate
b
p-value Estimate
b
p-value Estimate
b
p-value
Total Cholesterol 11.3 0.04* 11.0 0.05 12.8 0.04*
LDL-Cholesterol

9.4 0.04* 9.03 0.06 10.5 0.047*
NO 2= nitrogen dioxide; O 3=ozone 8-hour maximum daily; PM 2.5=particulate matter with aerodynamic diameter <2.5 μm; LDL=low density lipoprotein.

a
Associations reflect change in outcome measure (effect estimate (β)) scaled to 1 standard deviation of prior 1-year average ambient NO 2 with 3.9 ppb.
b
Linear regression model was used to estimate the associations of 1-year NO 2 and total cholesterol or LDL-cholesterol in multipollutant models after adjusting
for age, sex, race/ethnicity, occupational status of subject, parental education, self-reported exercise, current cigarette smoking, e-cigarette use (ever/never), total
body fat %, diet (total calories), season (warm/cool), and historic air pollution exposure.
*p<0.05

150

Supplemental Table 4.7 Association
a
of Long-Term PM2.5 Exposures with Insulin AUC in Multipollutant Models.
Long-term Exposures 1-Year PM2.5
1-Year PM2.5 +
1-Year NO2
1-Year PM2.5 +
1-Year O3
Cardiometabolic Measures:
Glucose Metabolism
Estimate
b
p-value Estimate
b
p-value Estimate
b
p-value
Insulin AUC 33.6 0.03* 30.3 0.09 24.7 0.17
PM 2.5=particulate matter with aerodynamic diameter <2.5 μm;NO 2= nitrogen dioxide; O 3=ozone 8-hour maximum daily; AUC=area under the curve.

a
Associations reflect change in outcome measure (effect estimate (β)) scaled to 1 standard deviation of prior 1-year average ambient PM 2.5 with 2.5 μg/m
3
.
b
Linear regression model was used to estimate the associations of 1-year PM 2.5 and insulin AUC in multipollutant models after adjusting for age, sex,
race/ethnicity, occupational status of subject, parental education, self-reported exercise, current cigarette smoking, e-cigarette use (ever/never), total body fat %,
diet (total calories), season (warm/cool), and historic air pollution exposure.
*p<0.05

151

Supplemental Table 4.8 Associations
a
of Short-Term NRAP Exposures with Obesity and Cardiometabolic Measures in 158
Participants Enrolled in the Meta-AIR Study from 2014-2018.

NO x= nitrogen oxides; BMI=body mass index; SAAT= subcutaneous abdominal adipose tissue volume/ total abdominal volume; VAT=visceral adipose
tissue volume/ total abdominal volume; HFF= Liver Fat Fraction; AUC= area under the curve; HOMA-IR= Homeostatic model assessment-insulin resistance;
LDL= low-density lipoprotein; HDL= high-density lipoprotein; VLDL= very low-density lipoprotein.
a
Associations reflect change in outcome measure (effect estimate (β)) scaled to 1 standard deviation of prior 1-month average near-roadway freeway NOx with
6.4 ppb, non-freeway NOx with 1.3 ppb and total NOx with 6.9 ppb.

1-Month
Non-Freeway NOx
1-Month
Freeway NOx
b

1-Month
Total NOx
Obesity Measures: Estimate
c
p-value Estimate
c
p-value Estimate
c
p-value
BMI 0.22 0.54 -0.6 0.15 -0.31 0.40
Body Fat (%) -0.61 0.34 -0.52 0.49 -0.089 0.90
SAAT -0.013 0.13 -0.01 0.30 -0.003 0.75
VAT -0.005 0.14 -0.001 0.86 -0.002 0.67
HFF
†
-0.094 0.11 -0.086 0.19 -0.025 0.68
Cardiometabolic Measures: Estimate
c
p-value Estimate
c
p-value Estimate
c
p-value
Systolic Blood Pressure -0.47 0.64 0.87 0.46 0.15 0.89
Diastolic Blood Pressure -0.51 0.49 0.21 0.81 -0.16 0.84
Lipid Metabolism Estimate
c
p-value Estimate
c
p-value Estimate
c
p-value
Triglycerides
†
-0.097 0.07 -0.07 0.28 -0.075 0.21
Total Cholesterol -4.21 0.29 5.19 0.27 2.48 0.58
HDL-Cholesterol -0.92 0.34 0.78 0.50 0.5 0.64
LDL-Cholesterol -1.58 0.64 5.56 0.17 3.17 0.41
VLDL-Cholesterol
†
-0.097 0.07 -0.07 0.28 -0.075 0.21
Glucose Metabolism Estimate
b
p-value Estimate
b
p-value Estimate
b
p-value
Fasting Glucose -0.57 0.46 0.47 0.59 0.031 0.97
Glucose 120 min. -0.19 0.94 -0.83 0.78 -0.69 0.80
Glucose AUC 2.64 0.57 2.30 0.67 2.45 0.62
Fasting Insulin
†
-0.19 0.07 -0.024 0.85 0.063 0.61
Insulin AUC 7.18 0.59 0.75 0.96 -0.049 1.00
HOMA-IR
†
-0.21 0.06 -0.015 0.91 0.068 0.59
Matsuda Index
†
0.097 0.21 -0.016 0.87 -0.048 0.59

152

b
Freeway NO x models include historic exposure which is the average childhood NRAP exposure from (5/2003 through 2011).
c
Linear regression model was used to estimate the associations of 1-month non-freeway, freeway and total NO x with obesity- and cardiometabolic-related
outcomes after adjusting for age, sex, race/ethnicity, occupational status of subject, parental education, self-reported exercise, current cigarette smoking, e-
cigarette use (ever/never), total body fat % (not included in SAAT, VAT, HFF), diet (total calories), season (warm/cool), and historic air pollution exposure (only
for freeway NO x).
†
log-transformed variable.

153

Supplemental Table 4.9 Associations
a
of Long-Term NRAP Exposures with Obesity and Cardiometabolic Measures in 158
Participants Enrolled in the Meta-AIR Study from 2014-2018.

1-Year
Non-Freeway NOx
1-Year
Freeway NOx
b

1-Year
Total NOx
Obesity Measures: Estimate
c
p-value Estimate
c
p-value Estimate
c
p-value
BMI -0.42 0.28 -0.51 0.25 -0.20 0.59
Body Fat (%) 0.84 0.23 -1.23 0.12 -0.29 0.67
SAAT 0.005 0.57 -0.02 0.06 -0.006 0.50
VAT 0.001 0.74 -0.001 0.87 -0.001 0.74
HFF
†
-0.006 0.92 -0.11 0.14 -0.01 0.87
Cardiometabolic Measures: Estimate
c
p-value Estimate
c
p-value Estimate
c
p-value
Systolic Blood Pressure -0.34 0.76 0.54 0.67 -0.20 0.85
Diastolic Blood Pressure -1.28 0.12 0.49 0.60 0.05 0.95
Lipid Metabolism Estimate
c
p-value Estimate
c
p-value Estimate
c
p-value
Triglycerides
†
-0.12 0.05 -0.075 0.29 -0.058 0.34
Total Cholesterol -7.03 0.12 3.56 0.51 -0.16 0.97
HDL-Cholesterol -0.017 0.99 0.47 0.72 -0.16 0.89
LDL-Cholesterol -5.11 0.19 3.97 0.38 0.59 0.88
VLDL-Cholesterol
†
-0.12 0.06 -0.075 0.29 -0.058 0.34
Glucose Metabolism Estimate
c
p-value Estimate
c
p-value Estimate
c
p-value
Fasting Glucose 0.82 0.39 0.56 0.56 0.21 0.80
Glucose 120 min. -2.60 0.35 -1.53 0.63 -0.80 0.77
Glucose AUC -6.76 0.19 0.50 0.93 2.64 0.59
Fasting Insulin
†
-0.25 0.07 0.061 0.67 0.12 0.33
Insulin AUC 3.98 0.82 -4.56 0.79 -2.073 0.89
HOMA-IR
†
-0.24 0.09 0.072 0.62 0.13 0.31
Matsuda Index
†
0.13 0.2 -0.047 0.64 -0.079 0.37
NO x= nitrogen oxides; BMI=body mass index; SAAT= subcutaneous abdominal adipose tissue volume/ total abdominal volume; VAT=visceral adipose tissue
volume/ total abdominal volume; HFF= Liver Fat Fraction; AUC= area under the curve; HOMA-IR= Homeostatic model assessment-insulin resistance; LDL=
low-density lipoprotein; HDL= high-density lipoprotein; VLDL= very low-density lipoprotein.
a
Associations reflect change in outcome measure (effect estimate (β)) scaled to 1 standard deviation of prior 1-year average near-roadway freeway NOx with 6.3
ppb, non-freeway NOx with 1.4 ppb and total NOx with 6.9 ppb.
b
Freeway NOx models include historic exposure which is the average childhood NRAP exposure from (5/2003 through 2011).

154

c
Linear regression model was used to estimate the associations of 1-year non-freeway, freeway and total NO x with obesity- and cardiometabolic-related
outcomes after adjusting for age, sex, race/ethnicity, occupational status of participant, parental education, self-reported exercise, current cigarette use, e-cigarette
use (ever/never), total body fat % (not included in SAAT, VAT, HFF), diet (total calories), season (warm/cool), and historic air pollution exposure (only for
freeway NO x).
†
log-transformed variable.

155

Chapter 5: Summary and Future Research
Over the past several decades, deleterious health effects related to air pollution have been
highlighted through epidemiological studies (1). Though it is well known that air pollution
affects respiratory (2-4) and cardiovascular (5-7) systems, others like metabolic systems (8, 9)
may also be perturbed. Factors that increase childhood obesity risk are of great concern as
obesity in childhood persists into adulthood (10) which in turn may result in adverse health
consequences. Recently, growing evidence from animal (11-13) and human (14-17) studies has
shown the negative effects of air pollution on obesity. However, there are many gaps in the
current literature that need to be addressed. First, epidemiological studies looking at the
longitudinal association of childhood obesity and air pollution exposure during critical
developmental periods like in utero and first year of life are lacking. Next, there have been no
studies that have examined gene-environment interactions of genetic determinants of obesity and
early life air pollution exposure on childhood obesity risk. Furthermore, studies showing short-
and long-term effects of air pollution on adiposity and cardiometabolic factors are needed. The
goal of this dissertation was to bridge the gaps in literature while leveraging data from a well-
established cohort of children in the Children’s Health Study (CHS). In the following section,
summaries of the dissertation projects are presented, followed by a brief discussion on future
research that may build on this work and further our understanding of air pollution and childhood
obesity associations.

5.1 Summary
The first project (chapter 2) evaluated the longitudinal association of early life air
pollution exposures on childhood obesity by using linear mixed effects models to estimate the

156

effect of in utero or first year of life near-road air pollution (NRAP) exposure on childhood body
mass index (BMI) trajectory. A 2 standard deviation (SD) difference in the first year of life
freeway near-road nitrogen oxide (NOx) exposure was associated with a 0.1 kg/m
2
(95%
confidence interval (CI): 0.03-0.2) faster increase in BMI growth per year and a 0.5 kg/m
2
(95%
CI: 0.02-0.9) higher attained BMI at age 10 years independent of confounders and later
childhood NRAP exposures. For in utero exposures, a 2-SD difference in freeway NOx was
associated with faster increase in BMI growth and higher BMI at age 10 years however, these
estimates did not reach statistical significance after adjusting for confounders. Though in utero
exposures were not statistically significant, a smaller sample size in the in utero (n=2072) model
versus first year of life (n=2318) model may have produced wider confidence intervals therefore
not reaching statistical significance of p<0.05. Overall, these findings suggest that higher early
life NRAP exposures increases childhood BMI trajectory that results in higher attained BMI at
10 years of age which may in turn contribute to increased obesity risk.
The second project (chapter 3) was an extension on the first project’s linear mixed model.
As statistically significant associations with first year of life freeway NOx and childhood BMI
growth were found in project 1, further investigation of genetic determinants of obesity
potentially modifying these responses to the environment (i.e. air pollution exposures) through
gene-environment interactions was warranted. A genome-wide association studies (GWAS)
derived, obesity related genetic risk score (GRS) was created to evaluate a GRS-environment
(GRSxE) interaction on the longitudinal association of early life NRAP exposure and childhood
BMI growth. First, the association of the obesity related GRS on childhood BMI trajectory
without any air pollution measures was examined. The GRS was statistically significantly
associated with attained BMI at age 10 years as well as BMI growth after adjusting for age, sex,

157

race/ethnicity, parental education, and Spanish questionnaire. An increase of 1 risk allele in
unweighted GRS was significantly associated with 0.006 kg/m
2
faster increase in BMI per year
(p=0.02) which resulted in a 0.07 kg/m
2
higher BMI at age 10 years (p<0.0001). In simpler
terms, children in the 90
th
percentile of GRS (i.e. the highest 10%) compared to those children in
the 10
th
percentile (i.e. lowest 10%) yielded a 0.27 kg/m
2
faster increase in BMI growth and a 2.6
kg/m
2
higher attained BMI at age 10 years, p<0.0001 and p=0.002 respectively. In the full model
with early life exposures of either in utero or first year of life NRAP, GRS and GRSxE, there
were no significant associations with GRSxE and BMI growth or attained BMI at age 10 years.
Though the GRSxE interaction was not statistically significant in the models, GRS was
statistically significantly associated with BMI growth and attained BMI at age 10 years given
average early life NRAP exposure (all p<0.05).
Lastly, in the third project (chapter 4), short- and long-term effects of air pollution on
obesity and cardiometabolic health in young adults was explored. The Meta-AIR Study
examined the associations of prior 1-month (short-term) and prior 1-year (long-term) average
NRAP and ambient air pollution exposures on various measures related to adiposity, glucose
metabolism and lipid metabolism. Several associations with short- and long-term ambient air
pollutants were found. First, a 1-SD change in long-term NO2 exposure was associated with a
11.2 mg/dL higher level of total cholesterol (p=0.04) and 9.4 mg/dL higher level of low-density
lipoproteins (LDL)-cholesterol (p=0.04). Among obese participants, associations between long-
term NO2 and total cholesterol and LDL-cholesterol were magnified compared to associations in
non-obese participants (pinteraction=0.008 and 0.03, respectively). Other notable associations found
were increased short-term O3 exposure and higher triglyceride and very-low-density lipoprotein
cholesterol levels as well as lower high-density lipoprotein cholesterol levels; short-term O3

158

exposure was also associated with higher hepatic fat levels; lastly, long-term PM2.5 exposure was
associated with higher levels of insulin area under the curve. This study shows that both short-
and long-term ambient pollution exposures may potentially alter lipid metabolism in young
adults.
Taken together, all three projects of this dissertation have contributed to the literature of
air pollution and childhood obesity. Our first study showed that early life NRAP exposures may
alter childhood BMI growth particularly those associated with first year of life further
reinforcing the handful of studies that have shown associations of early life air pollution
exposure and increased childhood obesity risk. Along with environmental risk, our data showed
that genetic susceptibility to obesity was also associated with altered childhood BMI growth;
however, our data did not support any gene-environment interactions. Lastly, the final project
showed associations across both short- and long-term exposures to ambient pollutants and
adverse lipid and hepatic fat levels in young adults. This data suggests that recent exposures
(prior month and prior year) to air pollutants may alter lipid metabolism pathways which
downstream may potentially affect the hepatic fat levels. This dissertation work has shown that
air pollution exposures during different periods of the life course have adverse consequences on
human health stemming from first year of life exposures that may adversely alter childhood BMI
trajectories as well as more recent exposures that may modify lipid profiles in young adults
particularly those who are currently obese.

5.2 Future Direction
Collectively the three projects from this dissertation addressed several gaps in the current
literature on air pollution and childhood obesity, but it has also brought to light the need for

159

continued research in this area. Additional studies in early life air pollution exposures and
childhood obesity and health are needed as these are critical periods of development that are
highly susceptible to environmental influences. As in any air pollution study, more accurate
exposure measures that reflect time spent in multiple locations is needed, though this may be
difficult to scale in large epidemiological studies. Here are several ideas for future studies that
may build on this dissertation work.
The first project’s longitudinal model can be extended to later childhood years as the
CHS has continued to take height and weight measurements at follow up visits. However, instead
of using a linear mixed model, splines must be considered to account for non-linear growth
during puberty and beyond. Additional effects of childhood air pollution exposures versus early
life exposures can also be explored in this cohort. Moreover, replication of these associations can
be performed using the new Maternal and Developmental Risks from Environmental and Social
Stressors (MADRES) cohort where pregnant mothers and their infants are currently being
followed. Since MADRES is a pregnancy cohort, there should be better measures of pregnancy-
and postnatal-related obesity risk factors (i.e. social stressors, maternal gestational weight gain,
infant birth weights, breastfeeding practices, etc.) that can be explored in the longitudinal model.
Aside from replication studies, intervention studies are needed to establish causal
pathways and mechanisms by which air pollution exposures increases childhood obesity risk.
Several animal models have shown that early life exposures to air pollution result in enlarged
adipose tissues and increased weight gain in rodent models primarily through inflammatory or
oxidative stress (11-13) pathways. These animal studies showed increased toll-like receptor
activation and lipid oxidation in the lungs which resulted in systemic inflammation (13),
neuroinflammation through increased microglial activation (11), and increased macrophages

160

with increased expression of proinflammatory genes in visceral adipose tissue (12) from
increased air pollution exposures. In human studies, using a birth cohort and looking at pro-
inflammatory markers (i.e. IL-6, TNF-α, CRP, etc.) and oxidative stress markers (i.e. MDA,
MPO, OxLDL, etc.) in samples of cord blood and or blood draws at birth may indicate early
signs of systemic inflammation and oxidative stress. Additionally, using a similar school-based
model of CHS, children across Los Angeles County can be recruited by air pollution exposure
status by selecting schools that are close to freeways and major roads as well as schools that are
farther from freeways and major roads. Blood draws can be made in children from respective
schools and the levels of pro-inflammatory markers and oxidative stress markers can be
compared to BMI measures.
The full GRSxE model in the second project was limited to a smaller sample size of CHS
participants. Though there were no significant associations with the interaction of genetic risk
and early life air pollution exposure, effects of childhood NRAP exposures and GRS on
childhood obesity in a larger group of children may be examined. Due to the statistical model
design with several different exposure windows (in utero, first year of life and childhood
exposures), the first and second project were limited to movers to avoid collinearity of exposures
from early life and childhood. However, childhood air pollution exposures may be examined in a
larger cohort of children which would include both movers and non-movers with genetic data.
Additionally, this model can similarly be replicated in the MADRES cohort where biological
samples as well as air pollution exposures are currently being collected.
Lastly, the third project of Meta-AIR rigorously collected adiposity measures alongside
glucose metabolism and lipid profiles, however these were only collected at one time point. It
would be interesting to see if changes in short- or long-term air pollution exposure elicit changes

161

in adiposity, glucose metabolism and lipid profiles over time. This age range of young adults is a
unique population as disease onset may be potentially be observed if these Meta-AIR
participants are followed into adulthood. Obesity and air pollution may work synergistically as
seen in the Meta-AIR study where effects of air pollution were stronger in obese versus non-
obese participants on lipid measures. Results from these additional projects could further
highlight the importance of the deleterious effects of air pollution on obesity and cardiometabolic
health. To further understand the associations of air pollution and lipids found in this study,
exploration of the effects of air pollution exposure on lipid metabolism should be further
explored in animal models.
In conclusion, this body of work has shown that near-roadway NOx is a potential risk
factor for childhood obesity and that short- and long-term ambient pollutants may alter lipid
metabolism in young adults. Though first year of life was statistically significantly associated
with childhood BMI growth, in utero effects remain uncertain from our data. Additional research
is warranted to see if one period is more susceptible to the effects or air pollution than the other.
Building on this idea, continued exploration of GRSxE interaction on childhood obesity is also
needed. Finally, a longitudinal study design extending the Meta-AIR study’s objectives would
greatly enrich the data that has already been collected. These future studies can continue to build
on this dissertation work which will provide much needed evidence to emphasize and support the
absolute need for air quality standards in our communities.

162

5.3 References

1. Brunekreef B, Holgate ST. Air pollution and health. Lancet 2002;360(9341):1233-42.
2. Gauderman WJ, Avol E, Gilliland F, et al. The effect of air pollution on lung
development from 10 to 18 years of age. N Engl J Med 2004;351(11):1057-67.
3. Dockery DW, Pope CA, 3rd. Acute respiratory effects of particulate air pollution. Annu
Rev Public Health 1994;15:107-32.
4. Neophytou AM, White MJ, Oh SS, et al. Air Pollution and Lung Function in Minority
Youth with Asthma in the GALA II (Genes-Environments and Admixture in Latino
Americans) and SAGE II (Study of African Americans, Asthma, Genes, and
Environments) Studies. Am J Respir Crit Care Med 2016;193(11):1271-80.
5. Shanley RP, Hayes RB, Cromar KR, et al. Particulate Air Pollution and Clinical
Cardiovascular Disease Risk Factors. Epidemiology 2016;27(2):291-8.
6. Chuang KJ, Yan YH, Cheng TJ. Effect of air pollution on blood pressure, blood lipids,
and blood sugar: a population-based approach. J Occup Environ Med 2010;52(3):258-62.
7. Poursafa P, Mansourian M, Motlagh ME, et al. Is air quality index associated with
cardiometabolic risk factors in adolescents? The CASPIAN-III Study. Environ Res
2014;134:105-9.
8. Thiering E, Cyrys J, Kratzsch J, et al. Long-term exposure to traffic-related air pollution
and insulin resistance in children: results from the GINIplus and LISAplus birth cohorts.
Diabetologia 2013;56(8):1696-704.
9. Brook RD, Jerrett M, Brook JR, et al. The relationship between diabetes mellitus and
traffic-related air pollution. J Occup Environ Med 2008;50(1):32-8.
10. Singh AS, Mulder C, Twisk JW, et al. Tracking of childhood overweight into adulthood:
a systematic review of the literature. Obes Rev 2008;9(5):474-88.
11. Bolton JL, Smith SH, Huff NC, et al. Prenatal air pollution exposure induces
neuroinflammation and predisposes offspring to weight gain in adulthood in a sex-
specific manner. FASEB J 2012;26(11):4743-54.
12. Xu X, Yavar Z, Verdin M, et al. Effect of early particulate air pollution exposure on
obesity in mice: role of p47phox. Arterioscler Thromb Vasc Biol 2010;30(12):2518-27.
13. Wei Y, Zhang JJ, Li Z, et al. Chronic exposure to air pollution particles increases the risk
of obesity and metabolic syndrome: findings from a natural experiment in Beijing.
FASEB J 2016.
14. Jerrett M, McConnell R, Wolch J, et al. Traffic-related air pollution and obesity
formation in children: a longitudinal, multilevel analysis. Environ Health 2014;13:49.
15. McConnell R, Shen E, Gilliland FD, et al. A longitudinal cohort study of body mass
index and childhood exposure to secondhand tobacco smoke and air pollution: the
Southern California Children's Health Study. Environ Health Perspect 2015;123(4):360-
6.
16. Rundle A, Hoepner L, Hassoun A, et al. Association of childhood obesity with maternal
exposure to ambient air polycyclic aromatic hydrocarbons during pregnancy. Am J
Epidemiol 2012;175(11):1163-72.
17. Mao G, Nachman RM, Sun Q, et al. Individual and Joint Effects of Early-Life Ambient
PM2.5 Exposure and Maternal Pre-Pregnancy Obesity on Childhood Overweight or
Obesity. Environ Health Perspect 2016.

Abstract (if available)

Abstract Childhood obesity continues to be a serious public health problem worldwide. In the United States, prevalence of obesity has been increasing at an alarming rate over the past 40 years and has somewhat tapered since early 2000s (1). Global prevalence of childhood obesity has also steadily increased with nearly a quarter of children being overweight or obese in developed countries (2). The childhood obesity epidemic has grave public health implications including a rise in obesity-related health burdens. Obese children are at a greater risk for developing health complications including insulin resistance, type 2 diabetes, metabolic syndrome and cardiovascular disease (3-5). ❧ Despite decades of diet and physical activity interventions, prevalence of childhood obesity has yet to decline suggesting that other factors may play an important role in obesity prevention. Emerging studies show the importance of environmental pollutants like air pollution as obesogens, chemicals that disrupt metabolic processes

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Street connectivity and childhood obesity: a longitudinal, multilevel analysis

PDF

A cohort study of air-pollution and childhood obesity incidence

PDF

Native American ancestry among Hispanic Whites is associated with higher risk of childhood obesity: a longitudinal analysis of Children’s Health Study data

PDF

Air pollution, mitochondrial function, and growth in children

PDF

Effects of near-roadway air pollution exposure on obesity, obesity-related behavior, and neurobehavioral deficits during peripuberty

PDF

Effects of stress and the social environment on childhood asthma in the children' s health study

PDF

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

PDF

Airway inflammation and respiratory health in the Southern California children's health study

PDF

Prenatal environmental exposures and fetal growth in the MADRES cohort

PDF

Ambient air pollution and lung function in children

PDF

Early life risk factors for childhood asthma

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

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

PDF

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

PDF

Investigating factors that influence peer relationships and obesity during middle childhood

PDF

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

PDF

Metabolic consequences of obesity-associated inflammation during puberty and perinatal development

PDF

Air pollution and breast cancer survival in California teachers: using address histories and individual-level data

PDF

Long term effects of air pollution on lung function in the European community respiratory health survey

PDF

Smoking in pregnancy: from effects to solutions

Asset Metadata

Creator Kim, Jeniffer S. (author)

Core Title Air pollution and childhood obesity

School Keck School of Medicine

Degree Doctor of Philosophy

Degree Program Epidemiology

Publication Date 10/29/2019

Defense Date 08/08/2019

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

Tag Air pollution,ambient air pollution,cardiometabolic health,childhood body mass index,childhood obesity,early life exposures,gene-environment interactions,genetic risk score,in utero exposures,near-roadway air pollution,OAI-PMH Harvest,Young adults

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Gilliland, Frank (committee chair), Berhane, Kiros (committee member), Eckel, Sandrah (committee member), Goran, Michael (committee member)

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

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-229819

Unique identifier UC11673758

Identifier etd-KimJeniffe-7890.pdf (filename),usctheses-c89-229819 (legacy record id)

Legacy Identifier etd-KimJeniffe-7890.pdf

Dmrecord 229819

Document Type Dissertation

Rights Kim, Jeniffer S.

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